Popular in Course
verified elite notetaker
Popular in Business Management
This 49 page Document was uploaded by an elite notetaker on Monday December 21, 2015. The Document belongs to a course at a university taught by a professor in Fall. Since its upload, it has received 9 views.
Reviews for The-Future-of-Personnel-Economics-by-Edward-Lazear
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 12/21/15
The Future of Personnel Economics for IZA European Summer Symposium in Labour Economics Ammersee, Germany September, 1999 Edward P. Lazear Hoover Institution and Graduate School of Business Stanford University October, 1999 This research was supported in part by the National Science Foundation. Abstract Personneleconomicshasgrownoverthepasttwentyyearstobecomeamajorbranchoflabor economics. Beginning primarily as a theoretical field, recent empirical analyses have provided supportforearliertheories. Personneleconomicsisdistinguishedfromtraditionalpersonnelanalysis inthatitiseconomics. Assuch,personneleconomistsassumemaximizingagents,invoketheconcept of equilibrium, and focus on economic efficiency. Althoughmuchhasbeenlearned,manyimportantquestionsremain. Forexample,areworker wage profiles dependent on individual attributes or is the firm more important in determining wage growth? Whyareexecutivessohighlypaidandwhydoesthepaytaketheformthatitdoes? Whyhas the use of stock and stock options grown and why is stock sometimes given even to lower level employees? How can cross-country differences in pay patterns be explained? Does variable pay provide better incentives than do fixed hourly wages? Under which circumstances is one form of compensationusedoveranother? Thesequestionsandothersareinvestigatedandsomeconjectures are offered. Edward P. Lazear The Future of Personnel Economics September, 1999 Personnel economics is defined as the application of microeconomic principles to human resources issuesthatareofconcerntomostbusinesses. Thefield,nowabouttwentyyearsold,arose forthreereasons. First,thoseofuswhowereteachingstandardlaboreconomicstobusinessstudents encountered a disinterested audience. Business students believe that labor supply and demand, unemployment, investment in education, and other topics, which are of primary concern to labor economists, are almost irrelevant to their future business careers. Their boredom with what we had to sellsignaledthatachangeintheproductmightbewarranted. Second,theissuesstudiedbyhuman resources specialistswereofinteresttoeconomists,buttheapproachtakenbythenon-economistwas unsatisfying. It was loose, unfocused, ad hoc, and lacked the general, rigorous framework to which economists have grown accustomed. The economist's reaction was to enter an industry where it was believed rents existed. Third, the technology of economics changed. As a result of some breakthroughs, particularly those dealing with agency and contract theory, economists were better equipped to tackle problems that had evaded them in the past. 1 Personnel economics is an attempt to look inside the black box. It is an imperialistic attempt by economists to do what Alfred Marshall (1890) said that “economists do not do.” Marshall’s famous statement that it is not the economist’s business to tell the brewer how to brew beer has not been adhered to when it comes to personnel economics. Personnel economists are attempting to do 1Early papers in the agency literature include Johnson (1950), Cheung (1969), and Ross (1973). Contract theory grew out of macroeconomic inquiries, reflected in papers by Baily (1977), Azariadis (1983) and Gordon (1990). Finally, Alchian and Demsetz's (1972 ) important work on monitoring helped frame the discussion. 1 Edward P. Lazear The Future of Personnel Economics September, 1999 precisely that; namely, to use the tools of economics to understand and sometimes even to guide practitioners and consultants in their trade.2 This essay will attempt do three things: first it will describe the philosophy of personnel economicsandpointoutthedifferencesbetweenpersonneleconomicsandpersonnel.Second,itwill argue that personnel economics is real; that is, that personnel economics is useful in helping us to understand factual behavior in the real world. Third, since personnel economics is a young and growingfield,itwilldiscussanumberofnewquestionsandissuesinhopeofinspiringotherstojoin the field. Personnel Economics or Personnel? Like other sub-fields of economics, personnel economics differs from other social sciences in three ways. First, personnel economics assumes that the worker and the firm are rational maximizingagents.Constrained maximizationisthebasicbuildingblockofalltheoriesinpersonnel economics. Our empirical analyses in the field are attempts to test models that are based on maximizing behavior. Even when evidence suggests that the theories are wrong, we do not drop the assumption of maximization. Instead, our approach is to think more carefully about the nature of the modelsetup,butnotabouttherationalityoftheindividualsmakingthechoices. Economistsarerarely willing to assume that individuals simply do not know whatthey are doing. The economic approach allowsforimperfectinformation,transactioncosts,andotherinterveningvariableswhichmakethings somewhat more complicated, but the essence of personnel economics is to assume that behavior is determined primarily by the interaction of the agents and not by forces beyond their control. The 2This is the theme of a general essay that I have written entitled “Economic Imperialism” Lazear (2000). 2 Edward P. Lazear The Future of Personnel Economics September, 1999 success of personnel economics is in large part a result of simply assuming maximization because doing so allows the analyst to express complicated concepts in relatively simple, albeit abstract, terms.Thelanguageofeconomicsallowsthepersonneleconomisttoremovecomplexity.Detailsmay add to the richness of the description, but the details also prevent the researcher from seeing what is essential. Inmanyrespects,thisispersonneleconomics’mainsellingpoint.Thetypicalhumanresources textisverboseandshortongeneralprinciple.Indeed,manybookseschewgeneralization,arguingthat each situation is different. The economist’s approach is the opposite. Rather than thinking of each human resources event as separate, the scientific method that economists use places a premium on finding the underlying general principle and on downplaying other factors. Thesecondfeaturethatdistinguishespersonneleconomicsfromotherformsofhumanresource analysis is that personnel economists focus on equilibrium. Like the physical sciences, almost all theoriesinpersonneleconomicsareconsistentwithsomenotionofequilibrium.Forexample,workers in firms are assumed to react in a particular way where each side generally takes the actions of the othersintoaccount.Whenthisisdone,aparticularequilibriumresults,which,again,assistsinmaking very specific predictions about outcomes in the real world. Third, efficiency is a central concept of personnel economics. Adam Smith’s early notion of the invisible hand makes its way into personnel economics. Individuals who maximize their own utilityandinteractwithfirmsthatmaximizeprofitsgeneratebehaviorthatusuallymakesbothparties better off. When efficiency suffers, say, as a result of moral hazard problems that arise in the agency literature, the economist pushes the analysis to another level, asking what actions might firms and/or 3 Edward P. Lazear The Future of Personnel Economics September, 1999 workers take to alleviate such inefficiency. Taking this further step assists in makingbetter positive predictions and also normative prescriptions for the business student. Much of personnel economics is quite consistent with old style labor economics. Agency theoryconcentrateson inducingworkerstoputfortheffortandbearsaverycloseresemblancetothe theoryof labor supply. There are two differences. First, labor supply usually refers to hours of work or theproportionofthepopulationthatisinthelaborforce.Inthecaseofpersonneleconomics,much ofthediscussionfocusesoneffort.Hoursworkedismerelyonemetricofeffort.Second,becausethe focus is on effort, the inability to observe effort is central to much of the discussion. The tension betweentheinterestsoftheworkersandthefirm,althoughpresentinthestudyoflaborsupply,ismore problematic when effort is considered because firms cannot simply pay on the basis of observable 3 input or output. Personnel economics differs from personnel in that, as in all branches of economics, there is no freelunch. Firmshireworkersinacompetitivelabormarketandcannotsimplytakeadvantageof them. Workers cannot be induced to do things that they do not want to do without appropriate compensation, either in the form of money or some other non-monetary reward. Furthermore, personnel economics is willing to express all compensation in terms of money, even if money is not the only or most important factor incompensation. Non-monetary factors can be expressed in terms of money simply by finding its monetary equivalent through a standard hedonic approach. 4 In an analogous vein, personnel economists think in terms of substitution, where other human 3This is what my paper discussing the choice of payment of salaries and piece rates was about. “Salaries and Piece Rates,” Lazear (1986). 4See Rosen (1974). 4 Edward P. Lazear The Future of Personnel Economics September, 1999 resources specialistsdonot. Forexample,mostfirmshaveabenefitsdepartmentthatisdistinctfrom the compensationdepartment.Inmostfirms,compensationisdefinedspecificallytoincludemonetary remunerationonly.Thereisnoexplicitrecognitionoftrade-offs,andnon-economistsfrequentlythink in terms of providing some market level of each job attribute rather than thinking in terms of a total package of utility. To see how economists differ from industrial psychologists and organizational behavior scholars, consider the following paragraph from O’Reilly and Pfeffer (2000): Moreover, the fact is that economics is not very helpful in managing people. This is because the fundamental economic theory of motivation is based on assumptions of effort aversion (people will not expend effort unless paid to do so), opportunism(people,inthepursuitoftheirowninterests,willoftenmisrepresenttheir true preferences and engage in guile and deceit), and a lack of goal alignment (employees in organizations have different agendas than the owners and, therefore, incentive systems need to be designed to force people to do what is right for the good of the organization). In the economists’ view, people are assumed to be lazy, dishonest, and at odds with the goals of the managers. Althougheachoftheseassumptionsmaybevalidinaspecificsituation,orfor a particularindividual(forinstance,whenmanagingeconomiststhemselves),noneis likely to be right in most settings with normal human beings. Worse, if you begin by designing systems to protect against the small minority, you end up by alienating the majority. Yet, under economic logic, managers are encouraged to operate as though theseimplicitassumptionsarealwaysandeverywheretrue. Managersareencouraged to carefully design monitoring systems to check on people so that employees don’t misbehave; to define jobs in ways that reduce individual autonomy and maximize standardization to increase the efficiency with which employees can be selected and monitored; to craft systems that rely on money (either current or deferred compensation) to ensure that employees put in a fair day’s effort; and to otherwise ensure that management can check and control the behavior of subordinates. 5 The criticism of the economic approach that is discussed in the preceding two paragraphs reflects deep differences in the way that economists and non-economists think about motivational 5O’Reilly and Pfeffer (2000). 5 Edward P. Lazear The Future of Personnel Economics September, 1999 issues and human resources in general. First, consider the statement that people are assumed to be lazy, dishonest and at odds with the goals of the managers. It is true that economists make this assumption, but theydosoaboutbehavioronthemargin.Itisthemarginalbehaviorthatisofinterest to economists, and to personnel economists in particular, because the things that people want to do, do not require motivation. For example, many people take pride in their work and are willing to perform many tasks even in the absence of compensation. Most of us would continue to do some of our research and teaching even without pay. Few of us would choose to teach five courses per semester voluntarily. Firms are willing to pay to get workers to do things that they otherwise would notdo because those activities have value to the firm. It is not that people are assumed to be lazy, but rather that, on the margin, they are pushed to the point where the chosen task becomes a “bad” rather than a “good.” This is a proposition of equilibrium and one that others who study personnel ignore. To see the point more clearly, refer to Figure 1. Indifference curves are shown as solid U- shaped curves, where initially courses taught are a “good.” An individual would be willing to give up some salary to be able to teach additional courses. But this changes at around three and one-half courses, where additional courses taught become a “bad,” and must be compensated. The zero profit iso-profit curve is the dotted curve shown with an upward slope. It slopes upward because universitiesarewillingtopayhigherwagesifaprofessorwillteachadditionalcourses,whichourses generate revenue for the university. The equilibrium is at Point A. At that point, there is a tension betweenwhat the universities would like the professors to do and what the professors would choose to doifleftontheirown.Themovementfromthree-and-ahalftofour-and-a-halfcoursescanonlybe induced by additional compensation. Even if every single professor in the school had these 6 Edward P. Lazear The Future of Personnel Economics September, 1999 preferences,itwouldstillbethecasethatalltheinterestingactionisintherangeofpreferenceswhere people are “lazy” and have goals that are at odds with those of the managers. Indeed, from the point ofview of understanding behavior at either the normative or positive level, only the upward-sloping portion of the indifference curve is relevant. None of the variation observed inthe market is placed in the downward-sloping portion because universities are willing to pay more salary to get more courses taught. Similarly, if one were to give advice to managersabouthowtoinduceprofessors to “do the right thing” it would not be about getting them to move to three-and-a-half courses, since individuals would go there voluntarily. The part over which there is contention and interest is the movement from three-and-a-half courses to four-and-a-half courses. Personnel economics, unlike organizationalbehaviororindustrialpsychology,recognizesthispointandstatesitinaclean,precise fashion. Beyond this basic point, O’Reilly and Pfeffer state that personnel economics encourages managers to “carefully design monitoring systems to check on people so that employees do not 6 misbehave.” Again, this is only a result of equilibrium. Most employee behavior is in line with the interests of management and shareholders, but that is not the part that needs addressing. It is only the partofemployeebehaviorthatisatoddswiththeinterestsofmanagementandshareholdersthatneeds to be dealt with. As before, all of the variation occurs in the range where there is tension between managersandworkers,notintherangewhereinterestsarealigned.Incidentally,theoutcomedoesnot necessarily result in strict monitoring systems. Sometimes the equilibrium is one where behavior is self-enforcing. For example, allowing taxi drivers to be full residual claimants induces them to 6O’Reilly and Pfeffer (2000). 7 Edward P. Lazear The Future of Personnel Economics September, 1999 behave appropriately with no monitoring at all. It is the job of personnel economics to describe situations underwhichcompletefreedomisanappropriateincentivestructurethatservesbothparties well. To be fair, non-economists are more sanguine about understanding taste variations among agents than are most economists. If one knew which factors changed the shape of the indifference curves to make work more palatable, then tastes would be a variable that would require explaining and could be manipulated. Some of social and industrial psychology deals with precisely this: the attempt to understand preferences or manipulate them in ways that benefit the firm or the larger society. Unfortunately, neither the theory nor empirical analyses have been very satisfactory. 7 A final example drawn from another paper illustrates well the difference between the way psychologists and economists think about a problem. Consider the issue of labor mobility. Schkade and Kahneman (1998) claim that despite the common perception that living in California is pleasant, their study finds that, as measured by responses to a particular survey, Californians are no happier than are non-Californians. Let us ignore potential criticisms of the study that might argue that utility and statementsaboututilityaresubjective.Letusassumeinstead,asSchkadeandKahnemanpropose to show, that utility is an objective and quantifiable concept. Even so, the result that psychologists consider anomalousandinconsistentwithutilitytheorywouldbethoughtofbyapersonneleconomist inquite a different way. Because we focus on equilibrium, we would argue that the marginal person outside of California can be no less happy than the marginal person who lives in California. Wages, housing prices, and other living and working conditions adjust to make this so. The existence of free 7See Lazear (2000). 8 Edward P. Lazear The Future of Personnel Economics September, 1999 mobility ensures that happiness must be equated for the marginal individual across locations. Wages and pricesadjusttobringthisaboutinequilibrium.ThefactthatCaliforniaisapleasantplacetolive does not mean that everybody in the world rushes to live in California. As a result of its amenities, land pricesinCaliforniarisetomakethemarginalindividualindifferentbetweenlivinginCalifornia and living elsewhere. Economics makes sense out of what seems to others to be puzzling behavior. This is the strength of economics in general and of personnel economics in particular. Personnel Economics is Useful This section makes the argument by way of a number of examples that personnel economics helpsusunderstandhumanresourcespracticesinbusinessandalsogivesusawaytomakenormative statements. Some examples follow. 8 Worklife Incentives: My first paper on personnel economics was published in 1979. The initial purpose of the paper was to explain mandatory retirement, but the main contribution was that of suggesting that deferred compensationcouldserveasamotivator.Iarguedthatyoungworkersarepaidlessthantheir worth and older workers more than their worth in order to provide appropriate incentives over the life cycle. The basic idea can be explained as follows. Aworkercanchoosetoworkatahighlevelofeffortorhemayshirk,puttingforthalowlevel of effort. A worker who works at a high level produces output given bytheVprofileinFigure2. If he shirks, his output is VN. ThW˜ curve is the value of the worker's alternative use of time, in this 8The next section draws directly from Lazear (1999). 9 Edward P. Lazear The Future of Personnel Economics September, 1999 contextmost easily thought of as the value of his leisure. TimeTisthedateofvoluntaryandefficient retirement. If workers received compensation V, they would choose to retire voluntarily at time T, becausethatisthepointatwhichthealternativeuseoftimejustequalstheworker'smarginalproduct or payment. Consider two schemes,Wand V where Wis constructed such that the present value of theW path from 0 to T equals the present value of the V path over the same period. With perfect capital markets, workers would be indifferent between paths W and V if all else were equal, but all else is notequal. Consider a worker who is being paid according toV andisnearingtimeT. Atthatpoint, the incentivestoshirkbecomeoverwhelming. OnthedaybeforeT,theworkermayeitherworkatthe high level of effort or he may shirk. If he shirks, the worst thing that canhappenisthathegetsfired. If he gets fired, he does not receive wage V duringthenextperiod,buthedoesgettoenjoythevalue ofhis leisure, which is equal to his wageVattimeT. Nothingislostbyshirking.Ifinsteadtheworker were paidW, things wouldbedifferent. Undersuchcircumstances,shirking allowshimtoenjoythe value of his leisureW˜ next period, but he forgoes wage W. SinceWissetsuchthatitiswellabove V at time T, a worker would forfeit quasi-rents by shirking. Thus, a sufficiently steep W profile induces workers to perform at a higher level of effort than they would were they paid their marginal 9 product at each point in their careers. The keypredictionofthismodelisthatwagesrisemorerapidlythanmarginalproduct. There wassomeearlycorroborationofthisidea. MedoffandAbraham(1980)usedsubjectiveperformance 9Actually, a discrete payment after retirment is necessary to prevent shirking in the last moment on the job. See Akerlof and Katz (1989) for a good discussion of this point. 10 Edward P. Lazear The Future of Personnel Economics September, 1999 data and found that wages rose more rapidly with tenure than performance scores. Their work has been criticized as being less-than-definitive because the subjective ratings might reflect overall ability, relative position within job, or some other factor that was not well correlated with productivity levels. The theory begged for quantitative productivity measures to prove the case. Spitz (1991) had more objective data. She was able to examine the productivity of supermarketretail clerks and found that their productivity did not rise as rapidly with experience as their wages. A dataset from an American company, Safelite Glass, allows a detailed examination of the relation between experience-earnings and experience-productivity profiles. Safelite is the nation’s largest autoglass installer, headquartered in Columbus, Ohio. Safelite has a very sophisticated informationsystem and keeps detailed machine-readable records of weekly output for each installer in the company. ThedatausedcovertheperiodfromJanuary,1994toJuly,1995.Duringtheperiod, Safelite switchedfrompayinginstallershourlywagestopayingpiecerates. Pieceratecompensation is an alternative to career motivation schemes, so most of the discussion in this section will focus on the period during which workers were paid time rates. More will be said about the time rate / piece rate distinction below. The data on output per week and compensation per week were used to estimate the relation ofbothproductivityandpaytotenure. Table1reportstheresultsthatareunambiguous. Irrespective of the specification, the tenure coefficient in the pay regression is always higher than the tenure coefficient in the output regression. Twoequationsareestimated. Oneisthestandardlogofearnings(pay-per-day)ontenureand 11 Edward P. Lazear The Future of Personnel Economics September, 1999 tenure-squared. The other has the same right-hand variables, but uses actual daily output as the dependent variable. 10 The key result is that tenure has a greater effect in the wage regression than it does in the outputregression. 11 Using the coefficients from the hourly wage regime, figure 3 shows the slopes of the two estimated profiles. The estimated diagram (figure 3) resembles the predicted diagram (figure 2) in that the slope of tenure in the earnings regression is steeper than that of the productivity regression. The model is estimated in a number of different ways. Fixed effects are included, dummies are omitted, and a different definition of tenure is used. The results are unchanged: tenure has a greater effect on pay than on output. Input or Output-based Pay? General discussions of compensation involve the distinction between paying on the basis of input or paying on the basis of output. In Lazear (1986), the choice of compensation scheme is explored and a discussion of incentive and selection issues is provided. At the heart of much of personnel economics is the discussion of response to incentive. 10 The period is a month so that the figures are the average daily figures in a given month. The reported coefficients are estimated only on the period during which workers are paid hourly wages. This is the regime in which the upward sloping profile would be expected to be used to generate incentives. 11 Output is measured in windshield units, whereas pay is in dollars. In order to make them comparable, windshield units had to be converted to a dollar value. In order to do this, it was assumed that workers were paid their value of marginal products over a ten year work life. Thus, the value of a windshield was chosen to set present value of the pay equaled the present value of output over a ten year period. The results are insensitive to assumptions about the length of the work life. 12 Edward P. Lazear The Future of Personnel Economics September, 1999 Incentive responses are difficult to test. In order to do so, it is necessary to observe workers intwo different situations and compare the outcomes. But this presents two problems. First, in most circumstancesonlyoneschemeislikelytobeobservedforeachworkerifoptimalityobtains. Second, in the context of salaries and piece rates, the major reason for choosing input-based pay is that some or allcomponentsofoutputarenotperfectlyobservable. 12 Thisisproblematicbecauseitisdifficult to test a theory in which the predictions are about components that cannot be observed. Sometimes, it is possible to test behavioral predictions by examining related factors that are observable. In rare cases, however, the researcher is presented with data that are well-suited to the 13 purpose and allow direct testing of the theory. Recently, a number of cases of this sort has arisen. I have personally been the fortunate recipient of what is, for the purpose of testing whether workers respond to incentives, perhaps the best of the available datasets. A change in management induced an autoglass firm to switch from paying hourly wages to paying piece rates. The available data set provides detailed information on each worker’s output of autoglass installed, both before and after the change to piece rates. This allows testing of the theory set out in Lazear (1986). This case is instructive not only because it reveals the importance of compensation policy in affecting productivity, but because it demonstrates the power of economics in explaining the real world. Thepredictionsthatcomefromthetheoryarequitepreciseandareborneoutbytheevidence. 14 The theory emphasizes two effects of compensation scheme choice: incentives and sorting. 12 See Baker (1992). 13See Fernie and Metcalf (1999), Paarsch and Shearer (1996), Prendergast (1999). 14The sorting approach was suggested earlier in another context by Salop and Salop (1976). 13 Edward P. Lazear The Future of Personnel Economics September, 1999 A simple diagram, taken from Lazear (1999b), illustrates both mechanisms and the predictions. Safelitegraduallyswitcheditsworkforcefromanhourlywagepaymentstructuretopiecework over about a one year period between 1994 and 1995. Autoglass installers were previously paid hourly wages. They were moved to asystemofpaymentbynumberofunitsofglasswithaminimum hourly wage guarantee. If the piece rate pay fell short of what the worker would have earned under the old system, he was paid according to the old hourly wage formula. Implicitineveryhourlywageofferisarequiredminimumstandardofeffortoroutput. Firms do not continue to pay workers who do not meet some minimum standard. Safelite’s plan guaranteed W to anyone who would have earned less thanW under the piece rate, and paid the piece rate to all of those whose compensation by the piece rate formula would have exceeded W. The scheme used is Compensation = max [W,be&K]. Here, W is the guaranteed wage, b is the piece rate based on number of units of output, e, and K is a constant term to satisfy the individual rationality constraint. The situation is shown in figure . Indifference curves shown in the diagram represent individuals with different tastes or abilities. Low ability workers have steep indifference curves because additional effort must be compensatedbylargeincreasesinincome. Solidindifferencecurvesarethoseforrelativelowability workers and dotted indifference curves are those for higher ability workers. The hourly wage schedule is shown by the function that starts at zero, becomes vertical at e 0 and then horizontal at point A. The piece-rate-with-guarantee schedule is the same, except that compensation rises with output above e*, as shown by the upward sloping segment. 14 Edward P. Lazear The Future of Personnel Economics September, 1999 If workers are offered the hourly wage schedule, then everyone chooses point A since there is no value to working at higher levels of effort. But high ability workers choose to move from A to Bwhenofferedthepiecerateschedulewithaguarantee. TheleastableremainatpointA. Thereare three implications. 15 First, average effort does not decrease and generally increases whenthe firm switches from hourly wages to piece rates. As longassomeworkersputforthenoughefforttobeinthepiecerate range, then average output rises. This is the incentive effect associated with moving from hourly wages to piece rates. Second, average ability of the work force increases because the ability of the lowest quality workerdoesnotchangeasaresultoftheswitchincompensationscheme,buttheabilityofthehighest qualityworkerrises. Now,someworkerswhowerepreviouslyunwillingtoworkatSafelitebecause hourly wages were too low, given their ability and alternatives, now find that the piece rate allows themto work harder and receive more from the job. The least able worker is indifferent between the two schemes. Switching to piece rates has the effect of improving retention and recruitment of high quality workers. This is the sorting effect. Third, variance of worker ability and output rises after the switch to piece rates. Even if underlying ability levels did not change, variance in productivity would rise because some workers remain at A, whereas others work in the piece rate range, with output levels exceeding e*. This, coupled with the fact that the maximum ability level increases under a piece rate implies that the increase in output variance becomes even greater. The evidence on these points is presented in table 15The formal derivation of this material is contained in Lazear (1999b). 15 Edward P. Lazear The Future of Personnel Economics September, 1999 2, which comes from Lazear (1999b). The two main points can be summarized as follows: 1. Overallproductivityincreasedabout44%(anincreaseinthelogof.368)asaresultofthe switch from the hourly wage contract to piecework. 2. The increase can be split into two components. Because individual workers can be followed and their output can be measured before and after the switch, the incentive effect can be taken out. It is defined as the increase in output for a given worker after the switch to piece rates occurred, averaged across all workers. This amounts to a log increase of .197, or a 22% increase in productivity for the average worker, as shown by the coefficient in the second row of table 2. The sorting effect shows up most clearly in the coefficient on new regime. New regime is a dummy equal to individuals who are hired after the piece rate scheme has been implemented throughout the firm. Note that these individuals are 0.243 log units more productive than the corresponding individuals who are not hired under the new regime. Although there is some evidence that separation rates for high-ability people fell slightly after the switch to piece rates, the strongest effect on selection seems to have occurred through the hiring of individuals who were higher output producers than the ones whom they replaced. Tournaments: Another form of incentive compensation depends on a worker’s relative performance rather than his or her absolute performance. The earliest statement of incentives based on relative performance is Lazear and Rosen (1981). The analysis featured there has come to be called “tournamenttheory.”Theessenceoftournamenttheoryisthatindividualscanbemotivatedsimplyby 16 Edward P. Lazear The Future of Personnel Economics September, 1999 comparisonwith their peers or by comparison to some standard. The basic idea is that compensation is fixedinadvanceanddependsonone’srelativepositioninthefirm. Thelargerthespreadbetween the “winner” and “loser,” the greater are the incentives. There is an optimal level of incentive that is the result of trading off the value of additional effort against the compensation necessary to induce workers to supply that level of effort. At some point, increasing the spread raises the equilibrium effort and disutility from it workers to such an extent that it is better to settle for a lower amount of effort rather than to pay workers enough to induce them to perform. Tournament theory has a number of testable implications, some of which have already been validated in the literature. One type of test examines the underlying structure of compensation to see whether it is consistent with tournament theory. A good recent example of such a study is Eriksson (1999) Another literature tries to examine whether the hypothesized effects ofspread on effort are actually found inthe real world. Some of these look at sports tournaments. Ehrenberg and Bognanno (1990) is a fine example. Fernie and Metcalf (1999) look at jockeys and their performance as performance relates to pay. Outside the sports arena are two examples. One is a series of studies by Knoeber (1989) and Knoeber and Thurman (1994), who examines the supply of chickens in a tournament-like environment. Another, by Drago and Garvey (1998), finds evidence of tournament- like behavior in Australian firms. All of these papers provide consistent evidence of the effect of spread on output and are very much in line with hypothesized tournament behavior. One can look at the importance of relative performance more directly. To do this, data were acquired from a large financial service firm that had approximately 42,000 employees at the end of 16See Tor Erikkson (1999). 17 Edward P. Lazear The Future of Personnel Economics September, 1999 thesampleperiodin1994.Informationisavailableonallemployeesinthefirmfromtheperiod 1986 through 1994, not only those who were present throughout the entire period. The file consists of a series of snapshots of the payroll data on the same date of each year. This data set provides information on individual performance ratings done by the supervisors annually. The ratings take on numerical values between zero and five and can be used to make relative comparisons. For the purposes here, the reference group is defined as a job-year; that is, each year of the job the average score on the performance evaluation is computed. So, for example, all secretary IIs in 1991 would formonecomparisongroupandthemeanperformancescoreforthoseindividualswouldbecomputed. Then,avariablethatmeasuresthedifferencebetweenanindividual’spersonalperformancescoreand the mean for his or her job is computed. This variable is called “difference in performance.” The tournament model predicts that those with high relative performance in their job are the ones most likelytobepromoted. Thejobisrelevantbecauseallindividualscompetingatthesameroundofthe tournament, should, ex ante, have the same estimated ability. At the empirical level, the job-year has been used as the group. All individuals in the same job title in a given year are assumed to be in the comparisongroupforthepurposesofthisanalysis. Deviationsfromthemeanofthatgroupinagiven year define relative performance. The results of the estimation are reported in table 3, where the probability of promotion is posited to depend on absolute and relative performance scores. As is clear from table 3, “difference inperformance,” defined as the difference between the absolute level on the performance rating and the job-yearmeanismoreimportantthantheabsolutelevelofperformanceindeterminingpromotion probability. (The standard deviation of the level of performance is .61 whereas that of the difference 18 Edward P. Lazear The Future of Personnel Economics September, 1999 is .50. The means are 3.48 and zero, respectively. The probability of promotion is .20.) Two individuals, A and B, who each receive, say, 4s on their evaluations are only equally likely to be promoted if the mean for their comparison groups are the same. If individual A is in a job where the average score is 3 and B is in a job where the average score is 2, then B is 6 percentage points more likely to be promoted thanA. The six points are almost one-third of the average promotion rate per year, so the effect is quite large. These findings are supportive of the tournament view of promotion. Note that the analysis is done using firm-based data, which allow the relative comparisons to be 17 made. Salary Compression: Wheneveranypartofcompensationisrelative,alargespreadbetweenthewagesof“winners” and “losers” results in a competitive, rather than cooperative work environment. For this reason, firms and workers may choose to adopt a more compressed wage structure. 18 Compressionmustbedefinedrelativetosomethingandtheappropriatemetricisproductivity. Again, the Safelite data permit an examination of productivity directly. Since there is also detailed information on compensation, it is possible to examine the relation between compensation and productivity to assess the amount of compression. The Safelite data are not ideal for this purpose because the pay compression argument has its 17 The same qualitative results are obtained for wage growth instead of promotion as the dependent variable. Furthermore, using rank on the performance evaluation, rather than the deviation from the mean, has the same type of effect on determining promotion in a linear probability model with job-specific fixed effects. 18 See Lazear (1989). 19 Edward P. Lazear The Future of Personnel Economics September, 1999 greatest force when individuals work together in teams. To the extent that installers are working individually, thereismuchlessmotivationforusingacompressedpaystructure. Still,itisinstructive to examine the Safelite data. The data are split into two regimes. Data are examined from the period when workers were paid hourly wages and from the piece rate period. The hourly wage period would appear to be more relevant because the piece rate structure, almost by construction, implies that compensation and productivityarematched. Onceafirmchoosestobasepayonthepiece,ithasmovedawayfrompay compressionsotheonlyquestionwouldbe“Whydoesthefirmchoosetopayonthebasisofoutput?” But this is not quite true. Because the firm has choice over the piece rate formula, wages can be compressed even under piece rates. For example, if compensation is given by Pay = a + ß e where emeasuresoutput,settingß=0 anda equaltosomepositivenumberwouldresultincompletely compressed pay; a straight salary structure. As ß goes to 1, the correlation between pay and productivity increases. Table 4 presents some data on wage andproductivityvariation. Firstnotethattheamountof variation in wages is always less than the amount of variation in productivity. This is true both in absolute terms and relative to the means. At some level, this result is almost guaranteed duringthe hourly wage period. Since the hourly wage does not change from day to day, even as hourly output inunitsvaries,itisnecessarilythecasethatproductivityvariationwillexceedwagevariationatleast on a hourly basis. The kindofvariationthatisofmoreinterestisvariationacrossworkers,ratherthanvariation 20 Edward P. Lazear The Future of Personnel Economics September, 1999 for a given worker over time. To get at this, fixed effects on output were computed from the fully specified regressionusedinLazear(1999b),whichrelateslogofoutputtomonthandyeardummies, tenurevariables,apieceratedummy,andinteractioneffects. Fixedeffectsonwageswerecomputed from an analogous regression with log of pay-per-day as the dependent variable. Again, the result is the same. The amount of variation in output fixed effects exceeds that in 19 wage fixed effects. Interpersonal variations in output are larger than interpersonal variations in compensation. As before, this is true for the whole sample period and also for each regime taken independently. The results of table 4 are clear evidence of pay compression relative to productivity. They mayor may not support the view that firms compress wages so as to bring about more harmony at the workplace. There is a fundamental difficulty in testing the latter proposition. In order to provide evidence on this point, it is necessary to be able to observe individual output so that output can be compared to productivity. But when individual output is observable, it is less likely that team production and cooperation are important. As a result, the evidence presented should be viewed as consistent with, but not conclusive proof that firms compress wages to foster cooperation. Peer Pressure: Kandel and Lazear (1991) analyze the effects of peer pressure on output. It is possible to 19The mean of the fixed effects differ (slightly) from zero for two reasons. First, only individuals who have values of fixed effects on both output and wages are included in the table, whereas each regression includes all individuals who have values only for the variables in that regression. Second, the weighting is different because table 4 computes the mean of each individual’s fixed effect and does not weight by the number of observations used in computing each fixed effect. 21 Edward P. Lazear The Future of Personnel Economics September, 1999 evaluate the effect of peer pressure by comparing the change in output of individuals who never produced enough to be in the piece rate range to the change in output to those who do make it into the piece rate range. Let us define two groups of workers. A dummy variable, L, for low output, is set equal to 1 ifaworkerneveraveragedmorethanfourunits-per-dayforanymonthduringwhichhewasemployed. The piece rate range does not start until a worker has produced at least 4.5 units, so these workers probably assumed that they would be earning the guaranteed hourly wage. In terms of figure 4, they are the workers who should have chosen to be at point A. The fixed-effectsregressionreportedintable3(row2)wasre-runontwogroupsseparately, namely, those having L=1 and those having L=0. To determine whether there is implicit pressure on those who are not affected directly by the piece rate structure, it is only necessary to compare the effect of switching from hourly wages to piece rates across groups. Because the group with L=1 is never in or very close to the piece rate range, there should be no direct effect of the switchonthose workers. The results are reported in table 5. Workers who are or have potential to be in the piece rate rangeexperiencedanaverageincreaseinoutputof.55units. Thosewhoneverreached(orwereclose to reaching) the piece rate range experienced an increase in output of .32 units. The model as illustrated by figure 4, taken literally, implies that there should be no change in output for those who are goingtobeoutofthepieceraterange. Workersshouldchooseeithertobeontheupwardsloping segment of the compensation schedule or to locate at point A. 20 20Note that the coefficients in table 5 are the effect only of the switch to piece rates. Thus, output of low output workers increased by .32 units net of all other effects when the store in which 22 Edward P. Lazear The Future of Personnel Economics September, 1999 Workerswhoareoutofthepieceraterangeprobablyincreasedtheireffortlevelbecausethey sensedthate ,0heminimumacceptableoutputdependedonwhatotherworkerswereproducing. Since the outputofotherswentupsubstantially,thelowoutputworkersmayhavefeltcompelledtoincrease their output. This is best interpreted as the effect of peer pressure. Because the increase in output is associated directly with the actual switch to the new compensation scheme and not merely with the announcement that a switch is going to occur, it seems reasonable that the increase among the low outputworkersresultswhentheyseeothers’outputsrise. Theriseinpeeroutputmakesmorecredible the implicit increase in eo. The conclusion ofthis section is that personnel economics provides a series of theories that have normative value, but more important, that are borne out by real-world observation. In the past, economists have been limited in testing the theory of personnel economics because the appropriate data have not been available. Recent data acquisitions, some of which have actual productivity information reported, have made it possible to test the theories and to provide verification or refutationofthem.Atleastfromthelimiteddataavailablesofar,itseemsthatthereisgoodevidence that worklife incentives incorporated in upward-sloping wage-earnings profiles, tournament style incentives,paycompressiontheoriesandpeerinteractioneffectsfindsupportinthedata.Furthermore, simple models that predict how switching from payment by input to payment by output will affect worker behavior are confirmed. the individual worked switched to piece rates. Since this is net of any time dummies, the .32 is not the effect of improvement in general conditions at Safelite in general. 23 Edward P. Lazear The Future of Personnel Economics September, 1999 Issues for the Future As a result of the changes that have taken place at the theoretical level and of the recent availabilityofnewfirm-baseddata,personneleconomicscanaddressissuesthatwereignoredinthe past. Inthissectionasampleofissuesthatremainunresolvedarepresented,bothatthetheoreticaland empirical levels. Intrafirm Mobility: The mainadvantageoffirm-baseddataisthatanalysescanbedonethatrequireunderstanding the relation of one worker to other workers within the context of the firm. For example, in order to understand whethermostofanindividual’swagegrowthisassociatedwithidiosyncraticcomponents thatare person-specific, or whether it is associated with general phenomena that affect the firm as a whole, itisnecessarytohavedataontheexperiencesofallworkerswithinthegivenfirm.Onemight put the question as “Are wages determined by outside competitive labor pressures or does a rising tide lift all boats?” There are currently a number of data sources that will allow these hypotheses to be distinguished.Swedencollectsdataonallworkersinmanufacturingestablishmentsandthesedata are available from the period of 1970 to 1991 on a quarterly basis. In future work, those data will be used to test the hypothesis. But it is possible to use the financial services firm data to get a preliminary idea of the importance of within-firm transition. One such analysis was presented in Lazear (1999a). The transition matrix shown in table 6 enables us to answer these questions. Columns reflect quintiles inearningsin1994. Rowsreflectquintilesinearningsin1990. Thus,theentryof457with 24 Edward P. Lazear The Future of Personnel Economics September, 1999 5.09 under itin row 1, column 2 is interpreted as 457 individuals or 5.09 percent of the total sample consistedofindividualswhowereinthelowestquintileinearningsof1990andinthesecondlowest quintile of earnings in 1994. It is obvious that the lowest level individuals tend to move up in the firm. This is simple arithmetic. They cannot move down so they either all remain in the same quintile or some move up. Infact, conditional on remaining with the firm for the period, half of the individuals who were in the lowest quintile in 1990 have moved up by 1994. Within-firmmobilitycanbemeasuredinanotherway. Thecellsabovethediagonalallreflect upward mobility. Those on the diagonal reflect no movement out of the quintile. Those below the diagonalreflectdownwardmobility. Itisuseful,therefore,tolookattheproportionineachcategory. Theresultsarethat57%stayedinthesamequintile,but43%movedoutoftheirquintileduring the four year period. Twenty percent of the initial sample moved up at least one quintile and 23% moved down at least one quintile. There is a great deal of upward and downward mobility in this firm. This is not surprising, given the rapid growth during the period. But the same pattern is observed when the transition matrix for 1986-90 is examined. In this earlier period, not shown on 21 the table, 61.6 percent remained in the same quintile; 17% moved up and 21.4% moved down. The fact that more move down than up suggests that new hires come in at higher wages or levels than incumbents and their position erodes over time. Perhaps a more important point is that the chances for moving up and down in this firm are 21 The number who move up does not equal the number who move down because quintiles are defined relative to those in the firm at a point in time. In theory, all could move up if every new entrant came in at a salary below those of the incumbents and there were a large enough number of new entrants. 25 Edward P. Lazear The Future of Personnel Economics September, 1999 substantial. Since most personnel economics models postulate the necessity for internal reward or punishment, it is encouraging to see that mobility is common. Were all workers locked into a particular position, most of the theories thatdiscuss schemes for addressing moral hazard would be suspect. The transition matrix can be used in a large variety of ways, but in the analysis just presented there is a major defect; namely, it refers to the experience of only one firm. In the Swedish data set, mentioned earlier, and in those from Finland and Norway, a large number of firms can be examined so that general patterns can be described and related to underlying characteristics of the firms and workforce.ThedatasetfromDenmarkhastheadvantagethatitincludestheentirelaborforceandnot just those involved in manufacturing, so that intersectoral mobility can be analyzed. Are Executives Paid Too Much? Manyoftheunresolvedquestionsinpersonneleconomicsrevolvearoundtheissueofwhether paypracticesarerational.Mostobvious,bothinthepopularpressandtosomeextentintheacademic literature,isthequestionofwhetherCEOsandhigh-levelexecutivesareoverpaid.Over-paymustbe defined first of all, because in a tournament context pay today does not necessarily relate to productivitytoday.Thesameistrueifwork-lifeincentivesoperatebecausehigh-levelexecutivestend to be more senior workers in the firm than those who would be in the part of the life cycle during which quasi-rents are being earned. Some of the empirically-observed growth in CEO pay in the U.S. over the past twenty years reflects ex post rather than ex ante increases, to the extent that a large fraction of CEO has taken the 26 Edward P. Lazear The Future of Personnel Economics September, 1999 form of stock options. Ex post compensation is related to market performance, which has been exceptional during the 80s and 90s. This aside, there is some evidence that even ex ante pay where stock options are valued at Black-Scholes, American CEO compensationstillseemstoohigh,given the behavior ofother countries and of other workers within the United States. The question, then, is whether this reflects market forces or some enhanced ability of managers to transfer rents from shareholdersandotherworkerstothemselves.Thisbegsthequestion,however,astowhythisgrowth wouldhaveoccurred.Presumably,managerswerealwaysgreedy.Whyaretheybetterabletoexercise their desires now than they were in the past? It is certainly interesting to know about differences in levels of CEO pay, especially across countries. Perhaps more interesting, though, is whether the form of CEO pay matters. Fernie and Metcalf (1999) andLazear(1999b)andfindthattheformofcompensationcanhavedramaticaffects 22 on production worker productivity. It is possible that the same is true for CEOs. Re-pricing Options: Related to this is the issue ofre-pricing of stock options. Often, after a significant decline in the priceofstock,whichrendersanexecutive’soptionsignificantlyoutofthemoney,theoptionsare re-priced; that is, the exercise price is lowered so as to make the option valuable again. The logic often given for the re-pricing is that the move is necessary in order to restore the incentive effects of the option value. This logic is valid, as shown by the following model. 22 CEOs are selected in large part for their willingness to put forth high levels of effort. Thus, the elasticitiy of CEO effort to compensation may not be so high. Of course, the value of incremental CEO effort may be much higher than that for production workers. 27 Edward P. Lazear The Future of Personnel Economics September, 1999 First note that in most respects, stock is a special case of stock options. A stock option is an optiontobuyashareatsomeexercisepriceK. IfKweresettozero,thenthevalueofthestockwould always be higher than the exercise price and the owner of the option would always exercise it; that is, the stock option would always be "in the money." This is equivalent to saying that the holder of 23 the optionownsthestockwithcertainty. Itisusefultothinkofstockthisway.Then,wecanalways think in terms of options and simply determine the optimal exercise price. A price of zero is the special case that implies the granting of stock. Now we will derive the expression which gives the value of a simplified option to a risk neutralholder. Consideranoptionthatcanbeexercisedononedayandonedayonlysometimeinthe future. Suppose that current value of the stock is V and the exercise price is K and e is a random variable that reflects the change inthe value of the stock between now and the exercise date. Let the discount rate be zero. Then the expected value of the stock option is 4 (1) Z ’ [V%e&K] f(e)de m K&V since the stock option is not exercised unless V + e> K, or equivalently, when e > K - V. For a givenstrike price, increasing the value of the firm, V, increases the value of the option by the probability that the option is in the money. That is, 2There are a couple of differences, however. An owner of an option does not receive dividends on that stock until he becomes an actual owner of the stock. Furthermore, the owner of an option does not have voting rights. But the option owner might have tax reasons to avoid exercising the option early. In the U.S., options are not taxed as income until they are exercised. 28 Edward P. Lazear The Future of Personnel Economics September, 1999 MZ (2) MV ’(V%K&K&V)f(K&V)% 1 & F(K&V) or (3) MZ ’ 1 & F(K&V) MV which is greater than zero. The first term of (2), which reflects the increased likelihood of being in themoney,dropsoutbecauseatthepointwheretheindividualismorelikelytobeinthemoney,being in the money has no value. The second term, which survives, reflects the fact that a one dollar increase in the value of the firm means one dollar to the option holder every time the option is in the money, which happens 1 - F(K-V) of the time. It is this point that generates the logic behind the repricing of stock options. If a manager’s optionisfaroutofthemoney,theoptionhaslittlevalueandtheincentiveeffectissmall. Themanager has greater incentives to increase V, the higher is the probability that the option is in the money. Although providing a rationale for repricing of options, repricing creates other incentive problems. If the manager knows that the option will be repriced, he will notput forth as much effort initially. There are two at least partial solutions to this problem. One is to index to exogenous changes in stock prices (i.e., market events) that do not reflect manager activity. An alternative is to allow repricing, but to charge the manager a fee each time repricing occurs. The fee, which depends on the amountofrepricing,assistsinaligninginitialincentives,andtherepricingensuresthatexpost incentives are correct. The fee and repricing pattern that provides optimal incentives can be worked out. 29 Edward P. Lazear The Future of Personnel Economics September, 1999 Personnel economists still do not have a clear understanding of why stock and options are given at all in light of such large free rider effects. It would seem that bonuses that relate more directly to a manager's contribution would provide better incentives than stock or stock options. Options,however,doprovideleverageandallowsomeoffsettingofeffort-dampingfreeridereffects. This can be seen in the following example. A share of stock is currently worth V and will be worthV+e on the date at which an option can be exercised. Supposethattherearetwopossibilities for e: good luck, in which case, e = 1, or bad luck, in which case e = -1, each of which occurs with probability ½. Suppose further that if the manager puts forth the anticipated effort, V = $10. Then, since the expected value of e is 0, the stock price would be $10, reflecting that half of the time the firm's value would be $11 and half of the time it would be $9. An option with a strike price of $10.50 would be worth $.25 because the option is in the moneyhalf of the time and when it is in the money, it returns $11-10.50 or $.50. Thus, 40 options at an exercise price of $10.50 are equivalent to one share of stock (stock being an option with an exercise price of zero). Now,suppose that the CEO can do something to raise the value of each share by $.01. If the CEOownedoneshareofstock,hewouldtakethemoneysavingactionifandonlyifithadapersonal cost of less than 1 cent to him. Suppose instead that he had been given 40 options with an exercise price of $10.50. Initially, this has value equivalent to that of the one share of stock. Raising the value ofa share of stock is now worth $.20 as opposed to the $.01 because the option will be in the money half of the time and when it is, each of the 40 options returns $.51 instead of $.50. Options provide leverage and allow the CEO to capture more of the return. As such, options offset free rider effects. 30 Edward P. Lazear The Future of Personnel Economics September, 1999 Personnel economics is providing a starting point for understanding variable pay, including the granting of stock and stock options, not only to managers but to employers much further down in the organization. Many questions remain unanswered. Cross-Country Comparisons: All the previous theory was quite general and does not speak to the differences in CEO pay across countries. The large expansion of American CEO pay relative to that of other countries has attracted particular attention. As mentioned earlier, part of this is merely a question of timing, in that expostandexantereturnsaresomewhatdifferent.The“beta”onAmericanCEOcompensationseems to be higher than that on compensation of CEOs in other countries. Because of the nature of the compensation contract, American CEO payvaries more with market conditions than does the pay of Asianand European CEOs. An important question is “Why should this be the case?” We do not have goodtheoriestoexplainthespecificcompositionofmanagerialcompensation,althoughearlierwork inpersonneleconomicshaslaidthegroundworkforsuchdiscussion.Still,littleapplicationhasbeen made of these theories, and this remains
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'