The Impact of the Intersectoral Reallocation of Labor

5. Labor Productivity and Labor Reallocation

5.2. The Impact of the Intersectoral Reallocation of Labor

Structural transformation in Singapore came with changes in labor com- position and shifts of labor across sectors. Changes in the composition of production factors may result from different sources. In the case of Sin- gapore, a small open economy heavily dependent on trade, it is possible to argue that one source of these changes is changing demand for its products in world markets. In fact, the purpose of restructuring efforts in Singapore is to respond positively to changing comparative advantages. Certain sectors and industries lead productivity in the island economy. These sectors and industries are those characterized by high learning capabilities. This section investigates the effect of the changes in labor allocation across the sectors of the economy on the ability of output generation by way of productivity growth.

In many countries structural changes coexisted with gains in manu- facturing productivity. Some researchers in the past associated structural changes with aggregate productivity growth (e.g., Kaldor, 1963, 1966;

Cheneryet al., 1986; Syrquin, 1995). Kaldor (1963, 1966) argued that there is often surplus labor in some industries in the manufacturing sector and labor supply in the manufacturing sector is elastic. He asserted that a major source of labor supply in manufacturing is the flow of labor from low- to high-productivity industries and associates aggregate labor productivity growth in manufacturing sector with labor shifts. He also argues that labor shifts across industries increase average manufacturing productivity due to two reasons. First, labor absorbing industry is growing with increasing returns. Second, in the industry that is losing labor, productivity level rises with the withdrawal of labor. Kaldor emphasized increasing returns and externalities in manufacturing. Harberger (1998) further differentiated between what he calls a “mushroom-process” (i.e., innovative activities in a particular industry improve productivity and cause resource shifts from relatively low-productivity activities to itself), and a “yeast-process” (i.e., gains in productivity spreads across industries).4Long-run growth is a mix of these two processes.

Now, there is a large literature on the impact of changes in labor com- position on productivity for developing as well as developed countries (e.g., Salter, 1960; Syrquin, 1984, 1986; Timmer and Szirmai, 2000). These studies focus on the shifts of labor and capital from primary sectors (e.g., agriculture) to manufacturing and services sectors. They point to the positive contribution of resource reallocation from low-productivity sectors (most likely agriculture and traditional manufacturing industries such as food, wood, and textile manufactures) to sectors and industries that exhibit higher productivity (such as electronics, basic metals, and transport equipment).

In this section, we measure the impact of intersectoral labor shifts to pro- ductivity growth using the static shift-share method described in Timmer and Szirmai (2000). This method decomposes productivity growth into pro- ductivity growth arising from within sectors and from intersectoral labor shifts and has been used in various studies to analyze the impact of labor

4Harberger (1998) describes mushroom and yeast effects as follows: “… yeast causes bread to expand very evenly, like a balloon being filled with air, while mushrooms have the habit of popping up, almost overnight, in a fashion that is not easy to predict.”

shifts on labor productivity (e.g., Fagerberg, 2000; Timmer and Szirmai, 2000; Jalavaet al., 2002; van Ark and Timmer, 2003).

We start with the following equation:

LPt = Qt

Lt =

i

Qi,t

Li,t ãLi,t

Lt

, (5.5)

whereLPstands for aggregate labor productivity,Lfor total employment, Qfor total output in the relevant sector or industry, and the subscripttfor time. Labor input here is total number of workers. Terms without subscripts refer to the values for the aggregate economy. The termLi,t/Lt in Eq. (5.5) refers to labor share of the sector or industryiin total labor and the term Qi,t/Li,t refers to labor productivity for the same industry. Renaming the former as sli and the latter asLPi, Eq. (5.5) can be rewritten as follows:

LP =

i

LPiãsli. (5.6)

Equation (5.6) implies that aggregate labor productivity level is a weighted sum of individual industries or sectors. The weights are the respective shares of industries or sectors in total labor. Changes in labor productivity are defined for any time period [0,1], where 0 and 1 stand for the beginning and the end years of the period, respectively. Here the analysis is confined into two distinct time periods: pre-1985 (1965–1985) and post-1985 (1985–

2003) periods. The analysis starts from 1965 because during the import- substitution period (1959–1965) the limitations on labor and size of the market that affect structural changes are different from the post-1965 years.5 The change in labor productivity level can be written simply by sub- tracting the level of labor productivity at the end of the period (1)from that of the beginning of the period (0):

LP1−LP0=

i

LPi,1ãsli,1−

i

LPi,0ãsli,0. (5.7)

5Due to its irrelevance to the present analysis, pre-1965 period is ignored. During 1959–

1965, Singapore implemented the import-substitution policy of federal Malaya and eco- nomic policies were biased towards exploiting the Malayan market. In 1965, Singapore abandoned import substitution.

Rearranging with some algebraic manipulations and dividing each side by LP0 to rearrange (5.7) in growth terms, the above decomposition finally takes the following form:6

LP1−LP0

LP0

= 1 LP0

i

(LPi,1−LPi,0)ãsli,o+ 1 LP0

i

(sli,1−sli,0)

×LPi,o+ 1 LP0

i

(sli,1−sli,0)ã(LPi,1−LPi,0). (5.8)

6The derivation of (5.8) from (5.7) requires an algebraic manipulation. Since (5.7) holds, LP1−PL0=

iLPi,1ãsli,1−

iLPi,0ãsli,0, then, we get:

LP1−PL0= i

LPi,1ãsli,1− i

LPi,0ãsli,0

=

i

LPi,1ãsli,1− i

LPi,0ãsli,0+ i

LPi,1ãsli,0

−

i

LPi,1ãsli,0+ i

LPi,0ãsli,1

−

i

LPi,0ãsli,1+ i

LPi,1ãsli,1

−

i

LPi,1ãsli,1+ i

LPi,0ãsli,0− i

LPi,0ãsli,0. Rearranging the terms, Eq. (5.8) can be obtained as follows:

LP1−PL0=

i

LPi,1ãsli,0− i

LPi,0ãsli,0

+

i

LPi,0ãsli,1− i

LPi,0ãsli,0

+

i

LPi,1ãsli,1− i

LPi,1ãsli,0

+

i

LPi,0ãsli,0− i

LPi,0ãsli,1

=

i

LPi,1−LPi,0

ãsli,0

+

i

sli,1−sli,0

ãLPi,0+ i

LPi,1−LPi,0

ã sli,1−sli,0 .

The first term on the right-hand side of Eq. (5.8), which is labor share of the beginning year of the period multiplied by labor productivity change during the period, describes internal productivity growth within individual industries and measuresintra-industry productivity growth. Sectoral labor shares are used as weights. Therefore, intra-industry effect measures the change in aggregate labor productivity growth that would have resulted if labor shares remained constant over time.

The second term (change in labor share multiplied by the labor produc- tivity of the beginning year of the period) measures labor shift based on the labor productivity level of the beginning of the period. In other words, this effect measures the changes in aggregate labor productivity resulting from the movements of labor across industries with differing productivity levels had the labor productivity levels of individual industries remained constant over time. When the employment shares of industries with high produc- tivity levels rise, this means a reallocation of labor towards industries whose productivity is growing rapidly. Following Timmer and Szirmai (2000), we name this component thestatic shift effect.

The third term, that measures the cross-effect of the changes in labor productivity and labor shares, is the most difficult one to interpret. When the industries whose productivity levels grow rapidly also increase their share of employment, labor is reallocated to industries with rapid growth in productivity. Since it takes into account both labor productivity and labor share changes in the selected period, this term will be named thedynamic shift effect, again following Timmer and Szirmai (2000).7

The two shift effects measure the impact of structural change on aggregate labor productivity. One can measure the impact of sectoral shifts of labor on the aggregate productivity level in alternative but similar ways as well (Syrquin, 1986). The method adopted here is capable enough to sum- marize the impacts of labor reallocation on aggregate productivity. Consid- erably large positive sum of the two shift effects implies a favorable impact on aggregate labor productivity. Note that increases in labor quality reflect not only the improvement in the quality of labor due to in-house training by companies or restructuring within the firms, but also the changes in available

7In their analysis of productivity slowdown in the United States, Beebe and Haltmaier (1980) named the intra-industry and shift effects as “rate” and “level” effects, respectively.

capital per labor. Higher capital-labor ratio leads to higher labor productivity level. In turn, the shift effects can be expected to reflect the restructuring efforts of the government starting from 1979 aiming at the reallocation of resources in order to increase capital intensity of local industries. It is important to note that the shift effects are related to average productivity, not marginal product of labor. It is assumed here, for simplicity, that all workers in the same sector have the same productivity, i.e., average produc- tivity remains unchanged by inter-sectoral employment shifts. In addition, labor is assumed to be homogenous.8 Of interest here is the average pro- ductivity growth.