The Causes and Consequences of the Dependence of Quality on Price Author(S): Joseph E

American Economic Association

The Causes and Consequences of The Dependence of Quality on Price Author(s): Joseph E. Stiglitz Reviewed work(s): Source: Journal of Economic Literature, Vol. 25, No. 1 (Mar., 1987), pp. 1-48 Published by: American Economic Association Stable URL: http://www.jstor.org/stable/2726189 . Accessed: 13/08/2012 12:34

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to Journal of Economic Literature.

http://www.jstor.org Journal of Economic Literature Vol. XXV (March 1987), pp. 1-48

THE CAUSES AND CONSEQUENCES OF THE DEPENDENCE OF QUALITY ON PRICE*

By JOSEPH E. STIGLITZ Princeton University

If the legal rate . . . was fixed so high . . . , the greater part of the money which was to be lent, would be lent to prodigals and profectors, who alone would be willing to give this higher interest. Sober people, who will give for the use of mnoneyno more than a part of what they are likely to make by the use of it, would not venture into the competition. . . . Adam Smith, Wealth of Nations, 1776. A plentiful subsistence increases the bodily strength of the laborer and the comfortable hope of bettering his condition and ending his days perhaps in ease and plenty animates him to exert that strength to the utmost. Adam Smith, Wealth of Nations, 1776. Low wages are by no means identical with cheap labour. From a purely quantitative point of view the efficiency of labour decreases with a wage which is physiologically insufficient . . . the present-day average Silesian mows, when he exerts himself to the full, little more than two-thirds as much land as the better paid and nourished Pomeranian or Mecklenburger, and the Pole, the further East he comes from, accomnplishesprogressively less than the German. Low wages fail even from a purely business point of view wherever it is a question of producing goods which require any sort of skilled labour, or the use of expensive machinery which is easily damaged, or in general wherever any greater amount

* I am greatly indebted to my several coauthors, do not mean cheap production, and that the best with whom I have worked on the analysis of the instructed and best paid labor proves itself to be causes and consequences of the dependence of qual- the most productive.... Similarly,Thomas Bras- ity on price: Carl Shapiro, Barry Nalebuff, Andy sey, Jr., is quoted by John H. Habbakuk(American Weiss, and Bruce Greenwald. I am also indebted and British Technologyin the Nineteenth Century, to helpful conversationswith George Akerlof, Janet 1962) as saying, "The cheap labour at the command Yellen, FranklinAllen, Bill Rogerson, Mark Gerso- of our competitors seems to exercise the same ener- vitz, Jonathan Eaton, Partha Dasgupta, among vating influence as the delights of Caphua on the others. Larry Summers and Gary Fields provided soldiers of Hannibal" (Work and Wages, 1872, p. helpful comments on an earlier draft. 142). The opening quotationswere supplied by Michael Some of this earlier literatureis reviewed by Greg- Perelman, Gavin Wright (second and fourth quota- ory Clark,in "ProductivityGrowth without Technical tion), GracielaChichilnisky, and FranklinAllen. Change: EuropeainAgriculture before 1850" (1986), Gavin Wright has drawn my attention to the fact and by A. W. Coats, in "ChangingAttitudes to La- that there was a large literature in the eighteenth bour in the Mid-eighteenth Century," Economic and nineteenth centuries arguingfor a link between History Review (August 1958, 2(11), pp. 35-51). wages and productivity. For instance, he cites the Coats traces the idea that high wages lead to high former U.S. Secretary of State Thomas F. Bayard's productivityback to JacobVanderlint, in MoneyAn- "IntroductoryLetter" to Jacob Schoenhof (in The swers All Things (London, 1734). Economy of High Wages New York, 1893) as saying, Financialsupport from the Hoover Institutionand "The facts you have adduced and your deductions the National Science Foundation is gratefully ac- irresistibly establish the proposition that low wages knowledged.

I Journal of Economic Literature, Vol. XXV (March 1987)

of sharp attention of of initiative is required. Here low wages do not pay, and their effect is the opposite of what was intended. Max Weber, The Protestant Ethic and the Spirit of Capitalism (Scribner, New York, 1925, p. 61). . . . highly paid labour is generally efficient and therefore not dear labour; a fact which though it is more full of hope for the future of the human race than any other that is known us, will be found to exercise a very complicating influence on the theory of distribution. Alfred Marshall, Principles of Economics, 1920. . . . the landlord who attempted to exact more than his neighbour . . . would render himself so odious, he would be so sure of not obtaining a metayer who was an honest man, that the contract of all the metayers may be considered as identical, at least in each province, and never gives rise to any competition among peasants in search of employment, or any offer to cultivate the soil on cheaper terms than one another. J. C. L. Simonde de Sismondi, Political Economy, 1814, cited in John Stuart Mill, Principles of Political Economy, 1848.

CONVENTIONAL competitive economic basis of observable characteristics (say, theory begins with the hypothesis sex or age) or the (inferences made from) of price-taking firms and consumers, actions undertaken by individuals, e.g., buying and selling homogeneous com- the job for which the individual applies modities at well-defined marketplaces. (George Akerlof 1976), the wage struc- In many situations, these assumptions ture chosen by the individual (Joanne are implausible: in insurance markets, Salop and Steven Salop 1976), or the firms know that some risks are greater quantity of insurance purchased (Michael than others (some individuals' life expec- Rothschild and Joseph Stiglitz 1976). In tancy is greater than others'; some indi- these models, the choices made convey viduals are more likely to have an auto- information; individuals know this, and mobile accident than others), but cannot this affects their actions. tell precisely who is the greater risk, Willingness to trade may itself serve even within fairly narrowly defined risk as a sorting device. In insurance markets, categories. In labor markets, firms know firms have recognized that as the price that some workers are better than others, of insurance increases, the mix of appli- but at the time they hire and train the cants changes adversely. Akerlof (1970) worker, they cannot tell precisely who has shown how such adverse selection will turn out to be the more productive. effects may also arise in other markets, In product markets, consumers know including the market for used cars (see that some commodities are more durable Charles Wilson 1979, 1980). His analysis than others, but at the time of purchase, has subsequently been applied to the la- they cannot ascertain precisely which are bor market (Bruce Greenwald 1986) and more durable. In capital markets, banks to capital markets (Greenwald, Stiglitz, know that the probability of bankruptcy and Andrew Weiss 1984; Stewart Myers differs across loans, but cannot tell pre- and Nicholas S. Majluf 1984). In each cisely which loans are better. of these instances, the uninformed party This heterogeneity has important con- (the seller of insurance, the used car sequences, some of which have long been buyer, etc.) forms rational expectations recognized. There are strong incentives concerning the quality mix of what is be- to sort, to distinguish the high risk from ing offered on the market; the price the low risk, the more able from the less serves as a signal or as a screening device. able. This sorting can be done on the The fact that an employee is willing to

2 Stiglitz: The Dependence of Quality on Price 3

work for $1 an hour suggests that she plications. Firstly, demand curves, may knows of no better offers; others who under quite plausible conditions, not be have looked at her have evidently de- downward sloping. When the price of cided that she is worth no more than $1 some security is higher, uninformed buy- an hour. ers may infer that the expected return Insurance firms have also long recog- is higher, and their demand may increase nized that the terms at which they write (Jerry Green 1973; Sanford Grossman insurance contracts may affect the actions and Stiglitz 1976, 1980). An increase in undertaken by individuals; that is, the the wage may so increase the productiv- likelihood of the insured against an event ity of the workers that the demand for occurring is an endogenous variable, labor may actually increase. A decrease which the insurance firm can hope to af- in the price of used cars results in a de- fect. The fact that individuals can under- crease in the average quality of those be- take unobservable actions that affect the ing offered and this may decrease de- likelihood of an accident is referred to mand (Akerlof 1970). In each of these as the moral hazard problem. The term instances, one can think of the change has come to be used to describe a wide in the price as having two effects: a move- range of incentive problems. In particu- ment along a fixed-information demand lar, employers know that the incentives curve, and a shift in the demand curve workers have to work hard may be af- from the change in information (beliefs). fected by the wage paid; and the incen- Secondly, one may not be able to sepa- tives borrowers have to undertake risky rate out neatly the analysis of demand projects may be affected by the interest and supply. Individuals' demands are rate charged on the bank's loan. Thus, based on inferences they are making from the characteristics of what is being traded prices, and these inferences are critically again depend on the price at which the dependent on the nature of the supply trade is consummated, though now be- responses. Thus, any alteration, say, in cause of incentive effects rather than se- the probability distribution of supply lection effects. characteristics will, in general, lead to These are instances in which price an alteration in the demand functions. serves a function in addition to that usu- Thirdly, markets may be thin. At the ally ascribed to it in economic theory: equilibrium price of insurance, some risk It conveys information and affects behav- averse individuals choose not to purchase ior. Quality depends on price. Of course, insurance, even though with perfect in- in standard economic theory, higher- formation, they would have. The mar- quality items will sell at higher prices: ginal person buying insurance is, in ef- Prices depend on quality. But here, be- fect, subsidizing other purchasers; he is liefs about quality, about what it is that not obtaining actuarially fair insurance, is being traded, depend (rationally) on as he would in a (risk neutral) competitive price. I market with full information.2 This dependence of beliefs about quality on price has some fundamental im- 2Akerlof (1970) investigated a case where equilibrium entailed no trade. As the price of cars de- 1 Some 40 years ago, Tibor Scitovsky (1945) wrote creased, supply decreased and demand decreased a brief but important paper discussing the conse- (the deterioration in car quality was so great that quences of the habit of judging quality by price. An- the quality-adjusted price actually increased): Inter- other antecedent of the recent literature is Alvin section occurred only at zero trade. But this is a Klevorick and Roger Alcaly (1970), who explore the special case which arises primarily because of the implications for the traditional theory of consumers' limited incentives to trade in his model. The effect behavior. he noted in the used car market is the same as had 4 Journal of Economic Literature, Vol. XXV (March 1987)

These models, while departing from mation. Finally, Part IV discusses some the traditional paradigm in denying the of the more important applications of the validity of the hypothesis of homoge- theory, including those in macroeconom- neous markets, retain the assumption of ics (Part IV.1) and in development eco- price-taking firms and consumers. In nomics (Part IV.2). many situations, firms are not price or wage takers. Banks do not simply take I. The Fundamental Implications of the the interest rate that they charge on loans Dependence of Quality on Price as given. 1. Repeal of the Law of Supply and This paper is concerned with situations Demand where firms not only recognize the dependence of quality on price (of pro- No law in economics has such stand- ductivity on wages, of default probability ing as the "Law of Supply and Demand." on the interest rate charged), but also There is an old joke about being able to attempt to use what control they have teach a parrot to be an economist-and over price (wages, interest rates) to in- a good economist at that-simply by crease their profits. The recognition of teaching it to repeat the words "demand this possibility has important implica- and supply." The law holds that competi- tions for economic theory, which have tive market equilibrium is characterized recently been explored in a large number by demland equaling supply. It asserts of papers in several disparate fields. The that the way to analyze changes in market objective of this paper is to survey these equilibrium is to isolate the changes in papers and to draw out the central the demand function and in the supply themes of this literature. function. This paper is divided into four parts When quality depends on price, mar- In Part I, we discuss the most important ket equilibrium may be characterized by implications of the dependence of quality demand not equaling supply.3 on price for competitive equilibrium the- Consider the labor market. Assume ory-the repeal of the law of supply and there is an excess supply of labor. The demand (Part I. 1), the repeal of the law conventional story is that in the face of of the single price (Part 1.2), the exis- excess supply, unemployed workers go tence of discriminatory equilibria (Part to potential employers and offer their ser- 1.3), the comparative static consequences vices at lower wages. The wage is bid (Part 1.4), and the inefficiency of market down. As the wage decreases, the de- equilibria (Part 1.5). Part II discusses al- mand for labor increases and the supply ternative explanations for the depen- decreases. This process continues until dence of quality on price in labor, capital, the wage is bid down to the level where and product markets. Part III explains supply equals demand. more precisely how these models differ If, however, the firm believes that the from standard competitive models and workers who offer their services at a from other models with imperfect infor- lower wage are less productive, then- if they are sufficiently less productive- long been recognized in insurance markets. As the price of insurance increases, the adverse selection 3And as we have already noted (p. 2), because effect implies that the mix of applicants changes ad- the inferences that can be drawn from observing a versely; thus the premium required for the insurance given price-and hence the demand at a given firms to break even must increase. Nonetheless, price-depend on the nature of supply, it may not there may exist insurance markets in which trade be possible in some cases to isolate demand and sup- occurs. ply disturbances. Stiglitz: The Dependence of Quality on Price 5

Cost per Efficiency Productivity / Unit // /

/ I slope X/w / / I Inverse of / / @ wage costs / / I per efficiency / I unit

w* Wage

w* Wage -increases as w increases to w*, and Cost per efficiency unit is minimized at wage w* w then decreases. Figure 1 Figure 2. Wage Productivity Curve it will not hire the lower-wage workers, wage. The curve depicted in Figure 2 for the cost per effective unit of labor shows X, the productivity of the worker, service will actually be higher with the increasing with the wage; the essential lower-wage worker. feature of this productivity curve is that Thus, in Figure 1 we have depicted there is initially a region of increasing the cost per effective unit of labor service returns, where an increase in the wage as initially falling as the wage is in- leads to a more than proportionate increased. There is a wage, w*, at which crease in the productivity. Many of our wage costs per effective unit of labor ser- results depend critically on the existence vice are minimized. This is referred to of some region for which this is true; in as the efficiency wage. For this section, the models that we investigate in detail, we assume the same costs per efficiency we establish that this is in fact the case. unit schedule faces all firms and char- In Figure 2, the cost per efficiency unit acterizes all workers. (In the following (cost per effective unit of labor) is given sections we shall consider the conse- by the (inverse of the) slope of a line quences of differing costs per efficiency through the origin to a point on the wage unit schedules.) productivity curve. Because the slope is If, at the efficiency wage, there is an increasing as we increase the wage from excess supply of laborers, no firm has an O to w*, the cost per effective unit of incentive to lower its wage, or to hire a labor is decreasing. Beyond w* the slope worker who offers his services at a lower decreases and hence the cost per effec- wage, for to do so simply increases labor tive unit of labor increases. It is clear, costs. Thus, when at w*, the supply of then, that the wage-productivity curve labor equals or exceeds the demand, w* generates a cost per effective unit of la- is the equilibrium wage. bor curve of the form depicted in Fig- This curve, giving the relationship be- ure 1. tween the cost per effective unit of labor The existence of wage-productivity re- and the wage rate, is derived from the lationship has long been recognized, as more fundamental wage productivity the quotations at the beginning of this curve, giving the productivity of the rep- article indicate. The more recent revival resentative worker hired at a particular of interest may be attributed to Harvey 6 Journal of Economic Literature, Vol. XXV (March 1987)

p = Expected Return per 1/Success Dollar Probability Lent

slope 1)(1 + r)p

r* Interest Rate Expected return per dollar lent is maximized 1 + r* 1 + r at r*. p(l + r) increases as r increases to r*, and then decreases. Figure 3 Figure 4. Default Curve Leibenstein (1957) who discussed it in the context of less developed countries bank will refuse to lend to anyone offer- and the subsequent developments by ing to borrow at a higher interest rate; James Mirrlees (1975) and Stiglitz its expected return would be lower than (1976b). There, however, the relation- what it obtains by lending at r*. ship is based on nutritional considera- The curve in Figure 3 may be derived tions.4 Near subsistence, workers are not from the default curve. The probability very productive; increases in wages may of a default is postulated to increase with lead to marked increases in efficiency. the rate of interest charged. Let p = Though our analysis focuses on alterna- probability that the loan is repaid; we tive explanations of the relationship, the assume, for simplicity, that when the consequences are very similar. loan is not repaid the lender receives Exactly the same analysis applies to nothing. Thus, the expected return to a the capital market (Stiglitz and Weiss lender is 1981). Assume that as the bank increases = + the interest rate, the "quality" of those p p(l r). who apply decreases; that is, those who In Figure 4, we have plotted 1/p as a apply have, on average, a higher proba- function of 1 + r. Thus, the slope of any bility of defaulting, of not repaying their line from the origin to the default curve loans. The safest borrowers are unwilling is 1/(1 + r)p. The bank wishes to maxi- to borrow at high interest rates. Then, mize p(l + r), i.e., find the point on the expected return on a loan, p, may the default curve with the lowest slope; actually decrease as the interest rate, r, this is clearly just the point of tangency increases, as depicted in Figure 3; r* is of a line through the origin with the de- the efficiency interest rate. If, at r*, there fault curve. Note that the slope decreases is an excess demand for loans (credit is as r increases up to 1 + r* (i.e., p in- rationed), r* is still an equilibrium. The creases) and decreases thereafter (i.e., p decreases), just as depicted in Figure 3. 4For discussionsof the validityof this relationship, (We postpone until later a justification see Christopher Bliss and Nicholas Stern (1978, 1981). See also P. H. Prasad(1970), G. B. Rodgers for the shape of the default curve.) (1975), and Dasgupta anidRay (1986b). Again, the conventional story for why Stiglitz: The Dependence of Quality on Price 7

there cannot be credit rationing in equi- ing nonprice terms7 of the contract x, librium is that those who are willing to where q(p, x) represents a description borrow at the given interest but denied of the (expected value of the) quality of credit go to the bank and offer it a higher the object (service) being purchased, a interest rate; this bids up the interest function of p and x, then the individual rate. As the interest rate is bid up, the chooses p and x to supply of credit is increased and demand Max U (1) decreases. The process continues until equilibrium is reached at the point where (p, x) the demand for loans equals the supply; subject to the constraint of being able there is no credit rationing. But now, to obtain the item, i.e., if V[p, x, q(p, the bank realizes that if it charges a x)] is the (expected) utility of an individ- higher interest rate, the probability of ual (firm) selling an item of quality q and default increases; and it may increase so V* represents the reservation utility much that the expected return to the level,8 then bank actually decreases. Thus, no bank will ever charge an interest rate above V[p, x, q(p, x)] ' V*. (2) r*.5,6 In a variety of circumstances, the solution In each of these cases, the story is the to this problem may entail the constraint same: Because quality (labor efficiency, (2) not being binding, at least for some bankruptcy probability) changes as the potential sellers.9 When that happens, price (wage, interest rate) changes, ex- market equilibrium will be characterized cess supply or demand may persist with- by demand not equaling supply. 10 out any tendency for price (wages, inter- 2. Repeal Qf the Law of the Single Price est rates) to move to correct the market imbalance. The law of supply and demand is of General Formulation. In each of the course, only one of the fundamental instances described above, an individual "laws" of conventional economics. An- (firm) sets the terms of the contract with other central aspect of the traditional par- another individual or firm to maximize its utility (profits), subject to offering 7Nonprice terms in credit markets include collat- that make the eral requirements, specifications of circumstances terms contract acceptable; under which credit is terminated, etc. under the circumstances described, the 8The reservation utility level also may vary across optimal contract will be such that the sellers. constraint on is not 9In the incentive (moral hazard) versions, q is a acceptability binding. result of an action taken by the individual; the action That is, if U[p, x, q(p, x)] is the utility taken is a function of the terms of the contract. In of the buyer, paying a price p, and offer- the selection version, q is the average quality of those offering goods (services) at the price p and x; the constraint (2) is to be read as saying that at least 5A cartoon appeared in several newspapers in the one item is offered for sale at the given terms. In early eighties, when interest rates were soaring, in general, in adverse selection models, the constraint which a banker is seen leaning over his desk, asking (2) is binding for only some individuals offering to the loan applicant, "What kind of person would be sell the commodity. willing to borrow at the interest rates we charge?" 10It is worth noting that while (2) may be viewed, 6 Note too that in the circumstances with which from the perspective of the buyer, as what has come we have just been concerned, where quality depends to be called the "individual rationality constraint," on price, the quantity demanded at any price de- V* is in general itself endogenously determined (it pends on the quality supplied: Thus, for instance, a represents the individual's best alternative opportu- change in the mix of loan applicants will affect the nity); and in the adverse selection models, (2) also supply of loans available at any interest rate; a change can be viewed as a self-selection constraint, differen- in the mix of job applicants will affect the demand tiating between those for whom the job (loan) is ac- for labor at any wage. ceptable and those for whom it is not. 8 Journal of Economic Literature, Vol. XXV (March 1987)

adigm is codified in the "Law of the Sin- gle Price." This law holds that all objects with the same observable characteristics should sell at the same price. When there is a relationship between quality and x~~~~~~ price, the price itself becomes a relevant characteristic; market equilibrium may be characterized by a price distribution (or a wage distribution, or a distribution of interest rates) for objects that cannot be distinguished (before purchase) other than by price. It has been widely observed that some firms have a high-wage policy; others pursue a low-wage policy. It is important w to bear in mind that the differences we The wage-productivity relationship may differ are concerned with here are not differ- across firms. Different firms may find it optimal ences in whether the firm hires those to pay different wages. with more education or less education, Figure 5 those with more work experience or those with less work experience. Rather, Wage Distributions When Costs per we are comparing the wages paid by Effective Unit Are Not Monotonic. Wage firms for workers with a given set of ob- (price, or interest rate) distributions also servable qualifications. arise when costs per effective unit are There are several reasons that econo- not monotone, as we suggested that they mies in which productivity depends on might not be. wages (or more generally quality on Assume, in particular, that the reason prices) may be characterized by a wage that productivity increases with wage is (price) distribution. nutritional;1" though productivity always Differences in Firms. If the wage- increases with wage, it may increase productivity relationship differs across more or less than proportionately. When firms, as in Figure 5, the efficiency wage it increases more than proportionately, may differ. More generally, firms where labor costs per efficiency unit decline net productivity is more sensitive to with wage increases. Thus, if over some wages (with higher turnover costs, higher ranges, say, for very low wages and for monitoring costs, or where shirking an intermediate range of wages, pro- workers can do more damage) will find ductivity increases more than proportion- it desirable to pay higher wages for work- ately with wages, one would obtain a ers of identical characteristics (Steven curve describing wage per efficiency unit Salop 1973). This is consistent with the as in Figure 6. observation that more capital intensive We define the Walrasian wage, tw, as firms tend to pay higher wages, because that wage at which demand for labor the "damage" a worker can do in such jobs may be higher. If monitoring costs " We choose this example because in more general are higher in firms employing large num- models, with adverse selection or incentive effects, bers of workers, one might expect to see the productivity of a worker at one firm depends on the wages paid at other firms and on the unem- such firms paying higher wages (other ployment rate. We wish to avoid these complications things being constant). in this simple exposition. Stiglitz: The Dependence of Quality on Price 9

Cost per Cost per Effective Effective Unit of Unit of Labor Labor

(wlA)** ___S\ (wlA)* __I_ --l W l l

l l l I WI W2 W l l l I I I I Wage distribution where cost per effective unit l l l I of labor is the same at w, and w2. l l l I Figure 7 Market equilibrium is characterized by excess supply of labor whenever Walrasian wage is lower tion of workers hired at wband w** at than w*. When Walrasian wage is between w and which demand is equal to supply for the w some workers are hired at w**, and other low-wage jobs, though there is excess workers are hired at wt. supply of labor at the high-wage firms. Figure 6. Wage Distribution Similar arguments apply to other markets. equals the supply. However, the Walra- Wage Distributions When Costs per sian wage is not the equilibrium wage Effective Unit Are Minimized at Several whenever there exists a wage higher than Different Wages. The previous subsec- the Walrasian wage with a lower cost per tion considered a situation where the cost efficiency unit. For it would pay any firm per effective unit was not monotonic. to increase its wage. Assume, in Figure Equilibrium wage distributions also arise 6, that the Walrasian equilibrium oc- when the curve describing the cost per curred at some wage between wt and effective unit of labor, depicted earlier, w**. One might be tempted to suggest that has several peaks, and all of the peaks w**, the wage greater than the Walrasian generate exactly the same level of costs wage at which costs per efficiency unit per effective unit of labor, as illustrated are minimized, is the market equilib- in Figure 7. Such a configuration would rium; but at w**, there is unemploy- seem to be an anomaly. Even if there is ment; and by lowering their wage enough no reason that the productivity curve has (to any wage between wt and w*), workers a single peak, why should the peaks occur can be hired with a cost per efficiency at the same level? One can show that unit that is lower than at w**. The equi- this may, in fact, occur quite easily. The librium now entails full employment with curve that we have depicted is the curve a wage distribution; some workers are facing a particular firm, given the wages hired at w** and others at wt. Costs per paid by all other firms. Stiglitz (1974b, efficiency unit are exactly the same. If 1985) shows, in the context of the labor- all workers were paid a wage of wt, there turnover model, that there is a wages would be (by assumption) an excess de- distribution (and in fact many such distri- mand for labor; that is why wt is not an butions) with the property that the prod- equilibrium. If all workers were paid uctivity curve has many peaks, each of w**, there would be an excess supply which yields exactly the same cost per of laborers. There is a particular propor- effective unit of labor. The reason for this 10 Journal of Economic Literature, Vol. XXV (March 1987)

is that the turnover costs facing any firm C are a function of the fraction of firms pay- Cost per ing a higher wage, which itself is an en- Effective B Unit of dogenous variable. The low-wage firm Labor A faces a higher turnover. The fraction of high-wage firms is such that the total labor costs-wages plus turnover costs- I I are the same at low-wage firms and high- l I l I l wage firms (see also Phillip Dybvig and I Gerald Jaynes 1980). I I

Similar results can be obtained in the wB UA W context of selection models. If different Group C is excluded from labor market. individuals have different costs associated with queuing, or with not selling their Figure 8 commodity (their labor), then high-wage firms may find that they face a longer and Stiglitz 1982) has not provided an queue and a higher quality of applicants. explanation of job discrimination or red In these models, high-wage labor looks lining-the differential access to certain (in terms of observable characteristics) jobs or credit of certain groups. The the- the same as low-wage labor, but in fact ories we are concerned with here do it is more productive with differences in this. productivity corresponding (in equilib- Assume that in the labor market there rium) precisely to differences in wages. are a number of different identifiable Wages together with queues are acting groups, each with its own "cost per effec- as self-selection devices. 12 (See Barry tive unit of labor" curve, as depicted in Nalebuff and Stiglitz 1982 and Stiglitz Figure 8. Then, equilibrium will be char- 1976a.) acterized by all groups hired having the same cost per effective labor unit; those 3. Discrimination with lower costs per effective labor ser- It has long been noted that, in the vice at any wage are paid a higher wage. absence of perfect information, there (There is, thus, wage discrimination.) may be statistical discrimination: Mem- But there are some groups, such as bers of a group will be paid the mean Group C in Figure 8, for whom the cost marginal product of the group, or will per labor service, at the optimum wage, be charged an interest rate correspond- and hence at any wage, exceeds that of ing to the mean default probability of the market equilibrium; such workers their group or charged an insurance pre- will not be hired. "Discrimination" takes mium corresponding to the mean proba- the form not of paying the workers in bility of accident, illness, or death of their group C a lower wage, but of refusing group. But the traditional theory of statis- to hire them. tical discrimination (see, e. g., Dennis While all of those in category A will Aigner and Glen Cain 1977 or Rothschild be hired, and none of those in category C, in general only some of those in cate- 12 Note, however, that the length of queue is not gory B will be hired. Other apparently chosen by the firm, but is an endogenous property identical workers will not be employed. of the equilibrium. In this respect, these models dif- When productivity depends on the un- fer in an important way from those in which the terms of the contract are determined either by the employment rate, there may be large dif- uninformed agent or the informed agent. ferences in equilibrium in group specific Stiglitz: The Dependence of Quality on Price 11

unemployment rates as well as wage rates. 13 The theory not only predicts that some groups may be rationed out of the market, but also suggests that the brunt of changes in the demand for labor (aggre- p~~~~~~~~ gate demand) will be felt by some groups in increased job rationing. Moreover, some versions of the theory predict which groups will be rationed out of the A market. In particular, in the incentive B version of the model (Carl Shapiro and Stiglitz 1984), the wage that must be paid rA rB r to a worker to induce him not to shirk Group C is excluded from capital market. depends on the cost of being fired. This 9 cost is likely to be lower for part-time Figure workers, for those close to retirement, and for secondary participants in the la- tential borrowers is called red-lining. bor force (e.g., low-wage workers with (Some of those in category B, where the high-income spouses). maximal expected return is equal to p*, Exactly the same phenomenon can oc- receive loans, while others do not.) cur in credit markets. Assume that the 4. Comparative Statics function relating the expected return to the interest rate differs across individuals With "normally" shaped demand and who differ in certain observable charac- supply curves, an increase in supply (that teristics, as depicted in Figure 9. There is, a shift in the supply curve such that will be an equilibrium rate of return on at every price a greater quantity of the a loan; loans to all categories that are of- good is supplied) leads to an equilibrium fered yield the same expected return. 14 with a lower price and a greater quantity The interest rate charged for categories traded. Similarly, a decrease in demand of loans, such as A, whose maximal ex- (that is, a shift in the demand curve such pected return exceeds p*, are charged that at every price a smaller quantity of lower interest rates (rA < rB). On the the good is demanded) leads to an equi- other hand, loan category C is such that librium with a lower price and a lower at no interest rate does it yield a return quantity traded. Now, in the class of equal to p*; all those in this category are models being examined in this paper, the simply denied loans. The practice of de- major effect of such a shift may be a nying credit to certain categories of po- change in the magnitude of rationing, with little effect on prices. Thus, in the labor market, when there is an equilib- "3If the productivity of group Xi is a function of rium with excess supply of labor, a small its wage wt and its unemployment rate, Ut then in increase in the demand for labor has no equilibrium, for all groups hired (for which Ui < 1) effect on the wage rate, but increases em- Xi(wi, UN)/Wi ployment. An increase in the supply of is the same, and credit available may have no effect on Xi(wi, Ui)lwi = Xi(wi, Ui)lwi. interest rates, but may simply lead to 14 Assuming that bankers are risk neutral, and thus more loans (less credit rationing) at the care only about the expected return of the loan. old interest rates (Figure 10). 12 Journal of Economic Literature, Vol. XXV (March 1987)

D' D s S S

D ,, S T~~~~~~~~I I A B C Labor A B C labor leave An increase in the demand for may An increase in the supply of credit may leave increase wage rate unchanged, but simply interest rate unchanged, but simply decrease employment. extent of credit rationing. Figure lOa Figure lOb

I B In markets where there are many would be reduced, and no group of work- groups in the population,a decrease in ers would be excluded (though some the demandfor labormay result in some groups of workers may decide to drop groups being completely excluded from out of the labor market) (Figure lla). the market, with only a slight lowering Similarly, in markets with credit ra- of the wage of those remaining.This is tioning, a decrease in the supply of avail- distinctlydifferent from the conventional able credit may result in some groups models, in whichthe wage for all workers being excluded from the credit market,

(w/X)~ ~ ~ ~~~

A decrease in the demand for labor may lead to ,-B A a decrease in the equilibrium cost per effective unit of labor, resulting in exclusion of some groups A decrease in supply of credit may lead to an from the labor market. (When wage per efficiency increase in the equilibrium return, resulting in unit falls from just above (w/A)* to just below it, the exclusion of some groups from the credit group B is excluded.) market. Figure la Figure llb Stiglitz: The Dependence of Quality on Price 13

w Supplyof labor w** _ _

Old demand

New demand

L A technological change may lead to a higher wage and less employment. Figure 12b

w* w** w A change in technology may result in a large the changed expectations affect the de- change in wages, but a relatively small change mand for in the cost per efficiency unit of labor. credit. But there are no clear implications for the interest rate charged Figure 12a by the bank; it may either increase or decrease (Figure 13).16 with relatively small increases in the in- 5. terest rates charged to those who are not Implications for Welfare Economics excluded (Figure llb). 15 One of the crowning achievements Shifts in the Productivity Curves. The of competitive equilibrium theory during other important source of changes in the the past half century has been the proof equilibrium comes from changes in the of the fundamental theorem of welfare productivity (or return) curves. In the economics-providing a precise state- labor market, a change in technology may ment of the meaning of Adam Smith's result in a change in the efficiency wage, invisible hand conjecture. as depicted in Figure 12a. If the effi- An implicit assumption in the standard ciency wage increases markedly, but the proofs of the Fundamental Theorem of cost per effective unit of labor decreases Welfare Economics is that there is per- only slightly, as in Figure 12b, then the fect information. 17 Obviously, economies new equilibrium may be characterized by a lower level of employment and a higher wage. 16In the simple models presented here, only one group in the populationis rationed;those above the Similarly, if banks' expectations of de- critical cutoff level receive all the credit they wish, faults increase, then the expected return, or can sell all the labor they wish, while others are at any interest rate, will decrease. This completely excluded from the market. But it is easy to construct models in which rationing is extended will normally result in a decrease in the to many groups. See Barry Nalebuff and Stiglitz supply of credit; the effect on the extent (1982) and Stiglitz and Weiss (1986a, 1986b, 1987). of credit rationing will on how 17 The analysescan be extended to situationswhere depend there is uncertainty, provided there is a complete set of risk markets, and provided that the nature of '5Indeed, in some circumstances a decrease in the commodity being traded (the quality of labor, supply may actually lead to a decrease in the the probabilityof bankruptcy,etc.) does not change weighted mean interest rate charged. See Stiglitz as prices (wages, interest rates)change. It is the latter and Weiss (1986a, 1986b). which concerns us here. 14 Journal of Economic Literature, Vol. XXV (March 1987)

Supply with greater pessimism

supply

r** - y9 Demand

Quantity r More pessimistic expectations may result in more More pessimistic expectations may lower expected credit rationing and lower interest rates. return and lower interest rate charged. Figure 13b Figure 13a with perfect information are likely to provement can be attained by subsidiz- function better than economies with im- ing commodities whose consumption in- perfect information:'8 That is an irrele- creases the mean quality of labor sold vant comparison. The relevant question on the market and by taxing commodities is, in a market characterized by incom- whose consumption decreases the mean plete information, are there interven- quality. The mean quality of labor offered tions that can attain a Pareto improve- on the market resembles the mean qual- ment? Is the market, in other words, ity of air: There is an important external- constrained Pareto efficient, taking into ity in each individual's decision. account the imperfections of information Their earlier paper was limited to and costs of obtaining more information? economies with market clearing. In a se- Greenwald and Stiglitz (1986a) have quel (Greenwald and Stiglitz 1986b), shown that competitive economies with they identify the inefficiencies that arise incomplete markets and/or imperfect in- in economies in which quality is depen- formation are essentially never con- dent on prices and markets do not clear. strained Pareto efficient. These authors There are inefficiencies both in setting develop a general framework within wages (interest rates) and determining which a variety of informational imper- the level of employment (the number of fections can be analyzed. They limit loans of different types). themselves to tax and subsidy interven- A direct consequence of the Funda- tions. They apply their analysis to one mental Theorem of Welfare Economics of the models that are the subject of dis- is the decentralizability of efficient re- cussion in this paper, namely, the Akerlof source allocations; and a direct conse- adverse selection model, where the qual- quence of the failure of the Fundamental ity offered is a function of the price but Theorem-of the externalities we have markets clear. In the context of the labor noted-is that the scope for decentraliza- market, they show that a welfare im- tion may be limited. 19

" This section focuses on the welfare economics 9One should perhapsdistinguish between two sit- of the imperfect information models. For a discussion uations:that where the decentralizedallocation with- of the welfare economics of the nutrition-based effi- out government intervention is inefficient, but ciency wage models, see Partha Dasgupta and Debraj government intervention, in the form of taxes and Ray (1986a). subsidies, can induce a decentralizedefficient alloca- Stiglitz: The Dependence of Quality on Price 15

Some rationing may be consistent with private market allocations and there may Pareto efficiency. Assume the govern- be other grounds for government inter- ment has no more information about the vention. Taxes and subsidies may affect quality of workers or potential borrowers the consumption vector of individuals, than do firms (or banks). It has to allocate and thus indirectly the effort exerted by workers to different jobs. It has to allo- the worker (or the wage required to in- cate capital among different firms. The duce the worker not to shirk). Such taxes wage at which workers are willing to and subsidies may thus be welfare en- work or the interest rate at which firms hancing (Arnott and Stiglitz 1985). are willing to borrow conveys informa- Shapiro and Stiglitz (1984) point out a tion to the government, just as it does variety of other inefficiencies that arise to the employer or the bank, and it will, in their model where high wages are used in general, wish to use this information, to reduce shirking. These relate to the even though to use it necessitates ration- intensity of monitoring and policies that ing-unemployment or credit rationing. affect quits and shirking. For instance, But the objectives of the government, if workers are very risk averse, unem- what it wishes to glean from the informa- ployment insurance may be desirable tion, are different from those of firms and even though in their model private firms banks in the private market equilibrium. will never supply unemployment insur- The latter are simply concerned with ance; if there is an excess supply of labor, maximizing their own profits. Thus, in firms can obtain the desired labor force choosing a wage, the government would at the going wage; any increases in unem- be concerned with ascertaining that ployment compensation simply increase those with a comparative advantage in a the wage that a firm must pay to induce job get assigned to that job; the firm is workers not to shirk.20 simply concerned with cost per efficiency These models also have some more ba- unit. (Similarly, the bank is not con- sic implications for how we think about cerned with how its actions affect the the welfare properties of market econo- profitability of investors, only with how mies. One of the important conse- it affects the profitability of the bank.) quences of the Fundamental Theorem of Private firms will not only set the wage Welfare Economics was that it enabled incorrectly; they will also hire an incor- a neat separation of efficiency and equity rect number of workers. Because all issues. In particular, whether the econ- workers whose opportunity cost is less omy is or is not Pareto efficient did not than the wage offer themselves for work, depend on the distribution of wealth. In the mean opportunity cost of an individ- the models under examination, this is not ual randomly hired by a firm is less than true, for two reasons. First, the ineffi- the wage. Thus firms will not hire work- ciencies with which we have been con- ers up to the point where the wage equals cerned arise because of the asymmet- the mean opportunity costs, as a govern- ric information between two parties to ment concerned with maximizing na- a transaction-between landlord and tional income would. worker, or between worker and capital- The inefficiencies we noted above are ist. But whether these transactions arise not the only inefficiencies associated with 20This assumes that the firm cannot treat quits tion; and those where even with this formof interven- and fires differentially. There are good reasons for tion, a decentralizedallocation, cannot attain certain this: If quits are treated preferentially, a worker who Paretoefficient allocations.Both problems may arise knows he is about to be fired will quit. For a more here. extended discussion, see Shapiro and Stiglitz (1985a). 16 Journal of Economic Literature, Vol. XXV (March 1987)

is, at least partly, determined by the dis- 1. Selection Models tribution of wealth. For instance, share- Labor Market. The reasons for the cropping arises largely because of the presence of adverse selection in the labor concentration of wealth. Secondly, the market are clear (Stiglitz 1976b; Weiss distribution of ownership of factors deter- 1976, 1980; Greenwald 1979, 1986). One mines whether it is, in fact, feasible to of the inferences I can make from the design Pareto improvements. In the Sha- fact that a worker is willing to work for piro-Stiglitz model, for instance, national me for 50 cents an hour is that he does income can be increased by taxing capital not have (know of) a better offer and subsidizing wages; if wealth were elsewhere.2' If other firms are screen- equally distributed, the losses individu- ing workers, keeping high-productivity als would suffer qua capitalists would be workers and letting go of low-productiv- more than offset by the gains that would ity workers, or adjusting their wages to be received qua workers. But if there reflect their lower productivity, the fact are two distinct groups in the population, that no other firm has offered the worker capitalists and workers, there may be a wage in excess of 50 cents an hour con- no way of improving the welfare of the veys a considerable amount of informa- workers without simultaneously hurting tion. (Obviously, the inferences I make the capitalists-the market equilibrium, depend on a number of details of the while not maximizing net national prod- surrounding circumstances; if the worker uct, is Pareto efficient (see Shapiro and is a newly arrived worker to the United Stiglitz 1984; Dasgupta and Ray 1986). States, the fact that he is willing to accept a job at 50 cents an hour is more likely II. Explanations for the Dependence of to reflect his limited opportunities for Quality on Price search and his lack of knowledge of the job market; if a job has a number of ex- tremely attractive nonpecuniary advan- In the introduction, we described two tages, the inferences I make should take broad classes of models giving rise to a that into account.) dependence of quality on price based on More productive workers are likely to incentive and selection effects. In the have better wage offers from other firms, preceding sections, we showed that if the but even if they are self-employed, they dependence of quality on price took on are likely to be more productive (that particular forms (e.g., the cost per effec- is, if those who are more able in one tive unit had an interior minimum) then production task are more able, on aver- there might exist an equilibrium in which age, in other production tasks as well). demand did not equal supply, or in which In the simplest formulation of the ad- there might be, in equilibrium, a price verse selection-labor model, a worker (wage, interest rate) distribution. We whose reservation wage is v has a pro- now consider in more detail not only why ductivity of a(v); if all workers whose res- quality may depend on price in certain ervation wage v is less than or equal to situations, but why the dependence w apply for a job with wage w,22 then should take on a form that may give rise to nonmarket-clearing equilibria and/or wage/price/interest rate distributions. 21Unless a worker's productivity is completely The analysis is divided into three sec- firm-specific, then no information is conveyed to one firm by another firm's refusal to hire him. tions, describing selection models, in- 22 More generally, the reservation wage of an indi- centive models, and nutritional models. vidual will depend on his fallback (self-employment) Stiglitz: The Dependence of Quality on Price 17

the mean productivity of the applicants Mean Productivity is of Applicants X a(v)f(v)/F(v) where F(v) is the distribution function of the population by reservation wages, and f(v) the density function. It is clear a, that if a' > 0, then mean productivity will increase with w. To see that it may have the shape depicted in Figure 1, consider a two group model; low-productiv- V1 V2 W ity individuals with productivity a, have a reservation wage of v1 while high-pro- Figure 14a. Wage Productivity Curve in Two Group Model ductivity individuals with productivity a2, have a reservation wage of V2. a is the mean productivity; obviously, at high likely to be less concerned about the in- wages, when everyone is willing to work, terest rate they have promised to pay a = a. The resulting productivity curve (when they do not default) than those is depicted in Figure 14a. who are undertaking safe projects and A variant of this model arises when will always repay. there are job-specific skills and positive More formally, Stiglitz and Weiss search costs; the larger the applicant (1981) consider a set of projects that the pool, the higher the expected productiv- bank has identified as "similar."They all ity, simply because the firm can find a yield the same mean return, and require larger number of individuals that fit in the same amount of bank finance. They well with the needs of the firm. show (a) the riskier projects23 yield a Capital Market. The intuitive reasons lower return to the bank; (b) at any inter- for the presence of adverse selection in est rate charged by the bank, firms with the capital market follow along parallel riskier projects apply, those with safer lines to those in the labor market (Stiglitz projects do not; and (c) as the bank in- and Weiss 1981). The fact that an individ- creases the interest rate charged there ual is willing to borrow from a bank at a is an adverse selection effect, with firms 25 percent interest rate implies that he with the best projects (the least risky, has found no one else willing to lend to i.e., those yielding the bank the highest him at lower interest rate, and this fact expected return) no longer applying.24 conveys considerable information. More- They show that the adverse selection ef- over, even if the individual had not been fect may outweigh the direct gain from turned down by other banks, those who an increase in the interest rate. This may are undertaking very risky projects, with little prospect of repaying the loan, are 23A project is said to be riskier than another if its return is a mean preserving spread of that of the other. income and the probability that he will obtain a 24The first result is a direct consequence of the higher wage, and this depends on the probability concavity of the payoff function of the bank (in the distributionof abilities and wage offersin the popula- absence of collateral, this is max [R, (1 + r)B], where tion, on the nature of the search technology, and R is the return to the project, (1 + r)B the amount on the costs of quitting a job once it has been ac- the firm has promised to repay); the second result cepted. See Nalebuff and Stiglitz (1982). Schlicht is a direct consequence of the convexity of the payoff (1986) argues for a link between reservation wage function of the borrower (min [R - (1 + r)B, 0] in and productivitybased on the shorterexpected dura- the absence of collateral); the third result is an imme- tion of holding a job by less competent individuals. diate consequence of the second. 18 Journal of Economic Literature, Vol. XXV (March 1987) be seen most easily in the case where Expected there are only two kinds of firms; each Return project costs B dollars, and is entirely

bank financed. The firm is required to [(1 + r)-C]+C put up c dollars worth of collateral per dollar loaned, which it forfeits in the event of a default. The projects of type /2 [(I + r)-C]+C i firm yield a return of R' if successful, and nothing otherwise; the probability of success is pi. The expected return to 1+ r* 1+ r the bank per dollar loaned from a loan Figure 14b. Expected Return as Function of to a firm of type i is just pi (1 + r) + Interest Rate Change: Two Group Model (1 - pi)c, where r is the interest rate charged. Thus, if pl > p2, one is the safe project and yields a higher return to the bank, provided c < (1 + r), which it ship. For the threat of termination to be always would be (otherwise, the bank an effective incentive, the terms of the faces no risk). contract (relationship) must be such as The expected return to a firm of type to make that contract strictly better than i is p' [Ri - B(1 + r)B] - cB(I - p'). It that of the next best opportunity; e.g., follows then that if the two projects have price must be in excess of marginal cost the same mean return, the riskier project or the wage paid must be higher than yields the higher return to the firm. The the minimum wage required to recruit safe firms no longer apply when the worker. We now describe in greater detail the - + - cB( - = 0. pl[Rl (1 r)B] p') effects of wages, prices, or interest rates Thus, for interest rates below [p'1Rl - on economic incentives. cB(l - pl)]/p'B - 1 = r*, both types Capital Markets. An increase in the of firm apply; for higher interest rates, interest rate charged on a loan may in- only the risky firms apply. Even if at r* duce the borrower to undertake greater there is an excess demand for funds, risks, lowering the expected return to the banks may not raise the interest rate (see lender (William R. Keeton 1980; Stiglitz Figure 14b). and Weiss 1981). This may be seen most 2. Incentive Effects simply in the case where the firm has two projects that it can undertake, each We are concerned here with the vari- of which costs B. As before, project i has ety of circumstances under which the a return, if successful, of R', and the quality of the product (the likelihood of probability of success is pi; if unsuccess- bankruptcy, the productivity of a worker) ful, the project yields no return. For sim- is affected by actions of the seller (bor- plicity, we ignore collateral. Then, the rower, worker), and those actions are af- expected return to the firm of undertak- fected by the price (interest rate, wage). ing project i when the interest rate is r Two broad categories of models will be is on whether there discussed, depending - + is or is not a long-term relationship. The p[R' (1 r)B]. latter are of particular interest; bad per- As Figure 15 illustrates, the safe project formance is (in these models) often pun- 1 has a higher expected return to the ished by a termination of the relation- firm for r ' r*, where Stiglitz: The Dependence of Quality on Price 19

Expected Return pl [R -(1 + r)B]

I \ ~~~~~~~~~p2[R2 (I + r BI

1 + r* 1 + r Firm chooses safe project if and only if r ' r*. Figure 15a

that the threat of terminating the credit (1 + r*)B = (p'R' - p2R2)/(p' - p2). relationship may have beneficial incen- Thuis, the bank will not raise the interest tive effects and that terminations are a rate above r*, even if there is an excess better incentive device than threats to demand for funds at r*, because to do charge higher interest rates (pay lower so would induce firms to undertake the wages). risky project, lowering the bank's return. Their analysis is similar to the explana- This is an example of what has come tion of why the expected return may de- to be called a principal agent problem.25 crease with the rate of interest charged The principal (here the bank) can exert provided by the models of Jonathan Ea- only indirect control over the actions of ton and Mark Gersovitz (1980, 1981a, the agent (the firm) and, he does so 1981b), Eaton (1985), and Franklin Allen through the design of the payoff schedule. Changing the nominal price (here the interest rate) may have adverse (in p terms of the interests of the principal) effects on the actions undertaken by the agent. The discussion so far has focused on a single period model. Stiglitz and Weiss pl (1 + r) (1983) have extended the analysis to a multi (two) period model. They establish

25A vast literaturehas developed on the principal agent problem since the early papers by Steve Ross (1973), Mirrlees (1974), and Stiglitz (1974d). We 1+ r* 1+ r make note of only those papers that are directly rele- Expected returns are maximized at r*. vant to the subject of this review, the dependence of quality on price. Figure 15b 20 Journal of Economic Literature, Vol. XXV (March 1987)

(1980, 1981, 1983). They are concerned Steve Stoft 1982.) If there were no unem- with situations where contracts are not ployment and if all firms paid the market- enforceable. We might say that there are clearing wage, then the threat of being implicit contracts, which have to be de- fired would not lead individuals to reduce signed to be self-enforcing. The terms their shirking: they would know that they of the contract determine whether, and could costlessly obtain another job. But under what circumstances, it will pay a if the firm pays wages in excess of that borrower to refuse to repay a loan (for a of other firms, or if there is unemploy- sovereign to repudiate his debt).26 ment (so that a fired worker must spend In these models, it is again the threat a period in the unemployment pool be- of termination of the relationship that in- fore he again obtains a job) then workers duces the borrower to repay the loan. have an incentive not to shirk; there is (In the Eaton-Gersovitz model, access to a real cost to being fired.29 One of the international capital markets enables the immediate consequences, then, of costly country to smooth out income vari- monitoring is that equilibrium must be ability.) For any given amount loaned, characterized by unemployment and/or the higher the rate of interest charged, wage dispersion. the more likely it is that the loan will It also implies that the productivity of be repudiated.27 the worker hired by the ith firm, Xi, is Labor Markets. There are several dif- a function of the wage it pays, the wage ferent explanations for why the wage paid by all other firms, w-i, and the un- might affect the productivity of workers employment rate, U: and the profitability of the firm. (a) Shirking. (Walter J. Wessels 1979, Xi= Xi(wi, W_i, U). 1985; Guillermo Calvo 1979; Calvo and Phelps 1977; Calvo and Stanislaw Wellisz In the simplest version of the shirking 1979; Shapiro and Stiglitz 1984; Samuel model, workers either work or shirk; Bowles 1985;28 Stiglitz and Weiss 1983; there is a critical wage below which the worker shirks. This critical wage is an 26in the Stiglitz and Weiss (1983) analysis, con- increasing function of the employment tracts are explicit, but enforceable only if it is in level or of the wage differential between the interests of at least one party to the contract to and In the case do so. In the papers discussed in this paragraph, this firm other firms. there is no legal enforcement mechanism. "Moral where all firms pay the same wages, the hazard"issues arise both in the compliance with the so-called no-shirking constraint, giving contract, and in the actions borrowers take which the minimum wage below which shirking affect the likelihood they will wish to comply. 27 Repudiationcould always be avoided if the re- occurs, is depicted in Figure 16a. The payments could be made state dependent, and if all demand for labor, given that workers do states were perfectly observable and verifiable by both parties. For a more general discussion, see Ea- ton, Gersovitz, and Stiglitz (1986). 28There is a large "radical"literature emphasizing 29These models are thus long-term models. The the effect of the employment relationshipon worker "punishment"is provided "publicly"by the period productivity. Other contributions to this literature of unemployment(rather than privately, throughre- include Herbert Gintis and Tsuneo Ishikawa(1985), duced wages or other means). Stiglitz and Weiss Tom Weisskopf, Bowles, and David Gordon (1983), (1983) provide conditions showing that termination Gerry Oster (1980), and Geoff Hodgson (1982). The is in fact the optimal punishment. distinctionbetween this literatureand the nonradical These models are constructed with identical indi- literature is not always readily apparent. For in- vidualsso there is no reputationeffect associatedwith stance, the Bowles (1985) and Shapiro and Stiglitz being fired. In any case.,it is often hard to distinguish (1984)models appearto be essentiallyidentical; some among voluntary and involuntary termination, in of the interpretationsgiven to the model, and the which case, the reputation effect of separationsmay lessons drawn, differ. be minimal (Shapiroand Stiglitz 1985a). Stiglitz: The Dependence of Quality on Price 21

Demand for labor

Wage

No shirking constraint Full Employment Unemployment

Employment High wages and unemployment are necessary to induce workers not to shirk. Figure 16a. Shirking Model not shirk, is a decreasing function of the form of higher wages)3' then the greater wage. Equilibrium, at the intersection the quit rate, the greater the firm's ex- of the demand curve and the no-shirking penditures on training and hiring costs. constraint, always entails unemploy- Increasing the wage rate (relative to ment. wages paid by other firms) will, in gen- (b) Labor turnover. (John Pencavel eral, lead to a reduction in the quit rate; 1972; Wessels 1979; Stiglitz 1974b, there is a wage that minimizes total labor 1974c, 1985; Robert Hall 1975; Salop costs of the firm (Figure 16b). Moreover, 1973, 1979; Dybvig and Jaynes 1980; the greater the unemployment rate, the Ekkehart Schlicht 1978.) A second im- less likely it is that the worker will find portant way that workers' behavior affects a better job. Because the effect of higher the productivity of firrnsis through labor quit rates is to decrease the "net" pro- turnover.30In most jobs, there are costs ductivity (net of turnover costs), we ob- to hiring and training that are specific tain a productivity wage relationship of to the firm. So long as individuals do not the above form. pay their full costs at the moment they "' Richard Arnott and Stiglitz (1985) and Shapiro are hired (recouping them later in the and Stiglitz (1985a)provide explanationsfor why individuals do not bear the full costs of training; these have to do with worker risk aversion, incomplete 30 importanceof this had, of course, long been insurance, and imperfect capital markets (though recognizedby laboreconomists. See SumnerSlichter these market imperfections in turn, can be related (1919). to informationimperfections). 22 Journal of Economic Literature, Vol. XXV (March 1987)

Turnover Costs

Total labor costs = wage and turnover cost

Turnover costs decline with wages

(450)

w* w Totallabor costs are minimizedat w*. Figure 16b. Labor Turnover Model

(c) Morale (James E. Annable effects. The formal analysis of this model follows 1977, 1980, 1985; Stiglitz 1973, 1974a; closely along the lines of that of the shirk- Pencavel 1977; Akerlof 1984; Whiteside ing model.33 (It may, in fact, be possible 1974.) Employers often allege that pay- to derive at least some variants of this ing higher wage results in higher pro- morale model from a standard utility ductivity, and not simply because of the maximizing model, in which utility de- greater penalty associated with being pends not only on effort and wages, but fired. A worker who believes he is being on relative wages.34 Obviously, quit rates treated more than fairly may not only get more job satisfaction from his job, "The observationthat it is individuals'perceptions but also may put out more for his em- of whether they are being treated fairly that affects ployer. And notions of fairness are closely behavior has one important consequence. The pro- how one perceives others of ductivity curve of a group that believes it is being related to treated unfairly will (if our argument is correct) lie similar ability being treated. Thus, we below that of groups of identical abilities that do not can postulate that an individual's efforts believe they are being treated unfairly.Such individ- depend not only on his own wage, but uals either will not be hired, or will be hired at lower wages than someone of the same ability (let alone on the wage of others in his reference someone of the ability that they believe that they group, w, the monitoring intensity, m, have). The fact that they are hired at lower wages reconfirms their belief they are not being treated and the cost of being fired (the unem- fairly. The employer who pays the lower wage does ployment rate):32 not believe he is discriminating,only that he is paying wages that are in accord with productivity, which ei = e(wi, w, m, U). depends both on (statistical projections of) ability and effort. 34Akerlof (1980, 1982) and Schlicht (1981a, 1981b) 32These effects have, of course, long been dis- develop alternative theories of the employer-em- cussed by labor economists. For an early formaliza- ployee relationshipin which psychologicaland socio- tion of the notion of interdependence, see Dan logical considerationslead to a dependence of pro- Hamermesh (1975). ductivity on wages. Akerlof (1984) describes some Stiglitz: The Dependence of Quality on Price 23

too will depend on morale effects: Indi- shirking and who is thereupon fired de- viduals may spend more resources pended on the wage relative to the wage searching for a better job if they believe paid by other firms and the cost of getting that they are being unfairly treated on another job (which depends, in part, on their present job.) the unemployment rate). Similarly, the Sharecropping. (Stiglitz 1974d; David penalty a customer levies on a firm that Newbery and Stiglitz 1979; Avishay has cheated him is to terminate the rela- Braverman and T. N. Srinivasan 1981; tionship; and the penalty associated with Braverman and Stiglitz 1982, 1986; Allen this depends on how high the price is 1985a). It has long been recognized that relative to the production costs (the profit increasing the share provided to the ten- the firm attains from the relationship) and ant worker could have beneficial effects how hard it is to recruit another, similar on his work incentives, and thus could customer. actually increase the receipts of the The essential insights are provided by landlord.35'36 It is possible that at the the following simple model. Assume that share at which the landlord's expected the cost of production for a high-quality income is maximized, there is an excess commodity is ch, for a low-quality com- supply of tenants.37'38 modity is cl, and the price is p. Assume Product Markets. (Joseph Farrell that when a customer observes a bad- 1979, 1980; Shapiro 1983; Dybvig and quality commodity, he infers that the Chester Spatt 1983; Allen 1984; Ben seller will always continue to sell a bad Klein and Keith B. Leffler 1981). quality. The bad quality commodity is In the product market, incentive ef- worth nothing to the customer, so if he fects similar to those described earlier believes that the commodity is bad, he arise: Buyers often cannot ascertain the will terminate the relationship. Then, quality of a commodity before they pur- the value to the firm of continuing to chase it. Earlier, we noted that the pen- produce good-quality commodities is alty associated with a worker caught (1 + r) (p - Ch)Qlr experiments in which differentworkers are assigned where Q is the quantity sold, r the rate identical jobs and paid identical wages, but believe of interest. The value of cheating and that they are either being under-or overpaid. These perceptions affect productivity. producing a bad quality commodity is 35Obviously, if the landlordcould monitor the ac- - tions of the tenant, the share contractwould specify (p c)Q. the level of workrequired, and these incentive effects would not arise. See Steve Cheung (1969). Thus, for it to be worthwhile for the firm 36Braverman and Srinivasan(1981) have identified to produce good-quality commodities, circumstancesin which in equilibrium there cannot exist an excess supply of labor. (1 + r) (p - ch)lr > p -cl 37If part of the fixed payment is paid after the crop, then it is as if the landlord lent the funds to or the worker; an increase in the amount to be paid has precisely the same effects as an increase in the p > (1 + r)ch - rc1 = p*. rate of interest detailed earlier. An increase in the fixed fee can, accordingly, adversely affect the ex- Accordingly, any individual who saw a pected return of the landlord(-cum-lender) (D. Gale firm attempting to sell a commodity for Johnson 1950). Allen (1985a) has shown that at low shares, tenants may have an incentive to "cheat"on less than p* would infer that that com- the contract, abscondingwith the entire produce of modity was a low-quality commodity.39 the land. ' The quotationfrom Sismondiin the introduction 39These beliefs are "rational";given the beliefs of to this paper makes it clear that selection effects may the individual, it would not pay the firm to produce also arise in the context of sharecropping. a high-qualitycommodity, if it ever produced a low- 24 Journal of Economic Literature. Vol. XXV (March 1987)

Wage/ Wage Efficiency Unit

w* B A B A

Effective Labor Supply LaborSupply At wages below w*, the effective labor supply is With nonconvexities in preferences, labor supply zero; at w* it increases discontinuously. may be discontinuous, but with continuum of individuals equilibrium exists. Figure 17a Figure 17b

disutility associated with effort up to a 3. Nutritional Models. critical level X, and infinite disutility This paper focuses on the dependence thereafter) and that the output per unit of quality on price arising from imperfect time with the maximum effort level e is information (adverse selection, moral a function of nutrition. Then the number hazard). But, as we noted, in one of the of labor units supplied will be a function earliest sets of efficiency-wage models of the wage rate. There is no way of ob- the quality of labor depended on the taining any labor at a cost per efficiency wage for nutritional reasons. How are unit of less than w*/X (w*). At wages in these nutritional models related to the excess of w* effective labor supply in- information models? creases, as depicted in Figure 17. Al- If the output of workers were perfectly though, by assumption at w*, each observable, then presumably all workers worker supplies either zero effective la- would be paid on a piece rate basis. In bor units or X(w*), all workers strictly equilibrium, some workers would be em- prefer to supply the latter; at w* there ployed (at the efficiency wage), and other is an excess supply of labor. identical workers would be unemployed. Although nutritional models can give Assume that workers get zero disutility rise to unemployment even if workers from work, up to 40 hours per week, and are paid on a piece rate basis, they obvi- infinite disutility thereafter (there is no ously can also give rise to unemployment quality commodity; and given that the firm indeed when piece rates are not feasible; thus, produces a low-qualitycommodity, the beliefs of in- in the example given above, there are dividuals are consistent with firm behavior. More no incentive problems, whether workers formally, the equilibriumcan be shown to be a per- are on a rate not.40 fect equilibrium. See Dybvig and Spatt (1983). paid piece basis or This is a "reputation"model based solely on incen- Thus, the nutritional efficiency-wage tive considerations. Other reputation models entail models, while differing in fundamental a mixture of incentive and adverse selection effects. ways from the informational efficiency- There are good producersand bad producers. When a buyer purchases a commodity that turns out to be wage models in their microfoundations, a low-quality commodity, then he infers that the yield similar conclusions: They generate seller must be a low-qualityseller. It is their concern similar form relationships be- about being so labeled that induces good producers reduced to produce high-qualitycommodities. One does not tween productivity and the wage paid by necessarily need many "bad"firms to induce good behavior on the part of the good firms, as David Kreps and Robert Wilson (1982) show in a rather 40It is straightforward to construct models that in- differentcontext. corporate both incentive and nutritional effects. Stiglitz: The Dependence of Quality on Price 25

the firm, wages paid by other firms, and ket for homogeneous commodities is nei- the unemployment rate. It is worth not- ther plausible nor innocuous. Markets in ing, however, that in one polar form of which commodities are completely ho- the efficiency-wage model, where indi- mogeneous-with respect to location and viduals do not share any of their income the date as well as other characteristics with others, productivity depends only -are almost inherently sufficiently thin on the absolute value of the wage paid so that the postulate of perfect competi- by the firm; accordingly the equilibrium tion is inapplicable. Markets that are suf- wage will be independent of the unem- ficiently "thick"to be competitive are al- ployment rate. (This stands in contrast most always nonhomogeneous. Consider with several versions of the informational the market for labor. If we define a sub- efficiency-wage model, where relative market, say, J. E. Stiglitz's labor, then wages and the unemployment rate are it may be homogeneous, but hardly com- the primary determinants of productiv- petitive; if we take a broader definition ity.) of the market, say, those with PhDs in economics, it may be fairly competitive, III. The Implications of the Dependence but it is hardly homogeneous. of Quality on Price for Economic Theory The problems with which we are con- 1. Differences Between Economies in cerned are central in capital markets and Which the Law of Supply and insurance markets. Both markets are es- Demand Is Repealed and Those sentially intertemporal: In the capital Where It Still Holds market, the lender lends the borrower money today, in return for a promise to It is useful to clarify the differences repay $(1 + r) next period, provided the between the assumptions in our analysis and those that are (often implicit) in the borrower can; in the insurance market, traditional competitive paradigm, e.g., of the insurer agrees today to pay the in- the model of Arrow and Debreu. Arrow sured a given amount next period if a and Debreu assume that each employer particular event occurs. The lender cares has perfect information concerning the about the likelihood that the borrower quality of his labor, that each buyer of a will default; the insurer cares about the commodity has perfect information con- likelihood that the insured against event cerning the quality of the commodities will occur. Borrowers (the insured) differ, which he purchases, and that each lender but lenders (insurers) cannot tell who is has perfect information concerning the more likely to default (to have an acci- characteristicsof those to whom he lends. dent). Thus, when they speak of a market for Moreover, the probability of default a commodity, they have in mind a market (of an accident) can be affected by the for a collection of objects, all of which actions of the borrower (insured). are identical, at least in all relevant as- The Arrow-Debreu model does not ac- pects. The person buying a commodity tually require perfect information. What or hiring on the market is completely in- it does require is that the nature of the different about which commodity he ob- "commodity"be fixed (i.e., that the aver- tains; the firm hiring a worker is indiffer- age quality of labor be unaffected by ent about which worker he obtains, and prices or wages and that average default no additional information would change (accident) probability be unaffected by his indifference. the terms of the loan (insurance policy)). This assumption of a competitive mar- There can be neither adverse selection 26 Journal of Economic Literature, Vol. XXV (March 1987) or moral hazard effects; that is, if all indi- Because most workers are not paid a viduals (commodities) are not identical, piece rate equal to the value of their mar- at least the mix of types is fixed and any ginal product, and because there are actions which they might take that affect costs associated with hiring workers, their productivity (or default, or acci- some of which are borne by the firm, dent, probabilities) are observable. firms are concerned both about the qual- Many economic relationships involve ity of the workers they hire, their pro- an element of insurance and/or loan. This ductivity on the job, and their turnover is the case, for instance, with most em- rate.43 The wage affects all of these vari- ployment and rental arrangements. ables. And so long as that is the case, There is one set of circumstances un- the possibilities with which we have been der which a firm does not care about the concerned here- that equilibrium will characteristics of those it hires. If the firm be characterized by an excess supply of can perfectly and costlessly monitor the labor, and/or by a wage distribution-are actions41 of its employees, and pay the real possibilities.44 employee for the services performed, and there are no fixed costs associated Albert Hirschman'sinsightful terminology. But labor with hiring a worker, then the firm is mobility is costly, and thus workers and firms are unconcerned whether the worker is a concerned about how the piece rate is set; it deter- low-ability worker pulling out three me- mines the division of the "ex post surplus"-given that the workersare workingfor the firm, the differ- dium sized weeds a day, or a high-pro- ence between the worker'sincome and his opportu- ductivity worker pulling out three hun- nity cost; and the difference between the profits of dred a day. It pays a fixed price per the firm from the current employees and what the medium sized weed out. profitswould be if the employer had to replace them. pulled 43 To the extent that firmsbear these turnovercosts In fact, relatively few workers are paid or pay workersnot solely on the basis of their actions, even partially on a piece rate basis (which firms can be thought of as providing insurance. ' There is one other situationwhere moral hazard is not to say that performance goes unre- problems do not arise in labor markets:When work- warded, either in terms of promotions ers are risk neutral (so there is no need for an insur- or salary). The literature on compensa- ance component in the relationship),then they can rent the machineson which they work(rent the land); tion schemes details a number of reasons because they receive all the residual, they will have for this, all of which attest to the impor- "correctincentives"; and because what the capitalist tance of the informational concerns that (landlord)receives is independent of the action of the worker, he is indifferentto the actions they un- are the center of our analysis (Stiglitz dertake. But even this is not correct if the worker 1975) but are completely ignored in the cannot pay the rent ahead of time; then there is some traditional competitive paradigm.42 probabilitythat he will fail to pay the promised rent, and the likelihood of this is, in general, a function of the actions he undertakes:There is still a moral 41 Where "actions" are sufficiently precisely speci- hazardproblem. When rents are paid ex post, then fied to imply a particular outcome in a particular it is as if the owner lent the money to the renter situation, regardless of who performed the action. (see Johnson 1950; Stiglitz and Weiss 1981). More- 42The monitoring required by piece rate systems over, if how the worker uses the machine affects its is costly; for it to be effective, there can be little future productivity and if it is difficult to verify variation in the quality of what is produced, or it whether the deteriorationin the machine is due to needs to be easy to write a contract that specifies misuse, there is a furthermoral hazard problem limit- the relationship between quality and price, and then ing the extent of rental markets. The contractualar- to observe and verify the quality of what is produced. rangementsbetween the partieswill try to limit these Moreover, as technology changes, the piece rate moral hazard problems, by stipulating restrictions needs to change; but this is often a contentious pro- on the use of the rented property.Thus, a landowner cess. If, of course, there were no costs of labor mobil- may stipulate what crops are to be planted, and he ity (no costs of hiring workers and no costs to the may impose restrictionson grazing.These moralhaz- worker of moving to another firm) then if the firm ard problems can be avoided by selling the machine; offered too low a piece rate, resulting in an income but this increases the risk of the worker; and it is below the worker's opportunity cost, then the worker even less likely that he will have the resources to would simply quit; exit would replace voice, to use pay for the purchase than to rent. Stiglitz: The Dependence of Quality on Price 27

It should be noted that most of the 2. Alternative Equilibrium Concepts issues with which we have been con- In Section I we argued that when cerned here would arise so long as indi- quality depends on price, equilibrium viduals are not paid on a piece rate, re- may be characterized by markets not gardless of the reason. Similarly, if the clearing. In traditional economic theory, firm must pay a uniform wage to all work- equilibrium is defined as market clear- ers, even though it knows which workers ing. Clearly, different notions of equilib- are more productive, it will worry about rium are being employed. Indeed, even the adverse selection effects of lowering within the literature recognizing the de- wages. pendence of quality on price, more than The conventional models assume not one equilibrium notion has been em- only that there is not imperfect informa- ployed, with contrasting results. In this tion concerning what is being bought, section, we review and contrast several but also, in trades which occur over time, of the more important equilibrium con- that there are no enforcement problems. cepts that have been used. In fact, of course, there are important Walrasian Versus Rationing Equilib- enforcement problems, arising both from ria. In models where quality depends on the incompleteness of contracts and from price, there may exist a market-clearing the costs of enforcing the contract terms price, for instance, a wage at which de- that are explicit.45 mand We referred to this The nutritional efficiency wage models equals supply. as the Walrasian equilibrium, but this can give rise to unemployment even may not be a market equilibrium. When- without informational problems, as we ever in the/labor market there is a region have already noted. These models differ where cost per effective labor unit de- from the conventional Arrow-Debreu creases with an increase in the wage, or model in two technical respects, which in the capital market there is a region turn out to be important: There is a non- where the expected return decreases convexity associated with the productiv- with an increase in the interest rate, ity wage relationship, and in equilibrium there may exist an equilibrium with ex- individuals are on the boundary of their cess supply of labor or excess demand feasible set.46 for capital. Indeed, there always exists 5 These enforcementproblems are, of course, cen- some level of demand for labor or supply tral both in internationallending, in implicit contract of funds for which this is true.47 theory, in sharecroppingmarkets, and in product markets. 46Thus what appears to be a technical assumption pears as in Figure 17a; the individualstrictly prefers in GerardDebreu's Theoryof Value becomes of cen- B to A. This explains why, with nonconvex prefer- tral importance in this analysis. Loosely speaking, ences, an equilibriumexists with a continuumof indi- to prove the existence of a market-clearingequilib- viduals, but a market-clearingequilibrium does not rium, one must show that the supply correspon- exist in the efficiency-wagemodel. Moreover, while dences are continuous, and to do this one must show with standardnonconvexities those who supply labor that the budget sets are continuous functions of and those who do not have the same level of utility, prices. To recast the nutritional efficiency-wage this is not true in our model. For a more extended model in standardterms, we let w# denote the wage discussion, see Dasgupta and Ray (1986a). per efficiency unit. While at w# = w*/A*, the indi- The demand curve is derived as follows. Assume vidual supplies X* efficiency units of labor, and at the firm's production function is Q = F[X(w)L]. It higher wages, the individual can supply, say, up to chooses w and L to maximize Q - wL where we A(w) units of labor, at lower values of w# the feasible have chosen output as the numeraire.From the first- set shrinks to zero. order conditions, A/A' = w meaning that the wage Note that while with nonconvex preferences, say, is chosen to minimize laborcosts per unit of effective for leisure, the supply of laborcorrespondence might labor (the efficiency wage). Employment is chosen look as in Figure 17b, with the individualbeing indif- so that, at the efficiency wage, the real wage equals ferent between points A and B; with the efficiency the value of the marginal product of labor: w = wage model, the labor supply correspondence ap- F'(w)A. 28 Journal of Economic Literature, Vol. XXV (March 1987)

It is important to emphasize that we existing workers, the firm would hire him are not arguing here that equilibrium is (assuming there were no further reper- never characterized by the equality of de- cussions of what might be viewed by mand and supply, only that it may not other workers as antisocial behavior).50 be. 48 Under these circumstances, equilibrium The Walrasian wage (interest rate) is would be characterized by demand the market equilibrium if and only if equaling supply. there exists no higher wage (lower inter- The fact that under these circum- est rate) at which costs per efficiency unit stances equilibrium is characterized by (expected returns) are lower (higher)49 demand equaling supply is thus a theo- The Choice of Appropriate Equilib- rem (admittedly trivial) to be proven; the rium Concepts. Our analysis thus differs equality of demand and supply should from the standard Arrow-Debreu model not be taken as a definition of equilib- not only in its informational assumptions, rium, but rather as a consequence follow- but also in its definition of equilibrium. ing from more primitive behavioral pos- Traditional theory has taken the equality tulates. What we have established in this of supply and demand to be part of the paper is that, under plausible behavioral definition of equilibrium. This, I think, postulates, equilibrium may not be char- is wrong. acterized by demand equaling supply. At Equilibrium is defined, loosely, as a the same time, it should be emphasized state where no economic agents have an that our economy is competitive in the incentive to change their behavior. conventional sense in which that word Whether a particular configuration of the is used: We are concerned with atomistic economy is an equilibrium depends, equilibria, in which all agents are small then, on agents' perceptions of the conse- relative to the market though in spite of quences of changes in their behavior. If this they are not price takers.5' employers believed that at a lower wage Passive Versus Active Sellers and Buy- they would obtain exactly the same qual- ers. While the models presented here ity of laborers as they obtained at a higher differ from the conventional competitive wage, then clearly, if a worker offered paradigm both in informational assump- to work for a wage lower than that of tions and in the equilibrium concept, these models differ from Akerlof's analy- 48In paticular, if the Walrasianwage exceeds the sis of adverse selection only with respect efficiencywage, the marketequilibrium is the Walra- to the equilibrium concept. In Akerlof's sian wage, and the law of supply and demand holds. Any firm that attempted to lower the wage to the model, as in the traditional perfect infor- efficiencywage would not be able to obtain any labor. mation model, however, both buyers and 49In this example, the cost per effective unit of sellers act completely passively. For in- labor curve has taken on a simple shape: At wages below w*, cost per effective labor unit is decreasing, stance, although firms know the statisti- and at wages greater than w*, it is increasing. There cal relationship between the wage paid is no intrinsic reason why the cost per effective unit of labor curve should take on such a simple shape as we have alreadynoted (see Figure 6). Then, there 'These repercussionshave been at the center of will be an excess supply of labor if the Walrasian recent literaturefocusing on the distinctionbetween wage is below w* or between tw and w**. It always insiders and outsiders. See AssarLindbeck and Den- pays the firmsto pay the wage in excess of the Walra- nis Snower (1984a). sian wage, which minimizes cost per effective unit "The concept of equilibrium employed by Das- of labor. Thus, for Walrasianwages in the interval gupta and Ray (1986a)in their analysis of the nutri- between wb and w** it pays the firm to increase tional efficiency wage model correspondsto a quasi- the wage to w**. (In these circumstances, equilib- equilibrium in Debreu (1959) and to the concept of rium may be characterizedby a wage distribution, a compensated equilibrium in Kenneth Arrow and as we noted earlier.) Frank Hahn (1971). Stiglitz: The Dependence of Quality on Price 29

and the productivity of the workers they tracts, perfect information concerning hire, they do not try to use this knowl- their borrowers. By contrast, Stiglitz and edge to increase their profits, e.g., by Weiss (1986a, 1986b) argue that as long setting wages at other than a market- as there is a residual of imperfect infor- clearing level. We would argue that there mation there may be scope for credit is, in most situations, no persuasive rea- rationing. 52 son to limit out uninformed agents to the In the Stiglitz-Weiss models, each bor- passive role that conventional theory has rower borrows the same amount. An- assigned to them. other term of the contract that can adjust Nonprice Rationing. A number of is the amount lent. Under certain cir- writers have argued that when, in the cumstances, reducing the amount lent capital market, interest rates cannot fall, reduces the risk faced by the bank. Thus, there are other methods by which mar- adjustments in the loan size can eliminate kets can be made to clear; similarly, in credit rationing (H. Milde and John Riley labor markets, when wages cannot fall, 1984). Again, the issue is not whether there are other methods by which mar- one can construct examples in which ra- kets can be made to clear. In markets tioning does not occur, but rather, are with imperfect information, contractual there alternative, plausible structures arrangements involve more than a single under which it does. In a multiperiod term; and it must be shown that none context, Stiglitz and Weiss (1981) have of these can adjust in a way as to re- shown that reducing the size of the loan store market clearing. There may be ad- may have an adverse effect on the risks verse selection and incentive effects from undertaken by the borrower; they under- changes in each of the contract terms. take projects that in effect, "force" the In the capital market, emphasis has lender to ante up more money in subse- been placed on the role of collateral. In- quent periods, if they are to recover their creasing collateral does induce firms to initial loans (see also Martin Hellwig undertake less risky projects, but may 1977). have adverse selection effects (Stiglitz In the labor market, discussions have and Weiss 1981; Hildegard Wette 1983; focused on the role of bonding and "job" Gerhard E. Clemenz 1984, 1985; Chang- purchases. That is, a worker could put Ho Yoon 1984, 1985). Several authors down a sum of money that he surrenders (e.g., Helmut Bester 1985) have con- in the event of being caught shirking. structed models in which, if banks can Bonding, it is argued, can alleviate the design contracts with varying interest incentive problems. B. Curtis Eaton and rates and collateral, there will be no William White (1982), Shapiro and Stig- credit rationing. litz (1984, 1985a), Edward Lazear (1982), It is important to recall that our con- and Stoft (1985) have argued against this tention has been not that equilibrium would always be characterized by credit 52They constructed a simple model in which they rationing, but that it may be, under plau- combine incentive and selection effects; there are sible conditions. Indeed, it is easy to con- two groups in the population, and each group has struct examples in which equilibrium is two activities (a safe and a risky project). Even though the bank is able to sort individuals perfectly, it cannot not characterized by credit rationing. raise interest rates, because to do so would induce Bester (1985) and David Besanko and An- greater risk taking. Hence, they show that there may jan Thakor (1984) provide examples with be credit rationing at one or both credit contracts. Credit rationing may also arise if individuals differ the peculiar property in that bankers can in more than one dimension, e.g., with respect to obtain, through offering a set of con- risk aversion and wealth. 30 Journal of Economic Literature, Vol. XXV (March 1987) on several grounds: Firstly, they note dividuals to pay for their full costs of that with young individuals having lim- training).54 ited capital there is, at least for these Advocates of the efficiency wage mod- workers, incomplete bonding, so that els claim, in the end, that the central firms are, in fact, concerned with the re- point is that firms do care about quit lationship between wages and productiv- rates, they do care about the incentives ity. (Though they could borrow to put of their workers, and they do use wage up the bond, this simply shifts who bears policies to affect the net profitability of the risk that the worker fails to perform, their employees. from the firm to the lender. Though non- 3. Other Imperfect Information Models vested pensions can be viewed as a form of bonding, it takes individuals a number The last two decades have seen a bur- of years to accumulate enough within geoning in imperfect information mod- their pension fund to serve as a suffi- els. It is worth noting the relationships ciently effective bond to eliminate the between the models that are the center need for firms to be concerned with of discussion here and some of these whether their workers shirk.) Secondly, other models. they note the "double moral hazard" Prices Versus Quantities Earlier liter- problem, the incentive of the firm to de- ature (Rothschild and Stiglitz 1976; Aker- clare that the worker has shirked when lof 1976; Michael Spence 1974) stressed he has not. This may be alleviated by the role of quantities in conveying infor- making the bond forfeiture go to a third mation; the literature this paper is con- party: again, the empirical relevance may cerned with stresses the role of prices be questioned; moreover such arrange- in conveying information. ments are sensitive to complicity by two The quantity of education obtained by of the parties against the third. 53 Thirdly, an individual conveys information be- they note that many individuals may not cause it is more costly for a less able indi- have funds to post a bond; to borrow the vidual to acquire education than for a funds would entail all the adverse effects more able individual. But it is no more noted earlier in our discussion of capital costly for a less able individual to an- markets; and to restrict applicants to nounce that he is willing to work only those who could finance the bond them- at a higher wage than a more able indi- selves would have the adverse selection vidual. How then can workers (informed effects noted earlier in connection with sellers) use prices to convey information collateral. about themselves? There must be some Similar problems arise with job pur- cost to announcing a higher reservation chases (or, what is equivalent, in the con- price: The price is the lower probability text of labor turnover, with requiring in- of being employed (or selling one's commodity). Stiglitz (1976a) and Nalebuff and ' These problems would also be ameliorated if Stiglitz (1982) have constructed models firms could establish a reputation. Another mechanism for providing incentives for in which higher-quality workers are will- workers that does not sufferfrom the "double moral ing to face a higher probability of not hazard"problem are contests (Lazear and Sherwin obtaining a job because their fallback Rosen 1981; Green and Nancy Stokey 1983; Halebuff and Stiglitz 1983b;Sudipto Bhattacharya1983). They do impose some risk for workers, and workers have 5' Some of the "problems"with job purchases ex- to believe that the contests are fairly administered, plain why the price for jobs would be low (e.g., and that they are evenly matchedwith their competi- workers'risk aversion, limitationson workers'access tors. Intrafirm contests may also have deleterious to the capital market), but not why the market for morale effects. jobs does not clear. Stiglitz: The Dependence of Quality on Price 31

wl w2 Wage

Indifference curve of Probability high ability of Employment

g* L Indifference curve of low ability

0 Wage and probability of employment acting as self-selection device. Figure 18 wage is higher. (Wilson [1979, 1980] has In the simple models analyzing the de- constructed a similar model for the prod- pendence of quality on price, quantity uct market.)55 Thus, if there are two variables are not observable. (There is types of workers, whose reduced form no variable, like education, to reveal the utility function can be represented by = labor markets;it may have more to do with the num- Ui Ui(w,g), where g is the probability ber of hours worked by a Harley Street doctor than of obtaining a job, then equilibrium may it does with unemployment among the low skilled. be characterized by both groups obtain- It is importantto note, however, that in this model ing wages commensurate with "unemployment"does not have the same interpreta- their abili- tion that it does in the national income statistics:It ties (though employers cannot observe means that the individualis not employed by others. them directly), but with the low-ability For the low skilled, this may indeed correspond to individuals having a unemployment; for the high skilled, it may corre- probability g* just spond to self-employment. large enough to induce the low-ability "This model has been extended by Nalebuffand not to apply, i.e., letting superscript 1 Stiglitz (1982)to the case where individualscan apply denote the for several jobs. In that case, the nature of the equi- low-ability workers librium depends critically on whether contracts are U'(wl, 1) = binding-whether once a job has been accepted, the Ul(w2, g*) worker can quit when he is offered a better job. If (See Figure 18).5657 contractsare binding, the opportunitycost of accepting a low-wage job is the foregone possibility of a high-wage job, the likelihood of which depends on 55Similarlythe probabilityof consummatinga deal both the job offersof differentfirms and the behavior may serve as a self-selection device within a bargain- of other individuals(how high they set their reserva- ing model. tion wages). They show that there exists a particular 3Technically this analysis should be contrasted wage distribution,such that when all workersration- with that of Rothschild and Stiglitz (1976), where ally set their reservation wages, the cost-per-effi- each firm (uninformedagent) sets the price and quan- ciency unit of the firm is the same for all firms. (In tity; here firms set only the price; the unemployment their equilibrium, while the quality of those who rate is determined as part of the equilibrium. This apply at each wage are, in fact, different, it is only is also true of the Stiglitz-Weiss (1986a) model of the wage-cum-unemploymentrate that distinguishes the capital market. them: in all other respects the workerslook the same. Some questions have been raised about the rele- Moreover, at each wage, except the lowest, there vance of this particularmodel for the description of is an excess supply of applicants.) 32 Journal of Economic Literature, Vol. XXV (March 1987)

le * Wage productivity curve facing / / second firm

Productivity /

/ / Wage productivity curve before / / first firm hires its (finite) I / number of workers I

Wage Figure 19. The Advantages of Being Late

individual's true ability.) But all that is Information Revealing Prices in the required for wages/prices/interest rates Capital Market. There is a closely re- to play a role in conveying quality infor- lated literature to that discussed here, mation given whatever variables that are where prices also convey information. In observable is that there remains some capital markets, the price at which a se- residual uncertainty of either the adverse curity sells may convey information con- selection or moral hazard sort.58 cerning the expected return of the security (or the likelihood of the occurrence of various states). (See Green 1973; By contrast, if the first firm to enter the market Grossman and Stiglitz 1976, 1980; Mar- had simply hired at what it thoughtwas the efficiency In wage, because it would have hired only lower-ability garet Bray 1981; Roy Radner 1979.) workers, those with reservationwages below the effi- structure, these models are most like the ciency wage, the "new"efficiency-wage schedule fac- Akerlof "Lemons" (1970) model, in which ing the later entrant would lie above the old one, both sellers and act nonstrategi- for w > w*, and hence there would be an advantage buyers to being late. See Lewis Guasch and Weiss (1980). cally: They take prices, and the informa- 5 Thus, models in which there are only limited tion conveyed by the price, as given. sources of imperfect informationmay be very mis- differ in while there is only leading. For instance, if individuals differ only in They that, one dimension, then education may serve to sort one seller (and usually only one buyer) individualsperfectly: There will be no residual im- of any particular automobile, or any par- perfect information, and hence no need to lower wages to improve the mix of laborers.But if individu- ticular person's labor services, there are als differin several dimensions, then educationalone many potential buyers and sellers of any will not sufficeto provide complete information.Sim- ilarly, in the capital market, if borrowersdiffer only in one dimension, and there is no moralhazard prob- in several dimensions and/or there are moral hazard lem, there may exist a fully revealing self-selection problems, there may exist credit rationing. Though equilibrium, and there will be, as a consequence, self-selection devices may reveal some information, no role for credit rationing. But if borrowers differ they will not be perfectly revealing. Stiglitz: The Dependence of Quality on Price 33

particular security. (Indeed, in the ab- (a) The Pattern and Form of Unem- sence of transactionscosts, with risk aver- ployment. A theory purporting to ex- sion and imperfectly correlated securi- plain cyclical fluctuations in unemploy- ties, virtually everyone is a buyer or ment should not only show that it can seller.) Thus, while the price of a car sold generate unemployment, but also ex- at an auction reflects the valuation placed plain the pattern and form of unemploy- on it by those who value it most highly, ment. This efficiency-wage theories in a security market, those who believe do.60 Less productive workers-those for that a security is overpriced will sell it whom the minimum wage per efficiency short, and their beliefs are, accordingly, unit is below some critical level-cannot reflected in the market equilibrium get jobs, though they might at higher price. levels of effective demand. (These may include young workers, part-time work- IV. Applications ers, or others for whom the total surplus from work is small, so that the no-shirk- 1. Implications for Macroeconomics ing wage, at any unemployment level, These models have direct and impor- is high.) tant implications for macroeconomics.59 Moreover, there is no work sharing, This section is divided into four sub- with the associated income reductions, sections, dealing respectively with labor for this would simply reduce the quality markets, capital markets, product mar- (productivity) of the labor force. In the kets, and the relationship of these theo- incentive efficiency-wage model, it is the ries to other recently advanced theories. total surplus (the amount by which the Labor Markets. The fact that these value of wage payments exceeds the fore- theories have yielded competitive mar- gone leisure) that determines whether ket equilibrium in which wages do not workers shirk; in the turnover model, it fall in the face of unemployment, in is the total surplus (relative to that offered which there is an equilibrium level un- by other firms) that determines whether employment, immediately suggests the a worker quits; in the selection models, possibility that these theories may pro- it is again the total surplus that deter- vide an important part of the explanation mines the individual's choice of one job of involuntary unemployment. This is an over another. Work sharing reduces the area of active ongoing research and con- surplus available to any individual, and troversy. What I wish to do here is to thus adversely affects the effort (quality, explain why some economists find these labor turnover). See, for example, Ar- models more persuasive than at least sev- nott, Hosios, and Stiglitz (1983) or Mi- eral of the competing theories, and to chael Hoel and Bent Vole (1986). present what appear to be at the present (b) Criticisms. Critics have raised sev- time the major issues, both the criticism eral objections, concerning both the and the defenses, pointing to certain un- quantitative significance of the efficiency- settled controversies. 'This should be contrastedwith the standardim- 59It is important to emphasize that although we plicit contract theories (e.g., Costas Azariadis1983; believe that these models provide an important part Grossmanand Oliver Hart 1981; for recent surveys, of the microfoundations for macroeconomics, they see Azariadisand Stiglitz 1983;and Hart 1983)which, do not provide the whole story: Other models (many at best, provide an explanationof work sharing (see of them also based on considerations of imperfect Stiglitz 1986). Arnott, Arthur Hosios, and Stiglitz information) are required. (1983) incorporate some aspects of efficiency-wage This section owes much to the comments of Bruce theory in their analysisof labor contractswith costly Greenwald and Larry Summers. labor mobility. 34 Journal of Economic Literature, Vol. XXV (March 1987)

wage effects and the consistency of the ployed, and thus even if individuals could theory with certain observed macroeco- have obtained a low-wage job, they nomic phenomena. choose not to do so.63 1. Can efficiency wage theory explain Greenwald has provided several alter- involuntary unemployment? Perhaps the native explanations for why individuals most widely cited criticism is that, unless may not accept a low-wage job, based efficiency wage considerations are impor- on information theoretic considerations: tant in all sectors, the theory cannot ex- Accepting a low-wage job may convey plain unemployment; that is, if there is information about the individual's ability; some sector (such as agriculture) where with high-ability individual's expecting to workers can be paid on a piece rate basis, get a "good"job sooner than a low-ability and the piece rate is flexible, then that individual, an individual who readily ac- sector should absorb all workers laid off cepts a low-wage job signals that he from those sectors where efficiency-wage thinks of himself as low-ability, and this considerations are important. Thus, effi- signal will lower his future wages. 64 ciency-wage theory might be able to ex- Moreover, accepting a job creates an in- plain wage differentials (the secondary la- formation asymmetry-the firm's new bor market), but not unemployment. employer will know more about the indi- (Note that this objection can be raised vidual's ability than other prospective against implicit contract theory explana- employers. Just as Akerlof showed that tions of unemployment as well.)6" these information asymmetries lead to Several of the efficiency-wage models thin markets for used cars, Greenwald have attempted to incorporate a flexible has shown that they result in thin mar- wage sector. In Stiglitz (1974c), for in- kets for "used labor."65Finally, he has stance, the agriculturalsector has flexible attempted to relate unemployment to wages, while efficiency-wage consider- capital market imperfections, which ations are important in the industrial sector. Individuals must choose in which sector to locate between the 63 Similarly, Arnott, Hosios and Stiglitz (1985) as- (mobility sume that the costs of search for the unemployed sectors is not costless and instantaneous). and the employed may be different and show that Unemployment in the urban sector is optimal contracts may then entail some individuals such as to income being layed off even if it means a finite probability equilibrate expected of being unemployed for a period. (of the marginal migrant) in that sector 'AAs is often the case, there may be another equi- to that in the rural sector.62 librium in which accepting a low-wage job does not Bulow and Summers signal one's ability: When there are significant trans- Jeremy Larry actions costs, it is more plausible that the acceptance (1985) assume that individuals can search of a low-wage job will serve as a signal. Alternatively, for a (high-wage) job only while unem- the argument to be given next serves to explain why accepting a job would lower one's lifetime wage prospects quite apart from signaling considerations: 61 Including theories of staggered contracts (Taylor Signaling considerations would then serve to 1980), which need to explain why labor is not ab- strengthen the magnitude of the effect of the accep- sorbed into those sectors whose contracts are up for tance of a low-wage job on lifetime income. negotiation. ' This result depends, in part, on the firm's not 62John Harris and Michael Todaro (1970) have de- being able to commit itself to paying higher wages veloped a similar analysis to explain urban unemploy- in the future. Those commitments themselves are, ment in LDCs. The major difference between the at best conditional on the firm's surviving. Thus, a Harris-Todaro model and the Stiglitz model is that worker who accepts a low wage now from a firm in in the former, the wage is exogenous and in the lat- bad financial straits with the promise of a high wage ter-as in all the work under examination here-it in the future takes, in effect, an equity position in is endogenous. Robert Hall (1975) has used a similar the firm. The reasons why workers may not be willing model to explain differences in unemployment rates to do so have been set forth in Greenwald, Stiglitz in different urban areas. and Weiss (1984). Stiglitz: The Dependence of Quality on Price 35

themselves can be explained by informa- some groups of workers or for some in- tion theoretic concerns. There are, in dustries than for others. Thus, consider general, some training costs that firms the argument that bonding is not em- must bear when they hire a worker; ployed because workers have insufficient moreover, hiring a worker represents a capital. This argument seems more ap- risky investment. With imperfect capital plicable to young workers and to low- markets, firms cannot divest themselves wage (unskilled) workers than to older of this risk; as a result, the implicit cost workers, and therefore might suggest of capital may be very high in a recession, that the effort-efficiency wage model is and hence it is possible that in a reces- more relevant for these workers than for sion, firms are willing to hire workers other groups of laborers.66 only if the lifetime wages are lower than Finally, we note that because of the they would be if the worker were hired presence of the important externalities in a later period, when the risks faced discussed in Part 1.5., the private costs by the firm are less. of pursuing, say, high-wage policies Critics might say at this juncture, (leading to unemployment) and of not "Aha, so unemployment is really volun- employing alternative strategies of sort- tary." We think little is gained from a ing and providing incentives may be semantic debate over whether unem- much less than the social costs. This has ployment is, in this sense, voluntary or been stressed by Akerlof and Janet Yellen involuntary. What is critical is (a) in the (1985).67 In particular, if firms are risk market equilibrium, some individuals, averse68 then they may not revise their with a given set of characteristics, have wages or, may not change other policies a distinctly higher level of (expected) util- even in the presence of some disturbance ity than other similar individuals; (b) the to the economy which leads to an excess equilibrium has, for one reason or an- supply of labor, even if were they to do other, some individuals in the unemploy- so would be welfare enhancing. ment pool who, under other circum- 3. Can efficiency wage theory explain stances, would be working; and (c) the nominal as well as real wage rigidities? market equilibrium is not (constrained) Although the criticism that these models Pareto efficient. provide an explanation of real-wage rigid- 2. Are there alternative mechanisms ities, not of nominal-wage rigidities, is for ensuring quality? A second criticism valid against most versions of the effi- is that if these efficiency-wage consider- ciency-wage models-as it is against most ations were really important, alternative mechanisms would be found that would ' This argument was put forward by Robert Hall not be anywhere near as socially costly in his discussion of Yellen's survey of efficiency-wage as the to which it models at the San Francisco meetings of the Ameri- unemployment gives can Economic Association, December 1984. rise. There are several answers to this 67This is in fact a general property of economies objection: Firstly, the fact of the matter that are not (constrained) Pareto efficient. There then is that firms are the exist large classes of perturbations to the economy concerned about that effect pareto improvements. If the firm sets its quality of their labor force and their rates wage in a privately optimal way, there exist some of turnover. Secondly, we have already pertubations of this wage that have a second order detailed some reasons at least some effect on the profits of the firm but have a first order why effect on the welfare of other agents in the economy. of the proposed alternatives that would See Greenwald and Stiglitz (1986a). eliminate unemployment (credit ration- 68 Again, the fact that firms behave in a risk averse be manner can be explained in terms of certain capital ing) may ineffective. Some of these market imperfections which, in turn, can be related arguments may hold with more force for to imperfect information. 36 Journal of Economic Literature, Vol. XXV (March 1987)

versions of implicit contract theory cal movements in real wages. This criti- (where, in principle, contracts should be cism takes on two forms. First, it is ob- indexed)-it is worth noting that the la- served, that in a recession, increases in bor-turnover efficiency-wage model is unemployment (reductions in relevant also consistent with money-wage rigidi- opportunities, including self-employ- ties: The turnover rate facing any firm ment income) should result in lower depends on the wages set by other firms; wages in the effort-efficiency wage mod- if each firm believes that other firms will els. Whether this is in fact the case is leave the wages unchanged in nominal debatable; if the quality of workers laid terms, it pays it to leave its wage un- off in a recession is, on average, lower changed in nominal terms. The Akerlof than those retained (a pattern that itself (1984) morale model is also consistent is consistent with efficiency-wage the- with nominal-wage rigidities. ory), then even if the average product 4. Can efficiency wage theory explain wage does not fall much, the quality ad- rationing among several groups? Simpler justed product wage may. Moreover, versions of the efficiency-wage theory what is relevant for the incentive-based yielded rationing for only one group in efficiency-wage theory is the lifetime re- the population. For all other groups, ei- turn from holding a job; with low dis- ther no one was hired or there was full count rates and long lifetimes, distur- employment. But with more general bances to the economy may leave this specifications, with productivity depend- relatively unaffected. Though the value ing on (each group's) unemployment rate of alternative opportunities may decrease then in equilibrium, there may be unem- in a recession, thus enabling a myopic ployment among several groups. And firm to lower its wage without inducing even if there were a continuum of shirking, nonmyopic firms will realize groups, with only one group rationed, that the value of the job is enhanced by the discontinuity in utility (between simi- the implicit insurance provided through lar groups) would remain. It is this dis- income smoothing."70'7 (On the other continuity-with similar individuals be- hand, because unemployment benefits ing treated discretely differently-which are typically lengthened during reces- is so much at odds with standard compe- sions, and the stigma associated with be- titive theory. ing laid off is decreased, it may actually 5. Can efficiency wage theory explain be worthwhile for firms to raise wages wage and employment dynamics? The fi- in recessions.) nal criticism69is that while the efficiency- wage theory may provide an 70This formulation entails, in effect, a synthesis explanation of implicit contract theory with efficiency-wage the- of the "natural unemployment rate," it ory. A firm that cheats on the implicit contract will does not provide an explanation of cycli- find that their workers leave later, when job opportunities become better. In this view, the appropriate time unit for the analysis of the worker-employer 69This list of criticisms is not meant to be exhaus- relationship is longer than just a month or even a tive. Another objection is that these models require year. not only that productivity increase with wages, but Moreover, selection (quality-efficiency wage) con- also that there be a range of wages over which in- siderations may strengthen these concerns, because creases in wages lead to more than proportionate it is likely to be the better workers who will have increases in productivity. This objection has been the easiest time finding alternative opportunities and dealt with where we showed how several simple mod- will have the highest propensity to quit. els exhibiting these properties can be constructed. 71 Moreover, wages are affected not only by de- It then becomes an empirical question whether in mand disturbances but also by any shock that affects practice the wage-productivity relationship has the the no-shirking constraint. Thus, the expectation of required shape. Unfortunately, there is insufficient an upturn in the economy would shift the no-shirking empirical evidence to date to provide a convincing constraint upward, necessitating an increase in to- answer. day's wage and a decrease in current employment. Stiglitz: The Dependence of Quality on Price 37

Still, the criticism that the efficiency- contracts cannot explain the persistence, wage theories have yet to provide a dy- even for a short while, of unemployment: namic theory is, for the most part, valid: At the first instance at which a contract The models constructed to date have comes up for renewal, the wage should been static. At a heuristic level, these fall to the market-clearing level. models do provide an explanation for why Moreover, the explanation of the pro- when there is a sudden shift in, say, the cess by which wages change is markedly demand for labor, it is not quickly re- different from the conventional story. We flected in a change in the wage, but have argued that in general the quality rather is reflected in a decrease in em- of the labor force (its productivity) de- ployment; the new equilibrium may en- pends on the unemployment rate as well tail unemployment, but even if in the as the wage. At higher unemployment long run equilibrium entails full employ- levels, the efficiency wage may be lower. ment of labor, the adjustment process For a variety of reasons, firms may be may entail unemployment as part of the slow to fire workers, and thus the unem- transition. ployment rate increases only slowly in The intuitive reason for this is easy to response to a decrease in the "long-run" see. Under a variety of conditions, the demand for workers. In such circum- quality of the workers obtained (or their stances, the fall in the wage rate can be productivity on the job) depends on the viewed both as a consequence of the in- wage paid by the given firm relative to creased unemployment and as mitigating that of other firms. Thus, if the different the extent to which unemployment in- firms in the economy do not simultane- creases. But in this theory the wage does ously adjust their wages, given the high not fall because of the "pressure" of ex- wages of other firms, it will not pay any cess supply directly on the labor market, single firm to lower its wage very much.72 but only because of the indirect effect, Our analysis of dynamics stands in through the effect on the efficiency marked contrast to the standard kind of wage. dynamics, which simply assumes slow- Capital Markets. We have already ness in the adjustment of wages and noted how the theory we have presented prices, and thus derives the transitional provides a theory of credit rationing. unemployment as a consequence of the Credit rationing, in turn, may provide ad hoc dynamic adjustment assumptions. both an explanation of why, and the Consider, for instance, the staggered mechanism by which, government policy wage contract model (Taylor 1980). The affects the macroeconomic equilibrium assumption of nonsynchronous long-term (Alan Blinder and Stiglitz 1983), and a contracts explains why wages at any par- part of the explanation of why the effec- ticular firm do not fall instantaneously tive cost of capital actually rises in the in the face of a decrease in the demand recession (Greenwald and Stiglitz 1986c, for labor, and why average wages fall only 1986d).73 gradually. But in the absence of some kind of efficiency-wage story, staggered 73This analysis can also be related to the recent studies (RobertBarro 1974; Stiglitz 1982c, 1983; Neil 72The analysis is somewhat more subtle than this Wallace 1981) arguing for the irrelevance of public suggests: Each firm's wage is dependent not only financialpolicies. A criticalassumption in these anal- on other firms' wages, but also on the unemployment yses is that public and private borrowingare perfect rate; nevertheless, in at least some versions of the substitutes for each other; in particular,individuals efficiency-wage model, the unemployment effect is are assumed to have unlimited access to the capital dominated by the wage effect, and wages do not fall, market. If they do not (and the theories presented or fall only slowly in the face of a downward shift in here explain why they do not), then public financial the demand curve for labor. policies are not irrelevant. 38 Journal of Economic Literature, Vol. XXV (March 1987)

Information concerning the nature of ipate facing, a liquidity crisis, one that borrowers is firm-specific and not easily may result in their bankruptcy: For them transferable. Accordingly, when the the effective cost of capital has increased. monetary authorities reduce the supply Thus, recessionary periods may be char- of high-power money, banks reduce the acterized by high effective interest rates credit that they make available. Mone- (in contrast to low nominal interest rates, tary policy, in this view, has effects not emphasized in the traditional literature); through the rate of interest (variations high effective interest rates can partly ex- in real rates of interest until recently plain many of the anomolous features of have been too small to account for much the business cycle, including patterns of of the variability in investment or sav- inventory accumulation (with a low ings), but through its effects on the sup- shadow price on labor and low interest ply of credit. This theory agrees with the rates, there should be much more pro- monetarists that the quantity of money duction smoothing than does in fact oc- may, as a result, play a more important cur) and some aspects of pricing role than the interest rate, but unlike policies. 76 the monetarists, provides a plausible Product Markets. Allen (1985b) has mechanism by which these effects are emphasized an alternative explanation realized. 7,7 for why firms may not lower prices in The increased uncertainty, for exam- the face of a downward shift in the de- ple, associated with recessionary periods, mand for their products: To do so could results in some groups being rationed out be interpreted as a signal of a deteriora- of the market who previously had not tion of quality. been. Moreover, other groups anticipate Relationship to Other Macrotheories. that the likelihood that they may be ra- We have already discussed briefly some tioned out of the market sometime in the of the contrasting implications of effi- future has increased. Firms face, or antic- ciency-wage theory and implicit contract theory. (a) Insider-Outsider Theory. Another 7 For this to be a cogent explanationof the cyclical variabilityin investment, one must explainwhy firms important recent development is com- do not turn to other sources of funds. The explanation monly called insider-outsider theory for this is provided in Greenwald, Stiglitz, and Weiss (1984), where they argue that asymmetries of infor- (Lindbeck and Snower 1984a, 1984b, mation make the cost of raisingcapital on the equity 1986, 1987; Robert Solow 1985). This marketvery high, prohibitivelyhigh for many firms. theory stresses the asymmetries between (Their model is essentially a direct applicationof the Akerlof-Lemonsmodel to the equity market; their insiders (the firm's current employees) result on the thinness of equity marketscorresponds and outsiders (potential employees). to Akerlof's results on the thinness of used car mar- These theories provide alternative ex- kets.) Equity rationingcan give rise to investment fluctu- planations for why net profits of the ations, even without credit rationing;for the absence firm might be reduced if a firm hired of futures marketscombined with limitationson the new workers at much lower wages; for ability to sell equity imply that a decision to produce forces managers/ownersto absorb risk. A reduction example, the current workers, recogniz- in their workingcapital will thus lead to a reduction in their equilibrium production levels. If, as is in fact the case, debt is not indexed, monetary shocks 76As Edmund Phelps and Sidney Winter (1970) can have real effects throughtheir effects on working emphasized, there may be a trade-off involved in capital recruiting customers today by lowering prices, with For other studies of monetarypolicy in the pres- higher future profits, but lower current profits. An ence of credit rationing,see Jackmanand John Sutton increase in the effective cost of capital tilts the trade- (1982), Smith (1983), Vale (1986), and Lindbeck off toward higher current profits, i.e., higher prices (1963). today. Stiglitz: The Dependence of Quality on Price 39

ing the threat that these workers pose provides the microfoundations of these to them, may refuse to provide them price rigidities.79 training. These theories are thus per- 2. Implications for Development fectly consistent with efficiency-wage Economics theories. 77 (b) Fixed Price Models and Efficiency We noted earlier that much of the Wage Theories. One of the reasons for recent interest in the unemployment interest in the efficiency-wage theories consequences of the dependence of pro- is that they provide an explanation of ductivity on wages originated in the de- wage and price rigidities, which play velopment literature, where the rela- such a central role in the fixed wage-price tionship was attributed to nutritional models that have enjoyed such popular- considerations. Since than a large litera- ity during the past decade. Though the ture has explored a variety of other two approaches are, to that extent com- causes and consequences of the wage- plementary,78 the efficiency-wage mod- productivity nexus within LDCs.80 els can also be seen as providing a cri- Mirrlees (1975) and Stiglitz (1976b) tique of the fixed-wage models, or at least showed that wage-productivity relations of the relevance of those models for pol- of the form depicted in Figure 2 would icy purposes. Though wages do not fall give rise, even within utilitarian families to a market-clearing level, government (maximizing the sum of the utility of the policies can affect the level of wages and members of the family) to inequality in thus the equilibrium level of employ- consumption; some members would re- ment. (In contrast, the fixed-wage price ceive a low consumption level, others a models simply assume that wages and high consumption level; those with a prices will remain unchanged.) low consumption level would have a (c) Multipliers. Multipliers have had low productivity. What they consumed a long and noble history in macroeco- would exceed their marginal product, nomic analysis. It has not been widely but be less than their marginal product recognized how difficult it is to obtain plus a pro rata share of the rents. Con- multipliers in conventional models: Usu- versely for the high consumers. Indeed, ally price responses in stable systems even if the family is Rawlsian (maximizes dampen the effect of any initial distur- the welfare of the worst off individual) bance. It is precisely because of price there may be consumption inequality. rigidities that traditional macroeconomic Because productivity depends on con- models yield multipliers. Our analysis sumption, individuals with landholdings will be more productive than landless 77They emphasize, however, at least one important aspect to which efficiency-wage theory has paid insufficient attention: The worker may not be able 79 The externalities associated with incomplete to commit himself to receiving low wages in the fu- markets and imperfect information also give rise to ture; after he is trained he may be in a position to multipliers. These multipliers should not be confused extract a high wage. What is relevant, of course, with the welfare multipliers that arise whenever the for the firm's decision is the relationship between economy is not (constrained) Pareto efficient. Then lifetime wages and productivity. (Notice that while there always exist some perturbations to some indi- Greenwald has emphasized the consequences of the vidual or firm that have a second order effect on inability of firms to commit themselves to pay high the welfare of the individual or firm but a first order wages in the future, insider-outsider theory has em- effect on social welfare (see Greenwald and Stiglitz phasized the consquences of the inability of workers 1986b). to commit themselves to accepting low wages in the 80In this section, we focus our discussion on the future.) consequences of the nutritional-wage productivity 78For an explicit attempt to integrate the two ap- nexus for LDCs. For a discussion of alternative expla- proaches, see Karl Moene (1985). nations, see Stiglitz (1982b, 1974c). 40 Journal of Economic Literature, Vol. XXV (March 1987)

workers, and will, accordingly, receive valorem wage subsidies have distinctly higher wages. Dasgupta and Ray (1986a, different effects on wage setting and em- 1987) have explored the implications of ployment policies. Because there are two inequality in land ownership for wages objectives that the government wishes and output. In particular, they note that to achieve (the correct wage level and the very poor, those with very small land- the correct urban employment level), it holdings, may be completely excluded requires two instruments; the two forms from the market because their minimum of wage subsidies/taxes provide the req- wage costs per efficiency unit is too high. uisite instruments. In their model, whether the economy is The wage-productivity nexus also has in an unemployment regime or a full em- important implications for the determi- ployment regime will depend on the ag- nation of the shadow wage of labor (par- gregate land supply (relative to the labor ticularly in the context of models with supply), in effect, on whether the Walra- Harris-Todaro and related migration sian wage is below or above the efficiency mechanisms). The opportunity cost of la- wage; it may also depend on the distribu- bor is not zero (in spite of the presence tion of land. A land reform may thus have of unemployment) or even the rural a significant effect on national output. wage. In some central cases, the shadow The dependence of productivity on wage is the urban wage, independent of wages in the urban sector results in urban attitudes toward future generations; in wages being set at levels in excess of the other cases, it lies between the urban rural wage. This, by the familiar Harris- wage and the rural wage (see Stiglitz Todaro migration mechanism (Todaro 1982a). 1968, 1969; Harris and Todaro 1970) and its generalizations (Stiglitz 1974c; Gary 3. Further Applications Fields 1975; Raaj Sah and Stiglitz 1984, 1985), results in urban unemployment. The dependence of quality on price As noted earlier (Part 1.5 above) the wage has a large number of other implications, and employment levels set by private two of which we briefly note here. firms do not maximize national income Technological Change and Competi- (and are not Pareto efficient). On the tive Entry. The fact that a lower price other hand, even if the government could is associated with lower quality has im- directly control the urban wage, it would portant implications for technological not set it at the rural wage: Some level change. Normally, we argue that if a firm of unemployment is optimal. Moreover, develops a cheaper way of making a policy prescriptions to reduce the real mousetrap it will be able to undercut its wage indirectly, through increasing the rivals, and thus to capture the whole mar- prices of commodities, are, at best, mis- ket for itself. When prices convey infor- guided. For private firms would respond, mation, the firm may not be able to un- say, to an increase in the price of food dercut its rivals; lowering the price may by increasing the wage. Indeed, if pro- simply lead potential customers to be- ductivity is more sensitive to the con- lieve that it is selling a lower-quality sumption of food than to the consump- mousetrap. (Farrell 1984, 1985, has dis- tion of other commodities, government should subsidize the consumption of consumption, workers could resell the subsidized food.8' In this context, specific and ad food, and the firm would thus not directly gain from the food subsidy. 81 Firms would, in such circumstances, find it in For an analysis of optimal taxation and subsidies their interests to subsidize the consumption of food; in the presence of productivityeffects, see Sah and but unless the subsidy is provided for on-premise Stiglitz (1984). Stiglitz: The Dependence of Quality on Price 41

cussed the entry barriers that arise in nopoly, the motive is to discriminate, to these models.) capture as much of the consumer surplus The Consequences of the Wage-Qual- from each worker/customer as possible. ity Relationship in Noncompetitive Mark- Concluding Remarks ets. The analysis so far has been confined to competitive markets. But similar re- There is little doubt that the observa- sults apply to noncompetitive markets as tion that quality may depend on price well. (productivity on wages; default probabil- The standard theory maintains that a ity on interest rates) has provided a rich monopsonist would never ration. For at mine for economic theorists: A simple the given price, if there is an excess de- modification of the basic assumptions re- mand he can increase his profits simply sults in a profound alteration of many of by increasing his price. Similarly, a mo- the basic conclusions of the standard par- nopsonist in the labor market would adigm. The Law of Supply and Demand never pay a wage in excess of the mini- has been repealed. The Law of the Single mum wage required to hire the number Price has been repealed. The Funda- of workers he wishes; there would never mental Theorem of Welfare Economics be an excess supply of workers (though has been shown not to be valid. the number of workers hired may be less More than that, the theories that we than in a competitive market-clearing describe here provide the basis of pro- equilibrium). But if the firm recognizes gress toward a unification of macroeco- that by raising the wage it increases the nomics and microeconomics. They pro- quality of its labor force, it may minimize vide an explanation of unemployment its wage costs by paying a wage that is and credit rationing, derived from basic above the minimum required to obtain microeconomic principles. It is a theory the amount of labor that it wishes to ob- in which the extensive idleness that peri- tain. Similarly, in the capital market, a odically confronts society's resources, hu- monopoly bank may charge an interest man and capital, is seen as but the most rate below the market-clearing level, rec- obvious example of market failures that ognizing that in doing so it increases its prevasively and persistently distort the expected returns. allocation of resources. Earlier, we showed how, in a competi- Several caveats should, however, be tive market, wages cum queues would borne in mind. Firstly, the repeal of the screen individuals; more able individuals Law of Supply and Demand (and the Law might be willing to take the risk of apply- of the Single Price) is a selective repeal: ing for a high-wage job, with a low proba- We have not contended that equilibrium bility of getting the job, when an (obser- is never described by the equality of de- vationally identical) individual of lesser mand and supply, only that it need not ability would not. The monopsonist be, and will not be in some important would similarly attempt to use wages circumstances. cum queues to differentiate among work- Secondly, though the basic outlines of ers. the general theory appear now to be well Still, it should be emphasized that the established, there remain several impor- motives for differentiating among indi- tant extensions and developments. We viduals (customers or workers) are mark- have referred to several of these within edly different under monopoly than un- the text. The models presented here der competition: Under competition, the were, for the most part, static; it is impe- motive is simply to identify workers who rative to develop an explicitly dynamic differ in productivity, while under mo- model if these theories are to provide 42 Journal of Economic Literature, Vol. XXV (March 1987)

part of the foundations of a theory ex- survey will spur continuation of efforts plaining cyclical fluctuations in employ- in that direction. ment. Moreover, on several occasions we This paper has, however, attempted contrasted efficiency-wage theory with to show how similar ideas have found ap- implicit contract theory, arguing that effi- plication in the analysis of labor, capital, ciency-wage theory provides a far better and product markets. These models pro- explanation of unemployment than does vide an explanation of several phenom- implicit contract theory. In fact, individ- ena within these markets that cannot be uals do have long-term implicit contract easily explained within the more conven- relationships with their employers; these tional paradigm. relationships are affected in fundamental REFERENCES ways by efficiency-wage considerations. The integration of implicit contract the- AIGNER, DENNIS AND CAIN, GLEN. "StatisticalTheo- .ries of Discrimination in Labor Markets," Ind. ory and efficiency-wage theory thus is a Lab. Relat. Rev., Jan. 1977, 30(2), pp. 175-87. second important topic for a research AKERLOF, GEORGE A. "The Market for 'Lemons': agenda. Qualitative Uncertainty and the Market Mecha- nism," Quart. J. Econ., Aug. 1970, 84(3), pp. 488- We have also noted that the problems 500. with which we have been concerned are . "The Economics of Caste and of the Rat mitigated, but not eliminated, by moni- Race and Other Woeful Tales," Quart. J. Econ., Nov. 1976, 90(4), pp. 599-617. toring and bonding (among other instru- . "A Theory of Social Custom, of Which Un- ments that may be available to the firm). employment May Be One Consequence," Quart. The limitations on monitoring and bond- J. Econ.," June 1980, 94, pp. 749-75. Labor Contracts as Partial Gift Exchange," ing have, however, received only limited Quart. J. Econ., Nov. 1982, 97(4), pp. 543-69. theoretical scrutiny. Finally, we have, for . "Gift Exchange and Efficiency Wage Theory: the most part, analyzed incentive models Four Views," Amer. Econ. Rev., May 1984, 74(2), pp. 79-83. and selection models in isolation from AKERLOF, GEORGE A. AND STIGLIIZ, JOSEPH. "Capital, each other.82 We have noted, however, Wages and StructuralUnemployment," Econ. J., that there may be important interactions June 1969, 79(314), pp. 269-81. AKERLOF, GEORGE A. AND YELLEN, JANET. "TheMac- between the two, and these require fur- roeconomic Consequences of Near Rational Rule ther study. of Thumb Behavior." Mimeo, U.C. Berkeley, Thirdly, we have focused our attention 1983. . "A Near-rational Model of the Business Cy- on the burgeoning theoretical literature cle, with Wage and Price Inertia,"Quart, J. Econ., (but have made no pretense of being Supp. 1985, 100(5), pp. 823-38. even complete within the scope of the ALLEN, FRANKLIN. "Loans, Bequests and Taxes Where Abilities Differ:A TheoreticalAnalysis Us- topics covered). This is partly a conse- ing a Two-abilityModel." D. Phil. thesis, Oxford, quence of the author's comparative ad- 1980. vantage, but partly a consequence of the . "The Prevention of Default," J. Finance, May 1981, 36(2), pp. 271-76. fact that the models presented here have . "Credit Rationing and Payment Incentives," not been the subject of extensive empiri- Rev. Econ. Stud., Oct. 1983, 50(4), pp. 639-46. cal testing.83 We hope, in fact, that this recent discussion of the relationship between pro- 82An exception is the Stiglitz and Weiss (1986a, ductivity and wages, see, for example, James Medoff 1986b) studies. and KatharineAbraham (1981). Macroeconomicanal- 83 There have, however, been numerous studies yses of the relationshipbetween unemploymentand addressing particular aspects of the theories de- productivity include James Rebitzeer (1985), and scribed in this paper. For an early investigation of Weisskopf, Bowles, and Gordon (1983). the price-qualityrelationship, see Gabor and Gran- There is a growing literature attempting to test ger (1966). For a recent examinationof the empirical the credit-rationing models. See, for instance, evidence on whether the labor market clears, see Charles Calomiris and R. Glenn Hubbard (1985), ThomasKniesner and ArthurGoldsmith (1986). For a Leonard Nakamura(1985), and P. Kugler (1985). Stiglitz: The Dependence of Quality on Price 43

._ 'Reputation and Product Quality," Rand J. ductivity, Wages and Nutrition,"J. Devel. Econ. Econ., 1984. 15(3), pp. 311-27. Dec. 1978, 5(4), pp. 331-97. . "On the Fixed Nature of Sharecropping . Palanpur-studies in the economy of a Contrasts," Econ. J., Mar. 1985a, 95(377), pp. 30- North Indian village. New Delhi: OxfordU. Press, 48. 1981. . "A Theory of Price Rigidities When Quality BOWLES, SAMUEL. "The Production Process in a Is Unobservable." Mimeo, U. of Pennsylvania, Competitive Economy: Walrasian, Neo-Hobbes- 1985b. ian, and MarxianModels," Amer. Econ. Rev., Mar ANNABLE, JAMES E., JR. "A Theory of Downward- 1985, 75(1), pp. 16-36. Rigid Wages and Cyclical Unemployment," Econ. BRAVERMAN, AvISIIAY AND SHINIVASAN, T. N. "Credit Inquiry, July 1977, 15(3), pp. 326-44. and Sharecroppingin AgrarianSocieties, "J. Devel. . "Money Wage Determination in Post Econ., Dec. 1981, 9(3), pp. 289-312. Keynesian Economics,"J. Post Keynesian Econ., BRAVERMAN,AvIsIIAY AND STIGLITZ,JOSEPli E. Spring 1980, 2(3), pp. 405-19. "Sharecroppingand the Interlinkinigof Agrarian . "Another Auctioneer is Missing." Mimeo, Markets,"Amer. Econ. Rev., Sept. 1982, 72(4), Economics Department, The First National Bank pp. 695-715. of Chicago, 1985 (earlier version presented at . "Landlords,Tenants and Technological In- meetings of the International Atlantic Economic novations,"J. Devel. Econ., 1986. Association, Rome, 1985). BRAY, MARGARET. "FuturesTrading, Rational Expec- ARNOTT, R.; Hosios, A. AND STIGLITZ, J. "Implicit tations, and the Efficient Markets Hypothesis," Contracts, Labor Mobility and Unemployment." Econometrica, May 1981, 49(3), pp. 575-96. Mimeo, Princeton U., 1983 (revised version of a BULOW, JEREMY I. AND SUMMERS, LAWRENCE H. "A paper presented at NBER-NSF conference, Dec. Theory of Dual Labor Markets with Application 1980). to Industrial Policy, Discriminationi and Keynesian ARNOTT, R. AND STIGLITZ, J. Labor Turnover, Wage Unemployment." NBER Working Paper No. 1666, Structures and Moral Hazard," J. Labor Econ., July 1985. 1985, 3(4), pp. 434-62. CALOMIRIS,CIIARLES AND HUBBARD, ROBERT GLENN. . 'Moral Hazard and Optimal Commodity "Price Flexibility, Credit Rationing, and Economic Taxation," J. Public Econ., Feb. 1986, 29, pp. 1- Fluctuations: Evidence from the U.S., 1979- 24. 1914." Mimeo, Northwestern U., Oct. 1985. ARRow, KENNETII J. AND HAIIN, F. H. General com- CALVO, GUILLERMO. Quasi-Walrasian Theories of petitive analysis. Edinburgh: Oliver and Boyd, Unemployment," Amer. Econ. Rev., May 1979, 1971. 69(2), pp. 102-07. AZARIADIS, C. "EMPLOYMENT WITlI ASYMMETRICIN- CALVO,GUILLERMO AND PIIELPS, EDMUND. "Indexa- FORMATION," Quart. J. Econ., Supplement 1983, tion Issues: Appendix: Employment Continigent 98(3), pp. 157-72. Wage Contracts," in Stabilization of the domestic AZARIADIS, C. AND STIGLITZ,J. E. "Implicit Contracts and international economy. Vol. 5. Eds.: KARL and Fixed Price Equilibria," Quart. J. Econ. Sup- BRUNNER AND ALLAN H. MELTZER. Carniegie-Roch- plement, 1983, 98(3), pp. 1-22. ester Conference Series on Public Policy. J. Monet. BALTENSPERGER, ERNST AND MILDE, HELLMUTII. Econ., Suppl., Series 1977, 5, pp. 160-68. "Loan Rate Flexibility and Asymmetric Default In- CALVO, GUILLERMOAND WELLISZ, STANISLAW. "Hier- formation," Geld, Banken und Versicherungen, archy, Ability and Income Distribution," J. Polit. 1982, pp. 1165-78. Econ., Part 1, Oct. 1979, 87(5), pp. 991-1010. BARRO, ROBERT. "Are Government Bonds Net CIIEUNG, S. N. S. The theory of share tenancy. Chi- Wealth?" J. Polit. Econ., Nov./Dec. 1974, 82(6), cago: U. of Chicago Press, 1969. pp. 1095-1117. CLARK, GREGORY."Productivity Growth without BECKER, GARY S. AND STIGLER, GEORGE J. "Law En- Technical Change: European Agriculture before forcement, Malfeasance, and Compensation of En- 1850." Unpub. ms., 1986. forcers," J. Legal Stud., Jan. 1974, 3(1), pp. 1- CLARK, KIM AND SUMMERS, LAURENCE. "Labor Mar- 18. ket Dynamics and Unemployment: A Reconsidera- BESANKO, DAVID AND TilAKOR, ANJAN. "Collateral and tion," BrookingsPap. Econ. Act., 1979, 1, pp. 13- Rationing: Sorting Equilibria in Monopolistic and 60. Competitive Markets." Northwestern University CLEMENZ, G. "Credit Rationing in the Absence of Discussion Paper, 1984. Direct Observability of Efforts and Abilities of Bor- BESTER, HELMUT. "Screening vs. Rationing in Credit rowers." Institut fur Wirtschaftswissenschaften der Markets with Imperfect Information, " Amer. Econ. Universitait Wien, Working Paper No. 8405, June Rev., Sept. 1985, 75(4) pp. 850-55. 1984. BIIATTACIIARYA,SUDIvTro. "Tournaments, Termina- . "Credit Markets with Asymmetric Informa- tion Schemes and Forcing Contracts." Mimeo, U. tion and the Role of Collateral." Mimeo, U. Wien, of California, Berkeley, June 1985. 1985. BLINDER, A. AND STIGLITZ, J. E. "Money, Credit DASGUPTA, PARTIIA AND RAY, DEBRAJ. "Inequality as Constraints and Economic Activity," Amer. Econ. a Determinant of Malnutrition and Unemploy- Rev., May 1983 73(2), pp. 297-302. ment: Theory," Econ. J., Dec. 1986a, 96. BLISS, CIIRISTOPIIERJ. AND STERN, NICilOLAS. "Pro- . "Adapting to Undernourishment: The Clini- 44 Journal of Economic Literature, Vol. XXV (March 1987)

cal Evidence and Its Implications."Mimeo, U. of rium,"in J. Japaneseand InternationalEconomies. Cambridge, 1986b. Forthcoming. . "Inequalityas a Determinantof Malnutrition . "The Theory of Production and Price in Con- and Unemployment:Policy," Mar. 1987, 97. tingent Renewal Markets." Mimeo, U. of Massa- DEBREU,GERARD. Theory of value. NY:John Wiley, chusetts, Nov. 1985. 1959. GREEN,JERRY. "Information, Efficiency and Equilib- DEVINNEY,TIMOTIIY M. "Iicentives anidMulti-Period rium." Harvard Institute of Economic Research, Rationingin Loan Contracts,in Geld, Banken und Discussion Paper No. 284, Mar. 1973. Versicherungen,Eds.: HERMANNGOPPL AND Ru- GREEN, JERRY AND STOKEY, NANCY. "A Comparison DOFHENN, VVW Karlsruhe,pp. 629-46. of Tournaments and Contracts," J. Polit. Econ., DYBVIG, PIIILIP H. AND JAYNES, G. "Output-supply, June 1983, 91(3), pp. 349-64. Employment and Intra-Industry Wage Disper- GREENWALD, BRUCE.Adverse selection in the labor sion," Cowles Foundation Discussion Paper No. market. NY: Garland, 1979. 546, 1980. . "Adverse Selection in the Labour Market," DYBVIG,PIIILIP H. AND SPATT,CIIESTER S. "Does It Rev. Econ. Stud., July 1986, 53(3), pp. 325-47. Pay to Maintain a Reputation? Consumer Informa- GREENWALD,BRUCE AND STIGLITz,JOSEPII. "Exter- tion and Product Quality." Working paper, 1983. nalities in Economics with Imperfect Information EATON, B. CURTIS AND WIIITE, WILLIAMD. "Agent and Incomplete Markets," Quart. J. Econ., May Compensationand the Limits of Bonding," Econ. 1986a, 101(2), pp. 229-64. Inquiry, July 1982, 20(3), pp. 330-43. GREENWALD,BRUCE AND STIGLITZ,J. "Externalities EATON,JONATIIAN. "Lending with Costly Enforce- in Economies with Information," Quart. J. Econ., ment of Repaymentand PotentialFraud."J. Bank- Feb. 1986a. ing and Finance, 1986, 10, pp. 281-93. . "The Inefficienicy of Competitive Equilibria EATON,JONATIIAN AND GERSOVITZ, MARK. "LDC Par- with Rationing." Mimeo, 1986b. ticipation in International Financial Markets: Debt . "Information, Finance Constraints and Busi- and Reserves," J. Devel. Econ., Mar. 1980, 7(1), ness Fluctuations." Paper prepared for the Semi- pp. 3-21. nar on Monetary Theory, Taipei, Jan. 3-8, 1986c. . "Debt with Potential Repudiation: Theoreti- . "Money, Imperfect Information, and Eco- cal and Empirical Analysis," Rev. Econ. Stud., nomic Fluctuations." Paper parpared for the Apr. 1981a, 48(2), pp. 289-309. Seminar on Monetary Theory, Taipei, Jan. 3-8, . "Poor Country Borrowing and the Repudia- 1986d. tion Issue." Princeton Studies in InternationalFi- GREENWALD, BRUCE; STIGLITZ, JOSEPII AND WEISS, nance No. 47, Princeton, NJ, June 1981b. ANDREW. "Informational Imperfections and Macro- EATON, JOIIATIIAN;GERSOVITZ, MARK AND STIGLITZ, economic Fluctuations," Amer. Econ. Rev., May J. E. "The Pure Theory of Country Risk," Euro- 1984, 74(2), pp. 194-99. pean Econ. Rev., 1986, 30(3), pp. 481-513. GROSSMAN, SANFORD AND HART, OLIVER. "Implicit FARRELL,JOSEPII. "Prices as Signals of Quality." M. Contracts, Moral Hazard, and Unemployment," Phil. dissertation, Oxford, 1979. Amer. Econ. Rev., May 1981, 71(2), pp. 301-07. . "Repeat Sales, Quality and Prices." Mimeo, GRoSSMAN, SANFORD AND STIGLITZ, JOSEPIi E. "Infor- M.I.T., Dec. 1980. mation and Competitive Price Systems," Amer. . "Moral Hazard in Quality, Entry Barriers, Econ. Rev., May 1976, 66(2), pp. 246-53. and Introductory Offers," M.I.T. Working Paper, . "On the Impossibility of Informationally Effi- May 1984. cient Markets," Amer. Econ. Rev., June 1980, . "Moral Hazard and an Entry Barrier." Rank 70(3), pp. 393-408. J. Econ., Fall 1986, 17, pp. 440-49. GUASCIi, LuIS AND WEISS, ANDREW. "Adverse Selec- FIELDS, GARY."Rural-Urban Migration, Urban Un- tion by Markets and the Advantage of Being Late," employment and Underemployment, and Job- Quart. J. Econ., May 1980, 94(3), pp. 453-66. Search Activity in LDCs," J. Devel. Econ., June HALL, ROBERT. "The Rigidity of Wages and the Per- 1975, 2(2), pp. 165-87. sistence of Unemployment," Brookings Pap. Econ. FITZRoY,F. "Contests." Mimeo, International Insti- Act., 1975, 2, pp. 301-49. tute of Management, West Berlin, 1981. . "Employment Fluctuations and Wage Rigid- FREIMER, MARSIIALLAND GORDON, M. J. "WhyBan-k- ity," Brookings Pap. Econ. Act., 1980, 1, pp. 91- ers Ration Credit," Quart. J. Econ., Aug. 1965, 123. 79, 397-416. HAMERMESII, DANIEL S. "Interdependence in the FRIEDMAN, BENJAMIN M. "The Roles of Money and Labour Market," Economica, Nov. 1975, 42(168), Credit in MacroeconomicAnalysis." in Macroeco- pp. 420-29. nomics, prices, and quantities: Essays in memory HARBERGER, A. C. "On Measurinig the Social Oppor- of Arthur M. Okun. Ed.: JAMES TOBIN. Washing- tunity Cost of Labour," Int. Lab. Rev., June 1971, ton, DC: BrookingsInst., 1983, pp. 161-89. 103(6), pp..559-79. GABOR,ANDRE AND GRANGER, C. W. J. "Price as HARRIS, JOIIN R. AND TODARO, MICIIAEL P. "Migra- an Indicator of Quality: Report of an Inquiry," tion, Unemployment & Development: A Two-Sec- Economica, N.S., Feb. 1966, 33, pp. 43-70. tor Analysis," Amer. Econ. Rev., Mar. 1970, 60(1), GINTIS, HERBERT AND ISIIIKAWA, TSUNEO. "Wages, pp. 126-42. Work Discipline and Macroeconomic Equilib- HART, OLIVER. "Optimal Labour Conitracts Under Stiglitz: The Dependence of Quality on Price 45

Asymmetric Information: An Introductioni,"1Rev. Uniemploymenit as anl Inisider-Outsider Dilemma." Econ. Stud., Jan. 1983, 50(1), pp. 3-35. Seminiar Paper No. 282, Inistitute for Initerniational HEAL, GEOFFREY. "Do Bad Products Drive Out Econiomic Studies, U. of Stockholm, 1984a. Re- Good?" Quart. J. Econ., Aug. 1976, 90(32), pp. vised as "Wage Rigidity, UniioniActivity anid Uni- 499-502. employmenit," in Wage rigidity and unemploy- HELLWIG, MARTIN F. "A Model of Borrowing and miient.Ed.: WILFREi) BECKERMAN. Duckworth anid Lendinig with Bankruptcy," Econometrica, Nov. Johnis Hopkinis U. Press, 1986, ch. 5. 1977, 45, pp. 1879-1906. . "Labor Turniover, Inisider Morale and Inivol- HODGSON, GEOFF. "Theoretical and Political Implica- unitary Unemploymenit." Seminar Paper No. 310, tions of Variable Productivity," Cambridge J. Inistitute for IniterniationialEconiomic Studies, U. Econ., Sept. 1982, 6(3), pp. 213-26. of Stockholm, 1984b. HOEL, M. AND VALE, BENT. "Effects of Reduced . "Explaniationis of Uniemploymenit," Oxford Working Time in an Economy Where Firms Set Rev. Econ. Policy, 1986, 1(2), pp. 34-59. Wages," Eur. Econ. Rev., 1986, 30, pp. 1097-1104. . "Efficienicy Wages Versus Inisiders anid Out- HOLMSTR16M, BENGT. "Equilibrium Lonig-Term La- siders," Eur. Econ. Rev., 1987, (Paper presenited bor Contracts," Quart. J. Econ., 1983, 98(5), pp. at EEA Conigress, Vienna, Aug. 1986). 23-54. MALCOLMSON, JAMES. "Unemploymenit anid the Effi- JACKMAN, R.; LAYARD, R. AND PISSARIDES, C. "Poli- cienicy Wage Hypothesis," Econ. J., Dec. 1981, cies for Reducing the Natural Rate of Unemploy- 91(364), pp. 848-66. ment," in Keyneseconomic legacy. Eds.: J. L. BUT- MANoVE, MIC,IAEL. "Job Responisibility anid Promo- KIEWICZ,K. J. KOFORD AND J. B. MILLER. Praeger, tioni: Anl Efficienicy-Wage Anialysis." Mimeo, Dept. 1986, pp. 111-52. of Econiomics, Bostoni U., Bostoni, MA, Apr. 2, JACKMAN, RICilARD AND SUTTON, JOHN. "Imperfect 1986. Capital Markets and the Monetarist Black-Box: MARSIIALL, ALFRED. Principlesof economnics.8th ed. Liquidity Constraints, Inflation, and the Asym- Londoni: Macmillan, 1920. metric Effects of Interest Rate Policy," Econ. J., MEDOFF, JAMES L. AND ABRAIIAM, KATIIARINE. Are Mar. 1982, 92(365), pp. 108-28. Those Paid More Really More Productive?"J. Hu- JAFFEE, DWIGIIT AND RUSSELL, TIoMAS. "Imperfect man Res., Spring 1981, 16(2), pp. 186-216. Information, Uncertainty, and Credit Rationinig," MELNIK, A. AND3 PLAUT, S. "LoaniCommitmenit Coni- Quart. J. Econ. Nov. 1976, 90(4), pp. 651-66. tracts, Terms of Lenidinig, anid Credit Allocation," JOIINSON, D. GALE. "Resource Allocationi Unider J. Finance, Junie 1986, 41 pp. 425-35. Share Contracts," J. Polit. Econ., Apr. 1950, 58, MILDE, HELLMUTII AND RILEY, JOIIN. "Signialinigin pp. 111-23. Credit Markets," UCLA Working Paper 334, 1984. KEETON, WILLIAM R. Equilibriumncredit rationing. MILL, J. S. Principlesof political economy. Lonidon: NY: Garland, 1980. J. W. Parker, 1848. KLEIN, BENJAMIN AND LEFFLER, KEIThI B. "The Role MIRRLEES, JAMES. "Notes oni Welfare Econiomics, In- of Market Forces in Assuring Contractual Perfor- formationi, anid Uncertainity," in Contributions to mance,"J. Polit. Econ., Aug. 1981, 89(4), pp. 615- economic analysis. Eds.: M. S. BALCIi, D. L. 41. McFADDEN AND S. Y. Wu. Amsterdam: North- KLEVORICK,ALVIN K. AND ALCALY, ROGER E. "Judg- Hollanid, 1974. ing Quality by Price, Snob Appeal, and the New . "A Pure Theory of Uniderdeveloped Econio- Consumer Theory," Z. Nationalokon., 1970, mies," in Agriculture in development theory. Ed. 30(1-2), pp. 53-64. L. A. REYNOLDS. New Haven: Yale U. Press, 1975, KNIESNER, TlhOMAS AND GOLDSMIThi, ARThiUR H. pp. 84-106. "Does the Labor Market Clear? A Survey of the MIYAZAKI, HAJIME. "Work Norms anid Inivolunitary Evidence for the U.S.," Res. Labor Econ. Forth- Unemploymenit." Quart. J. Econ., May 1984, pp. coming. 297-311. KREPS, DAVID M. AND WILSON, ROBERT. "Reputation MOENE, KARL 0. "A Note oni Keyniesiani Unemploy- and Imperfect Information, "J. Econ. Theory, Aug. ment as a Worker Discipline Device," Econ. Lett. 1982, 27(2), pp. 253-79. 1985, 18(1), pp. 17-19. KUGLER, P. "Credit Rationing: Evidence from Dis- MOOKIIERIJEE, DILIP. "Iivolunitary Uniemploymenit equilibrium Interest Rate Equations." U. Basel and Worker Moral Hazard." Staniford U., Graduate Working Paper No. 32, 1985. School of Business, May 1985. LAZEAR, EDWARD P. "Agenicy, Earninigs Profiles, MYERS, STEWARTC. AND MAJLUF, NIcIhOLAS S. "Cor- Productivity and Hours Restrictions," Amer. Econ. porate Finaniciniganid Investmenit Decisionis When Rev., Sept. 1982, 71(4), pp. 606-20. Firms Have Informationi that Inivestors Do Not LAZER, EDWARD P. AND ROSEN, SIIERWIN. "RanikOr- Have," J. Finan. Econ., Junie 1984, 13(2), pp. 187- der Tournaments as Optimum Labor Contracts," 221. J. Polit. Econ., Oct. 1981, 89(5), pp. 841-64. NAKAMURA, LEONARD ISAMU. "Customer Credit, Fi- LEIBENSTEIN, HARVEY. Economiiicbackwardness and nianicialIntermediaries anid Real Income: Prelimi- economic growth. NY: Wiley, 1957. nary Evidenice That Credit Matters." Mimeo, LINDBECK, ASSAR. A study in miionetaryanalysis. Dept. of Economics, Rutgers, Oct. 1985 Stockholm: Almquist and Wiksell, 1963. NALEBUFF, BARRYT. AND STIGLIYZ,JOSEPhi E. "Qual- LINDBECK, ASSARAND SNOWER, DENNIS. "Involunitary ity and Prices." Econiometric Research Program 46 Journal of Economic Literature, Vol. XXV (March 1987)

Memorandum No. 297, Princeton U., May 1982. of taxation for developing countries. Eds.: D. . "Information, Competition and Markets," NEWBERY AND N. STERN. Oxford U. Press, Forth- Amer. Econ. Rev., May 1983, 73(2), pp. 278-83. coming. . "Prizes and Contests: Towards a General . "The Social Cost of Labor and Project Evalu- Theory of Compensation and Competition," Bell ation: A General Approach," J. Public Econ., Nov. J. Econ., Spring 1983b, 14, pp. 21-43. 1985, 28(2), pp. 135-63. NEWBERY, DAVID AND STIGLITZ, JOSEPli E. "Share- SALOP, JOANNE AND SALOP, STEVEN. "Self-Selection cropping, Risk Sharing and the Importance of Im- and Turnover in the Labor Market," Quart. J. perfect Information," in Risk, uncertaintyand ag- Econ., Nov. 1976, 90(4), pp. 619-27. ricultural development. Eds.: J. A. ROUMASSET, SALOP, STEVEN C. "Wage Differentials in a Dynamic J. M. BOUSSARD, AND I. SINGII. SEARCA, A/D/C, Theory of the Firm." J. Econ. Theory, Aug. 1973, 1979, pp. 311-41. 6(4), pp. 321-44. . The theory of commodity price stabilization: . "A Model of the Natural Rate of Unemploy- A study in the economicsof risk. Oxford: Clarendon ment," Amer. Econ. Rev., Mar. 1979, 69(1), pp. Press, 1981. 117-25. OSTER,GERRY. "Labour Relations and Demand Rela- SCIILICIIT, EKKEIIART."Labour Turnover, Wage tions: A Case Study of the 'Unemployment Ef- Structure and Natural Unemployment," Z. ges. fect,'" CambridgeJ. Econ., Dec. 1980, 4(4), pp. Staatswiss., June 1978, 134(2), pp. 337-46. 337-48. . "Reference Group Behaviour and Economic PENCAVEL,J. H. "Wages, Specific Training, and La- Incentives: A Remark," Z. ges. Staatswiss., Mar. bor Turnover in U.S. Manufacturing Industries," 1981a, 137, pp. 125-27. Int. Econ. Rev., Feb. 1972, 13(1), pp. 53-64. . "Reference Group Behaviour and Economic . "Industrial Morale," in Essays in labor mar- Incentives: A Further Remark," Z. ges. Staat- ket analysis in memoryof YochananPeter Comay. swiss., Dec. 1981b, 137(4), pp. 733-36. Eds.: 0. ASiIENFELTER AND WALLACE E. OATES. . "Dismissal vs. Fines as a Discipline Device: NY:Wiley, 1977, pp. 129-46. Comment on Shapiro-Stiglitz." Mimeo, Institute PhIELPS, EDMUND S. AND WINTER, SIDNEY G. "Opti- for Advanced Study, Princeton, NJ, Dec. 1985. mal Price Policy Under Atomistic Competition, . "A Link Between Reservation Wage and in Microeconomicfoundations of employmentand Productivity." Mimeo, Institute for Advanced inflation theory. Ed.: EDMUND S. PIIELPS. NY: Study, Princeton, NJ, Feb. 1986. Norton, 1970, pp. 309-37. SCITOVSKY,TIBOR. "Some Consequences of the Habit PRASAD,P. H. Growth with full employment.Bom- of Judging Quality by Price," Rev. Econ. Stud., bay: Allied Publishers, 1970. 1945, 12(2), pp. 100-05. RADNER, Roy. "Rational Expectations Equilibrium: SIIAPIRO, CARL. "Premiums for High Quality Prod- Generic Existence and the Information Revealed ucts as Returns to Reputations," Quart. J. Econ., by Prices," Econometrica, May 1979, 47(3), pp. Nov. 1983, 98(4), pp. 659-79. 655-78. SIIAPIRO, CARL AND STIGLITZ,JOSEPIh E. "Equilibrium REBiTZEER, JAMES B. "Unemployment, Long-term Unemployment as a Worker Discipline Device," Employment Relations, and Labor Productivity Amer. Econ. Rev., June 1984, 74(3), pp. 433-44. Growth." Mimeo. U. of Texas, Nov. 1985. . "Equilibrium Unemployment as a Worker REPULLO,RAFAEL. "A Simple Model of Interest Rate Discipline Device: A Reply," Amer. Econ. Rev., Deregulation." Mimeo, London School of Eco- Sept. 1985a, 75(4), pp. 892-93. nomics and Bank of Spain, Nov. 1985. . "Can Unemployment be Involuntary: Re- RILEY, J. AND ZECKIIAUSER, R. "When to Haggle," ply?" Amer. Econ. Rev., Dec. 1985b, 75(5), 1215- Quart. J. Econ. Forthcoming. 17. RODGERS, G. B. "Nutritionally Based Wage Determi- SIMONDE DE SISMONDI, J. C. L.. Political economy. nation in the Low Income Labour Market," Oxford NY: Kelley, [1814] 1966. Econ. Pap., Mar. 1975, 27(1), pp. 61-81. SLICIITER, SUMNER H. The turnover offactory labor Ross, STEPIIEN A. "The Economic Theory of Agency: NY: Appleton, 1919. the Principal's Problem," Amer. Econ. Rev. May SMITlI, BRUCE. "Limited Information, Credit Ration- 1973, 63(2), pp. 134-39. ing, and Optimal Government Lending," Amer. ROTIISCIIILD, MICIIAEL AND STIGLITz, J. E. "Equilib- Econ. Rev., June 1983, 73(3), pp. 305-18. rium in Competitive Insurance Markets: An Essay SOLOW,ROBERT. "Another Possible Source of Wage on the Economics of Imperfect Information," Stickiness,"J. Macroecon., Winter 1979, 1(1), pp. Quart. J. Econ., Nov. 1976, 90(4), pp. 630-49. 79-82. . "A Model of Employment Outcomes Illus- . "On Theories of Unemployment," Amer. trating the Effect of the Structure of Information Econ. Rev. Mar. 1980, 70(1), pp. 1-11. on the Level and Distribution of Income," Econ. . "Insiders and Outsiders in Wage Determi- Letters, 1982, 10, pp. 231-36. nationi," Scand. J. Econ., 1985, 87(2), pp. 411- SAII, RAAj KuMAR AND STIGLITZ, JOSEPII. "Taxationi 28. and Pricing of Agricultural and Industrial Goods." SPENCE, A. MICIIAEL. Market signaling: Informa- National Bureau of Economic Research, Working tional transfer in hiring and related processes. Paper No. 1338, Cambridge, 1984. In The theory Cambridge, MA: Harvard U. Press, 1974. Stiglitz: The Dependence of Oualitu on Price 47

STIGLITZ, JOSEPli E. "Approaches to the Economics tions to the Credit and Labor Markets,"Amer. of Discrimination," Amer. Econ. Rev., May 1973, Econ. Rev., Dec. 1983, 73(5), pp. 912-27. 3(2), pp. 287-95. . "Credit Rationing and Collateral," in Recent . "Theories of Discrimination and Economic developmentsin corporatefinance. Eds.: JEREMY Policy," in Patterns of racial discrimination. Ed.: EDWARDS, JULIAN FRANKS, COLIN MAYER AND STE- VON FuRSTENBERG. Lexington, MA: Lexington, PIIEN SCIIAEFER. NY: Cambridge U. Press, 1986a, 1974a, pp. 5-26. pp. 101-35. . "Equilibrium Wage Distributions." IMSSS . "Macroeconomic Equilibrium and Credit Technical Report No. 154, Stanford U., 1974b. Rationing." Mimeo. Bellcore, 1986b. Econ. J., Sept. 1985, 95(379), pp. 595-618. . "Credit Rationing With Many Borrowers," . "Alternative Theories of Wage Determina- Amer. Econ. Rev., Mar. 1987, 77(1). tion and Unemployment in L.D.C.'s: The Labor STOFT,STEVEN. "Cheat Threat Theory: Anl Explana- Turnover Model," Quart. J. Econ., May 1974c, tion of Involuntary Unemployment." Mimeo, Bos- 88(2), pp. 194-227. toni U., 1982. . "Incentives and Risk Sharing in sharecrop- . "Wages, Unemployment and Piece Rate: ping," Rev. Econ. Stud., Apr. 1974d, 41(2), pp. Double Asymmetric Information." Boston U. Dis- 219-55. cussion Paper No. 113, July 1985. . "Incentives, Risk and Information: Notes STRAND, JON. "Efficiency Wages, Implicit Contracts Towards a Theory of Hierarchy," Bell J. Econ. and Dual Labor Markets: A Theory of Work Habit Management Sci., Autumn 1975, 6(2), pp. 552- Formation." Mimeo, Dept. of Economics, U. of 79. Oslo, 1986. . "Prices and Queues as Screening Devices TAN, TOMMY CIIIN-CIIIU AND DA COSTA WERLANG, in Competitive Markets." IMSSS Technical Report SERGIO RIBEIRO. "Life Cycle Credit Rationing. No. 212, Stanford U. Aug. 1976a. Mimeo, U. of Chicago, June, 1985. . "The Efficiency Wage Hypothesis, Surplus TAYLOR, JOIIN B. "Aggregate Dyniamics anid Stag- Labour and the Distribution of Income in gered Contracts,"J. Polit. Econ., Feb. 1980, 88(1), L.D.C.s," Oxford Econ. Pap., July 1976b, 28(2), pp. 1-23. pp. 185-207. TODAARO,MICIIAEL. "The Urban Employment Prob- . "Lectures in Macro-economics." Mimeo, lem in Less Developed Countries: An Analysis of Oxford U., 1978. Demand and Supply," Yale Econ. Essays, 1968, . "The Wage-Productivity Hypothesis: Its 8(2), pp. 331-402. Economic Consequences and Policy Implications." . "A Model for Labor Migration and Urban Paper presented at the New York Meetings of the Unemployment in Less Developed Countries," American Economic Association, 1982a. Modern Amer. Econ. Rev., Mar. 1969, 59(1), pp. 138- developments in public finance: Essays in honor 48. of Arnold Harberger. Ed.: M. BOSKIN. Basil Black- VALE, B. "Effects of Bank Reserve Requirements well, Forthcoming. with 'Grey' Credit Markets Under Asymmetric In- . "Alternative Theories of Wage Determina- formation." Mimeo No. 15, U. of Oslo, Aug. 19, tion and Unemployment: The Efficiency Wage 1986. Model," in The theory and experienceof economic WALLACE, NEIL. "A Modigliani-Miller Theorem for development'sessays in honor of Sir W. Arthur Open Market Operations," Amer. Econ. Rev., Lewis. Eds.: MARK GERSOVITZ ET AL. London: Allen June 1981, 71(3), pp. 267-74. & Unwin, 1982b, pp. 78-106. WEISS, ANDREW. "A Theory of Limited Labor Mar- . "On the Relevance or Irrelevance of Public kets." Ph.D. dissertation, Stanford U., 1976. Financial Policy." Paper presented to a conference . "Job Queues and Layoffs in Labor Markets at Rice University, NBER Working Paper No. with Flexible Wages,"J. Polit. Econ., June 1980, 1057, Apr. 1982c. 88(3), pp. 526-38. . "On the Relevance or Irrelevance of Public WEISSKOPF, TIIOMAs E. BOWLES, SUMNER AND GoR- Financial Policy: Indexation, Price Rigidities and DON, DAVID M. "Hearts and Minds: A Social Model Optimal Monetary Policy," in Inflation, debt and of U.S. Productivity Growth," Brookings Papers indexation.Eds.: RUDIGER DORNBUSCII AND MARIO Econ. Act., 1983, 2, pp. 381-450. SIMONSEN. Paper presented to a conference at Rio WESSELS, WALTER J. "The Conitributioni by Firms de Janeiro, Dec. 1981. Cambridge, MA: MIT to Unemployment: A Dynamic Model," Southern Press, 1983, pp. 183-222. Econ. J., Apr. 1979, 45(4), 1130-S50. . "Theories of Wage Rigidity," in Keynes' eco- _ "The Uses and Limits of Unemployment as nomic legacy. Eds.: J. BUTKIEWICZ,K. KOFORD AND a Disciplining Device in the Efficiency Wage J. MILLER. NY: Praeger, 1986, pp. 153-221 (also Model." Mimeo. North Carolina State U., Nov. available as NBER Working Paper 1442). 1986. STIGLITZ, JOSEPII E. AND WEISS, ANDREW. "Credit WETTE, HILDEGARD. "Collateral in Credit Ra- Rationing in Markets With Imperfect Informa- tioning in Markets with Imperfect Information: tion," Amer. Econ. Rev., June 1981, 71(3), pp. Note," Amer. Econ. Rev., June 1983, 73(3), pp. 393-410. 442-45. . "Incentive Effects of Terminations: Applica- WIIITESIDE, HAROLD D. "Wages: An Equity Ap- 48 Journal of Economic Literature, Vol. XXV (March 1987)

proach," J. Behav. Econ., 1974, 3(1), pp. 64-84. YELLEN, JANET. "Efficiency Wage Models of Unem- WILSON, CHARLES A. "Equilibrium and Adverse Se- ployment," Amer. Econ. Rev., May 1984, 74(2), lection," Amer. Econ. Rev., May 1979, 69(2), pp. pp. 200-05. 313-17. YOON, CHANG-HO. "A Reexamination of the theory . "The Nature of Equilibrium in Markets with of Credit Rationing." Mimeo, Stanford U., 1984a. Adverse Selection," BellJ. Econ., 1980, 11(1), pp. . "On the theory of Credit Rationing: Further 108-30. Analysis." Mimeo, Stanford U., 1984b.