«Industrial Policies, the Creation of a Learning Society, and Economic Development1 B. Greenwald and J. E. Stiglitz Industrial policies—meaning ...»
18 Similar results obtain if learning is related to investment, as in Arrow’s original 1962 paper. See Greenwald and Stiglitz (forthcoming).
(a) Learning elasticity—how much sectoral productivity is increased as a result of an increase in labor input.
hk is the elasticity of the learning curve in sector k.
(b) Learning spillovers—the extent to which learning in sector i spills over to sector j.
Hk/Ltj 0, j k, if there are learning externalities, while Hk/Ltj = 0, j k, if there are no learning externalities.
Full learning externalities. One interesting case is that where there are full learning externalities, i.e. knowledge is a public good, so Hk = Hj = H.
Then we choose Lt to
so Ui – v’ + Hi[Σ Lt+1k Uk (Lk t+1 H[Lt]) = 0.
If we assume homotheticity, U = u(Φ (x)), with Σ Φk (x)xk = Φ, then we can rewrite the above as u’ Φi – v’ + hi Uυ = 0 where υ = dln U/dln Φ We can generate the optimal allocation by providing a subsidy of τi on the ith good, for with such a subsidy an individual maximizes U(x) – v(Σxi(1 - τi)) We can get the optimal allocation by setting hi Uυ = v’ τi or τi= hi Uυ/v’ Consumption should be subsidized the more the value of future consumption (the larger ), and the higher the learning responsiveness hi.19 Optimal subsidies with no cross sectoral spillovers, full within sector spillovers. Similar results hold in the case where there are no spillovers across sectors, but there are full spillovers within the sector. A competitive firm again will take no account of the learning benefits—learning is a sectoral public good. We illustrate with the case with separable ulility. With separability of utility across goods (so U = Σui), the first order condition for welfare maximization becomes
The sensitivity of the subsidy to the learning elasticity or to depends on the proportionality variable Uυ/v’. Later discussions in the case of separable utility functions will provide some sense of the factors that determine that variable. See also Stiglitz, 2012.
The optimum can be achieved by setting a subsidy on the consumption of good i at τi= hi ηi uit+1 /v’t Again, it is apparent that, as before, consumption should be subsidized the higher the value of future consumption (the larger ), and the higher the learning responsiveness hi. Now, there is a third factor—the elasticity of marginal utility. If the elasticity is low, then the benefits of learning diminish rapidly.20 The case of full symmetry. In the case of full symmetry (both in consumption and in learning), the only distortion is in the level of output, i.e. if there are n commodities, 1/nth of income will be spent on each, but in a competitive market with full spillovers within the sector, whether or not there are spillovers to other sectors, no attention is paid to the learning benefits. Hence, the market equilibrium will entail too little production (labor) the first period.
Monopolistic competition. In the case of monopolistic competition, where there is a single firm in each sector, and no learning spillovers, the firm will fully take into account the learning benefits, but now, because of imperfections of competition, output will be restricted. There is again less than the socially desirable level of learning.
Differential spill-overs. The formal analysis so far abstracts from the third determinative factor—the extent of spillovers—for we have assumed that there are either no cross-sector spillovers or perfect spillovers from every sector.
There are a variety of reasons that learning may be higher in one sector than another, and why spillovers from one sector may be greater than in other. Historically, the industrial sector has been the source of innovation. The reasons for this are rooted in the nature of industrial activity. Such activity takes place in firms that (relative to firms in the other sector) are (1) large; (2) long-lived; (3) stable; and (4) densely concentrated geographically. Agricultural/craft There is a complicated fourth factor uit+1 /v’t = (uit+1 / uit ) / (v’t / uit) = (uit+1 / uit )(1 - τi), so τi/(1 - τi )= hi ηi (uit+1 / uit ). uit+1 / uit reflects the diminution of marginal utility as a result of increased consumption of good i over time. See Stiglitz (2012).
production, by contrast, typically takes place on a highly decentralized basis among many small, short-lived, unstable firms.
In the following paragraphs we describe in more detail some of the reasons for the comparative advantage of the industrial sector in learning and why that sector is more likely to give rise to learning externalities.
(1) Large enterprises. Since particular innovations are far more valuable to large organizations that can apply them to many units of output than to smaller ones with lower levels of output (see Arrow, 1962b), there is far greater incentive to engage in R&D in the industrial sector than in the agricultural/craft sector. The result will be higher investments in innovation in the former sector than the latter. This can be looked at another way: Large firms can internalize more of the externalities that are generated by learning.21 Moreover, innovation is highly uncertain, and firms and individuals are risk averse. Large enterprises are likely to be less risk averse, and thus better able to bear the risks of innovation. Moreover, because of information imperfections, capital markets are imperfect, and especially so for investments in R & D, which typically cannot be collateralized. Capital constraints are less likely to be binding on large enterprises.
(2) Stability and Continuity. The accumulation of knowledge on which productivity growth is based is necessarily cumulative. This, in turn, greatly depends on a stable organization for preserving and disseminating the knowledge involved and on continuity in jobs and personnel to support these processes. In large organizations, with the resources to provide redundant capacity where needed, the required degree of stability and continuity is much more likely to be present than in small dispersed organizations where the loss of single individuals may completely compromise the process of knowledge accumulation. As a result, steady productivity improvement will be much more likely to arise from industrial than agricultural/craft production. There is another As we noted earlier, it is these learning benefits that help explain an economies industrial structure—the boundaries of what goes on inside firms. In general, the diseconomies of scale and scope (related, for instance, to oversight) are greater in agriculture than in industry. In the case of modern hi-tech agriculture, there are increased benefits of learning, and that will affect the optimal size of establishments.
way of seeing why stability/continuity contributes to learning: As we noted earlier, the benefits of learning extend into the future. Long lived firms can value these distant benefits—and because industrial firms are typically larger, longer lived, and more stable than, say, firms in other sectors, they can have access to capital at lower interest rates.
They are likely to be less capital constrained, act in a less risk averse manner, and to discount future benefits less.22 (3) Human Capital Accumulation. Opportunities and incentives for accumulating general human capital are likely to be far greater in large complex industrial enterprises with a wide-range of interdependent activities than in small, dispersed narrowly-focused agricultural /craft enterprises. (There is, for instance, a greater likelihood of benefits from the cross-fertilization of ideas.) Long-lived stable firms may have a greater incentive to promote increased human capital that lead to greater firm productivity, better ability to finance these investments, and more willingness to bear the risks. The resulting human capital accumulation is a critical element in both developing the innovations on which productivity growth depends and in disseminating them as workers move between enterprises and across sectors.
(4) Concentration and Diffusion of Knowledge across firms. Diffusion of knowledge among densely collocated, large-scale industrial enterprises (often producing differentiated products)23 is likely to be far more rapid than diffusion of knowledge among dispersed small-scale agricultural/craft enterprises. (Recall that earlier we had emphasized the importance of the diffusion of knowledge, and stressed the key role that geographical proximity plays. More recent discussions of the role of clusters have re-emphasized the importance of geographical proximity. See Porter, (1990).
(5) Cross border knowledge flows. While learning is facilitated by geographical proximity, especially developing countries (where many firms are operating far below “best practices”) can learn from advances in other countries. While agricultural conditions The importance of these factors has clear implications for the conduct of macro-economic policy, which we discuss later in this paper.
The fact that they are producing different products enhances the likelihood that they will make different discoveries. The fact that they are producing similar products enhances the likelihood that a discovery relevant to one product will be relevant to another.
may differ markedly from one country to another, the potential for cross-border learning may be greater in the industrial sector; and the existence of large, stable enterprises with the incentives and capacities to engage in cross-border learning enhances the role of that sector in societal learning. Indeed, it is widely recognized that success in the industrial sector requires not just knowledge, but also the ability to acquire knowledge, that is common across borders. Again, some of this knowledge and these abilities are relevant to the agricultural sector, and disseminate to it.
(6) The Ability to Support Public Research and Development. Learning by one firm or subsector spills over to other firms and sub-sectors within the industrial sector, through, for instance, the movement of skilled people and advances in technology and capital goods that have cross-sector relevance. But the benefits spill-over more broadly, even to the agricultural sector, and in the following paragraphs we describe some of the ways that this occurs, especially as a result of the tax revenues that a growing industrial sector can generate. Large-scale, densely concentrated activities are by this very nature far easier to tax than small-scale dispersed activities. Thus, economies with large accessible industrial sectors will be far better able to support publicly sponsored R and D than those consisting largely of dispersed, small-scale agricultural/craft production units.
This factor may be especially important in the support of agricultural research, like that undertaken by Agricultural Extension Service in the United States. These activities directly contribute to agricultural productivity growth, but could not be supported without a taxable base of industrial activity.
(7) Public Support for Human Capital Accumulation. Just as in the case of R and D, private capital market failures may mean that public support in the form of free primary and secondary education is a critical component of general human capital accumulation.
Moreover, the high returns to education in the industrial sector lead to a greater demand for an educated labor force. Again, the greater susceptibility of concentrated industrial enterprises to taxation is key to funding. And again, as workers migrate across sectors, ultimately higher productivity growth in the agricultural/craft sector will be engendered as well.
(8) The Development of a Robust Financial Sector. Greater investment in the industrial sector leads to higher levels of productivity both directly through capital deepening and the embodiment of technical progress (Johansen, 1959, and Solow, 1960), and indirectly through the capital goods industry, which is often a major source of innovation. Some of the innovations here (such as those relating to mechanization) have direct spillovers to the agriculture sector. But so do the institutional developments that are necessary to make an industrial economy function. The heavy investment of a modern industrial economy requires finance, It is not surprising then that an industrial environment should be characterized by a more highly-developed financial sector than an agricultural/craft environment. Once developed, a strong financial sector facilitates capital deployment throughout the economy, even in the rural sector.24 The implication of this analysis is that it pays government to take actions (industrial policies) to expand sectors in which there are more learning spillovers (in the above analysis, the industrial sector; within the industrial sector, there may be subsectors for which the learning elasticity is higher and from which learning spill overs are greater).
Exploitation by money lenders in the rural sector led to the development of rural cooperatives, e.g. in the United States and in Scandinavia.
II. Finance and Industrial Policy One of the reasons that markets fail to allocate resources efficiently to "learning" are capital market constraints. R & D is hard to collateralize, and optimal learning entails expanding production beyond the point where price equals short run marginal costs.
Imperfections of information often lead, especially in developing countries, to credit and equity rationing. Interestingly, a key instrument of industrial policy in East Asia was access to finance, often not even at subsidized rates. (Stiglitz and Uy,1996).
There are several aspects of "learning" in the design of financial policy. The first, emphasized by Emran and Stiglitz, is learning about who is a good entrepreneur. The problem, as we noted earlier, is that because of "poaching" the benefits of identifying who is a good entrepreneur may not appropriated by the lender. There will be too little lending to new entrepreneurs.
Secondly, information is local, which means foreign banks may be at a disadvantage in judging which entrepreneurs or products are most likely to be successful in the specific context of the particular less developed country. Foreign banks are more likely accordingly to lend to the government, to other multinationals, or to large domestic firms. Financial market liberalization may, accordingly, have an adverse effect on development.25 (Rashid 2012).