Unemployment, Structural Change and Globalization

M. Pianta and M. Vivarelli

1. Introduction

While technological change may have adverse quantitative effects on employment, which can be more or less counterbalanced by market compensation mechanisms (see Unemployment and Technical Change), it also has qualitative effects on the composition of the employed . In particular, during the ‘90s the debate is focused on the so-called "skill bias" which seems to characterize the present ICT revolution. The idea is that new technologies imply a change in the relative ratio between skilled and unskilled workers with the demand for labour shifting in favour of the former. This tendency in the labour market can imply either lower wages for the unskilled (this seems the tendency in the US and recently in the UK) or higher unemployment rates (and also long-term unemployment rates, see Unemployment and the Labour Market) among the unskilled (this tendency seems to prevail in continental Europe).

The skill bias hypothesis is generally contrasted with the idea that globalization and not technical change is the important cause of the relative displacement of the unskilled workers. This issue will be fully treated in a separate entry: Unemployment and International Competitiveness.

Moreover, the skill bias approach is also an interpretation of current trends in income distribution and can be useful in understanding current forms of social exclusion. The discussion in the following sections will try to take into account these additional dimensions of the problem.

Finally the debate about the displacement of the unskilled and their likely social exclusion has generated different attempts to give an answer to this problem in terms of economic and social policies. In this context, many policy proposals are based on some forms of incentive in favour of the unskilled workers. The idea here is to push firms to hire unskilled through subsidies able to affect the relative costs of the different components of the labour force. Section 3 will deal with these policy prospects, trying to underline pros and cons of different proposals.

At the end of this introduction a caveat has to be put forward: differently from other issue discussed in other entries, empirical evidence about the skill bias is very rare and often not precise. Unfortunately, statistics about skills and tasks are very hard to collect and so skills are always proxied by either educational level or by a rough distinction between blue and white collars. Moreover, even these proxies are available for a very limited number of countries and for short time spans. So, although some tentative empirical discussion will be put forward , the section about the empirical evidence will suffer from these severe statistical constraints. In order to partially compensate the limits of the empirical section, previous empirical literature on the subject will be surveyed in detail in the following section.

One way to deal with the current debate about the skill bias hypothesis is to single out three different strands of literature. The first one is mainly theoretical and try to investigate possible solutions to the displacement of the unskilled workers. The second one is mainly empirical and here the purpose is to see whether and how much job losses are due to technical change and not to other causes, like for instance globalization and decentralization away from industrialized countries. Finally, the third one is devoted to detect and to measure the wage differential which can be originated by a shift of the demand for labour away from unskilled labour force.

From a theoretical point of view, a skill-biased technical change can be represented by an asymmetric movement of the isoquant towards the origin: this means that the same amount of output can be produced with less workers and that this reduction in labour coefficients is more severe for the unskilled workers (see Cooper-Clark, 1982, p.108; Vivarelli, 1995 p. 80-81). When technical change involves an increase in the average proportion between skilled and unskilled workers, even assuming favourable demand conditions is not enough for assuring a return to full employment. In fact, although demand conditions can potentially counterbalance productivity gains, structural unemployment can occur as a consequence of scarcity of skilled labour. An example can be of some help to illustrate this issue.

Assume that current technological conditions implies full employment of both skilled and unskilled workers: for instance, 20 skilled workers and 100 unskilled workers are employed in a given economy. Then, a skill biased technical change occurs: the same output can now be produced with 10 skilled workers and 30 unskilled ones (the relative coefficient in the use of labour has increased from 1/5 to 1/3 because of the skill bias). Now, even if an unlimited demand expansion is assumed, the economic expansion can lead to the full utilisation of skilled labour (20) and to under-utilisation of the unskilled (60); as a consequence, 40 unskilled workers remain unemployed. In other words, a limited supply of skilled labour implies unemployment among unskilled workers.

From a theoretical point of view, two possible solutions can cope with this kind of situation. The first one is to allow a wage adjustment which implies a drop in the relative wage of the unskilled; if such is the case, unskilled labour becomes cheaper and employers can move along the isoquant and decide to hire a bigger proportion of the unskilled workers. This is what is currently occurring in the US and partially in the UK. Three are the limits of this solution. First of all, a reduction in wages can imply a decrease in aggregate consumption and so in the effective demand (see also Unemployment and Wages), if such is the case, demand expansion can be unable to compensate productivity gains due to technical change. Secondly, the reaction of the relative demand for labour to wage adjustments can be very limited when technology is relatively rigid: if a given production process requires a fixed skilled/unskilled utilization ratio, employers may be completely insensible to wage adjustments. Thirdly, unskilled workers may exit from unemployment but they may enter into an income trap leading to a working poor situation.

The second possible theoretical solution is training, that is to transform some unskilled into skilled workers. In the example above, it would be enough to train 10 unskilled workers for reaching full employment (retaining the optimistic hypothesis of no demand constraints, 10 newly skilled workers can be combined with the remaining 30 unskilled). Of course, also the implementation of this measure is not immune from possible hindrances. Training can be either inadequate or redundant and in this case "workfare" becomes only a way to disguise open unemployment (as it seems to have occurred, at least partially, in Nordic countries, see Calmfors, 1994 and Lindbeck et al., 1994). Moreover, biased technological change is a fast and continuous phenomenon whilst training is difficult to plan, update and implement in time; as a consequence, training programmes can be insufficient to avoid a renewing high rate of unemployment among the unskilled (this seems the situation within the European Union where, notwithstanding the huge efforts and expenses in national and international training programmes - think, for instance, to the European Social Fund - unemployment among the unskilled is far from being reduced).

From an empirical point of view, the main point is to single out the different causes of a decrease in the relative labour demand for the unskilled. In this case the main distinction is between worker displacement "between industries" and "within industries". In the first case, the relative reduction in the use of unskilled labour is due to a demand shift - which can be triggered by consumers, investors, government procurements or international trade - towards sectors whose production processes are more intense in skilled workers; in the second case the shift is within a given sector and is directly linked with technical and organizational changes. In particular, this distinction is important to contrast the skill bias hypothesis with the idea that unskilled workers are displaced because of international trade and globalization.

For instance, Berman - Bound - Griliches, (1994) have found that less than one-third of the shift of employment from production to non production workers can be accounted for "between industry" shifts (which are in turn caused by changes in the relative composition of defence procurements and trade in the 1980s); while most of the shift to nonproduction employment has occurred within four-digit US manufacturing industries. This important residual can be related to skill bias labour-saving technological change; in fact, this paper also shows that skill upgrading is positively correlated with investment in computers and with R&D expenditures.

Using three different US data-sets, Doms - Dunne - Trotske, (1997), try to test the skill-biased technical change hypothesis through the investigation of a possible positive correlation between the adoption of ICT manufacturing technologies and the up-grading of skills and wages. Indeed, their cross-sectional results (358 plants) support both the 2 correlations: on the one hand, they find a quasi-monotonic correlation between worker education degrees and the adoption of new technologies; on the other hand, they run regressions looking at qualification instead of education and their results are consistent with the previous ones. Finally, their time series estimates (3.300 plants) show that the most technologically intensive plants shifted their employment towards more skilled workers (measured as the fraction of non production workers on total employment).

In other empirical Canadian case studies, Baldwin - Diverty - Johnson, (1995 and 1996), show that the introduction of ICT technologies increases the skill requirements in most cases (from 44 to 46%, according to the different ICT innovations), while reduces them in few firms (from 6 to 14%). Moreover, survey results point out that technology adoption creates a need for more training: between 2/3 and 3/4 of technology-using firms have reported that the adoption of new technologies has increased their education and training costs (this is consistent with the theory discussed above). In another study based on Canadian manufacturing data, Betts (1997) has found that in 10 out of 18 industries technical change has not been skill-neutral: in each case of skill-biased technical change, innovation has been skill-intensive.

Turning the attention to an European country, Machin, (1996) shows that UK manufacturing employment considerably shifted away from manual to non-manual labour and the same pattern emerges when educated vs less educated workers are compared. Interestingly enough, about 80% of the relative shifts are due to within industry changes and only 20% to the between industry component; so, the author’s interpretation is that skill upgrading has occurred mainly because of skill-biased labour-saving technical change, while import competition seems to have played a minor role. If attention is focused on author’s econometric estimates, industry-level regressions are run using three different measures of technical change: R&D expenditures (Business Monitor survey database), major innovations (SPRU database) and use of computers (WIRS database). In all the regressions the hypothesis that more innovating sectors are more likely to have increased their non-manual share is significantly confirmed.

Finally, Lauritzen - Nyholm - Jorgensen - von Sperling, (1996) present empirical evidence supporting the positive relationship between ICT diffusion and the need for up-skilling and show how marginalisation can be worsened by the lack of education and skill. Using questionnaires from 515 Danish manufacturing firms with more than 20 employees, empirical results show on the one hand a markedly negative impact on employment opportunities for unskilled workers when firms start to intensively use ICT technologies; on the other hand, job opportunities for civil engineers and non university engineers substantially increase from investment in ICT. A second analysis - based on a total sample of 200.000 Danish firms and 125.000 persons - clearly shows that the ratio of unskilled labour in gross job creation flows has steadily declined during the ‘80s, while the ratio for unskilled workers in gross job destruction has exhibited an upward trend. The reverse is true for the skilled portion of total labour force. Interestingly enough, these trends are common to all industries; indeed, in many industries which traditionally employ a high percentage of unskilled workers, the average education intensity has increased more than the average.

On the whole, both theoretical considerations and empirical evidence suggest that the skill bias hypothesis plays an important role in determining the level and the internal composition of unemployment. This findings do not exclude other determinants like globalization (see Unemployment and International Competitiveness and Wood, 1994), but the empirical evidence discussed above seems to suggest that other causes are quantitatively less important than skill biased labour-saving technological change. While the latter mainly involves unemployment in Europe, in the US it has a large effect on income distribution.

In fact, the impact of a skill biased technical progress is partially absorbed in terms of higher unemployment and partially in terms of a higher wage gap between skilled and unskilled workers (this is also the wage adjustment which should counterbalance the initial displacement of the unskilled, see the theoretical discussion above).

In a seminal paper, Krueger (1993) focuses on the issue of whether employees who use computers at work earn more as a result of applying their computer skills. This hypothesis is tested through econometric estimates based on data from the US Current Population Survey (CPS) conducted in 1984 and 1989. The main result is that workers who use a computer earn roughly 10-15 percent higher pay, other things being equal. In addition, because more highly educated workers are more likely to use computers at work, the estimates imply that the proliferation of computers can account for between one-third and one-half of the increase in the rate of return to education observed between 1984 and 1989.

In another paper, Dunne - Schmitz (1995) have found that the share of production workers in plants that use computer-based equipment is smaller than that in plants that do not use this equipment, although this difference is small (4%). In addition, production workers in ICT plants earn a 14% wage premium compared with workers in more traditional plants (for some plants’ size categories, the technological regressor explains 60% of the wage differential).

On the whole, these empirical studies clearly show that, at least in the US, the computer revolution has enlarged the wage premium for more educated people; conversely, the less educated workers face a higher risk to be either excluded (unemployment) or trapped in a working-poor situation.

However, recent studies on the impact of ICT upon wage differentials have cast some doubts on the hypothesis that technical change is entirely responsible of wider wage gaps. For instance, Bartel and Lichtenberg (1987) correctly pointed out that skilled workers have a comparative advantage with respect to learning and implementing new ICT technologies and found out that the demand for skilled workers declines as the capital stock ages. In other words, the main empirical question is whether workers who use ICT are better paid (that is the hypothesis discussed above) or whether abler workers are more likely to use ICT. If such is the case, the wage differential is not a consequence of computer use but is simply the effect of better education and skills which are personal and not technological features.

For instance, using German data, DiNardo and Pischke (1997) have found out that wage differentials are correlated with the use of computers but also with the use of telephones, with sitting down jobs and even with the use of pens and pencils. This evidence suggest that the computer wage effect is not robust to the inclusion of the worker’s educational level or skill. In other words, a worker gets a higher wage and uses a computer because he is better skilled and more educated, but there is not any causal relationship between the use of a computer and the wage differential.

Chennells and Van Reenen (1997) has used data drawn from the British Workplace Industrial Relation Surveys (2000 plants) and they conclude that "simultaneous determination of technology and earnings leads to the conclusion that higher earnings exert a positive influence on the probability of introducing technical change, but that technical change per se has little direct influence on the earnings of manual employees".

Finally, Entorf and Kramarz (1997) deals with longitudinal French data and they find that computer-based new technologies are used by already better paid abler workers; yet, these workers appear to become more productive after the ICT introduction and so their wage differential increases an extra 1% per year.

On the whole, while the skill bias effect seems to be demonstrated in terms of workers displacement, the bias in terms of wage differential seems to be quite less important than what had emerged from the first studies on the subject (Krueger, 1993 and Dunne - Schmitz, 1995)

The political debate in Western countries is now focused on the issue of the "employability" of unskilled workers. While in Europe the unskilled workers are more than 50% of the unemployed and the vast majority of long-term unemployed, in the US virtually all the open unemployed are unskilled workers and this category largely overlaps with the socially excluded: drug-addicted, alcoholic, homeless, criminals and so on.

Shifting the attention to the employability issue is a way to contest a welfare approach to the problem. The idea is to move resources away from the welfare system, especially form unemployment benefits, in favour of subsidies which can increase the likelihood for an unemployed unskilled worker to be hired. In terms of recipients, the subsidy goes to the firm and not directly to the unemployed (unless unemployment benefit is transformed in a hiring voucher, see Snower, 1994 and 1995), Of course, these arguments open the way to the much more general discussion about the welfare reform. Here it is impossible to discuss such a complex issue and attention will only be focused on the pros and cons of public subsidies devoted to increase the employability of the unskilled.

These subsidies can be designed in different ways: they can be related to the educational degree (higher subsidies for lower educated labour force), to the working task (higher subsidies for productive less skilled jobs), to the length of the unemployment period (higher subsidies for the long-term unemployed which are generally the less skilled and more vulnerable in terms of social exclusion). From this respect, probably the most sophisticated proposal has been put forward by Phelps (1997). This author has proposed a wage subsidy which takes the form of a support to the overall employment (not only the additional one); this subsidy should be declining along the wage ladder. In other words, the subsidy should range from 3 dollars per hour supporting a minimum wage of 4 dollars per hour to 6 cents supporting a maximum wage of 12 dollars per hour (ibid., p.113). In this way, the subsidy should support the employability of the unskilled workers, whose "private" productivity would be lower than a decent wage. This measure would assure both firm’s profitability and the avoidance of the "working poor trap". On the other hand, this kind of subsidy would be linked to real jobs and so it would be very different from both the "negative income tax" - proposed by Milton Friedman in the early 1960s - and from the "basic income", which is proposed nowadays as a possible outcome of the welfare reform. Finally, in Phelps’ opinion, the initial increase in the budget deficit would be compensated both by the additional public revenues associated with taxation of the additional employment and economic output and by saving on welfare, unemployment benefits, health care and so on.

Although this and similar policy proposals deserve to be taken into account, at least the following four factors cast severe doubts about the effectiveness of a generalized wage subsidy:

1. The usual dead-weight, substitution and displacement effects fully apply to this measure. There is a dead-weight effect when the vacancy associated with the subsidy would anyway have been filled (even without the subsidy). There is a substitution effect when the vacancy would otherwise have been filled by other unemployed people outside the targeted group but with higher productivity (in other words, the subsidy induces a substitution effect in favour of low-skilled/low-productivity workers and this trend may have adverse effects in the long-run). Finally, there is a displacement effect when the subsidised firms displace competitors, leading to a market distortion (in this case firms whose production processes are more intense in low skilled labour would benefit more in comparison with rivals).

2. The risk of "moral hazard" by the recipient firms is extremely high: employers would tend to undervalue the skills of their employees or even to substitute skilled with unskilled workers in order to get larger subsidies; in the latter case long-run productivity perspectives would be severely compromised.

3. The role of technical change has to be taken into account: in the presence of a fast, continuous and skilled-bias technical progress, a subsidy does little against the tendency to make unskilled workers redundant: even when the unskilled labour is made cheaper - if it is not matchable with skilled labour and advanced capital - it will be dismissed in any case;

4. While the welfare system and the basic income are measures which operate "erga omnes", a wage subsidy requires a minimum level of productivity (equal to 4 dollars per hour in Phelps’ proposal) and this is an endowment that some unemployed do not have. For these ones, there would not be any chance of getting out of social exclusion.

For all this reasons, Phelps’ and similar proposals should be considered very cautiously and probably combined with other more general measures.

Empirical evidence

If one looks at employment composition over the last two decades, a common trend emerges in industrialized countries (see table 1): white collars present a rate of change higher than the average while the opposite occurs for blue collars (whose rate of change turns out to be negative in countries like Australia and Germany). As a consequence, white collars share is increasing everywhere (reaching more than 70% of the labour force in North American countries) while the share of blue collars shows a dramatic decrease, which is slower in highly industrialized countries like Germany and Japan. In addition, within white collars workers, the sharpest increases are detectable in the professional and managerial categories, whose rates of change are in general more than twice the overall average.

These data confirm the historical tendency to an increase in the ratio between high and low skilled workers. Unfortunately, no data are available in order to assess the role of technology in this long-term process, but it is plausible that the skill-bias, together with educational progress and social customs, has played an important role in determining these historical trends.

However, a different evidence emerges when attention is turned towards the newly industrialized countries: while Mexico shows a recent trend (1988-95) which is similar to those characterizing North American countries, Turkey, and especially Philippines and Egypt turn out to be still characterized by an extensive pattern of industrialization in which the blue collar share still show an increasing (or stable) trend.

While the internal composition of employment in the industrialized countries shows an increasing importance of the skilled component; the opposite happens if one looks at the unemployment rates. In table 2 unemployment rates by level of education are reported for a group of OECD countries. As far as men are concerned (no overall data are available), the differential between the rate of unemployment among the lower educated workers and the highly educated ones is increasing in all the countries but the less developed Turkey (see above), Germany (where data are heavily affected by the reunification process), the Netherlands (where massive part-time absorbs low skilled workers) and Ireland (where the difference is decreasing but still very large). As far as women are concerned, the only exceptions are Germany and the Netherlands again and Belgium (where the gap is still over 10%), Italy and Switzerland. Thus, it is possible to conclude that, all over the industrialized world, unemployment is characterized by an increasing skill bias which tends to widen the gap between the rate of unemployment among the unskilled workers and the same rate among the skilled one.

These results are confirmed by the relative patterns presented in table 3, which only refers to the US. As it can be seen, the gaps between the different unemployment rates by educational attainment are clearly widening over time.

Thus, the evidence presented in tables 2 and 3 is quite consistent with the skill bias hypothesis, at least with regard to the relative larger displacement of the unskilled workers. Turning the attention to the earning differentials, table 4 reports data about the trend of the ratio between the earnings of the highly educated workers and the earnings of the lower educated ones (ratio equal to 100 in the late ‘70s).

As it can be seen, larger inequality has clearly occurred in the US (going from 99 to 127) and in the UK (going from 108 to 121) but not in other countries: for instance, in the European countries the ratio is stable or even decreasing. In other words, that portion of the skill bias hypothesis concerning the emergence of wider earning differentials cannot be supported by data, at least in some OECD countries. Once again, the evidence presented in table 4 is very general and it is not directly related to any technological variable. Yet, as a matter of fact, the ICT era is not everywhere characterized by larger income inequality. One common interpretation is that the US (and partially the UK) have reacted to the skilled biased technical change through wage adjustments while the European countries have reacted through higher unemployment among the unskilled workers (see Unemployment and Wages and Unemployment and the Labour Market).

Finally, table 5 presents some evidence which is consistent with this interpretation: in the US the cost of labour concerning the blue collars has increased by 61% from 1982 to 1995, while the cost of labour concerning the white collars has increased by 76%. Thus, in the US, the skill bias has involved a wage/cost adjustment adverse to the unskilled workers.

Table 1: Panel 1 Employment growth by occupation Total percentage changes

Table 1: Panel 2 Employment shares by occupation

Table 2: Unemployment rates by level of education

Table 3: USA: Unemployment rates by educational attainment

Table 4: Earnings differential by education

Table 5: USA: Employment Cost Index

* The following contributions deal with the relative displacement which can affect low skilled workers as a consequence of the introduction of new technologies (especially Information and Communication Technologies and particularly computers). It is important to note that these papers (apart from the first two) do not discuss technological unemployment, that is the overall possible labour-saving effect of new technologies (see Unemployment and Technical Change), but only the relative bias against low skilled workers and in favour of highly qualified and better educated workers. The seminal contributions on this issue are Berman et al., (1994) and Doms et al., (1997).

Cooper, C.M.- Clark, J. (1982), Employment, Economics and Technology, Wheatsheaf, Brighton.

Vivarelli, M. (1995), The Economics of Technology and Employment, Elgar, Aldershot, (pp. 79 and ff.)

Calmfors, L. (1994), Active Labour Market Policy and Unemployment: A Framework for the Analysis of Crucial Design Features, OECD Economic Studies, n.22.

Lindbeck , A. - Molander, P. - Persson, T. - Petersson, O. - Sandmo, A. - Swedenborg, B. - Thygesen, N. (1994), Turning Sweden Around, MIT Press, Cambridge (Mass.).

Berman, E. - Bound, J. - Griliches, Z. (1994), Changes in the Demand for Skilled Labor within U.S. Manufacturing: Evidence from the Annual Survey of Manufactures, Quarterly Journal of Economics, vol. 109, pp. 367-98.

Baldwin, J.R. - Diverty, B. - Johnson, J. (1995), Success, Innovation, Technology, and Human Resource Strategies - An Interactive System, paper presented at the conference on "The Effects of Technology and Innovation on Firm Performance and Employment", Washington, May 1 and 2, 1995.

Machin, S. (1996), Changes in the Relative Demand for Skills, in Booth, A.L. - Snower, D.J. (eds.), Acquiring Skills: Market Failures , Their Symptoms and Policy Responses, Cambridge University Press, Cambridge, pp. 129-146.

Lauritzen, F. - Nyholm, J. - Jorgensen, O. - von Sperling, J.U. (1996), Technology, Education and Unemployment, in OECD, Employment and Growth in the Knowledge-based Economy, Paris, pp. 359-381.

Johnson, J., Baldwin, J.R. - Diverty, B. - (1996), The Implication of Innovation for Human Resource Strategies, Futures, vol.28, pp.103-119.

Doms, M. - Dunne, T. - Trotske, K. (1997), Workers, Wages, and Technology,Quarterly Journal of Economics, vol. 112, pp. 253-89.

Betts J. (1997), The Skill Bias of Technological Change in Canadian Manufacturing Industries, Review of Economics and Statistics, vol.79, pp. 146-50.

* The following articles are instead focused on the possible wage premium which can be associated with the use of new ICT (computers) technologies. The papers included in the first list support the hypothesis that the skill bias increases wage differentials. The seminal contribution on this subject is Krueger, (1993).

Krueger, A. B. (1993), How Computers Have Changed the Wage Structure: Evidence from Microdata, 1984-1989, Quarterly Journal of Economics, vol.108, pp.33-60.

Dunne, T. - Schmitz, J.A. (1995), Wages, Employment Structure and Employer Size-Wage Premia: Their Relationship to Advanced-technology Usage at US Manufacturing Establishments, Economica, vol.62, pp. 89-107.

* Instead, the following paper presents a survey and some descriptive evidence that challenges the view that there is a close link between computerisation, aggregate trends in skill composition and relative earnings. An alternative to the "skill-biased technical change hypothesis" is the view that many US employers have begun to adopt low-wage human resource strategies since the late 1970s. In other words, the deterioration of low-skill wages would have more to do with industrial relations than with biased technological change.

Howell, D.R. (1996), Information Technology and the Demand for Skills: A Perspective on the US Experience, in OECD, Employment and Growth in the Knowledge-based Economy, Paris, pp. 291-305.

* Finally, in this final list the Krueger conclusion is explicitly challenged (see Section 2). The seminal contribution on this subject is DiNardo and Pischke (1997).

Bartel, A.P. - Lichtenberg, F.R. (1987), The Comparative Advantage of Educated Workers in Implementing New Technology, Review of Economics and Statistics, vol. 69, pp. 1-11.

Chennells, L. - Van Reenen, J. (1997), Technical Change and Earnings in British Establishments, Economica, vol. 64, pp. 587-604.

DiNardo, J. E. - Pischke, J.S. (1997), The Returns to Computer Use Revisited: Have Pencils Changed the Wage Structure Too?, Quarterly Journal of Economics, vol.112, pp.291-303.

Entorf, H. - Kramarz, F. (1997), Does Unmeasured Ability Explain the Higher Wages of New Technology Workers?, European Economic Review, vol.41, pp. 1489-1509.

* About the vast policy debate on the skill bias, it is important to quote the following works (see Section 3).

Edmund S. Phelps, Rewarding Work: How to Restore Participation and Self-support to Free Enterprise, Harvard University Press, Cambridge (Mass.), 1997, pp. 198.

Snower, D. (1994), Converting Unemployment Benefits into Employment Subsidies, American Economic Review, vol. 84, pp. 65-70.

Snower, D. (1995), Unemployment Benefits: An Assessment of Proposals for Reform, International Labour Review, vol. 134, pp. 625-47.

* The proponent of the alternative view that trade (and not technology) is the main cause of the displacement of the low skilled workers is:

Wood, A. (1994), North-South Trade, Employment and Inequality: Changing Fortunes in a Skill-Driven World, Oxford Clarendon Press, Oxford.