Abstract

Using industrial census data for 2000, and smaller sets of survey data for 2001–2004, this paper examines the extent of wage differentials between medium-large (20 or more workers) foreign multinational enterprises (MNEs) and local plants in Malaysia's manufacturing industries. On average, wages in sample MNEs were higher than in local plants by two-fifths or more. In addition to being more capital-intensive and relatively large, MNEs also hired higher shares of workers in highly paid occupations and with moderate or high education. Results from large samples of 17 manufacturing industries combined suggest that statistically significant MNE–local differentials of 5%–9% persisted even after accounting for differences in worker occupation, education, and sex, plant capital intensity and size, as well as the influences of yearly fluctuations, industry affiliation, and plant location on the constants estimated. When MNE–local differentials and all slopes are allowed to vary among the 17 industries, positive and significant differentials were observed in all estimates for six industries: food and beverages, chemicals, rubber, general machinery, electrical machinery, and furniture. Positive and significant differentials were also observed in most estimates for another five industries. However, the size and significance of these differentials often varied depending on the industry and sample examined, as well as the estimation technique used.

I.  Introduction

There is now substantial evidence that foreign multinational enterprises (MNEs) tend to pay wages that exceed those of corresponding local firms or plants. There is also evidence that positive MNE–local wage differentials persist even after controlling for related differences between foreign MNEs and local firms or plants (e.g., differences in factor intensity, size, industry affiliation, and location). However, most previous studies of MNE–local wage differentials have been unable to account for the fact that MNEs tend to hire higher quality workers than local firms or plants. Lipsey and Sjöholm's (2004a) study of Indonesian plants in 1996, which finds that MNE–local differentials persist even after accounting for worker education, is the major exception. Moreover, aside from limited evidence in studies by Lim (1977) and Lee and Nagaraj (1995), and the descriptive statistics presented in Ramstetter (1995, 1999) and Ramstetter and Haji Ahmad (2011), there is very little evidence regarding MNE–local wage differentials in Malaysia.1

The Malaysian case is important because MNEs have been heavily involved in the manufacturing sector for many decades and Malaysian data include information on worker occupation in addition to worker education, which allows for more complete measurement of worker quality than in the Indonesian case. Data constraints have been an important reason for this important gap in the literature. This paper begins to fill that gap by using plant-level data from Malaysia's manufacturing census for 2000 and related surveys for 2001–2004 that allow for examination of MNE–local wage differentials after accounting for two important aspects of labor quality, educational background, and occupation.

The paper begins with a short literature review and a description of methodologies used to examine the question of whether MNEs pay higher or lower wages than their local counterparts (Section II), before examining the data and some descriptive statistics (Section III). Results of estimating Mincer-type equations, which reveal the extent of wage differentials after accounting for differences in labor force quality and other related plant characteristics, are then analyzed (Section IV). Finally, some conclusions and suggestions for future research are offered (Section V).

II.  Literature Review and Methodology

Lipsey and Sjöholm's (2004a) study of manufacturing plants in Indonesia in 1996 is perhaps the most sophisticated study of MNE–local wage differentials and their relationship to labor quality available for host, developing economies.2 They estimate Mincer-type equations for white-collar and blue-collar workers by industry and account for the influence of worker educational background, the share of female workers, as well as other plant-level characteristics related to wages (size, energy per worker, inputs per worker). They found that MNEs paid higher wages than local plants and that statistically significant wage differentials persisted after accounting for the educational background of the plant's work force and the other plant-level characteristics.

Ramstetter and Phan (2007) also found that positive wage differentials between MNEs and local, private firms in Viet Nam persisted after accounting for firm size, factor intensity, shares of technical workers, and shares of females in the firm's workforce. In contrast, results from Lee and Nagaraj's (1995) sample of workers in the Klang Valley of Malaysia in 1991 suggest that foreign ownership of a plant had no significant effects on wages of either male or female workers, after accounting for the influences of several measures of labor quality (education, experience, occupation, training) and numerous other worker-level and plant-level variables.3

Other studies of Malaysia (Lim 1977), Thailand (Movshuk and Matsuoka-Movshuk 2006, Ramstetter 2004), and Venezuela and Mexico (Aitken et al. 1996) have found that MNE–local wage differentials tended to persist after accounting for similar plant-level or firm-level characteristics, but were unable to account for the influences of labor force quality. In addition, Ismail and Haji Mat Zin (2003) and similar studies of workers in other economies reveal significant returns to human capital when measured by worker education, training, and experience, for example. Thus, there is substantial evidence that both plant ownership and worker quality have important influences on worker earnings. It is clear that relatively well-educated, experienced, and well-trained workers generally expect relatively high returns to their work efforts. Firms or plants hiring high-quality workers usually expect relatively high productivity from them and offer commensurate compensation.

The primary reason that MNEs pay higher wages than local plants is probably the well-documented tendency for MNEs to be relatively technology-intensive or skill-intensive compared to non-MNEs (Caves 2007, Dunning 1993, Markusen 2002). However, even relatively sophisticated studies like Lipsey and Sjöholm (2004a) probably fail to fully account for MNE–local differences in skill intensity. For example, in addition to differences in worker education, there may be important differences in worker occupation, training, background, and experience that are often accounted for in studies of wage determination among individuals. In this study of Malaysia, for example, it is possible to account for differences in education and occupation, but the available data do not contain information on worker background (e.g., race, nationality), training, or experience.

Other reasons for MNE–local differentials are perhaps less clear, but there are at least three important possibilities. First, there is substantial evidence that MNEs often find it difficult to identify and retain suitably qualified workers. For example, in 1998, securing adequate quantity and quality of labor was the third most common of 27 possible problems for Japanese affiliates operating in the ASEAN-4 (the four largest developing economies in the Association of Southeast Asian Nations, namely Indonesia, Malaysia, the Philippines, and Thailand), this problem being cited by 8.5% of these MNEs (METI Japan 2001, pp. 536–37).4 Other surveys also indicated that securing labor supply was the third most frequently cited of 14 investment motives of Japanese affiliates in Malaysia, being cited by 11%–13% of these firms during 2000–2004 (Toyo Keizai, various years).5 Correspondingly, many of the aforementioned studies suggest that MNEs may pay relatively high wages to secure or retain labor in economies like Malaysia.

Second, workers in host economies are often relatively familiar with management practices in local firms and may therefore be relatively reluctant to work for MNEs that often use less familiar management styles. This may lead them to demand a premium for working in the relatively unfamiliar MNE environment. Unfortunately, there is relatively little empirical evidence on this point, though many of the studies reviewed above mention it. On the other hand, relatively recent surveys of Malaysian university graduates suggest that MNEs are actually among the more popular employers for educated workers in this economy.6

Third, MNEs are often hypothesized to have important firm-specific assets in relatively large amounts compared to non-MNEs.7 These firm-specific assets are generally intangible, and many of them are related to worker quality. However, even when an MNE's intangible assets are not directly related to worker skills, they may facilitate higher worker productivity by improving a firm's marketing and management, for example. In other words, the MNE's possession of firm-specific assets has the potential to make workers more productive in MNEs than in non-MNEs, even if labor quality is identical in MNEs and non-MNEs. In such cases, MNEs may find it profitable to pay relatively high wages to compensate for their relatively high productivity, especially when the ability to utilize firm-specific assets is related to workers’ firm-specific experience or motivation, for example.

Partially reflecting differences in firm-specific assets, MNE–local wage differentials are thought to result from differences in other plant-level characteristics that might affect labor productivity and/or wages. For example, much of the literature reviewed above suggests that firms or plants that are relatively large or capital-intensive often pay relatively high wages and have relatively high labor productivity. In addition, location and industry affiliation are found to have important influences on the wage levels in firms or plants. Plants with relatively large shares of female workers often tend to pay relatively low average wages because females generally earn less than males, and Malaysia is no exception.8

Thus, this paper will follow the previous literature summarized above and estimate earnings equations that account for the influences of worker quality and gender, plant size, factor intensity, location, and industry affiliation, as well as ownership (MNE vs. local owners). Particular attention will be paid to labor quality, which can be measured in two ways with the data obtained. The industry dimension will also be carefully considered by the use of industry dummies in samples of all plants in 17 industries combined and by estimating separate equations for each industry (thereby allowing both intercepts and slopes to vary across industries).

III.  Data and Descriptive Statistics

This study employs the micro data underlying Malaysia's census of manufacturing plant activity in 2000 (Department of Statistics 2002) and smaller surveys of stratified samples for 2001–2004 (Department of Statistics various years). If samples are limited to plants with viable basic data (i.e., positive values of paid workers, output, worker compensation, and fixed assets), there were 18,799 plants in the 2000 census, but corresponding samples included only 11,898 to 13,197 plants in 2001–2004 (Ramstetter 2013, Appendix Table 1c). Most of the difference between the census and survey samples results from the census’ inclusion of small plants with limited production or employment. For example, if samples are limited to medium-large plants with 20 or more paid workers and viable basic data (positive output, worker compensation, and fixed assets), the 2000 census contained 8,540 plants and the subsequent surveys 7,406 to 7,581 plants. These medium-large plants accounted for 95%–97% of the paid workers (Table 1) and 97%–98% of the output (Ramstetter 2013, Appendix Table 1b) by all plants with viable basic data.

Table 1. 
Number of Paid Workers in All Plants with Viable Data and MNE Shares
Paid Workers in All Plants (1,000s)MNE Shares of Paid Workers (%)
Industry2000200120022003200420002001200220032004
Medium-large plants, 1,498 1,341 1,441 1,452 1,501 40 41 40 40 40 
  20+ workers           
17 sample industries 1,397 1,246 1,339 1,352 1,393 41 42 41 41 42 
Food and beverages 113.945 113.523 120.215 123.549 124.806 14 14 13 12 13 
Textiles 44.005 39.219 43.427 37.962 36.242 48 47 50 43 45 
Apparel 69.244 64.769 74.281 68.214 77.000 30 35 42 34 49 
Wood products 123.493 112.469 108.890 111.324 114.278 14 15 15 17 17 
Paper products 32.185 29.457 32.898 30.494 34.040 16 21 17 15 16 
Chemicals 45.406 42.147 47.076 48.989 51.367 38 41 41 38 39 
Rubber products 72.505 67.061 66.594 69.072 72.318 36 37 39 41 39 
Plastics 91.455 82.441 98.369 101.143 100.176 23 24 28 30 31 
Non-metallic 55.508 55.623 55.732 54.158 54.182 22 23 24 23 24 
  mineral products           
Basic metals 39.878 37.959 40.128 40.835 42.081 24 24 25 27 22 
Fabricated metals 57.730 52.093 61.743 63.662 68.045 28 26 24 27 24 
General machinery 46.561 45.091 45.420 44.943 51.062 52 47 41 38 40 
Office, computing, and 101.030 95.658 82.802 81.936 89.155 82 85 92 91 93 
  precision machinery           
Electrical machinery 77.121 60.444 65.876 67.848 67.425 54 62 50 60 59 
Radio, TV, and 330.140 250.308 279.046 292.317 285.123 70 78 74 73 73 
  communication           
  machinery           
Motor vehicles 33.700 38.999 46.739 47.776 50.597 16 12 10 15 16 
Furniture 63.286 59.226 69.289 67.987 75.369 20 18 17 14 14 
7 excluded industries 101 94 103 100 107 20 21 19 16 21 
Small plants, 73 48 45 43 39 
  1–19 workers           
Paid Workers in All Plants (1,000s)MNE Shares of Paid Workers (%)
Industry2000200120022003200420002001200220032004
Medium-large plants, 1,498 1,341 1,441 1,452 1,501 40 41 40 40 40 
  20+ workers           
17 sample industries 1,397 1,246 1,339 1,352 1,393 41 42 41 41 42 
Food and beverages 113.945 113.523 120.215 123.549 124.806 14 14 13 12 13 
Textiles 44.005 39.219 43.427 37.962 36.242 48 47 50 43 45 
Apparel 69.244 64.769 74.281 68.214 77.000 30 35 42 34 49 
Wood products 123.493 112.469 108.890 111.324 114.278 14 15 15 17 17 
Paper products 32.185 29.457 32.898 30.494 34.040 16 21 17 15 16 
Chemicals 45.406 42.147 47.076 48.989 51.367 38 41 41 38 39 
Rubber products 72.505 67.061 66.594 69.072 72.318 36 37 39 41 39 
Plastics 91.455 82.441 98.369 101.143 100.176 23 24 28 30 31 
Non-metallic 55.508 55.623 55.732 54.158 54.182 22 23 24 23 24 
  mineral products           
Basic metals 39.878 37.959 40.128 40.835 42.081 24 24 25 27 22 
Fabricated metals 57.730 52.093 61.743 63.662 68.045 28 26 24 27 24 
General machinery 46.561 45.091 45.420 44.943 51.062 52 47 41 38 40 
Office, computing, and 101.030 95.658 82.802 81.936 89.155 82 85 92 91 93 
  precision machinery           
Electrical machinery 77.121 60.444 65.876 67.848 67.425 54 62 50 60 59 
Radio, TV, and 330.140 250.308 279.046 292.317 285.123 70 78 74 73 73 
  communication           
  machinery           
Motor vehicles 33.700 38.999 46.739 47.776 50.597 16 12 10 15 16 
Furniture 63.286 59.226 69.289 67.987 75.369 20 18 17 14 14 
7 excluded industries 101 94 103 100 107 20 21 19 16 21 
Small plants, 73 48 45 43 39 
  1–19 workers           

Note: Plants with viable data are those with positive paid workers, output, worker compensation, and fixed assets. Excluded industries are tobacco, leather and footwear, publishing, petroleum products, other transport equipment, miscellaneous manufacturing, and recycling.

Source: Author's compilations from micro data underlying Department of Statistics (2002, various years).

Three types of ownership are identified in these data: majority-local plants, 50-50 joint ventures, and majority-foreign plants. MNEs are thus defined rather narrowly as plants with foreign ownership shares of 50% or more.9 MNEs are predominantly medium-large plants which differ from small, predominantly local plants in important ways.10 Thus, it is more meaningful to limit analyses of MNE–local wage differentials to medium-large plants than to include all plants. In addition to making sample size more consistent over time and allowing more meaningful MNE–local comparisons, dropping small plants has the important advantage of removing most outliers from the samples.11 The analysis also excludes seven relatively small industries with few MNEs, heterogeneous definitions, and/or heavy government regulation.12

The 17 industries included in the sample consist of one combination of two 2-digit categories (office, computing, and precision machinery), three 3-digit categories (rubber, plastics, furniture), and 13 2-digit categories (the remaining industries in Table 1). The criteria for defining the 17 industries were: (i) inclusion of competing plants in the same industry (because there are many multi-product plants in Malaysia, definitions are rather broad); (ii) industry size measured in terms of paid workers or output, and (iii) insuring sufficient samples of MNEs and local plants in each industry to facilitate reasonable comparisons. These 17 industries employed the vast majority of the paid workers in medium-large plants with viable basic data (89%–91%, Table 1).

In 2001, paid employment in the 17 sample industries was 11% lower than in 2000, partially because coverage was lower in survey years than in the census year. However, most of the decline in 2001 probably resulted from the dot.com-related recession, which led to a sharp fall in real manufacturing GDP.13 Its effects on the three electronics-related industries (office, computing, and precision machinery; electrical machinery; and radio, television, and communication machinery) were particularly large. In these three industries, paid employment declined by 101,881 workers, which was over two-thirds of the decline in all 17 industries (150,705 workers). These three industries accounted for about one-third or more of all paid employment in the 17 industries (36% in 2000, 32%–33% in other years) and are very important in Malaysia.

Radio, TV, and communication machinery was the largest of the 17 industries, employing 330,140 paid workers in 2000, 250,308 in 2001, and 270,046–292,317 in subsequent years (Table 1). MNEs accounted for 70%–78% of these workers and 82%–93% in office, computing, and precision machinery, which was another industry where maximum paid employment topped 100,000 (in 2000). Electrical machinery was a smaller employer (maximum 77,121) with relatively large MNE shares (54%–62%). However, in other industries with large paid employment (over 100,000 in at least one year), MNE shares were much smaller, 12%–14% in food and beverages, 14%–17% in wood, and 23%–31% in plastics. MNEs also played relatively large roles in several of the smaller industries examined in this study. For example, they accounted for 40% or more of the paid workers in at least one year in textiles, apparel, chemicals, rubber, and general machinery. In the 17 industries combined, MNEs accounted for just over two-fifths of paid employment (41%–42%), somewhat larger shares of manufacturing employment than in most of East Asia's large developing economies (Ramstetter 2012b).

MNEs generally paid higher wages, defined as compensation per worker (including payments in kind, social insurance payments, and other items), than local plants (Table 2).14 If all 17 industries are combined, this definition of wages was an average of 41%–46% higher in MNEs than in local plants. However, there was a large variation in wage levels and MNE–local wage differentials among industries. MNE wages were highest in chemicals in all years and increased from just over RM36,000 (just under $1,000) in 2000–2001 to RM45,241 in 2004. MNEs also paid relatively high wages in general machinery, non-metallic mineral products, food and beverages, and basic metals. At the other end of the scale, MNE wages were lowest in apparel, wood, and furniture, ranging from just under RM14,000 to about RM18,500.

Table 2. 
Mean Annual Compensation per Paid Worker in Medium-Large MNEs and MNE-Local Differentials
MNEs (current ringgit/year)MNE–Local Ratios
Industry2000200120022003200420002001200220032004
17 sample industries 23,205 25,128 25,987 27,092 27,910 1.407 1.464 1.451 1.440 1.432 
  combined           
Food and beverages 28,428 28,909 31,257 31,585 32,886 1.702 1.835 1.931 1.852 1.878 
Textiles 19,341 20,210 20,166 20,366 19,744 1.307 1.263 1.239 1.218 1.118 
Apparel 13,986 15,085 14,508 14,969 16,142 1.222 1.279 1.193 1.210 1.380 
Wood products 15,390 15,853 15,986 15,274 16,006 1.177 1.150 1.120 1.035 1.037 
Paper products 21,329 23,240 23,357 26,591 24,786 1.184 1.213 1.148 1.284 1.203 
Chemicals 36,959 36,524 40,813 43,183 45,241 1.473 1.401 1.463 1.469 1.502 
Rubber products 19,973 20,996 21,374 21,891 23,556 1.302 1.320 1.276 1.289 1.321 
Plastics 19,187 21,252 22,820 23,429 23,020 1.217 1.272 1.307 1.290 1.157 
Non-metallic mineral 28,669 30,594 29,863 34,416 34,133 1.614 1.675 1.570 1.673 1.589 
  products           
Basic metals 27,426 30,284 29,044 32,200 30,599 1.285 1.298 1.218 1.337 1.196 
Fabricated metals 24,148 24,538 24,138 25,411 26,015 1.398 1.330 1.297 1.288 1.269 
General machinery 29,093 32,604 34,257 37,092 36,321 1.409 1.417 1.435 1.552 1.467 
Office, computing, and 22,177 22,823 25,037 24,159 25,862 0.990 0.866 1.005 0.819 0.928 
  precision machinery           
Electrical machinery 19,207 21,746 21,907 22,944 23,312 1.117 1.169 1.073 1.094 1.041 
Radio, TV, and 21,109 24,046 22,766 23,790 24,426 1.145 1.261 1.205 1.111 0.992 
  communication           
  machinery           
Motor vehicles 25,510 23,890 26,642 29,720 28,633 1.424 1.204 1.281 1.397 1.271 
Furniture 15,141 16,507 16,839 18,530 18,455 1.188 1.263 1.180 1.307 1.247 
MNEs (current ringgit/year)MNE–Local Ratios
Industry2000200120022003200420002001200220032004
17 sample industries 23,205 25,128 25,987 27,092 27,910 1.407 1.464 1.451 1.440 1.432 
  combined           
Food and beverages 28,428 28,909 31,257 31,585 32,886 1.702 1.835 1.931 1.852 1.878 
Textiles 19,341 20,210 20,166 20,366 19,744 1.307 1.263 1.239 1.218 1.118 
Apparel 13,986 15,085 14,508 14,969 16,142 1.222 1.279 1.193 1.210 1.380 
Wood products 15,390 15,853 15,986 15,274 16,006 1.177 1.150 1.120 1.035 1.037 
Paper products 21,329 23,240 23,357 26,591 24,786 1.184 1.213 1.148 1.284 1.203 
Chemicals 36,959 36,524 40,813 43,183 45,241 1.473 1.401 1.463 1.469 1.502 
Rubber products 19,973 20,996 21,374 21,891 23,556 1.302 1.320 1.276 1.289 1.321 
Plastics 19,187 21,252 22,820 23,429 23,020 1.217 1.272 1.307 1.290 1.157 
Non-metallic mineral 28,669 30,594 29,863 34,416 34,133 1.614 1.675 1.570 1.673 1.589 
  products           
Basic metals 27,426 30,284 29,044 32,200 30,599 1.285 1.298 1.218 1.337 1.196 
Fabricated metals 24,148 24,538 24,138 25,411 26,015 1.398 1.330 1.297 1.288 1.269 
General machinery 29,093 32,604 34,257 37,092 36,321 1.409 1.417 1.435 1.552 1.467 
Office, computing, and 22,177 22,823 25,037 24,159 25,862 0.990 0.866 1.005 0.819 0.928 
  precision machinery           
Electrical machinery 19,207 21,746 21,907 22,944 23,312 1.117 1.169 1.073 1.094 1.041 
Radio, TV, and 21,109 24,046 22,766 23,790 24,426 1.145 1.261 1.205 1.111 0.992 
  communication           
  machinery           
Motor vehicles 25,510 23,890 26,642 29,720 28,633 1.424 1.204 1.281 1.397 1.271 
Furniture 15,141 16,507 16,839 18,530 18,455 1.188 1.263 1.180 1.307 1.247 

Note: Plants with viable data are those with positive paid workers, output, worker compensation, and fixed assets. Excluded industries are tobacco, leather and footwear, publishing, petroleum products, other transport equipment, miscellaneous manufacturing, and recycling.

Source: Author's compilations from micro data underlying Department of Statistics (2002, various years).

MNE–local differentials tended to be largest in the industries where MNEs paid relatively high wages: food and beverages (70%–93%), non-metallic mineral products (57%–68%), chemicals (40%–50%), and general machinery (41%–55%). On the other hand, negative differentials were only observed in office, computing, and precision machinery (2000–2001 and 2003–2004); and radio, TV, and communication machinery (2004). Wage differentials were positive but relatively small in electrical machinery and wood (4%–18%). MNE–local wage differentials were positive in 15 of the 17 industries in all years, exceeding 10% in the vast majority (14–16) of sample industries and 20% in a majority (10–13 industries).

Not surprisingly, similarly large variation across industries is observed in several measures of labor quality, measured either in terms of worker occupation or educational achievement, though variation of worker quality is less pronounced over time. If all 17 industries are combined, the mean share of paid workers in highly paid occupations, defined as management (including proprietors), professionals, and technicians, in paid workers, was 20% in MNEs in 2000 and 22%–23% thereafter (Table 3).15 These shares were 28%–37% higher than in local plants (ratios of 1.28–1.37 in Table 3). Consistent with the notion that wages are closely related to labor quality, the industry distribution of shares of highly paid workers was similar to the industry distribution of wages. For example, in MNEs, the share of workers in highly paid occupations varied from 9%–10% in apparel, 9%–11% in wood, and 12%–17% in furniture to 27%–32% in general machinery, 23%–26% in food and beverages, 33%–36% in chemicals, and 26%–30% in basic metals. MNE–local differentials for shares of highly paid workers were positive (ratios in excess of 1) in all 17 industries in 2000, 14–15 industries in 2001–2003, and 11 industries in 2004. However, these differentials were generally smaller than corresponding wage differentials. Relatively large differentials exceeding 10% were observed in only 8 industries in 2004 and 11–13 industries in other years.

Table 3. 
Mean Shares of Paid Workers with Highly Paid Occupations in Sample MNEs and MNE–Local Differentials
MNEs (% of paid workers)MNE–Local Ratios
Industry2000200120022003200420002001200220032004
17 sample industries 20 22 22 23 22 1.369 1.338 1.282 1.360 1.276 
  combined           
Food and beverages 23 25 26 25 24 1.456 1.513 1.474 1.504 1.436 
Textiles 17 19 19 19 16 1.150 1.205 1.061 1.128 0.978 
Apparel 10 1.104 0.904 0.969 0.886 0.555 
Wood products 11 10 11 10 1.022 0.949 0.908 0.856 0.932 
Paper products 19 20 22 24 21 1.212 1.250 1.262 1.365 1.244 
Chemicals 33 33 35 36 34 1.378 1.272 1.281 1.277 1.270 
Rubber products 15 16 16 15 16 1.129 1.107 1.115 1.114 1.117 
Plastics 19 22 20 20 18 1.286 1.234 1.137 1.221 1.036 
Non-metallic mineral 22 22 21 26 24 1.468 1.333 1.185 1.567 1.410 
  products           
Basic metals 26 25 27 30 26 1.407 1.185 1.228 1.395 1.317 
Fabricated metals 20 20 21 22 20 1.286 1.129 1.106 1.183 1.045 
General machinery 27 29 30 32 31 1.401 1.505 1.295 1.450 1.329 
Office, computing, and 20 20 22 24 24 1.084 0.905 1.129 1.034 0.946 
  precision machinery           
Electrical machinery 18 20 21 20 20 1.083 1.087 1.063 0.983 0.952 
Radio, TV, and 20 23 23 22 23 1.239 1.150 1.028 1.019 0.962 
  communication           
  machinery           
Motor vehicles 20 20 22 24 22 1.177 1.067 1.107 1.321 1.198 
Furniture 12 12 13 17 13 1.080 1.096 1.111 1.353 1.028 
MNEs (% of paid workers)MNE–Local Ratios
Industry2000200120022003200420002001200220032004
17 sample industries 20 22 22 23 22 1.369 1.338 1.282 1.360 1.276 
  combined           
Food and beverages 23 25 26 25 24 1.456 1.513 1.474 1.504 1.436 
Textiles 17 19 19 19 16 1.150 1.205 1.061 1.128 0.978 
Apparel 10 1.104 0.904 0.969 0.886 0.555 
Wood products 11 10 11 10 1.022 0.949 0.908 0.856 0.932 
Paper products 19 20 22 24 21 1.212 1.250 1.262 1.365 1.244 
Chemicals 33 33 35 36 34 1.378 1.272 1.281 1.277 1.270 
Rubber products 15 16 16 15 16 1.129 1.107 1.115 1.114 1.117 
Plastics 19 22 20 20 18 1.286 1.234 1.137 1.221 1.036 
Non-metallic mineral 22 22 21 26 24 1.468 1.333 1.185 1.567 1.410 
  products           
Basic metals 26 25 27 30 26 1.407 1.185 1.228 1.395 1.317 
Fabricated metals 20 20 21 22 20 1.286 1.129 1.106 1.183 1.045 
General machinery 27 29 30 32 31 1.401 1.505 1.295 1.450 1.329 
Office, computing, and 20 20 22 24 24 1.084 0.905 1.129 1.034 0.946 
  precision machinery           
Electrical machinery 18 20 21 20 20 1.083 1.087 1.063 0.983 0.952 
Radio, TV, and 20 23 23 22 23 1.239 1.150 1.028 1.019 0.962 
  communication           
  machinery           
Motor vehicles 20 20 22 24 22 1.177 1.067 1.107 1.321 1.198 
Furniture 12 12 13 17 13 1.080 1.096 1.111 1.353 1.028 

Note: Plants with viable data are those with positive paid workers, output, worker compensation, and fixed assets. Excluded industries are tobacco, leather and footwear, publishing, petroleum products, other transport equipment, miscellaneous manufacturing, and recycling. Highly paid occupations are (i) proprietors, business partners, (ii) managers, professionals, executives, and (iii) technicians, professionals.

Source: Author's compilations from micro data underlying Department of Statistics (2002, various years).

Combining all 17 industries, highly educated workers (those with some kind of tertiary education) accounted for 16% of all workers in MNEs in 2000 and this share rose to 19% in 2003–2004 (Table 4). Similar to shares of highly paid workers, shares of highly educated workers were highest in chemicals and general machinery (averaging 30%–31%), followed by non-metallic mineral products; radio, TV, and communication machinery; and basic metals. These shares were often substantially higher in MNEs than in local plants. If all 17 industries were combined, these differentials computed to 61% in 2000, 83%–90% in 2001–2003, and 71% in 2004. At the industry level, MNEs had higher shares in 12–15 of the 17 industries. Moreover, differentials exceeded 10% in 11–14 industries and 20% in 9–13 industries. In other words, MNEs clearly hired relatively large shares of highly educated workers compared to local plants.

Table 4. 
Mean Shares of All (Paid and Unpaid) Workers with High Education in Sample MNEs and MNE–Local Differentials
MNEs (% of all workers)MNE–Local Ratios
Industry2000200120022003200420002001200220032004
17 sample industries 16 17 18 19 19 1.611 1.879 1.901 1.828 1.706 
  combined           
Food and beverages 17 17 19 19 19 2.113 2.435 2.658 2.563 2.370 
Textiles 10 0.933 0.928 0.824 0.914 0.889 
Apparel 1.068 1.271 1.030 0.956 0.629 
Wood products 1.426 1.488 1.488 1.201 1.143 
Paper products 16 15 17 18 16 1.312 1.424 1.442 1.368 1.225 
Chemicals 28 29 31 32 34 1.517 1.677 1.626 1.582 1.557 
Rubber products 11 10 10 10 1.200 1.283 1.129 1.074 1.139 
Plastics 14 16 18 17 15 1.292 1.593 1.625 1.545 1.209 
Non-metallic mineral 19 19 19 25 21 2.000 2.288 2.203 2.797 2.029 
  products           
Basic metals 18 17 18 20 18 1.268 1.337 1.426 1.464 1.269 
Fabricated metals 16 16 17 18 18 1.258 1.578 1.762 1.694 1.643 
General machinery 26 28 31 35 31 1.996 2.207 2.086 2.448 1.930 
Office, computing, and 16 15 16 19 21 0.975 0.712 0.820 0.695 0.731 
  precision machinery           
Electrical machinery 14 17 18 17 17 0.973 1.288 1.233 0.972 0.969 
Radio, TV, and 16 19 18 18 20 1.048 1.234 1.026 0.983 0.953 
  communication           
  machinery           
Motor vehicles 15 14 19 20 18 1.182 1.124 1.340 1.548 1.375 
Furniture 0.986 0.911 1.236 1.379 1.152 
MNEs (% of all workers)MNE–Local Ratios
Industry2000200120022003200420002001200220032004
17 sample industries 16 17 18 19 19 1.611 1.879 1.901 1.828 1.706 
  combined           
Food and beverages 17 17 19 19 19 2.113 2.435 2.658 2.563 2.370 
Textiles 10 0.933 0.928 0.824 0.914 0.889 
Apparel 1.068 1.271 1.030 0.956 0.629 
Wood products 1.426 1.488 1.488 1.201 1.143 
Paper products 16 15 17 18 16 1.312 1.424 1.442 1.368 1.225 
Chemicals 28 29 31 32 34 1.517 1.677 1.626 1.582 1.557 
Rubber products 11 10 10 10 1.200 1.283 1.129 1.074 1.139 
Plastics 14 16 18 17 15 1.292 1.593 1.625 1.545 1.209 
Non-metallic mineral 19 19 19 25 21 2.000 2.288 2.203 2.797 2.029 
  products           
Basic metals 18 17 18 20 18 1.268 1.337 1.426 1.464 1.269 
Fabricated metals 16 16 17 18 18 1.258 1.578 1.762 1.694 1.643 
General machinery 26 28 31 35 31 1.996 2.207 2.086 2.448 1.930 
Office, computing, and 16 15 16 19 21 0.975 0.712 0.820 0.695 0.731 
  precision machinery           
Electrical machinery 14 17 18 17 17 0.973 1.288 1.233 0.972 0.969 
Radio, TV, and 16 19 18 18 20 1.048 1.234 1.026 0.983 0.953 
  communication           
  machinery           
Motor vehicles 15 14 19 20 18 1.182 1.124 1.340 1.548 1.375 
Furniture 0.986 0.911 1.236 1.379 1.152 

Note: Plants with viable data are those with positive paid workers, output, worker compensation, and fixed assets. Excluded industries are tobacco, leather and footwear, publishing, petroleum products, other transport equipment, miscellaneous manufacturing, and recycling. High education is defined as education beyond the Malaysian Education Certificate (SPM), an exam taken by all students in the fifth year of secondary school (i.e., some level of vocational school, college, university, or graduate school).

Source: Author's compilations from micro data underlying Department of Statistics (2002, various years).

MNEs also tended to hire relatively large shares of moderately educated workers (those who completed the Malaysian Education Certificate or SPM but did not continue to tertiary education, Table 5). If all 17 industries are combined, these workers accounted for just under two-fifths of all employees in MNEs, and these shares were 20%–37% higher than in local plants. The highest MNE shares were observed in motor vehicles; radio, TV, and communication machinery; office, computing, and precision machinery; paper products; and electrical machinery. In other words, MNEs were relatively large employers in the three electronics-related industries, and they also hired relatively large shares of moderately educated workers in these industries.

Table 5. 
Mean Shares of All (Paid and Unpaid) Workers with Mid-level Education in Sample MNEs and MNE–Local Differentials
MNEs (% of paid & unpaid workers)MNE–Local Ratios
Industry2000200120022003200420002001200220032004
17 sample industries 36 39 39 39 39 1.201 1.369 1.355 1.341 1.283 
  combined           
Food and beverages 35 37 35 34 38 1.295 1.372 1.344 1.236 1.328 
Textiles 32 34 40 36 38 1.075 1.013 1.219 1.134 1.137 
Apparel 16 22 20 20 24 0.781 1.066 1.012 0.884 0.816 
Wood products 17 22 21 22 21 0.921 1.180 1.083 1.162 1.100 
Paper products 39 41 44 45 43 1.109 1.250 1.290 1.282 1.179 
Chemicals 41 41 42 42 40 1.104 1.161 1.131 1.074 1.033 
Rubber products 28 32 34 31 35 1.049 1.178 1.309 1.203 1.282 
Plastics 35 42 41 40 38 1.024 1.316 1.337 1.332 1.181 
Non-metallic mineral 33 35 34 34 34 1.275 1.392 1.425 1.416 1.412 
  products           
Basic metals 38 45 46 39 37 1.127 1.344 1.336 1.230 1.160 
Fabricated metals 39 41 40 38 38 1.111 1.378 1.394 1.275 1.277 
General machinery 39 41 40 40 40 1.104 1.085 1.098 1.088 1.044 
Office, computing, and 44 43 42 48 47 0.937 0.948 0.864 1.205 1.219 
  precision machinery           
Electrical machinery 34 43 41 45 45 0.794 1.101 1.042 1.116 1.101 
Radio, TV, and communication 43 48 46 49 48 1.007 1.294 1.119 1.146 1.064 
  machinery           
Motor vehicles 47 51 48 49 47 1.217 1.360 1.231 1.249 1.185 
Furniture 20 24 20 24 21 0.711 1.015 0.877 0.992 0.809 
MNEs (% of paid & unpaid workers)MNE–Local Ratios
Industry2000200120022003200420002001200220032004
17 sample industries 36 39 39 39 39 1.201 1.369 1.355 1.341 1.283 
  combined           
Food and beverages 35 37 35 34 38 1.295 1.372 1.344 1.236 1.328 
Textiles 32 34 40 36 38 1.075 1.013 1.219 1.134 1.137 
Apparel 16 22 20 20 24 0.781 1.066 1.012 0.884 0.816 
Wood products 17 22 21 22 21 0.921 1.180 1.083 1.162 1.100 
Paper products 39 41 44 45 43 1.109 1.250 1.290 1.282 1.179 
Chemicals 41 41 42 42 40 1.104 1.161 1.131 1.074 1.033 
Rubber products 28 32 34 31 35 1.049 1.178 1.309 1.203 1.282 
Plastics 35 42 41 40 38 1.024 1.316 1.337 1.332 1.181 
Non-metallic mineral 33 35 34 34 34 1.275 1.392 1.425 1.416 1.412 
  products           
Basic metals 38 45 46 39 37 1.127 1.344 1.336 1.230 1.160 
Fabricated metals 39 41 40 38 38 1.111 1.378 1.394 1.275 1.277 
General machinery 39 41 40 40 40 1.104 1.085 1.098 1.088 1.044 
Office, computing, and 44 43 42 48 47 0.937 0.948 0.864 1.205 1.219 
  precision machinery           
Electrical machinery 34 43 41 45 45 0.794 1.101 1.042 1.116 1.101 
Radio, TV, and communication 43 48 46 49 48 1.007 1.294 1.119 1.146 1.064 
  machinery           
Motor vehicles 47 51 48 49 47 1.217 1.360 1.231 1.249 1.185 
Furniture 20 24 20 24 21 0.711 1.015 0.877 0.992 0.809 

Note: Plants with viable data are those with positive paid workers, output, worker compensation, and fixed assets. Excluded industries are tobacco, leather and footwear, publishing, petroleum products, other transport equipment, miscellaneous manufacturing, and recycling. Mid-level education is defined as successful completion of the Malaysian Education Certificate (SPM), an exam taken by all students in the fifth year of secondary school, but no further education.

Source: Author's compilations from micro data underlying Department of Statistics (2002, various years).

MNE–local differentials were again positive in most industries, 12 in 2000 and 15–16 industries in subsequent years. Differentials exceeded 10% in eight industries in 2000 and 11–13 industries in subsequent years. Thus, MNEs hired relatively large shares of both moderately and highly educated workers, and MNE–local differentials were substantially larger for highly educated worker shares than for moderately educated worker shares. Conversely, Tables 4 and 5 also make it clear that MNEs hired relatively small shares of workers who did not complete the SPM and are considered relatively unskilled in Malaysia.16

Information on the race of Malaysian workers and the number of and nationality of foreign workers are two other indicators potentially related to labor quality which were collected in the 2000 and 2005 censuses (and probably in the surveys for interim years).17 Probably because this information is sensitive in Malaysia, it was not included in the data sets provided for this and related studies. The lack of information on foreign workers is probably of most consequence because the number of foreign workers in manufacturing plants rose rapidly during the period studied, from 219,633 or 14% of all manufacturing workers in 2000 to 366,095 or 22% of the total in 2005 (World Bank 2013, pp. 24 and 193).

Because foreign workers tend to be relatively unskilled and paid less than Malaysian nationals, MNE–local wage differentials may also be related to differences in foreign worker shares in MNEs and local plants. Many foreign workers are also known to work in the relatively labor-intensive, MNE-dominated electronics-related machinery industries (Bormann et al. 2010). However, I know of no published information on the extent to which shares of foreign workers differ between medium-large MNEs and local plants. Moreover, it is important to note that information on race, nationality, and sex of workers reflects both aspects of worker quality and the extent to which a group is discriminated against in the workplace or in educational institutions, for example.

In summary, these industry-level compilations suggest MNEs tend to pay relatively high wages, as well as hire relatively large portions of workers in high-paid occupations and with moderate or high education. Previous compilations also suggest that MNEs tend to be much larger and somewhat more capital-intensive than local plants during this period (Ramstetter and Haji Ahmad 2011). These differences can all explain at least part of the observed wage differentials between MNEs and local plants. It is thus of interest to examine whether wage differentials persist after accounting for these differences between MNEs and local plants.

IV.  Results of Estimating Earnings Equations

Using a specification similar to that of Lipsey and Sjöholm (2004a), the extent to which MNE–local wage differentials persist after accounting for the influences of plant-level capital intensity and size, as well as worker occupation, education, and sex, can be investigated by including a dummy variable identifying MNE plants in a typical earnings equation such as the following:

formula
1

where

LCE = log of compensation per employee in a plant (2000 ringgit)

LKE = log of fixed assets per employee in a plant (2000 ringgit)

LO = plant size, measured as the log of output in a plant (2000 ringgit)

SH = share of paid workers in highly paid occupations in a plant (%)

S3 = share of paid and unpaid workers with high education in a plant (%)

S2 = share of paid and unpaid workers with mid-level education in a plant (%)

SF = share of paid workers that are female in a plant (%)

DF = dummy variable identifying MNE plants (=1 if MNE, 0 otherwise)

As in Table 2, the dependent variable is defined to include all compensation to employees including all bonuses, payments in kind, social insurance payments, and other compensation. Nominal wages are converted to real values using the consumer price index. Reflecting previous discussion, signs of the coefficients on capital intensity (a1) and size (a2) are generally expected to be positive because capital-intensive and larger plants are generally expected to have more productive and better paid workers than smaller, more labor-intensive ones. Capital intensity and output are converted to real values using GDP deflators for 24 industries, which were generally defined at the 2-digit or 3-digit level (Department of Statistics 2011a).18

Plants with higher quality workforces are expected to pay relatively high wages and the coefficients a3, a4, and a5 are thus expected to be positive. In contrast, the coefficient a6 is expected to be negative because females generally earn less than their male counterparts.19 If data were available, it would be interesting to include the share of foreign workers in this equation, but this is not possible. To the extent that there are MNE–local differences in shares of foreign workers, worker experience, and training (which are not available), estimates of Equation (1) may face an omitted variable problem.20 Finally, if MNE–local wage differentials persist after accounting for capital intensity, size, as well as worker occupation, education, and sex, a7 will be positive.

Plant-level panels are compiled using plant identity codes in the data and estimated equations also include sets of year, region, and industry dummies to account for year-specific, region-specific, and industry-specific influences on the constant that are not captured by the plant-level variables. Year-specific dummies use the first year in each sample as the base and estimates are performed for both 2000–2004 and 2001–2004. These alternative samples are used to examine the sensitivity of the results to the inclusion of the census year and to facilitate comparisons of a contemporaneous specification with a lagged specification where all independent variables are lagged one year. The lagged specification is used because it is less likely to be affected by simultaneity issues. Although simultaneity may not be a large problem in Equation (1) because wage levels are not likely to be an important determinant of the right-hand side variables, the lagged specification provides an important robustness check.21

Region dummies are usually defined at the state level, but the lack of observations makes it necessary to combine some states when performing industry-level estimates. In three cases, states with similar population densities and nearby locations are combined (Perlis and Kedah; Kelantan, Terengganu, and Pahang; and finally Sabah, Sarawak, and Labuan). As a result, there are usually 10 regions with nine dummies defined using Kuala Lumpur as the base.22 Industry dummies are defined at the 3-digit level. Some industries are defined at the 3-digit level and other 3-digit subcategories contain too few plants to facilitate estimation; thus, industry dummies are excluded from several of the industry-level regressions.23 Because the descriptive data and previous research for Indonesia and Thailand suggest large variations in MNE–local wage differentials among industries, estimates are performed individually for 17 industries and compared to results for the 17 industries combined. In general, the industry-level results are thought to be more reliable than results for all 17 industries combined.

Results of pooled ordinary least squares (OLS) and random effects panel estimates are also compared to evaluate the robustness of the results to alternative econometric assumptions. Fixed effects panel estimates are often used with data such as these, but they cannot reveal the extent of MNE–local wage differentials because ownership is time invariant for most plants. Hence, if a fixed effects estimator is used, the coefficient on the MNE dummy a7 measures the wage differential between plants changing ownership and those with constant ownership, not the MNE–local wage differential, which is the primary concern in this paper. The pooled OLS results are the easiest to interpret because the coefficient a7 is approximately equal to the percentage difference in wages between MNEs and local plants. When a random effects estimator is used, the presence of individual effects complicates the interpretation of this coefficient, but it is still important to examine how sensitive the results are to alternative econometric assumptions.

Table 6 presents the slope coefficients, a goodness of fit measure (R2), and the results of the Breusch-Pagan test for random effects for estimates in combined samples of all 17 sample industries. The Breusch-Pagan test is always highly significant at the 1% level or better, both in these combined estimates and in the industry-level regressions discussed below. This implies that random effects should be preferred to OLS, but I remain interested in checking the robustness of the results to alternative assumptions. Robust standard errors are used to address potential heteroskedasticity, and there were no indications of multicollinearity, which might be suspected among shares of workers with high education and shares in highly paid occupations, for example.24

Table 6. 
Multinational-Local Wage Differentials, Other Slope Coefficients, and Equation Indicators from Estimates of Equation (1) for all 17 Sample Industries Combined
Pooled OLSRandom Effects
LaggedContemporaneousLaggedContemporaneous
Slope coefficient
variable, indicator2001–20042001–20042000–20042001–20042001–20042000–2004
 0.0242*** 0.0329*** 0.0338*** 0.0183*** 0.0360*** 0.0367*** 
 0.1071*** 0.1178*** 0.1187*** 0.1032*** 0.1229*** 0.1264*** 
 0.0074*** 0.0070*** 0.0082*** 0.0037*** 0.0061*** 0.0074*** 
       
 0.0064*** 0.0072*** 0.0060*** 0.0042*** 0.0064*** 0.0049*** 
       
 0.0011*** 0.0011*** 0.0005*** 0.0006*** 0.0007*** 0.0001 
       
       
 –0.0039*** –0.0035*** –0.0036*** –0.0032*** –0.0026*** –0.0025*** 
       
 0.0890*** 0.0809*** 0.0913*** 0.0749*** 0.0525*** 0.0658*** 
       
       
R2 0.5591 0.5735 0.5638 0.5454 0.5683 0.5579 
Observations 21,671 26,855 34,491 21,671 26,855 34,491 
Breusch-Pagan Test – – – 8,254*** 10,202*** 14,135*** 
Pooled OLSRandom Effects
LaggedContemporaneousLaggedContemporaneous
Slope coefficient
variable, indicator2001–20042001–20042000–20042001–20042001–20042000–2004
 0.0242*** 0.0329*** 0.0338*** 0.0183*** 0.0360*** 0.0367*** 
 0.1071*** 0.1178*** 0.1187*** 0.1032*** 0.1229*** 0.1264*** 
 0.0074*** 0.0070*** 0.0082*** 0.0037*** 0.0061*** 0.0074*** 
       
 0.0064*** 0.0072*** 0.0060*** 0.0042*** 0.0064*** 0.0049*** 
       
 0.0011*** 0.0011*** 0.0005*** 0.0006*** 0.0007*** 0.0001 
       
       
 –0.0039*** –0.0035*** –0.0036*** –0.0032*** –0.0026*** –0.0025*** 
       
 0.0890*** 0.0809*** 0.0913*** 0.0749*** 0.0525*** 0.0658*** 
       
       
R2 0.5591 0.5735 0.5638 0.5454 0.5683 0.5579 
Observations 21,671 26,855 34,491 21,671 26,855 34,491 
Breusch-Pagan Test – – – 8,254*** 10,202*** 14,135*** 

*** = significant at the 1% level, ** = significant at the 5% level, * = significant at the 10% level.

LKE = capital intensity, LO = output scale, SH = highly paid share of paid workers, S3 = highly educated share of all workers, S2 = moderately educated share of all workers, SF = female share of paid workers, DF = MNE–local differential (ratio less 1).

Note: Robust standard errors (clustered by plant for random effects) are used to account for potential heteroskedasticity. Results of the Breusch-Pagan Test (null of no random effects) is always rejected at the 1% level. These results come from estimates that also include year, industry, and region dummies. Full results including constants and coefficients on year, industry, and region dummies are available from the author.

Source: Author's estimates from microdata underlying Department of Statistics (2002, various years).

In this combined sample, coefficients on capital intensity, output, and the three measures of labor quality were positive as hypothesized and highly significant in almost all estimates. The only exception was in the contemporaneous specification for 2000–2004, when the coefficient on the share of moderately educated workers became insignificant. The coefficient on the female share was negative as hypothesized and highly significant. R2 ranged from 0.55 to 0.57, indicating that these estimates described the variation in the dependent variable reasonably well. The coefficient a7 was positive and highly significant in all estimates, indicating that MNE–local wage differentials among all medium-large plants remained positive and meaningful statistically, even if the influences of capital intensity and size, as well as worker occupation, education, and sex, are accounted for. Not surprisingly, however, accounting for these influences greatly reduced the size of MNE–local wage differentials from over 40% (Table 2) to 8%–9% when estimated by pooled OLS and 5%–7% when estimated by random effects.

Space constraints prevent the presentation of all slope coefficients for all 17 industry-level estimates but they are available in Ramstetter (2013, Appendix Table 6). The output variable was the only one that was consistently positive and significant at the 5% level or better in all 102 estimates performed. Positive and significant coefficients were also common for shares of highly paid and highly educated workers as well as capital intensity (91, 85, and 79 of the estimates, respectively). The coefficient on the female share was also negative and significant in most industry-level estimates (74). The equation fit the data the worst in furniture, where R2 ranged from 0.30 to 0.38. R2 was 0.39 or higher in all other industries. The smallest sample was 322 observations in the lagged specification for office, computing, and precision machinery, and most samples exceeded 1,000. In short, for these 17 industries, all samples were sufficiently large to allow relatively reliable analysis, and the estimates explained the variation of wages relatively well.

Table 7 presents estimates of the MNE–local wage differentials from industry-level estimates of Equation (1). As in Table 2, wage differentials often differ greatly across industries. These wage differentials were positive and significant at the 5% level or better in all estimates for six of the 17 industries: food and beverages, chemicals, rubber, general machinery, electrical machinery, and furniture. Estimated differentials tended to be largest in rubber (21%–23% if estimated by pooled OLS and 11%–15% if estimated by random effects). Estimated differentials were also large and significant in five of the six estimates for motor vehicles (11%–16% and 13%–16%, respectively). Consistently significant differentials were smallest in chemicals (9%–10% and 7%–9%, respectively) and furniture (10%–11% and 7%–11%, respectively). Differentials were also positive and significant in four of the six estimates for another four industries: apparel; plastics; non-metallic mineral products; and radio, TV, and communication machinery. In another three industries (textiles, paper, and basic metals) OLS results suggested positive and significant differentials, while random effects results suggested no significant differentials. Finally, there were no significant differentials in two industries (wood and office, computing, and precision machinery), and there was only weak evidence of positive differentials in fabricated metals. Negative and significant differentials were never observed.

Table 7. 
Industry-level Estimates of Multinational-Local Wage Differentials after Controlling for Capital Intensity, Size, Labor Occupation, Labor Skill, and Labor Gender from Equation (1)
Pooled OLSRandom Effects
LaggedContemporaneousLaggedContemporaneous
Industry2001–20042001–20042000–20042001–20042001–20042000–2004
Food and beverages 0.1069*** 0.1154*** 0.1176*** 0.1104*** 0.1241*** 0.1075*** 
Textiles 0.1084*** 0.1155*** 0.1196*** 0.0269 0.0804* 0.0627 
Apparel 0.0764*** 0.0433 0.0499** 0.0922** 0.0551 0.1138*** 
Wood products 0.0040 –0.0084 0.0047 0.0203 –0.0170 0.0047 
Paper products 0.0805** 0.0661** 0.0710*** 0.0715 0.0496 0.0206 
Chemicals 0.0950*** 0.0858*** 0.0958*** 0.0679** 0.0858*** 0.0906*** 
Rubber products 0.2259*** 0.2076*** 0.2195*** 0.1470*** 0.1121*** 0.1215*** 
Plastics 0.0798*** 0.0749*** 0.0906*** 0.0579** 0.0266 0.0409* 
Non-metallic mineral 0.1077*** 0.0728** 0.0845*** 0.0865** 0.0142 0.0317 
  products       
Basic metals 0.1023*** 0.1153*** 0.1124*** 0.0436 0.0477 0.0462 
Fabricated metals 0.0408* 0.0316 0.0692*** 0.0597* 0.0336 0.0743** 
General machinery 0.0911** 0.0762** 0.0755*** 0.1141** 0.1282*** 0.1365*** 
Office, computing, and –0.0140 –0.0503 –0.0273 0.0047 –0.0234 –0.0379 
  precision machinery       
Electrical machinery 0.1020*** 0.0903*** 0.1126*** 0.0604** 0.0789** 0.0862*** 
Radio, TV, and 0.0754*** 0.0633*** 0.0460*** 0.0719*** 0.0435* 0.0284 
  communication machinery       
Motor vehicles 0.1091** 0.1361*** 0.1617*** 0.1267** 0.1270* 0.1605** 
Furniture 0.0993*** 0.1057*** 0.1090*** 0.1103*** 0.0690** 0.0702** 
Pooled OLSRandom Effects
LaggedContemporaneousLaggedContemporaneous
Industry2001–20042001–20042000–20042001–20042001–20042000–2004
Food and beverages 0.1069*** 0.1154*** 0.1176*** 0.1104*** 0.1241*** 0.1075*** 
Textiles 0.1084*** 0.1155*** 0.1196*** 0.0269 0.0804* 0.0627 
Apparel 0.0764*** 0.0433 0.0499** 0.0922** 0.0551 0.1138*** 
Wood products 0.0040 –0.0084 0.0047 0.0203 –0.0170 0.0047 
Paper products 0.0805** 0.0661** 0.0710*** 0.0715 0.0496 0.0206 
Chemicals 0.0950*** 0.0858*** 0.0958*** 0.0679** 0.0858*** 0.0906*** 
Rubber products 0.2259*** 0.2076*** 0.2195*** 0.1470*** 0.1121*** 0.1215*** 
Plastics 0.0798*** 0.0749*** 0.0906*** 0.0579** 0.0266 0.0409* 
Non-metallic mineral 0.1077*** 0.0728** 0.0845*** 0.0865** 0.0142 0.0317 
  products       
Basic metals 0.1023*** 0.1153*** 0.1124*** 0.0436 0.0477 0.0462 
Fabricated metals 0.0408* 0.0316 0.0692*** 0.0597* 0.0336 0.0743** 
General machinery 0.0911** 0.0762** 0.0755*** 0.1141** 0.1282*** 0.1365*** 
Office, computing, and –0.0140 –0.0503 –0.0273 0.0047 –0.0234 –0.0379 
  precision machinery       
Electrical machinery 0.1020*** 0.0903*** 0.1126*** 0.0604** 0.0789** 0.0862*** 
Radio, TV, and 0.0754*** 0.0633*** 0.0460*** 0.0719*** 0.0435* 0.0284 
  communication machinery       
Motor vehicles 0.1091** 0.1361*** 0.1617*** 0.1267** 0.1270* 0.1605** 
Furniture 0.0993*** 0.1057*** 0.1090*** 0.1103*** 0.0690** 0.0702** 

*** = significant at the 1% level, ** = significant at the 5% level, * = significant at the 10% level.

Note: Robust standard errors (clustered by plant for random effects) are used to account for potential heteroskedasticity. Results of the Breusch-Pagan Test (null of no random effects) is always rejected at the 1% level. Other slope coefficients and equation statistics are presented in Appendix Table 6. These results come from estimates that also include year, industry, and region dummies. Full results including constants and coefficients on year, industry, and region dummies are available from the author.

Source: Author's estimates from microdata underlying Department of Statistics (2002, various years).

The results are thus more or less consistent with the observations from the descriptive statistics analyzed in the previous section. MNEs tended to pay higher wages than local plants. However, MNEs also tended to be relatively capital-intensive, large, and have relatively large shares of workers in highly paid occupations and with higher or moderate education. Thus, accounting for these influences greatly reduces the scope of conditional MNE–local wage differentials in most industries. Perhaps the most conspicuous difference had been in rubber products, where conditional differentials were relatively large and statistically significant, but unconditional differentials were relatively small compared to other industries. This suggests that the measures of worker occupation, education, and sex, as well as capital intensity and size were not as strongly related to MNE–local differentials in this industry as in others.

In contrast, the lack of significant differentials in wood suggests that worker quality, worker sex, capital intensity, and size accounted for the vast majority of the substantial unconditional differentials observed in this industry. In office, computing, and precision machinery, differentials were similarly small and/or insignificant, whether measured conditional on worker quality and other plant characteristics or unconditionally. This probably reflects the heavy MNE dominance of this industry and its labor-intensive nature. As emphasized above, conditional and unconditional differentials were either both relatively high or low in a number of other industries. However, it is very difficult to explain the precise reasons for the inter-industry variation in the size of MNE–local differentials.

V.  Conclusions

This paper has investigated the extent of wage differentials between medium-large MNEs and local plants in Malaysia in the early 21st century. A brief literature review highlighted the important fact that MNEs have often been found to pay higher wages than non-MNEs. The main cause of these wage differentials is probably the fact that MNEs tend to hire relatively large shares of highly skilled workers. However, MNEs may also be motivated to pay higher wages to facilitate recruitment, reduce turnover, and compensate for higher worker productivity that results from the MNE's possession of firm-specific, generally intangible assets related to technology, marketing, and management, for example. MNE–local wage differentials can also result from workers’ reluctance to work for MNEs.

In the 17 industries examined, medium-large MNEs paid wages that were on average about two-fifths higher than wages in corresponding local plants. Consistent with previous evidence summarized in the literature review, shares of highly educated workers (about three-fifths to 90%) were also substantially higher in MNEs than in local plants. Shares of moderately educated workers and workers in highly paid occupations were also about one-fourth to one-third higher in MNEs. Previous evidence also indicates that MNEs tended to be more capital-intensive and larger than local plants during this period. Estimates of earnings equations suggest that differences in worker quality and plant characteristics are strongly correlated with plant-level wages.

After accounting for differences in worker occupation, education, and sex, as well as the size and capital intensity of plants, and the effects of industry affiliation, plant location, and annual fluctuations on the constants estimated, estimates of MNE–local wage differentials became much smaller, an average of 5%–9% in the 17 industries combined, but these differentials were highly significant statistically. When all slopes are allowed to vary among the 17 industries (by estimating separate equations for each industry), results varied across industries and estimation method. Significant differentials were more common in the pooled OLS estimates than in the random effects estimates. There were consistently significant differentials in six industries: food and beverages, chemicals, rubber, general machinery, electrical machinery, and furniture. These two key results are also robust to alternative specifications of the occupation, education, and sex variables and observed in shorter subperiods (2000–2002 and 2002–2004; Ramstetter 2012a). There was also relatively strong evidence of positive, conditional differentials in motor vehicles; apparel; plastics; non-metallic mineral products; and radio, TV, and communication machinery, but these results are not as consistent when estimated with alternative specifications or in subperiods.

These findings are an important addition to the literature on MNE–local wage differentials in Southeast Asian developing economies, largely because the Malaysian data allow more detailed measurement of worker quality than similar data for other countries. The evidence is consistent with previous studies of Indonesia, for example, suggesting that MNEs tend to pay their workers more than local plants even accounting for differences in worker quality and related plant characteristics in Southeast Asia. This in turn implies that MNEs impart important benefits on the workers who are fortunate enough to work for them. Many policy makers in Malaysia and other developing economies recognize these benefits, and it is important to know that this understanding is supported by empirical evidence.

Although these results are important, this paper leaves several related questions unanswered. For example, how do MNE takeovers of local plants (or vice versa) affect wage differentials? Alternatively, does MNE presence affect wage levels in local plants in Malaysia? Although such questions are of keen interest, their answers require use of statistical techniques that are not compatible with the evaluation of average wage differentials and are left for future research. Analysis of the latter question also requires aggregation of industries and the industry-level results from this paper suggest aggregation may bias results. It is also important to reemphasize that this study fails to account for several potentially important aspects of worker quality often accounted for in similar studies of earnings by individuals. These include worker background (e.g., race or nationality), experience, and training. There are also potentially important econometric issues relating to potential simultaneity and sample selection bias, which this and previous papers in this literature have not been able to address fully and should be considered further.

On the other hand, by comparing results of estimating specifications using lagged independent variables, this paper addresses simultaneity issues to a greater extent than previous estimates for Indonesia, which could only be performed in cross sections. Moreover, worker occupation and education are arguably among the two most important aspects of worker quality, and the ability to control for both of these aspects is an important advance over the research on Indonesia, for example, where data on occupation are not available. In conclusion, this paper provides important evidence that MNE–local wage differentials persist in Malaysian manufacturing, even after accounting for these key measures of worker quality, as well as worker sex and the size and capital intensity of plants.

References

References
Aitken
,
Brian
,
Ann
Harrison
, and
Robert E.
Lipsey
.
1996
.
Wages and Foreign Ownership: A Comparative study of Mexico, Venezuela, and the United States
.
Journal of International Economics
.
40
(
3–4
):
345
371
.
Bormann
,
Sarah
,
Pathma
Krishnan
, and
Monika
Neuner
.
2010
.
Migration in a Digital Age, Migrant Workers in the Malaysian Electronics Industry, Case Studies on Jabil Circuit and Flextronics
.
Berlin
:
World Economy, Ecology and Development
.
Buckley
,
Peter J.
and
Mark
Casson
.
1992
.
The Future of the Multinational Enterprise
. 2nd ed.
London
:
Macmillan
.
Casson
,
Mark
.
1987
.
The Firm and the Market: Studies on the Multinational and the Scope of the Firm
.
Cambridge, Mass
:
MIT Press
.
Caves
,
Richard E
.
2007
.
Multinational Enterprise and Economic Analysis
. 3rd ed.
London
:
Cambridge University Press
.
Chapman
,
Bruce J.
, and
J. Ross
Harding
.
1985
.
Sex Differences in Earnings: An Analysis of Malaysian Wage Data
.
Journal of Development Studies
21
(
3
):
362
376
.
Department of Statistics
.
2002
.
Census of Manufacturing Industries 2001
.
Putrajaya
:
Department of Statistics
.
Department of Statistics
.
2011a
.
Annual National Accounts, Gross Domestic Product (GDP) 2000–2010
.
Putrajaya
:
Department of Statistics
.
Department of Statistics
.
2011b
.
Malaysia Economic Statistics—Time Series 2011
.
Putrajaya
:
Department of Statistics
.
Department of Statistics
.
Various years
.
Annual Survey of Manufacturing Industries
.
Kuala Lumpur
:
Department of Statistics
.
Dunning
,
John
.
1993
.
Multinational Enterprises and the Global Economy
.
Workingham, UK
:
Addison-Wesley Publishing Co
.
Ismail
,
Rahmah
and
Haji Mat Zin
,
Ragayah
.
2003
.
Earnings Differentials Determinants between Skills in the Malaysian Manufacturing Sector
.
Asian Economic Journal
17
(
4
):
325
340
.
Lee
,
Kiong-Hock
and
Shyamala
Nagaraj
.
1995
.
Sex Differences in Earnings: An Analysis of Malaysian Wage Data
.
Journal of Development Studies
31
(
3
):
467
480
.
Lim
,
David
.
1977
.
Do Foreign Companies Pay Higher Wages than Their Local Counterparts in Malaysian Manufacturing
.
Journal of Development Economics
4
(
1
):
55
66
.
Lipsey
,
Robert E.
and
Fredrik
Sjöholm
.
2004a
.
Foreign Direct Investment, Education, and Wages in Indonesian Manufacturing
.
Journal of Development Economics
73
(
1
):
415
422
.
Lipsey
,
Robert E.
and
Fredrik
Sjöholm
.
2004b
.
FDI and Wage Spillovers in Indonesian Manufacturing
.
Review of World Economics
40
(
2
):
321
332
.
Lipsey
,
Robert E.
and
Fredrik
Sjöholm
.
2005
.
Host Country Impacts of Inward FDI: Why Such Different Answers?
In
Does Foreign Direct Investment Promote Development?
Theodore
H. Moran
,
Edward M.
Graham
, and
Magnus
Blomstrom
, eds. pp.
23
43
.
Washington, DC
:
Institute for International Economics
.
Lipsey
,
Robert E.
and
Fredrik
Sjöholm
.
2006
.
Foreign Multinationals and Wages in Indonesia
. In
Multinationals in Indonesia and Thailand: Wages, Productivity and Exports
.
Eric
D. Ramstetter
and
Fredrik
Sjöholm
, eds. pp.
35
53
.
Hampshire, UK
:
Palgrave-Macmillan
.
Markusen
,
James R.
2002
.
Multinational Firms and the Theory of International Trade
.
Cambridge, Mass
:
MIT Press
.
METI Japan
.
2001
.
Dai 29 Kai Wagakuni Kigyou no Gaikai Jigyou Katsudou (The 29th Survey of Overseas Business Activities of Japanese Companies)
.
Tokyo
:
Ministry of Economy Trade and Industry
.
Milanovic
,
Branko
.
2006
.
Inequality and Determinants of Earnings in Malaysia
.
Asian Economic Journal
20
(
2
):
191
216
.
Movshuk
,
Oleksandr
and
Atsuko
Matsuoka-Movshuk
.
2006
.
Multinationals and Wages in Thai Manufacturing
. In
Multinationals in Indonesia and Thailand: Wages, Productivity and Exports
.
Eric D.
Ramstetter
and
Fredrik
Sjöholm
, eds. pp.
54
81
.
Hampshire, UK
:
Palgrave-Macmillan
.
Ramstetter
,
Eric D.
1995
.
Characteristics of Multinational Firms in Malaysia: A Time Series Perspective
. In
Ajia Kogyoken no Keizai Bunseki to Yosoku IV (Projections for Asian Industrializing Region IV)
.
Mitsuru
Toida
and
Daisuke
Hiratsuka
, eds. pp.
95
171
.
Tokyo
,
Japan: Institute of Developing Economies
.
Ramstetter
,
Eric D.
.
1999
.
Comparisons of Foreign Multinationals and Local Firms in Asian Manufacturing over Time
.
Asian Economic Journal
13
(
2
):
163
203
.
Ramstetter
,
Eric D.
.
2004
.
Labor Productivity, Wages, Nationality, and Foreign Ownership Shares in Thai Manufacturing, 1996–2000
.
Journal of Asian Economics
14
(
6
):
861
884
.
Ramstetter
,
Eric D.
.
2012a
.
Do Multinationals Pay High Wages in Malaysian Manufacturing? Working Paper No. 2012-05
.
Kitakyushu, Japan
:
International Centre for the Study of East Asian Development
.
Ramstetter
,
Eric D.
.
2013b
.
Foreign Multinationals in East Asia's Large Developing Economies.
Working Paper 2012–06.
Kitakyushu
:
International Centre for the Study of East Asian Development
.
Ramstetter
,
Eric D.
.
2013
.
Wage Differentials between Foreign Multinationals and Local Plants and Worker Quality in Malaysian Manufacturing
.
Working Paper No. 2013-22
.
Kitakyushu, Japan
:
International Centre for the Study of East Asian Development
.
Ramstetter
,
Eric D.
.
2014
.
Exporting, Education, and Wage Differentials between Foreign Multinationals and Local Plants in Indonesian and Malaysian Manufacturing
.
Working Paper No. 2014-03
.
Kitakyushu, Japan
:
International Centre for the Study of East Asian Development
.
Ramstetter
,
Eric D.
and
Phan Minh
Ngoc
.
2007
.
Employee Compensation, Ownership, and Producer Concentration in Viet Nam's Manufacturing Industries
.
Working Paper No. 2007-07
.
Kitakyushu, Japan
:
International Centre for the Study of East Asian Development
.
Ramstetter
,
Eric D.
and
Shahrazat Binti Haji
Ahmad
.
2011
.
Foreign Multinationals in Malaysian Manufacturing After the Crisis
.
Thammasat Economic Journal
29
(
4
):
124
160
.
Rugman
,
Alan M
.
1980
.
Internalization as a General Theory of Foreign Direct Investment: A Re-Appraisal of the Literature
.
Weltwirtschaftliches Archiv
116
(
2
):
365
379
.
Rugman
,
Alan M
.
1985
.
Internalization is Still a General Theory of Foreign Direct Investment
.
Weltwirtschaftliches Archiv
121
(
3
):
570
575
.
Schafgans
,
Marcia
.
2000
.
Gender Wage Differences in Malaysia
.
Journal of Development Economics
63
(
2
):
351
378
.
Sjöholm
,
Fredrik
and
Robert
Lipsey
.
2006
.
Foreign Firms and Indonesian Manufacturing Wages: An Analysis with Panel Data
.
Economic Development and Cultural Change
55
(
1
):
201
221
.
Toyo
Keizai
.
Various years
.
Kaigai Shinshutsu Kigyou Souran: Kaisha Betsu Hen (A Comprehensive Survey of Firms Overseas: Compiled by Company)
.
Tokyo, Japan
:
Toyo Keizai
.
World Bank
.
2013
.
Immigration in Malaysia: Assessment of its Economic Effects, and a Review of the Policy and System
.
Washington, DC
:
World Bank, Human Development Social Protection and Labor Unit East Asia and Pacific Region
.

Notes

*

This paper is one output of the research project “Multinationals, Wages, and Human Resources in Asia's Large Developing Economies” funded by the International Centre for the Study of East Asian Development (ICSEAD) in fiscal year 2013 (ending March 2014). I thank ICSEAD for financial support and ICSEAD staff for logistic assistance. I also thank Niny Khor and two anonymous referees for detailed comments on previous versions. Responsibility for all opinions expressed and any remaining errors or omissions are the author's alone.

1

With the exception of Lee and Nagaraj's (1995) analysis of a limited sample of workers in the Klang Valley, which focused on male–female differentials, previous studies of Malaysia do not analyze MNE–local wage differentials.

2

These authors also examine other aspects of wage differentials and how they change over time in Lipsey and Sjöholm (2004b, 2005, 2006) and Sjöholm and Lipsey (2006).

3

These variables were union membership, marital status, migration status, total hours worked, plant size, and plant export-orientation.

4

The most commonly cited problems were (i) competition for product markets (11.2%) and (ii) political instability (8.6%).

5

The most commonly cited motives were (i) development of local markets and (ii) strengthening of international competitiveness, which were cited by 21%–31% of affiliates.

6

For example, seven of the top 10 employers in 2008 were foreign companies in Malaysia (http://malaysias100.com/media/foreign-firms-the-favorite), referenced December 2013.

7

Some theorists (especially Dunning 1993) view the possession of firm-specific assets or ownership advantages as a key necessary condition for a firm to become an MNE (in addition to internalization and location advantages). Other theorists (Buckley and Casson 1992, Casson 1987, Rugman 1980 and 1985) dispute this view, choosing instead to emphasize the role of internalization as the key distinguishing characteristic between MNEs compared to non-MNEs. However, the important point is that all agree that MNEs tend to possess these kinds of firm-specific assets in relatively large amounts.

9

This definition is narrower than usual because MNEs are usually defined as plants or firms with foreign ownership shares of 10% or more. Unfortunately, it is impossible to use the standard definition because the precise foreign ownership share is not collected.

10

Medium-large plants accounted for 90%–95% of all MNE plants compared to only 41%–60% of local plants (Ramstetter 2013, Appendix Table 1c). Medium-large plants also accounted for 99.7% or more of the paid workers (Table 1) and the output (Ramstetter 2013, Appendix Table 1b) of all MNEs, while corresponding shares for local plants were 93% or more.

11

For example, in the entire 2000–2004 sample, there were 4,592 observations of plants paying extremely low average compensation per worker of less than RM100 (about $26) per year. However, only 20 medium-large plants reported similarly low wages.

12

Five of these industries (tobacco, leather and footwear, petroleum products, other transport equipment, and recycling) had only a dozen or fewer medium-large MNE plants in at least 4 of the 5 years, two were heterogeneous and relatively small industries (other transport equipment and miscellaneous manufacturing), and one industry was closely regulated by the government (printing and publishing).

13

In 2001, real manufacturing GDP growth was –5.9% if measured at 1987 prices and –4.3% if measured at 2000 prices (Department of Statistics 2011b). This compares to 18% in 2000 (1987 prices), 4.3% (1987 prices) or 4.1% (2000 prices) in 2002, 8.6% (1987 prices) or 9.2% (2000 prices) in 2003, and 9.8% (1987 prices) or 9.6% (2000 prices) in 2004.

14

In sample plants, narrowly defined wages accounted for an average of 82%–84% of total compensation, while payments in kind accounted for 5%–6%, social insurance payments for 8%, and other items for the remainder.

15

Occupations that tended to be lowly paid were defined as clerics and related workers, elementary occupations, plant and machine operators, assemblers, and part-time workers.

16

In all 17 industries combined, shares of workers not completing SPM were 48% for MNEs in 2000 and 42%–44% thereafter, or 20%–31% below corresponding shares in local plants.

17

I have a copy of the survey form for the 2005 census and have been told that the same form was used in 2000–2004 by Malaysian officials.

18

This is reasonable for output but not very accurate for capital because changes in asset prices are not reflected. Unfortunately, I know of no deflators for fixed assets in Malaysia.

19

Alternatively, Ramstetter (2012a) disaggregates shares of workers in highly paid occupations, highly educated workers, and moderately educated workers by sex. This paper uses the specification in Equation (1) because it simplifies interpretation of coefficients on the occupation, education, and gender variables, and is more comparable to the specification in Lipsey and Sjöholm (2004a). Most importantly, estimates of MNE–local wage differentials from Equation (1) are very similar to corresponding estimates of the alternative specification in Ramstetter (2012a).

20

An anonymous referee also pointed out that export orientation of plants might affect the scope of MNE–local wage differentials, implying that the omission of export orientation might also create omitted variable bias. I am cognizant of this possibility and have examined the issue closely for both Indonesia and Malaysia in Ramstetter (2014). Because the literature suggests that exporters have to incur sunk costs related to the creation of export networks and related firm-specific assets, which are similar to the firm-specific assets possessed by MNEs, this makes it difficult to include both export orientation and ownership as independent variables in Equation (1). When Equation (1) was estimated separately for exporters and non-exporters in Malaysia, MNE–local differentials were slightly larger for exporters in samples of all industries combined (8.8%–9.2% vs. 6.2%–7.5% in pooled OLS estimates and 7.2%–7.8% vs. 4.7%–6.7% in random effects estimates), but industry-level results were weaker and inconsistent.

21

Ramstetter (2012a) also estimates an alternative, contemporaneous specification (see note above) for 2000–2002 and 2002–2004, in addition to 2000–2004. Results for the subperiods suggest that significant, MNE–local wage differentials were more common in the earlier period. This paper focuses on longer panels because they are thought to be relatively reliable and can facilitate more meaningful estimates of the lagged specification.

22

In the regression for office, computing, and precision machinery, one less region is used because of the lack of plants.

23

Industry dummies are not included in estimates for apparel, wood, paper, rubber, plastics, motor vehicles, and furniture (Ramstetter 2013, Appendix Table 6).

24

Because of the large number of samples examined, it is not possible to show all correlation matrices, but the author can make these available upon request.