Industrial Relations Section Princeton University Working Paper No. 63 February 1S75 Work Effort and Alternative Methods of Remuneration John H. Pencavel Stanford University Work Effort and ALterr.at.ive Methods of Remuneration John H. Pencavel* The Incidence of Wage Payment Methods It has become customary in economics to measure the input of labor services in production over a period of time by some combination of the number of workers employed and their hours of work adjusted in some fashion for variations in acquired skills. This has been a mo3t useful procedure in assisting our understanding of a large class of behavior in the labor market. In this paper, however, another dimension of the input o£ labor services receives particular attention and emphasis, namely, the intensity and diligence of work effort manifested by each employee working individually and by all employees working together as a team. The neglect by economists oi this appect of the exchange of labor services is a relatively recent phenomenon. For instance, Alfred Marshall stressed how the effective supply of productive factors depends upon "...the wilMingness of those, in whose charga it is, to apply it in production" and he wrote, therefore, of the "supply of efficient wcrk.'^" He censured Bohm-Bawerk for believing that wages were determined by decisnd alone "for even if ths number of [working]] hours in the year were rigidly 2/ fixed, which it is not, the intensity of work would remain elastic — Yet analytical work in economics of a more recent vintage has tended to ignore the intensity of work effort and the subject has become the almost _ Among the many people who have assisted me In the preparation of this manuscript, I am especially indebted to Yoram Barzel, Albert Raa3, a\i.d Joseph S. Stigiits. By awarding me a National Fellowship, the Hoover Institution at Stanford University and W. Glenn Campbell in particular made it possible for me to embark on thi9 project. -^Marshall [1920], pp. 437-39. 2/ -Marshall [1920], p. 438, n. 1. -2- exclusive province of industrial relations researchers. As one such student has observed, "...while the economist thinks primarily of an increase in the number of workers when he speaks of a change in the supply of labor, managers in practice—especially in times of full employment---think in terms of increasing the supply of effort from their existing labor force.'^ Much of this paper, therefore, is concerned with rehabilitating the notion of work effort into analytical economics. The manner in which individual and team work effort is modelled below derives from considering aspects of the operation of alternative methods of remunerating employees. This is a natural context in which to examine work effort since essentially it is to elicit a response from this dimension of the input of labor services for which incentive methods of remuneration are designed. In this way, attention is directed to particular characteristics associated with the exchange of the services of labor and the resources consumed in effecting such transactions. An important class of these resources consists of the factors spent by management in converting a body of dicparate and parhaps grudging individuals into an integrated and productive labor force. In such team work, each employee has the occasion of pursuing his own interests at the expense of all and this self-seeking will partly take the form of each worker consuming more leisure on the job. This shirking is detectable by management spending resources to monitor the behavior of employees so that the incidence of slovenly work behavior will be related to the properties of the function relating enforcement costs to the level of the firm's policing activities.—'' Evidence that important. coat3 are involved in monitoring workers is supplied by the fact that firms sometimes choose to hive off that activity and pay someone else to discharge that function. For instance, in many cases, labor contractors are paid not merely to recruit workers, but also to supervise them and to apportion rewards 21 among them.— Whether in the direct pay of the firm itself or the employees of an intermediary, the incomes of the supervisors are to be understood as the marginal return to the firm for their monitoring the 3/ work effort of the firm's labor force.— Tills policing activity Includes "metering input productivity and metering rewards."—^ If the price nuachanism is to allocate resources efficiently, the rewards and penalties to different activities must accurately reflect their marginal productivities and marginal costs, /.'his correspondence between market signals and demands and supplies is accomplished through, devices that monitor and register claims on and contribution —^For a more thorough statement of this view of the firm, Gee Alchiau and Demsetz [1972], 2/ — This is true of the contractors in the California harvest labor market some of whose functions have been taken over recently by officers of employee unions; of the Italian padrone at the turn of the century who not merely advanced credit to poorly-informed immigrants and occupied the role of the go-between for laborers and employers, but also sometimes acted as foreman and paymaster; nr.d of the jamadar in India who enlists and supervises workers for firms in the construction industry. For more information cn the work performed by labor contractors, consult Epstein and Monat [1973]. 3/ — The nature of these net returns is discussed in Becker and Stiglar [1974] hi — Alchian and Demsetz [1972], p. 778. to output. One class of these devices consists of the various arrangements by which individual employees are remunerated for their work. Essentially, there are two such arrangements: remuneration may be distributed according to the input o-_- inputs that each worker supplies tc his job or according to the performance or output produced by a worker or group of workers. The first type of arrangement commonly takes the form of time payments—wage per hour, salary per month or per year, etc-while the second encompasses pure piecework and a great variety of incentive payments schemes. Since many individuals classified as operatic on incentive methods have some fraction of their total pay related to their working time, the menu of choices available to workers does not simply consist of receiving either all their earnings in the form of tine payments or all a3 piecework. Instead, the market has generated a spectrum of wage payment methods in which the weight attached to working time in the metering system ranges from zero in the pure piece-rate system to unity for workers on time-rate methods. The distinction between tha use of time input, on the one hand, and output or performance, on the other hand, as a metering device is not always a clear one, however. For instance, a number of payment-by-results schemes measure performance by the difference between the time taken by a worker to complete a task and the "normal" time scheduled as gauged from work study methods. Here pay is related to work performance and the index of performance it3elf is linked to the input of time, namely, to the extent to which the time spent by the worker on an activity falls short of the rated tiir.2 "alliiwed." The desire to combine the advantages of each general class of payment mechanism has led to "mixed contracts" like measured daywerk under which a worker (or group of workers) is paid a basic time rate and receives a bonus (or set of bonusec) providing a certain standard is net. This is a contract in which part of the payment is contingent upon the satisfaction of some condition. Moreover, this classification into time-related and output-related metering devices involves grouping a wide assortment of payment schemes into two sweeping categories. Thus, in what follows, the terms "incentive: systems," "payment-by-resu'lts methods," and "piecework contracts" are used interchangeably and we do injustice to the rich variety of different types of incentive schemes that the market has generated. No doubt, there are occasions on which lack of discrimination among different types of time payments and incentive payments methods will hinder an understanding of a particular problem. Nevertheless, the main purpose of this paper is not to look for differences within tha two payments system, but rather to identify some general principles that distinguish between the two. Since the lines of demarcation between f-.ir.ie payments and incentive payments methods are blurred and since this categorization has sometimes appeared as no more distinctive as that snong different types of incentive schemes, the information available on the incidence of payment systems is sometimes difficult to interpret for changes over time or differences between countries. Consider the data on United States manufacturing in Table 1: the first column shows the percentage of factory workers on piecework as given in the Census of Population of 18S0 and the second column shows the percentage of production workers on incentive pay in 1958. — -•-Sea Lebergott [1972]. Tr.ble 1 Percentage of Production Workers on Piecework in 1890 and Incentive Pay in 1958 in United States Manufacturing Industry 1890 1958 All manufacturing industry 17.9 27.0 Tobacco 61*.1 30.9 Furniture 5U.2 25.0 Apparel 51.3 59.3 Leather UU.O 63.2 Paper boxes »+5.3 20.0 Printing lit.6 3.8 Textiles 13.U 39.7 Food 10.1+ 11.5 Chemicals 5.6 8.8 Lumbar h.l 6.3 Instruments 1*5.2 29.2 Toys, sporting goods 21+.5 21+.0 Nonelectrical machinery 21.2 25.9 Fabricated metals 18.6 23.1 Electrical machinery 15.6 1|0.3 Primary metals 10.1 1*6.1* Jewelry, silverware 10.0 35-0 Transport equipment 1+.6 10.1* Stone, clay, and glass 8.U 25.1 From U. S. Bureau of the Census, Census of Population i89o: Manufactures, Part 1, Table k L. Earl Lewis, Monthly Labor Review, May I960, pp. 1*60-63. Table 2 Percentage of Production Workers Paid on an Incentive Basis, 19lt5-56 and 1958 in United States Manufacturing Industry 19^5-^5 1958 1. men's + boy's dress shirts and nightwear lh 71 2. drugs + medicines 11 17 3. industrial chemicals 3 5.5 k. paints and varnishes 6 13 5. aircraft engines + engine parts 20 k 6. ferrous foundries 29 27 7. nonferrous foundries 20 23 8. iron and steel forgings 31* 33 9- cotton textiles 35 36 10. rayon + silk textiles 35 29 11. textile dyeing + finishing 22 Ik .12. woolen and worsted textiles 3^ 29 13. bakeries < 0.5 5 ik. cigarettes 6 3 15. cigars 73 66 16. footwear (except slippers + rubber products) 69 70 17. costume Jewelry 22 17 18. precious jewelry lh 35 19. structural clay products 26 31 all manufacturing industries studied 30 27 From Joseph M. Sherman, "Incentive Pay in American Industry, .19^5-^6," Monthly Labor Review, i'lov. 19^7, pp. 535-38. L. Earl Lewis, "Extent of Incentive Pay in Manufacturing," Monthly Labor Review, May I960, pp. i>60-63. Table 3 Percentage of Manual Workers on Payments-by-Resuits in British Industry for October 1938 and April 196l males aged 21 females aged 18 ?i over & over all workers 193b I96I 1938 1961 1938 lgol 1. minimg & quarrying 27 23 __ U3 27 23 (except coal) 2. treatment of nonmetals 15 30 30 37 15 29 3. bricks, pottery & glass 34 1*5 51 >+9 35 H5 It. chemicals, paints, & oils 5 21 2k 20 9 21 5. metal manufacture 50* 62 k2* kk 1+9* 59 6. shipbuilding & marine engineering Uo* 67 20* 2h 1+0* 65 T. vehicles 53* 52 56* 5k 52* 52 8. other metals & engineer- ing ui+* 1+5 53* 53 IfS* lt6 9. textiles 29 39 6h 61 lt5 51 10. clothing 36 3U Ul 50 35 hi 11. leather, fur, etc. 39 23 30 27 33 12. food, drink, & tobacco 3 lit 27 28 lit 20 13. woodworking 13 27 26 3fc 15 26 lit. paper, printing, & stationery 5 17 29 25 12 19 15. building & construction 0 15 — 7 1 11+ 16. transport & communica- tions (except rail) 3 7 3 2 2 7 17. public utilities 1 2 1 0 1 2 all industries surveyed 18 30 U6 it It 25 33 * = data relate to April 19i+7 rather tban to October 1938 From the Ministry of Labour Gazette for 191+7 and for 196l Table It I-ercent of Production Workers Paid on an Incentive Basis in American and British Manufacturing Industry U.S.A. (1958) Britain(ip6l) 1. Food and drink 11.2 18 2. Tobacco 32.0 U2 3. Chemicals and allied industries 8.8 21 It. Primary metals 46.5 59 5. Electrical, and nonelectrical machinery 28.9 hi 6. Shipbuilding and marine engineering 20,0 65 7. Vehicles 9.5 52 8. Other metal goods 23.1 41 9. Textiles 39.8 51 10. Leather and fur 36.8 33 11. Clothing 58.3 45 12. Footwear 70.Q 56 13. Bricks, pottery, glass + csinent 24.5 ItO lU. Timber and furniture 12.3 26 15. Paper, printing and publishing 7.7 19 16. Toys and sporting goods 2k.0 11 all manufacturing industry 27.0 >t2 Hotes to Table k The American data pertain tc those production and related workers eligible for incentive payments in May 1958, They appear in Earl Lewis, "Sxtent of Incentive Pay in :• lanufsxtrxing," Monthly Labor Review, Vol. 83, Mo. 5, May I960, pp. U6O-63. The British data are defined as wage-earners paid partly or wholly under payment-by-results systems in April 196l and are given in the Ministry of Labour Gazette, Vol. 69, Ho. 9, September 196l, pp. 369-73. From what can "be gathered from the commentary accompanying these data, both surveys cover essentially the same sort of payment systains, namely, piecework arrangements, bonus schemes, and systAa; in which remuneration is related to the output of a group of workers. laspaction of these numbers indicates contrasting experiences in different manufacturing industries, but overall the incidence of those employees working at paynont-by-resv.lt» methods appears to have increased by almost tea percentage points. Unfortunately, this inference may not be htl;1 with a great degree of confidence since the comparability of these two series is somewhat suspect. Firs., in the 1890 Census, the employer was requested to list operatives, officer;? cr firm members, clerks, unskilled workers, and finally pieceworkers "cot included in foregoing statement." This taxonomy may mean that soma pieceworkers are classified among the other groups cf workers and that the series in the first column of Table 1 is something of a residual category. Second, the 1390 data pertain to factory worker:, only; yet, in apparel and textile manufacturin; in particular, a number of employees worked at home and all of these, of course, were remunerated at a piece-rate basis. For both of these reasons, the incidence of piecework ic 1890 is probably under-stated by these figures. Greater confidence may t.j • '. jr.ej in the comparability cf the data in Table 2 which show the percentage of production workers on incentive pay in 1945-56 and in 1958: the use of incentive methods fell over this period in "aircraft engines and engine parts" and rose for "bakeries" and "precious jewelry," but there appears to be no general dominant trend Similarly, the criteria used in classifying the British data in Table 3 have been pretty Much the sane over cats period from 1936' to 1961. Bare, the incidence of incentive pay ay^c-.rs to have been approximately constant for adult female workers, but has increased for adult isale workers. The only comparable British study since 1961 was coivJ"Cted by the Nation;-.; Board for Prices and Incomes In December 1967 from which is concluded "...there is no clear evidence of any overall movement to, or away froa, payment-by-results since the Ministry of Labour's 1961 enquiry,"—^ The data given in Table 4 compare the incidence of incentive pay in manufacturing industry in the U.S. with that In Britaiii. Though these methods are more widely used in B/iuain than in the United States, typiccll; in those industries where incentive systems are common in one country the-y are also frequent in the other country. (The simple correlation coefficient between the two series in Table 4 is 0.52.) In fact, thevre is considerable casual evidence suggesting that the United States is among those countries that make least ur.e of paycent-by-results methods for remunerating manual workers in industry. For instance, in Che late 1940's, it was estimated that the percentage of total hours worked by manual workers at some form of piecework was 37 in West Germany, 40 in 2/ Denmark, 52 in Norway, 58 in Sweden, and 70 in Hungary.— As of Augu3t 1965, some 63 percent of production workers in the Soviet Union were paid by some form of piecework with almost one-quarter working at 3/ straight piecework methods.-- In the 1950's, up to 42 percent of China's industrial force were on piecework, but it is thought that the use of these methods subsequently declinrd and premium payments with bonuses paid for exceptional group or individual performance were substituted in their stead in the 1960's. —^National Board for Prices and Incomes, [Hay 1968]. 2/ —See International Labour Office, [1951], Ch. 3. 3/ —See Kirsch, [1972], Ch. 2. Piecework systems, (particularly of the "progressive" form in which the piece-rate increases with rising output) were, of course, extensively applied in the Soviet Union under Joseph Stj?iii, Such economies in which the state figures so prominently tend also to make considerable use of exhortative tochniques to stumulate work effort: the moat faxous of these was the Stakhanovist u.o\etnent in the 1930's Jn the U.S.S.R., named after a coal-face worker who was said to have yielded 1C2 tons of coal in one shift, fourteen timas above the norm! -'See Howe, [1973], pp. 118-27. -8- The Employer'3 Choice Among Wage Payment Systems At any given moment of time, each employee's contribution to the firm's output and sales is incompletely known to management. In addition, the employer is not fully cognizant of the labor services whi<:h will be forthcoming at different levels of remuneration that he may offer.. Put differently, since information is a scarce resource, management has the problem of disseminating rewards among and of posing incentives for its employees. A firm's response to these two interdependent problems takes the form of adopting a particular wage payment system. Consider th?. manner in which each of these two problems is treated by various wage payment methods. In assessing each employee's contribution to production and thus in allocating wage payments among workers, the method of time payments distributes remuneration according to some measure of working time while incentive pay systems involve constructing some index of work performance and relating pay to this index. The costs of precisely metering the output of an individual (or a group) compared with identifying his working hours (or days) may lead to the total absence of piecework from certain types of jobs. Presumably this is why piecework is less common among white-collar employees. The exceptions are provided either by those employees whose remuneration- is based in part upon the value (or increases in the value) of their sales (e.g., retail salesmen of durable goods, insurance, and security) or by those for whom "tips" constitute an important fraction of their income (e.g.^ taxi cab drivers, restaurant waiten, -9- hotel page boys). In the first case, something readily measurable is .iv&ilable; the second category of workers have their performance evaluated directly by the consumer. There are, in fact, few jobs where some index of performance could not be constructed and used in remunerating workers though it may not be desirable to do so.—' Teaching was once invoked as the standard example of a profession in which a payment-by-results method could not work profitably, but the debate over the operation of educational performance contracting under which a teacher's earnings are linked to increases in the measured scholastic aptitudes of their students has rendered the once classic example a little less convincing. Even where the output is measurable with little cost, the resources used up in policing an incentive payments system would be considerable if careful, time-consuming, checking of the output is requir-.i in order that performance does not fall below a quality constraint. This illustrates the point that under incentive schemes shirking takes the form of a tendency for the quality of the product to fall, while workers on time-rates shirk simply by putting as little effort into their work as goes undetected. Hence supervisory personnel are used to reduce shirking —^The payments scheme introduced for a short while by the British Natiorr.l Health Service to reward dentists presides a vivid illustration of the customer bearing (presumably) heavy costs. Dentists' pay wae related to the number of cavities filled with the predictable result that the incidence of tooth decay identified by the dentist took a suspiciously dramatic leap upwards and the speed with which cavities were filled increased painfully from eighteen to six minutes per filling! Subsequently, wore thorough devices were introduced to monitor the performance of dentists. -10- by workers on tiiae-rates while with payment-by-results systems more resources are devoted to inspect the quality of output, to ascertain in the case of contingent contracts like measured daywork whether the conditions have been satisfied under which a bonus is to be paid, and to prevent employees from maltreating physical capital owned by the firm. (Thus under incentive schemes the employer will be less inclined to "loan" capital equipment to the worker.) Hence the function of supervisovy personnel tends to differ under the two payment systems: in each case. they serve as agents communicating information between upper management and production workers; but with time payments their concern is directed towards monitoring the £§ce at which individuals work while with payment- by-results it is the care and assiduousness exhibited by the workers that is particularly subject to deterioration and thus which requires the supervisors' attention. Clearly the supply prices of these different monitoring functions will affect the choice of payment systems to adopt for rewarding production workers.— The second important class of problems í:o which the wage payment method is addressed arises from the fact that, at any given moment of time, knowledge of market clearing prices is not freely available to market participants. As a consequence of opportunities and preferences continually changing, information concerning the terms at which goods and services are being offered for sale is not costless to obtain. In the context of the labor market, employers are not certain what quantity —^In their examination of wage payment methods, the British National Bočiť1. for Prices and Incomes "...came across a strong minority opinion [*3;ong managements] that payaent-by-rasults systems were primarily concerned, not to raise productivity, but ť? deal with the problem of work supervisior National Board for Prices and Incomes, [Hay 1968] , p. 25. -11- of labor services will be forthcoming at different offered prices and workers do not have complete information of all wage offers. Though the employer's ignorance of the labor supply schedule facing him obtains whether it is a set of piece-rates or of time-rates that is being determined, different elements of the problem are associated with the two payment systems. Thus, payment-by-results has the advantage that, once piece-rates have been settled for a given class of workers, the more able and more industrious individuals will naturally receive higher earnings as the simple consequence of the system's operation. Put differently, incentive systems tend to circumvent some of the employer's problem of accurately screening his employees. So, given a cost to detecting each employee's work performance, the use of piece-rate methods is prompted by a wide dispersion of productivities among different workers. Under time-payments, such differential rewards for diligence and hard work are more difficult to determine since such characteristics of workers are Incompletely known both prior to and after being hired. So under incentive schemes, employers may permit the rage payment mer-.hanism itself to discriminate among workers of different productivity. On the other hand, there is a good deal of evidence to the effect that the determination of the piece-rate is particularly difficult (costly). Individuals schooled in the techniques of scientific management and industrial engineering try to determine that piece-rate which will yield the flow of output from workers that the employer is seeking and their procedures normally take the form of ascertaining what an "average" employee will produce over a ,;normal" period of working time. If the piece-rate is set i;toc" low, the required number of workers will not be forthcoming to replace t.hoce kuo -12- have quit unless the employer engages in costly methods of recruitment; if it is set "too" high, the fear of "rate-cutting" induces workers to form formal or informal cartels to prevent some individuals from working ss hard as they would like.—^ If the production technology and individuals' preferences and opportunities were constant over time, by trial and error the employer vzill grope towards the optimal piece-rate but such a stationary environment is not, of course, the one in which the employer operates. In particular, rapid technical change involves especially frequent piece-rate revisions with all the attendant costs so that we should expect incentive methods to affect and to be affected by the pace of technological change: where keen competition obliges firms to be innovators or swift initators in production technology, there will be a tendency for a firm to make use of time-rate methods that do not necessitate a revision of wage rate schedules each tine a production process is reorganized; alternatively, piece-rate methods throughout an industry add a cost to and thereby discourage the development of new 2/ technologies in that industry.— Where the firm employs workers of various skills, a change in the piace-rate necessarily involves questions of the relative remuneration of workers of different skill classes. These problcra-3 —^Such cartels go by the name of "restrictions of output," These are discussed below. 2/ — There is, indeed, evidence to this effect: "An incentive plan is also helped by stable technology. The manufacturing industries where the value of incentive plans is most frequently questioned appear to have had continuous and significant technological change. Those industries where incentive payment is taken for granted appear to have had relatively stable technology," Sumner II. Slichter, James J. Healy, and E. Robert Livernash, [1960], p. 519. -13- cf ttatermining the relative wage structure also obtain when workers are 1/ employed at time-rates— though the tendency for earnings differentials to display marked anomalies over time under an incentive system heart; witness to the problem not only of determining the piece-rata for a give--' grcup of workers, but also of establishing skill differentials that do not create feelings of inequity on injustice that affect work perforate: Ihe conpacuuiary aspects of alternative employments that derive from the fact that (in Marshall's words) "the. seller of labour must deliver himssi include, inter alia, the fairness aud justice with which each worker sees himself as being accorded by tha management. Arbitrary or, sometimes, inequitable treatment of euploy&es on tha part of maaageaeat is offen J-^B.ees [I970T haa put the general point in the following succinct me.rmer: "...every change in wages tf»rids to raise questions of internal equity tba can i>« very troublesome to solve. Even if it is decided to raise all •rage uniformly, it must be decided whether this is to be by n uniform percentage or by a uniform amoj-.t, and whether the increase Is to exi~e =ill the way up the structure of wageo and salaries or only part way. The problems are not unlike those faced ;y the United States Congress in separating problems of the level of the income tax from is°u«.s of tax reform,".. -Cf. 'Wa have in shoi t foviu? considerable disenchantment among workers with the constant time-consuming process of shop floor haggling, inequity falsification of work records, inversion of customary ekili differentials and lack of security, which are so often associated with conventional paymont-by-results systems...workers are as concerned with the equity and stability of earnings as they =»re with their absolute amount, and are therefore likely to be willing to 'tcade-off* those features which permi; a rapid and uncontrolled rise in earnings in exchange for alternative advantages," National Board for Prices and Incomes, [May 1968], p. 29. -14- followed by uncooperative and unaccommodating work behavior. Not only do incentive methods lead to disparities in earnings that sometimes in-iu,> discontent among employees, but the fewer supervisory personnel with *.iic' payment-by-results systems typically operate also implies that the costs of managing these wage payment usthods are particularly sensitive to the prevailing climate of management-* ".ployae relations.—^ The Supply of Working Hours and r:ork Effort Consider now the supply of labor services under different wage payment systems and, for this purpose, suppose the output or performance. of an individual employed over any given period of time (say, a week) cn an incentive payments scheme may be indexed by Z. For this present analysis, all variables that might affect the level of Z are excluded save two: the numbers of hours worked at piecework (H^) and the amount of effort expended at that work (E). The relationship between these variables describes an individual's on-the-job production function (assumed to be strictly concave): Z = Z(H ,E). P Consider first the association between working hours and output. Most of the research that has been undertaken on this relationship has been conducted not by observing tn individual's performance, but the work performance of a group of employees at a plant or factory. A convenient summary of this research is provided by Brown [1965] from whose paper the —^There is a large amount of evidence from the industrial relations literature of how incentive systems tend to generate substantial or anomalous interskill earnings differentials which contribute to considerable management-employee discord. These are described by the National Board for Prices and Incomes as "decayed" systems while Slienter, Healy, a>v3 Livernash talk of "demoralized" plans noting that "Serious deasoraliaatioa typically involves not only poor union-management relations and a high rate of grievances but continuous vse of wildcats and slowdowns," p. 493. -15- lable 3: A Summaiy of the Studies Invcstiyitigg the Association between Hours of Work and Output no. of case length of workweek* studies % of cases in which % of case3 in whir'-' hours and output are hours and ou positively asso- unrelated or negi- clated_ tively related_ more then 60 hours 238 53 47 52-60 hours 894 66 3i less than 52 hours 91 82 1G *Length of workweek is measured at original hours in situations T 0. There is a long history of highly respectable, research into work fatigue and, while some commentators seem to think that with the shorter working hours (per day cr per week) of today fatigue is a thing of the past, there remains an awareness of the fact that boredom aid the monotony of the job affect work performance. It is variations in the amount of effort expended by workers -16- that is invoked to account for the shape of the "work curve", as it is commonly called in the literature, the time path of workers' output during the day or week. This curve reveals interesting differences from case study to case study and clearly a careful analysis of this variety would require examination of a number of explanatory factors. However, if a typical curve can be said to exist, then it has the characteristics first of rising during the course of continuous work and then of falling off, sometimes quite steeply, towards the end of the working period.—^ This decline in the rate of output that appears to set in after a period of continuous labor is generally attributed to the growing disinterest and fatigue of workers. It was relationships such as these that led earlier writers to reserve a particular role for effort in the supply of labor services .— — An excellent discussion of "work curves" is contained in P. Sargant Florence [1924] who writes, "This curve cannot be attributed to any physiological fluctuation that would occur even if no work were done, since such fluctuations as have been measured seem to be diurnal in span, i.e.,, from a curve spread over the whole day and not over particular hours corresponding with the spell of work in the factory. Further, this curve can be spoken of as typical; it is not a mere compromise between two or more divergent curves but tends to be the curve most frequently shown. Deviations therefore shown by particular workers and particular days seem distributed around it as might be expected from chance errors. Nor, as far as my evidence goes, do the less productive individuals have the more precipitous curves", p. 234. 21 — And the treatment at that time of this dimension of labor services corresponds to some suggestions that the deceleration of output per man from the 1890's in certain trades in British industry is attributable "...to a certain slackening in the men's individual labour effort," E. J. Hobsbawm, [1964], p. 351. Or, to illustrate with another instance earlier in that century which also draws attention to the adjustment of work effort to rewards, consider this description of English navvies (railway constructio-a laborers) working in France in 1843: "The navvies worked this way for six weeks and earned fifteen francs a day each. Then the French engineers began to see that the Englishmen were making them pay an enormous sum for the work and reduced the price. The navvies, having no other work to go to, had to take it. When the next pay day came round there was another reduction, this time to five francs a day. Then the. EngliohmsnU mutter fid that', they would do net more than five francs worth of work, and they thereupon slackened off ixid took things easy," Coleman, [1968], pp. 203-209. -17- The employee's production function will, of course, contain other arguments. For instance, the vital role of work experience in accounting for variations in labor incomes across individuals suggests a "learning-by-doing" effect in the function 2. Such learning effects have been -e.corded not only for individual work performance, but also for group performance,—'' There are also characteristics of the individual's working environment that will affect his output—the machinery with which he works, the organization of the work processes, social interactions among e^ployo^.r, !he climate of management-employee relations, and so on. However, for our present purposes, the presence of two variables, working hours and work effort, in the individual's production function is sufficiently general to highlight the substitution that may take place in the supply of labor services for individuals working en payment-by-results schemes. Consider now the implications for the standard utility analysis of the supply of labor services when the individual's work-leisure choice includes an option to be remunerated by incentive methods at work rather than by time-rate methods.— ilis (wcil-behave-i) utility function (U) take:; the form: U = U(Ht,H ,E,Q> W where H measures hours of work in the time-rate sector, K hours of work t P in the piece-rate sector, E the amount of effort spent at piece-work, and Q the consumption of commodities. Increases in Q are assumed to yield greater utility in the individual (i.e., 3U/3Q > 0) while small increases •■-'see, for instance, the study by Blackburn, [1936]. ?/ — For expositional convenience, the individual is taken to be the decision-t aker. The extension to the context >r the household is straightforward. -18- ln working hours and -work effort ai'e presumed to be associated with more disutility (i.e., W/dK , W/dK^, 3U/3E < 0). The assumption of ;iU/3E < 0 may well not be the case over certain ranges of effort: when the individual is bored, greater -work effort may reduce his feeling of ennui and yield him greater satisfaction. Under such circumstances, greater effort will be expended at work until the marginal utility of •rork effort turns negative. The budget constraint may be written as follows: I + wH. + irZ(H ,E) = pQ (2) t p ■vhere I measures nonlabor inccsie, p the price of consumer goods , w the wage per hour worked in the time-rate sector, and 7r the piece-rate for work performed in the payrrent-by-results sector. In the spirit in which this analysis is normally conducted, all prices (including the piece-rate) are taken as parametric to the individual. The total hours (K°) available to an individual over a given period of time (say, a week) nre assumed to be divided entirely between household activities or "leisure" (HD), work paid at time-rates (H ), and work in the piece-rate sector (Hp). The individual's behavior may now be described by the maximization of his utility, equation (l), subject to the budget constraint given in equation (2), to the form of his on-the-job production function, and to the implicit constraint on hours (H° = H. + H, + H ). X. u p The form in which effort is modelled here is only one of a number of possible characterizations and it describes a particular dimension of the expenditure of effort. Here, effort is spent at piece-work, but not <*t work in the time-rate sector nor in the process of consuming "leisure" off the Job, Of course, effort is spent at time work hut variations in this expenditure of effort are not immediately and automatically reflected in variations in a worker's earnings. Instead, an employee on time paynen who consistently and permanently works diligently and whose effort is detected by supervisors will tend to be promoted to a highsr wage rate class more rapidly than other worker?. Under time rates, therefore, h*bit vark effort is rewarded through promotion and the appropriate incentives involve steeper earnings-experience profiles than will be the case for piece-rate workers. Work effort that is transitory and volatile tends not to be rewarded in time rate schemes but is compensated with piece-rate methods through the mechanical workings of the payment system. Formally., the level of habitual work effort has the properties of an investment decision: an expenditure of a resource (effort) is made prior to the payment of the rewards. It is the expenditure of transitory work effort that is described by the relationships in equation (2). As far as the use of effort at nonmarket activities is concerned, en alternative characterisation might recognize that an energy or effort constraint obliges the individual to allocate his energies between that expended on the job and that used in household or leisure activities and the appropriate formulation would involve not merely the consumption, but also the production of energy during non-labor market time. In this way, the model links up with the economics of health. This would lead us away from the main purpose of this essay which is to examine labor supply behavior under different forms of the wage payment contract, but it seems -■20- a natural and perhaps useful development.—^ The first-order conditions for a maximum involve the constraint shown as equation (2) plus t p p ••rhere ,\ is a Lagrange multiplier to be interpreted as the marginal utility of consumption (that is, Xp = 3U/3Q), If the individual works entirely in the time-rate sector, then = E = 0, > 0, and there is the familiar aU/SH^. - -Aw condition; alternatively, for the worker specializing in the piece-rate rector, H+ = 0, > 0 and E > 0 and the marginal rate of substitution of work effort for working hours in hi utility function equals the marginal rate of substitution of effort fox-hours in his on-the-job production function: 3U/3E _ 3Z/3E_ . (3) 3U/2H ~ 3Z/3H P P The absence of the piece-rate from this condition arises i'rom ine postulated foiin of the incentive payment system under which neither working hours nor work effort are rewarded directly, but rather only as they con tribute to vork performance (Z). "^The budget constraint also abstracts from factors like the existence of a premium rate for overtime and from the fact that very many workers Lhe piece-rate sector do not receive all their remuneration on a payment by-results basis, but are also paid in part on a time-rate basis. Those modifications in the wage payment system may be grafted on to the basic riodel with little difficulty. -21- Totally differentiating the first-order conditions and then solving for the individual's decision variables yields equations for the supply of working hours (H. and H ) and effort (E). As usual, the effects of a "C p change in the wage rate upon hours worked and of a change in the piece-rate upon hours and upon effort expended at work may be decomposed into income effects and utility-constant substitution effects. In the case of hours worked in the time-rate sector, the substitution effect OH./cw; is ambiguously positive, the standard result,, (The asterisk here and below denotes the income compensation that enables the individual to remain on the same indifference surface.) However, the unusual Implicatic of this model is that, within the piece-rate sector, the sign of each of the substitution effects is ambiguous. If changes in work effort did not affect the individual's production (i.e., if 8Z/8E = 0), then, in the case of the H equation, the substitution effect would be unenibiguously ST positive, as is the standard result. Correspondingly, in the E equation if the individual were not free to vary his hours of work (i.e., if 3Z/8Hp = 0), alterations in work effort would be positively associated wit. income-compensated changes in the piece-rate. But once a substitution between hours of work and work effort is permitted in the piece-rate sector, we can no longer be certain of the signs of these utllity-constan substitution effects. The economic rationale for these results is straightforward: at piece-work, effort and hours are rewarded only inso/_ as they contribute to an individual's output (Z); an income-compensated increase in the piece-rate affects simultaneously the return from hours of work and from work effort and so may encourage the substitution of effort for hours of work depending upon ths nature cf the production fume ind of the utility function, both functions containing hours of work of and work effort as arguments. A necessary condition for these abnormal substitution effects is that hours at piecework and work effort are strong substitutes in the individual's utility function: that is, if an individual elects to work more hours, then he iH.ll tend to reduce his fatigue (effort) rrom working. Although the signs of the substitution effects (SH^/Stt)* and (3S/3tt)* are indeterminate, we can be sure that the income-compensated increase in ir will induce an increase in net labor input and, therefore, given the marginal products of these inputs, an increase in output Z. The reason for this is that, since a small change in ir leaves the relative return on work effort and working hours unchanged within the piece-rate sector, we may invoke Hicks' composite commodity theorem, treat effort and hours "as a single commodity", and derive the result that the income-compensated effect of an increase in tt on the individual's output or performance (Z) is unambiguously positive. In short, in payment-by-results work, the familiar conditions on substitution effects apply to changes in a worker's output (Z), not to changes in his working hours, as a function of changes in the piece-rate (if). Careful empirical research needs to be undertaken into the structure of the individual worker's on-the-Jcb production function since, inter alia, this structure will affect his labor supply behavior. The research that involves estimating cross-section wage equations may well be useful in this regard though their interpretation needs to proceed with caution. For instance, annual or weekly earnings equations for workers on incentive payments systems wi31 only shed light on the on-the-job production fuucti.-wi -23- if the determinants of the piece-rate (it) can be separately identified and held constant, otherwise the right-hand variables in such equations will tend to compound the factors that affect ir with the arguments of the production function. This problem also exists for equations in which tho dependent variable is defined as average hourly earnings (or, more frequently, it natural logarithm). Concerning the inferences that may be drawn on the individual's on-the-job production function, this procedure of dividing earnings by hours worked is, of course, unexceptional where, under payment-by-results schemes, the elasticity (in the case of the logarithmic specification) of an individual's output with respect to hours worked is unity (i.e., Sin Z/31n H =1). But presumably, this should not always be treated as a maintained hypothesis. Many incentive payments schemes reward various dimensions of a worker's output at different rates. The door-to-door salesman with a suitcase of products selling at various prices is an obvious illustration. There are other cases: slaughtermen in Britain are paid different rates according to the type and the size of animal they slaughter;—'' a chemical, firm investigated by the National Board for Prices and Incomes [Dec. 1969] awarded points for workshop cleanliness and linked pay to the number of points collected while a synthetic fibres plant related one clement of earnings to the quality of output by reducing pay as the number cl defective items produced increased; ^ and, in educational performance contracting, the contractor's (including the teacher's) payment often depends upon increases in pupil's recorded mathematical abilities, gains —^Consult the interesting study by Greenwood, [1969]. in 'i.-.l.\;.ir verbaL test pupils have acquired (like a higher attendance rate).— In each of these illustrations, several dimensions of output are distinguished and each is rewarded at a different rate. The individual then has to allocate his time not simply between piece-work, work on time rates, and leisure, but also among the various types of piece-work activities, each activity corresponding to a different dimension of output. If i indexes the different work activities and leisure, then the sup/ply equations for H„. and Ei to each activity will contain as arguments the incentive rates, r in all the activities and we arrive at a general system of economic incentives that determines the allocation of an individual's time and effort among many activities. Once it is granted that an employer values a number of different attributes of an employee's work performance, this case of a multidimensional output may be regarded as the typical one. And if for analytical convenience the situation is characterized as one in which the employee applies his time and his effort to merely two work activities (say, Za and Z^), the two-sector model of general equilibrium that was originally devised to handle problems in international trade suggests 2/ itself as an appropriate representation.— Suppose that a worker's supply of hours and of effort is made in a two-stage process such that, at the first stage, the division of hours and effort between market work and nonmarket activity is determined and, at the second stage, this given quantum of working hours and effort is then allocated among the two different work activities. This separability allows us to treat the endowment — See the summary of eight performance contracting programs in Hall [1972]. •'- Alternatively, if effort were expended in nonmarket as well as in market ?;fltiviti.?s, again a tvo-sector (market versus nonmarket) model of the allocation of tins oaA effcnt is involved. 711 of working hours and of work effort as fixed for a given individual although, of course, these endowments will normally vary between employe.'.;-according to variations in individual preferences. These input endowments are spscific to an individual and cannot be transferred between their use as inputs in producing Z^ and (providing specialization in either Zg or Z^ is ruled out). Consider two employees with the same on-the-job production functions (linear and homogeneous) yet differed relative endowments of working hours and working effort to apply in the production of both Z and Z. . Though the per unit returns to Z and a b a Z. are likely to differ (i.e., tt f- n ) both individuals face the same piece-rate in a given activit;^. One final and crucial condition is that, for all relative input shadow prices, Z and Z, have different and invariarri input intensities, say, Z being hours-intensive and Z, effort-intensive a b (i.e., Ha/Es > H^/E^). Under these postulates, the following "factor price; equalization'/ theorem may be asserted: the two individuals have the identical shadow price of time at work in the two activities and have the same shadow price of work effort. This does not mean, of course, that they receive the same earnings (or hourly earnings) since, although their per unit valuation of working hours and effort is the same, they bring to work different endowments of these inputs.—'' The usefulness of this result-is that, if two individuals are observed working the same number of hours, then in certain circumstances differences in their earnings may be attribu;;i to differences in the amount of effort they apply to their work and not to differences in their respective valuations of effort and time that would —^On the assumptions stated, an individual's labor earnings (Y) over a period of time is equal to the sum of his earnings in producing Zp and Z^ (i.e., Y — tt Z + i\Z.) and this in turn is divided into that attributable to hours andatfiat to effort [i.e., Y = u(H^ + H^) + v(Ep * E^) where u and v are the shadow prices of hours and si'fort reopsc tjyolyj. -26- derive from dissimilar input marginal productivities (production functions}, This sort of model can also be employed to yield implications concerry'.isfT the effects of technological change. It is frequently asserted that with modern factory processes the employee becomes little more than a machine tender and his output is unresponsive to changes in factors under his particular control. Put differently, it is argued that changes in the technical organization of production have tended to raise the marginal product of hours relative to the marginal product of effort, an effort-saving technological change. Suppose a quantity and a quality dimension to the output of each individual are identified, the index of quantity being given by Z and the measure of the quality of work performance by £L ' Also, as before, suppose that Z^ is relatively intensive in effort snd Zg in hours (i.e., Ha/Ea > H^/E^): in other words, the quality of work performance requires comparatively greater effort. Under these conditions, an effort-saving (hours-using) technical change in the performance of the effort-intensive output, Z^, leads to hours being substituted for effort in both Z and Z. and, of course, the shadow a o price of effort rises and that of hours falls. The output of the effort-intensive commodity, Z^, will rise and that of Z& will fall: the quality of work performance will increase. Conversely, an effort-saving technological change in the production of Z , the hours-intensive quantity of output, induces a substitution of effort for hours in Z^ though not necessarily in the quantity of output, Z . In this case, the effects EL on the outputs Z& and Z^ are indeterminate. —^In the example of the synthetic fibres plant above, Z^ was measured by the (inverse of the) number of defective items produced. -27- Restriction of Output One of the most consistent and well-researched findings of industrisl relations specialists examining the operation of incentive payments schemes is that production workers in a number of plants operate under another constraint that, in this literature, goes by the name of "restrictions of output" The role of unions in operating "make work" or "featherbedding" arrangements is familiar to economists, but the extent of such practices among nonunion employees is not typically appreciated. Starting with Mathewson's classic work in 1931 on the Restriction of Output among Unorganized Workers, researchers have discovered that social pressure and custom may operate as effectively as formal union rules to 1it 2/ the production of employees.—' Though such restrictions of output exist where pay is related solely to the input of time, they seem particularly prevalent under incentive payment systems where a fear is that a great expenditure of effort leading to a bulging pay packet will induce manajemet't — Synonyms include rattening, soldiering, striking on the job, goldbricki::g.s government work, scamping, skulking, and Ca'Canny. ?' — Thus, as Leiserson observed in a chapter of Mathewson's book: "...nonunionists are far from helpless in limiting the amount of work they will turn out. They have developed methods of collective action as effective as trade-union practices...Trade-union restrict?oi is ordinarily enforced by agreements with employeia -which usually come to the public notice. In fear of the condemnation of public opinion, responsible leaders of labor organizations use their influence to reduce the restrictive practices of their members. Non-union workers have no such restraining influence, except as the management is on the watch; but this is offset by an amazing amount of encouragement of restriction by the workers' immediate foremen and supervisors,1' Mathewson, [1931], pp. 162-63. And Leiserson footnotes .T.H. Willits who is quoted as writing, "In short I don't think you can do anything but say that restriction occurs also abundantly among non-unionists and that the causesP therefore, must be considered as deeper than the trade-union organization." This behavior is sometimes manifested not in a restriction in output of a given quality, but a fall in the quality of a given output so that the inspection procedure results in rejecting a higher fraction of output as faulty or defective. -28- to respond by cutting the piece-rate. The ■workers' cartel may thus be interpreted as a device for redistributing income among vorksrs in one period as an insurance against rate-cutting in the future. There are, naturally, other explanations—to stabilize earnings over vagaries in output, to reduce interpersonal frictions by narrowing the dispersion in individual earnings, to assert an independence from management-initiated administrative schemes, end so on—thoa£.h the possibility of rate-cui-tiug appears to be the most frequently cited reason for the existence of these restrictions on output. Moreover, there are aany historical precedents fcr rate-cutting so the employees' fears are by no means groundless. Once the cartel has been formed it may veil be directed not merely at narrowing the dispersion of earnings at each point in time, bat also at raising the average. This may be achieved either by affecting the settir.^ of the piece-rate or by altering the indices (the Z's) to which the piece-rate is linked. In these laxter activities, -whatever else the employees' cartel may be doing, it is both supplying information to management about work processes and euployee work performance and also auditing management so that management's commitments in the labor contract are being fulfilled. Insofar as the restriction of output is concerned, the fraction of workers (unionised and nonunionised) in American and in British industry for whom such a constraint on their earnings is operative is not known, but the industrial relations literature referred to suggests thct it la by no means peculiar. Presumably, the existence of this constraint (if known) will be an element in the individual's Job choice and workers who -29- do not wish to he bound by such a constraint will tend to congregate in ou»?2oyment.n where it is not operative. Over a longer period of time, therefore, this constraint loses nuch of its force. On the other hand, in the short run, such customs and traditions should be incorporated into our economic models, as are the constraints of the legal system, if they operate meaningfully to affect behavior. In fact, the effect of these restrictions of output on hours worked and effort spent is fully tractable by conventional economic analysis. For suppose that the piece-rats (ir) is given to the individual so that a restriction on the output (Z) he may produce is equivalent to a constraint cn his piecework earnings. Let (tt£)^ be the earnings limitation imposed by social pressure or union rules so that for any worker employed wholly on an incentive basis, in addition to the standard budget constraint, he is subject to the constrain-1, thatirZ £ (uZ)°; this is his earnings allotment in the cartel.—'' The Lcgrs-psdJi thus takes the following form: a = U(H, ,H ,E,Q) + X[I + wH + ifZ(H .E) - pQ] + y[(TrZ)° - irZ(H ,E)J t> p op p -:- This characterization of the problem is adopted because it is simple and permits us (as shown below) to draw upon the literature on the behavior of the consumer under rationing. Nonevboless, it is a little misleading since breaking the constraint is a feasible option. Akin to the effect of a law on behavior, the individual worker's decision whether or not to act in accordance with the community-imposed limitation on his output will be affected by the penalties that would follow from his noncompliance. These normally take the form of verbal antagonism from his fellow employees, the imposition of fines, social ostracism, and "hinging" (striking on the upper erwl—see Roethlisberger and Dickson, [1939], PP. ^21-23). vhere p < 0 is a multiplier to be interpreted as the increase in the individual'3 utility (other things equal) from small decreases in working hours and/or work effort that accompany changes in the cciKriunity-imposed piece-work earnings constraint. If, in fact, the individual elects to earn less than the constraint (irZ)°, then u = 0 ar.d the analysis in the previous section of this -paper-" rquires no amendment. If, on the other hand, the constraint is binding (u < 0), the analysis becomes formally identical to multiple constraint problems like the consumer's choice unde.-rationing.—'' There will be some reallocation of labor as those individuals formerly near the margin of indifference "between working in the piece--rat-? sector and working in the time-rate sector now evade the piece-T;orfc earnings constraint by working at tin? payments. For those individuals remaining at piece-work, their hours of work (or v,-ork effort) is cleter^i-i.cc at that point at which its marginal disutility eiv.als a weighted average of the piece-rate times the marginal product of working hours (or work effort), the weights being the Lagrange an multipliers: ?U/3H^ = w(3Z/aH^) (u - X); and 3U/3E = ir(DZ/3K)(ii -• X). Dividing the one equation ly tha — In fact, treating the constraint as an equality rather than an inequality also seems descriptively accurate in those cases in which social pressure i levied on those whose output (or earnings) falls substantially short of tl-e prescribed level. Thus consider the following description of one workshop' customs: "To work too quickly was to be labelled a 'teararse' and to be at least partly shut out from the friendly give and take of the shop and from the spontaneously formed 'scrounging groups' which, in defiance of management rulings, assembled in secluded corners of the shop for unofficial tea breaks and discussions...If to be a "teararse1 was to court exclusion from the social life of the workshop, it was also regarded as a breach of workuh custom to be too much of a 'scrounger'. The man who persistently ir(3Z/3Hp; + 3Z/3E) and in the piece-rate sector when the inequality is reversed.—'' Where tt and w are given to each individual, a postively-inclined relative market supply curve would be generated if physical and intellectual aptitudes cause 3Z/3Hp and 3Z/3E to vary across individuals. Clearly, the greater the similarities in these marginal products among individuals in a given market, the greater the elasticity of the relative supply curve. Differences in the position of the relative supply function across, say, occupations will be related to differences in the frequency distribution of the marginal productivities over individuals in each occupation: where the distribution of marginal products in one market lies to the right of that in another market, so the relative supply curve in the former market will lie to the right of the relative supply curve in the latter market. — Note that the inequality may switch for a giver individual as his marginal productivities alter with increasing inputs of time and effort. An absence of total specialization of work on the two wage payment systems, may be manifested in "moonlighting" or in individuals seeking out jobs in which remuneration is partly on a time-rate and partly on a piece-rate basis. -3h- This section uses this framework for an empirical analysis of the relative pay and employment ox* the two classes of workers. Of course, since returns to workers in the piece-rate sector are measured in terms of dollars per unit of output (it) and in the time-rate sector in terms of dollars per unit of time (w), there is little purpose in comparing ir with w alone. The pay differentials analyzed in this section are the hourly earnings of these workers. Thus, if denotes the weekly earnings of timeworkers and since the data are adjusted to omit overtime work, the hourly earnings of these workers measure Vj./Hj. = w. On the other hand, for pieceworkers producing a t.'eekD.y output indexed by Z and ignoring the minimum time payments guaranteed, their hourly earnings are given by Vp/Hp = (ir?.)/H ■ By way of illustration, when Z is Cobb-Douglas in hcuxs and effort, the logarithm of the hourly earnings of these workers is given by y Intrr2-) = constant + In ir ■>■■ (a-l)ln H + bin E n p P where a and b are the elasticities of output with respect to hours worked and effort respectively. All workers from the Chicago labor market study discussed below were observed at the same number of working nours so H P is a constant. Of course, E is unobserved though will be higher in the payment-by-resuits sector since tx-ansitory changes in work effort are unrewarded at time-payments. It is well known that the average hourly earnings of incentive payments workers typically exceed those for timeworkers within particu!! a.r skill-industry categories. To examine whether the relatively higher earnings of employees on incentive psy may be attributed to identifiable personal or plant characteristics requires a body of data that contain information on variables in addition to the earnings of the tvo types of workers. Such information was collected for one semi-skilled occupation, namely, punch press operators, by Albert Rees and George Shultz [1970] as part of their Chicago labor market stu&vM Of the 183 male operators in 12 establishments, 8k work on time-rates and 99 on incentive methods; only one of these 12 establishments employed both male time- and piece-workers together. Of the 120 female operators in 8 establishments, 51 worked on time-rates and 69 on incentive pay; no establishment employed the two types of female workers together. A summary of characteristics of these workers is given in Table 6. As far as the male employees are concerned, there is the typical situation of the (unadjusted) hourly earnings of workers on incentive pay being higher on average and having a greater dispersion across individuals than time-wor The frequency distribution of hourly earnings approximates a lcgnormal for each class of workers though the dispersion is greater for those on piece-rates : both the highest paid and the lowest paid workers were on incentive schemes. Of the other characteristics of these male employees, workers i^I am most indebted to Mary Hamilton for going to considerable trouble to provide me with these data. "A punch press operator sets up and operate a machine that forms or shapes metal or plastic parts by punching them out of a sheet or strip of stock. The work pace is usually under the control of the operator. Punch presses differ greatly in size and in the heaviness of the stock used," [1970], p. 57. This last characteristic, the variation in the type of machine, means one cannot assume the piece-rate is the same even among pieceworkers in a single establishment. Thus, with these data, the factors that affect each pieceworker's output are not distinguishable from those that determine his piece-rate. on incentive methods tend to have longer job tenure, are more likely to be black or have a Spanish surname and to bo covered by union-negotiated collective bargaining agreements, and are working in plants characterised by high turnover rates for all workers in the plant.—'' These characteristic of time and incentive workers also hold true for the female employees srsvs for the relatively lower earnings and smaller dispersion of pieceworkers. This aberration from the normal situation of relatively higher earnings for piece-workers raises the question of whether this sample of female workers is, indeed, representative of the population of female employees on the two wage payment systems to whi r.h an assured answer cannot be given with our present state of knowledge. The use of piece-rate methods as a self-screeing device where a wide dispersion of productivities is thought to obtain will account for the higher incidence of black and female employees at work on payment-by-results. Are the hourly earnings of workers on incentive pay significantly higher then those on time payments after adjusting for other identifiable differences between the two groups of employees? To answer this question requires, of course, the use of some maintained hypothesis concerning the causes of earnings differentials. For this purpose, we turned to the recent —^The turnover variable (Xg) measures the percent of all workers in the establishment who have been hired in the previous three years and the private agency variable (X„) measures the percent of all workers hired through advertisements or private agencier,. Of the variables in this table, all measure characteristics of the individual employees except for these variables X^ and X„. The latter two variables measure characteristics of the establishments in which each employee works' '• and each individual is given the values of Xg and X_ that correspond to the plant in which he or she works. Table 5 Characteristics of Time-Rate and Piece-Rate Workers: Punch Press Operators in Chicago male workers female worker;: t ime-ratc piece-rate time-rate plC number of individuals üh 99 51 Y, hourly earnings ($s): mean standard deviation 2.35 0.26 2.55 0.50 2.h3 0.30 2.2'-} 0.?.:. X^, years of schooling 9.71 9.16 9.22 9 Ti- Xg, years on present job 6.08 10. Ik 9.17 ll. 33 X^ = 1 if occupational experience prior to present job 0.U9 0.1(9 0.27 0.% X^ ~ 1 if black or Spanish surname .08 .25 .16 .73 Xj- 1 if occupation unionized .06 .53 .0; Xg, percent turnover in estab- lishment 7.72 27.50 7.27 16.13 X„, percent hired through private agency or advertisement. 25.35 11-90 52.67 ?2 >■: literature on human capital and labor earnings which has provided evidence that variations in investments in schooling and in on-'cne-jcb training remove a large fraction of earnings differentials across irdividual workers. The particular regression equation which was fitted to the observations on these male rccriers takes the following form: la Y = constant + c.X., + o.J2 ■'■ s„X., + a, X^ + acX.,X„ + nrX„ + 11 2 l .-2 h 2 5 12 6 j a^X + a8X2X3 + Q-^n + e2X5 + B3X6 ':" ^kXJ + (L > ■yP + SGP + e, where In Y measures the (raUir?!) logarithm of the hourly, earnings of each worker before any overtime, P is a dicb.otomous variable taxing the value of urity if the worker was p-id c:i incentive methods said of zero on time-ratec, G is a dichotomous v.rr:r-'ble to be explained below, e stands for a stochastic disturb-.-Jic2 term, and where the other variables ore indicated in 'Cable 6.—^ T'.i'.' variables , Xg, X^ are interpreted as measuring human capital investments end have be:-i entered into the equation in a iao«t general form, namely, as if the function relating these variables to the services of the human capital stock may be approximated by a second-order Taylor series expansion. This specification 21 has been used before in accounting for earnings differentials— and for this sample it involved a substantial improvement over the narrower —^All these individuals, in effect, are working the same number of weekly hours. More precise definitions of these variables and information on the sample design and siirvey are contained in Rees and Shultz, [1970]. 21 — See Mincer, [197U] i form that involves a quadratic term in expedience hut net in schooling *>nd no interaction terms bstweea schooling eaid pc&tach-.oling investment -he turnover and use of private tgenvies variables (Xg and X^.) respectively) are included to capture the effects of management's policies tu vrtgco whenever labor supply curves to :Iio firm are less than co-npletely elastic. Thus each aansgement may economize oa the costs cf labor turnover and cf searching for productive employees by payiaG relatively high (conspicuous) wages so that our hypotheses a:..e that both 3^ ssi 3^ 1/ possess nege'.iive signs,— As noted above, only one of the twelve establishment- employing the maJ.3 punch press operators (end none of the eight establishments employing the female workers) operated both time-rate and piece-rate methods. This non-specialized establishment supplies ten observations to the sample of 183 male workers of whom five are paid at time rates and 5 at piece-rates. The estimating equation has been specified to discover the particular relative hourly earnings differential between piece and time workers in this singular estublitifcaent. Thus., C- is a dichotomous variable taking the value of unity if a worker is employed in this esteblishment that operates both time-rate and piece-rate methods and taking the value of zero for workers in all other establishments. An estimate of the proportionate earnings differential in this establishment is given by the estimates of y plus <5. —^On the employer's wage-turnover strategy, see Pencavel [1972] and on the employer's wage-search strategy, see Stigler, [1962]. Before turning to the results frosa estimating this equation with ovv sample of 183 male punch press operators, recall that previous stu-He i.:ito the earnings of individuals vith different employers have made use of sample sizes many times larger thra what is available to us here. For instance, Jacob Mincer's earnings functions were fitted to a sample of over 28,000 individuals and George Johnson and Frank Stafford's study of the earnings of academic economists msJie use of over 3,000 observations -: Our sample of 183 male employees in 12 different establishments looks pretty meagre by comparison and we should not expect the same sort of precision in our estimates of the regression coefficients as has been obtained by other researchers using equations similar to that above. •Xhe consequences of estimating the above equation by the method 2/ of ordinary least-squares— are shown in the column headed equation (l) in Table 7. Almost three-quarters of the variance in the logarithm of hourly earnings is removed by this linear combination of variables and our sign hypotheses on the regression coefficients are borne out by tne estimates. The absence of a strong positive effect of unionism (X^) on wages may be attributed to the threat effect in an area with a strong union movement: that is, in setting the wage in his own plant, an -See Mincer, [1972] and Johnson and Stafford [197U], 2/ — The appropriateness of this estimating procedure can, of course, be challenged. For example, an argument involving the simultaneous determination of wages and one or two of the right-hand variables is not difficult to construct. Treating the presence of piecework systems (the variable P) as simultaneously determined with In Y and applying the method of instrumental variables yields an estimate y that is practically identical to that given in Table 7. Estimate* Regression Equations with ■cho Logarithm of Hourly Earnings for Male Workers as the Dependent Variable oq-i.aticn (i) equation (2) 2quj^icn_J rte an of dependent variable .888 .850 .919 rcandard deviation of dependent ariable .159 .107 .ICS . ciirib ar of observations 183 81+ 99 "srtinates oi\.. constant .85'+ 1,005 U0P3) (.099) (.183: , schooling . 0h22 .0155 (.0169) (.0166) (.0258) (schooling)'" ~. 00?1 •-,0009 -,00f:'+ (.0009) (.0007) ( . 001 ■;■) experience .0099 .0138 .0077 2 (.0052) (.0069) (.0075) (experience) -.000001 -.00013 .00003 (schooling x experience) (.00009) (.00031) (.00013) -.00028 -.ooo-'a -.oooov (.0003)) (.0001+1+) (.00050) previous experience .0379 .0711 . oi+60 (=0557) (.0516) (.0931) (schooling x previous -.0089 -.00)18 -.0088 experience) (.0052) (.001+5) (.0009) XX (experience x previous .0035 -.0023 . 0033 experience) (.0013) (.0026) (.0025) Bls.ck or Spaní sh surname -.010'+ .00002 (.0151+0) .00001 (.0265Ó; unionism -.0'. '40 -■.oi+87 • •.. ?P2? J (.0288) (.01+96) ( , 121+0) V6 turnover -.0031 (.0009) -.0008 (.0037) -.0087 (.0031+) JT private agency or ad. -.0030 -.0063 .0019 incentive pay (.0006) .0681+ (.0399) (.0011) (.0021) . -i (incentive pay x non- -...061+2 -.0967 specialized, establishment) (.Qhlh) (.05^3) R2 .71+3 . 910 .735 SEE .08U .035 .10>i ■-■tandard errors are given in parentheses beneath estimated coefficients. 2E is the standard error of estimate of the regression equation.) er:;plc;..;r of nonunion labor will he induced by the threat of ^'.ion-organizing activity to prevent the appearance of a large gap between th-i wage rates he pays and those received by comparable union labor; such a gap normally implies that his workers would be extremely receptive to a union membership drive.—^ A test of the null hypothesis that the variables measuring the human capital investments of these workers have no effect on their earnings (i.e., a joint test on the a coefficients) is A 2/ decisively rejected: F = 8.1U > I?0 Q^(8,l68) = 1.99. The estimated coefficient on P suggests that, holding other factors constant, the proportional difference betveen the hourly earnings of incentive payments workers and those paid at time-rates is approximately 7 percent. This is, in fact, not significantly different from the figure we would arrive at by a comparison of the unadjusted wage differential of these 3/ workers (namely, 8.8 percent).— This positive coefficient on P may be interpreted as the value of the self-screening service provided to the employer by the operation of the piece-work system. It reflects the face — For further details of this in the Chicago labor market, see Rees and Shultz, [1970], pp. kk-6, 181-8U, and 220. Perhaps it should be added here that the estimated coefficients on other variables (such as distance to work and location) used by Rees and Shultz to account for the earnings differentials of these workers never approached conventional levels of significance and, more important, did not alter any of the qualitative conclusions from our estimates. 2/ — Since Xg measures the years spent by each individual on his present job and X_ indicates whether or not he had any occupational experience prior to starting his current job, the reader may wish to interpret the coefficients attached to these variables as isolating the effects of (post-schooling) specific training and of (post-schooling) general training, respectively, on earnings. The marginal rates of return to schooling are approximated by 31n Y/3X^ where an interpretation in terms of gros? investment and depreciation would seem to be appropriate. 3/ — The mean hourly earnings of timevo-}-erc and of piettrroi-ly-rvj not employed in the establishment that, oyeral^s both payment methods are $."!.'> 7 •••nd ;;2.>75 respectively. -la- that incentive payments workers are rewarded for transitory variations in their work effort, a variable that is omitted from this regression. One check on this explanation takes the form of estimating this regression equation for the 81* time-workers and for the 99 payaent-by-results workers separately: since the earnings of workers on time-races are independent of transitory variations in their effort, we would be omitting an effective constant from this regression equation for time-workers; on the other hand, since tr&nsitory variations in work effort are rewarded under an incentive payments scheme and since we should expect individuals to respond in different ways to this incentive system, a regression equation describing the earnings of payment-by-results workers would, indeed, be omitting an important determining variable. Equation (2) in Table 7 describes the earnings of workers on time-rates—''' and equation (3) of workers on incentive pay. The standard error of estimate (SEE) of regression equation (2) is about one-third of the standard deviation -of the logarithm of earnings of time-workers while the ratio of the SEE to the standard deviation of the dependent variable in equation (3) is almost twice that number. Alternatively, the independent variables included in the regression equation will normally predict the earnings of time-workers with much more precision than they will the earnings of workers paid on incentive schemes. This is consistent with the conjecture; —^The only unionized timeworkers are employed in the establishment that operates both time-rates and piece-rates. Hence the timeworker regression equation founders on collinearity between and G. This is why equation (2) in Table 7 omits the G- variable. ' -1+2- that an important explanatory variable (namely, work effort) has been omitted fron oqration (?).—^ As was pointed out above, the rsward for diligent vork offc?.t performed under time payments is provided by steeper earnirgs~e:-q>eriencc profiles than operates for payaent-by--results workers. Ms .Is evident from the regression ecuetions (2) and (3) in Table J. Compare the (logs/.-of) earnings-experience profiles for the two classes of workers tit the approximate mean values of schooling and previous occupational typerie;?.i.>i (nsmely, X., = 10 and X., = 0,50 for 'both groups) , For the time-rate workers, the slope of the logarithm oi' emings-experience relationship i given by =• .01675 - .00026 x2 while for payiaent-by-resu.lts workers it is = .0Q3?-> ■■■ .00000 3X2 The proportionate rise cf earnings with on-the-job experience is, indeed greater for time-rate workers than for piece-rate workers: the incentive to habitual work effort for employees on time payments involve greater rewards to on-the-job learning. The slight concavity (from below) in thti logarithm of earnings-experience relationship for time-rate workers and the even more imperceptible convexity in the relationship for piece-rake —test of the null hypothesis that the regression coefficients in equation (2) are no different frou those in equation (3) may be rejoet-.-d- -1*3-- ^orkers implies that, after a number of ye arc on the job, the proportionate return to experience becomes the same for the two classes of workers. For these employees, this obtains at 2k years of uninterrupted eyvnerience on the job. However, from these data not too much should be made of this convergence in the profiles: only rome 13 of all 133 male workers were observed with at least 2k years of experience so, at this range of values, inferences are being drawn from a very small sample of observations. The hourly earnings differential between these two groups of workers in the establishment that manages both wage payment methods is given by the sum of the estimates of -y and 5, namely, .068 - ,06k - .00k which, to all intents and purposes, is zero. Hence, an earnings differential between piece and time workers only obtains between workers employed ou the two payment systems in different establishments; in the only establishment that employs the two groups of vcrkers alongside one another, there is no such pay differential. Since according to equation (3) of Table 7 pieceworkers in this establishment earn almost ten percent less than the other male incentive workers in this sample, the absence of a piece-time pay differential in this establishment may be attributed simply to less work effort on the part of these pieceworkers,'-''' Altr;r;i:L: .:.. -V , the management of this plant may believe that a small interpersonal —''of course, 5 may be picking up other affects on earnings peculiar to this establishment that have little to do with the workings of the wage payment systems. An alternative procedure is to estimate an equation omitting the establishment variables (X_, Xg, and X„) and replacing them with separate dichotomous variables for each of the twelve establishments. The general drift of the results from this estinatin-"? equation does not conflict with those in Table 7 though the latter are typically estimated with greater statistical precision. dispersion of earnings encourages harmonious and productive work effc:. t. Note also that this is a unionized plant and the union itself may be operating effectively to narrow the pay differential of timevorkers and 1/ pieceworkers*. .— Trade Unions aid Wage Payment Systems This raises the general issue of the role of trade unions in the operation of different wage payment systems. In the popular Kind in the United States, labor unions are associated with strong opposition to incentive payments schemes. Wo doubt, this is attributable to the consp: cuous stand against piecework methods on the part of some unions, the United Automobile Workers in particular. Thus Walter Reuther of the U.A.W. spoke in I9U7 of the opposition to payineat-by-results as "...the cornerstone of our basic union policy" and continually bargained for the application of an hourly wage rate plan. Although some observers ha questioned whether union opposition provides a sufficient explanation for the move away from incentive pay in the automobile industry since the early 1930's, the importance of the U.A.W. policy as a contributory cau, 2/ is unchallenged.— However, there is abundant evidence of other unions being not merely sympathetic to, but actually encouraging of the use of 3/ incentive methods.— In many cases, presumably, the existence of the —^Equations similar to those in Table 7 were estimated for the female time and piece-workers. A 2^s per cent hourly earnings differential was measured in favor of piece-workers though this was not estimated with great precision. Not too much should be made of these results given our inability with these data to take account of the discontinuous market experience of these female workers prior to their starting on the pre-cer 2/ — See the measured discussion in Maedonald, [1963] , pp. 112-31. ■.tiiica introduces no element into the choice and operation of t'rxe *i~s^ payment system that would not alto be present if the workers were unorganized: the union does no more than register, though perhaps in a better publicized fashion, the attituf.es of its members. There may well be ether cases, however, in which the methods of wage payment afford diff. rent opportunities for unions to negotiate pay levels that are above competitive rates. For instance, it has been well documented that the costs of operating incentive payments methods are particularly sensitive to the climate of management-employee relations. Insofar as organized workers can threaten to be more effective in dislocating production than unorganized workers, the employer may have to offer a higher price to purchase union cooperation in the administration of the incentive pay scheme. This suggests the hypothesis that the impact of the union on wages is greater for workers on payme:;.b-by-results than for timevorkors, On the other hand, mention was made in the previous sa.,tlK.'i&';of circinrstnnees in which each individual employee may be working under a community-imposed output constraint: typically, these restrictions on output operate where workers fear that high earnings will irduce employer to cut piece-rates. If union and nun-union plants are equally effective in restricting output in this way, then, as far as this factor is concev:i there will be no union-related pay differential between timeworkers end pieceworkers. On the other hand, if the apparatus of union organization permits a more effective enforcement of output restrictions in union establishments, then, when all other variables are held constant, the average hourly earnings of pieceworkers will tend to be lower in vnioix -1+6- l>lexx'cr. than in nonunion plants:—'' put differently, unions are nore effective in eliminating observations in the upper tail of the frequency distribution of pieceworkers' hourly earnings. This argues for the earnings differential between pieceworkers and timeworkers being lower in unionized establishments. Consider the data in Table 8 on the proportionate difference in the hourly earnings of pieceworkers and timeworkers in union establishments and in nonunion establishments. If Y and Y. denote the hourly earnings P * of pieceworkers and of timeworkers respectively, then the third column in Table 8 lists r where r is defined as where the subscript u indicates unionized establishments and n nonunion plants. These data do not indicate a persistent tendency for the piecevorkers-timeworkers earnings differentia?, to be either higher or love-, in union as compared with nonunion establishments though unionism clearly — Some illustrations of the policing arrangements of the union are given in Van Dusen Kennedy, [19^+5]. Kennedy writes, "In some cases the union imposes penalties for serious infractions, in the form of a brief suspension from work or a money fine. An interesting arrangement is the "kitty" system encountered in a few cases. A limit on production or earnings by the hour or by the day is agreed upon and workers forfeit any amount earr.o'. over the limit into a common fund or kitty. This fund is expended periodically for a group function or celebration. That less formal and more direct methods of enforcement are sometimes used is indicated by the statement of a committeeman at one plant that if a worker persisted in exceeding the limit established some of the boys would "take him out in the alley and beat hell out of him," p. 118. Table 8 Average Hourly Earnings of Incentive and Tims i^.-^rs i:i American Machinery Manufacturing Industries in 19^-2 proportionate hourly earnings difference in occupation urion plants nonunion plants r male assemblers, bench, class A -.007 .163 _ .Ihb male assemblers, bench, class B .137 .209 - .060 male assemblers, bench, class C .lh2 .307 -, .126 female assemblers, bench, class C ,296 .091 .187 male boring-mill operators, class A .086 .390 -, .219 male boring-mill operators, class B .162 .300 -, .105 nale buffers and polishers „221 .Ull -, .135 male casting cleaners .192 .1+02 -, .150 wale craters, class B .169 .315 -, .110 male drill-press operators, class A .081 .265 -, .1^5 male drill-press operators, class B .163 .130 .029 male drill-press operators, class C .182 .037 .1U0 male heat treaters, class A .163 .516 -, .232 male lathe operators, engine, class A .012 .251 . 19.1 male lathe operators, engirt, .062 class B .179 .110 male lathe operators, turret, class A' .023 .278 .200 male lathe operators, turret, class B .187 .122 .053 male metal-saw operators .189 .207 -. .015 male milling-machine operators, class A .016 .332 -. .237 male milling-machine operators, class B .197 .321+ -, ,096 male packers ,3kl .175 .1U1 female packers .118 -.018 .139 male spray painters .238 .106 .119 male planer operators .063 .022 .Oho male power-shear operators .120 .0^5 .072 male screw-machine operators, class A .072 .26U -. ,152 male screw-machine operators, class B .128 .119 .008 male screw-machine operators, class C ,180 .158 . o.-.e occupation union plants nonunion plants . female testers, class C .276 .075 .1^7 male welders, hand, class A .163 .232 ~,0S6 male welders, hand, class B .376 .127 -221 male machine welders .290 .380 -.065 female winders, class C .228 .130 unweighted arithmetic mean r = -.057 number of occupations in union plants nonunion plants incentive earningstime-work earnings by 0- h.9% 5.0- 9.9/5 10.0-14.9* 15.0-19.9£ 20.0-2''.. 9;^ 25.0-29.9/i 30.0-34.9? 35.0-39.9# k0% an.d over total number of occupations 1 1 3 3 It 3 r j O 12 3 3 3 3 h x 1 2 0 3 33 33 From Effect of Incentive Payments on Hourly Earnings, BLS Bulletin Wo. 7 1943. -1*7- narrovs the inter-plant dispersion in (Y - Y, )/Y., Alternatively. p t, t using the Kees-Shulta data on male punch press operators, when an interactive term X^P (unionism dummy variable times the incentive pay dummy variable) is added to the regression equation (1*) above, the resulting coefficient estimate is positive (.093), but with a standard error larger than the estimated coefficient. Though different conclusions may well be extracted from cth?r bodies of data, there is little in these results to suggest that trad ft unions matter as far as the relat' pay of pieceworkers and time-workers is concerned. Cozjcl'jsionh:. The questions focused upon in tMs pe.per have been cenewjed viVh ulie problem of eliciting effort free: :Jiployees, Tnis problem arises because of the conflict cf interest latwten the owners of capital (or their representatives) and the Mrad labor and because malingering is costly to detect and prevent. This gives rise to tha issue of pr^viliag the appropriate set of incentives and penalties and Lho difiV.vetvfc -."age payment mechanisms are directed to that end. In particular, piece-rate methods tend to be preferred when to.e employer's problem of accurately screening his workers assumes especial importance. The seven per cent pay differential between male pieceworkers and tlmeworkers that has been estimated above measures the value of the signalling function provided to the employer by the operation of the piece-worker system. Further work c this theme should move beyond the simple piece- -rate/time-rate dichotomy of payment methods that has been fcMowed in this paper and in&tos-d exami the nature of different sorts of ; .c.yraeat iMchaniKo. References Alchian, Armen A., and Harold Demsetz, "Production, Information Cocts, and Economic Organization," American Economic Review, Vol. 62, No. 5, Dec. 1972, pp. 777-95. Becker, Gary S., and George J. Stigler, "Law Enforcement, Malfeasance, and Compensation of Enforcers," Journal of Legal Studies, Vol. 3, No. 1, January 1974, pp. 1-18. Behrend, Hilda, "The Effort Bargain," Industrial and Labor Relations. Review, Vol. 10, No. 4, July 1957, pp. 503-15. Blackburn, J.M., "The Acquisition of Skill: An Analysis of Learning Curves," Industrial Health Research Board (of Great Britain), Report No. 73, hTm.S.O., London, 1936. Brown, David G., "Hours and Output" in C.E. Dankerts F.C. Mann, and H.R. Northrup, eds., Hours of Work, Harper and Rows New York, 1965- pp. l'*'i- 60. Coleman, Terry, The Railway Navvies, Pelican Books, London, ly68. Epstein, E., and J. Monat, "Labour Contracting and Its Regulation: I," International Labour Review, Vol. 107, No. 5, May 1973, pp. 451-70. Florence, P. Sargant, Economics of Fatigue and. Unrest and the Efficiency of Labour in English and American Industry, George Allen and Unwin, London, 1924. Greenwood, J.A., "Payment by Results Systems: A Case Study in Control at the Workplace with a National Piecework Price List," British Journal of industrial Relations, Vol. 7, No. 3, Ncvoiiber 1969, pp. 399--H3 . Hall, G.R., P. Carpenter, S.A. Haggart, M.L. Rapp, and G.C. Sumner, A Guide to Educational Performance Contracting, Rand Corporation, R-955/1 - HEW, March 1972. Hobsbawm, E.J., "Custom, Wages, and Work-Load" in Labouring Men: Studien in the History of Labour, Weidenfeld and Nicolson, London, 1964. Howe, Christopher, Wage Patterns and Wage Policy in Modern China 1919-1972; Cambridge University Press, 1973. International Labour Office, Payment by Results, Studies and Reports, New Series, No. 27, Geneva, 1951. Johnson, George J5., and Frank P. Stafford, "Lifetime Earnings in a Professional Labor Market: Academic Economists," Journal of Political Economy, Vol. 82, No. 3, May/June 191k, pp. 549-69. Kem^.y, "'?jq Dusen, Unf.~>rt Policy e --I Tacan^jWap'e Methods, IT?v York.. Columbia University Press, 19^5. ICirseh, L.J., Soviet tfaftee:__Oh?-tif-;es in Structure and Administer:,!::"'.-.:-. j;2§i> M-I.tTpi'sss, 1972. Lcherrott, Ptenlcy. "The AT.ericcr. Labor Forte,." Lance E. T):: • Is et -?J._., .American Fcononi.c G-.'oi/th.' Ar- h:-. io:a.ist's Hi:-"-:or;r of the I ■•»-.:• ted States, Harper and Row, 1972. r.cT>i,>n, Tr-m, On the Pfcr/o Floor1 r- "t'ldHes of Workshop Organisation mid Output, Pergamon, 1963. Macdonald, Robert M. . Collective Bo:gaining hi the Aubomobila Industry, .'ale University Frens, New Kaveti .'T-i r-r.i'*orj, 1953» Marshall, Alfred.. Principles of Ecor.oaics, Macmillan London, oighth edition, 1920. " Mathews 011, Stanley B., Restriction of Output, .Ar.iong Unorganised Workers, The vUtiiig ih^fcwCi, I-lew Yorh, 1931. hincer, Jacob, Schooling, Experience, and TSandngs . National Bureau of Econonic Research, Columbia University Press, view York, 197^- National Board for Prices and Incomes, Payatent by Results Systems, Report No. 65, H.M.S.O., London, Car.a , fcy lySS end _rt, Report Wo. 65 (Supplement), Cur.d. ;5fi2','-I, D jcrraber 196S. l'encavel, J.H,? "Wgrss, Specific T.aining, and Labor Turnover in U.S. Manufacturing Indusxries,Intcrr, xhional 5<*oc .-vie 'iXan.e^. Vol 13, Ho. l, February 1972, pp. 53-"?^. Rees, Albert, "The Phillips Curve as a Menu for Policy Choic?," Economic Vol. 37, No. 1U7, August 1970, rp. 227-38. Rees, Albert, and George P. Shulta. "io'-rkers and Wa-^es in an Urban Labor Market, University of Chicago Press, 1970. Roethlisberger„ F.J,, and W.J. Dickson, Management and the Worker, Harvard University Press , Cambridge , Mass. , 19":>9. Slichter, Sumner H., James J. Healy, and E. Robert Livernash, The Impact-of Collective J^'gotoing on Mar.afieff.-T-t, Brookings Institution, Washingto i.e. I i960.""" Stigler, George J., "Informs.tion in the j>bor Market," Jcurat-1 of Folitl Economy, Supplement, Vol. L]:'X, October 196'i, pp, 9U-.1C5.