5. Labor Productivity and Labor Reallocation
5.1.2. Trends in labor productivity and real wages
Two quantitative techniques are utilized in analyzing aggregate labor pro- ductivity growth and its sources. First, the connection between labor produc- tivity, patterns of employment, and labor cost is traced. In theory, integration with global markets and outward orientation are expected to benefit an economy as the absorption of foreign technology and productivity gains may facilitate rises in real wages and total employment levels. Industries exhibiting these characteristics, such as electronics, would also serve as the engine of growth for the rest of the economy. Here, the so-called Hodrick–
Prescott filter is used to dismantle the cyclical variations in productivity growth and real wages from their respective historical trends. This method disassociates the trend component and helps us to understand better the wage cycle and long-term productivity patterns. On a second level of analysis, a decomposition exercise on the nature and sources of productivity growth within and across the sectors of the economy is conducted. Here, we have extensively used the methodology described in Timmer and Szirmai (2000).
Labor productivity is a partial productivity measure, which is defined as the amount of output per unit of input (Figure 5.1). This can be represented as
PX= Q
X, (5.1)
9,50 10,00 10,50 11,00 11,50 12,00 12,50 13,00
1959 1961 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001
Log Scale
Manuf Agr Util Const Com Tr- Com Fin- Bus Total
Fig. 5.1. Labor productivity (labor:number of employees) (1959–2002).
Note: The data are smoothed with the Hodrick–Prescott filter.
Source: Author’s calculations.
wherePX,Q,Xrepresent productivity, output, and input amounts, respec- tively. We are rather interested in the growth of productivity, which is calculated as follows:
PX
PX
= Q Q −X
X , (5.2)
where the operator denotes change in a relevant item in a given time period. Equation (5.2) states that the growth rate of productivity of inputX equals the difference between the growth rates of output and inputX. In the case of labor productivity,Xrefers to labor input.
In labor productivity analyses, we have used the computed real value- added data for the sectors and manufacturing industries. Labor is repre- sented by both labor hours and number of employees. Consequently, two measures for labor productivity are derived: value-added per worker (Q/L) and per working hour (Q/H).Q,L, andH refer to real output, number of employees, and working hours, respectively.
Labor productivity growth rates for both labor productivity measures at the major sector and industry levels are presented in Tables 5.4 and 5.5.
11:539inx6inB-655b655-ch05
LaborProductivityandReallocationinSingapore111
Table 5.4. Labor Productivity Growth Rates (Labor Input: Number of Employees).
Sectors/
Industries 1959–1965 1966–1978 1979–1985 1959–1985 1965–1985 1986–1997 1998–2002 1986–2002 1959–2002 1965–2002 AGR 0.6 −16.3 9.2 −4.3 −6.5 0.5 −10.1 −2.6 −3.6 −4.8
MIN 6.8 9.3 6.5 8.0 7.1 −2.4 −13.7 −5.7 2.6 1.4
MANUF 3.6 0.0 3.8 2.8 2.3 5.1 6.7 5.5 3.9 3.8
Food 2.4 1.9 4.7 3.0 2.9 2.8 −8.4 −0.5 1.6 1.4
Tex 15.1 4.1 3.8 9.0 2.5 4.8 −1.8 2.9 6.6 2.7
Wear −2.4 5.0 −1.1 0.2 1.6 2.1 9.2 4.2 1.8 2.7
Leat −0.9 1.0 3.0 1.1 0.2 5.5 −0.7 3.7 2.1 1.7
Wood 2.4 5.8 −2.1 2.0 2.6 4.8 −5.9 1.6 1.9 2.2
Furn 3.3 −1.8 4.8 2.0 −0.6 2.3 −23.8 −5.4 −0.9 −2.8
Paper −3.7 3.0 13.3 4.7 7.8 1.8 −8.1 −1.1 2.4 3.8
Pub −0.8 0.7 5.7 2.1 2.7 5.0 −4.2 2.3 2.2 2.5
Chem 18.0 8.4 5.4 12.4 7.9 4.3 22.6 9.7 11.3 8.7
Petr 14.1 2.6 −5.9 5.9 0.7 5.1 7.0 5.7 5.8 2.9
Rub 9.8 3.5 3.4 4.3 2.5 2.7 −1.0 1.6 3.3 2.1
Non-met 2.9 5.6 −0.5 4.4 3.1 5.3 −10.0 0.8 3.0 2.1
(Continued)
11:539inx6inB-655b655-ch05
DevelopmentinEastAsia
Table 5.4. (Continued) Sectors/
Industries 1959–1965 1966–1978 1979–1985 1959–1985 1965–1985 1986–1997 1998–2002 1986–2002 1959–2002 1965–2002
Met 12.7 6.8 −0.7 6.9 3.6 −0.2 −2.0 −0.7 3.9 1.6
Fab-met −0.8 1.8 3.0 1.7 2.4 3.0 −2.3 1.5 1.6 2.0
Mach −0.5 4.3 1.6 2.9 4.8 −1.1 7.7 1.5 2.3 3.3
Elec 0.9 −1.2 8.0 2.5 1.6 5.1 8.0 5.9 3.9 3.5
Prec 1.5 −1.1 7.8 5.2 2.0 4.9 8.8 6.0 5.6 3.8
Tran 3.4 5.3 4.0 4.0 3.7 2.9 −4.0 0.8 2.8 2.4
Oth-man −0.9 2.6 15.8 5.4 8.2 2.6 −19.0 −3.7 1.8 2.9
UTIL 2.2 10.5 9.1 7.8 9.3 4.5 5.9 4.9 6.7 7.3
CONST 12.4 2.8 1.8 4.1 2.7 5.0 −3.1 2.6 3.5 2.7
COM −18.7 −4.8 1.0 −7.9 −2.6 6.1 −4.1 3.1 −3.5 0.0
TR-COM 3.4 2.7 7.7 4.1 4.4 4.3 5.5 4.7 4.3 4.5
FIN-BUS 10.1 5.6 4.9 4.4 5.5 0.5 −7.0 −1.7 2.0 2.3
OTH-SERV 6.0 −0.5 2.4 2.6 0.4 4.2 −1.2 2.6 2.6 1.4
Total −0.8 0.0 3.0 0.3 1.0 4.4 2.9 4.0 1.7 2.3
Source: Author’s calculations.
11:539inx6inB-655b655-ch05
LaborProductivityandReallocationinSingapore113
Table 5.5. Labor Productivity Growth Rates (labor Input: Total Labor Hours).
Sectors/
Industries 1959–1965 1966–1978 1979–1985 1959–1985 1965–1985 1986–1997 1998–2002 1986–2002 1959–2002 1965–2002 AGR 6.3 −17.0 9.3 −4.5 −7.1 −0.1 −10.0 −3.0 −3.9 −5.3
MIN 3.3 8.8 8.3 7.4 7.3 −2.7 −13.9 −6.0 2.1 1.3
MANUF 3.1 2.7 4.9 3.4 3.2 4.9 −1.1 3.1 3.3 3.2
Food 3.8 1.4 5.0 2.9 3.3 2.9 −8.8 −0.5 1.5 1.6
Tex 24.0 3.6 4.0 8.4 2.2 4.3 −0.9 2.8 6.2 2.4
Wear −9.3 4.6 −0.7 −0.1 1.6 1.8 8.4 3.8 1.5 2.6
Leat −0.6 0.1 3.4 0.8 −0.3 5.2 −1.0 3.4 1.8 1.4
Wood −1.6 5.1 −1.2 1.9 2.5 4.3 −4.3 1.8 1.8 2.2
Furn −2.6 2.2 5.7 2.1 −0.8 1.9 −22.9 −5.4 −0.9 −2.8
Paper −1.3 2.4 13.9 4.6 7.5 1.8 −8.2 −1.2 2.3 3.6
Pub 1.2 0.1 6.3 2.0 2.4 5.0 −4.2 2.3 2.1 2.4
Chem 29.5 7.6 5.7 12.1 7.5 3.9 24.7 10.0 11.3 8.6
Petr 27.1 1.8 −5.6 5.6 0.3 4.7 8.4 5.8 5.7 2.8
Rub 7.8 2.7 3.7 4.1 2.1 2.3 −1.1 1.3 3.0 1.7
Non-met 7.2 4.4 −0.1 3.8 3.0 4.5 −8.0 0.8 2.6 2.0
(Continued)
11:539inx6inB-655b655-ch05
DevelopmentinEastAsia
Table 5.5. (Continued) Sectors/
Industries 1959–1965 1966–1978 1979–1985 1959–1985 1965–1985 1986–1997 1998–2002 1986–2002 1959–2002 1965–2002
Met 15.5 6.0 −0.5 6.4 3.3 −0.6 −0.9 −0.7 3.6 1.5
Fab-met −0.3 1.6 3.2 1.6 2.5 2.4 −2.2 1.1 1.4 1.9
Mach 0.6 4.0 1.5 2.6 4.7 −1.3 8.0 1.5 2.1 3.2
Elec 4.2 −1.2 8.7 2.7 1.9 4.2 8.8 5.5 3.8 3.5
Prec 15.7 −1.8 9.0 5.2 2.1 4.5 9.3 6.0 5.5 3.8
Tran 0.7 5.4 4.2 4.0 3.8 2.8 −3.7 0.9 2.7 2.5
Oth-man −3.9 3.3 15.9 5.1 8.2 2.3 −20.7 −4.5 1.3 2.5
UTIL −1.3 11.3 9.4 7.9 9.8 4.5 6.9 5.2 6.8 7.7
CONST 8.6 2.8 1.2 3.7 2.4 4.9 −2.6 2.7 3.3 2.5
COM −24.8 −4.7 1.3 −7.7 −2.4 6.0 −3.8 3.2 −3.4 0.1
TR-COM 3.0 2.6 9.2 4.5 5.4 4.0 6.1 4.6 4.5 5.0
FIN-BUS 0.9 5.9 5.1 4.5 5.7 0.4 −6.9 −1.7 2.0 2.3
OTH-SERV 9.1 0.1 3.3 3.1 1.0 3.6 −1.0 2.2 2.7 1.6
Total −2.5 0.1 3.6 0.4 1.3 4.1 3.4 3.9 1.8 2.4
Source: Author’s calculations.
Table 5.4 presents the growth of labor productivity measured as output per labor and Table 5.5 presents the growth of labor productivity measured as output per working hour. In these tables, major sectors are denoted by capital letters. The tables show that in long periods (i.e., 1959–1985, 1965–1985, 1986–2002, 1959–2002, 1965–2002) the growth of labor productivity using any of the two measures converge to each other. In short, subperiods (e.g., 1959–1965, 1979–1985), however, they tend to diverge. The remarkable productivity growth performance of the chemicals industry is observable from both tables. This performance was accompanied in some periods by precision equipment and electrical and electronic machinery industries.
In this section, labor productivity and real wages are decomposed into a trend component and to cyclical deviations using the so-called “Hodrick–
Prescott filter” developed by Hodrick and Prescott (1997). This idea was borrowed from a study by Voyvoda and Yeldan (1999) on Turkey. The, long- run movements of real wages and productivity can be calculated using this method.
Suppose that series xt is composed of trend τt({τt}Tt=1) and cyclical ct({xt−τt}Tt=1)components:
xt =τt+ct, (5.3)
wheret denotes time (t = 1,2, T). Hodrick and Prescott (1997) suggest that with the following minimization method, the cyclical componentctcan be isolated fromxt:
{minτt}Tt=1
T
t=1
(xt−τt)2+λ
T−1
t=2
(τt+1−τt)−(τt−τt−1)2
, (5.4) whereλis the smoothing parameter (also named penalty parameter). The first term in the minimization function (5.4), the sum of the squares of devi- ations, penalizes the variance ofct. The second term, the summation of the second differences of the trend componentτtmultiplied by the smoothing parameterλ, places a penalty to the lack of smoothness inτt, i.e., a penalty on the variations in the growth rate of the trend component with the degree of penalization directly proportional to the value of the parameterλchosen.
Although Hodrick–Prescott filter is easy to use, selecting an appropriate value forλis a major drawback. Ifλapproaches to 0, the trend component is almost equal to the original series, and if diverges to an infinitely large
number, a linear trend is obtained. It is recommended to set the value ofλ to 100 for annual data in Hodrick and Prescott (1997). Following this tra- dition, we have set the value ofλto 100.
The results of the calculations are portrayed in Figures 5.1–5.6. In Figures 5.1, 5.3, and 5.5, the long-run trends of average labor productivity and real wages at the sectoral level are presented. Agriculture and mining sectors are ignored in these tables due to their unimportance for the economy.
Figures 5.2, 5.4, and 5.6 provide the same figures for major manufacturing industries. All values are smoothed with Hodrick–Prescott filter. For con- venience, the values in the figures are presented on a logarithmic scale.
Aggregate labor productivity, defined as real output per unit working hour, declined during the first phase (import-substitution era) and remained almost stable until the late 1970s, when the government initiated a large- scale restructuring to stimulate higher value-added activities via large-scale increases in wages. From then on, average labor productivity has an ever- increasing trend. This needs to be compared with the trend of real wages in Figure 5.5. The trend in real wages is increasing although slightly until the late 1970s and from then on it has an increasing tendency similar to
9,00 9,50 10,00 10,50 11,00 11,50 12,00 12,50 13,00 13,50
1959 1961 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001
Log scale
Food Chem Petr Fab- met Mach Elec Prec Tran
Fig. 5.2. Manufacturing labor productivity (labor:number of employees) (1959–2002).
Note: The data are smoothed with the Hodrick–Prescott filter.
Source: Author’s calculations.
1,75 2,00 2,25 2,50 2,75 3,00 3,25 3,50 3,75 4,00 4,25 4,50 4,75 5,00
1959 1961 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001
Log scale
Manuf Agr Util Const Com Tr-Com Fin-Bus Total
Fig. 5.3. Labor productivity (labor:working hours) (1959–2002).
Note: The data are smoothed with the Hodrick–Prescott filter.
Source: Author’s calculations.
1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 5.50 6.00
1959 1961 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001
Log scale
Food Chem Petr Fab- met Mach Elec Prec Tran
Fig. 5.4. Labor productivity (labor:working hours) growth rates of major manufacturing industries (1959–2002).
Note: The data are smoothed with the Hodrick–Prescott filter.
Source: Author’s calculations.
8,00 8,50 9,00 9,50 10,00 10,50
1959 1961 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001
Manuf Const Com Tr-Com Oth-serv Total
Fig. 5.5. Real labor remuneration levels (1959–2002).
Note: The data are smoothed with the Hodrick–Prescott filter.
Source: Author’s calculations.
8,50 9,00 9,50 10,00 10,50 11,00 11,50
1959 1961 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001
Log scale
Food Chem Petr Fab- met Mach Elec Prec Tran
Fig. 5.6. Real labor remuneration levels of major manufacturing industries (1959–2002).
Note: The data are smoothed with the Hodrick–Prescott filter.
Source: Author’s calculations.
aggregate productivity growth. Large increases from the late 1970s to the mid-1980s are a result of the high-wage policy. From the mid-1980s on, real wages follow a uniform increasing path with the rate of increase getting gradually lower. As seen in Figure 5.3, value-added per hour worked for the aggregate economy increased at a slowing rate after the mid-1980s. A similar movement can also be observed for the average real wage level in the economy (see Figure 5.5).
If the number of employees is used as the labor input measure, the highest productivity growth rate in the pre-1985 period was recorded by the utilities sector, followed by the services sectors (see Figure 5.1).
Primary sectors (mining and agriculture) can be ignored due to their very small shares in employment. For the post-1985 period, the best per- former is the manufacturing sector, chemicals, and electrical and electronic appliances industries in particular. Note that electrical and electronic appli- ances industry increased its share in total employment steadily throughout the entire period, until it remained more or less stable in the late 1990s.
Taking total working hours as the relevant labor input measure, there is not much change, but the best performer in the post-1985 becomes the utilities sector and transport and communications services sector exhibits a higher productivity growth rate than the manufacturing sector. Higher productivity levels of the utilities and transport and communications services sectors can be explained by the fact that despite decreases in their both value-added and employment shares, the decline in their employment share was sharp whereas the decline in their value-added share was moderate, hence giving rise to increased productivity levels.
To see the differences between the productivity growth rates across industries better, one should consider international trade as a major eco- nomic activity in Singapore. As pointed out by MTI (2001), the degree of competition and openness for each sector is highly important in this respect.
The sectors with inferior productivity growth rates (such as services and construction) are, by their nature, producers of less tradable goods and ser- vices and are inward-oriented. However, other sectors with high produc- tivity growth performances such as the manufacturing and transport and communications sectors are more open to free trade and hence are subject to competition with foreign rivals, which is a stimulant for upgrading and restructuring.
The trends in real wages reflect the tight labor market conditions in Sin- gapore. The increase in real wages accelerates from the early 1970s when Singapore achieved full employment. It is important to note that real wages increase was slow after the mid-1980s. There seems to be a strong relation between real wages and labor productivity, especially in the manufacturing sector. The initial suppression of wages during the early years of industri- alization (until the restructuring efforts of the late-1970s) as represented by relatively little real wage increases in Figures 5.5 and 5.6 was followed by the government’s adjustment policies in the labor market to stimulate higher value-added activities and facilitated further increases in real wages. The change in real wage trends to an increasing one from the late 1970s was not accompanied by a rising trend in labor productivity. The trends in pro- ductivity rather favored real wage stability or small increases in real wage growth over productivity growth. These findings suggest a weak relation between gains in labor productivity and real wage earnings. Thus, the strong influence of the government in wage determination until the mid-1980s and governmental efforts in improving labor productivity went head-to-head and reinforced each other. Following the restructuring efforts, productivity gains allowed producers to offset production costs brought about by increases in real wages to a large extent.