The dollar, trade, technology and inequality in the USA.

Anderton, Bob ; Brenton, Paul

1. Introduction

The USA experienced a considerable increase in inequality during the 1980s, with the major increase in inequality occurring within industries.l Although several studies have investigated the possible causes of this decline in the relative economic fortunes of the less-skilled in the USA their conclusions differ quite considerably. For example, Feenstra and Hanson (1995 and 1996) claim that increased imports explain much of the rise in US inequality; Machin and Van Reenen (1998) find that the main cause is skill-biased technological change; and Haskel and Slaughter (1997) argue that it is the sectoral bias of skill-biased technological change that matters. This article contributes to the debate by focussing on the relationship between US inequality, US trade with low-wage countries and large movements in the dollar. In contrast to previous studies - which investigate the impact of US imports on inequality but do not distinguish between import suppliers - we examine whether the impact of imports from industrialised countries differs from that of imports from low-wage countries. Section two looks at the relationship between aggregate movements in US inequality and the appreciation of the dollar in the early 1980s. Section three examines movements in trade and technology indicators for three industry groups - representing high and low-tech sectors - and section four econometrically estimates the extent to which these factors explain the trends in US inequality. This is followed by some conclusions, policy recommendations and suggestions for further work.

2. Movements in US inequality and the dollar

It is now widely held that the main cause of the decline in the economic fortunes of the less-skilled seems to be a shift in demand towards higher skilled workers, Two main explanations are frequently offered for such a demand shift: first, that labour-saving technical progress has reduced the relative demand for less-skilled workers; second, that increased international trade with Low-Wage Countries (LWCs) - ie, nations with an abundant supply of low-skill and low-wage labour - has decreased the demand for low-skilled workers in the advanced industrialised countries. These impacts from trade may come about via Stolper-Samuelson effects or by mechanisms such as 'outsourcing'.(2) There are various routes by which skill-biased technical progress may reduce the relative wages and employment of the less-skilled. For example, technical progress which is biased towards reducing the use of unskilled labour will tend to increase the share of skilled, relative to unskilled, labour in production. This fall in demand for unskilled workers will tend to push down their wages and employment relative to the skilled.

The decline in the demand for the less-skilled in the US has not occurred at a constant rate. Feenstra and Hanson (1996) first pointed to the puzzle of the 'lumpiness' of the rise in US inequality. Using non-production workers as a proxy for higher-skilled labour, and production workers to represent the less-skilled, they found that the wage share of non-production workers in the USA showed a particularly large increase in the early 1980s. Given that this period corresponds with a recession in the United States, the behaviour of the wage share is not surprising as the relative demand for nonproduction workers is generally countercyclical. However, two mysteries remain: why was the change in the wage share so abnormally large in the early 1980s; and why did it not return to its previous level after the recession?

The hysteresis-type behaviour of the wage share of non-production workers corresponds to a period when the US dollar temporarily appreciated by around 40 per cent which, in turn, corresponds to a period of possible hysteresis in trade performance.(3) Baldwin (1988) and others argue that the high level of the dollar during the early 1980s caused a surge in US imports, and a fall in US import prices (in dollars), neither of which were reversed when the dollar depreciated back to its previous level from 1986 onwards. Consequently, one theme of this paper is whether the large appreciation of the dollar during the first five years of the 1980s, and the associated 'monetarist' policies, help to explain the rise in the wage bill share of high-skilled workers in the US.

Table 1 shows values at key points in time for the wage and employment shares of non-production workers, total import penetration and R&D expenditure as a percentage of GDP.(4) The latter variable is shown as it is frequently used in inequality analysis as a proxy for technological change and its behaviour over time lies behind many of the claims that technology has caused an increase in inequality in a number of countries.(5) The table clearly shows that the major rise in US inequality - proxied by the wage share of non-production workers (which captures movements in both relative wages and employment) - occurred between 1978 and 1986 and roughly corresponds with the appreciation of the dollar. Furthermore, it is clear that it is the rise in the relative employment of skilled workers, rather than a rise in relative wages, which explains the bulk of the increase in wage share.(6) Similarly, US import penetration rose at a more rapid rate during this period, but carried on rising - albeit at a much slower pace - even though the dollar depreciated by around 40 per cent from 1986 onwards (which is consistent with hysteresis-type behaviour).

Table 1. US non-production worker's wage and employment shares,
import penetration and R&D((a))

Year Non-product- Non-product- Import R&D/
 ion wage ion employ- penetra- out-
 put
 share((b) ment share((c)) tion((d)) ratio((e))

1974 34.5 25.4 5.8 2.19
1978 35.1 26.1 7.9 2.13
1986 41.3 31.2 12.3 3.51
1993 42.5 30.9 13.8 2.94

Notes: (a)All figures are in percentages. (b) Wage bill of
non-production workers divided by total wage bill for manufacturing
sector. (c) Employment of non-production workers divided by total
employment. (d) Imports divided by US imports plus domestic
production of manufactures. (e)IR&D expenditure in manufacturing
divided by manufacturing output.

Finally, R&D expenditure as a percentage of GDP follows a similar profile - it seems that technological change accelerated extremely rapidly during the early 1980s and then slowed down somewhat from the mid-1980s onwards, but R&D expenditure then remained at a significantly higher level relative to the previous decade (which is again consistent with hysteresis-type behaviour). The movements of the dollar - and the close relationship between the dollar appreciation and the rise in R&D expenditure in the early 1980s - are shown in Chart 1.

What can we conclude from the above table and chart? If our choice of explanations for the rise in US inequality is only between trade or technology then the above evidence seems to suggest that there is more support for the trade-based explanation than the results of previous studies suggest. This is not only because import penetration increased when inequality increased but also because the rise in the dollar, and the associated deterioration in the trade competitiveness of US industry, may explain the rapid rise in R&D expenditure via various mechanisms. For example, less-competitive firms - most likely comprising low-tech companies offering low quality products, perhaps associated with minimal R&D spending and a high proportion of low-skilled workers in their labour force - would be squeezed out of business (as the dollar appreciation made US imports much cheaper). These possible compositional effects imply that, after a considerable 'shake-out' brought about by the dollar appreciation, US industry would subsequently consist of a higher proportion of high-tech firms and the average R&D/output ratio would therefore rise (and be associated with a higher proportion of high-skilled workers if the technology is skill-biased). Moreover, the deterioration in competitiveness may have encouraged US manufacturers to 'innovate defensively', le, faced with strong competition from low-cost imports, firms may attempt to escape fierce import price competition by upgrading the quality of their manufactures via 'product innovation' which, in turn, is achieved by spending more on R&D etc.(7)

3. Trade, technology and inequality within high and low-skill-intensive sectors

Traditional trade theories can help explain movements in relative wages across industries, whereas what needs to be explained is the dramatic fall in the economic fortunes of less-skilled workers within US sectors. One possible mechanism which may explain how trade with low-wage countries may cause increased inequality within US sectors is 'outsourcing'. 'Outsourcing' is where firms take advantage of both the low-wage costs of the LWCs and modern production techniques - where the process of manufacturing a product can be broken-down into numerous discrete activities - by moving the low-skill-intensive parts of production abroad to the LWCs but continue to carry out the high-skill-intensive activities themselves. Once the low-skill activities have been performed the goods are then imported back from the LWCs and either used as intermediate inputs or sold as finished goods. Hence, trade with the LWCs via this route will shift demand away from less-skilled towards skilled workers in countries such as the USA, and put downward pressure on the relative wages and employment of low-skilled workers within industries. 'Outsourcing' is claimed to be an important activity in industries such as footwear (Yoffie and Gomes-Casseres, 1994, case 7) and textiles (Waldinger, 1986; Gereffi, 1993), etc. The above articles also illustrate that outsourcing applies to finished goods as well as intermediate inputs. But, aside from complex mechanisms such as outsourcing, it would not be surprising to find that industries which are more prone to import penetration from LWCs experience larger increases in wage inequality (particularly if we assume some degree of labour immobility in terms of the ability of the less-skilled to switch jobs between sectors).

One explanation for a possible link between exchange rate movements and 'outsourcing' may be provided by Orcutt (1950) who argues that the costs of switching from domestic to foreign suppliers may cause the price elasticity of imports to be bigger for large price changes than for small changes and a similar argument can be made for disproportionately large increases in 'outsourcing'. For example, when considering whether or not to 'outsource', US producers have to take into account the costs incurred when switching from in-house, or other domestic, supplies to foreign suppliers. For instance, when switching to foreign suppliers US producers may have to modify production techniques to be compatible with the newly-imported products and spend time ensuring that the new supplier is both reliable and produces a product of the required specifications and quality (as there is always some degree of uncertainty regarding the characteristics - such as quality and reliability - of previously untried imported goods and suppliers). Consequently, small changes in the prices of foreign goods will not be acted upon as the change in price differential will not cover switching costs. In contrast, a large appreciation of the dollar will result in a substantial differential between the costs of producing 'in-house' (or domestic) goods and imports - which may be at least sufficient to cover the costs of switching. In summary, switching costs may cause a disproportionate increase in 'outsourcing' during large exchange rate appreciations, which may partially explain the 'lumpiness' of changes in the economic circumstances of the less-skilled in the USA. Furthermore, the increase in 'outsourcing' will be difficult to reverse, even if the large appreciation of the dollar is fully reversed, as US producers now have a greater understanding of the benefits of 'outsourcing' since they are now familiar with the quality of goods not previously imported. Consequently, the substantial temporary appreciation of the dollar may have encouraged US purchasers permanently to switch from domestic to foreign goods (which is consistent with a disproportionate increase in 'outsourcing' at a time when the economic fortunes of the less-skilled deteriorated very rapidly and is also consistent with the trade-hysteresis literature).

Our method is to estimate econometrically the impact of trade with LWCs on the wage bill share [TABULAR DATA FOR TABLE 2 OMITTED] of the less-skilled by using a proxy variable for 'outsourcing' similar to Feenstra and Hanson (1996). Feenstra and Hanson proxy 'outsourcing' by US imports from all countries, which implicitly captures 'outsourcing' of US production to advanced industrialised countries as well as LWCs. However, there is no obvious reason why firms would 'outsource' low-skill-intensive activities - which is the mechanism by which 'outsourcing' affects the demand for the less-skilled - to advanced industrialised countries which are relatively abundant in skilled labour. Consequently, a major objective here is to investigate whether the source of imports matters by disaggregating US imports according to individual supplier countries and constructing US import share terms for both high and low-wage countries. Therefore, by explicitly identifying imports solely from low-wage countries and using this as a variable to explain changes in the wage share of the less-skilled in the USA, we are more likely to accurately capture 'outsourcing' to low-wage countries.

In previous work on the UK, Anderton and Brenton (1998b) find that the impact of trade with LWCs differs considerably between high- and low-skill-intensive sectors. Hence in the following analysis we distinguish between groups of industries which we classify as intrinsically high- or low-skill. In table 2 above we look at two groups of industries which can be classed as low-skill-intensive (abbreviated as LSA and LSB) and one group of high-skill-intensive sectors (HS). The first three columns of table 2 show that the largest rise in US inequality occurred in all three sectors during the period of substantial dollar appreciation, but that inequality continued to increase, albeit more gradually, through the rest of the 1980s and early 1990s.(8) The last three columns of table 2 show that R&D expenditure expressed as a proportion of output is extremely small in the low-skill sectors (less than a quarter of 1 per cent in LSA and not much above one half of 1 per cent for most of the period in LSB). Given that the R&D ratios are extremely small in the low-skill sectors - indicating that these are low-technology-intensive industries - one obvious question to ask is how can movements in R&D expenditure/technology explain the change in the wage share of non-production workers in these sectors? On the other hand, the technology explanation corresponds to movements in R&D expenditure in the high-skill sectors, particularly the large rise in R&D during the period of the dollar appreciation in the early 1980s. Table 2 also shows US imports from LWCs as a proportion of total sectoral imports. Although the relationship between the import share of LWCs in the low-skill sectors and the wage share of non-production workers is unclear in the early 1970s, there is a large increase in US imports from LWCs during the period when inequality rose more rapidly and the dollar appreciated. Conversely, imports for the high-skill sector group remained static between 1978 and 1986 - perhaps indicating that defensive innovation succeeded in reducing import competition from LWCs in this sector (however, the relatively high import share of LWCs in this high-skill sector suggests that the degree of low-wage country competition may be sufficient to be a plausible cause of defensive innovation).

4. Econometric results

In this section we estimate econometrically the impact of trade with LWCs and R&D spending on the US wage share for non-production workers. We use highly disaggregated US wage and production data - converted from US SIC to ISIC REV2 - and define non-production workers as skilled and production workers as less-skilled (source: US Census of Manufactures and Annual Surveys). Technological change is proxied by R&D expenditure as a proportion of GDP (source: OECD ANBERD database). The capital stock data are from the OECD's International Sectoral Database (ISDB). The bilateral US imports data were obtained from the OECD on an SITC basis and converted to the ISIC REV2 classification. Trade, production and wage bill share data are all disaggregated to the 4-digit ISIC level (hence all variables are on an ISIC basis and further details of the 4-digit sectors used in the analysis are given in the data appendix). In order to provide enough observations for separate 'panel estimation' of our three sectoral groupings, we pool the data across 4-digit ISIC sectors within the LSA, LSB and HS broad groupings using annual data for the sample period 1973-93 (imposing, in effect, the same parameters across the different 4-digit sectors).(9)

We adopt the wage-share specifications used by Machin et al. (1996 and 1998) which are based on the derivations in Berman et al. (1993, 1994). Following Machin et al. (1996 and 1998), we begin with a simple restricted variable translog cost function from which the wage bill share equation (in first differences) can be derived, as shown in (1) below:(10)

[Delta]S[W.sub.it] = [Alpha][Delta]1n [K.sub.it] + [Beta][Delta]1n [Y.sub.it] + [Rho][(R&D/Y).sub.it-1]

+ [Lambda][Delta]1n M[S.sub.it] + l[Delta]1n[[W.sup.hs]/[W.sup.ls]).sub.it] + [Gamma][D.sub.it] + [U.sub.it] (1)

where

S[W.sub.it] is the share of the wage bill of the high skilled

[Mathematical Expression Omitted]

[Mathematical Expression Omitted] is the wage bill of the higher skilled (ie, non-production workers).

[Mathematical Expression Omitted] is the wage bill of the lower skilled (ie, production workers).

[W.sup.hs]/[W.sup.ls] = relative wage rates of high and low-skilled workers.

[K.sub.it] is the capital stock.

[Y.sub.it] is real output.

R&[D.sub.it-1] is research and development expenditure.

M[S.sub.it] is the share of the value of domestic demand for the output of industry i accounted for by imports,

[D.sub.it] is a set of time dummies included to capture any company preferences for non-production or production workers common across industries for a given year,

[U.sub.it] is an error term.

Subscript represents industry i.

First differences are denoted by [Delta].

The MS term represents US imports and can be interpreted as a proxy for outsourcing. In this paper, we follow the approach of Feenstra and Hanson (1995, 1996) and justify the inclusion of the MS term in the wage bill share equation by arguing that merely including the factors derived from a traditional translog production function will not capture other factors - such as outsourcing - which may influence a firm's demand for skilled labour. Given that outsourcing to low-wage countries is claimed to push the range of activities performed by domestic industry away from low-skill towards high-skill tasks, the MS term can be interpreted as representing a reduced-form relationship between outsourcing and a firm's unit input requirement for skilled labour. We experiment with two different versions of MS:

1. MSO = US imports from high-wage countries (which we define as OECD countries).(11)

2. MSNO = US imports from LWCs (which we define as Non-OECD countries).

Our final specifications are shown in table 3 below. Note that they do not include the relative wage rates for the two types of labour in our estimated wage bill share equations mainly because relative wages are unlikely to be exogenous. However, the equation includes a set of macro time dummies which will capture any firm-level changes in preferences for higher-skilled workers due to absent variables such as relative wages.(12) We estimate two equations for each industry group - the first equation uses the US imports from OECD countries term (ie, MSO) and the second uses imports from LWCs (ie, MSNO). The results show that the change in output is negatively signed and statistically significant (with the exception of industry group LSA) and conforms with our prior that a short-run decline in output tends to reduce the demand for the less-skilled relative to the skilled. The capital stock term is not statistically significant in any of the equations, which may not be surprising for the low-skill-intensive sectors as they are extremely low-capital-intensive industries. Although the capital stock term has the correct positive sign for the high-skill sector grouping - as we expect complementaries between capital and skill - it is not statistically significant (perhaps because it is dominated by the strongly significant R&D term).

The statistical significance of the MSNO terms in the LSA and LSB sectors suggest that increased trade with LWCs tends to increase the wage share of non-production workers in the low-skill sectors. But the strong significance of the R&D terms in the HS sectors suggest that technological change rather than trade partly explains the rise in US inequality in the high-skill sectors. For the LSA sector grouping, there is also some limited-but less-convincing - evidence that any increase in [TABULAR DATA FOR TABLE 3 OMITTED] inequality from increased trade may be partly due to increased imports from the higher-wage OECD countries, whereas only imports from LWCs increase inequality in the LSB sectors.(13)

It is important to note that the above results may underestimate the impact of trade with low-wage countries on inequality as we do not include the import price in our specifications. Relative import price terms may capture other effects in addition to those captured by the import penetration terms such as the threat of increased competition from LWCs (for example, the fall in the import price of LWC products as the dollar appreciated may have made it easier for firms to obtain agreement from their workforce to restrain the wages, or terminate the employment, of less-skilled workers, and so on).(14)

As mentioned before, previous studies such as Machin et al. (1996) do not find a significant impact of trade on the relative wages and employment of the less-skilled in the USA. However, unlike our analysis, they do not use trade data which separately identify imports from low-wage countries - which is important as mechanisms such as 'outsourcing' influence inequality only via trade with low-wage countries - and their empirical work is at a more aggregate level. Although Feenstra and Hanson (1995) find that imports have increased US inequality they too do not distinguish between import suppliers. In contrast, we have shown that when assessing the impact of trade on inequality the source of imports matters, which is consistent with economic theory. For the USA, it seems that using aggregate imports to capture mechanisms such as outsourcing may be misleading and that disaggregation of imports in order to identify low-wage countries is necessary and that the impact of trade on inequality may vary across sectors of different skill intensities.

5. Conclusions and policy lessons

1. The dramatic and temporary appreciation of the dollar in the early 1980s was associated with a substantial deterioration in the economic fortunes of the less-skilled in the USA. The effects were wide ranging (ie, throughout the manufacturing sector), but the mechanisms through which the low-skilled were affected differ according to the skill-intensity of the sector in which they were employed.

2. An increase in US imports from LWCs, encouraged by the large appreciation of the dollar in the early 1980s, seems to explain some of the rise in US inequality in low-skill-intensive sectors. Rapid technological change does not seem to be an important determinant of inequality in these sectors - which is not surprising given the low-technology nature of these industries.

3. Technological change - proxied by R&D expenditure-seems to be strongly positively correlated with the rise in US inequality in our sample of high-skill-intensive sectors. However, given that the timing of the sudden rise in US R&D expenditure corresponds with the appreciation of the dollar, it may be the case that the deterioration in US trade competitiveness during this period contributed to the rapid increase in the rate of US technological change via mechanisms such as 'defensive innovation'. Hence one might also argue that the technology-based explanation for the rise in US inequality could actually be a trade-based explanation, but this remains a tentative claim as more research is required regarding the links between increased competition with LWCs and defensive innovation.

4. Both R&D expenditure and imports from LWCs increased substantially during the rapid appreciation of the dollar, but both remained high even when the dollar depreciated back to its previous level. Anderton and Brenton (1998b) show that the UK experienced similar changes during the temporary appreciation of sterling in the early 1980s which was associated with a permanent increase in inequality in the UK. These findings are consistent with hysteresis theories of the permanent impact of temporary shocks.

5. One possible interpretation of the above results is that exchange rate-induced recessions, combined with the increasing globalisation of international trade, may result in irreversibilities which can fundamentally change the nature of economies. Of course, all advanced industrialised countries are experiencing the increasing globalisation of international trade, but perhaps in a much more gradual fashion than the USA (and the UK) where the sheer rapidity of change may not have provided sufficient time for workers and firms to adjust in an optimal fashion. Although we believe that increased trade is generally 'good' - in terms of increasing welfare by reducing excess rents and encouraging production according to comparative advantage, and so on - the speed of this process may partly determine the magnitude of these benefits and their distribution. Policymakers should therefore be alarmed at the possible permanent effects of large but temporary appreciations, particularly those considering or committed to membership of the European single currency, since entering at too high a rate may result in social exclusion as well as causing a deterioration in competitiveness. On the other hand, not participating in EMU may leave countries exposed to the rapid adjustments experienced by countries with volatile exchange rates (eg, the USA and UK).

6. When considering policies to combat inequality, does it matter whether trade or technology is the cause of the decline in demand for the less-skilled? Regardless of the cause, the policy response centres around improving the aggregate skill level of the workforce by concentrating in particular on those at the lowest level so as to provide greater adaptability to the higher degree of technological change in skill-intensive sectors, and to ensure that workers can successfully move to new industries as mature industries increasingly become the domain of low-wage countries. Such changes in aggregate skill levels would enable the US economy to cushion itself more effectively from rapid changes in world demand for its products by moving into areas in which qualitative attributes outweigh the apparent labour cost disadvantages of the US as globalisation increases.

7. It is important to establish why skill-biased technological change has been so rapid as this will help us judge whether this trend will continue in the future (an associated question is whether firms would find 'skill-neutral' technological change to be as effective as 'skill-biased' technological change - in which case policies to encourage the former would solve the problems caused by the particularly sharp decline in demand for less-skilled workers). Conversely, if increased trade with LWCs increases inequality both directly and indirectly through skill-biased technological change via defensive innovation, then we know that inequalities will continue to widen as globalisation increases - unless there is an appropriate policy response.

DATA APPENDIX: 4-DIGIT SECTORS

We group together industries which we classify as high or low-skill-intensive. In particular, we form two low-skill groups, LSA and LSB, and one high-skill grouping we call HS. LSA consists of 4-digit ISIC sectors within 32, 33 and 34 (ie, Textiles, Apparel and Leather; Wood Products and Furniture; Paper, Paper Products and Printing). LSB consists of 4-digit ISIC sectors within 36, 37 and 381 (ie, Non-Metallic Mineral Products; Basic Metal Industries; Metal Products); HS consists of 4-digit ISIC sectors within 35, 382, 383,385 (Chemical Products; Non-electrical Machinery; Electrical Machinery; Professional Goods). We pool the data across eighteen 4-digit ISIC sectors for LSA, across ten 4-digit ISIC sectors for LSB, and across twenty-two 4-digit ISIC sectors for HS. Our annual sample period extends from 1973 to 1993. Given that we lose one observation because we estimate a first difference model, our estimation period 1974-93 therefore provides us with 360 observations for LSA (ie, 18x20); 200 observations for LSB; and 440 observations for HS. The specific 4-digit ISIC sectors used in the estimation are as follows:

LSA:

ISIC3211 Spinning, weaving and finishing textiles.

ISIC3212 Manufacture of made-up textile goods, except wearing apparel.

ISIC3213 Knitting mills.

ISIC3214 Manufacture of carpet and rugs.

ISIC3215 Cordage, rope and twine industries.

ISIC3219 Manufacture of textiles not elsewhere classified.

ISIC3220 Manufacture of wearing apparel except footwear.

ISIC3231 Tanneries and leather finishing.

ISIC3232 Fur dressing and dyeing industries.

ISIC3233 Manufacture of products of leather except footwear and apparel.

ISIC3240 Manufacture of footwear except rubber or plastic.

ISIC3311 Sawmills, planting and other wood mills.

ISIC3312 Manufacture wooden, cane containers, small cane ware.

ISIC3319 Manufacture wood and cork products N.E.C.

ISIC3320 Manufacture of furniture, fixtures except primary metal.

ISIC3411 Manufacture of pulp, paper and paperboard.

ISIC3412 Manufacture of containers and boxes of paper, paperboard.

ISIC3419 Manufacture of articles of pulp, paper and paperboard NEC.

LSB:

ISIC3610 Pottery, china and earthware.

ISIC3620 Glass and glass products.

ISIC3691 Structural clay products.

ISIC3692 Cement, lime and plaster.

ISIC3699 Non-metallic mineral products, NEC.

ISIC3710 Iron and steel basic industries.

ISIC3720 Non-ferrous metal basic industries.

ISIC3811 Cutlery, hand-tools and general hardware.

ISIC3812 Furniture and fixtures primarily of metal.

ISIC3819 Fabricated metal products except machinery and equipment NEC.

HS:

ISIC3511 Basic industrial chemicals.

ISIC3512 Fertilisers and pesticides.

ISIC3513 Synthetic resins, plastic materials, man-made fibres excluding glass.

ISIC3521 Paints, varnishes and lacquers.

ISIC3522 Drugs and medicines.

ISIC3523 Soap, cleansing preparations, perfumes cosmetics.

ISIC3529 Chemical products, NEC.

ISIC3530 Petroleum refineries.

ISIC3540 Miscellaneous products of petroleum and coal.

ISIC3551 Tyre and tube industries.

ISIC3559 Manufacture of rubber products, NEC.

ISIC3560 Plastic products, NEC.

ISIC3824 Manufacture of special industrial machinery and equipment except 3823.

ISIC3825 Office, computing and accounting machinery.

ISIC3829 Machinery and equipment except electrical not elsewhere classified.

ISIC3831 Electrical industry machinery and apparatus.

ISIC3832 Radio, telecommunications equipment and apparatus.

ISIC3833 Electrical appliances and housewares.

ISIC3839 Electrical apparatus and supplies.

ISIC3851 Professional scientific and control equipment.

ISIC3852 Photographic and optical goods.

ISIC3853 Watches and clocks.

Comments should be addressed to Bob Anderton at the National Institute of Economic and Social Research, tel. no. 0171 654 1928, e-mail: b.anderton@niesr.ac.uk or to Paul Brenton at the Centre for European Policy Studies, Belgium, e-mail: paul.brenton@ceps.be. We are grateful to the National Institute's Editorial Board and Valerie Jarvis for useful comments, but any errors remain the responsibility of the authors. This research forms part of the Globalisation and Social Exclusion project funded by the European Community under the Targeted Socio-Economic Research (TSER) Programme.

NOTES

(1) We define a rise in inequality as a deterioration in the relative wages and/or employment of the less-skilled.

(2) Anderton and Brenton (1998a,b) describe both of these trade mechanisms in detail.

(3) 'Hysteresis' denotes the situation where a temporary shock results in permanent effects.

(4) The years 1974 and 1993 correspond to the beginning and end of our sample period, whereas 1978 and 1986 roughly correspond to dates before and after the dollar appreciation.

(5) However, other papers - such as Haskel, 1996a,b for the UK - find that increased computer usage can also explain the rise in inequality.

(6) This seems to contradict the usual characterisation of the US labour market as a market where the flexible adjustment of wages prevents unemployment from rising.

(7) The experience of the UK during this period is very similar to that of the US. Between 1979-81, sterling temporarily appreciated by around 30 per cent and was associated with a rise in the UK manufacturing R&D/output ratio from around 1.5 per cent to 2 per cent and remained higher even when the appreciation was subsequently reversed.

(8) The higher wage bill share for non-production workers in the HS sectors relative to the other sectors is consistent with our claim that the former sectors are relatively high-skill-intensive. Also note that the sum of the sectors do not add up to the aggregate wage share in table 1 as not all sectors are included in table 2.

(9) F-tests from preliminary regressions show that it is acceptable to impose the same slope parameters across the relevant 4- digit industries.

(10) For full details of the assumptions underlying the restricted variable translog cost function and how the wage bill share equation is derived the reader is directed to Berman et al. (1993).

(11) The OECD countries are defined as those members up to and including 1993 (ie, excluding later members such as the Czech Republic, Hungary, Poland, South Korea and, probably most importantly, Mexico).

(12) As an alternative to sector specific relative wages we experimented with aggregate relative wages - as the latter should avoid the endogeneity problem - but aggregate wages were not statistically significant.

(13) However, note that only imports from the LWCs are actually statistically significant at the 5 per cent level of significance for the LSA sector grouping. The parameter for the MSO term also indicates that the trade impact from higher-wage countries is much smaller than the impact of imports from LWCs.

(14) A relative import price term may also capture the increased opportunities for decreasing labour costs via outsourcing.

REFERENCES

Anderton, R. and Brenton, P. (1998a), 'Trade with the NICs and wage inequality: evidence from the UK and Germany' in Brenton, P. and Pelkmans, J. (eds), Global Trade and European Workers, London: Macmillan.

----- (1998b), 'Did outsourcing to low-wage countries hurt less-skilled workers in the UK., in Brenton, P. and Pelkmans, J. (eds), Global Trade and European Workers, London: Macmillan.

Baldwin, R. (1988), 'Hysteresis in import prices: the beachhead effect', American Economic Review, 78, 4, pp. 772-85.

Berman, E., Bound, J. and Griliches, Z. (1993), 'Changes in the demand for skilled labor within US manufacturing industries: evidence from the Annual Survey of Manufactures', National Bureau of Economic Research Working Paper No. 4255.

----- (1994), 'Changes in the demand for skilled labor within US manufacturing: evidence from the Annual Survey of Manufactures', Quarterly Journal of Economics, CIX, 2, May.

Feenstra, R.C. and Hanson, G.H. (1995), 'Foreign investment, outsourcing and relative wages', National Bureau of Economic Research Working Paper No. 5121.

----- (1996), 'Globalization, outsourcing and wage inequality', American Economic Review, Papers and Proceedings, 86, 2, May, pp. 240-45.

Gereffi, G. (1993), 'The role of big buyers in global commodity chains: how US retail networks affect overseas production patterns', in Gereffi, G. and Korzeniewicz, M. (eds), Commodity Chains and Global Capitalism, Westport CT: Praeger, pp. 95-122.

Haskel, J. (1996a), 'The decline in unskilled employment in UK manufacturing', Centre for Economic Policy Research, Discussion Paper No. 1356.

----- (1996b), 'Small firms, contracting-out, computers and wage inequality: evidence from UK manufacturing', Centre for Economic Policy Research, Discussion Paper No. 1490.

----- and Slaughter, M. (1997), 'Does the sector bias of skill-biased technological change explain changing wage inequality?', unpublished paper.

Machin, S., Ryan, A. and van Reenen, J. (1996), 'Technology and changes in skill structure: evidence from a panel an international panel industries', Centre for Economic Policy Research No. 1434.

Machin, S. and Van Reenen, J. (1998), 'Technology and changes in skill structure: evidence from seven OECD countries', Institute for Fiscal Studies Working Paper No. W98/4.

Orcutt, G. (1950), 'Measurement of price elasticities in international trade', Review of Economics and Statistics, 32, 2, pp. 117-32.

Waldinger, R.D. (1986), Through the Eye of a Needle, New York: New York University Press.

Yoffie, D.B. and Gomes-Casseres, B. (1994), International trade and Competition, New York: Mcgraw-Hill.