Transcript Ec305 Lecture 2 - University of Hong Kong

Political Economy of
Economic Policy
Lecture 4-5
Government institution:
Organizational Forms and Incentives
Based on Qian and Xu, 1993 and Maskin, Qian and Xu, 2000
Chenggang Xu
Copyright @ Chenggang Xu
How are governments organized?
• Most countries in the world are too large to be ruled by their national
governments alone
• Regional and local governments are necessary
• The relationship between national and sub-national governments has
important impacts to incentives
• Unitary state
– A unitary state has a single ultimate sovereignty, the central government
possesses all governmental authority
– When delegating political power to regional governments, political power still
ultimately rests with the central government
• Any decisions they make can be overturned by the central government.
– Examples: France, Japan, Britain, Italy, Spain, Indonesia.
• Federal state
– A federal state has a dual location of sovereignty, with authority divided
between central and regional governments
– The self-governing status of the regions is typically constitutionally entrenched
and may not be altered by a unilateral decision of the central government
– Examples: US, Germany, Russia, India, Brazil
Fundamental Chinese institution:
Regionally decentralized authoritarianism
• Highly centralized personal controls – in contrast to a federal system
–
–
–
–
–
Provincial level officials are controlled by the central government
Municipality level officials are controlled by their upper provincial governments
Nested personnel controls over all lower level regional officials
Personal control as incentive instruments for the central government to lead
The central government can overrule subnational governments’ decisions
• Highly decentralized in resource allocation – in contrast to an unitary
state system
– Example: land is de jury owned by the national government; de facto owned by
subnational governments
• Subnational governments run the economy
– Most firms are either under direct control or under great influence of subnational
governments
• Regions (provinces, cities, counties) are relatively self-contained
– Provide conditions for regional competition and regional experiments
• The fundamental institution has been stable although the degree of
decentralization varies over different periods
Governance Structure of Chinese Economy
Government as bureaucracy
• Beyond property rights protection: the role of government
– Contain social disorder
• Mobilizing resources for development
• Continental Europe and Japan in 19th Century
• All governments are organized as bureaucracies
• Bureaucracies are not restricted to governments (firms, churches, etc)
• Centralized economies is an extreme for the role of government
– Extreme cases are helpful for understanding
Bureaucratic hierarchy
Principal
Agent 1
Agent 2
Agent 3
Centralized Economy vs. Market Economy
• A pure centralized economy
– No market
– Only bureaucratic coordination
• Within state-owned enterprises and between them
• A market economy
– Within firms: bureaucratic coordination
– Between firms and households: market coordination
• Transition from a centralized economy to a market
economy
– To replace bureaucratic coordination between firms by markets
• Product markets; labor markets; financial markets; market institutions
– Ownership of the firms
– Legal protection of the property rights and enforcement of contracts
Organizations of centralized
economies and firms
• A national economy should be organized as a single firm –
Lenin (The State and Revolution), before the 1917
revolution
• Stalin invited top Western executives of large firms to
advise setting up a centrally planned economy, in the
1920s
• All other centralized economies copied the Soviet model
of central planning system
• Importance in analyzing organizational forms
– To understand operation of centrally planned economies
– To understand transition
– To understand firms in market economies
Greatest possible role of government:
Centralized economies
-Under certain conditions, centralized economy help development
-Russia created centralized economy in the late 1920s
-From one of the poorest in Europe (1920s) to a super power (1950s-80s)
- China copied centralized economy from USSR since 1950
-From one of the poorest in the world to a regional power
-Under certain conditions centralized economy hinder development
-Stagnation in USSR and CEE countries since mid 1970s
-Reforms/transitions triggered by those problems
-Great differences in reforms:
Table: USSR vs. China GDP (million 1990 International Dollars)
USSR
China
1950
510243
239903
1960
1980
1989
1998
2001
843434
448727
1709174
1046781
2037253
2044100
1124868
3873352
1343230
4569790
Objective of Reforms
- Abolishing the vertical relationships between SOEs and
the government.
- Replacing the bureaucratic inter-firm coordination by
market coordination.
- How to do it?
- In economics, do we know how to create markets from
scratch?
- Privatization (change of ownerships); Reform/establish
financial institutions; Reform/establish legal institutions
- Why reforms results are so different in FSU, CEE and
in China?
Why do performances of reforms in
China and in EEFSU so different?
- Sharp difference between reform performances:
-
In 1978-2002 Chinese per capital GDP was increased by 500%
In 1989-98 FSU economies per capital GDP was reduced by 50%
- Why?
- Reform strategy difference: gradual (China) vs. Big bang (EEFSU)
- However, China studied gradual reforms in EE
- Development stage: backward advantage (China) vs. Overindustrialization (EEFSU)
- However, is backwardness always beneficial? Africa? 1800-1950 China?
- Current Chinese per capita GDP is similar to that of Russia
- not as backward as people might thought
Explain performance by organizational forms
• Both China and EE-FSU were centrally planned economies
• EE-FSU were organised on the principle of functional specialisation
- Every SOE was under the control of a specialized ministry (Uform)
- Each ministry’s function complements with other
ministries’functions.
-the Chinese economy was organised into an M-form.
-Most SOEs were under the control of regional governments.
-Each region is relatively self-contained.
Soviet Union (U-form)
Central Gov't
Steel Ministry
SOE AS
SOE BS
Chemistry Ministry
SOE AC
SOE BC
different organizational forms of centrally planned economies
China (M-form)
Central Gov't
Province A
SOE AS
SOE AC
Province B
SOE BS
SOE BC
Scale Economy & Planning Complexity
Firm Sizes: China, EEFSU and the West, 1988
(Employment per Firm)
Czechoslovakia
The Soviet Union
Hungary
China
Italy
2,930
806
460
145
96
Planning indicators, late 1970s
China
760
Always < 1000
Soviet Union
12,000,000
Fundamental institution:
Regionally decentralized authoritarianism
• Authoritarian regime with decentralized regional economies
– Central government controls key factors
• Appointment of governors; National law; Mass media; tax rule; plan quota
for land use, for environment, etc.
– Regional governments run the economy (Wong, 2005; WB, 2002)
• More than 99% SOEs are under regional government control
• Almost all firms in non state sector are under regional government control
(regulation and resource allocation)
– Path dependent: the fundamental institution – regionally
decentralized authoritarianism – is evolved in history before
reform started (Perkins, 1964, 1975, 1977; Granick, 1990; Qian
and Xu, 1993; Naughton, 1991, 1995)
• First wave of regional decentralization: late 1950s
• Second wave of regional decentralization: ‘The Cultural Revolution’
Decentralization: Chinas vs. Other Countries
(Wong, 2005)
Nations
China 2003
Developing Countries 1990s
Transition Countries 1990s
OECD Countries 1990s
Other Large Countries 1990s
Sub-National
Expenditure/National
Expenditure (%)
70
14
26
32
Germany
India
Japan
40
46
61
Russia
USA
38
46
Regional decentralization and
Successful earlier reforms
• Almost all successful earlier reforms are based on
regional experiments, regional competition
• Basic reform policy is for regional competition: ‘let
some regions become rich first’
• Regional experiments initiated by regional governments
– Household responsibility system initiated by Fengyang and
promoted to regions nationwide
• Regional experiments initiated by the central
government and promoted nationwide later
– Special economic zones
– Fiscal decentralization: started in Jiangsu
Impacts of regional decentralization
on growth
GDP Growth
Per capita GDP
(10000Yuan)
Per capita
GDP
Growth
%
1978
2004
%
9.4%
379
2880
8.1%
Jiangsu
12.4%
430
7024
11.3%
Zhejiang
13.1%
331
6586
12.2%
Guangdong
13.4%
369
6040
11.4%
Average of above
12.9%
377
6550
11.6%
Liaoning
9.0%
680
5214
8.1%
Jilin
9.7%
381
3068
8.4%
Helongjiang
7.9%
564
3321
7.1%
Average of above
8.8%
542
3868
7.9%
National
More Regionally Controlled
More Centrally Controlled
Growth and Regional Decentralization
(Lin and Liu, 2000)
• It is a reduced form regression model on impacts of major reforms
on growth
• Fiscal decentralization (FD) significantly affects growth
• Regional decentralization based reforms and variables
– Household responsibility system (HRS)
– The share of non-SOEs' output in the total industrial output (NSOESH)
– The growth rate of per capita investment (GI)
• Variables with insignificant impacts to growth
– Financial strength of a region: Fiscal capacity (FISCAP): 3-year moving
average of per capita GDP
– Impacts of urbanization and the size of the population on economic growth:
The percent of rural population (POPSHR) and the total population (TPOP)
– Price liberalization: relative price of farm products to non-farm products
(FPMP)
Growth and Regional Decentralization
(1970 to 1993. Lin and Liu, 2000)
Fiscal Decentralization and
Non-state Sector Development
(Jin, Qian and Weingast, 2005)
China vs. FSU-CEE Reforms:
With or without Higher Level
Competition
• Chinese regions are in tough competition
– ‘Getting rich first’ as a reform slogan
– Regions (provinces, cities) are ranked
• FSU ministries/regions were not in competition
– Competitions happened only at bottom level
– High level officers/managers incentives are more
important
• Why cannot FSU mimic Chinese practice?
• What prevents FSU from competition?
Organizational forms affect incentives (1)
• Chinese local governments were given greater incentives
to introduce reforms than their counterparts in the EEFSU
• M-form is better than U-form in providing managerial
incentives
• managers need incentives and should be awarded
conditional on performance.
• performance may not be perfectly correlated with
managerial effort due to the uncertainty faced by firms.
• if a similar region performs well a failing manager is hard
to find excuse => it is desirable to make the manager’s
reward conditional also on his performance relative to that
in other regions.
Organizational forms affect incentives (2)
• why cannot the same be done in a U-form?
• Comparing specialized ministries (U-form) is more
difficult than comparing self contained regions (M-form)
• e.g. Comparing apples with banans vs comparing a basketof-fruits with another basket-of-fruits.
Principal-agent model solution (one agent)
• The principle choose incentive scheme t(q(.)) to maximize
expected profit:
Max E [r(g(e, ε)) - t(g(e, ε))]
e, t(.)
s.t.
e ε argmax E [V(t(g(e', ε)))] - H(e)
E[V(t(g(e', ε)))] - H(e)  U
(IC)
(IR)
• The solution is:
t = T(q).
• Var(ε) and/or corr(e, q) => e  induced by t.
Principal-agent model setup (multi-agents)
• Principal's utility: Up(r(q)-t) = i {ri (qi)- ti}
• Agent's utility:
Ui (ti, ei) = V(ti) - H(ei)
• Optimal incentive scheme t(q(.))
Max i E{ri (qi(ei, εi)) – ti(qi(ei, εi))}
e, t(.)
ei ε argmaxE[V(ti (g(ei', εi)))]-H(ei)
e i'
E[V(ti (g(ei', εi)))] - H(ei)  U
(IR)
s.t.
(IC)
Principal-agent model setup (multi-agents)
• In general, ti = Ti (qi, qj, qk)
• In a linear case, ti = aiqi + ajqj + akqk
• ai  0, aj  0, and ak  0, if cov(εi, εj) > 0 , cov(εk, εi ) > 0 ,
cov(εk, εj) > 0 : agent i competes against agents j and k.
• aj = 0, and ak = 0 if cov(εi, εj) = cov(εk, εi ) = cov(εk, εj) =0.
• If εi εj and εk are perfectly correlated, a yard-stick
competition (rank agents by their relative performance) can
serve as a filter for these shocks such that performance can
be inferred from output independent from shocks. =>
achieve first best.
• Organization may affect information set εi εj and εk
Information and Incentives:
Intuition
• Hard to compare apples with bananas
• Easier to compare one basket of fruits with
another basket of fruits
• Organisation as a basket of different tasks
– U-form (specialisation in tasks): put all apples into
one basket and all bananas into another basket
– M-form (based on region): mix apples and bananas
into baskets
Organisational Forms & Incentives(U)
U-form Organisation
X1A
Top
Shock 
Output:
, X1B , X2A , X2B
Ministry 1
Shock 1
Output:
X1A , X1B
Plant 1A
Shocks 1, A
Output:
X1A
Plant 1B
Shocks 1, B
Output:
X1B
Ministry 2
Shock 2
Output:
X2A , X2B
Plant 2A
Shocks 2, A
Output:
X2A
Plant 2B
Shocks 2, B
Output:
X2B
Organisational Forms& Incentives(M)
M-form Organisation
Top
Shock 
Output:
X1A , X2A, X1B , X2B
Region A
Shock A
Output:
X1A , X2A,
Plant 1A
Shocks 1, A
Output:
X1A
Plant 2A
Shocks 2, A
Output:
X2A,
Region B
Shock B
Output:
X1B , X2B
Plant 1B
Shocks 1, B
Output:
X1B
Plant 2B
Shocks 2, B
Output:
X2B
Basic model
• Output of plant ir
• The cost of effort e is C(e) where C(0) = 0, and C’(e)>0,
C’’(e)>0
• A manager’s utility is U(t) – C(e)
– e is unobservable
– t is the compensation, which depends on output {xir}
• The organizational problem to choose
– a set M of managers and
– a set of reward schemes for each manager j so as to maximize the
expected net output
– Subject to IC: e* = argmax E{U(t)-C(e)}
– and IR : E{U(t)-C(e)} U
Information and Incentives
• The general central point: different organizational
forms give rise to different information, which
determines incentives
• The specific central claim: if there is less "variation"
in shocks across regions than across industries, the
M-form dominates the U-form from the standpoint
of incentives
U-form Organisation
X1A
Top
Shock 
Output:
, X1B , X2A , X2B
Ministry 1
Shock 1
Output:
X1A , X 1B
Plant 1A
Shocks 1, A
Output:
X1A
Plant 1B
Shocks 1, B
Output:
X1B
Ministry 2
Shock 2
Output:
X2A , X2B
Plant 2A
Shocks 2, A
Output:
X2A
Plant 2B
Shocks 2, B
Output:
X2B
M-form Organisation
Top
Shock 
Output:
X1A , X2A, X1B , X2B
Region A
Shock A
Output:
X1A , X2A,
Plant 1A
Shocks 1, A
Output:
X1A
Plant 2A
Shocks 2, A
Output:
X2A,
Region B
Shock B
Output:
X1B , X2B
Plant 1B
Shocks 1, B
Output:
X1B
Plant 2B
Shocks 2, B
Output:
X2B
Top/Bottom Level Managers’
Incentives in M/U-form
• Top manager’s reward depends on:
– M-form: {X1A , X2A, X1B , X2B}
– U-form: {X1A, X1B , X2A , X2B }
• Plant 1A manager’s reward depends on:
– M-form: {X1A  X2A, X1B , X2B }
– U-form: {X1A  X1B , X2A , X2B }
• Plant ir manager’s reward depends on:
– {Xir  Xjr, Xjs , Xjs }for industries i,j regions r,s
Different Level Managers’
Incentives in M/U-form
Prop 1: Given any incentive scheme tη(x1A, x2A, x1B, x2B) for the top
manager in the M-form, there exists an equivalent scheme tη'(x1A, x2A,
x1B, x2B) for the top manager in the U-form and vice versa. Similarly,
given any incentive scheme tir for the industry i manager under the
region r manager in the M-form, there exists an equivalent scheme tri'
for the region r manager under the industry i manager in the U-form,
and vice versa.
• The same is not true of middle-level managers.
• The middle managers’ sphere of control are partitioned differently for
M-form and U-form
– In the M-form, the region A and B managers affect
respectively, whereas
– in the U-form, the industry 1 and 2 managers affect
respectively.
Analyse Middle Level Managers’ Incentives:
Information set comparison
• Two music CDs (CD1 and CD2) with the same
Beethoven Symphony number 9
• One CD is noisier than another
– Noisier CD is worse than the other
• How to measure:
– If add noise to CD1 (by a noise generator) can make
CD1 sounds the same as CD2
– CD1 is better than CD2
• The same principle can be used to compare
information set for organizations
Further simplified model
U top
X1
X2
xi = ei + εi, i = 1, 2
(ε1, ε2): joint normal distribtn
M top
XA
XB
xr = er + εr , r = A, B
(εA, εB): joint normal
Middle Managers’ Incentives
output in industry i:
xi = ei + εi, i = 1, 2
ei is the effort of the manager in charge of shock εi
output in region r :
xr = er + εr , r = A, B
er is the effort of the manager in charge of shock εr
• (ε1, ε2), (εA, εB) follow joint normal distribution
•
•
•
•
• σr2 = Var(εr), r=A,B; σAB = Cov(εA, εB);
• Conditional variation: Var(εA εB) = σA2 - (σAB)2/σB2
• σi2 = Var(εi), i=1,2; σ12 = Cov(ε1, ε2);
• Conditional variation: Var(ε1 ε2) = σ12 - (σ12)2/σ22
• Suppose Var(εA εB) < Var(ε1 ε2)
• Managers’ preferences:
U(t) - C(e)
Can M-form imitate U-form?
U top
X1
X2
M top
XA
XB
M use information xA , xB, z such that
xA - αxB + z = x1 in distribution
βxB+γ = x2 in distribution
Middle Managers’ Incentives (2)
• Fix manager 2’s & manager B’s effort as e'
• Set noise generator as
–
–
–
–
z  N[αe' , Var(ε1 ε2) - Var (εA εB)]
α = σAB/σB2 - σ12/(σ22σB2)½,
β = (σ22/σB2) ½,
γ = (1-β)e’
• Show: random variables (x1, x2) and
the same distribution:
• E(xA - αxB + z) = e - αe' + αe' = e = Ex1
• E(βxB + γ) = βe' + (1-β)e' = e'= Ex2
(xA - αxB + z, βxB+γ) have
Middle Managers’ Incentives (3)
• Var(βxB + γ) = β2 Var(xB) = σ22
• Cov(xA - αxB +z, βxB +γ)
= βσAB2 - αβσB2
= σ122
• Var(xA - αxB + z)
= σA2 - 2ασAB + α2σB2 + [Var(ε1 ε2) - Var (εA εB)]
= σ1 2
• (x1, x2) = (xA - αxB + z, βxB + γ) in distribution
• M-form can imitate U-form by adding noise z
• Manager A in M-form can have incentives at least as good
as manager 1 in U-form
Middle Managers’ Incentives (4)
Proposition: If both
min { Var (εAεB), Var (εBεA) } <
min { Var (ε1ε2), Var (ε2ε1) }
and
max { Var (εAεB), Var (εBεA) } <
max { Var (ε1ε2), Var (ε2ε1) },
then both managers A and B can be given better incentives
than manager 1 and 2.
Empirical Test
• Test the conditional variations
• Chinese data: 520 state-owned enterprise from 1986 to
1991
• Compare the Chinese organizational form (M-form)
with a hypothetical U-form.
– In this U-form, all firms would be organized into
hypothetical industrial ministries
• Compare conditional variances of regional and
industrial shocks under M-form and U-form
arrangements.
Data and methodology
• The data set contains industry classification codes (30
industries) and location codes (20 provinces) for each
firm.
• Use log-linear Cobb-Douglas production function to
estimate industry-specific shocks (θ) and regionspecific shocks (δ).
• For every industry i, region r, and period t, dummy
variables DitI and DrtR: coefficients as proxies for θ
and δ.
Regression model
• Where,
Estimate region and industry shocks
• Due to an identification problem, regional and industrial
shocks cannot be estimated directly.
• By dropping the dummy variables of one region and one
industry, and estimate the coefficients of the dummy
variables for the remaining regions and industries relative
industry-specific and relative region-specific shocks
are estimated
•
: estimated relative shocks by using the shocks in one
region and one industry as a benchmark
• Take region C and industry 3 as benchmarks.
– A time series of relative shocks is obtained from the regressions
– Treat the estimated time series as if they were real shocks
– Use the estimated time series to calculate conditional variations
Comparing Industrial and Regional Variance
and Conditional Variance (by Province)
Evidence: existence of yardstick competition
• Previous evidence suggest China can benefit from yardstick
competition
• Were yardstick competition actually used in China?
• Need evidence on promotions of regional government
officials based on relative performance evaluation.
• The Chinese political system is still under one-party rule
– The representation of a region in the Party Central Committee
indicates the status and power of the regional government officials.
• A province's representation in the Party's Central Committee
is used as a proxy for the promotion chances of officials in
that province.
– Normalize the representation by using the "per capita number of
Central Committee members" as an index
• Measure economic performance of a province by its growth
rate in "national income"
Relative performance index &
relative reward index
• To investigate how the change of relative ranking in economic
performance lead to the change of relative ranking in political power
• Region r’s relative economic performance index measures the change
in rank in economic performance between 1976 and 1986
• Region r’s relative political reward index measures the change in rank
in political position between 1977 and 1987
• The convex third terms in EINDEX and PINDEX: staying at the top
requires greater reward than staying at the bottom
• Evidence: Exist Regional Yardstick Competition
PINDEXr = - 0.453 + 1.76 EINDEXr, R2 = 0.671.
(0.246)
Further Evidence of Regional Competition
Economic Performance v Political Turnover
(Li and Zhou, ‘Political Turnover & Personal Control in China,’ JPubE, 2006)
• Turnover of top provincial leaders, y
– Turnover data in China between 1979 and 1995
• Hypothesis: The probability of promotion for provincial
leaders increases with the provincial economic
performance
– Prob(y) = f(y*)
• Suppose central government makes promotion/termination
decisions based on a performance score of provincial
leaders, y*
• Assume y* = xβ + ε,
– x: provincial GDP growth rate; ε : follow a normal distribution.
Variables affecting promotion
• But only turnover of a provincial leader is observable, y
= 0 for a termination, if y* ≤ a1
= 1 for remaining at the same level, if a1< y* ≤ a2
= 2 for a promotion, if y* > a2
– Cutoff point 1: for remaining at the same level (or termination)
– Cutoff point 2: for promotion (or no promotion)
• Promotion may be related to connections with the center.
The central connection variable:
= 1, for provincial leaders having previous experience or holding
joint-appointment in the central government
= 0, otherwise.
Economic Performance v Provincial Leaders’ Turnover
Dependent variable: Prob(turnover) (0=termination, 1=same level, 2=promotion)
Caveats
• The MQX theory implicitly assumes that the government has a well
defined and measurable objective, such as GDP growth rate or
attracting FDI
– Yardstick competition on that objective can be organized
• Yardstick competition may not work
– If the government faces multiple tasks and some tasks are not well defined; or
– If some tasks are more difficult to be measured than others
• The evidence on yardstick competition condition in China is limited
– No data on Russia
– Its data has only 6 years whereas a formal test requires a much longer time series
• The political economy evidence is preliminary
– Although the performance is lagged there still exist potential endogeneity
problem:
– Past political power of a regional government may have long lasting impacts
that it may affect future economic performance of the region