We have seen that any realistic analysis of labour markets
must allow for both heterogeneities of labour and market imperfections. Indeed, one of the major defects of the simple marginal productivity approach is that it says very little about differentials in wages between different types of workers in different labour markets. To this problem we now turn. The most comprehensive treatment of the subject (Reynolds and Taft [77]) deals with five types of differential. Two of these — interpersonal and inter-firm — have already been discussed. What follows is concerned with occupational, inter industry and geographical wage differentials. The fragmentation of labour markets is most obvious along occupational lines. Mobility between occupations tends to be more impeded, by 'market' and 'non- market' forces, than mobility between industries and even regions. (At least as a generalisation — movement from floor-sweeper to labourer is clearly less difficult than mobility of floor-sweepers from northern Scotland to London.) Differences in pay between occupations form a large part of the differentials between industries, and industry differentials account for a major part of regional differentials. These rather pragmatic considerations determine our order of treatment.
OCCUPATIONAL DIFFERENTIALS
Adam Smith ([88] pp. 112—23) suggested five reasons for
differences in pay between occupations. Firstly, the 'agreeableness' of the job: net advantages, not merely cash payments, must be considered. Secondly, the degree of constancy of employment, the liability to unemployment in different occupations. Thirdly, the 'probability of success' differs between jobs, being less for barristers than for solicitors, for example. Both these factors suggest that discounted expected earnings over a lifetime are what is involved in rational job choice, rather than pay at any one point in time. Fourthly, 'the trust to be reposed in the worker', or responsibility requirements : goldsmiths in Srnith's day, workers maintaining expensive and delicate machinery in our own, are well paid on this account. Lastly, the 'cost of learning the trade', comprising tuition fees and earnings forgone while undergoing training. (In modern terms this is 'human investment' ; cf. Becker [3].) Abstracting from the first four factors, the fifth provides us with a simple long-run theory of occupational wage differentials. Assume that workers bear all the costs of investing in themselves. If natural ability is irrelevant, training open to all, and all non-pecuniary advantages identical, the long-run supply of every occupation will be perfectly elastic: pay differentials between jobs will equate the rates of return on the various amounts of human investment required. These assumptions may be modified — e.g. by allowing for variations in the suitability of workers for positions of responsibility — without substantially altering the conclusion that differences in pay depend largely on differences in the level of education and training required for different jobs. The demand for each occupation has no impact, and relative shifts in demand will have no effect on differentials. An alternative theory of occupational wage differentials can be developed by reversing one or two of these crucial assumptions. Again assume that non-pecuniary factors are identical, and that Smith's second and third elements are inoperative, so that the risks of failure and unemployment do not differ between occupations. Suppose, however, that natural ability is all-important, that part of the labour force is born as sheep and part as goats, and that their abilities cannot be altered by education or training. Then, since the supply of each is perfectly inelastic, their earnings will differ only because of differences in the strength of demand for their services. If goats are in greater demand, they will be paid more, and earn a rent equal to their differential over the sheep. Changes in this differential will occur only because of changes in relative demand. (For an exposition of both models, see Reder It is unnecessary — and unwise— to rely solely on genetic differences to derive a theory of occupational wage differentials. The division of society into non-competing groups may occur on other grounds: for example, social prejudice (the lack of Cockney ambassadors) or inequality in educational opportunity. Both factors may be eliminated in the long run as occupational supply curves become much more elastic. The consequent narrowing of differentials suggests that the goats were only earning quasi-rents. A process of this sort will often occur if training periods are long and changes in demand cannot easily be foreseen: the relatively high earnings of computer staff in the 1960s are an example. Demand factors tend to dominate occupational differentials in the short run, and underlying supply factors (above all human investment) in the long run. We know (Reynolds and Taft [77]) that there was a secular decline in occupational wage differentials in both the U.K. and the U.S. over the last half-century. This is true in general between both manual and white-collar workers (Routh [85]), and skilled and unskilled manual occupations. There is also some evidence that differentials have moved counter- cyclically, tending to rise in depressions and fall in booms. (Ozanne [64] denies this tendency, but his data are drawn from one firm only.) How far are these trends explicable in terms of our simple models, and what alternative explanations can be offered ? Part of the reduction in differentials must be attributed to the massive and general expansion in state education. This, with increased equality of access to human investment, has reduced the rate of return to it, partially breaking down the barriers between non-competing groups. In the U.S. the cessation of mass immigration of unskilled labour was an additional factor. It has also been suggested (Keat [39]) that, as incomes rise, the non-monetary factors in a job become increasingly important to the worker, encouraging him to prefer lower-paid but more enjoyable work. If occupational supply curves are imperfectly elastic even in the long run, shifts in relative labour demand may have affected differentials. Modern technology is often assumed to have stimulated the substitution of less skilled for more highly- skilled labour, and this would tend to reduce differentials. Reder [68] suggests that the loosening of hiring standards in booms increases the supply of skilled labour relatively to that of unskilled, and the reverse process operates when aggregate demand is low. He explains the failure of differentials to widen in some slumps by the existence of a conventionally accepted 'social minimum' standard below which the lowest-paid are not allowed to fall. Partial support for his thesis comes from the tendency for the relative severity of unemployment of lowlyskilled workers to increase in depressions. Oi's [63] theory of labour as a semi- fixed factor leads to similar conclusions via differential fluctuations in labour demand. One big problem is the evidence — clear for the U.K., less so for the U.S. — that major reductions in occupational differentials have been sudden, coming during and immediately after the two world wars, and breaking periods of long-run stability which in some cases lasted for centuries (Phelps Brown and Hopkins [11]). But changes in education, and in relative demand under the influence of technical progress, have been much more gradual. This gives rise to an identification problem, for these periods of major change have coincided with rapid inflation (Perlman [66]) and the growth of mass unionism among lower-skilled workers (Turner [97]). Either or both might be held responsible for narrowing differentials independently of broader economic factors. Routh [85] suggests that occupational differentials are inherently stable because of social conservatism, being displaced only by major shocks. A compromise position is possible: changes in the relative supply of and demand for different occupations play only a permissive role, requiring specific institutional forces to bring them into operation. Reynolds and Taft [77] make a similar point in arguing that American differentials were too high, on simple economic criteria, at the turn of the century, and that union pressure has produced a more optimal occupational wage structure. INTER-INDUSTRY DIFFERENTIALS
Assume, with Reder [69], a perfect labour market with
homogeneous labour in each occupation. All workers in each occupation are thus paid the same. Average wages can differ between industries only because of differences in the 'skillmix' : industry A's average wage will exceed industry B's if and only if A has a larger proportion of highly-skilled workers. If inter-industry differentials do exist, they will change only if there are changes in occuPational wage differentials, or differential changes in the skill-mix. (A's percentage wage advantage over B will widen if, for example, its skill-mix improves while B's remains the same.) Reder relaxes the assumption of perfect occupational labour markets with the postulate that the short-run supply curves of specific occupations to specific industries are upward-sloping. To expand its labour force, A must pay each occupation a premium over its earnings in B. Thus in the short run changes in inter-industry differentials will be positively correlated with relative changes in employment between industries ; in the long run this relationship disappears as equilibrium is restored. Evidence on this hypothesis is conflicting, and testing raises important analytical problems (Perlman [65]). Other empirical evidence strengthens our doubts as to the validity of either the long- or the short-run version of Reder's theory. The following have been found, although not unanimously, to be relevant to the structure and evolution of interindustry wage differentials: trade union activity (Lewis [51]) ; product market concentration (Weiss [99]) and the average size of plants; profitability and the level and rate of growth of productivity (e.g. Brown [8]). All these factors are inconsistent with Reder's simple models. Even productivity differentials present problems, for if labour mobility is perfect and the law of equal price prevails, either long-run marginal productivities are equal in all industries or equilibrium conditions are absent. The influence of trade union activity will be explored below. At least part of the contribution of product market concentration to inter-industry differentials seems to be explained by the higher-paying firms having 'the pick of the market', selecting the ablest members of each (heterogeneous) occupational group (Weiss [99]). The role of plant size and profits may be explained in the same way, though in view of our knowledge of substantial market imperfections even within localities this is unlikely to be the whole answer. It is unclear how far these factors influence the course of inter-industry differentials over time. One would not, for example, expect highly concentrated or well-unionised industries to have a long-term tendency to pull away from the others in the earnings league. The actual trend of inter- industry differentials is still disputed, though it seems that the ranking of industries by average wages has been relatively stable in the long term, while the percentage dispersion has probably narrowed (Cullen [16]). (Absolute cash dispersion is unlikely to have narrowed; a 50p weekly differential, huge in 1900, is probably within the margin of statistical error today !) The forces of inertia can easily be exaggerated, and major short-run fluctuations ignored; in one U.K. depression (1920— 4) money-wage reductions ranged from zero to 60 per cent. Likewise there is still little agreement generally on the factors influencing changes in the inter-industry wage structure. This would perhaps be no great loss, except that wage statistics are generally collected on an industry basis (in the U.K. to the exclusion of almost all others).
GEOGRAPHICAL DIFFERENTIALS
One reason for the plight oflow-wage areas is the
concentration of low-paying industries that they have collected (Ministry of Labour [60]) — the industry-mix. This in turn is partly a reflection of regional variations in the skill- mix. Low-wage regions are weak in their endowments of highly-trained, well-educated labour. Regional wage differentials may thus be seen as the sum of two elements: an excess of low-paying jobs, due to unfavourable skill- and industry-mixes; and a tendency for relatively low earnings in each job. The typical research procedure has been to standardise for one or more of these explanatory variables, and to treat the residual differential as reflecting exploitation in the technical sense of unequal pay for equal work. These techniques have been widely used to analyse the North—South differential in the U.S. Fuchs [26] obtains a very crude approximation of the effects of regional differences in the skill-mix by assuming that the South's stock of human capital was the same as that of the rest of the U.S., and found that one-third of the unadjusted wage gap between the South and the national average was thus explained. A further third was found, by a similar procedure, to be due to the smaller average size of Southern towns and cities (reflecting, perhaps, greater employer monopsony power in small labour markets). Much of the residual third seems due to racial discrimination: not only do Blacks form a major part of the Southern labour force, but racial discrimination in employment is more severe in the South than in the U.S. as a whole. Fuchs and Perlman [27] used a similar procedure in showing that the South's inferior industry-mix was a major cause of the differential. They estimate that the South's relative wage level will rise only slowly, for the industries which it is attracting have been precisely those with a relatively low average wage in the U.S. as a whole. It is not clear how this finding can be reconciled with Scully's [87] report that the capital—labour ratio — and hence, ceteris Paribus, labour productivity — is now above the national average in the South. This refers only to physical capital; Scully found that the South's deficiency in human capital was the single biggest factor in the region's continued wage disadvantages, with racial discrimination running it a close second. All researchers are agreed that the South's relative earnings level has slowly improved since 1900, a finding which is. repeated in fragmentary evidence for low-wage areas in the U.K. (Knowles and Robertson [44]). It is uncertain how far this has been due to the in-migration of capital predicted by theory and verified by Scully's results, and how far to institutional factors like the extension ofunionisation to the South, and Federal minimum wage legislation. (In Britain the emergence of industry-wide collective bargaining seems to have been the single most important factor.) Confusion is increased by the dispersion of regional differentials between industries: some Southern industries pay more than their Northern counterparts, and some very much less (Lester [50]). This evidence, confirmed by a rudimentary examination also for regional differentials in Britain, serves to underline the severe limits to our knowledge of wage differentials generally.