British and American Changes in Interindustry Wage Structure Under Full Employment

DURING the past twenty years considerable work has been done concerning the behavior of wage differentials and wage dispersion.' In many of these studies attention has been confined to a particular wage differential, e.g., the North-South wage differential, and an attempt has been made to ascertain the average change over time in that differential in several industries. mere description of this important structural aspect of our industrial economy has general interest. Moreover, changes in wage dispersion and differentials imply shifts in the personal distribution of income which, in turn, have an impact upon the consumption function and upon resource allocation.2 It has been observed that in the period subsequent to the mid-I930's the personal distribution of income has become more nearly equal.3 This shift toward greater equality may be attributed, in part at least, to the absorption of the unemployed and the fuller employment of the partially employed. But changes in wage differentials can either strengthen or tend to offset the movement toward greater equality of income. Some studies have led to the conclusion that fuller employment accompanied by changes in wage differentials that strengthen the shift toward greater equality. Dunlop, Lebergott, and Ober,4 for example, have provided some empirical basis for the hypothesis that has been a trend toward a narrowing of the percentage differential between skilled and unskilled workers during the past half-century, and that the trend accentuated in boom periods of full employment and reversed in depressions. 5 This hypothesis has not gone unchallenged. Bell concludes that there is no logical pattern for cyclical variations in the structure of occupational differentials. . . . These cycli* This paper based largely on an M. A. thesis on file in the Duke University Library. work was done under the direction of Professor Frank A. Hanna and benefited from the comments of Professor Frank T. de Vyver. 1 See, for example, J. Dunlop, Cyclical Variations in Structure, this REVIEW, XXI (February 1939), 3039; S. Lebergott, Wage Structures, this REVIEW, XXIX (November I947), 274-85; H. Ober, 1907-1947, Monthly Labor Review, LXVII (August I948), I27-34; P. Bell, Cyclical Variation and Trend in Occupational Differentials in American Industry since 19I4, this REVIEW, XXXIII (November 1951), 329-37; J. Bloch, Regional Differentials; I907-46, Readings in Labor Economics, ed. F. Doody (Cambridge, Mass., 1950), 8I-9I; R. Lester, Differentials: Developments, Analysis, and Implications, Southern Economic Journal, XIII (April I947) 38694; H. Ober and C. Glasser, Regional Differentials, Monthly Labor Review, LXIII (October 1946), 5II-25; D. Roberts. The Meaning of Recent Changes, Insights into Labor Issues, ed. J. Shister and R. Lester (New York, 1948), 226-37; F. Meyers, Notes on Changes in the Distribution of Manufacturing Earners by StraightTime Hourly Earnings, 194I-48, this REVIEW, XXXII (November I950), 352-55; D. Brady, Equal Pay for Women Workers, Annals of the American Academy of Political and Social Science, XXLI (May 1947), 53-60. 2 It should be noted that the wage structure considered here the structure of average hourly earnings. Since the accounting period only one hour, this structure not strictly comparable with the personal distribution of income, in which the accounting period typically one year. In order to render them comparable, factors such as the variation among workers in annual hours of work must be taken into account. Despite the discrepancy, a direct relationship between the wage structure of an industry and the personal distribution of income would seem to exist. It should also be noted that the portion of labor income emanating from an industry in the form of salaries not included in the wage structure. Since average hourly earnings are used, the subsequent analysis concerned actually with the dispersion of average hourly earnings rather than the dispersion of wage rates. See J. Dunlop, op. cit., for a discussion of differences between average hourly earnings and wage rates. Attention here confined to the basic monetary remuneration of workers. In order to include the effects of fringe benefits in the analysis, data on the value per hour of such benefits to each worker would be required. Such data would provide new frequency distributions which would show the dispersion of average hourly earnings and fringe benefits combined. No such data are available. With regard to the North-South wage differential in the cotton textile industry, it has been suggested that fringe benefits tend to increase differences between the North and South. See the Report on the New England Textile Industry (by Committee Appointed by the Conference of New England Governors, 1952), pp. I37-43. 'H. Miller, Factors Related to Recent Changes in Income Distribution in the United States, this REVIEW, XXXIII (August 95I), 214-I8. 4 Dunlop, op. cit.; Lebergott, op. cit.; Ober, op. cit. 'Paraphrased in Bell, op. cit., p. 329.

and rates,' thought by many economists to be a highly desirable match, do not appear to possess those characteristics of compatibility upon which the perfect marriage depends. Disconcertingly, in the short-run, they display a clear tendency to move either in opposite directions or, if in the same direction, at quite different rates. The postwar experience in this country and the possibility of continuing high levels of employment in the future, in an economy imperfectly protected from the raids of pressure groups, have caused intensified concern over the relationships of these two variables. We may appropriately inquire whether it is reasonable to expect man-hour output and money wages, from year to year or over moderately longer periods, to progress hand in hand.2 It is frequently suggested that they should move in close harmony over some period of time, since, if they did, price would be facilitated. Lord Keynes was particularly influential in popularizing this view.3 Sir William Beveridge ' and Professors Hansen 5 and Slichter,6 among others, have added the weight of their support. These normative proposals of economists have been elevated to the level of a national policy by the Council of Economic Advisers which advocates wage increases which are in line with productivity trends. 7 This article is not concerned with whether this norm is a desirable one or not, but rather with a realistic appraisal of the likelihood of its being followed. The time period under consideration is the short-run. Over relatively long periods the three quarters of a century before World War I and the quarter of a century after it manhour output and wages increased by approximately the same amount. The short-run movements of the two variables, however, are also important and have an effect on the operation of the economic system on costs, profits, prices, and real wages. Significant short-term departures from the pattern of the long-term movements will be shown to have occurred. It will be suggested that, given the business cycle and our present collective bargaining institutions and practices, the short-run gearing of wages to productivity in the future is unlikely to be achieved with any great precision. The distinct impact of the business cycle on productivity and on wages, a neglected consideration, and the realities of the setting process make implementation of the policy more difficult than commonly supposed. The two variables almost inevitably act 1 The term productivity will be used throughout to mean physical, not value, productivity; and rates to mean money, not real, wages. 2 Real wages per hour and man-hour output, unless there is some change in the earners' share of the national product, must change approximately together in the longrun and in fact they do. (See Jules Backman and Martin R. Gainsbrugh, Productivity and Living Standards, Industrial and Labor Relations Review, January I949.) This present article is not concerned with the relationship of real wages to man-hour output, nor with long-run historical relationships between money wages and productivity in the economy at large, nor with long-run relationships in individual industries. Professor Hansen has noted, for example, that over a period of about seventy-five years prior to World War I money wages rose commensurate with improved productivity. (A. H. Hansen, Wages and Prices: The Basic Issue, New York Times Magazine, January 6, I946.) Professor Dunlop has discussed the longrun impact of productivity on inter-industry differentials and demonstrated that it has had a pronounced effect. (John T. Dunlop, Productivity and the Wage Structure, Income, Employment and Public Policy, New York, I948.) aJ. M. Keynes, The General of Employment, Interest and Money (New York, I936), p. 27I. See also Seymour E. Harris, Attack on Laissez Faire and Classical Economics and Wage Theory in The New Economics: Keynes' Influence on and Public Policy, edited by Seymour E. Harris (New York, I947), pp. 554-56. For a discussion of earlier suggestions along the same lines as Keynes, see A. G. Pool, Wage Policy in Relation to Industrial Fluctuations (New York, I938), p. 286 ff. 'William H. Beveridge, Full Employment in a Free Society (New York, I945), p. 200. Op. cit. He views stability of efficiency-wages as the appropriate goal for policy. (Monetary and Fiscal Policy, New York, I949, Ch. 8.) 6Sumner H. Slichter, The Challenge of Industrial Relations (Ithaca, N. Y., I947), p. 89. See also Basic Criteria Used in Wage Negotiations, Chicago Association of Commerce and Industry, I947, pp. 44-48. 7Council of Economic Advisers, The Annual Economic Review-January 1949, p. 45.

COLLECTIVE bargaining results in the periodic negotiation of a structure of wage rates and of other conditions of employment. Variation in the structure may be thought of as a step function through time constant between the effective dates of bargained changes and rising or falling discretely on those dates. As illustrated below, the level of wages shifts at times t, t + I, and t + 2 and is otherwise constant. It is these shifts that economists Wage_Iordinarily think of as changes in the price of labor to the employing t t + I t + 2 firm. Time Because average earnings data are more readily available than are wage rate data, students of wage problems have often been forced to rely upon earnings as a measure of the price of labor. This measure, of course, reflects not only changes in basic wage rates but also changes in the distribution of workers throughout a given wage structure. For example, if at some time an industry increases its employment of highly paid, skilled workers, average hourly earnings will rise. But is such a rise to be interpreted as a rise in the price per unit of labor? The answer would appear to be no, if we rely upon deductions from the theory of the firm. Starting from an equilibrium situation, a firm that increases its employment of a factor is most likely responding to a rise in the derived demand for that factor's services. The derived demand may have risen because of changes in marginal physical product, the price of the product, the prices of other factors, and so forth.' Just as a change in the proportion of more highly paid workers will cause a rise in average hourly earnings while basic rates of pay remain unchanged, so, too, will other variations in the composition of the work force. Studies of the Bureau of Labor Statistics 2 cite the following additional influences that tend to raise earnings: increases in the proportion of workers who are male, who work in high-wage regions (e.g., the North and the Northwest), who are employed in more densely populated urban areas, who work at premium rates, and who are paid according to an incentive or piece-rate system. Variations in some of these influences are not so easily interpreted by applying microeconomic theory as is an increase in the employment of a single factor, assumed homogeneous. Thus, for example, if a firm opens a plant in the low-wage area of the country, thereby decreasing the average earnings of all its workers, should this be interpreted as an increase in the average supply of labor as a whole or as an increase in the demand for labor of a particular grade? Or is this the discovery of a resource that justifies our speaking of a different production function? We may say that the answers to these questions depend in part upon what was in the mind of the entrepreneur who decided on opening the new plant. But this does not carry our analysis far until we know what was in his mind. Fortunately, for the purposes of this paper we need not define unequivocally the theoretical meaning of such changes as those discussed. It is enough to remark that many investigators of wage movements have found it necessary to * I am indebted to Leonard S. Rhynus, Susan Crittenden, and Terry Allen for help in assembling materials for this study. The companies and unions included here were generous in answering detailed questions about their negotiations. Financial aid was received from Wesleyan University and from the Research Committee for Public Affairs at Wesleyan, a committee charged with allocating funds from a Ford Foundation grant. Philip Brown processed the data for the wage rate index. William Barber, Harry Douty, Melvin Lurie, and several economists in industry offered valuable comments on an earlier draft of this paper. Any errors in the work are, of course, my own. 1 While the foregoing is thought to be a fair general interpretation of employment changes, it is not meant to preclude the cases wherein union work rules can compel an increase in demand. In such cases, indeed, supply (pressure) creates its own demand! 2 Harry Douty, Union and Nonunion Wages, in W. S. Woytinsky and Associates, Employment and Wages in the United States (New York, I953), 493-50I.

This paper investigates the degree of earnings inequality in Great Britain over the period 1975-96 using individual-based data. It finds that the contribution of within-group inequality to both earnings inequality cross-sectionally and to its trend over time is substantially more important than the contribution of between-group inequality. Thus, the primary source of increasing inequality in the overall earnings distribution is increasing inequality within regions and not differences in average earnings between regions. A decomposition of the Gini coefficient is also adopted to illustrate how regional convergence in average earnings has been accompanied by increasing overall earnings inequality in the UK. A partir des données auprès des particuliers, cet article cherche à examiner l'importance de l'inégalité des revenus régionaux en Grande-Bretagne de 1975 à 1996. Il en résulte que la contribution de l'inégalité à l'intérieur du groupe à l'inégalité des revenus par échantillon et sur le temps s'avère nettement plus importante que ne l'est l'inégalité entre groupes. Ainsi, la source primordiale de la croissance de l'inégalité de la distribution globale des revenus est l'accroissement de l'inégalité intrarégionale et non pas l'écart des revenus moyens interrégionaux. On se sert aussi du coefficient Gini désagrégé afin de démontrer comment la convergence régionale des revenus moyens se fait conjointement avec une hausse globale de l'inégalité des revenus au Royaume-Uni. Dieser Aufsatz untersucht das Ausmaß der Unterschiede von Einkommen in Großbritannien im Zeitraum 1975-1996, wobei er sich auf Daten von individuellen Fällen stützt. Der Befund zeigt, daß der Beitrag der Unterschiede innerhalb einer Gruppe sowohl im Querschnitt als auch zu ihrer langfristigen Tendenz zum Einkommensunterschied weitaus bedeutender ist als der Beitrag zu Unterschieden zwischen verschiedenen Gruppen. Die Hauptquelle zunehmender Unterschiede in der allgemeinen Einkommensverteilung ist daher die zunehmende Ungleichheit der Regionen, und nicht die Unterschiede in durchschnittlichen Einkommen zwischen einzelnen Regionen. Eine Zerlegung des Gini Koeffizienten wird auch dazu benutzt, aufzuzeigen, wie regionale Konvergenz bei durchschnittlichen Einkommen mit zunehmenden allgemeinen Einkommensunterschieden im UK einhergeht.

This paper examines the changes in the inter-industry wage structure experienced by Russia since 1993, as part of its transition from a plan-based economy to a more \market oriented structure. Using two Russian household panel data sets, the RLMS and the RUSSET, we nd that since the transformation process began, the dispersion of inter-industry wage structure has increased. Moreover, Russia exhibits large movements in wage premia, as industries respond to massively changing demand conditions. The issue of wage arrears (unpaid wages or outstanding pay), which aects half of all employees, plays an important role in the determination of wages. Studies, which do not account for wage arrears, overestimate the overall inter-industry wage dispersion. Despite movements towards a privatized market economy, we still nd government ownership and Soviet network eects play an important role in determining the wage structure.

IT IS inconceivable that any one should not want full employment, or at least should express his dislike openly. The wage earner wants it, because it means the maintenance of his standard of living. The man wants it, because it means the continuous operation of his productive unit. The general public wants it, because it means an unbroken and maximum flow of the goods and services upon which it depends for the maintenance of the good life. True, there may be a small number of moss-backed, casehardened industrialists who would welcome a constant pool of unemployed as a means of breaking down labor resistance and keeping wages low. But in these days they are not likely to say much about it in public. It follows that those who are now opposing proposals to insure full employment, by legislative or other measures, must do so because they object to the means rather than the end. Almost invariably the means that are opposed represent an extension of governmental (usually federal) activity or authority, and it is upon this that the objections are based. This would seem to indicate that governmental measures to secure full employment are deprecated either on the ground that they would not be effective for the purpose indicated, or else that the extension of governmental activity into the field of is so abhorrent in itself that it is worth avoiding even at the high price of depressions, failures, and untold human suffering and degradation. For it is very generally admitted today, not only by far-seeing liberal leaders like Henry Wallace, but also by intelligent men and the orthodox economists who are their theoretical mentors.and practical protagonists, that free by itself cannot guarantee full and continuous employment for the total working force of such a highly capitalized country as the United States. Even the most assured and vociferous exponents of the leave everything to free enterprise school would hardly want to back up their protestations with an ironclad promise not to go running to the government for relief should a blue Monday arrive, nor even to accept the assistance voluntarily offered by a paternally-minded administration. The business cycle, with its recurrent extremes of boom and depression, is now recognized by the more emancipated of the conventional economists as an unavoidable feature of a capitalistic, or price-and-profit, system. This relationship is accepted with varying degrees of complacency or resignation by those who regard the capitalistic system as the perfect flower of social evolution and cannot conceive that any other arrangement could possibly be on the whole preferable. Explanations of the process, and of its manifestations, vary. Some students find the central cause in the essentially unworkable and paradoxical character of the price-andprofit system itself. It has been conclusively demonstrated,' that monetary profits on a society-wide scale are a physical and mathematical impossibility. Even some writers of essentially orthodox economics textbooks admit this, as, for example, Gemmill and Blodgett when they say that, In the long run, therefore, we expect to find that the profits of a competitive industry just about balance the losses of that industry.' Mr. R. H. Doane, in his book The Measurement of American Wealth, has demonstrated the proposition by the statistical approach, showing by carefully correlated figures that over the long stretch of time the profits of American industry as a whole are almost precisely balanced by the

What GAO Found: Household spending patterns varied by age, with mid-career households (those aged 45-49) spending more than older households. For example, according to 2013 survey data from the Bureau of Labor Statistics (BLS), mid-career households spent an estimated average of around $58,500, while young retiree households (those aged 65-69) spent about 20 percent less. While the share of spending was consistent for some categories, other categories had larger variations across age groups. For example, housing expenses comprised the largest share of spending regardless of age, while older households spent more out of pocket on health care than mid-career households. Spending was less variable across age for low-income households compared to other households. For example, there was not a significant difference in average spending between mid-career and young retiree households in the lowest income quartile, compared to an approximately $20,000 difference for the highest income quartile. These variations in spending patterns have implications for the resources households need to maintain their standard of living in retirement. Researchers and financial industry professionals develop target replacement rates — the percentage of income to aim for in retirement — based on certain key factors, including spending, household characteristics, and pre-retirement earnings. GAO's analysis of the literature found that calculating an appropriate replacement rate can be complex. For example, there is debate over whether households that have raised children should target a lower replacement rate than households that have not. In addition, a worker's pre-retirement earnings could be defined as earnings at the end of the worker's career or as average earnings over the course of the career. Despite these complicated considerations, target replacement rates cited in the articles and reports GAO reviewed typically range between 70 and 85 percent. Some financial industry professionals told GAO that they develop customized targets that take into account workers' assets and expected spending, while others questioned the usefulness of replacement rates. The information and tools on replacement rates that the Department of Labor (DOL) provides may be too limited to help workers understand how to use such rates for retirement planning. DOL's Employee Benefits Security Administration's (EBSA) website provides information and tools to help American workers better plan for retirement, including a tool to help workers calculate their retirement income needs as a percentage of preretirement income. While EBSA's materials note that a target replacement rate can vary based on individual circumstances, they do not include specific examples of demographic groups that research indicates can result in higher or lower income replacement needs, or how much a replacement rate might need to be adjusted for those groups or for other individual circumstances. Without additional information, workers may not understand how to adjust target replacement rates when planning for retirement. Further, EBSA's worksheet and online tool for calculating how much to save use a default replacement rate with no opportunity for a user to adjust the rate based on individual circumstances. Without the ability to adjust the replacement rates used in planning tools, workers may over- or under-estimate how much they need to save for retirement. Why GAO Did This Study: Part of DOL's mission is to promote the retirement security of America's workers, a goal that has become increasingly challenging. One tool for assessing the adequacy of retirement income is the replacement rate. However, recommendations for the replacement rate that a household should target vary widely, in part because of the diverse underlying assumptions used to develop the rates. GAO was asked to review what consumption in retirement looks like and how target replacement rates are developed. GAO examined (1) whether and how spending patterns vary by age, (2) key factors used to develop target replacement rates, and (3) the usefulness of information on such rates provided by DOL. GAO analyzed data from the BLS's 2013 Consumer Expenditure Survey, the most recent available; analyzed 59 articles and reports that discussed how to develop, calculate, or evaluate replacement rates; collected non-generalizable information from 14 retirement services firms and financial planners recommended by researchers and actuaries who have studied replacement rates; and reviewed DOL materials and interviewed officials. What GAO Recommends: GAO recommends that DOL provide additional examples and guidance on using a replacement rate for estimating retirement savings needs in its planning tools, and modify the planning tools so the rate can be adjusted. DOL generally agreed with our recommendations and plans to add information and provide options for adjusting replacement rates in its planning tools by June 2017.

Korean economy has changed rapidly since the 1997 Korean financial crisis. This article investigates whether the inter-industry wage differentials in the Korean manufacturing industry are consistent over time and how the inter-industry wage structure has changed after the financial crisis. We used the 1995 and 1999 survey report on the wage structure–Korea. Our empirical results provide evidences for a wider inter-industry wage differentials and changes of the wage structure. After the crisis, company size, region (live in capital), sex, tenure, education are more critical to determine wages. According to factor analysis, the job quality factor is most closely related to wages before the crisis, while the industry attributes factor does in 1999.

The Kerner Commission and the Irony of Antiracist Politics

We provide a comprehensive analysis of income inequality and income dynamics for Germany over the last two decades. Combining personal income tax and social security data allows us—for the first time—to offer a complete picture of the distribution of annual earnings in Germany. We find that cross‐sectional inequality rose until 2009 for men and women. After the Great Recession, inequality continued to rise at a slower rate for men and fell slightly for women due to compression at the lower tail. We further document substantial gender differences in average earnings and inequality over the life cycle. While for men earnings rise and inequality falls as they grow older, many women reduce working hours when starting a family such that average earnings fall and inequality increases. Men's earnings changes are on average smaller than women's but are substantially more affected by the business cycle. During the Great Recession, men's earnings losses become magnified and gains are attenuated. Apart from recession years, earnings changes are significantly right‐skewed reflecting the good overall state of the German labor market and increasing labor supply. In the second part of the paper, we study the distribution of total income including incomes of self‐employed, business owners, and landlords. We find that total inequality increased significantly more than earnings inequality. Regarding income dynamics, entrepreneurs' income changes are more dispersed, less skewed, less leptokurtic, and less dependent on average past income than workers' income changes. Finally, we find that top income earners have become less likely to fall out of the top 1 and 0.1%.

As was discussed in Chapter 4, the gross difference between male and female earnings (i.e. ‘statistical’ discrimination) will overestimate the extent of ‘pure’ discrimination. The whole of the statistical difference is unlikely to be solely the result of male. prejudice leading to women being paid less than men doing the same work and denying women equal opportunities for advancement. Part of the difference will be due to causes which are ‘economic’ rather than discriminatory in origin. In Chapter 3 we attempted to show the contribution of various factors to this differential and we were able to isolate occupational distribution, hours of work and age. Allowing for these factors the major source of the difference in average earnings was the fact that women were paid substantially less than men within each occupational category. In this chapter we attempt to isolate supply-side differences between the sexes which would ‘explain’ at least a part of the difference, so that the remainder can then be regarded as an upper estimate of the extent of discrimination.

ObjectivesRA is known to impact work ability, but much of this knowledge comes from historical comparisons vs the general population that neither reflects current RA management nor distinguishes between effects of RA and pre-existing socio-economic conditions of patients. We therefore aimed to examine earnings of patients before and after RA diagnosis, using recent data and sibling comparisons.MethodsSwedish register data were used including demographic information and healthcare utilization. Participants were patients with RA (aged 30–60 years, diagnosed with RA between 2006 and 2017) identified in the Swedish National Patient Register, and their same-sex siblings (n = 2433:2433; mean 48 years; 72% women). Earnings data for 2001–2019 were retrieved from Statistics Sweden and analysed from 5 years before to 5 years after RA diagnosis.ResultsNo differences in average earnings were observed between siblings during the 5 years before diagnosis, but during the 5 years after diagnosis, patients with RA earned on average 5.4% less annually [–1430€ (95% CI –2130, −720)] than same-sexed siblings. The change in earnings for the subgroup diagnosed between 2006 and 2010 was –8.2% [–2020€ (95% CI –2930, –1120)] but for patients diagnosed between 2011 and 2017, there was no statistically significant change in earnings compared with siblings [–1.5%; –420€ (95% CI –1490, 640)]. Subgroup analyses demonstrated a more negative impact on earnings for older individuals and those with lower education level.ConclusionRA diagnosis was associated with lower earnings in comparison with same-sex siblings, particularly for older individuals and those with lower education level. The negative impact of RA on earnings declined or disappeared over the study period.

The objective of this paper is to employ the Hurricane Katrina evacuation into Houston, TX as a natural experiment to estimate the effect of large scale in-migration on regional earnings. Given the characteristics of the evacuees, their influx would have caused the supply of applicants for lower skilled jobs to increase proportionately more than for higher skilled jobs. We utilize a differences-in-differences-in-differences methodology in which we compare differences in average earnings in the low-skill, non-traded goods industries to the corresponding differences in the set of high skill industries in the Houston and Dallas-Fort Worth MSAs before and after the Katrina-induced migration. Unlike previous studies, we include a measure of the post-storm increase in demand for local goods and services on the demand for labor. We find evidence that the relative average payroll in the low-skill non-tradable goods industries in Houston decreased by .7% when compared to the relative change for the same group of industries in Dallas. Our findings also suggest that failure to account for demand-side influences following the migration results in significant under-estimation of wage effects due to the shift in labor supply.

This study investigates the extent to which differences in average earnings between men and women may be the result of sorting by the sexes into jobs with different average levels of disagreeable and agreeable working conditions. An analysis of data from the 1977 Quality of Employment Survey shows that, on average, men and women hold jobs with substantially different working conditions and that these differences are of a pattern suggesting the need to pay higher wages to attract employees to the jobs held by men. Estimation of wage equations shows that these differences in working conditions contribute significantly to the ability to explain average earnings for each sex.

Somewhat divergent approaches to finding the components of group differences (for example, the difference in average earnings across gender or racial groups) have grown up in the literatures of economics and sociology. Both, however, are variants of the same underlying approach. This article illustrates different approaches and draws attention to the conditions under which two, three, or four components of the income gap between groups can be usefully distinguished. The two central issues are which standard one uses to evaluate endowment differences between groups, and whether the scales used to operationalize underlying concepts have arbitrary zero-points or not. This latter difficulty has often been neglected in empirical applications using variables such as education, region, occupation, industry, and marital status. In such cases, the choice of a comparison group is inherently arbitrary, and nothing other than an arbitrary decomposition of the residual group difference can result.