This action might not be possible to undo. Are you sure you want to continue?

com/locate/jet

**Knowledge hierarchies in the labor market
**

Gilles Saint-Paula, b,∗

a Université des Sciences Sociales, Toulouse, France b CEPR, CES and IZA, France

Received 16 April 2004; ﬁnal version received 8 September 2005 Available online 5 December 2005

Abstract Many workers produce intangible, knowledge-intensive inputs, rather than participating directly in the production process. We develop a model where the labor market organizes itself in a knowledge hierarchy. Skills are segmented into successive clusters. Each cluster buys knowledge from the next one and sells knowledge as a production input to the preceding one. The model is useful to study the impact of improvements in the technology of knowledge transmission on inequality. It is shown that inequality goes up at the top of the income distribution, but not at the bottom. © 2005 Elsevier Inc. All rights reserved.

JEL classiﬁcation: J3; I2; O3; O4 Keywords: Income distribution; Information technology; Computers; Worker displacement; Human capital; Overlapping generations; Knowledge; Worker assignment

1. Introduction Many workers do not directly participate in the production process. Rather, they produce various forms of intermediate inputs that are in turn transformed into other goods by other workers. Furthermore, an increasing share of workers do not produce tangible intermediate inputs; their contribution is better understood as expertise and know-how.

This paper has beneﬁted from suggestions by Michael Piore, Daron Acemoglu, Alessandro Lizzeri, four anonymous referees, as well as comments from seminar participants at Universitat Pompeu Fabra, Barcelona, Humboldt University, Berlin, and the NBER Summer Institute workshop on macroeconomic perspectives. It has been circulated as a working paper under the title “Information technology and the knowledge elite”. Allée de Brienne, 31000 Toulouse, France. Fax :+33 5 61 12 86 37. E-mail address: gilles.saint-paul@univ-tlse1.fr.

∗ Corresponding author. Université des Sciences Sociales, MF 521, GREMAQ-IDEI, Manufacture des Tabacs,

0022-0531/$ - see front matter © 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.jet.2005.09.010

G. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126

105

A consultancy ﬁrm makes creative suggestions about how a ﬁrm could better organize its business. A CEO can use its advice to set strategic goals for the corporation. Executives at a lower-level design the actual practical steps necessary to implement such strategic goals. At an even lower level, people actually process the orders so that a factory be built, supplies be delivered, etc. Production workers then perform the fabrication and shipping tasks. From the consultancy ﬁrm down to the production worker, a long production chain arises where at each stage, people use an intangible “knowledge input” provided by the preceding step to transmit another kind of knowledge input to the next one. Casual observation also suggests that people have greater skills, and are better paid, the higher their position in this ladder. In the academic world, the best researchers invent new theories and make discoveries. These are embodied in textbooks by (often less creative) ones. Other researchers work out the applications of the new theory, or its pedagogical exposition. The textbooks are then used by college teachers for their undergraduates, who will apply what they learned in their professional life. Here again we have an example of a knowledge ladder where more skilled (or talented) workers provide a knowledge input to less-skilled workers in the next stage of the production process. An important determinant of the income of knowledge providers is the number of people who can use their output: a journalist in a local television channel earns less than a journalist in a national one; an economist who explains his ideas in a book will make more money than one who teaches them to 30 students; a consultancy ﬁrm that could embody its recipes in a software will potentially generate a much larger income than if it continues to advise its clients directly. From the point of view of the users, buying a book is cheaper than attending classes, and buying a software is a better deal than paying expensive fees. At the same time, technologies that make the transmission of knowledge cheaper allow users to get it from higher quality providers, thus watching the best actors on TV rather than attending a local comedy theater, buying the stateof-the art accounting software rather than hiring a second-rate accountant, etc. An interesting question is: how do such technologies affect the distribution of income? What happens to the second-rate accountant? To the ﬁrst-rate ones who wrote the software? To the employees of the ﬁrms which use it? How is one’s income affected depending on one’s position in the knowledge ladder? This paper analyzes these issues, deriving predictions for the structure of the distribution of income. Workers have different skills and choose to be either a production worker or a knowledge worker. In both cases, they acquire knowledge from some supplier, which determines their productivity. An important assumption is that knowledge is a “quality” good in the sense that one cannot buy it from several low-quality producers instead of one high-quality one. Two bad consultants would not be able to give the same advice as a good one, and two bad researchers will not make the same discoveries as a good one. Another aspect is that the knowledge input provided by a worker to other workers can only be used by a ﬁnite number of workers. Thus, knowledge is not “non-rival”, but an increase in the productivity of knowledge transmission is captured by an increase in the number of workers that can use it, which makes it more “nonrival”. For example, the invention of printing increases the number of people who can read a given book. The Internet may increase the number of people who can attend a given educational program. I show that—for the parameter values of interest—the economy organizes itself in a succession of clusters of ability levels, called “knowledge ladders”, where a member of a given ladder buys knowledge from a worker in the subsequent ladder and sells it to a worker of the preceding ladder. The return to human capital increases as one moves up the knowledge ladder.

32]) and because of that enjoy a larger increase in wages than the least skilled. Rosen [31.3]. Murphy et al. Garicano [19]. [2. and their implications for the distribution of income and the allocation of talent. Cappelli [11]. These effects are also key to the present paper. such as substitutability between computers and unskilled workers. and at the same time leads to a displacement of the least able workers within each ladder to the inferior one. 5 See Bernard and Jensen [7] for some evidence. for example. as workers with intermediate skills. OECD [30]. Veneri [37]. which intensiﬁes competition among workers specialized in knowledge production. . it seems less related to IT and has perhaps more to do with other factors such as institutional change (as argued by DiNardo et al. I show that such a shift increases the gap in the rate of return to human capital between two consecutive ladders. and in particular is not systematically related to changes in the returns to ability. they gain in absolute terms because of their cheaper access to knowledge. access to software. 2 See Bound and Johnson [9]. improvements in literacy. Mishel and Bernstein [28]. The least skilled do not participate in this competition. but a gain with respect to those who are displaced. [16]. This includes Calvo and Wellisz [10]. Gottschalk [21]. 1 These predictions may shed light on some aspects of recent inequality trends in the US and a number of other countries. which differs from this literature in many respects. trade. [15]). this may capture shifts as diverse as changes in classroom size. Smeeding [26]. [29]. Gould et al. and how it is affected by the productivity of knowledge transmission. knowledge acquisition is cheaper. 4 That is consistent with the paper’s message that better productivity in knowledge transmission displaces some knowledge workers to lower levels. their wages fall. in particular in that it focuses on the organization of the labor market in successive knowledge ladders. One strand has insisted on “span of control” effects.32]. and outsourcing. 3 A body of empirical work challenges the ‘established view’ that the IT revolution has been mostly detrimental to the least skilled and suggests that it is not so much the least skilled who lose from computerization. Those who lose in such competition end up displaced to occupations with a lower knowledge intensity. 4 See for example Howell et al. the invention of printing. these are quite different from the “superstars” effects discussed here. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 Next I consider the impact of an improvement in the productivity of knowledge transmission. we may expect the effects studied here to be at work. 1 Production workers experience a relative loss with respect to knowledge workers who are not displaced to a lower level. the increasing returns they generate. 5 The paper is related to several strands of theoretical literature. While in the past decades workers at the bottom of the distribution have suffered even more. and that their wages may suffer. immigration. and Gould [22]). which makes them more productive. and related papers that focus on how occupational sorting by comparative advantage affect inequality (One may mention Sattinger [35]. Autor et al. [23] also ﬁnd that inequality at the bottom of the distribution of earnings is affected by different factors from inequality at higher levels. which tends to increase inequality. Those who win can spread their ability over a larger market (as in [31. the internet. [25]. as they are not specialized in knowledge production. its impact effect is to reduce wages for some workers who are displaced to lower knowledge levels. Another central result is that while an improvement in knowledge transmission may increase wages for all worker types in the long run. Entorf et al. That is. which reduces inequality between them and the least skilled. Doms et al.106 G. Clearly. Willis and Rosen [38]. etc. Cooke [14]. Gittleman and Howell [20]. 3 One should however keep in mind that there are other effects as well. Howell [24]. [17]. 2 To the extent that one of the effects of the computer revolution is to increase the productivity of knowledge transmission. Krugman [27]. which is modelled as an increase in the number of people who can buy knowledge from the same supplier—depending on the model’s interpretation.

Hence the technical change that we consider is not skilled biased by construction. a young worker must buy knowledge from a fraction s of an old worker. The size of each cohort is ﬁxed and normalized to one. in such a way that the poorest workers cannot be harmed. however. 6 See Saint-Paul [34] for details. In the second period of their life they work and consume. More precisely. Closer in spirit is the paper by Saint-Paul [34]. Section 4 extends the model by allowing for an elastic demand for knowledge. It is distributed over [0.39]. (1) In order to acquire human capital. Furthermore.18. given the assumption of a continuum of agents. Its density is given by f (g) = e− g . This determines their human capital. rather. and how the model can be enriched to allow for growth in productivity and/or population. it explicitly analyzes. Section 6 concludes. where people when young. 2.12. which determines their technology-speciﬁc skill when old. and an improvement in its efﬁciency induces people to buy knowledge from fewer. The overall impact on inequality is similar to my previous paper. In both cases. where s < 1. the young train themselves by buying knowledge. Here they acquire a given level of human capital. 6 Here. work in a given technology. .4. They can either work in the production sector or sell knowledge to the next generation of young. its effect on inequality is a general equilibrium consequence of the easier replicability of symbols and of the importance of knowledge quality. the present model captures the previously ignored phenomenon of worker displacement between knowledge ladders. 7 There is some similarity between this structure and that used by Chari and Hopenhayn [13]. knowledge is a spillover exerted by creative people within networks. which redistributes from “stars” to “superstars”. For notational simplicity we shall assume r = 0.G. 7 The young generation’s agents differ in innate ability g. but better people. where in order to acquire human capital one needs to buy knowledge of some quality from members of the previous generation. In Saint-Paul [34]. An improvement in the efﬁciency of knowledge transmission increases these spillovers. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 107 Another strand discusses how the introduction of new technologies that are more intensive in human capital affects the distribution of income [1. Section 5 discusses transitional dynamics to the steady state. which then determines their occupation and rank in the knowledge ladder when old. In the ﬁrst period of their life. the consequences for the distribution of income of knowledge being spread over more people. The present paper differs from that strand in that rather than assuming a change in key elasticities of the production function. open economy which can borrow and lend at a ﬁxed real interest rate r. which is an overlapping generations model of human capital accumulation. Model setup We consider an overlapping generation. knowledge is traded. Section 3 studies steady-state equilibria. using a single production function. Agents are atomistic and live for two periods. but the economic mechanism analyzed here is quite different. +∞) with an exponential distribution. Section 2 sets up the basic model. knowledge is an indivisible good which is more valued by more productive agents.33. furthermore inequality goes up and then down as knowledge becomes cheaper.

but this number is not inﬁnite. 11 We shall assume that > 1. etc. an old agent of human capital b who works in the production sector produces an output 8 Thus an ‘individual’ old worker of small measure d may supply knowledge to a measure 1/s. the smaller s. 1/s need not be an integer. 9 If b∗ is the human capital of that knowledge supplier. may capture different phenomena. the young worker’s human capital is given by the following human capital production function: b = Beg b∗ . 9 Thus. In the production sector. so that the human capital production function exhibits decreasing marginal returns to the knowledge supplier’s human capital. changes in s. This speciﬁcation implies that knowledge is not a homogeneous input: it has to be bought from a single provider. The parameter s captures the efﬁciency of knowledge transmission.108 G. the allocation of knowledge suppliers to knowledge users is represented by a mapping g ∗ (g). To simplify the exposition we will talk as if an old worker supplies knowledge to exactly 1/s people. Hence. the Internet. who devotes a fraction s of his time to each buyer. atomistic buyer gets knowledge from a single atomistic supplier. the greater the number of people who can use one person’s knowledge. We shall refer to it as “knowledge efﬁciency”. such as increases in classroom size. the production function is linear and given by Y = AH. Consequently. . progress in reproduction technologies. 10 Having e g instead of eg would yield no gain in terms of generality. where (. which guarantees that the cross-sectional moments of productivity are well deﬁned. which is perfectly competitive. Furthermore. although mathematically speaking that is an abuse of language. 10 Furthermore. the mass of people who can get knowledge from a given supplier is 1/s. Clearly in such a continuous setting. it must satisfy. thus transmitting knowledge to a mass 1/s of younger workers. which allows us to pick = 1. where H is the aggregate amount of human capital employed in that sector. for any measurable set S. the lower the working time of an old knowledge supplier that is needed in order to train one young worker—i. the invention of printing. progress in literacy. Depending on the context and on the model’s interpretation. which gives the ability of the knowledge supplier as a function of that of the knowledge buyer. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 that means that (i) a set of old workers of (Lebesgue) measure can supply knowledge to a set of young workers of measure /s > 8 (ii) we consider that a single. knowledge is assumed to be complementary with ability: people with greater ability get a higher marginal return from increasing the skills of the people from whom they learn. the property that (g ∗ (S)) = s (S). although our comparative statics results obviously apply if one restricts oneself to 1/s ∈ N ∗ . there is one degree of freedom in choosing the unit of measurement of g. One cannot substitute two mediocre teachers for a good one. as one can then simply replace g with g and with / . in other words. < 1. In equilibrium. (2) where B and are positive constants. 11 Note that the knowledge input we consider is not a pure public good: more people can beneﬁt from the same supplier as s falls.e.) denotes the Lebesgue measure.d of workers in the next generation.

Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 109 equal to Ab. Bewley [8] and Townsend [36] have developed models of imperfect access to credit markets whose structure is quite similar to OLG models. denoting by bt () the human capital function for the generation born at t. gt∗ () the knowledge assignment functions for the generation born at t. supervision.G. or equivalently g ∗ (g) = g. we must have b(g ∗ (·)) ≡ b∗ (·). ability) of the agent from whom a young agent with skill g buys knowledge. b would then be more broadly interpreted as a set of intangible. computed by applying the knowledge assignment function b∗ () to my own ability. the same aggregate amount of human capital can be obtained by hiring a few high-skill workers or many low-skill workers. (3) The left-hand side (LHS) of this condition is my “teacher’s” human capital. which tell us about the human capital (resp. Contrary to what occurs in the knowledge sector. 12 The questions we are interested in are the following: how does the equilibrium distribution of human capital look like? How are knowledge producers assigned to knowledge consumers? Who specializes in knowledge production? What is the distribution of income? How does a change in s affect these parameters? These characteristics are represented by an equilibrium wage schedule (b). . the right-hand side (RHS) is his human capital. 13 The young’s maximization problem is max b (Beg b ) − s (b). and an equilibrium pair of knowledge assignment functions b∗ (g) (resp. an equilibrium distribution of skills represented by a function b(g) which tells us what is the human capital of an agent with ability g. information.) that allow to boost productivity in the ﬁnal goods sector. (4) 12 One may argue that in such a case an inﬁnite horizon framework would be more appropriate. it may also apply to higher frequency phenomena and be used to understand the economy-wide organization of production. ∗ and bt (). the equilibrium condition must then be ∗ written bt−1 (gt∗ (·)) ≡ bt (·). Such a condition would be written b(g ∗ (·)) ≡ b(·). However. Also. as computed by applying the human capital function b() to his ability g ∗ . this is a model of human capital accumulation over the life cycle by successive cohorts. Steady state We ﬁrst characterize a steady state where these functions are invariant over time. human capital enters in a homogeneous way in the production of physical output. ideas. etc. g ∗ (g)). 13 This condition must not be confused with a one which would state that my knowledge supplier has the same human capital as me. We now analyze the equilibrium values of these functions.. At a more metaphorical level. out of steady state. In that sector.. both computations must match. however. the assignment functions are time-dependent. In this case. Literally speaking. intermediate inputs (organizational skills. 3.

Let us ﬁrst describe the equilibrium in regime I. and a knowledge “chain” arises in a way made clear below. (ii) All knowledge workers are indifferent between knowledge production and output production. ∀b. In such a case.and left. . Proposition 1. b is distributed over [b0 . Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 Assuming that (·) is locally differentiable. respectively. In regime I all workers are indifferent between knowledge production and output production. and the ﬁrst-order condition (5) must be replaced with the following pair of inequalities: R (Be g b ) Be g b −1 s R (b) and L (Be g b ) Be g b −1 s L (b). (iii) The knowledge assignment function g ∗ (g) is given by g ∗ (g) = g + (1 − ) ln . the ﬁrst-order condition is 14 (Beg b ) Beg b −1 =s (b). 14 It turns out that the equilibrium wage schedule we construct is not differentiable at its critical points. however. g (10) Consequently. it is right. In regime II workers are entirely specialized in knowledge production above some ability level. (7) (1− )−1 (1− ) (6) then there exists a steady state such that (i) The wage schedule is given by (b) = Ab. (5) Our central result is that depending on parameter values. (9) (v) The human capital of an agent with ability g is b(g) = B 1− 1 1− s e 1− . the economy can be in one of two regimes. +∞) with density f0 (b) given by f0 (b) = (1 − )b0 1 (1− ) − (1− )−1 b with b0 = b(0) = B 1− ( s ) 1− . s (iv) The knowledge assignment function b∗ (g) is b∗ (g) = Beg s 1 1− (8) .differentiable.110 G. where the R and L subscripts denote the right and left derivatives. Assume s and s.

affects both the mean and elasticity of the production . See appendix. Therefore. which makes it more likely that there are enough workers with ability g ∗ (g) to supply knowledge to workers with ability g. however. they want fewer suppliers but of higher quality. The explanation is as follows. Their wage will have to increase above Ab. since we are not able to characterize an equilibrium for s > . i. An improvement in knowledge efﬁciency. for workers with ability lower than gK = (1 − ) ln s . and they will no longer be indifferent between supplying knowledge and production work. the second effect—i.. 16 (1− ) (1− )−1 . knowledge producers are always indifferent between knowledge and goods production. gK is the ability of the knowledge supplier of lowest quality. i. the demand for knowledge workers is zero. for s below a critical value there will be excess demand for knowledge workers. This validates the assumption that it is equal to A. people buy knowledge from workers with a large ability level. the tail of f (·) is not thick enough. since a positive 15 That is because the analytics are then untractable. Results from a similar partial equilibrium model. In this case. let us ﬁrst take for granted that s . i. The parameter values for which this regime holds are given by conditions (6) and (7). The density of high-quality knowledge suppliers falls more quickly with s than the quantity demanded. (1− )−1 . 16 The role of is more difﬁcult to interpret. and there are fewer people available at this level. and the regime prevails for any value of s . the greater . there will always be enough workers to supply knowledge while earning Ab. which implies that the marginal return to human capital is pinned down by the production sector. the ability distribution has a “thick tail”. there is a “reserve army” of knowledge suppliers at marginal cost sA at any skill level. its marginal productivity in the output sector. homothetically. earning also A per unit of human capital.G.e. is higher.e.e. the lower density of knowledge workers of higher quality. these workers are entirely specialized in production. which is discussed below. as a change in function for human capital. there are more than enough workers willing to supply the amount of knowledge demanded by workers with ability g. At a constant marginal return to human capital equal to A. Consequently. That is. if knowledge transmission is not too If (1 − ) > 1. If the tail of the distribution of g is thick enough. Therefore.e. condition (6) holds for s efﬁcient. i.e. at the same time. the fall in the quantity demanded—is larger than the ﬁrst—i. the “thinner” the tail of f (·). In other words. i. All worker types earn more and relative inequality is unchanged. The economy will be in regime II. But. they require fewer knowledge producers. In this regime. This will be the case if is small enough. The marginal return to human capital is constant throughout the distribution of skills and equal to A. Note that as s becomes small. then condition (6) is always satisﬁed and regime I holds for all s . The critical value of s which (1− ) separates regime I from regime II. This is because while people buy knowledge from more able workers. suggest that for < s workers buy knowledge from less able people. 15 If (1 − ) 1. a reduction in s. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 111 Proof.e. as s is low.. The density of workers around g ∗ (g) is larger than s times the density of workers around g. in equilibrium these workers must be indifferent between supplying knowledge or working in the production sector. At the bottom of the distribution of income.e. To interpret these conditions. Therefore. shifts the distribution of human capital. and therefore income. workers with ability g want to purchase knowledge from workers with ability g ∗ (g) given by (8).

0 = Ab(0) 1 (1− ) −1 i+1 (14) and where k= s > 1. . b(gi+1 )] wages are given by i + Ak i (b − b(gi )). (11) Then there exists a steady state such that (i) The knowledge assignment function g ∗ (g) is given by g ∗ (g) = g + 1 1 ln . gi+2 ]. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 fraction of any skill type works in the production sector. . See appendix. (iv) The human capital of an agent with ability g is b(g) = (Beg ) 1− s (b) = where i − (1− )2 . . . (iii) The knowledge assignment function b∗ (g) is given by b∗ (g) = (Beg ) 1− s 1 1 − 1 (1− )2 . implying g0 = 0. the demand shift in favor of higher quality knowledge suppliers is matched by a corresponding movement from production to knowledge among high quality workers. s (12) (ii) There exists a sequence of critical ability levels (g0 . .) such that (a) gi = −i/ ln s. gi ] and buy it from workers in [gi+1 . can be recursively computed as = i + Ak i (b(gi+1 ) − b(gi )). gi . (c) people such that gi g gi+1 are entirely specialized in knowledge production. The next proposition describes the more interesting regime where some ability levels are entirely specialized in knowledge production. Therefore. 17 At the same time the least able knowledge producers (those whose ability is slightly above gK = (1 − ) ln s ) become entirely specialized in production. . (b) people such that 0 g g1 are entirely specialized in the production of physical output.112 G. Proposition 2. 17 It is easy to check that for g (1 − ) ln . . (15) Proof. Assume s 1 1− (1− ) . They sell knowledge to workers in [gi−1 . (13) (v) Over the interval [b(gi ). a constant fraction s (1− ) −1 (1− ) of workers at any skill level g are s specialized in knowledge production. . without any change in wages. g1 .

If the worker is specialized in knowledge. Hence the marginal return to human capital goes up as one moves up the distribution of skills. This is because the marginal willingness to pay for knowledge tends to accumulate as one moves up the knowledge chain. This pushes up the marginal price of human capital for high quality workers. this marginal return itself reﬂects the marginal willingness to pay for knowledge quality of workers in the preceding ladder. Proposition 2 tells us that the economy then organizes itself into a “knowledge chain”. as illustrated on Fig. The wage schedule is piece-wise linear and convex in the (b. and so on. The quantity k captures the marginal willingness to pay for knowledge quality relative to one’s marginal return to human capital. the marginal willingness to pay of a given worker for an improvement in the quality of his knowledge input is proportional to his own marginal return to human capital. b b*(g) (b) b(g) g (g) g Fig. as one’s position in the knowledge chain is more remote from direct productive activity. importantly. the agents of an economy in regime I try to buy knowledge from workers with an ever greater quality relative to themselves. 1. while workers at the lowest interval of skills [0. a sequence of adjacent intervals of skill levels such that in any interval each worker buys knowledge from a previous generation’s worker of the next interval and sells it to a worker of the next generation in the preceding interval.e. and eventually bump into a supply constraint for quality workers. the marginal return to human capital is increasing as one moves up the distribution of skills. i.e. the wage schedule is piece-wise exponential. and. i. As s becomes low. . i. so that the economy must now be in a regime where the wage schedule is no longer linear. g1 ] are entirely specialized in output production. ) plane. log wages are again piece-wise linear (Fig. (1− )−1 ).G. i. In the (g. greater than one. 1.e. where k is given by (15).e. ) plane. 1). Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 (1− ) 113 Regime II holds if (1 − ) > 1 and s ∈ [0. As one moves up one ladder in the knowledge chain the marginal return to human capital increases by a factor k. This cumulative effect is reﬂected in the k i factor in (14). Hence it tends to prevail for greater knowledge efﬁciency than regime I.

this may capture changes such as the invention of printing. The following proposition characterizes the long-run impact of a fall in s on the distribution of income and human capital. goes up. Proposition 3 analyses the various effects exerted by a fall in s on the long-run distribution of income. It means that the total mass of workers buying knowledge from a given producer. More efﬁcient information technologies induce people to buy knowledge from fewer and better quality suppliers. Consider a reduction in s. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 As is also illustrated on Fig. Assume (11) holds. g ∗ (g)−g would not be constant across the ability distribution. s + .g. Bénabou [5]. Then. implying that each ladder covers a larger segment of the distribution of skills. (iv) There exists some critical level of s. At our level of abstraction. 18 One can deﬁne the “knowledge distance” between two different skill levels as the number of intervals between them. Proof. the human capital acquisition function b(g) and the knowledge assignment function b∗ (g) are exponential. the Internet. (16) then at the bottom of each ladder i 1 there exist worker types who are displaced from segment i to segment i −1 (lower knowledge distance from direct production activity). e. We now study the impact of an improvement in the technology of transmitting knowledge. However. and that the number of workers specialized in output production increases. as in. See appendix. for other distributions. 1/s. 1. These effects are the following. i. (v) There exists s − such that if s s−. increases in classroom size. as captured by a fall in s. Thus the ability gap between a supplier 18 Such properties lie at the heart of most of the results in the literature on segregation. People buy knowledge from more skilled workers (by a ﬁxed additive constant). (ii) Inequality as measured by the relative difference in the marginal return to skill between two consecutive ladders increases. that the knowledge distance between any two types falls or remains constant. etc. the fact that workers with greater ability have a greater marginal propensity to pay for their supplier’s human capital. periods) it takes for knowledge to be transmitted from the interval of the highest level to that of the lowest level. the introduction of television. e-mail. This is also clearly equal to the number of generations (or. The clustering of workers into skill intervals. Proposition 3. This convenient property is due to the exponential distribution. in the long-run steady state: (i) The distribution of human capital shifts up homothetically.114 G. they would eventually beneﬁt from further improvements in information technology. that if s > s + then the wages of all worker types increase. would presumably remain: It is essentially due to the “single-crossing” properties of the human capital accumulation function.e. though. and ladders would have different lengths. and therefore their human capital is below that of their knowledge supplier (by a ﬁxed multiplicative constant). (iii) gi increases homothetically. under our more general interpretation. . and whose wages fall.

In other words.e. Finally. Consequently. and for s small enough workers who are displaced to lower levels in the knowledge chain suffer long-run wage losses (Fig. and a consumer of knowledge goes up. 2. The ﬁrst parts of Proposition 3 imply that inequality between segments increases. when they are stabilized in the directly productive activity.G. Whether one gains or loses depends on whether one is displaced or not. We can use some approximation to say more about the net effect on inequality. 2). i. However. inequality between individuals need not increase. 3). all worker types who were originally specialized in knowledge production experience a marginal wage loss as they cross the boundary to move to a lower ladder. as a low value of s gives them access to ever-better knowledge producers. because segments are wider. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 115 b b(g) (b) g (g) g Fig. so that each knowledge ladder is wider. This process has limits. their income ends up rising without bound. Consider inequality between two people in the ﬁrst segment specialized in . which strengthens the cumulative effects of knowledge demand on the marginal return to human capital at a given ladder. there is a uniform multiplicative shift to the distribution of human capital. The greater ease with which they accumulate human capital compensates for possible wage losses at a given level of human capital. as in equilibrium all workers buy knowledge from more able workers. there is a greater willingness to pay to improve the quality of one’s knowledge supplier. Proposition 3 tells us that if knowledge transmission is initially not too productive. On the other hand. As s goes to zero from s − . so that inequality among knowledge ladders go up. workers at the bottom of a ladder are displaced to the preceding one. depends on one’s relative position in one’s segment. as the knowledge distance between two given types falls. then in the long run all worker types gain from an improvement (Fig. However. however. k goes up.

given that their rank i can be obtained from their skill level g using the formula i = − g/ ln s. with x = AB 1− s (1− ) (s (1− ) − 1). 3. which does not depend on s. as segment 0 gradually absorbs segment 1. Therefore. it ends up reducing inequality among these people. In contrast. if one considers say workers originally located in the two bottom segments. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 b b(g) (b) g (g) g Fig. 20 Interestingly. ﬁrst note that b(g (1− ) s (1− )2 (s − (1− ) − 1). this is what Mishel and Bernstein [28] tend to ﬁnd when looking at correlations between wage changes and investment in technology. Next. The difference equation (14) can be solved to get 19 AB 1− s (1− = −s 1 − )2 g i −s 1 1− (1− ) + s 1 1− (1− ) −s i s . so that (14) can be written i+1 ) − b(gi ) = B 1− s 1 1 − − 1 2 − as i+1 = i + xy i . . more productive knowledge transmission clearly increases inequality at the top of the distribution of income. consider two knowledge producers belonging to different ladders. This quantity is clearly increasing when s falls. Then look for a solution of the form i = 0 + 1 y i . Thus. Their relative wage. According to Proposition 3. (17) Assume that our two producers are located at critical points with values of i large enough for the ﬁrst term in brackets to be dominated in (17). it does not affect inequality between two workers in the bottom segment. expressed in logs. their income is simply proportional to e 1− . Then. the relative log wage of these two workers is (1 − ln / ln s) g. 1 pure production. and y = ks (1− ) = /s. is therefore equal to g/(1 − ).116 G. 20 1 1 − i − 19 To solve it.

actually increases when s falls. As a result a fall in s always reduces the number of knowledge producers. with the only difference that the cut-off points gi are translated to the right by an amount g0 . the number of people with ability above g1 . so that clusters of the same length as in Proposition 2 arise. this depends on the sign of dg1 /ds. dg0 /ds > 0. several studies document the rise in the proportion of non-production workers. is to perform comparative statics with respect to a fall in C. The demand for knowledge has become inelastic as the whole population has been included in the “knowledge economy”. such a change may capture various phenomena: obsolescence of old technologies as they are replaced by new technologies with higher requirements of human capital. Workers in [g0 .e. g1 ] specialize in production but acquire human capital. displacement to the preceding segment does occur. since the effect of the increase in the length of each segment. on whether the increase in the length of each knowledge segment is larger or smaller than the fall in g0 . To do so. In regime II.G. For C high enough some workers prefer not to buy human capital—the “unskilled”. and the previous section’s analysis applies: further improvements in knowledge efﬁciency reduce the number of knowledge producers and workers at the bottom of each ladder are displaced to the preceding ladder. migration of physical capital from old to new technologies. Furthermore. i. the reverse occurs. In practice this may not happen because as knowledge is cheaper some people who previously did not buy it may start doing so. g0 ] arises. workers such that g < g0 . which 21 See Berman et al. An additional cluster of workers with ability g ∈ [0. for large enough knowledge distances i from ladder 0. we extend the model by assuming that workers have the option of not buying it at all. However. Empirically. implying that the total number of knowledge consumers increases when its efﬁciency improves. . Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 117 4. the model’s realism is improved if one introduces some elasticity in the demand for knowledge. i. in the context of this extension. [6] and other references mentioned in the introduction. We can prove that as long as g0 is strictly positive. g0 is eventually equal to zero. As s becomes small enough. where g0 > 0. these do not purchase knowledge and work as production workers. This implies that the total number of knowledge producers. in which case their resulting human capital is Ceg . Workers above that level supply knowledge to the preceding ladder. 23 I am indebted to an anonymous referee for this suggestion. Therefore. the net effect on the number of knowledge producers may be either positive or negative. is multiplied s by i. 22 An interesting application.e. Elastic demand for knowledge A limitation of the model is that the demand for knowledge producers is by construction totally inelastic and always equal to s. and that rather than having people displaced from the bottom of ladder 1 to the top of the output producing segment. the productivity of workers who do not acquire knowledge. as before. 22 These dynamics are similar to what happens to other new technologies such as consumer appliances while their diffusion from the top to the bottom of the distribution of income is completed. 21 Therefore. i. In regime II. where C is a constant. 23 Intuitively. although these are not directly modelled. workers with ability above g0 purchase knowledge from the same supplier g ∗ (g0 ) deﬁned by (12).e. One can then show (see Web appendix) that in both regimes I and II. these will be the least able in society. dg1 /ds > 0. 1 ln 1 .

Such a price fall is captured by a fall in C here. and so do the wages of those who now do but did not prior to the change (otherwise. which would then operate through a fall in the price of goods intensive in unskilled labor. except those who switch from unskilled to the bottom ladder of skilled production workers. This phenomenon unravels up the distribution of income: at the top of each ladder. by an additive amount which is the same for all workers within the same ladder ω b ′0 b0 b ′1 b1 b ′2 b2 b ′3 b3 b Fig. they would have been better-off buying knowledge even before the fall in C). 24 One can show that a fall in C reduces all the gi ’s. 25 See the appendix for the proofs. competition from foreign countries better endowded in unskilled labor. . What happens to wages? The wages of those who do not acquire knowledge fall. The wages of production workers who used to buy knowledge and still do are unchanged. 24 Modelling that as a fall in C is obviously a short-cut. 25 The best endowed unskilled now prefer to acquire human capital. as the alternative option is less attractive. Consequently. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 reduces the productivity of workers who still use the old technology (as in [12]). as one would need several goods to capture this mechanism. hence all the gi ’s fall.118 G. the total number of knowledge workers has to go up (and the number of production workers accordingly falls). One can conveniently show that. 4. and get the same steady-state level of human capital. by a ﬁxed amount. But the most skilled production workers now become knowledge workers to satisfy the increased demand from the previously unskilled. people now supply human capital to a worker who is more highly ranked in the knowledge hierarchy. including g0 . implying a fall in g1 . workers acquire human capital from the same supplier as before. One can show that the wages of all knowledge workers go up.

Thus. on the distribution of income and the occupational pattern. a reduction in the required working time input in the production of human capital. 2. is concentrated at the bottom of the distribution of income. Intuitively. for whom it is too late to acquire human capital. the lower gi s imply that those sufﬁciently high in the knowledge hierarchy indirectly supply human capital to a greater number of workers. in a world where knowledge is a quality good entering as an input in the production of human capital. they may be worse-off in consumption terms. Two important results are as follows. a fall in s cannot be Pareto-improving. however. though. The model allows to analyze the effect of an improvement in information technology. One may also introduce growth in total factor productivity as well as in population size. Concluding remarks This paper has developed a theory of the distribution of income. First. Other extensions One can also implement two other extensions of the model (these are worked out in the Web appendix). Second. 1.e. As workers have to pay more to acquire human capital. The shift in the wage structure implied by a fall in C is illustrated in Fig. ) plane. and if their ladder i is high enough. where worker types who are displaced are compensated by a higher level of human capital. people are willing to pay more to increase the quality of their knowledge supplier. in response to a fall in s. and thus earn the same wage. because of lags in the positive supply response of human capital. The central result of this paper is that the impact of an improvement in information technologies on a given worker’s income depends on this worker’s position relative to his or her ‘peer group’ 26 It is not possible to represent the wages of the unskilled in this plane: while they all have the same human capital. 26 Wages are different from consumption. they have different abilities and different wages. Because of that. 5. What about knowledge workers? One can show that consumption goes up for workers who are above a certain level in the knowledge hierarchy. A key result is then that TFP growth increases the demand for human capital by boosting its future return relative to its present cost. inequality overshoots its long-run level. the loss in physical output triggered about by the fall in the unskilled’s productivity and the reduction in the number of production workers. which includes the steady-state distributions for any parameter set.G. and of the assignment of workers between output and knowledge production. . 4 in the (b. represents the increase in the number of workers who acquire human capital. but they have to pay more for this human capital. That stands in constrast to the comparison across steady states. x 6. which makes the wage schedule steeper and more conve˙ . The left-ward shift of the ﬁrst point in the curve. but falls for those who are below that level. That is not true of the ﬁrst generation of old. that effect is stronger than that of their own higher cost of acquiring knowledge. This is clearly the case for production workers: they acquire the same amount of human capital. workers at the top consume more despite the decline in total physical output. because the initial generation of old who are displaced from segment 1 to segment 0 unambiguously loses. i. The transitional dynamics to the steady state can be fully worked out. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 119 but goes up exponentially with the rank of that ladder. provided the initial distribution of human capital belongs to a certain class.

one should ﬁnd other culprits for what has happened at the low end of the distribution of income. First. one indeed has an equilibrium. that the supply of old workers in any given interval is actually larger than the demand for knowledge producers coming from the corresponding young workers. Conversely.e. Substituting (9) into (2) proves the ﬁrst part of (v). Proof of Proposition 2. Then. Workers at the bottom of the distribution of income gain. Thus. Proofs Proof of Proposition 1. then the young’s optimization problem is concave and the ﬁrst-order condition (5) is necessary and sufﬁcient. Finally. Their ability must lie between g ∗ (g) and g ∗ (g) + g ∗ (g) dg. . that the knowledge supplier of the lowest ability is indeed available on the market.120 G. one has to check that b∗ (0) b0 . Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 of other workers in the same knowledge ladder. Those who are potentially harmed are found at the bottom of each ladder. but these are people at intermediate and high income levels. Using the ∗ (g) it is easy to see that this is exactly equal to s e− g dg.1. we have to show that this is indeed an equilibrium. It will be greater than the demand for knowledge producers of that ability range if and only if s e− g e− g − (1− ) s . s. while substituting (10) into the density of g proves its second part. applying (3) yields (iii). Let us start from the knowledge assignment function in (i) and show that it is supported by a competitive equilibrium which satisﬁes (ii)–(v). This is equivalent to B s i. If (b) = Ab throughout. i. note that in such an equilibrium the supply of knowledge produced by any type g ∗ matches the demand coming from the corresponding type g. while those below g1 produce output. as they are consumers rather than producers of intangible goods. It is easy to check that it is equivalent to (9). and there are e− g g ∗ (g) dg such workers.e. while information technology may have contributed to the observed overall increase in inequality. i. They buy knowledge from s e− g dg old agents. They buy it from ∗ workers between g ∗ (g) and g ∗ (g + dg). which proves (iv). we see that the total supply of such workers is e− g ( s )− (1− ) . If this paper message’s is to be believed.e. Appendix A A. To see this. There are e− g dg young workers between g and g + dg. To complete the proof. which is clearly equivalent to (6). it is straightforward to check that if all these conditions hold. note that there are e− g dg workers between g and g + dg. all workers above deﬁnition of g g ∗ (0) = g1 = −(ln s)/ are entirely specialized in knowledge. 1 1− B 1− 1 1− s . and that these workers buy knowledge from s e− g dg suppliers. Using (iii).

these two conditions are equivalent to (19). that (iii) holds).G. Given that g < gi+1 . individual g must pay a marginal wage at least equal to Ak i+2 . individual gi+1 buys knowledge from individual gi+2 . and one has a local optimum if and only if the right-derivative (resp. Consequently. ﬁrst note that if g ∈ [gi . b(gi+2 )]. In the special case where g = gi . Next. This is because over that interval the ﬁrst term in (18) is linear in its argument.e. the objective function is not differentiable. But since b(gi+1 ) ∈ [b(gi ). These two conditions are Ak i+1 Beg b∗ and Ak i Beg b∗ −1 −1 − sk i+2 0 − sk i+1 0. (19) This is equivalent to b (g) = ∗ Beg sk 1 1− . Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 121 Next. i.. b(gi+1 )] and since b∗ (g) maximizes g’s welfare for the resulting b in [b(gi ). (18) Our ﬁrst step is to show that b∗ (g) satisﬁes the ﬁrst-order conditions. individual g must buy knowledge from some g > gi+2 . b(gi+1 )]. b(gi+1 )]. then b∗ (g) ∈ [b(gi+1 ). . i. i.. in order to get the desired level of human capital. b(gi+2 )]. it must be the case that Ak i+1 Beg b −1 − sAk m < ˜ 0 for all b such that b ∈ [b(gi+1 ). along with (i). which. which gives him a level of human capital equal to b(gi+1 ). gi+1 ]. with m i + 2. People elect the human capital level of their knowledge supplier b∗ by maximizing max ∗ b (Beg b∗ ) − s (b∗ ). b(gi+1 )] it is clearly an optimum among all the values of b∗ such that the resulting human capital of the individual with ability g remains within [b(gi ). b(gi+2 )] interval. to get a marginal wage equal to Ak i+1 . ˜ Consider now what happens if individual g tries to get a level of human capital b in the [b(gi+1 ). Together. we prove that this is a global optimum. given the deﬁnition of k. positive) or zero. implying that the FOC is Beg b∗ −1 − sk = 0. is equivalent to (iii). this clearly generates less utility than picking b∗ (g). Under our assumed knowledge assignment function. we show that this allocation of resources is the outcome of individual optimization if wages are determined by (v) and if human capital is related to skills by (iv) (which implies. equal to Ak m . Given the convexity of the wage schedule and its linearity within each interval [b(gi ). left-derivative) is negative (resp. Given that b > b∗ (g). so that the objective function is concave.e. from some b > b(gi+2 ).e. To see this. To satisfy the ﬁrst-order condition (FOC) it must maximize Ak i Beg b∗ − sAk i+1 b∗ . which implies (by continuity and convexity of the ˜ wage schedule) that the best individual g can do is to pick up b = b(gi+1 ). that it is a local optimum.

Thus.e. all these optima are inferior to the one where the agent picks b∗ (g).O ]. Using the recursive law (14).O ]. i+1 i +AkO bO (gi+1. assume that it is satisﬁed for i. (20) where subscript O (resp. Let us now prove (iv). We prove this inequality by induction. N) refers to variables and functions corresponding to s = sO (resp.. ˜ A similar induction argument can be made with respect to values of b in inferior intervals.O − Ak0 [bO (giN ) − biO ] . which yields a utility no greater than the optimum in the preceding interval [b(gj ). j i.O i+1 + AkO [bO (gi+1. given this knowledge assignment function and this wage schedule.N − biN ] i AkO [bi+1.122 G. This implies again that the optimum in that interval is b = b(gj +1 ). and biO = bO (giO ). we see that this is equivalent to iN i + AkN [bi+1. we have to prove that iN iO i + Ak0 [bO (giN ) − biO ].N ) − bi+1. This in turn validates (iii) and (iv). implying that the marginal wage he pays is greater than the one he gets by a factor at least as large ˜ as k. individual g must buy human capital from an agent with human capital greater than b(gj +2 ). (i)–(iii) follow directly from Proposition 2. Consider a fall in s from sO to sN < sO .N − biN ] iO i + AkO [bi+1. while the wage schedule implies the specialization pattern in (ii).N ) − bi+1. Let us assume (11) holds. we prove that the new critical types gi all gain.N ) − bi+1.O − biO ] i+1 +AkO [bO (gi+1. then a sufﬁcient condition for (21) to hold is i AkN [bi+1. Therefore. We would like to show that it is then satisﬁed for i + 1. by induction. First. To prove that all worker types gain. that i+1.O − biO ] i+1 i = AkO [bi+1. b(gj +2 )]. it is enough to prove that its kinks are all above its previous location to prove that it shifts up everywhere (otherwise convexity would be violated). it is indeed optimal for each g to buy knowledge from b∗ (g).O ]. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 ˜ This argument can be extended by induction as follows: to reach some b in [b(gj +1 ).. s = sN ). Next. This formula says that the new wage at kink i must be greater than the wage of the same type in the old wage schedule (where this type was in the interior of ladder i). Because the wage schedule is piece-wise linear and convex in eg/(1− ) .N ) − bi+1. (21) If (20) holds. Therefore. Proof of Proposition 3. Thus.O − bO (giN )] + AkO [bO (gi+1. note that it is trivially satisﬁed for i = 0. we do have an equilibrium. b(gj +1 )].N i+1. i. where sN is arbitrarily close to sO .

(22) holds for s ∈ [s + . has 1 − (s (1− ) 1 1 1− (1− ) 1 1− (1− ) 1 < it holds since . s1 ]. Thus we have kN ≈ k O 1 − giN ≈ giO + i / . Consequently. as well as . with 0 < s0 < s1 < ∞. His wage is simply equal to N (b1N ) = 1N 1 1 − 2 1− s (1− ) . one 1 − (1− ) − ). = AB N = Ab1N (23) 27 This is done by expressing each new variable as a function of the corresponding old one. 28 Remember that i . (22) This will hold for all i 0 if the term in bracket is always positive as well as the sum of all the other terms. For the term in brackets. note that for s = 1 1 1− (1− ) the LHS is equal to and the RHS is smaller than . bO (giN ) ≈ biO 1 + biN ≈ biO 1 + etc.G. and ignoring any second-order terms in . Next. so that this deﬁnes a (possibly empty) interval [s0 . s0 < < s1 and this inequality holds over some interval of values of s greater than zero but smaller than the maximum for which (11) holds. We ﬁrst consider the wage of agent g1N . this is equivalent to s 1 1− (1− ) s 1− 1 (1 − ) + (1 − ) . (1 − ) −1 . so that the RHS is a linear combination between s and . 28 Therefore. i + (1 − ) (1 − )2 1 1 1− (1− ) is the highest possible value of s for which (11) holds. . Let −1− (1− −1 + )s 1 (1− ) ) [s + . substituting into (21). (1 − ) . we get −1− (1 − ) s −1 + 1 (1 − ) + 1 (1 − )2 s 1 − (1− ) − 0. who deﬁnes the ﬁrst kink in the new wage schedule. 1 1 1− (1− ) > 0. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 123 Using the formulae of Proposition 2 and assuming that sN = sO (1 − ). Turning now to the last term (s 1 1− 1 ] be such an interval. with that this is equivalent to 27 i s 1 − (1− ) 1. This completes the proof of (iv). always positive as s < 1. Note also that the LHS is increasing and concave in s. 1 Note that (11) implies (1− ) < 1. it is (1− ) ]. In this interval. Let us now prove (v).

and if the LHS of (22) is negative at i = 1. The skill content of recent technological change: an empirical exploration. which guarantees that the ﬁrst term in brackets in (22) is negative. To prove that the wages of agents displaced from higher segments also fall for s small enough.09. Changes in unemployment and wage inequality: an alternative theory and some evidence. and that his human capital b(g) then goes to inﬁnity. Using again sN = sO (1 − ). Consequently. Murnane. Please visit doi: 10. [3] D. Appendix B. It can be unraveled provided the LHS of (22) is negative for any i. this always holds provided s is lower than some s2 . then the wages of this type of agent falls. NBER Working Paper.jet. Econ. with sO 1 1− (1− ) 1 we get that this is equivalent to (1 − ). Supplementary data The online version of this article contains additional supplementary data. to prove that wages eventually rise. Finally. Re-assessing the revisionists. Autor. (1 − )2 (1 − ) Given that s 1 (1− ) < 1. 2004. 89 (5) (1999) 1259–1278. Autor. This will work provided s < s + .2005. F. Thus (v) holds for (1− ) s − = min(s2 . Comparing this with (23).010 References [1] D.124 G. M. one can just replicate the above proof by induction. Acemoglu. inverting signs. R. Amer. . [2] D. Rev. [ (1 − )] (1− )−1 ). (2003). if this inequality is violated.1016/j. s + . or equivalently s 1 1− (1− ) 1+ 1 (1 − )2 +s (1 − 2 ) <s+ . Katz. note that any ﬁxed worker type g ends up in the directly productive activity as s goes to zero. Quart. as does his wage Ab. we see that the wage of this type of worker will increase if and only − 1 (1− )2 1 − (1− ) sN − sO 1 − (1− ) − sO sO (1−2 ) −1 (1− )2 . Kearney. meaning that sO is too small. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 His wage prior to the fall in s can be computed using the relevant formula for ladder 1: O (b1 ) = 1O + AkO [bO (g1N ) − b1O ] 1 − 1 (1− )2 = AB 1− sO if sN − 1 (1− )2 +A sO 1 (1− ) −1 [B 1− sO 1 − 1 (1− )2 − (1− ) sN − B 1− sO 1 − 1 (1− )2 ]. Econ. J. Levy. L. The preceding condition guarantees that the induction process can be initiated.

skill upgrading. and the diffusion of new technology. (2000) (http://www. (1993) 515–530. R. Rosen. University of British Columbia. 1994. J. in: Reducing Unemployment: Current Issues and Policy Options. Smeeding. and income distribution. Bernard. M. J. Econometrica 43 (3) (1975) 455–468.B. Economic Policy Institute WP #217. The information technology workforce. Economy 87(5) (Part 1) (1979) 991–1010. Cooke. 1994.H. [13] V. 1997. Labor Econ. Technological Revolutions. 1994. [26] A. J.V. J. Econometrica 64 (5) (1996) 1001–1044. Amer. Polit. Low wages in the US and high unemployment in Europe: a critical assessment of the conventional wisdom.cisp. 49–80. A. Econ. Rev. B. Shleifer. Quart. pp. B. Econ. J. Polit. Gittleman. Amer. wage inequality and growth. ability.-A. Hierarchy. D. DiNardo. wages. Wellisz. Moav. Bernstein. [35] M. and the distribution of earnings. growth. Beaudry. Sattinger. N. The economics of superstars. M. Vintage human capital. Changes in the structure of wages in the 1980s: an evaluation of alternative explanations.R. T. Howell. Are skill requirements rising? Evidence from production and clerical jobs. (1993). Is the technology black box empty? An empirical examination of the impact of technology on wage inequality and the employment structure.D. Workings of a city: location.B. Krugman. comparative advantage. [23] E. Centre for Economic Policy Research Discussion Paper: 1972. with an application to ideas and networks. Center of Economic Policy Analysis Working Paper #5. 106(2) 503–530. J. Weinberg. [5] R. and production. 13 (2) (1982) 311–323. [14] S. 1998. Vishny. Berman. Kareken. . K. Dunne. [19] L. [20] M. 89 (1) (1999) 78–102. Coping with technological progress: the role of ability in making inequality so persistent. Gottschalk.R.htm). Jensen. Gollac. Chari. Jackson Hole. J.M. H. New technologies. [8] T. pp. Polit. Econ.R. inequality. Fortin. Labor market institutions and the distribution of wages. [10] G. J. [31] S. Rubinstein. Calvo. Ind. 1998. Econ. J. 20 (2002) 105–147.). Bénabou. Bell J. Cappelli. Economy 99 (6) (1991) 1142–1165. J. Rev. 1980. Exporters. [24] D. Changes in the structure and quality of jobs in the United States: effects by race and gender 1973–1990. Quart.R. Rev. Changes in the demand for skilled labor within US manufacturing industries: evidence from the Annual Survey of Manufacturing. Garicano. Econ. Ability biased technological transition. Institutional failure and the American worker. Murphy. J. Howell. Econ. S. Economy 108 (5) (2000) 874–904. Labor Econ. The allocation of talent: implications for growth. Doms. Gould. Wages and employment in the US and Germany: what explains the difference?. Quart. 17 (3) (1999). Griliches. Rev. G. Econ. Minneapolis. Duncan. control. [21] P. Have computers changed the wage structure?. and the wage gap. Caselli. [7] A.PDF) [29] K. and technological progress. Quart. [25] D.W. Green. in: J. Cross-national comparisons of earnings and income inequality. Saint-Paul. Workers. T. Federal Reserve Bank of Kansas City. Authority. Entorf. 48 (3) (1995) 420–444. Econ. Growth 6 (2001) 285–316. 112 (1) (1997) 253–290.com/epi/virlib/Technical/1994/isthetechnology.org/imp/april_2000/ cooke/04_00cooke. [6] E. Hierarchies and the organization of knowledge in production. Rev. Rising wage inequality. O. 1998. Lit. Jerome Levy Institute Brief #29. 82 (3) (1992) 371–392.G. Gould. wages and technology. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 125 [4] P. [9] J. 109 (1994) 367–397. mimeo. D.M. Lemieux. J. Johnson. (1993). and the growing importance of general skills in the United States. Labor Relat. Bound. Mishel. Moav. Polit. T. Econ. [33] Y. [11] P. [30] OECD. (http://www. [12] F. education. 2000. On the distribution of income and worker assignment under intra-ﬁrm spillovers. 169–210. and worker selection. Bewley. Models of Monetary Economies. 42 (1997) 3–31. Hopenhayn. Z. D. Int. Econ. The optimum quantity of money. Harrison. [27] P. Quart. 71 (5) (1981) 845–858. N. [34] G. [28] L. [18] O. [15] J. Precautionary demand for education. [17] H. Bound. Past and prospective causes of high unemployment. Galor. Troske. Tel Aviv Foerder Institute for Economic Research and Sackler Institute for Economic Research Working Paper: 27/98. Tsiddon . J. Howell. Labor Relat. OECD Jobs Study. J. Comparative advantage and the distributions of earnings and abilities. [32] S. 1973–1992: a semiparametric approach. Krueger.lights. J. Economy 109 (1) (2001) 1–37. Kramarz. J. O. Federal Reserve Bank of Minneapolis. [22] E. IMP Mag. Ind. Econ. Wallace (Eds. Rosen. Econ. Amer. F. [16] M. 35 (1997) 633–687. J.

Veneri. Economy 113 (4) (1999) 1091–1117. J.M. . 1980. 265–304.J. Federal Reserve Bank of Minneapolis. Models of Monetary Economies. jobs of tomorrow: opportunities in information technology. (1998). Wallace (Eds. Quart. [39] J. Townsend. Education and self-selection. J. S. Polit. pp.). N. Models of money with spatially separated agents.H.126 G. Rosen. Outlook Quart. Here today. [37] C. machines and economic growth. Minneapolis. Kareken. in: J. Willis. [38] R. Occup. Economy 87 (5) (1979) S7–S36. Zeira. Saint-Paul / Journal of Economic Theory 137 (2007) 104 – 126 [36] R. Workers.

- BP Human Capital
- Human Capital 1
- Human Capital
- km
- Human Capital n Corporate University_Dr. Donny Syafardan
- 659338
- Intangibility and Portability
- Westphalen
- Dp 06585
- Human Capital and Performance
- Intro to Economics - Productivity and Human Capital
- Human Capital
- Human Capital Investment- Theodore W
- Competency Approach to Human Resources Management
- sem II eco
- Full Text
- Human Capital Fertility and Economic Growth
- Introduction to Business
- Sinisa Zaric and Vojislav Babic Faculty of Economics University of Belgrade, Samara Text Eng
- Chap 004
- pianalto_20030510.pdf
- Information Technology Capability and Value Creation
- Chapter 4
- HumanCapitalWhitePaper_2007_01_26
- Industry-Specific Human Capital
- 996.full
- Human Capital Profile
- Chapter6 US
- HR in Thirukural
- Thanks to the Accenture Human Capital Development Framework
- Inequality

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.

scribd