IN SEARCH OF THE HOLY GRAIL: POLICY CONVERGENCE,
EXPERIMENTATION AND ECONOMIC PERFORMANCE
September 2003.
Abstract
We consider a model of policy choice in which appropriate policies depend on a country’s own
circumstances. The presence of successful countries (“leaders”) generates informational externalities
for latecomers. On the positive side, some corrupt governments are reined in. On the negative side,
some honest governments are disciplined inefficiently, resulting in too little “policy experimentation.”
Our model predicts a U-shaped pattern in economic performance as we move away from the leaders in
the relevant space of characteristics: close neighbors should do very well, distant countries moderately
well on average with considerable variance, and intermediate countries worst of all.
I. Introduction
There has been a remarkable convergence of views over the last two decades among both scholars
and policy makers on what constitutes desirable policies to promote economic growth. Yet economic
performance is more heterogeneous across the world than it has ever been. Growth rates in the
developing world have been lower on average in the past two decades than they were in the 1960s
and 1970s. And the dispersion in growth rates across countries has become wider (see Mukand and
Rodrik, 2002, Table 1). Why this disappointing performance if we have apparently learned so much
about what we need to foster economic convergence?
One possible answer is that the divergence in performance is due to the failure of governments
around the world to adopt the consensus policies. Yet countries that have in fact tried hard to implement Washington Consensus-style reforms have often done poorly. Consider Latin America, for
example. Most Latin American countries have undertaken signiÞcant amount of structural reform,
including trade and Þnancial liberalization, tax reform, privatization and opening up to capital ßows
since the middle of the 1980s.1 Yet economic performance in Latin America has been quite disappointing over this period. Virtually all Latin American countries have grown at slower rates during
the 1990s than they did prior to the 1980s (with Chile being the only clear exception). Of course,
the relationship between ”structural reform” and growth is a complex one, and there is only a limited
conclusion that one can draw from such a broad-brush look at the evidence. But the Latin American experience does suggest that the growth payoff reaped by the most ambitious reformers has been
meager at best.
Another important strand of evidence comes from the last two decades’ success cases. China
and India stand out here. These two large countries have experienced signiÞcant increases in their
growth rates - China since the late 1970s, and India since the early 1980s - in a period when most other
developing countries have gone the opposite way. Indeed, China’s and India’s performance overshadows
the disappointing outcomes elsewhere, making the last two decades a developmental success on a
population-weighted basis. The policies that have enabled this performance present a very awkward
Þt with the usual list that has been on the agenda of reformers in Washington and elsewhere. China’s
reforms have been marked by partial liberalization, two-track pricing, limited deregulation, Þnancial
restraint, an unorthodox legal regime, and the absence of clear private property rights. India’s reforms
have been less distinctive, but still marked by signiÞcant departures from the rulebook. Even after
the trade reforms of the early 1990s, for example, India remained one of the world’s most protected
economies. One could argue that these economies would have grown even faster had they embarked
on a more orthodox reform agenda. But the difficulty with this argument is that those countries that
adopted the orthodox policies - such as in Latin America - for the most part did worse than China
and India, not better.
We entertain a different possibility in this paper. We consider a world in which appropriate policies
1
Morley et al. (1999) provides a quantitative picture of these reforms. Their index of structural reform for Latin
America rises from around 0.47 in the early 1970s (out of a maximum of 1) to around 0.55 in the early 1980s, and then
jumps up to 0.82 by 1995.
1
and institutional arrangements have a large element of speciÞcity, and experimentation is required to
discover what works locally. Reforms that succeed in one setting may perform poorly or fail completely
in other settings. Two-track reform may work well in Deng’s China but not in Gorbachev’s Soviet
Union. Gradualism may be appropriate to India, but not Chile. Import-substitution may foster
competitive industries in Brazil, but not in Argentina. Industrial policy may produce results in South
Korea, but not in much of Africa.
Such speciÞcity could arise from differences in historical trajectories, geography, political economy,
institutional settings, or other initial conditions. It could help explain why successful countries - China,
India, South Korea, and Taiwan among others - have almost always combined unorthodox elements
with orthodox policies. It could also account for why important institutional differences persist among
the advanced countries of North America, Western Europe, and Japan - in the role of the public
sector, the nature of the legal systems, corporate governance, Þnancial markets, labor markets, and
social insurance mechanisms, among others.
We do not mean to suggest that economic principles work differently in different places, or that
economics itself needs to be tailored to local conditions. We make a distinction instead between
economic principles and their institutional embodiment (Rodrik 2003). Most Þrst-order economic
principles come institution-free. Incentives, competition, hard-budget constraints, sound money, Þscal
sustainability, property rights are central to the way that economists think about policy and its reform.
But these principles do not map directly into institutional solutions. Effective property rights can be
implemented through common law, civil law, or, for that matter, Chinese-type socialism. Competition
can be maintained through a combination of free entry and laissez-faire, or through a well-functioning
regulatory authority. Macroeconomic stability can be achieved under a variety of Þscal institutions.
Institutional solutions that perform well in one setting may be inappropriate in other setting without
the supporting norms and complementary institutions. In the words of North:
“ economies that adopt the formal rules of another economy will have very different
performance characteristics than the Þrst economy because of different informal norms
and enforcement. The implication is that transferring the formal political and economic
rules of successful Western economies to third-world and Eastern European economies
is not a sufficient condition for good economic performance.” (North 1994, 8)
In addition, since policy makers always operate in second-best environments, optimal reform trajectorieseven in apparently straightforward cases such as price reform - cannot be designed without due regard
to prevailing conditions and without weighting the consequences for multiple distorted margins.
A paper that takes these issues seriously has a triple burden. It must explain why countries
do converge on “consensus” policies even when their circumstances call for different arrangements.
Second, to the extent that some countries choose to experiment rather than imitate, it must provide
a reason why they do so. Finally, it must provide a plausible accounting of the pattern of economic
performance that emerges when there is a tendency for countries to converge on similar policies. This
paper contains some Þrst steps in all three directions.
2
We consider a model in which there is uncertainty about which policies (or equivalently, institutional arrangements) represent the best Þt for a country’s circumstances. Policies that are well
matched with circumstances produce higher output than policies that are not. Policy makers are
better informed than the electorate about the requisite policies, but they may be also tempted to
implement corrupt policies that siphon rents to themselves. The voters observe the policy choice of
their government, but not the resulting output or the rent transfer (which do not materialize until the
end of the government’s full term in office). They have an opportunity to kick the government out of
office midstream. Since they can potentially save on the economic rents that would otherwise accrue
to a corrupt government, voters are more likely to do so the higher the probability that the policy
in place is corrupt. Policy makers like to remain in office (i.e., they derive “ego” rents from holding
office) and, everything else being the same, would rather stay in office than be kicked out.
We focus on the stage game where the governments and their electorates in a “follower” country
have already observed a successful policy in a neighboring country (called the “leader”). The follower
government faces the following dilemma. It can imitate the leader’s policy, and thereby signal that
there will be no corruption, but at the cost of adopting a policy that may not be appropriate to
domestic circumstances. Or it can choose a policy that targets as closely as possible its private signal,
in which case it incurs the cost of being perceived as corrupt (along with the cost of experimentation).
We show that the informational externality created by successful leaders results in both an upside
and a downside. Countries whose underlying state is close to the leader - i.e., the leader’s neighbors by
the relevant metric - choose to mimic the leader’s policies, even when their governments may otherwise
have followed corrupt policies. This yields a double beneÞt to the neighbors: they can forego the
costs of experimentation, while reaping the beneÞts of the discipline that is imposed on potentially
corrupt governments. Countries in the far periphery are unaffected by the leader’s example, as their
governments’ incentives to experiment or adopt corrupt policies remain unaltered. The cost is borne
by honest governments in the near periphery - i.e., by countries that are too far from the leader for
mimicking to be welfare improving, but close enough for the informational externality to generate an
incentive for honest governments to mimic. The last case entails inefficient disciplining of government
policies. The shadow of corrupt governments imposes a cost on honest governments.
Therefore our model yields distinct predictions about the patterns of policy imitation, corruption,
and economic performance as a function of a country’s position vis-a-vis successful leaders. In particular, it predicts a U-shaped pattern in economic performance as we move away from the leader in
the relevant space of characteristics: close neighbors should do very well, distant countries moderately
well on average with considerable variance, and intermediate countries worst of all.
We believe this framework helps account for some of the salient features of the economic landscape
of the last few decades. Countries in the vicinity of growth poles such as Japan and later the East
Asian tigers have tended to do very well, in part by imitating many of the policies followed by these
leaders. Countries whose underlying characteristics or geographic distance place them very far from
the leaders have experienced highly variable fortunes. Superstars like China, Botswana and Mauritius,
which have achieved success on the back of a great degree of policy experimentation, have occasionally
3
emerged from this group. Some of these experimenters in turn have become examples for others to
follow, such as China in the case of Vietnam. And we would claim that many countries in between
have been “inefficiently disciplined,” adopting policies that are less corrupt and more transparent than
in the past, but also perhaps less appropriate to their circumstances. The Latin American economies
of the 1990s may constitute the chief examples of this last group.
The perspective we adopt in this paper has many antecedents in the literature. Economic historians have long emphasized that the fragmentation and diversity of early modern Europe was a source
of economic strength that allowed Europe to eventually overtake the centralized empires in China,
the Middle East, and the Indian sub-continent. Central here is the idea that decentralization enabled
competition, not only in the market for goods and services, but also in market for institutional arrangements (see Landes 1998, North 1993, and Diamond 2001; the latter emphasizes the geographical
sources of these historical trajectories). In the recent literature on development strategies, Evans (2001)
and Rodrik (2000) have argued for the importance of local deliberation in ensuring a good Þt between
institutions being adopted and local conditions, while Besley (2001) notes that “adopting a policy
is analogous to adopting new technologies with uncertain beneÞts.” Dewatripont and Roland (1995)
and Roland (2000) underscore the importance of uncertainty and experimentation in the context of
transition economies.
Our paper also relates to the large literature on Þscal decentralization and interjurisdictional competition, as surveyed for example by Oates (1999). This literature considers policy experimentation as
a possible beneÞt of Þscal decentralization, and touches on the information externality that is generated
by the adoption of innovative policies (see Rose-Ackerman 1980). This literature differs from the approach adopted here in that it typically presumes policy makers are rewarded or punished based solely
on their policies’ success. Besley and Case (1995) provide an interesting analysis of interjurisdictional
“yardstick competition” within the U.S. where voters make comparisons with neighboring states to
overcome political agency problems. The public administration literature on policy innovation among
U.S. states is discussed and evaluated in Berry (1994).
Perhaps most directly relevant to our research is a recent paper by Berkowitz, Pistor, and Richard
(2003). This paper Þnds that countries that developed their formal legal orders internally, adapted imported codes to local conditions, or had familiarity with foreign codes ended up with signiÞcantly better
legal orders than those that acquired their formal legal order from abroad without much adaptation
and/or familiarity. Our paper is also related to Kremer, Onatski, and Stock (2001), who consider the
role of experimentation in driving the pattern of incomes around the world. In their model, countries
search among policies until they reach a satisfactory income level, at which point they stop experimenting. Our framework puts more structure on these issues, and also, more importantly, endogenizes
the choice between experimentation and imitation.
The plan of the paper is as follows. Section II presents the model which we analyze in Section III.
Section IV concludes.
4
II. The Framework
Countries and policies: We consider a world with a large though discrete number of countries. Countries differ from each other in terms of their geography, culture and other local conditions which make
any given policy more or less “appropriate” across countries. We capture these country speciÞc local
conditions by assuming that each country i has a unique “state of the world” zi , where zi ∈ R and is
drawn from a distribution F (z).
The government gets to choose a policy aj that persists for the entire period and affects national
income, where aj ∈ R. There are two distinctive features of policies that we emphasize. The Þrst is the
state-contingent nature of policy making, with the impact of a policy on national output depending
on the state of the world. It is useful to keep in mind that in our analysis a “policy” has very
broad connotations and is not necessarily just related to the economic policy choice per se - it might
well consist of institutional and political innovations. This broader interpretation of policies becomes
particularly useful in understanding the second aspect of a policy that we highlight. We assume that
some policies are conducive to the inefficient siphoning away of resources due to corruption and wastage
(corrupt policies), while other policies encourage ‘discipline’ (honest policies).
Accordingly, if a country with the underlying state of the world zi implements policy ajh ∈ Ah
where Ah is the set of “honest” policies, then private-sector income is yi = −θ(zi − ajh )2 . Observe
that the closer or more “appropriate” is the policy to the state, the higher is private income. In
contrast, if the government chooses a policy ajc which belongs to the set of corrupt policies Ac and
the government is not kicked out of office midstream (see below), then private-sector income is given
by yi = −θ(zi − ajc )2 − R, where the economic rents siphoned off by the government equal R. We also
incorporate the fact that a corrupt policy results in a dead-weight loss to society as a whole, given by
δic , which is drawn from G(δ) - a symmetric single-peaked distribution with support [δL , δH ] where
δL > 0. We capture this effect by assuming that there is a country-speciÞc corrupt policy ajc that
minimizes the dead-weight loss of corruption, where ajc = zi + δic . This dead-weight loss is privately
observed by the government before it makes its policy decision.
Governments and the citizen: We assume that all policymakers are identical with an objective function
given by Vi = yi +(φ−1)R+² where φ > 1 and ² > 0. Here ² is the “ego” rents from being in office and
R is the “economic” rents from enacting a policy that allows corruption. Importantly, both are earned
are reaped only if the government remains in power the full term. If the government is kicked out
mid-term, there is a “caretaker” government that takes office and continues the incumbent’s policies,
with the associated deadweight loss (if any), but does not capture the economic or ego rents.2
Finally, each country i has citizens of unit mass, who live one period, with payoffs given by ui = yi .
Midway through the government’s term, citizens make the decision of whether to organize a coup, a
revolt or force midterm elections and replace the incumbent at a stochastic cost ci ∼ U [0, c̄].
Information and experimentation: The citizens of a country do not know the location of their under2
This is a particularly simple mechanism that ensures that the probability of re-election is an increasing function of
the perception that policies pursued are corrupt. Alternative more elaborate political underpinnings can also be provided,
as in some variant of the career-concerns model.
5
lying state of the world, other than zi ∼ F (z). However on assuming office, a government learns about
the country’s location zi . We are interested in capturing the uncertainty inherent in the policy making
process when experimentation is carried out. We capture this cost of experimentation by assuming
that even though the government i knows its “ideal policy”, it can determine the actual policy implemented only up to a random error term, ai = zi + η, with η ∼ N (0, ση2 ). This is a relatively simple
way of capturing the costs as well as the gains from experimenting with an untried policy. If, instead
of experimenting, a country i picks up a policy off-the-shelf that has been tried in some other country,
then there is no uncertainty associated with its implementation. We assume that δc2 > ση2 for all δ —
implying that experimentation generates a higher expected output than a corrupt policy.
The timing of events: The world consists of a large number countries where the entire set of countries is
divided into two cohorts - the “leader” cohort makes policy choices at the beginning of period one and
the “follower” cohort makes policy choices at the beginning of period two. At the beginning of period
one, the governments in the leader cohort privately observe their location zi and the dead-weight loss
of a corrupt policy δic and then choose the publicly observable policy ai . Citizen-voters observe ai , ci ,
and then decide whether to retain or replace the incumbent with a caretaker government (as described
above). At the end of the period, yi is realized, and if the original government has survived, R and ²
get “paid out”. Information and the sequence of decision making in the follower cohort are identical
to the above, except in one important respect. At the start of period two, citizens and governments
in the second “follower” cohort observe policies and private-sector incomes in the leader cohort, i.e.
they observe the period one history h1 ∈ H1 where H1 is the set of all possible period one histories,
and h1 = {(a1 , y1 ), ..., (aN , yN )} is one element in this set (H0 = Φ for leader cohort).
In the next section we search for a Perfect-Bayes’ equilibrium of the political game facing each
country’s government with objective function Vi . Such an equilibrium will consist of a government’s
strategy which is a function of zi , δi , and Aj and that determines whether to experiment, imitate
or choose a corrupt policy. Citizens in each country will observe ai , ci and ht in making decision of
whether to replace the government. Finally, the government’s strategy and the citizen’s beliefs have
to be consistent with each other and derived using Bayes’ rule.
III. Discipline and Corruption in the Search for Prosperity: Analysis
We begin our analysis by Þrst examining the “leader” cohort. Consider the government of country i
which has privately observed its true location zi . The expected payoff of a government that chooses
to experiment with an “honest” policy aih = zi is given by,
Vihxpmt = E(−θ(zi − aih )2 + ²) = [−θση2 ] + π X ² if ai ∈ Ah ,
(1)
where π X is the probability that a government which experiments with an honest policy ah , remains
in office for the entire period. A government faces uncertainty when it experiments with an untried
policy, and this uncertainty results in an expected loss of output - the Þrst term in the square brackets.
However, this choice of experimentation (potentially) also affects the expected rents that may accrue
6
to the government - the second term above. Similarly, the expected payoff from a corrupt policy is,
2
Vicorr = E[−θ(zi − aic )2 + (φ − 1)R + ²] = −θδic
+ π C [(φ − 1)R + ²] if ai ∈ Ac .
(2)
The choice of a corrupt policy aic (= zi + δic ) results in not only a rent transfer but also a deadweight
loss - the Þrst term on the right hand side. The choice of a corrupt policy has an impact on the
expected rents earned (the second term on the right hand side), by affecting (again, potentially) the
probability of remaining in power(π C ).
2 − σ2 ) >
This implies that a government prefers ai ∈ Ah iff Vihxpmt > Vicorr , which is true when θ(δic
η
C
X
π [(φ − 1)R + ²] − πh (²) = π(φ − 1)R. The latter equality arises because of the citizen’s inability to
distinguish between a corrupt and an honest policy, since output and rents are only observed at the
end of the period, which implies that πhX = π C ≡ π. Therefore in the leader cohort a citizen’s reelection rule is independent of the policy choice observed, and is driven solely by whether the realized
cost of replacement (ci ) exceeds the expected gains. The gains of replacement are the economic rents
that would have been consumed by government, had it been allowed to complete its full term in
office. Therefore, the probability of remaining in office for a government in the leader cohort equals
π = prob(ci ≥ ψj1 R), where ψj1 is the probability that any policy aj chosen by the leader cohort, is
corrupt. Notice that this probability that any chosen policy is corrupt, is endogenously determined
and is a function of the distribution across countries of the dead-weight loss δc . In the lemma that
follows we characterize this relationship between policy choices and the country’s deadweight loss from
corruption and demonstrate that it is a function of a cut-off δ1∗ .
Lemma I.There exists an equilibrium in the “leader” cohort’s political game, such that all countries
with δ ≤ δ1∗ choose corrupt policies while all those with δ > δ1∗ prefer to experiment.
Proof: See Appendix.
Decision making in the “follower” cohort constitutes the heart of our analysis - and the remainder
of our discussion focuses on it. As with the leader cohort, governments in the follower cohort continue
to have the choice of either experimenting with an untried honest policy or adopting a corrupt policy.
However, they now have an additional option - imitation of a policy that has been successfully triedand-tested by a country in the leader cohort. Not only does imitation of such previously tried policies
eliminate the uncertainty associated with experimentation, but it also potentially enables an incumbent
to signal to the citizen that it has chosen a policy that is not corrupt - increasing his chances of
remaining in power. In what follows, for simplicity, we restrict the set of policies that can be imitated
to AhL — the set of policies that are perceived to be honest, since they have resulted in a sufficiently high
income in the leader cohort, i.e. where AhL = {ahL |yiL (ahL ) ≥ ȳ = −θδL2 − R}, and AhL ⊆ Ah . We label
all such high income countries in the leader cohort as “successful” leaders, with nL being the number
of such leaders.3 Therefore the additional option confronting a government that has privately observed
3
At the cost of some work, it is possible to analyze the more general case, where we do not limit our analysis
to the case of such “successful” leaders. This generalization follows from the fact that output follows a chi-square
distribution (since η is normally distributed and the square of a normally distributed random variable is chi-square).
7
zf is that of imitating any of the “successful” leaders. The expected payoff from such imitation of a
successful leader’s policy aiL ∈ AhL , equals,
Vfimit = −θ(zf − aiL )2 + π I [²] = −θ∆2f L + π I [²]
(3)
Here π I is the probability of the government completing its term in office if it imitates the leader.
Observe that the “successful” leader’s policy is aiL = ziL + ηiL . We deÞne z̃iL ≡ aiL = ziL + ηiL as the
“effective” location of the successful leader i and ∆f L = zf − z̃iL as the “effective distance” between
the imitator and this leader.4 As is clear from the above formulation, a government’s payoff from
imitating the policies of a leader is declining in the distance.
Policy choices made by the follower cohort are a function of the Þrst period history h1 , i.e. the
number and location of “successful” leaders (if any). To facilitate our analysis of decision making in
the follower cohort, we propose the following equilibrium to the follower country’s political game.5
A country at a relatively small distance will always imitate and adopt the leader’s policy. Countries
at some intermediate distance will prefer to imitate if their government has a relatively high δ and
choose a corrupt policy if δ is low. Countries at a greater distance will choose between a corrupt
policy and experimenting with an honest one. Finally, if there is more than one “successful” leader,
then a country which prefers to imitate does so by adopting the policy of the leader from whom it
has the smallest effective distance. Therefore, under this proposed equilibrium, the payoffs to zf from
imitating (Vfimit ), experimenting (Vfxpmt ) and choosing a corrupt policy (Vfcorr ) are (using (1), (2) and
(3) ),

2
I

if
af ∈ AhL
 −θ(zf − z̃iL ) + πf [²]
Vfk =
if
aj ∈ Ah
−θση2 + πfX [²]


if
ai ∈ Ac ,
−θδf2c + πfC [(φ − 1)R + ²]
where (as earlier), π k is the probability of remaining in office as a function of whether the government
imitates, experiments or chooses a corrupt policy. In comparing the payoffs, recollect that due to the
citizen’s inability to distinguish between a corrupt or honest policy, we have πfX = πfC .
In what follows, we demonstrate the existence of an equilibrium where government decisions are a
function of (δ, ∆).
Further we can check to conÞrm that a chi-square distribution satisÞes the MLR property (see Milgrom, 1981), i.e. if
the conditional density function of private-sector output is given by f (y|aj ), where aj can be corrupt or honest, then
according to this property, it must be that [f (y|ac )/f (y|ah )] is decreasing in y. The fact that the MLR property is satisÞed
ensures that when a follower country (and its citizens) observes and ranks output across countries in the “leader” cohort
y1 (a1 ) ≥ y2 (a2 ) ≥ y3 (a3 )... ≥ yN (aN ) - a higher output is associated with a greater probability of the policy being honest
(1 − ψ). A country observes its zf and conditional on imitating, will adopt the policy of the “leader” which maximizes
its payoff i.e. V imit = (1 − ψ)[−θ∆2f + πI ] + ψ[−θδ 2 + π C ((φ − 1)R + ²)]. We conjecture that the lower the output of a
leader, the smaller the associated “basin” of imitation.
4
Observe that if there is more than one ‘successful leader’ country, then the effective leader will be the one whose
“effective distance” is smaller. In other words if a∗f = zi , where a∗f = argmaxai {Vfimit (ai ) : ai ∈ AhL }.
5
If there are no “successful” leaders in the Þrst period, then there is no externality imposed on the follower cohort.
This implies that the distribution of policy outcomes between experimentation and corruption across the two cohorts is
similar.
8
Proposition I. There exists a Perfect Bayesian Equilibrium where governments in the follower cohort
make policy choices as a function of h1 , zf , δf c and ∆f i and citizens in each of these countries retain
or replace the incumbent as a function of observed af , ci , AhL if nL ≥ 1, such that:
(a) if ∆f i ≤ ∆1i , then ∀δ, the government imitates and adopts the policy af = aiL of the “successful”
leader with the smallest distance to it, ∀ ai ∈ AhL
(b) if ∆f i ∈ (∆1i , ∆f i ] then governments with high δ imitate, those with intermediate δ experiment
and those with small δ choose corrupt policies
(c) if ∆f i > ∆f i then countries with relatively small δ will choose corrupt policies while those with
high δ will experiment.
Proof: See Appendix.
We label as “neighbors” all countries whose location from the leader is sufficiently small, such that
they all prefer to imitate the leader, irrespective of their δ (see Proposition I(a)). If a follower country
falls within this “neigborhood” of two or more successful leaders, then it is deÞned to be the neighbor
of the leader with whom its effective distance is smaller. In contrast, we label all countries that are
located beyond this distance, as being located in the “periphery” of that particular leader country. We
futher distinguish between those in the “near-periphery” or the “far-periphery”. A country is deÞned
as being located in the “far-periphery” if it is sufficiently distant, such that it will prefer not to imitate
under any circumstances(see Proposition I(c)). Finally, countries that lie at an intermediate distance
between the “neighbors” and the “far-periphery”, are said to be located in the “near-periphery”(see
Proposition I(b)).
The above proposition establishes the existence of an equilibrium where follower countries in the
proximity of a successful leader are “disciplined” into imitating its policies. Such imitation has both
a positive and a negative aspect to it. The upside of imitation is that it minimizes uncertainty and
corruption while the downside is that a country may adopt policies that are not appropriate for its
individual circumstance. In our framework, the decision to imitate a leader has a political payoff a government may complete its term in office by signaling to its citizens that corruption is unlikely.
Therefore, while political considerations result in governments getting “disciplined”, it is at the cost of
adopting policies inappropriate to the country’s local conditions. Indeed, there is the possibility that
this disciplining may turn out to be inefficient.
This makes it especially important to systematically evaluate the welfare impact of this disciplining
of nations. We do this in the next proposition.
Proposition II. There exists a Perfect Bayesian Equilibrium where the relationship between a follower
country’s location zf and its national income with respect to a leader zL is such that
(a) “neighbors” are efficiently disciplined into imitating policies that enhance national income ∀δ,
(b) some of the countries that lie in the “near periphery” are inefficiently disciplined into imitating
the leader’s policies, thereby resulting in lower expected income than in the “far-periphery” or the
“neighbors”,
(c) countries that are sufficiently distant and are located in the “far-periphery” do not get disciplined,
choose between experimentation and corrupt policies, and have higher expected income than the inefficiently disciplined countries of the “near-periphery”.
Proof: See Appendix.
9
The above proposition establishes that countries within an intermediate distance of a leader are
adversely affected by the informational externality generated by its economic success.
Despite the simple structure of our model, it yields several rich implications (where needed, these
are formally demonstrated in the Appendix).
(i) History and location: In our framework, policy and institutional choices in the follower cohort
of countries are determined by historical accident. The success or failure of experimentation with
institutions amongst the leader cohort inßuences the pattern of institutional adoption and economic
performance amongst the follower set of countries. In emphasizing a novel channel of institutional
change, we throw light on some of the factors underlying institutional adoption by countries in the
vicinity of successful leaders in Western Europe, Japan or China. In contrast, part of the difficulty
in developing appropriate institutions in countries in Africa may well be location - the absence of a
demonstrably “successful” leader in their vicinity.
(ii) Neighbors: free riding and leapfrogging: A follower country in the immediate proximity of a
leader (i.e. with ∆f L < ∆1L ) has the advantage that it for all δ it can free ride on this leader’s
previous experimentation. The downside of such free-riding is that the institutions it adopts may not
be perfectly appropriate for its own individual circumstance. Imitating a leader’s policies results in
(expected) output that is decreasing with greater “effective distance”. However ex post, some of these
neighbors may leapfrog and achieve output that is even higher than the leader whose policies they
mimicked. This is because depending on the realization of ηL and the “effective distance” ∆f L , a
leader country’s output yL = −θ(ηL )2 may be lower or greater than that of a follower country, which
equals yf = −θ(zf − z̃iL )2 .
(iii) Near-periphery — imitation and inefficient disciplining: Countries that are located at some
intermediate distance from the successful leader, may get inefficiently disciplined into adopting the
leader’s policies. Indeed the greater the distance from the leader, the larger the proportion of countries
in this region that get inefficiently disciplined. This inefficient disciplining can be of two kinds. First,
there exist some countries that would otherwise have chosen to efficiently experiment (i.e. have high
δ), but now end up imitating for political reasons alone. Second and more striking is the possibility
of inefficient disciplining of otherwise corrupt governments. Here the economic gains of a reduction
in corruption are more than offset by the fact that the government is following a leader who is too
far (see Appendix). This inefficient disciplining of countries will result in expected national income
that is lower than that of countries in the neighborhood of the leader, or that of countries in the
“far-periphery”.
(iv) Far-periphery — experimentation and corruption: Countries in the follower cohort who lie in
the “far-periphery” of all successful leaders, are unaffected by the informational externality provided
by the preceding cohort’s institutional experimentation. Accordingly, some of these governments
experiment with honest policies, while others adopt corrupt policies. Therefore, there is higher variance
in outcomes, with the potential for some new spectacular performers, as well as some with disastrous
economic outcomes.
(v) Distance, growth and economic performance: The most striking empirical implication of our
framework is that it yields a U-shaped relationship between distance from the leader and economic
performance. Follower countries in the close proximity of the leader get efficiently disciplined achieving
a reasonably high level of economic performance with a “growth pole” around the leader.6 As distance
6
The mass of countries in each “growth pole” may differ depending on the “successful” leader’s location and the
underlying distribution of countries F (z). Therefore, a country such as Hong Kong may have fewer imitators than
Western Europe.
10
from the leader increases national income declines, as the social cost of being disciplined into imitating
inappropriate policies increases. However, despite this economic cost a government prefers to imitate
because of its positive political payoff. Nevertheless, beyond a certain threshold distance the decline in
output is sufficiently acute to outweigh any political gain from inefficient imitation. Therefore, in this
last region some of these governments prefer to experiment resulting in a jump in expected national
income. Taken together, the result is a U-shaped relationship between distance and expected income.
In the working paper version of this paper (Mukand and Rodrik 2002), we report a simple empirical
test of our framework that focuses on the experience of post-socialist countries. These countries
were forced to search for alternative policies once they abandoned socialism, so they constitute a
useful sample for our purposes. For most of them, the model to emulate, if any, was the Western
European example. We hypothesize that the geographic distance between each of these countries and
Western Europe is an adequate (inverse) proxy for the suitability of European-style institutions to
their circumstances. We then demonstrate that there is a robust U-shaped relationship between these
countries’ distance from Brussels and their post-transition growth rates. The U-shaped relationship
survives when we control for a variety of other determinants, including the extent of “structural reform”
undertaken, initial income, and years under socialism.
IV. Concluding Remarks
The principal innovation of this paper is a richer conception of policy making which allows for countryspeciÞcity of policies as well the possibility of learning from successful leaders. In our framework
experimentation and imitation both have a useful role to play, but they also each have a downside.
Experimentation allows countries to discover policies closer to their “ideal,” but it necessarily involves
taking risks. Imitation avoids those risks, but creates the possibility that imported policies will prove
inappropriate. By endogenizing policy choice, we have shown that the informational externality generated by successful leaders beneÞt those countries that have a “close” degree of similarity to the
leaders in their underlying circumstances, while it hurts countries that have an “intermediate” degree
of similarity.7
In our framework, the informational externality generated by successful countries results in a
sub-optimal level of policy experimentation. The reason is that the shadow of corrupt governments
restricts the options available to honest governments. But by making the distinction between honest
and corrupt policies clearer to observe, it also reduces the scope for corruption. We end by emphasizing
that countries may be able to escape this tradeoff in the longer run by establishing political systems
with a better track record of accountability and honesty.
References
Berkowitz, Daniel, Katharina Pistor, & Jean-Francois Richard, “Economic Development, Legality, &
the Transplant Effect,” European Economic Review, 47(1), Feb. 2003, 165-195.
7
We note that there may be other mechanisms that generate an “excessive” incentive to imitate. For instance, suppose
international capital markets expect that policies that constitute the Washington consensus are more likely to work for
developing countries. If a government has information that suggests an alternative course of action might be superior,
it is likely to be punished in the form of low capital inßows. In turn these lower capital ßows, by lowering employment,
may have a negative impact on the political survival of the government. Therefore, fear of lower capital ßows (and its
adverse political and economic consequences) may prevent governments from pursuing policies they know are likely to
work best. See Mukand (1999) for an elaboration of this argument.
11
Berry, Frances Stokes, “Sizing Up State Policy Innovation Research,” Policy Studies Journal, 22(3),
1994, 442-456.
Besley, Timothy, “Political Institutions and Policy Competition,” in Institutional Foundations of a
Market Economy eds. Kochendorfer-Lucius & Pleskovic, World Bank, 2001
Besley, T. and A.Case, “Incumbent Behavior: Vote Seeking, Tax Setting and Yardstick Competition”,
American Economic Review, March 1995
Dewatripont, Mathias, and Gerard Roland, “The Design of Reform Packages under Uncertainty,”
American Economic Review, vol. 83, no. 5, 1995.
Diamond, Jared, “How to Get Rich” (2001) available at http://www.edge. org /3rdculture/
Evans, Peter, “Beyond ‘Institutional Monocropping’: Institutions, Capabilities, and Deliberative Development,” University of California at Berkeley, November 2001.
Kremer, M., A. Onatski, & J. Stock, “Searching for Prosperity,” NBER Paper No.8250, 2001.
Landes, David S., The Wealth and Poverty of Nations, W.W. Norton, New York, 1998.
Milgrom, Paul, “Good News and Bad News: Representations Theorems and Applications”, Bell Journal of Economics, vol. 12, pp 380-91, 1981
Mukand, Sharun, “Globalization and the ConÞdence Game”; Tufts University Working Paper, 1999.
Mukand, Sharun, and Dani Rodrik, “In Search of the Holy Grail: Policy Convergence, Experimentation, and Economic Performance,” NBER Working Paper 9134, August 2002.
North, Douglass C., “Economic Performance Through Time,” American Economic Review, Vol. 84,
No. 3. (Jun., 1994), pp. 359-368.
North, Douglass C., “The Paradox of the West,” available at http://netec.mcc.ac.uk/WoPEc,1993.
Oates, Wallace E., “An Essay on Fiscal Federalism,” Journal of Economic Literature, XXXVII,
September 1999, 1120-1149.
Rodrik, Dani, “Institutions for High Quality Growth: What they Are and How to Acquire Them”,Studies
in Comparative International Development, vol. 35, no.3, Fall 2000.
Rodrik, Dani, “Growth Strategies” manuscript, John F. Kennedy School of Government, Harvard
University, Cambridge, MA, June 2003.
Roland, Gerard, Transition & Economics: Politics, Markets, and Firms, Cambridge,MIT Press, 2000
Rose-Ackerman, Susan, “Risk Taking and Reelection: Does Federalism Promote Innovation?” Journal
of Legal Studies, 9, 1980, 593-616.
12
"!$#&%('*),+-#/.0'-1243657#987:<;'*=,>@?ACBDFE%HGIKJL?HJM2N?2-),)IO'*G>@?(1P=Q#9%=,>R?HJ6#S)Q#92-86#T1UIO'*J6'-1H?V=Q?HJ
A&WA DB ;1H#T.0#T12XIO'-1H1PG6;M?
;'*),=YIOZ-E-VJ=,)Q#[2-),)6IO'*G>@?(1P=Q#9%\V=Q?HJ]2XA&^A BD ;M1H#T.#T1_?('`#O5;#T1P=,ab#9>c?Tdef'*>%H=Y86#T1\31P%(?\?HJ6#
IT=Q?H=QgT#9>"hi%"+-'-?H=,>6j86#9IT=,%H=,'*>lk"=,>m;#T1P=Q'78bnNEN?HJ6#f=,>ITGa/o#9>c?"=,%"1H#T?H2-=,>6#98/V=Q?HJm;M1H'-oM2NoM=,)Y=Q?pZq DsrutMv-w-xNy{zK| ^~} DP`€_r
‚„ ƒ‚…6† E7%H=,>MIO# zˆ‡Š‰S‹ Œ7CzOŽ  d_' V‘%HG6;M;’'*%H#“?HJ2N??HJ6#T1P#ˆ#O57=Y%(?H%2&A D %HGIKJb?HJ2N?s} DfrutMv-w-xFy A&”A DK€r–•]y A DO€ d_—ˆJ=,%
=,a];˜),=Q#9%?HJ2N?_.1P'*a™?HJ6#ˆIT=Q?H=QgT#9>lhi%\;#T1P%H;’#9IO?H=,+-#s?HJ6#“;M1H'-oM2NoM=,)Y=Q?pZm?HJ2N?_?HJ#“j-' +-#T1P>ab#9>c?V=,),)@o’#[1H#T?H2-=Y>6#98]=,%
j*=Q+-#9>
oZR?HJ6#&1H#9),2N?H=Q'*>%PJ=Q;Lq D r nš †˜›‚ œž„Ÿ¡ El%HGMIPJ¢?HJ2N?m2-%H%('7IT=,2N?(#98£V=Q?HJ¢#T+-#T1HZ£A]=Y%2¤IO'-1H1H#9%H;’'*>M8=,>6j¥q D dU'-?H=,IO#
?HJ2N?s?HJ=,%=,ab;M),=Q#9%_2/>6#Tj*2N?H=Q+-#1P#9),2N?H=Q'*>%HJM=Q;]o#T?pVf#T#9>4q D 2->8¥Acd!$#>' V¦?HG61P>]?('/?HJ6#j-'C+-#T1P>a]#9>@?Thi%f;1H'-oM)Q#9aRd
§#T1H#&24j-' +-#T1K>a]#9>@?V=,),)lo#/=Y>8=ž¨#T1H#9>@?“o’#T?¡Vs#T#9>£#O57;#T1P=,a]#9>@?H2N?H=Q'*>¢'-1UIO'-1P1PG6;?H=Q'*>¢=Q.©mªH«9¬­ r © ‚<®°¯„¯ d—ˆJ=,%
),2N?(?(#T1ˆ#9±G2-),=,?pZ¥=,%[?(1KG6#m=Q.šˆ²*³lµs´ ¶ q·D¸ r šˆ²cAC´ ¶ q"D¹ ‹Qy<º š»n €° ¶ ¸  ds¼7=,a];˜),=Q.0Z=,>j6E-Vs#m'-o?H2-=Y>½q Dsr¿¾ œÄ„Ã*ÀKƒ ƒ D ¾( YÁ † ÂÀ d
Å oM%H#T1H+-#U?HJ2N?“?HJ6#mj-' +-#T1K>a]#9>@?Thi%ˆ'-;?H=,a¥=Qg92N?H=Q'*>¤j*=Q+-#9%“2S;'*%H=Q?H=,+-#1P#9),2N?H=Q'*>%HJM=Q;¥o#T?pVf#T#9>LA&2->8½q D dÆ'-1ˆ?HJ6#T1H#
?(']o#m2/35#987:<;'*=,>@?f#9±@G=Y),=Qo1P=YGa½E-Vf#X>6#T#98½2]ACBD ?HJ2N?ˆ%P=,a/GM)Q?H2->6#T'*G%H),Zb%('*)Q+-#9%[o'-?HJ¤?HJ6#Uj-' +-#T1P>Ma]#9>c?Thi%;'*),=,IOZ
IPJ'*=,IO#[1PG)Q#2-%
Vf#9),)2-%?HJ#fIT=Q?H=,gT#9>lhi%
'-;?H=,ab2-)c+-'-?H=,>6jU1PG)Q#-d\ÇU=Q+-#9>&?HJ6#[IO'*>c?H=,>G=Q?¡ZX'-.?HJ6#[G>8#T1P)QZ7=,>6j“.{G>IO?H=Q'*>%9E
%H=,>MIO#m?HJ6#X1H#9),2N?H=,'*>%HJ=Q;¤o#T?pVf#T#9>ÈA/2->M8Rq D 86#9%PIO1P=Qo#98¤o@Z¤?HJ6#/IT=Q?H=QgT#9>lhi%“;M1H'-oM)Q#9aÉJ2-%2b>6#Tj*2N?H=Q+-#&%H)Q'-;#UVJ=,),#
?HJ2N?ˆj*=,+-#9>½o@Z¤?HJ6#mj-'C+-#T1P>a]#9>@?Thi%;1H'-oM)Q#9aÉJM2-%ˆ2];'*%H=Q?H=Q+-#m%H),'-;’#-E6?HJ#T1H#X#O56=,%(?H%ˆ2]G>M=,±@G#UITG?„:<'N¨AFDB d
½½ÊM
ËTÌ0͒Ì0Î 7!$#U2N1H#U=,>@?(#T1H#9%(?(#98=,>b3M>M8=,>6jUVJ6#T?HJ#T1s?HJ6#T1P##O56=,%(?;˜2N1P2-a]#T?(#T1P%[%HGIKJb?HJ2N?f'*G61
;1H'-;'*%(#98&#9±G=,),=,o1P=,Ga–=,%\%H2N?H=,%(3#98"E-VJ6#T1H#[.0'-12Uj*=Q+-#9>]J=,%(?('-1PZ/Ï D ?HJ6#“j-'C+-#T1P>a]#9>@?Thi%86#9IT=Y%H=Q'*>&?('m#O5;#T1P=Ya]#9>c?TE
=,ab=,?H2N?(#m'-1IPJ'@'*%(#&2bIO'-1H1PG;?“;'*),=,IOZ4=Y%ˆ2].{G>IO?H=Q'*>½'-. yÑФ A € d“Æ6'-1#O57;'*%H=Q?H=Q'*| >|2-)lIO'*>@+-#9>=Q#9>MIO#-EMVf#`1H#T;1P'8GMIO#
?HJ6#;M29Z-'N¨l%[?('/j-'C+-#T1P>a]#9>@?“=,>b.0'*),)Q'CVs#T1[IO'*J6'-1H? y{ÒCÓ6 A Óc€ .01H'*aÔ=,ab=Q?H2N?H=,>j y © Ó ¬ ­ € E@#O57;#T1P=,a]#9>@?H=,>6j y © Ó ªH«T¬s­ € '-1
IPJ'@'*%H=Y>6jb2bIO'-1H1PG;?“;'*),=,IOZ y © Ó ‚<®°¯„¯ € E y %H#T#/2-),%('bÆ\=Qj*G1H#bn €
šˆ² Ð ´Ó9Ú ¶ qlÓÛ ‹ ¸ 
=,. Ü Ó]ݽÞSßÚ y0ఀ
Ù
×
Ø
‹

r
Ö
šˆ²*³ µ´ ¶ q Ó· ¸
=,. ÜFá ÝRÞ ß y0à<à„€
©`ÕÓ
šˆ²@A Ó´ ‚ ¶ q Ó¹ ‹Qy<º šân €° ¶ ¸  =,. Ü |_Ý¤Þ ‚Oã y0à<à<ఀ
—ˆJ6#¤'-;;'-1H?HGM>=Q?pZuIO'*%(?b'-.?HJ6#½j-' +-#T1P>Ma]#9>c?Thi%4IPJ6'*=YIO#R'-.2$IO'-1P1PG6;?S;'*),=,IOZä=Y%&=Y>IO1H#92-%H=,>jÈ=Y>åAcd~—ˆJ6#T1H#T.0'-1H#-E
=,>L'*G61;M1H'-;'*%(#98R#9±G=,),=,o1P=,GaRE@?HJ6#&;1H'-o˜2NoM=,),=Q?¡Z4?HJM2N?2->¢'-oM%(#T1H+-#98£;'*),=,IOZ y IO'*>8=,?H=Q'*>2-)'*>¢>6'=,ab=Q?H2N?H=,'*> € =,%
IO'-1H1PG;?TE=,%/86#9IO1H#92-%H=Y>6jL=,>Acd½—ˆJ6#¥%(#T?/'-.“2-),)_;'*%H%H=Qo˜)Q#S;’#T1K=Q'8È'*>#4J=,%H?('-1P=Q#9%Uæ D 2N1H#¥8=Q+7=,86#98È?HJ1H#T#bV[29Z7%X:
?HJ6#¥%H#T?`'-.[;#T1P=Q'78È'*>6#¥J=,%H?('-1P=Q#9%mVJ#T1H# y 2 € ?HJ#T1H#¥=,%X2L%('*)Q#äç(%HGMITIO#9%H%(.{G),èR)Q#92-8#T1 y0é Ú r n € E\=ÑdÙ#-dê D ^ ê Ž r
šˆ²cA Ú´ š  ^ëê á ì7íär î nNE y o € ?HJ6#T1H#]2N1H#S?¡Vs'R'-1Xab'-1H#$ç(%HGITIO#9%H%(.{G),è)Q#92-86#T1K%TE"=ÑdÙ# é Úuïñð E y I € ?HJ6#T1H#]2N1H#b>6'
ç(%HGITIO#9%P%(.0GM),è])Q#92-86#T1P%9E=ÑdÙ#-d é Ú r‘Œ dò°>¤VJ2N?ˆ.0'*),)Q' V%[Vs#X.0'7ITG%“'*> y 2 € 2->M8R%(ó-#T?HIPJ¢'*G6? y o € 2->8 y I € d
!$#½>6'CVÉo’#Tj*=Y>oZäab2N;;M=Y>6jÈ©`‚<®°¯„¯¥2->8~© ªH«9¬­ '*> y ©  A € %H;M2-IO#-d Å oM%(#T1P+-#4?HJ2N?$ôPõú ö¡÷„øù r ¸ ôHûC ü ” Œ
2->8¢?HJ2N? ôPõlý0 þpÿ<ÿ r š ð ²cA ¶ ‹Qy<º š¦n €° ¶ ¸  ô(û  ” Œ dX¼7=,>MIO#`?HJ6#SIT=Q?H=QgT#9>7:<+-'-?(#T1XIT2->M>6'-?m8=,%(ô ?H=,>j*G=,%HJ¤oô #T?pVf#T#9>
2$IO'-1H1PG;?&2-ô>M8~2->~J6'*>6#9%(?S;'*),=,IOZ-EVf#½J29+-#¢ô q· r q"¹ r q VJ=,IKJå=,a];M),=,#9%`?HJ2N? ôKõ
” ôPõ ýþpÿpÿ duò°>
'-1P86#T1?('b#9>%HG61P#m?HJ2N?ˆo'-?HJ½?HJ6#`8' V>@Vf2N1K8R%H)Q'-;M=Y>6j]ITG61H+-#9%=Y>c?(#T1P%H#9IO?=,>¤?HJ6#X1P#9)Q#T+F2->@?ˆ1P2->ô ö¡j-÷„ø#Xù '*> y ©  ô A  € %(;M2-IO#
y 1H#9IO'*),),#9IO?/?HJ2N?bA½=Y%Xo'*G>86#98äo#T?pVf#T#9>åA Ú 2->M8»A € E_Vs#¤2-%H%HGMa]# y<º šën €° ^Ô² y A Ú´ š~³lµ´ € dȗˆJ=,%m#9>M%HG61H#9%
IO'*>8=,?H=Q'*>%X'*>u?HJ6#4#9>87:<;'*=,>@| ?H%X| '-. y0ఀ š y0àpà<ఀ %HGIKJä?HJ2N?Tk4© ªH«T¬s­ y A r A Ú €`r ¸ šâ²*³ µ´ ”É©X‚p®¡¯„¯ y A r A Ú €`r
šˆ²cAC´ ¶ ¸ ¶ y<º šën €° ” © ¬ ­ r ¸ duÇm=Q+-#9>»?HJ6#½IO'*>c?H=,>G=Q?¡Z'-.ˆ?HJ6#¤;M2 Z-'N¨%¥2->8ä?HJ#1P#9%(?(1P=,IO?H=Q'*>M%`'*>»?HJ6#
#9>87:<;'*=,>@?H%[?HJ2N?“Vf#`J2 +-#`=,a];'*%(#98lE7Vf#XJ29+-#&2->R=,>@?(#T1P%(#9IO?H=Q'*>Ro’#T?¡Vs#T#9>R?HJ6#`ITG61H+-#9%© ‚<®°¯„¯ 2->M8¢©mªH«9¬| ­=,| >?HJ6#
1H#9)Q#T+N2->@? y ©  A € %H;M2-IO#:
VJ#T1H#Vf#m86#9>6'-?(#?HJ6#;'*=,>@?'-.
=,>@?(#T1P%(#9IO?H=Q'*> y Ф  A € dÆG61P?HJ6#T19E7>6'-?H=,IO#?HJ2N?© ¬ ­ y(y0ఀ
2No' +-# € =,%ˆ=,>8#T;’#9>M86#9>c?f'-.AS2->8RJ6#9>MIO#XIT2->Ro#X86#T;M=,IO?(#98L2-%2bJ6'-1K=QgT'*>c?H2-)),=,>6#-d—J6#T1H#T.0'-1H#mVs#/2N1H#`>6'CVŠ=,>½2
;'*%H=Q?H=Q'*>?('4j-1P2N;MJ y0ఀ š y0à<à<ఀ '*> y ©  A € %(;M2-IO#m:s2-%=,),),G%H?(1P2N?(#98¤=,>½Æ\=Qj*G61H#bnNd
!$#¥2N1P#b=,>c?(#T1P#9%(?(#98$=,>È'-o?H2-=,>M=,>6j?HJ6#]'-;?H=,ab2-)
;'*),=,IOZ¢IPJ6'*=,IO#].0'-1`2¤j-'C+-#T1P>a]#9>@?`2N? ÒCÓ 2-%`2¤.0G>MIO?H=Q'*>£'-.
yÑФ A € d_—ˆJ=,%l=,%"1H#9ab2N1HóN2NoM)QZU%H=,a];˜)Q#\?(''-o?H2-=,>Xo@Zmab2N;;˜=,>6j“.0'-1"#92-IPJ]A Ýä‹ A Ú  A  ?HJ#;'*),=,IOZÜ á IO'-1H1H#9%(;'*>8=Y>6j
n
Govt
payoffs
| ¬ | ­ ÓTÚ
‚<®°¯„¯ Ú
ªH«T¬­ Ú
´
´
Ú
| ¬ | ­ D Ú
‚p®¡¯„¯
Deadweight
loss
ªH«T¬s­
µ´
| ¬ | ­ ´
Æ\=Qj*G61P#SnNk±G=,),=,o1P=,Ga ;’'*)Y=,IOZ¥IPJ'*=,IO#9%
?('¤?HJ6#¥© á ?HJM2N?UIO'*>M%(?H=Q?HG6?(#9%U?HJ6#!"#%$#9>@+-#9)Q'-;#S'-.?HJ6#S;M2 Z-'N¨%TE"=<dÙ#-dm.'-1X2->@Z£AcE"Üá B r Ü v'&( Ü*),+-/.F© Ó á y ÜNá € k
ÜNá Ý . Þ ‚ (ÞSß7(Þ&ßÚ1| 02| 0 d43m#T#T;M=Y>6jb?HJ=,%=,>¢ab=,>8lE˜'-oM%(#T1H+-#&?HJ2N??HJ6#&;M29Z-'N¨».1H'*a =,ab=,?H2N?H=Q'*>L86#9IO1P#92-%(#9%m2-% Ð¥ÓTÚ
=,>IO1P#92-%(#9%ˆ=ÑdÙ#-d 5© ¬ ­ 6 5 ÐbÓTÚRr š ð ² ÐbÓTÚ ” Œ 8d 7 ç(%HGMITIO#9%H%(.{G),è])Q#92-86#T1 hi%¥ç(>6#9=Qj*J@o'-1P%Hè]2N1H#/86#T3M>6#98½2-%?HJ6#`%(#T?
'-.IO'*G>@?(1P=Q#9%VJ6'*%(#`#O¨’#9IO?H=Q+-#]8=Y%(?H2->IO#`=,%%HG 9¥IT=,#9>c?H)QZ½%Ha¥2-),)"%HGIKJL?HJM2N??HJ6#TZ½Vf'*G),8L;1H#T.0#T1?('4=Yab=Q?H2N?(#/±G=Q?(#
=Q1H1P#9%(;#9IO?H=Q+-#m'-.\?HJ6#X;M2 Z-'N¨.1H'*a 2-)Q?(#T1P>2N?H=Q+-#`;’'*)Y=,IOZIKD J6'*=,IO#9%.'-12-),)"A*E’=ÑdÙ#-dš² Ð ´ÓTÚ ï šˆ²cAC´ ¶ y<º šân €° ¶ ¸ d
—ˆJ=Y%X%H=,a];˜),=Q3#9%?('¢2¢8=,%(?H2->MIO# Ð ´ÓTÚ W Ð ´D r ACÚ´ š ¾ ‹Qy<º š‘n €° ¶ ¸  d½—ˆJ6#T1H#T.0'-1H#-E\2-),)%HGIKJ–ç(>6#9=,j*Jco'-1P%HèR'-.
?HJ6#X)Q#92-8#T1ˆ=,ab=Q?H2N?(#X=Q?H%[;'*),=YIOZ-d Å >¤?HJ6#m'-?HJ6#T1ˆ#O57?(1H#9a]#XIO'*>%P=,86#T1“?HJ6#XITG6?„:<'N¨$| .0'-1“| ?HJ6#Rç„.{2N1(:<;#T1P=Q;MJ6#T1PZè7d_Æ\=Q1P%(?TE
'-oM%(#T1P+-#4.01H'*a Æ
=,j*G61H#¢n¥?HJ2N?S=Q.“8=,%(?H2->IO#4.1P'*a ?HJ6#¤)Q#92-86#T| 1S=,| %/%HGIPJu?HJ2N?b© ¬ ­ W¿© ªH«T¬s­ y A r A €¥y =ÑdÙ#-| d£|=Q.
šˆ² Ð ´ÓTÚ ¶ ¸ r šˆ²*³lµ´ € E@Vf#U#9=Q?HJ6#T1[J29+-#X© ‚<®°¯ ^©`ªH«T¬s­_^ © ¬ ­
.0'-1[AS” Am '-1fVf#UJ29+-#X©`ªH«T¬­_^ © ‚<®°¯ ^¦© ¬ ­
.0'-1A ï A@ dsÇU=Q+-#9>½?HJ6#XIO'*>@?H=,>G=Q?¡Z4'-.
?HJ6#m;M2 Z-'N¨%' +-#T1?HJ#m1H#9)Q#T+N2->c?ˆ%(;˜2-IO#-E?HJ=,%f#9%(?H2NoM)Y=,%HJ6#9%[?HJ6#m#O56=,%(?(#9>MIO#m'-.
2XITG6?„:<'N¨R8=Y%(?H2->IO# Ð ÓTÚ Ec%HGMIPJ]?HJ2N?s.'-1s2-),)A Ýä‹ A Ú  A  Ecj-'C+-#T1P>a]#9>@?H%f)Q'IT2N?(#98¥o#TZ-'*>8]?HJ=,%_ITG?„:<'N¨L>6'X)Q'*>6j-#T1
=,ab=,?H2N?(#S:ˆj*=Q+=Y>6j½G%m?HJ6#ç„.{2N1(:<;#T1P=Q;MJ#T1HZè7dbef'*G>@?(1P=Q#9%U?HJM2N?`),=Q#]2N?`%('*ab#ç(=,>c?(#T1Ka]#98=,2N?(#9èR8M=,%(?H2->IO#]o#T?pVf#T#9>
?HJ6#Èç(>6#9=Qj*J@o'-1P%Hèb2->8¢?HJ6#£ç„.02N1(:<;#T1P=,;MJ6#T1HZ7è y 2-%U86#T3M>6#98¢2No' +-# € E"IO'*>%(?H=,?HG6?(#`?HJ6#Èç(>6#92N1(:<;#T1P=,;MJ6#T1HZ7è7:d 7%IT2->
o#U1H#92-8=,)QZ¥'-oM%(#T1H+-#98¤.01H'*aÉÆ\=Qj*G61H#]nNEMIO'*GM>c?(1P=,#9%“)Q'IT2N?(#98L=,>?HJ=,%f1H#Tj*=Q'*>lEa¥29Z¤IPJ6''*%(# y 2-%ˆ2S.{G>IO?H=,'*>4'-.?HJ6#9=Q1
A € =,ab=,?H2N?H=Q'*>lE#O57;#T1P=,a]#9>@?H2N?H=Q'*>~'-14IO'-1H1PG6;?H=,'*>â2-%b?HJ6#9=,1];’'*)Y=,IOZuIKJ6'*=,IO#9%Td¦!$#L>6#O57?¥IPJ#9IHóâ?HJ6#½'-;M?H=,ab2-),=Q?¡Z
ð
'-.U?HJ6#¢IT=,?H=QgT#9>7:<+-'-?(#T19hi%41H#O:<#9)Q#9IO?H=Q'*>1PG)Q#-d¦§#T1P#L?HJ6#LIT=,?H=QgT#9>IO'*a];˜2N1H#9%b?HJ6#¢1H#92-),=QgT#98IO'*%(?¥'-.U1H#T;M)Y2-IT=,>6j$?HJ6#
=,>ITGMa`o#9>@? z | EV=Q?HJ»?HJ6#½#O57;#9IO?(#98~j*2-=,>~=,>»1H#9>@?H%]=Q.?HJ6#¤;’'*)Y=,IOZä2-86'-;?(#98åoZ»?HJ6#¤j-' +-#T1P>ab#9>c?¥=Y%]IO'-1H1PG6;?Td
ÇU=Q+-#9>¦?HJ6#ÈG>=,.'-1Pa 8=,%(?(1P=,oMG6?H=Q'*>~.0'-1 z E?HJ6#£1H#O:<#9)Q#9IO?H=Q'*>¦;1H'-o˜2NoM=,),=Q?¡Z»qñ=,%IO'*>c?H=Y>@G6'*GM%Tdñ—J=,%¥=Y%¥o#9IT2-G%(#
?HJ6#£IT=Q?H=QgT#9>lhi%4;M29Z-'N
¨ | =,%486#9IO1H#92-%H=Y>6juIO'*>@?H=,>G6'*%H)QZâ=,>?HJ6#L;M1H'-oM2NoM=,)Y=Q?pZ»?HJ2N??HJ6#L;'*),=,IOZå=,%IO'-1H1PG6;M?¥%H=,>IO#
q råtMv-w-xNy{z | ï } ´ `€ dˆÆ6'-1U=,>%(?H2->MIO#-E˜=Q._Ü Ó¥Ý¢Þ ßÚ E˜?HJ6#9>L.1P'*
a [29Z-#9%Th1PGM)Q#X?HJ6#`IT=,?H=QgT#9>7:<+-'-?(#T1Uó>' V%.'-1%HG61H#
Œ
r
?HJ2N?ˆ?HJ#X;'*),=,IOZ¥=,%J6'*>6#9%(?[:s2->M8½J6#9>IO#X
} qlÛ nNd
' V~Vf#“%(ó-#T?HIPJS'*G6?_2-)Q?(#T1K>2N?H=Q+-#[J=,%(?('-1P=,#9%V=,?HJ`?¡Vs'U'-1_a]'-1H#]ç(%HGITIO#9%P%(.0GM),è),#92-86#T1P%TE-=ÑdÙ#-d é Ú ^ënNd\! =Q?HJ6'*G6?
)Q'*%H%ˆ'-.
j-#9>#T1P2-),=Q?¡Z-EM%HG6;;'*%(#U?HJ6#T1P#`2N1H#X?¡Vs')Q#92-86#T1P% Ò Ú Ÿ 2->M8 Ò Ú d—ˆJ6#T1H#X2N1P#X?pVf'¥;'*%H%H=,oM)Q#U%HG6o6:pIT2-%(#9%9dsÆ\=Q1P%(?TE
=,%?HJ6#;'*%H%H=,oM=,),=Q?¡Z`?HJ2N?f>6'*>6#'-.l?HJ6#IO'*G>@?(1P=Q#9%f=,>¥?HJ6#.0'*),)Q'CVs#T1[IO'*À J6'-1H?f),=Q#9%%P=,a/GM)Q?H2->6#T'*G%H),Z/V=,?HJ=,>]2&8=,%(?H2->IO#
Ð ÓTÚ '-.o'-?HJb?HJ6#)Q#92-86#T1P% Ò Ú d\ò°>¥?HJ=,%sIT2-%(#-E@?HJ#;1H#9IO#98M=,>6j`2->2-),Z%H=Y%IT2N1H1P=Q#9%' +-#T1“2->8 ÒCÓ hi%f86#9IT=,%H=Q'*>b;1H'-oM)Q#9a
=,%m2-%/=Q.?HJ6#T1P#¥=,%X2½%H=Y>6j*)Q#])Q#92-86#T19d4§'CVs#T+-#T19E\Vf#¥>6'-?(#b?HJM2N? ÐbŽ ÓTÚ
y0é Ú r ð € ^ Ð¥Ž ÓTÚy0é Ú r n € %P=,>IO#bJ=Qj*J#T1`A
IO'*G>@?(1P=Q#9%&2N1H#4ab'-1H#4),=,ó-#9)QZ¢?('È=,ab=,?H2N?(#-E?HJ6#T1P#To@Z)Q'CVs#T1K=,>6jL?HJ#429+-#T1K2Nj-#½A y 2->8uJ#9>IO#4q € '-.“?HJ6'*%(#¥VJ6'¢86'
>6'=,ab=Q?H2N?(#-dˆò„>L8M=,2Nj-1P2-abab2N?H=YIX?(#T1Pab%9EM?HJ=,%=,%#9±@G=,+F2-)Q#9>@?ˆ?('¤248' V>@Vf2N1K8L%HJ=,.?ˆ'-._?HJ#]© ‚<®°¯„¯  ©`ªH«T¬­sITG61H+-#9%
:“=,>IO1H#92-%H=Y>6j ÐbŽ ÓTÚ d Å oM%(#T1H+-#&?HJ2N?.'-1m2->@
Z y{Ò@€ ElVf#&IT2->È2-)QV[29Z7%UIPJ6''*%(#]2²4),2N1Pj-#/#9>6'*Gj*J£%H'2-%?('#9>%HG61H#
?HJ2N?“?HJ=Y%f8' V>@Vf2N1K8½%HJ=Q.0?['-.?HJ6#/©`‚<®°¯„¯2->8L© ªH«T¬­ =,%“>6'-?ˆ%H'S)Y2N1Hj-#X%(']2-%“?('bab2Nó-#`=Yab=Q?H2N?H=Q'*>?HJ6#m86'*ab=Y>2->c?
%(?(1P2N?(#Tj-Z.0'-1[2-)Y)˜IO'*G>@?(1P=Q#9%[=,
> y{Ò@€ d[¼#9IO'*>8lE6=,%s?HJ6#;’'*%P%H=QoM=,)Y=Q?pZ&?HJ2N? ÒCÓ ),=Q#9%so#T?pVf#T#9>½2->84V=Q?HJ=,>42S8=,%(?H2->IO#
ŽÐbÓTÚ '-._?¡Vs'¤'-1Ua]'-1H#/IO'*GM>c?(1P=,#9% Ò Ú 2->M8 Ò Ú E’=<dÙ#-d Ò Ú š Ò Ú ˜” ð Ð¥Ž ÓTÚ dò°>L?HJ=,%IT2-%H#-E2¥.'*)Y)Q' Vf#T1IO'*GM>c?(1HZ’hi%
;M2 Z-'N¨$=Y%“ab2F57=Yab=QgT#98¤oZ=,a¥=Q?H2N?H=,>6j&Ÿ ?HJ6#`)Q#92-8À #T1“V=Q?HJ¤VŸ J6'*aÉ=,À?ˆJ2-%[?HJ6#`)Q#92-%H?bç„#O¨’#9IO?H=Q+-#9è48=,%(?H2->MIO#X2->8¤?HJ6#T1H#
=,%[>6'&#O5;#T1P=Ya]#9>c?H2N?H=,'*>4oZ42->@Z¥'-.?HJ6#mIO'*G>@?(1P=Q#9%[)Q'7IT2N?(#98¤o’#T?¡Vs#T#9>R?HJ6#9%(#U?pVf'ç(%HGITIO#9%H%H.0G)Yè/)Q#92-8#T1P%TdÆ\=,>2-),),Z-E
=Q.
>6'ç(%HGITIO#9%H%H.0G)Yè])Q#92-86#T1P%“#9ab#T1Hj-#98R=,>¤?HJ6#X),#92-86#T1IO'*J6'-1H?TEM?HJ6#9>R86#9IT=,%P=Q'*>½ab2Nó7=,>6j]=,>¤?HJ6#X.0'*),)Q'CVs#T1IO'*J6'-1H?ˆ=,%
=,86#9>@?H=,IT2-)?('b?HJM2N?=,>¤?HJ6#`),#92-86#T1IO'*J6'-1H?Td
ÊMË9Ì͒Ì0ÎäTC 7%f2/o#9>IPJMab2N1Hó’EcVf#31K%(?s86#9%PIO1P=Qo#ˆ?HJ6#U%(#T?['-."IO'*G>@?(1P=Q#9%f.'-1fVJ6'*aÔ?HJ6#31P%(?„:<o#9%(?s;'*),=,IOZ
=,>@+-'*)Q+-#9%
=,ab=Q?H2N?H=Q'*>`2->8/2-8'-;?H=Q'*>X'-.=Y>%(?H=Q?HG6?H=,'*>%l86#T+-#9)Q'-;#98XoZX2),#92-86#T19d 7~IO'*G>c?(1PZmV=,),)-=,a¥=Q?H2N?(#s2)Q#92-86#T1 Ò Ú
1P2N?HJ6#T1_?HJM2->S#O5;#T1P=,ab#9>c?_=Q.’© Ó | ¬ | ­ ï © Ó ªH«T¬ E*VJ6#T1P#f?HJM=,%\=,>6#9±G2-),=Q?¡Z`=,%
?(1PG6#“%('X)Q'*>jU2-% Ð ´Ó Wâ³ µ´ š ¾ ‹ q · šSq Û  d
§' Vf#T+-#T19E7?HJ6# $#;'*),=YIOZ/#9>@?H2-=,),%=Yab=Q?H2N?H=Q'*>]'*>M)QZ&=,.’=Q?s=,%_'-;?H=,a¥2-)6?('`=,ab=,?H2N?(#=,>]?HJ6# #"! #'-.;'*),=Q?H=YIT2-)
IO'*>Ó9%HÓ =Y86#T1P2N?H=Q'*>% y =ÑdÙ#¥VJ#9> ¸ #9±G2-),%`gT#T1H' € dR—ˆJ=Y%X=,a];M)Y=Q#9%?HJ2N?`?HJ6#4%# 9¥IT=Q#9>@?/8=Y%(?H2->IO#b.0'-1/=,a¥=Q?H2N?H=Q'*>$#9±G2-),%
Ð r ÒFÓ $
š Ò# Ú r ³ µ d
ò„>]VJ2N?_.0'*),)Q'CV%_Vf#ˆo#Tj*=,>SoZ&IO'*>%P=,86#T1P=,>j?HJ#ˆJ=,%(?('-1HZ]Ï D EcV=Q?HJ é Ú r nNdÆ
=,1P%(?TE*?HJ6#4ç(>6#9=Qj*J@o’'-1K%HèU2N1H#ˆ?HJ6#
%(#T?s'-.IO'*GM>c?(1P=,#9%VJ6'*%H#8=,%(?H2->MIO#ˆ.1P'*a D ?HJ6#),#92-86#T1=,%s%HG 9¥IT=Q#9>@?H)QZ&IT)Q'*%(#%PGIPJ¥?HJ2N??HJ6#TZb=,a¥=Q?H2N?(#ˆ?HJ6#),#92-86#T1s2-),)
A y %(#T#UÆ\=Qj*G61H#`n € E7=ÑdÙ#-d_=Q. Ð ´D WAC´š Ã-ƒ¾  d 9¥IT=Q#9>@?8=,%HIT=,;M),=,>6#f=,a];M)Y=Q#9%\?HJ2N?s=Yab=Q?H2N?H=Q'*>]'-."2/)Q#92-86#T1s#9>J2->IO#9%
;1P=,+F2N?(#%(#9IO?('-1\>2N?H=Q'*>M2-)c=,>IO'*ab#-E #9=Q?HJ6#T1o@ZU1H#98MGIT=,>6j[?HJ6#IO'*%H?H%'-.6IO'-1H1PG;?H=Q'*>m'-1?HJ6#IO'*%(?'-.| #O57| ;#T1P=,a]#9>@?H2N?H=Q'*>ld
—ˆJ6#T1P#T.'-1H#-E-.0'-1\8=,%HIT=Q;˜),=,>=,>js?('Uo#s%# 94IT=Q#9>c?
.0'-1\2->@Z`IO'*G>@?(1HZ ÒCÓ6y A PÐ¥€ E@=Q?
=,%?HJ6#“IT2-%(#[D ?HJ2N?© Ó ¬ ­ ï ©`Ó ‚<®°¯„¯ d\—ˆJ6#
),2N?(?(#T1f=,%\?(1PG6#“VJ#9>lEMšˆ² Ð |´ Ú ¶ qlÓÛ ï šˆ²cACÓ´ ‚ ¶ q"Ó¹ ‹Qy<º šRn €° ¶ ¸ & % Ð ´ÓTÚ W . y A ´"š ¾ y q"¹ ‹Qy<º šLn €° ¶ ¸  ¶ ¸ € 0 ´*ã
' V 1H#9IO'*),)Q#9IO?4?HJ2N?¥#%9¥IT=,#9>c?=,ab=Q?H2N?H=Q'*>'ITITG1P%bVJ#9> ÐbÓTÚ W ³ µ E[VJ6#T1P#L³ µ ” A Ú d뗈J6#T1H#T.0'-1H#-E[.0'-1¥?HJ6#
=,>6.0'-1Pab2N?H=,'*>2-)l#O57?(#T1P>2-),=Q?¡Z½?('IT2-GM%(#/%# 94IT=Q#9>c?8=,%HIT=,;M),=,>=Y>6j6E@?HJ#`.0'*),)Q' V=,>6jb?¡Vs'½IO'*>8=,?H=Q'*>%ˆa&G%(?o’#&a]#T?Tk y 2 €
?HJ6#T1H#½#O56=,%(?H%]2$>'*>7:<#9a];?¡Zu%(#T?]'-.UIO'*G>@?(1P=Q#9%SVJ=YIPJå2N1H#R8=,%PIT=Q;M),=,>#98=,>@?('=,ab=Q?H2N?H=,>j£2$)Q#92-8#T1].'-1];'*),=Q?H=YIT2-)
1H#92-%('*>M%TEoMG6?“Vf'*G),8½>'-?86'b%('b'-?HJ#T1HV=,%(#-E2->M8 y o € .0'-12b%HGoM%(#T?“'-.\?HJ6#9%(#`IO'*GM>c?(1P=,#9%TE6;1P=Q+N2N?(#O:p%(#9IO?('-1U=,>IO'*a]#
=,>IO1P#92-%(#9%`8G#b?('R?HJ6#41H#98GIO#98IO'-1P1PG6;?H=Q'*| >| ?HJ2N?`1H#9%PG)Q?H%mo#9IT2-G%(#¥D '-.“j-1H#92N?(#T1&8M=,%HIT=Q;M)Y=,>=,>6j6dS' V .'-1 y 2 € ?('
J6'*),8"E;˜2N1P2-a]#T?(#T1P%ˆa&G%(?“o#X%HGMIPJ¤?HJ2N?© ¬ ­ ^ ©/‚<®°¯„¯=ÑdÙ#-d=,.
A ´ ¶ ¾ . ¸ y nˆš$q ¹ € š$q ¹ y<º š~n €° 0 ï Ð ´Ó9Ú d Å >
?HJ6#U'-?HJ6#T1“J2->8"E@.0'-1 y o € ?('SJ'*),8lE;M2N1P2-a]#T?(#T1P%“a&G%(?fo#%PGIPJ?HJ2N?mšˆ² Ð ´ÓTÚ ^뚈²@A ´š  d_ò¡?[=,%f#92-%(Z4?('SIKJ6#9IHó
?HJ2N?ˆo'-?HJ y 2 € 2->8 y o € 2N1H#/%H2N?H=,%(3#98R.'-12¥%HG9¥IT=Q#9>@?H)QZ)Q' Vëo'*G>M84'*>¢A Ú 2->8 ¸ d
!$#X>6'CVë86#9a]'*>%H?(1P2N?(#U?HJ2N?[?HJ6#T1H#U#O57=Y%(?[;M2N1P2-a]#T?(#T1P%“=Y>¥?HJ6#½ç(>6#92N1H:<;’#T1K=Q;MJ6#T1HZ7è7E?HJ2N?“8=,%(;M)Y29Zç(=,>6%# 94IT=Q#9>c?
8=,%PIT=Q;M),=,>M=,>6j*è7d"—ˆJ6#T1P#2N1H#U?¡Vs'S;'*%H%H=Qo˜=,),=Q?H=Q#9%9k y 2 € IO'*GM>c?(1P=,#9%s?HJM2N?fVf'*G),84'-?HJ6#T1HV=,%(#JM29+-#m;1H#T.0#T1H1H#984?('S#O5;#T1P=Q:
a]#9>@?“V=Q?HJ2->½J6'*>6#9%(?[;'*),=,IOZ y =ÑdÙ#-d_?HJ6'*%(#mV=,?HJJM=Qj*J½A € 2N1H#m=,>%# 9¥IT=Q#9>@?H)QZ¥8=,%PIT=Q;M),=,>#98b=,>@?('S=Yab=Q?H2N?H=,>6j&2])Q#92-86#T1
;1P=Yab2N1P=,)QZ¤.0'-1m;'*),=Q?H=,IT2-)1H#92-%('*>(% ' y o € IO'*G>@?(1P=Q#9%U?HJ2N?mVf'*G),8ÈJ2 +-#bIPJ6'*%(#9>äIO'-1H1PG6;?;'*),=,IT=Q#9%9E’2N1H#b8M=,%HIT=Q;M)Y=,>6#98
)
=,>@?('R2-86'-;?H=,>jRJ6'*>6#9%(?X;'*),=YIT=Q#9%TEloMG6?m?HJ#b)Q'*%H%X=,>$=Yab=Q?H2N?H=,>6j½2½;'*),=YIOZL'-.“2R8=,%H?H2->c?XIO'*G>@?(1HZÈ=,%mj-1H#92N?(#T1&?HJ2->
?HJ6| #Uj*| 2-=,>½8G#?('b2S1H#98MGIO?H=Q'*>¤=,>IO'-1H1PG6;?H=,'*>ld_Æ
=,1P%(?TE7Vs#X8#9%HIO1P=Qo#;˜2N1P2-a]#T?(#T1P%ˆ%PGIPJ?HJ2N? y 2 € =,%[%H2N?H=Y%(3#98lk_=ÑdÙ#-d
© ¬ ­ ï © ªH«T¬­ =ž¨È³ µ´ ¶ ¾ y q Û š$q · € Ó9ï Ó Ð ´ÓTÚ ds¼7=,>IO#mq Û r nm2->8½q · ”ën  Vf#`J2 +-#X=,>6%# 9¥IT=Q#9>@?“=,ab=Q?H2N?H=,'*>¤.'-1
2S),2N1Hj-#U#9>6'*G6j*J ¸ y 1P#9IO'*),)Q#9IO?“?HJ2N? Ð ”â³ µ € d y o € ' V–IO'*>%P=,86#T1[?HJ6#;’'*%P%H=QoM=,)Y=Q?pZS'-.=Y>6#%9¥| IT| =Q#9>@?f8M=,%HIT=Q;M)Y=,>=,>6j
'-.“IO'-1H1PG6;M?mj-' +-#T1P>ab#9>c?H%Td¤Æ6'-1&?HJ=,%U?('¢o’#]?(1KG6#-E8=Y%(?H2->IO# ÐbÓTÚ J2-%m?('¢o#b%HGIPJ?HJ2N?S© ¬ ­ ï ©X‚<®°¯„¯FE#T+-#9>
?HJ6| '*G6| j*J»šˆ² Ð ´ÓTÚ ”™šˆ²cAC´ˆD š  d—'¤%(#T#&?HJ2N?;˜2N1P2-a]#T?(#T1P%#O57=Y%(??HJ2N?U%H2N?H=,%H.Z½?HJ6#9%H#&IO'*>8M=Q?H=Q'*>%TE˜'-oM%(#T1H+-#&?HJ2N?
© ¬ ­ ï © ‚<®°¯°¯ % AC´ ¶ ¾ ‹ ¸ y qlۓšäq"D ¹ € šäq"¹ y<º šn €° ï Ð ´ÓTÚ d_ò°ab;’'*%P=,>6jS?HJ6#X2No' +-#&IO'*>8=Q?H=Q'*>¤=,ab;M),=Q#9%f?HJ2N?
2½%HG94IT=Q#9>c?mIO'*>8M=Q?H=Q'*>È=,%m=Q.sAC´ ¶ ¾ ‹ ¸ y qlۚ»q"¹ € šåq"¹ y<º š¦n €° ï Ð ´ÓTÚ ^™A ´ ¶ † ¾ ã¼6=,a];M),=,.Z7=,>6j]Vf#]'-o?H2-=,>
?HJ6#U.0'*),)Q' V=,>6j ¸ y qlۚ£q"¹ € ^ by n ¶ t ¹ y<º šun €(€ VJ6#T1H#UqlÛ r nm2->8¤q"¹å”ënNd_ò°?[=Y%s#92-%HZ4?('bIKJ6#9IHó?HJ2N?“?HJ=,%f=,%
?(1PG6#m.0'-1ˆ#Tj-'41H#9>c?H%ˆ?HJ2N?2N1H#`%PG9¥IT=Q#9>@?H)QZ4)Y2N1Hj-#m:2]1H#9%H?(1P=,IO?H=Q'*>½IO'*>%P=,%(?(#9>@?ˆV=Q?HJ¤'-?HJ6#T1K%=,>¤?HJ6#m;˜2N;’#T1 d
ò„>S'-1P86#T1_?('X%HJ' V~?HJ2N?_#O57;#9IO?(#98b=,>IO'*a]#“=,>&?HJ6#]ç(>6#92N1H:<;’#T1K=Q;MJ6#T1HZ7è=,%_)Q' Vf#T1_?HJ2->]=,>&?HJ6| #b| ç„.{2N1(:<;#T1P=Q;˜J6#T1HZ7è7E
#O562-ab=,>6#mÆ\=Qj*G61H#]nNdÆ1H'*a¿?HJ6#m;1H#9IO#98M=,>6j&;M2N1P2Nj-1P2N;MJ"E?HJ#T1H#X=,%[=,>6%# 9¥IT=,#9>c?“=,ab=,?H2N?H=Q'*>¤=Q.
© ¬ ­^ ©XªH«T¬­„E6#T+-#9>
?HJ6'*G6j*J½%PGIPJR=Yab=Q?H2N?H=Q'*>1H#9%HGM)Q?H%[=,>½2]86#9IO1H#92-%(#X=Y>¤;1P=Q+N2N?(#m>2N?H=Q'*>2-)l=,>MIO'*a]#m2-%ˆIO'*a];M2N1H#98R?(']#O5;#T1P=,ab#9>c?H=,>j6E
=ÑdÙ#-d‘#T+-#9> ?HJ6'*G6j*Jñšˆ² Ð ´u” šˆ²*³lµ´ d ˆ#9IO'*)Y)Q#9IO?¥?HJ2N?¤=,ab=,?H2N?H=Q'*>â=,%¥.'-1;'*),=Q?H=,IT2-)[1H#9Ó 2-Ó %H'*>% y =<dÙ#-d ¸ € VJ6#9>2
j-' +-#T1K>a]#9>@?X=,ab=Q?H2N?(#9%m86#9%H;M=Q?(#&?HJ6#S#O¨’#9IO?H=Q+-#b8M=,%(?H2->Ó9IOÓ # ÐbÓTÚ o#9=,>j¥j-1H#92N?(#T1X?HJM2-> Ð r ³ µ d`' V %HG6;M;’'*%H#
?HJ2N?%HGIPJ 2j-'C+-#T1P>a]#9>@?)Y=Q#9%¥2N?%('*a]# Ð # ^ Ð d Å oM%H#T1H+-#R?HJ2N?4.'-12->cZâIO'*G>@?(1H| Zâ| )Q'7IT2N?(#98¦2N?4%HGIKJ2
Ð # E";'*),=,?H=,IT2-)\IO'*>%H=,86#T1K2N?H=Q'*>%#9>%HG61P#]?HJ2N?m?HJ6#bIO'*G>@?(1HZÈ=,>6#%9¥IT=Q#9>@?H)QZ£=,ab=Q?H2N?(#9%TE=ÑdÙ#-d¥© ¬ ­ y Ð¥# € ^ ©`ªH«T¬­ y Ð¥# € d
—ˆJ6#9>?HJ6#T1H#4=Y%X=,>6%# 9¥IT=,#9>c?X=,a¥=Q?H2N?H=Q'*>$oZ$2-),)IO'*G>@?(1P=Q#9%`V=Q?HJäA ݖ‹ A #  A  d Å oM%(#T1H+-#b?HJM2N?/2-),)IO'*GM>c?(1P=,#9%XV=Q?HJ
AR=Y>ä?HJ=Y%/1K2->6j-#¤oMG6?&V=Q?HJ ÐbÓTÚ ^ Ð ÓTÚ V=,),)V=Q?H>6#9%H%&2->å=,>IO1H#92-%H#¤=Y>u=,>MIO'*a]#2-%]2-)Y)j-'C+-#T1P>a]#9>@?H%b=,>»?HJ=,%
1H#Tj*=Q'*>4;1H#T.0#T1f?('/#O57;#T1P=,a]#9>@?Td¼7=,>IO# Ð ´ÓTÚ ^³ µ´ .0'-1fIO'*G>@?(1P=Q#9%f=,>¥?HJ=,%1P#Tj*=Q'*>lE;1P=Q+N2N?(#>2N?H=,'*>2-)˜=Y>IO'*a]#.01H'*a
#O57;’#T1K=,a]#9>@?H2N?H=Q'*>¤=,%j-1P#92N?(#T1ˆ?HJ2->¤=,ab=Q?H2N?H=,'*>ld_—ˆJ=,%s=,a];M),=,#9%2SJ=Qj*J6#T1f#O57;’#9IO?(#98½=Y>IO'*a]#.'-1“IO'*G>@?(1P=Q#9%[=,>?HJ=,%
1H#Tj*=Q'*>"dfÆ\=,>2-)Y)QZ-E=Q?=,%ˆ#92-%(Z½?('4IKJ6#9IHó½?HJ2N?.'-1IO'*G>@?(1P=Q#9%=Y>½?HJ6#Lç„.{2N1(:<;#T1P=Q;MJ#T1HZè7E>'*>6#X'-._?HJ6#`IO'*GM>c?(1P=,#9%2N1H#
8=,%PIT=Q;M),=,>#98½=,>c?('#9>2-IO?H=,>j¤J'*>6#9%(?;’'*)Y=,IT=Q#9%Td—ˆJ6'*%(#&V=Q?HJ£2¤J=Qj*JÈA*E"%HGMIPJ£?HJ2N?`© ªH«9¬­ ^–©/‚<®°¯„¯ EVs'*GM),8¢JM29+-#
IPJ'*%(#9>½J6'*>6#9%(?[;'*),=,IT=,#9%s=,>¤2->@Z4IT2-%(#-ds§'CVs#T+-#T19E’%H=,>IO##O57;’#T1K=,a]#9>@?H2N?H=Q'*>¤=,%fo’#9=Y>6j/IT2N1H1K=Q#98'*G6?“=,>?HJ=,%f1H#Tj*=Q'*>lE
?HJ6#X+N2N1P=,2->MIO#X=,>½=,>IO'*a]#m'*G?HIO'*a]#9%=Y%“J=Qj*J½2->8½j*=,+-#9>½o@Z¤³ µ´ d