What Explains the Increase in the Labor Income Share in Malaysia?

Abstract

Labor income shares have been falling in many advanced and emerging economies within the last few decades, partly as a result of a combination of impacts from technology and increased global integration. This in turn is associated with the relatively slow growth of wages, especially for medium-skilled workers, and the worsening of the income inequality in these economies. In contrast, Malaysia’s labor income share has been increasing since 2005, together with a reduction in income inequality. We investigate this development by exploring the differences in trends of the labor income shares across different economic sectors and firm sizes and identifying factors that could explain the increase in the labor income share in Malaysia. We find that the increase is mainly due to the growing importance of more traditional service subsectors and SMEs in the economy. This in turn is associated with greater reliance on low-skilled foreign workers during this period. These findings have important policy implications for Malaysia, including the potential trade-off between driving labor productivity and fostering inclusiveness. This contrarian trend offers insights that could be relevant to the experiences of, and policy choices available to, other emerging economies facing deindustrialization.

Appendices

Appendix 1: Data Sources and Descriptions

The data that we use for calculating the various indicators come from various publications from the DOSM:

• 1. National Accounts: Gross Domestic Product Income Approach, 2010–2016

• 2. National Accounts: Annual National Accounts Gross Domestic Product (GDP), 2005–2016

• 3. National Accounts: Capital Stock Statistics, 2010–2016

• 4. Labour Force Survey, 2005–2016

• 5. Salaries & Wages Survey Report, 2010–2016

• 6. Economic Census Profile of Small and Medium Enterprises, 2010 and 2015

• 7. Economic Census, 2010 and 2015

• 8. Household Income Survey, 2016

We define the various intensities that we use as:

• 1. Foreign workers’ intensity: the ratio of the number of foreign workers to the total employed.

• 2. Capital intensity: the ratio of net real capital stock to the number of employments.

• 3. Machine and equipment intensity: the ratio of the machine and equipment component of the net capital stock to the total net capital stock.

Appendix 2: Shift-Share Analysis

$LIS$ = Labor income share

$W$ = GDP share

$i$ = Sector

$T$ = Final year (2016)

0 = Starting year (2005 or 2010)

Source: Adapted from Abdih and Danninger (2017).

Appendix 3: Twenty-One Economic Subsectors, Categorization of Workers by Skill Level, and Manufacturing and Service Subsectors

A. Twenty-One Economic Subsectors

Adapted from Malaysia Standard Industrial Classification (MSIC) 2008 version 1.0

• 1. Rubber, oil palm, livestock, and other agriculture

• 2. Forestry and logging

• 3. Fishing

• 4. Mining and quarrying

• 5. Food, beverages, and tobacco

• 6. Textiles, wearing apparel, and leather products

• 7. Wood products, furniture, paper products, and printing

• 8. Petroleum, chemical, rubber, and plastic products

• 9. Non-metallic mineral products, basic metal and fabricated metal products

• 10. Electrical, electronic, and optical products

• 11. Transport equipment, other manufacturing, and repair

• 12. Construction

• 13. Electricity, gas, steam, and air conditioning supply

• 14. Water supply, sewerage, waste management, and remediation activities

• 15. Wholesale and retail trade, repair of motor vehicles and motorcycles

• 16. Transportation and storage

• 17. Accommodation and food and beverage services

• 18. Information and communication

• 19. Financial and insurance/takaful activities

• 20. Real estate activities

• 21. Professional, scientific, and technical activities

B. High-, Medium-, and Low-Skilled Workers

Skill categorization is based on the DOSM and the Malaysia Standard Classification of Occupations (MASCO) 2013

High-Skilled Workers

• 1. Managers

• 2. Professionals

• 3. Technicians and associate professionals

Medium-Skilled Workers

• 1. Clerical support workers

• 2. Services and sales workers

• 3. Skilled agricultural, forestry, and fishery workers

• 4. Craft and related trade workers

• 5. Plant and machine operators and assemblers

Low-Skilled Workers

• 1. Elementary occupations

C. Low-, Medium (Mid-), and High-Tech Manufacturing

Modified in accordance with UNIDO’s classification, following the OECD technology classification based on R&D intensity relative to value added and gross production (ISIC categorization)

High-Tech

• 1. Electrical, electronic, and optical products

• 2. Transport equipment, other manufacturing, and repair

Medium (Mid-) Tech

• 1. Petroleum, chemical, rubber, and plastic products

• 2. Non-metallic mineral products, basic metal and fabricated metal products

Low-Tech

• 1. Food, beverages, and tobacco

• 2. Textiles, wearing apparel, and leather products

• 3. Wood products, furniture, paper products, and printing

Modern Services

Following the Asian Development Bank ( 2013 ), adapted from Eichengreen and Gupta ( 2009 ) based on labor productivity (ISIC categorization)

• 1. Information and communication

• 2. Financial and insurance activities

• 3. Real estate activities

• 4. Professional, scientific, and technical activities

Appendix 4: Formal Treatment of the Relevant Production Function

This section provides a detailed explanation of some key concepts underlying the channels through which the main drivers affect the labor income share, including the production function framework and the elasticity of substitution between capital and labor. The explanation below draws from the Estrada and Valdeolivas’ (2012) discussion.

The upward LIS trend observed in Malaysia—or the downward trend experienced elsewhere—signals the invalidity of unitary elasticity of substitution between capital and labor that conventional production functions, such as Cobb–Douglas, assume, as it implies a constant LIS. One way to rethink this is by considering a constant elasticity of substitution (CES) production function, which allows the elasticity of substitution between capital and labor to be different from one. In this case, should there be changes in the relative cost of either factor of production, the LIS would not be constant.

For this task, Arpaia et al. (2009) provided a comprehensive approach. Essentially, it considers and merges four production factors through a series of nested CES production functions, thus allowing for different elasticities of substitution among them.

Firstly, at the lower level of the production process is a CES function involving skilled labor ($L_{S}$) and capital ($AK$, where $A$ denotes a capital-augmenting technological process), which produces the composite input, denoted $X$, for the subsequent production function specified later.

$$X = \{ a\left( {AK} \right)^{{\frac{\eta - 1}{\eta }}} + \left( {1 - a} \right)(L_{S} )^{{\frac{\eta - 1}{\eta }}} \}^{{\frac{\eta - 1}{\eta }}}$$

η represents the elasticity of substitution between $L_{S}$ and $K$. If η is lower (higher) than 1, it implies that an increase in the supply of capital would increase (decrease) the share of skilled labor compensation (on the production of $X$). In other words, an η that is lower than one means that the two production factors are complements; if it is higher than one, they are substitutes.

The second CES function involves the combination of the previous composite input $(X$) and unskilled labor ($L_{U} )$ to generate value added ($Y$). $\rho$ is the new elasticity of substitution in this function, which allows for different degrees of complementarity between capital and the two types of labor.

$$Y = \{ \alpha \left( X \right)^{{\frac{\rho - 1}{\rho }}} + (1 - \alpha )(L_{U} )^{{\frac{\rho - 1}{\rho }}} \}^{{\frac{\rho - 1}{\rho }}}$$

Given the characterization of technology that the above CES production functions specify, we can infer that the LIS will depend, non-linearly, on four key variables, namely capital-augmenting technological progress, capital intensity, the unskilled–skilled labor ratio, and the capital–skilled labor ratio. How the LIS changes with respect to these variables depends on the degrees of substitutability between the different production factors laid out above. The remainder of this section focuses on explaining the conditions necessary to eventuate a positive impact on the LIS via these variables, which is what happens in Malaysia.

First, the condition for capital-augmenting technological progress to have a positive impact on the LIS is that composite input X and unskilled labor are substitutes. This implies that a negative shock of capital-augmenting technology increases the income share of unskilled labor. The income share of skilled labor, on the other hand, can either increase or decrease, since it is the product of the income share of the composite capital–skilled labor in the value added—which decreases under the previous condition—and the income share of skilled labor in the composite—the change of which depends on the elasticity of substitution between capital and skilled labor. When the two factors are complements, negative technological shocks will lead to a decrease in skilled labor’s income share. However, if the degree of complementarity is lower than the degree of substitutability between the unskilled labor and the composite, the decrease will not be enough to outweigh the increase in the unskilled labor income share. As a result, the overall labor income share will increase.

Second, the conditions to yield a positive impact on the LIS through capital intensity are essentially similar to those in the previous case, for the same reasons. In fact, the theoretical model indicates that the two variables should enter the model with the same parameter.

Third, for the unskilled–skilled labor ratio to affect the LIS positively, again, composite X and unskilled labor must be substitutes. In this case, it means that an increase in the ratio increases the unskilled labor income share. As for skilled labor, two counteracting forces are at play: less skilled workers will be employed but with a higher skill premium due to the decrease in supply. The two scenarios combine to result in an overall increase in the labor income share.

Lastly, the capital–skilled labor ratio has an unambiguously positive relationship with the LIS. In other words, when the capital supply decreases below the supply of skilled labor, the relative demand for skilled labor will drop correspondingly, exerting downward pressure on the wage premium and, thus, the labor income share of skilled labor. This mechanism, however, has no effect on the unskilled labor income share.

These conditions are important in understanding how changes in these variables led to the increase in the LIS in Malaysia during the past decade; the econometric analysis section in the paper formally establishes the relationship between these variables, except for the capital–skilled labor ratio due to the data limitation.¹³

Appendix 5: Further Details of the Econometric Analysis

This section explains in detail the econometric analysis that this study employs. The baseline estimation equation of the regression is as below:

$$Y_{it} = \alpha_{i} + \delta_{t} + \beta_{1} X '_{it} + \beta_{2} C_{it} + \varepsilon_{it}$$

where

• (i) $i$ denotes the sector and $t$ denotes the year;

• (ii) $Y_{it}$ is the dependent variable, that is, the labor income share;

• (iii) $X'_{it}$ is the vector of explanatory variables of interest, including foreign workers’ intensity, machine and equipment intensity, and capital intensity;

• (iv) $C_{it}$ is the additional independent variable, namely labor productivity;

• (v) $\alpha_{i}$ and $\delta_{t}$ are sector and year fixed effects, respectively; and

• (vi) $\varepsilon_{it}$ is the error term.

The main coefficients of interest are $\beta_{1}$, which capture the extent to which the corresponding variation in the potential determinants can explain the variation in the labor income share. The sector and year fixed effects essentially capture industry- and year-specific economic and social confounding factors.

Specification Tests

Because the time dimension, T, of the dataset is small, we perform the Harris–Tzavalis test¹⁴ to test for unit roots in the panel. For most of the variables, we cannot reject the hypothesis of the presence of unit roots at level; thus, we apply first differencing to obtain stationary series. We also estimate the model using standard errors clustered by industry to address the serial correlation concern in the panel. Besides, to detect the presence of random effects, we test the model for over-identifying restrictions—a Hausman-type test that is robust to heteroskedasticity and within-group correlation. The test finds no random effects in all the specifications of the model. We also test the joint significance of year-specific effects using the F-test. For all the specifications, we cannot reject the hypothesis that the year-specific effects are jointly statistically insignificant; therefore, we do not include year-specific effects in any of them.