Stiglitz on Second generation reform(1)

["Z.CUI" <zcui@xxxxxxx>]

>Distribution, Efficiency and Voice: Designing the Second Generation of
>                                          Reforms
>
>
>                                         Joseph Stiglitz
>                           Senior Vice President and Chief Economist
>                                       The World Bank
>
>                                     Brasília, Brazil, July 14, 1998
>
>
>
>     Introduction
>
>     One of the main insights of modern economics is overturning the
traditional neo-classical idea that issues of
>     efficiency and equity could be completely separated. In the 1950s the
Second Fundamental Theorem of Welfare
>     Economics established that any Pareto efficient allocation could be
supported as a competitive equilibrium, given
>     appropriate lump sum taxes and transfers. This result had profound
implications: it stated that we could let the
>     government achieve the equity it desired by redistributing resources
and then leave the market to achieve an
>     efficient equilibrium. But the Second Fundamental Theorem was also
based on extremely strong assumptions.
>     Most obviously, in practice the government cannot use lump sum taxes
and transfers to serve distributional goals;
>     and basing taxes and transfers on income and other variables that are
affected by individual action is necessarily
>     distortionary.
>
>     More profoundly, developments in the economics of information have
established that in the presence of
>     imperfect information, that is always, markets are not even
constrained Pareto efficient; there is some government
>     action that, while itself bound by the same informational limitations,
can make at least some people better off
>     while not making anyone worse off (Greenwald and Stiglitz 1986). A
direct implication of this result is that
>     distribution does matter; changing the distribution of income or
wealth can affect the efficiency of the economy. In
>     a sense this result was obvious: economists complain bitterly about
the distortionary effects of a 50 percent tax
>     on income, and it should have been equally clear that a sharecropping
contract that required that the tenant pay
>     50 percent of his returns to the landlord would discourage work,
investment in the land, and innovation. Enabling
>     sharecroppers to buy their land could result in greater agricultural
productivity, resulting in what would seem (at
>     least with compensation) a Pareto improvement. 1
>
>     In the first part of this talk I will generalize on this issue to make
some points about the effects of inequality in land
>     and other assets on efficiency. But efficiency is a relatively narrow
goal. The process of development is about
>     much more than just increasing measured GDP. In the second part of
this talk I will discuss the second
>     generation of reforms, which go beyond the so-called "Washington
Consensus" in seeking broader goals of
>     development and employing a wider range of instruments in doing so.
One of the widely recognized objectives of
>     the second generation of reforms is institutional development, the
topic of the third part of my talk. I will extend
>     the argument in the first part of my talk that equity affects economic
efficiency, to argue that participation and
>     voice affect institutional performance, including the goals of
institutions and their effectiveness in achieving those
>     goals.
>
>     Equity Affects Efficiency: The Case of the Distribution of Land
>
>     Inequalities in wealth will generally require the owner of assets, the
"principal," to delegate the use of assets to
>     another, the "agent." In the presence of incomplete information, it
will be impossible to write down a complete
>     contract to specify what actions the agent should undertake in each
contingency or at least impossible to monitor
>     (and thus enforce) that contract. As a result, a principal-agent
problem arises, creating what are sometimes called
>     "agency costs." Agency costs and the steps taken to mitigate them,
like engaging in costly monitoring, affect
>     efficiency in a number of ways. If the extent of inequality affects
the extent of agency costs, then it will also affect
>     the overall efficiency of the economy.
>
>     Perhaps the clearest illustration of the principal-agent problem is
the relationship between the landlord (the
>     principal) and his tenant (the agent).2 If the landlord could
costlessly monitor the tenant, his contract would
>     specify the amount of labor that the worker would provide. But
monitoring is costly, and hence the landlord
>     designs contracts like sharecropping3 that ensure that the worker has
an incentive to make an effort. These
>     contracts, however, are imperfect for several reasons:4
>
>        i.The sharecropper only receives a fraction (usually one-half to
two-thirds) of his or her marginal product
>          and will thus undersupply effort.
>        ii.The sharecropper will not enjoy the benefits of improvements to
the land (or bear the costs of
>          deterioration) and thus will underinvest in the land by, for
instance, not taking sufficient care to avoid
>          erosion. Similarly, the sharecropper will have insufficient
incentives to invest in other inputs, like better
>          seeds or fertilizer.5
>        iii.The sharecropper is typically more risk averse than the
landlord, but the sharecropping contract does not
>          allow him or her to transfer this risk. Also, the different risk
incentives mean that the tenant may use
>          techniques (such as the choice of crop or fertilizer) that have a
lower risk, but also a lower expected
>          return.
>
>     Thus, inequality in wealth or land distribution led to an economic
organization which had markedly adverse
>     effects on output.
>
>     These narrow economic concerns with efficiency may not have been the
primary motivation for land
>     redistribution, but they provide the hope that land redistribution
would lead to increases in output and standards
>     of living. Yet many land reform initiatives around the world have been
disappointing. In many cases output has
>     actually fallen. Also, the effects of land redistribution have often
proven only temporary, with land becoming more
>     concentrated again as time passed. This is why many recent reforms
have imposed restrictions on sales of land,
>     although, as I discuss later, they have often been unsuccessful in
stemming the reconcentration of land and have
>     brought their own efficiency problems.
>
>     The two problems are related, because the falling output induce
farmers to resell the land. To put it another way,
>     evidently the value of land under tenancy seemed to be higher than the
value of land on self-managed farms. How
>     can this be?
>
>     Increasing Output After Land Reform
>
>     Again, the concept of agency costs provides us insights into the why
this may be so. I have just been discussing
>     the distortions which arise from the agency problems associated with
the disparity in the ownership of "labor" and
>     land. But there are agency costs associated with the distribution of
wealth more generally. Successful agricultural
>     production requires not only labor and land, but also capital and
know-how (technology). Poor tenants are often
>     lacking in all three and simply granting them land does not resolve
the other two gaps. The agency problem
>     associated with capital is as serious as that associated with land.
Credit rationing (see Stiglitz and Weiss 1981),
>     resulting from the combination of limited liability (the borrower does
not bear the full cost of failure) and the
>     costliness of screening and monitoring for riskiness, ensures that
many poor tenants will not have access to credit
>     or will be forced to borrow from informal lenders at very high
interest rates.6 Thus, poor farmers have been
>     unable to buy the high-cost seeds, fertilizer, or small-scale tractors
necessary for productive farming.7
>
>     In the absence of government action, a more egalitarian distribution
of land can actually increase the agency costs
>     vis-à-vis technology. Knowledge, as is now widely recognized, has the
properties of a public good (when an
>     additional person uses the knowledge, it does not detract from what I
know; and the costs of exclusion are often
>     high). As such, it will be underprovided privately. If there were a
single large landowner, he could capture the full
>     benefits of investing in technology. It does not pay, however, for
each tenant to invest in knowledge on his own
>     and thus innovation will be inhibited. Each small farmer may wait for
his neighbor to try the new seed to see if it
>     works. Each wants to free ride off of the innovation efforts of
others. Land reform can exacerbate the free rider
>     problem, leading to less innovation and even delays in implementing
best practices. This is another example of a
>     way in which, contrary to neo-classical economics, distribution
affects efficiency.
>
>     Both of these problems can be remedied by the design of the land
reform program and by appropriate
>     government interventions. The World Bank, for instance, is supporting
the Brazilian government in accelerating its
>     land reform program through a market-based mechanism in which
communities of landless rural workers
>     themselves select and negotiate the purchase of land. Not only does
this approach increase the ownership,
>     literally and more broadly, of the new small landowners, but it also
eliminates the conflictual aspects of land
>     reform. Together with extended credit and an increased emphasis on
education, the initial indications are that this
>     project will be able to help thousands of families, increasing their
productivity and incomes, allowing them to
>     accumulate enough savings to repay the original loans used for the
purchase of the land.
>
>     A number of countries and programs, most notably the Grameen Bank in
Bangladesh, have demonstrated that
>     credit can be made available to small farmers on reasonable terms,
with high repayment rates.8 Peer monitoring,
>     whereby non-hierarchical monitoring structures are used to reduce
agency costs, can make an important
>     contribution to the success of these programs (see Varian 1990 and
Stiglitz 1993). Small credit programs in
>     Indonesia and Thailand have also had high levels of success, although
based on different principles.
>
>     Governments have long played an important role not just in the
development of new technologies, but also in their
>     dissemination, through extension programs. Well-designed extension
programs, which depend on the
>     circumstances of the country and the education level of the farmers,
can successfully transfer knowledge.
>     Furthermore, the farmers who benefit from these extension programs
often act as role models to others,
>     spreading the dissemination more widely9 - another example of the
public goods characteristics of knowledge.
>     There is also evidence that even a few years of schooling makes
farmers far more receptive to the new
>     technologies, and so government programs to extend education can play
an important indirect role in a successful
>     land reform program (World Bank forthcoming).
>
>     Sustaining Land Reform
>
>     The second concern about land reform is that it has not had
sustainable consequences. The dynamic processes of
>     the market have led to the reconcentration of land, undoing the land
reform.
>
>     Some years ago, Avishay Braverman and I (1989) formulated a general
model attempting to capture these
>     dynamics. Institutional arrangement that are constrained by costly
monitoring are typically characterized by
>     imperfect risk sharing,10 and as a result will lead to some
redistribution. We can divide the population into groups
>     according to the amount of land they own and describe the number of
people in, say, the jth decile by xj,t and the
>     entire distribution at time t by the vector xt. For simplicity, assume
that that the distribution at time t+1 is
>     governed by the transition matrix Ak where the superscript denotes the
institutional arrangement:
>
>     xt +1 = Akxt.
>
>     In this model with fixed coefficients in the transition matrix, the
prognosis for land reform could be bleak. For
>     almost all transition matrices11 there will be a unique equilibrium
distribution of land x* which is the eigenvector of
>     the transition matrix:
>
>     (1)x* = Ax*.
>
>     If the economy is perturbed from x* it will asymptotically return back
to its original distribution - any land reform
>     will eventually be undone, with a return to the original land
distribution.
>
>     But the institutional arrangements are themselves endogenous, and
depend on the land (wealth) distribution. In
>     circumstances of great inequality, the institution of sharecropping
will emerge, solidifying the unequal distribution.
>     In contrast, when land is distributed in a more egalitarian manner
farms will be owner-managed, resolving the
>     agency costs and making it easier for small farmers to maintain or
even extend their property. In our simple
>     algebraic exposition, we can think of the transition matrix itself as
a function of the land distribution:
>
>     (2)xt +1 = A(xt) xt.
>
>     In this case there may be multiple equilibrium land distributions:
>
>     x* = A(x*) x* and x** = A(x**) x**
>
>     If so, a land redistribution can lead to a change the institutional
arrangements and thus be sustained. To be sure, it
>     is unlikely that the government will, in its initial land
redistribution, hit upon an equilibrium. But this model also
>     holds out the promise that a small redistribution could eventually,
and without further intervention, lead to a lead
>     to a large shift from one equilibrium to another. Even if we see some
reconcentration after the land reform, the
>     final equilibrium may still be far more egalitarian than the pre-land
reform equilibrium.
>
>     These conclusions are strengthened if we treat the transition matrix
as being dependent on government policies,
>     A(G). For instance, the provision of micro-credit and extension
programs may increase the chance that a very
>     small-holder can increase his or her holdings. In this case the
equilibrium land distribution will be a function of
>     government policies:
>
>     (3)x*(G) = A(G) x*(G)
>
>     and changes in G can change the ultimate distribution of land in ways
that simply changing x* through land
>     redistribution, at least in the framework of equation (1), cannot.
>
>     If we combine (2) and (3) so the transition matrix depends both on the
distribution of land and on government
>     policy (which itself may be affected by the distribution of land) then
government policies can affect each of the
>     equilibria, and indeed, government programs can even create a second
more egalitarian equilibrium, where in the
>     absence of the government program there would only be a unique
equilibrium. As a result government credit and
>     technology programs not only enhance the effectiveness of the land
redistribution programs, they may be
>     necessary for its very success.
>
>     Active government programs, facilitating the provision of credit and
technology which have limited the success of
>     land reform programs in many parts of the world, have been far more
attractive than limiting land transactions, the
>     approach pursued by other countries.12 Farmers often find ways, like
leasing arrangements, to circumvent these
>     policies, which not only lead to effective reconcentration but are
associated with ancillary problems, like
>     underinvestment in the land. Even if these reforms manage to sustain
the egalitarian land redistribution they do so
>     at a high economic cost. Because land cannot be bought and sold, it
cannot be used as collateral. Because it
>     cannot be used as collateral, it exacerbates the capital market agency
problem. Policies to limit land transactions
>     also make it more difficult for the economy to adapt to changing
circumstances and the process of development.
>     Over time, it may be desirable for there to be some increase in the
average size of farms, as some workers move
>     off the land to higher paying jobs in other sectors. Frequently,
countries with these strong land regulations hold
>     that individuals who move off the land lose their rights to the land,
and thus these regulations inhibit the
>     development of more productive industries and the reallocation of
labor to more productive uses. The
>     consequences of these arrangements is thus in many cases a lowering of
standards of living after the land
>     redistribution program.
>
>     But even if we reject these strong forms of land restrictions, and
even if we engage in active government
>     programs designed to overcome limitations on credit and technology,
there remains a question: are some
>     restrictions on land transactions (such as taxes on the purchase of
land by large landowners) desirable? From the
>     perspective of efficiency, that is a social welfare function which
only takes into account individual well-being, with
>     no explicit account of distinctions, these restrictions might seem
hard to justify. After all, in a perfect market,
>     individuals would only enter into contracts to buy and sell land if
they were made better off than they otherwise
>     would have been, that is if the transaction were Pareto improving.
>
>     There are, however, several possible market failures that might
justify such interventions even from the
>     perspective of efficiency. First, individuals may not take fully into
account the benefit of the land bequest to their
>     children, including the benefits which arise from the reduction in
agency costs. Second, because of capital market
>     imperfections poor individuals have much higher discount rates than
richer individuals. As a result, even if the sale
>     to the higher-income farmer lowers the returns to the land (because of
the increased agency costs), evaluated at
>     the lower discount rate, its value might still rise. As a result,
although the transaction would be rational from the
>     perspective of the two participants, it could still lower national
income. Although the general theory of the second
>     best makes it ambiguous whether such sales might still be desirable
(even though output is lower, the value of
>     consumption by the tenant which it supports can be higher), it is
clear that if there are taxes related to output
>     (such as an export tax), then sales will occur which are undesirable.
>
>     But all of this takes a too narrow view of the object of land
redistribution: it is more than just a matter of
>     economic efficiency, or even equity as measured by a conventional
social welfare function. It is part of a broader
>     social and institutional change that is the central component of the
second generation of reforms, the topic I turn
>     to now in the second part of this lecture.
>
Zhiyuan Cui
617-253-2951(p)
617-258-6164(f)
http://web.mit.edu/polisci/www/faculty/Z.Cui.html