CHAPTER 2

PRODUCTIVITY CONCEPTS AND MEASUREMENT:

A LITERATURE REVIEW

2.1 Introduction

This chapter presents a review of the literature on the basic productivity concepts;

productivity measures employed in the construction industry; methods to measure TFP

growth in the construction industry; and productivity concerns in the construction

industry of Singapore.

Based on this literature review, the chapter attempts to answer the questions: (1) what

is the most appropriate productivity concept to measure construction industry-level

productivity performance? and (2) what is the most appropriate method to measure

TFP growth in the construction industry?

There exists a lot of productivity concepts and measures. Different measures serve

different purposes. Based on the production function, productivity is theoretically

defined in Section 2.2. Section 2.3 discusses the significance of productivity. Section

2.4 provides a historical view of the development and definition of basic productivity

concepts. Partial productivity and total factor productivity are discussed, respectively,

in this section. Section 2.5 reviews the existing productivity concepts and

measurement, focusing on the construction industry. Section 2.6 discusses what is an

appropriate productivity concept to measure construction industry-level productivity

19
performance. Section 2.7 provides answers to what is an appropriate method to

measure TFP growth in the construction industry through a comparison between the

Chau’s method and the Jorgenson’s method. Section 2.8 reviews the existing

productivity measures employed at the economy level of Singapore and in the

construction industry of Singapore. Section 2.9 highlights the current need for having

an appropriate measurement in the context of the construction industry of Singapore.

2.2 Definition of productivity

Broadly defined, productivity is the relationship between output produced and one or

more of associated inputs used in the production process (National Research Council,

1979). The divergence in opinion among researchers centers on the choice of concept

for a specific measurement purpose, and how to measure output and inputs. In essence,

the selection of an appropriate concept of productivity depends on the objective of

measurement, availability of data, and the preference of research.

2.3 Significance of productivity

Productivity can be evaluated with respect to level or growth rate. A high productivity

level represents good use of resources and high returns. High growth rates indicate a

dynamic and growing economy or industry with great potential. Hence, nations or their

individual industry, no matter the forerunner with high productivity or the fall-behinds,

in order to keep competitiveness in the international marketplace, will make every

effort to improve their productivity growth.

20
At the national level, productivity drives economic growth and development, and

indicates potential for improved material standards and increase in the society’s

general welfare as well as more leisure time for the inhabitants. The construction

industry can provide impetus for economic growth owing to its size and particular

linkage with the rest of economy. Increasing productivity in construction results in a

virtuous circle. It leads to decrease in relative prices on construction products, higher

wages for construction workers, and expanded profits for contractors and building

owners, which in turn leads to increase in investment, more working opportunity and,

enhanced industry’s competitiveness.

2.4 Basic productivity concepts

Productivity can be measured in terms of level and rates of change. Generally,

productivity can be studied at four levels: site/project, firm/organization, industry and

entire economy. Because productivity measures exist largely to be compared, people

are more interested in productivity change. Hence it is more meaningful to use

productivity measures as indices of performance (National Research Council, 1979).

The conceptual framework for the measurement of productivity, including changes in

productivity over time, should be systematically built on and derived from production

theory, by means of empirical production functions (National Research Council, 1979;

Gollop, 1985). When output and inputs have been measured in constant prices over

time, ratios of real output to individual real input can be calculated to obtain single

productivity measures; and ratios of real output to all associated real inputs can be

calculated to obtain a MFP or TFP. One should be aware that in most previous

21
productivity studies, no distinction is made between the use of terms MFP and TFP1.

However, this study will strictly classify the situations under which each of them

should be appropriately employed.

Single productivity measures are useful in showing the savings that have been

achieved over time in the use of each input per unit of output. Their changes however

reflect not only changes in productive efficiency but also factor substitutions that result

from changes in relative factor prices. In contrast, TFP has been widely accepted as a

much better indicator of productive efficiency than conventional partial productivity in

measuring the efficiency of utilization of resources.

2.4.1 Single productivity concepts

Historically labour productivity was the first type of productivity measure to be

developed. Labour productivity (Q/L) is a measure of the ratio of output produced to

labour input used. Since labour is only one of the factor inputs, changes in labour

productivity are affected by changes in factor substitution as well as by changes in

productive efficiency as measured by multi-factor productivity or total factor

productivity. Other single productivity concepts are capital productivity and

intermediate productivity, which measure the relationship between output and capital

input and the relationship between output and intermediate input, respectively. Among

the three single productivity concepts, the most widely used measure is labour

productivity, which is usually measured as value-added per worker.

1
The output-labour-capital-materials productivity is known as “MFP” in UK (Oulton, 1994). In
USA, it was known as “TFP” (Jorgenson, 1987). In HK, it is known as “TFP” (Chau, 1988;
1993). In Singapore, value-added labour capital productivity is known as “MFP” by DOS
(1997), by SPRING as “TFP”. In HK, it was known as “VATFP” (Chau, 1993).

22
As different levels of the study are associated with different kinds of production

functions, productivity measures as well as productivity indexes (or productivity

growth) of each level will be discussed separately.

2.4.1.1 Single productivity at the economy level

At the economy level, as the objective for production is to satisfy human wants

through deliveries to end users, the proper measure of output is final demand or

aggregate value-added, and the primary inputs are labour and capital. Intermediate

inputs can be viewed as internal, offsetting transfer. Consequently, the relevant

production function for the macro-economy can be expressed as:

V = f ( L, K , T ) (2-1)

where, V is value-added, L and K is labour and capital input, and T is time.

Hence, at the economy level, labour productivity is the ratio of V to L, which is usually

measured as value-added per worker. Capital productivity is the ratio of V to K, which

is measured as value-added per capital stock used. To develop single productivity

indexes, indexes are obtained as the difference between the level of productivity at

time T and T-1.

2.4.1.2 Single productivity at the industry level

The objective of individual industry is different from that of the entire economy.

While, the macro-economy is concerned with maximizing value-added, the individual

23
sectors satisfy human wants through the production of gross output. Therefore, at the

industry level, the proper measure of the industry’s output is its total production, the

sum of its deliveries to intermediate and final demands (Gollop, 1985). Consequently,

the volume of output (Q) is a function of the volume of services of the basic factors of

production, labour (L) and capital (K), of the intermediate goods and services

purchased from other industry (X), and of the level of productive efficiency, which

changes over time (t):

Q = f ( L, K , X , t ) (2-2)

Based on the production function, labour productivity can be defined as Q/L, the ratio

of the ratio of output produced to labour input used. Similarly, capital productivity can

be defined as Q/K, the ratio output produced to capital input used. It should be noted

that at the macro-economy level, value-added is the proper measure for output; while

at the industry level, gross output is the proper measure for output. However, in

practice, value-added per worker is a prevalent indicator to measure labour

productivity at the industry level. Again, to develop single productivity indexes,

indexes are computed as the difference between the level of productivity at time T and

T-1.

2.4.2 Multi-factor productivity (MFP) and Total factor productivity (TFP)

The development of measures of MFP has lagged behind that of labour productivity.

The increasing capital intensity in the 1950s forced economists to restructure their

measurement of productivity growth to incorporate both labour and capital inputs

24
(National Research Council, 1979). The development of national economic accounting

systems and production function theory in the 1950s has made the empirical analysis

of MFP ultimately possible.

Early estimates of MFP were conducted by Jan Tinbergan in 1942 and by George

Stigler in 1947, and Barton and Cooper in 1948. The basic concept of MFP underlying

their work is:

Q
MFP = (2-3)
aL + bK

In this formulation, Q is the real product of the industry; L is labour input, measured as

labour hours; K is capital input, assumed to changes proportionately to real capital

stocks in the various industry; and a and b are the percentage shares of labour and

capital in the factor income generated in the industry. However, this concept was

regarded as misleading because intangible inputs resulting from research and

development, education and training are not included with the tangible factor inputs

(National Research Council, 1979).

At the same time, MFP was being measured within the framework of an economic

accounts framework. Solow (1957) initiated the estimation of MFP using the Cobb-

Douglas production function:

Q = TLα K β (2-4)

25
In this formulation, the exponents α and β is the elasticity of output with respect to

labour and capital input or, when index numbers are used, the income shares of factors.

In time series analysis, the scalar T measures the change in MFP over time. The rate of

change in T is the difference between the rates of change in output and the weighted

rates of change in the inputs:

∆T ∆Q ∆L ∆K
= −α −β (2-5)
T Q L K

MFP is the residual part of the rate of growth of output which cannot be accounted by

the weighted growth rate of inputs.

Following Solow’s work, there has been considerable further development of the

production function approach. Contrary to the Cobb-Douglas formulation which

assumes neutral technological change, Jorgenson and his associates (1987) use a

translog (transcendental logarithmic) production function to measure industry-level

TFP by assuming “biased technological change”. The advantage of using the translog

production function is that it allows the researchers to estimate TFP growth without

having to estimate the parameters of the production function. Under constant returns to

scale and producer equilibrium, the average rate of productivity growth can be

expressed as the difference between the growth rate of output and the weighted

average growth rate of intermediate, capital and labour inputs. Biased technological

change means that the value-shares of the factors in total output may be changing over

time, which is more realistic than Cobb-Douglas formulation which assumes constant

value-shares of the factors in total output (i.e. neutral technological change). Moreover,

unlike Cobb-Douglas method which considers only capital and labour inputs,

26
Jorgenson’s method measures industry-level productivity growth in a more accurate

way by treating intermediate, capital and labour inputs symmetrically.

2.4.2.1 Multi-factor productivity (MFP) at the economy level

As stated before, the conceptual framework for productivity growth can be derived

from the production theory, by means of empirical production functions. At the

economy level, the relevant production function for the macro-economy is:

V = f ( L, K , T ) (2-6)

where, V is value-added, L and K is labour and capital input, and T is time.

Because aggregate productivity is concerned with how well the economy transforms

inputs into output, the rate of productivity growth can be defined as the rate of growth

of value-added with respect to time, holding labour input and capital input constant.

Here, the term MFP is used to represent the aggregate productivity growth of the

economy level. By taking the total logarithmic derivative of Equation (2-6) with

respect to time:

d ln V ∂ ln f ∂ ln V d ln L ∂ ln V d ln K
= + ⋅ + ⋅ (2-7)
dT ∂T ∂ ln L dT ∂ ln K dT

∂ ln f
where, ∆ ln MFP =
∂T

Under constant returns to scale and competitive factor markets,

p L L ∂ ln V
vL = =
qV V ∂ ln L

27
 p K K ∂ ln V
vK = = (2-8)
qV V ∂ ln K

and

pL L pK K
+ =1 (2-9)
qV V qV V

where, qV is the price associated with aggregate value-added, p L and p K are

economy-wide average prices for labour and capital inputs, respectively,

To rewrite Equation 2-7 in the equivalent form and one gets:

∂ ln f d ln V ∂ ln V d ln L ∂ ln V d ln K
∆ ln MFP = = − ⋅ − ⋅ (2-10)
∂T dT ∂ ln L dT ∂ ln K dT

Consider the data at any two discrete points of time, say t and t-1, the growth rate of

MFP can be expressed as the growth rate of value-added minus weighted average

growth rate of labour and capital inputs:

− −
∆ ln MFP = [ln Vt − ln Vt −1 ] − v L [ln Lt − ln Lt −1 ] − v K [ln K t − ln K t −1 ] (2-11)

2.4.2.2 Total-factor productivity (TFP) at the industry level

At the industry level, two different TFP concepts have been used. One is MFP (some

researchers like to call it TFP, but essentially it is MFP), which relates to value-added

production function and takes account of labour and capital inputs in its measurement.

Another is TFP, which relates to gross output production function and incorporates

three factors—labour, capital and intermediate inputs in its measurement.

28
At the industry level, the methodology currently employed to measure MFP is

basically similar to the one used at the economy level. The format is the same as

equation 2-11, except that each term is accordance with an industry scale instead of

economy scale. The following sections focus on developing an index of TFP.

Similarly, the microeconomic theory of production provides the basis for measurement

of productivity growth at the industry level. The proper measure of productivity can

be derived from its production function, which is:

Q = f ( L, K , X , t ) (2-12)

where, Q is gross output, L, K and X are labour and capital and intermediate inputs,

and t is time.

Again, as industrial productivity is concerned with how efficiently inputs are converted

into outputs, the rate of productivity growth can be defined as the rate of growth of

output with respect to time, holding labour, capital and intermediate inputs constant.

At the industry level, the term TFP is used to represent productivity growth. By taking

the total logarithmic derivative of Equation 2-12 with respect to time:

d ln Q ∂ ln f ∂ ln Q d ln L ∂ ln Q d ln K ∂ ln Q d ln X
= + ⋅ + ⋅ + ⋅ (2-13)
dT ∂T ∂ ln L dT ∂ ln K dT ∂ ln X dT

∂ ln f
where, ∆ ln TFP =
∂T

Under constant returns to scale and competitive factor markets,

p L L ∂ ln Q
vL = =
qQ ∂ ln L

29
 p K K ∂ ln Q
vK = =
qQ ∂ ln K

p X X ∂ ln Q
vX = = (2-14)
qQ ∂ ln K

and

pL L pK K p X X
+ + =1 (2-15)
qQ qQ qQ

where q, p L , p K , and p X denote the prices of output, labour, capital and intermediate

inputs, respectively.

To rewrite Equation 2-13 in the equivalent form and one gets:

∂ ln f d ln Q ∂ ln Q d ln L ∂ ln Q d ln K ∂ ln Q d ln X
∆ ln TFP = = − ⋅ − ⋅ − ⋅ (2-16)
∂T dT ∂ ln L dT ∂ ln K dT ∂ ln X dT

Consider the data at any two discrete points of time, say t and t-1, the growth rate of

TFP can be expressed as the growth rate of gross output minus weighted average

growth rate of labour, capital and intermediate inputs:

− − −
∆ ln TFP = [ln Vt − ln Vt −1 ] − v L [ln Lt − ln Lt −1 ] − v K [ln K t − ln K t −1 ] − v X [ln X t − ln X t −1 ]

(2-17)

2.4.3 Relationship between single productivity and TFP

As divergence exists in the use of productivity concepts, to clarify the difference in

these concepts, it is necessary to discover the relationship between single productivity

30
and TFP. Again, situations at the economy level and industry level are discussed

separately in the following sections.

2.4.3.1 Relationship between single productivity and MFP at the economy level

By dividing Equation 2-1 by L, one obtains:

V f ( L, K , T )
= (2-18)
L L

Taking the total logarithmic derivative of Equation 2-18 with respect to T, one obtains:

V
d ln
L = ∂ ln f + ∂ ln V ⋅ d ln L + ∂ ln V ⋅ d ln K − d ln L (2-19)
dT ∂T ∂ ln L dT ∂ ln K dT dT

V
d ln
L is labour productivity growth (LP growth) and ∂ ln f
where, is the rate of
dT ∂T

aggregate productivity growth (MFP growth).

Under constant returns to scale and competitive factor markets,

p L L ∂ ln V
vL = =
qV V ∂ ln L

p K K ∂ ln V
vK = =
qV V ∂ ln K

and

pL L pK K
+ =1
qV V qV V

31
where, qV is the price associated with aggregate value-added, p L and p K and

economy-wide average prices for labour and capital inputs, respectively,

Then Equation (2-19) can be written in the following equivalent form:

− K K
ln LP (T ) − ln LP (T − 1) = [ln MFP(T ) − ln MFP(T − 1)] + v K [ln (T ) − ln (T − 1)]
L L

(2-20)

At any two discrete points of time, say T and T-1, the growth rate of labour

productivity can be expressed as the growth rate of MFP plus the weighted average of

the growth rate of capital intensity K/L. Similarly, capital productivity growth can be

expressed as the growth rate of MFP plus weighted average of the growth rate of L/K.

− L L
ln KP (T ) − ln KP (T − 1) = [ln MFP(T ) − ln MFP(T − 1)] + v L [ln (T ) − ln (T − 1)]
K K

(2-21)

From Equation 2-20, labour productivity growth is attributed to two factors-- MFP

growth and growth of capital intensity. Labour productivity may rise as long as capital

intensity K/L rise, although technological change (MFP growth) declines. One way of

enhancing labour productivity is through substitution of capital for labour (rise K/L) in

the production process. However, labour productivity growth will not be driven so

much by higher capital intensity K/L because of the limit to which capital investments

can increase before diminishing returns set in. A great deal of historical evidence

suggests that labour productivity have depended primarily on the level of investment

32
rather than on the rate of rise of TFP in Singapore during the last twenty decades.

However, labour productivity growth will not be driven so much by higher capital

intensity K/L because of the limit to which capital investments can increase before

diminishing returns set in. As the capital-labour ratio rises, returns to new investment

will tend to decline (Radelet, Sachs and Lee, 1997). Thus, it can be concluded that at

the economy level, MFP, which represents general technological change of an

economy over time (i.e. qualitative improvements that allow output to increase without

any use of additional inputs), is a more fundamental productivity concept than single

productivity.

2.4.3.2 Relationship between single productivity and TFP at the industry level

By dividing Equation 2-2 by L, one obtains:

Q f ( L, K , X , T )
= (2-22)
L L

Taking the total logarithmic derivative of Equation 2-22 with respect to T, one obtains:

Q
d ln
L = ∂ ln f + ∂ ln Q ⋅ d ln L + ∂ ln Q ⋅ d ln K + ∂ ln Q ⋅ d ln X − d ln L (2-23)
dT ∂T ∂ ln L dT ∂ ln K dT ∂ ln X dT dT

Q
d ln
L is labour productivity growth (LP growth) and ∂ ln f
where, is the rate of
dT ∂T

productivity growth (TFP growth).

Under constant returns to scale and competitive factor markets,

p L L ∂ ln Q
vL = =
qQ Q ∂ ln L

33
 p K K ∂ ln Q
vK = =
qQ Q ∂ ln K

p X X ∂ ln Q
vX = =
qQ Q ∂ ln X

and

pL L pK K p X X
+ + =1.
qQ Q qQ Q qQ Q

Then, Equation 2-23 can be written in the following equivalent form:

− K K
ln LP(T ) − ln LP(T − 1) = [ln TFP(T ) − ln TFP(T − 1)] + v K [ln (T) − ln (T − 1)]
L L
− X X
+ v X [ln (T) − ln (T − 1)]
L L

(2-24)

At any two discrete points of time, say T and T-1, the growth rate of labour

productivity can be expressed as the growth rate of TFP plus the weighted average of

the growth rate of capital intensity K/L and intermediate intensity X/L.

Similarly, capital productivity growth can be expressed as the growth rate of TFP plus

the weighted average of the growth rate of L/K and X/K.

− L L
ln KP (T ) − ln KP (T − 1) = [ln TFP (T ) − ln TFP (T − 1)] + v L [ln (T ) − ln (T − 1)]
K K
− X X
+ v X [ln (T ) − ln (T − 1)]
K K

(2-25)

34
From Equation 2-23, labour productivity growth is attributed to three factors-- TFP

growth and growth of capital intensity and intermediate intensity. Again, it can be

concluded that at the industry level, TFP, which represents technological change over

time in the specific industry, is a more fundamental productivity concept than single

productivity.

2.4.3.3 Relationship between MFP growth at the aggregate economy level and
sectoral TFP growth

The rate of aggregate productivity growth for the economy as a whole can be

decomposed into a weighted sum of sectoral productivity growth and weighted sums

of rates of growth of value-added, capital input, and labor input, reflecting the

reallocations of value-added and primary inputs among sectors. According to

Jorgenson (1987), the weighted sum of sectoral productivity growth is the most

important component of aggregate MFP growth. It overweighs the reallocation of

value-added, capital input, and labour input among sectors.

−
− wj − j −
vT = ∑ ⋅ v T + [ln V(T) − ln V(T − 1) − ∑ w j [ln Vj (T) − ln Vj (T − 1)
− j
vV
− j
− vK − j − −
+∑wj ⋅ ∑ v Kk [ln K kj (T) − ln K kj (T − 1)] − v K ⋅ ∑ v Kk [ln K k (T) − ln K k (T − 1)]
− j
vV
− j
− vL − j − −
+ ∑wj ⋅ ∑ v Ll [ln L lj (T) − ln L lj (T − 1)] − v L ⋅ ∑ v Ll [ln L l (T) − ln L l (T − 1)]
− j
vV

(2-26)

35
where,

−
v T -----aggregate rate of productivity growth, i.e. MFP growth at the economy level;

−
w j -----average ratio of value-added in the jth sector to value-added in all sectors;

− 1
w j = [ w j (T ) + w j (T − 1)];
2
p Vj Vj
wj = ;
∑ p Vj Vj

− j
v V ----average value share of value-added in output of the jth sector;

− j
v T ----average TFP growth of jth sector;

V -----aggregate value-added for all sectors;

Vj ----value-added for the jth sector;

− j
v K ---- average value share of capital in output of the jth sector;

− j
v Kk ----average value share of kth type of capital of the jth sector in the total capital of

the jth sector;

K kj ----quantity of kth type capital in the jth sector;

−
v K ----average of value share of total capital in total output for all sector;

−
∑ v Kk ----average value share of kth type of capital for all sector in the total capital for

all sector;

K k ------quantity of kth type of capital for all sector;

− j
v L ---- average value share of labour input in output of the jth sector;

− j
v Ll ----average value share of lth type of labour inpur of the jth sector in the total

labour input of the jth sector;

36
L lj ----quantity of lth type labour input in the jth sector;

−
v L ----average of value share of total labour input in total output for all sector;

−
∑ v Ll ----average value share of lth type of labour input for all sector in the total labour

input for all sector;

L l ------quantity of lth type of labour input for all sector;

2.5 Productivity measurement in construction

In the construction industry, productivity can be measured at three levels, that is,

site/project, firm and industry. Productivity concepts that have been adopted in the

construction industry include labour productivity, multi factor productivity (MFP) and

total-factor productivity (TFP) concepts.

Many studies on productivity measurement in construction focused on the site or

project and most of them relate to the measurement of labour productivity. Shaddad

and Pilcher (1984) developed a causal research model at the project or site level. The

view was widely held that factors at the head office and site level are the main

constraints to productivity performance. However, given the multiplicity of operations

involved in any project, and the interfaces among them, this view is not necessarily

correct. A few authors have highlighted the merit of assessing construction

productivity at a higher level (Horner et al., 1987). Kellogg et al. (1981) proposed a

hierarchical model for the study of construction industry productivity, which is too

conceptual to be of immediate use for an empirical study. Due to the complexity of the

measurement method and lack of data required (Hillebrandt, 1984), only a few

37
empirical research studies at the industry-level have been undertaken and even then,

most of them relate to labour productivity.

Measurement difficulties in construction are (Hillebrandt, 1984; and Valence, 1996):

• No two products are not identical, thus it is difficult to apply a unitary

measurement method.

• The fact that construction work done by establishments not classified as being

in the construction industry is excluded from output data. There is no

consensus on the definition and scope of construction. The frequent

modifications of the Standard Industrial Classification make data inconsistent.

• Price effect of business cycles will artificially affect data on output and inputs.

The deflators used to obtain real output data are mainly developed from indices

of labour and material costs instead of output price indices. It may

systematically underestimate the real growth of output and, hence, the

productivity.

• Impact of technology is difficult to be completely quantified.

• Capital stock, capital utilization is inadequately estimated; and data on capital

stock has been unavailable until recent years.

38
Until 1980s, empirical research at the industry level has captured extensive attention as

the productivity slowdown in the construction industry plagued almost every country

after the 1973 oil crisis. In the UK, research was led by Hillebrandt (1984) and NEDO

(1989). In the US, series work included Solow (1957), Dacy (1965), Koch and

Maovenzadeh (1979), Stokes (1981), the Business Roundtable (1982), Schriver and

Bowley (1985), Allen (1985), and the Building Research Board (1986). However,

most of these studies relate to labour productivity.

The increasing capital intensity has forced economists to restructure their measurement

of productivity growth to incorporate both labour and capital inputs. Using translog

value-added production function and Tornqvist index, the major statistical agencies

(Department Statistics of Singapore, US Bureau of Labour Statistics, Australian

Bureau of Statistics) have published MFP indices at both the national and industry

level. In the construction industry, using translog value-added production function and

Tornqvist index, Tan (2000) estimated the index of MFP for the Singapore

construction industry.

While MFP, in general, takes into account two factors, a few studies in the

construction industry incorporated all the three factor inputs---labour, capital and

intermediate inputs in their productivity measure model. However, due to data

problems in the construction industry they tended to use indirect ways. Koch and

Maovenzaden (1979) estimated TFP and technology change in the highway sector of

the construction industry in the United States through its components, i.e. the

productivities of labour, capital and materials. Chau and Walker’s approaches (1988;

and 1993) are to measure TFP indirectly from construction cost and price data. By

39
defining TFP as the ratio of quantity of output to quantity of aggregate input calculated

by an input aggregator function, assuming Hicks Neutral Technical change and

constant to return, Chau and Walkers represented TFP growth as the sum of weighted

change in the price of factor inputs: labour, capital and materials, minus the change in

the price of output.

2.6 Appropriate measure for construction industry-level productivity

It has been suggested that the definition of industry-level productivity cannot be

arbitrary and results of empirical research can be affected greatly if different

productivity concepts are adopted (Wolff, 1981). The purpose of this section is to

answer the question: what is the appropriate productivity concept to measure

construction industry-level productivity performance? Or among the existing

productivity concepts (labour productivity, MFP and TFP), which is the better

indicator to evaluate construction industry-level productivity performance?

In fact, the debate on appropriate measurement of industry-level productivity had

already begun earlier in 1980s within the framework of macro economic productivity

studies. Bronfenbrenner (1985) argues that when two or more distinct inputs worked

together to produce a common output, it may be or often be misleading to concentrate

on the productivity of one input while neglecting the others. Gollop (1985) believes

that the appropriate productivity measurement should follow directly from economic

principles, i.e., the well-defined indexes of productivity growth must measure how

well the economy allocates and the industries use all scarce resources. The notion of

scarcity is no more appropriate to any one primary input than another. The

40
characteristics of sector productivity growth differ importantly from its aggregate

counterpart. Whereas the macro-economy is concerned with maximising value-added,

individual sectors satisfy human wants through the production of gross output. He

argued that the proper measures of sectoral productivity must be evaluated in the

context of total-factor productivity. All labour, capital and material inputs are equally

primary to sectoral producers. Jorgenson (1987) further tested that the value-added

function must exist in the economy as a whole, but there is no existence of a value-

added function at the industry level. Hence, he concluded that it would be

inappropriate to employ value-added function in analyzing patterns of production at

the sectoral level. Instead, he uses a translog production function which treats labour,

capital and intermediate inputs symmetrically to develop the TFP index.

In the construction industry, many researchers have recognized the limits of labour

productivity and the need to consider total factor productivity (Lowe, 1987;

Rapporteur, 1987; Chau, 1993). Rapporteur (1987) disputed that change in the value of

ratio of output/number of operatives employed are often interpreted as measuring

changes in labour efficiency when, in fact, some other factors such as mechanization,

prefabrication, and change in demand may well have influenced the equation. He

suggests that researchers ought to be concerned with the total process of conversion of

raw materials to finally positioned components in buildings. Chau (1993) argued that

the contribution of new materials that are of better quality and easier to handle with to

construction productive efficiency cannot be ignored.

Based on the literature review, one can conclude that the appropriate production

function for the construction industry should take gross output as a function of

41
intermediate, capital and labour inputs and time and thus TFP is the appropriate

productivity measure to evaluate the industry-level productivity performance in

construction.

2.7 Appropriate TFP measurement for the construction industry

In the construction industry, the existing TFP measure is Chau and Walker’s

approaches (1988; and 1993), which measure TFP indirectly from construction cost

and price data. In macroeconomics, the best-known TFP measure is that of Jorgenson

(1987). Therefore, these two approaches will be presented separately in the following

sections. To answer the question of which is the more appropriate method to measure

TFP growth, a critique of Chau’s approach (1988; and 1993) is conducted to draw

some conclusions.

2.7.1 Chau’s approach

In Chau’s approach, TFP is defined as:

Q0
T= (2-27)
g (QI1 , QI 2 ,..., QI n )

where, T is TFP and g (.) is an input aggregator function, and it is assumed to be

linearly homogeneous and continuously differentiable.

Under the assumption of Hicks Neutral Technical change and constant returns to scale

and using the Divisia index, TFP growth can be expressed as:

42
 n −
∆ (ln T ) t = ∑ s it ∆(ln PIi ) t − ∆(ln PO ) t (2-28)
i =1

where, the ∆ sign denotes the difference in the value of the corresponding variable

between two successive periods and the bar sign denotes the average of the same two

successive periods, Po is the price of output and the PIi the price of ith input.

Thus, Chau (1993) made an empirical estimation of TFP growth in the Hong Kong

building industry by computing the difference between the sum of weighted change in

the prices of factor inputs-- labour, capital and materials, and the change in the price of

output.

2.7.2 Jorgenson’s approach

Building on the pioneer works of Solow (1957), Denison (1967) and Jorgenson and

Griliches (1967), Jorgenson et al. (1987) developed a production function-based TFP

measurement. It assumes that for each industry, there exists a transcendental

logarithmic (translog) production function, giving output as a function of intermediate

input, capital input, labour input and time.

Z = F ( X , K , L, T ) (2-29)

where, Z is quantity of output, X, K, L are quantity of the intermediate, capital and

labour inputs and T is time.

43
The rate of productivity growth, called TFP growth, is defined as the rate of growth of

output with respect to time, holding intermediate input, capital input and labour input

constant:

∂ ln Z
vT = ( X , K , L, T ) (2-30)
∂T

Under the condition of constant returns to scale and producer equilibrium, the average

growth rate of TFP can be expressed as the growth rate of output less the sum of

weighted average of growth rate of intermediate, capital and labour inputs:

− −
VT = [ln Z (T ) − ln Z (T − 1)] − V X [ln X (T ) − ln X (T − 1)] −
− − (2-31)
V K [ln K (T ) − ln K (T − 1)] − VL [ln L(T ) − ln L (T − 1)]

Where

− 1
V X = [V X (T ) + V X (T − 1)]
2
− 1
VK = [V K (T ) + VK (T − 1)]
2
− 1
VL = [V L (T ) + VL (T − 1)] (2-32)
2
− 1
VT = [VT (T ) + VT (T − 1)]
2
p X p K pL L
VX = X VK = K VL =
qZ qZ qZ

− − −
where, V X , VK , VL represent the respective shares of intermediate, capital and labour

inputs averaged over time T and T-1, while q, px, pk, pL denote the prices of the output

and intermediate, capital and labour inputs.

Each input can be defined as a translog function of its individual inputs. Under

constant returns to scale and producer equilibrium, the growth rate of each input can be

44
expressed as a weighted average of growth rates of individual input, with weights

given by the average value shares:

−
ln X (T ) − ln X (T − 1) = ∑ V Xi [ln X i (T ) − ln X i (T − 1)] (2-33)

−
ln K (T ) − ln K (T − 1) = ∑ V Kj [ln K j (T ) − ln K j (T − 1)] (2-34)

−
ln L(T ) − ln L(T − 1) = ∑ V Li [ln Li (T ) − ln Li (T − 1)] (2-35)

2.7.3 A critique study of Chau’s approach

Due to limited data available, physical measurement of inputs and outputs is not

possible in HK (Chau, 1988). Chau then modified the method of measuring TFP to suit

the HK’s case: that is to measure TFP growth of HK’s building industry indirectly

through construction cost and price data using Hicks neutral Technology change. This

made empirical calculation easily activated. A critique of Chua’s method is discussed

as follows.

Firstly, Chau’s method is only used for estimating TFP growth for building industry

(Chau, 1988 and 1993). It is not suitable for measuring TFP growth for the

construction industry as a whole. It is because the output price index (OPI) developed

by Chua (1990) is a price index for building work only and it should not be used as a

general deflator for all types of construction works.

Secondly, Chau’s method is based on the existence of an aggregate input function

g(QI1, QI2, …, QIn), in which aggregate input depends on the intermediate, capital

and labour inputs and is independent of the level of technology T. An aggregate input

45
is equivalent to the assumption of Hicks Neutrality of technology change (Jorgenson et

al., 1987). According to Chau (1993), under Hicks neutrality assumption, input

proportions (its value-share) remain unchanged if there is no change in relative price.

In other words, the rate of productivity growth is independent of intermediate, capital

and labour inputs and depends only on time, or biases of productivity growth are equal

to zero (Jorgenson et al., 1987).

The following sections explains what does it mean that the biased of productivity

growth are equal to zero. Consider a specific translog production function:

Z = exp[α 0 + α X ln X + α K ln K + α L ln L + α T ⋅ T +

1
β XX (ln X ) 2 + β XK ln X ln K + β XL ln X ln L + β XT ln X ⋅ T +
2
1
β KK (ln K ) 2 + β KL ln K ln L + β KT ln K ⋅ T +
2
1 1
β LL (ln L) 2 + β LT ln L ⋅ T + β TT ⋅ T 2
2 2

Under necessary conditions for producer equilibrium imply that the value-shares of

intermediate, capital, and labour inputs are equal to the elasticities of output respect to

these inputs. Similarly, rates of productivity growth are equal to rates of growth of

output, holding all inputs constant:

where

1
ln X = [ln X (T ) + ln X (T − 1)]
2
1
ln K = [ln K (T ) + ln K (T − 1)]
2

46
 1
ln L = [ln L(T ) + ln L(T − 1)]
2
1
T = [T + (T − 1)]
2

Under Hicks neutrality, the biases of productivity growth are zero for all three inputs,

i.e. β XT = β KT = β LT = 0 . Jorgenson tested the restriction derived from Hicks

neutrality, that is, the hypotheses that β XT = β KT = β LT = 0 and the hypotheses were

rejected. It proves that biases in productivity growth are significant and cannot be

ignored. This made Chau’s method which is based on neutral technology change

assumption restrictive and vulnerable. Consequently, using Chau’s method to calculate

TFP growth may obtain inaccurate or even misleading results.

In contrast, Jorgenson’s method is based on solid production function theory which

defines output as a function of intermediate input, capital input, labour input and time

Q=F(X,K,L,T). It does not impose the assumption of Hicks Neutral Technology

Change (Jorgenson, 1987) and hence can produce more accurate results than that of

Chau’s. Furthermore, Jorgenson’s method is applicable to the construction industry as

a whole.

From the above reasons, it can be concluded that the appropriate and less restrictive

TFP measurement for the construction industry is that of Jorgenson (1987).

Unlike the HK case, the data required by Jorgenson’s method (such as quantity of

inputs and outputs) in Singapore are available. It enables Jorgenson’s method to be

used for Singapore’s case.

47
2.8 A review of productivity measures at the economy level and the construction
industry of Singapore

This section starts with the problems associated with construction productivity in

Singapore. Then a literature review of productivity measures at the economy level and

in the construction industry are conducted. Next, the significance of productivity is

stated. Finally, the importance of having a good measure for productivity of the

construction industry of Singapore is highlighted.

2.8.1 Historical development of productivity measures at the economy level of

The development of productivity measures at the economy level of Singapore was

associated with the changing patterns of economic growth. In the 1960s and 1970s,

Singapore’s economic growth was mainly driven by labour accumulation. By the late

1970s, labour employment reaches its full capacity while the economy is still booming.

The tight labour market situation made improving productivity more pressing. In 1981,

the government launched the Productivity Movement to propel the economy from the

labour-driven to the investment-driven phase of development. Productivity was

highlighted as the key strategy to achieve the targeted annual economic growth of 8 to

10%. At the same time, in the 1970s and early 1980s, there was a large capital

investment in the economy. As returns to infrastructure-spending are usually evident

only in the medium to long term, the economic growth entered into a capital-driven

pattern only after 1980. As such the government was concerned primarily with

reducing labour inputs. Hence, productivity measures centre on labour productivity. A

measure which incorporates other factors of production would be of limited interest to

policy makers at that time.

48
In 1995, the Productivity Movement reached an innovation-driven phase. Thus far,

labour productivity has been a major source of economic growth. As the investment-

driven S-curve was nearing the peak in 1995, productivity growth would no longer be

driven so much by higher capital investments because of the limit to which such

investments can increase before diminishing returns set in. Rather, Singapore’s

economic growth would depend very much on making the best use of labour and

capital resource to achieve greater output per unit input, also known as Total Factor

Productivity (TFP)2. Since 1995, TFP has become a very important productivity

measure in Singapore and it has been recognized by the government as a sustainable

and long-term source for economic growth. In the new millennium, the Productivity

Movement emphasizes innovation and value creation in all sectors of the economy and

segments of society.

2.8.2 Problems in the construction productivity of Singapore

In Singapore, low productivity in the construction industry has been identified as one

of the main problem areas of the country’s economy (Ofori, 1993). An earlier report by

the CIDB Task Force on Productivity (1992) indicated that compared with the

country’s manufacturing sector and other countries, the productivity of the

construction industry of Singapore is relatively low. Since 1995, construction

productivity has been registering a negative productivity growth with levels dropping

more than 13 per-cent. Alerted by this trend, in 1998, the Construction 21 Steering

Committee (1998) was initiated to deal with the problems facing the industry, in

particular, the low level of productivity and the heavy reliance on unskilled foreign

workers.
2
It is essentially MFP, that is value-added labour capital productivity.

49
The causes for this situation include insufficient use of buildable designs especially

prefabrication, heavy reliance on a large number of unskilled foreign workers,

segregation of design from construction, multi-layered subcontracting system

involving many small firms, lack of site management and inadequate mechanisation

and automation (Construction 21, 1998; and CIDB Task Force, 1992).

Owing to its labour-intensive nature, the construction industry in Singapore employs a

large share of the nation’s workforce, disproportionate to its contribution to the

economy. Since labour is in very short supply in Singapore while foreign workers are

quite cheap, it causes the construction industry to rely heavily on foreign workers

(about 80 per-cent of construction labour are foreign workers). This has brought about

a lot of social problems such as crimes and increased accident rates to the construction

industry. With the government policy of controlling the number of foreign workers in

Singapore, the construction industry is facing the pressure to reduce its labour

requirement, while meeting the nation’s construction requirements. Consequently, a

series of programmes have been launched by the government to increase construction

productivity. These include the practice of (Construction 21, 1998):

a) buildable-design and prefabrication by requiring a minimum buildability score

for different types of buildings;

b) the foreign worker levy to discourage contractor from using foreign workers;

c) Design& Build (D&B) to integrate the construction process; and

d) the Investment Allowance Scheme for contractors to accelerate mechanisation

in construction.

50
Another important concern raised by the Construction 21 Committee in 1998 is that

the current state of affairs in the construction industry cannot be sustained. There is a

need for a major transformation of the industry to achieve better resource utilization

and efficiency, and to reduce wastage and social costs.

2.8.3 Productivity measures in the construction industry of Singapore

In the construction industry of Singapore, the initial productivity measure is labour

productivity. Value-added per worker is the simplest and often used productivity

measures employed in the construction industry of Singapore. Recording of the

statistical data for this indicator in Singapore started in 1975. However, “output” here

is measured in monetary value, which may not reflect the quantity of the products.

Moreover, it is difficult to use this indicator to make comparison study with the

construction industry of other countries. Aware of these problems, in 1992 the Task

Force on Construction Productivity (TFCP) of Singapore introduced a quantity-based

measure, square metres of build-up area per manday, as a supplementary indicator to

value-added per worker. The indicator measures the number of mandays on site

required to put up a unit of completed floor area. Publication of data on this alternative

measure started in 1991.

The development of the MFP measure has lagged behind the estimation of labour

productivity. In recent years, interest in the issue of Singapore’s TFP performance

arose as the government recognized the importance of TFP as a source of sustainable,

long-term economic growth for the fast-maturing, resource-scarce economy

(Department of Statistics, 1997). In 1997, the Department of Statistics (DOS) of

51
Singapore published the concept and methodology for the computation of MFP growth

in Singapore. It assumes a translog production function whereby, under the

assumptions of constant returns to scale and competitive markets for two factor inputs

(labour and capital), MFP growth is estimated from the difference between the growth

of output and a weighted average of the growth of labour and capital. Following the

methodology of DOS, the Construction Industry Development Board (CIDB) began to

release the index of MFP growth in the construction industry of Singapore from 1996.

Meanwhile, using the Tornqvist index approach and assuming a translog value-added

production function, Tan (2000) estimated the MFP growth (value-added two factor

productivity) for the Singapore construction industry between 1980 and 1996. The

equation used for the computation of MFP by Tan (2000) is identical to the one

employed by DOS (1997).

2.9 Importance of having a good measure for construction productivity in

The current challenge in Singapore construction is that the industry is facing the

dwindling local demand. Construction orders are projected to average between $12

billion and $15 billion per year for the next 10 years. It pushes the industry to upgrade

its capabilities through greater innovation to compete against the more established

foreign construction firms. New innovative products have been developed to enhance

the use of pre-cast products in the industry. Pre-cast technology became one of the

important niches identified by the government to sharpen the local firms

competitiveness (EDB, 2002). However, the current two productivity measures

employed in the construction industry of Singapore, LP and MFP failed in measuring

the increasing use of labour-saving or high technological materials such as

52
prefabrication in productivity improvement. Thus, a more comprehensive measure

which incorporates all of the factors of production (labour, capital and materials

inputs) is critical to meet the current needs in the industry.

Meanwhile, the current state of affairs in the construction industry of Singapore cannot

be sustained (Construction 21, 1998). To achieve sustainable growth, it is imperative

to transform the industry from a labour-intensive nature to an innovation-driven one to

achieve better resource utilization and efficiency, and to reduce wastage and social

costs. Thus, a good measure of productivity to evaluate the efficient use of all

resources in the construction industry is vital.

The definition of industry-level productivity cannot be arbitrary (Wolff, 1981) as

results of the empirical study can be greatly affected if different productivity concepts

are applied. The three productivity indicators currently adopted in the construction

industry are misleading as they concentrate on the productivity of one or two factors

and neglect the others. A high measure of productivity from an input (labour) in

producing an output may result from a wasteful overuse of one or more associated

inputs (machine, materials). Hence, in the construction industry of Singapore, it is not

appropriate to treat any one input more primary than the others. Thus, it is important

that a good productivity measure is found to determine how well the industry uses all

its scarce resources.

Consequently, in order to evaluate, monitor and control the progress made by the

programmes launched by the government to enhance the low productivity in the

Singapore construction industry, it is imperative to seek a good productivity measure.

53
2.10 Chapter summary

This chapter has examined the basic productivity concepts including their

development, definitions, and methods of measurement at the economy and industry

levels, and specifically to Singapore. Corresponding reviews were conducted to answer

two important questions involving productivity. Based on these reviews, it was

concluded that TFP is the appropriate productivity measure to evaluate industry-level

productivity performance in construction and Jorgenson’s (1987) method is the

appropriate measure for TFP growth.

Productivity was defined in Sections 2.2 as the relationship between the output

produced and one or more of the associated inputs used in the production process.

Significance of productivity is highlighted in Section 2.3.

Single productivity and TFP were discussed separately in Section 2.4 and each was

dealt with, respectively, at the economy level and industry levels. Also, to avoid the

misuse of productivity concepts, derivations of equations which links single

productivity and MFP at the economy level, as well as single productivity and TFP at

the industry level, MFP at the economy level and TFP at the industry level, were

carried out. The equations suggested that TFP (MFP), which represents growth due to

technological change, is a more fundamental concept than single productivity.

A historical review of the development of major productivity measures in the

construction industry is provided in Section 2.5. Measurement difficulties originating

in construction were also presented in this section.

54
In order to answer what is the appropriate productivity concept to measure

construction industry-level productivity, a systematic examination was conducted

under Section 2.6 from the perspectives of production function theory, as well as

specific context of the construction industry. Combining the two, it can be clearly

concluded that TFP is the appropriate productivity concept to measure construction

industry-level productivity performance.

What is the appropriate TFP measure method in the construction industry was

discussed in Section 2.7. A critique study of the existing TFP measure in the

construction industry, i.e., Chau’s method was conducted. The advantages of the one

recognized widely in economic academia, i.e. Jorgenson’s method, were pointed out. It

was concluded that Jorgenson’s TFP measure is the appropriate method to measure

construction industry-level productivity. For (1)unlike the Chau’s method, Jorgenson’s

method does not impose on the assumption of Hicks Neutral Technological Change,

which is tested unrealistic in the construction industry; (2) Chau’s method only applies

to building industry, but Jorgenson’s method is for the whole industry.

Section 2.8 reviewed the productivity measures employed in Singapore’s economy as

well as problems and productivity measures applied in Singapore’s construction

industry. It first reviewed the historical development of productivity measures at the

economy level of Singapore. It then discussed the current productivity problems

encountered in the construction industry of Singapore. It then reviewed the three

productivity measures currently employed in Singapore construction. As the three

productivity measures have their limitations, the importance of developing a good

measure of productivity in Singapore was argued at section 2.9.

55