decentralization and political institutions in developing countries ekaterina zhuravskaya new...
TRANSCRIPT
Decentralization and Political Institutions in Developing
Countries
Ekaterina ZhuravskayaNew Economic School
ESNIE, May 19, 2008
Ekaterina Zhuravskaya ESNIE 2008 2 of 44
Motivation • Decentralization has been at the center of the
stage of policy experiments in the last three decades in a large number of developing and transition countries– Latin America, Africa, Asia, CE Europe
• On account of its many failures, centralized state everywhere has lost legitimacy
• The world bank, for example, has embraced decentralization as one of the major governance reforms on its agenda (World Bank 2000)
Ekaterina Zhuravskaya ESNIE 2008 3 of 44
Goal and RoadmapQuestions:
– “Is decentralization good or bad?” – “Does it result in more efficient governance, higher
economic growth, and/or better public goods?”• Virtue of decentralization is a very popular notion, but often this
notion is driven by a series of misconceptions – E.g., Dictators (Hitler, Stalin, etc) chose centralization and they are
vicious, but that does not necessarily imply any causality from centralization to outcomes
– Discuss: • Arguments for and against decentralization and (rather scares)
empirical evidence• Generalizations (if any) that one could make about “when” and
“for what”
– Background material: • Treisman (2007) “Architecture of Government” Cambridge UP• Bardhan (2002) JEP
Ekaterina Zhuravskaya ESNIE 2008 4 of 44
Defining terms
Lots of scholarly arguments about decentralization are characterized by different sides having different things in mind
One must define terms!
Types of decentralization:
1.Administrative decentralization
2.Political decentralization
1.Decision-making decentralization
2.Appointment decentralization
3.Constitutional decentralization
3.Fiscal decentralization
Ekaterina Zhuravskaya ESNIE 2008 5 of 44
Defining terms: Administrative decentralization
• At least one policy is implemented not by the central government directly but by locally based agents appointed by and subordinate to the central government
Ekaterina Zhuravskaya ESNIE 2008 6 of 44
Defining terms: Political decentralization
• Decision-making decentralization– At least one subnational tier of government has excusive authority
to make decisions on at least one policy issue
• Appointment decentralization– Government officials at one or more subnational tiers are selected
by local residents independent of higher level governments• via democratic elections or a non-democratic selection by local elites
– Federal state:a) Decision-making decentralization (e.g., China)b) Decision-making decentralization + Appointment
decentralization (at the same tier) (e.g., Argentina)
• Constitutional decentralization– Subnational governments (or their representatives) have a formal
right to participate in non-trivial way in central policy-making (e.g., US Senate)
Ekaterina Zhuravskaya ESNIE 2008 7 of 44
Defining terms: Fiscal decentralization
a) Theory: Decision-making decentralization on tax and/or expenditure issues
b) Empirics: Subnational governments account for a large share of total government revenues or spending
• Very little connection between the two
(Theoretical) reasons for and against various kids of
decentralization
Ekaterina Zhuravskaya ESNIE 2008 9 of 44
1st (potential) advantage: Administrative efficiency
“How could the general of an army be instantaneously obeyed by all its solders if the army were not divided into regiments, the regiments into companies, the companies into squadrons? But the effect of order is still more admirable in a state than in an army [because of higher territory and longer horizon]”
Charles Loyseau (1610)
Ekaterina Zhuravskaya ESNIE 2008 10 of 44
Administrative efficiency 1: optimal scale
• Technology of public goods provision varies – National defense vs. garbage collection (economies
of scale)
In a multi-tier structure, provision of different PG can be assigned to most efficient level
• Is political decentralization necessary?– If one concerns with just technical costs of PG provision political
decentralization is not needed, just administrative decentralization is sufficient
• Optimal scale argument implies, however, that all jurisdictions at a certain level should be the same
– which certainly is not true
Ekaterina Zhuravskaya ESNIE 2008 11 of 44
Administrative efficiency 2: Heterogeneous tastes and policy differentiation
• Another (potential) reason for decentralization – preference heterogeneity– Hayek (1945): local governments have more information about people’s
preferences and lower costs of acquiring it
• Alesina and Spolaore (2003) -- size of countries and degree of decentralization depends on:– a tradeoff b/w economies of scale (favors large size and central
provision) and heterogeneity (small size and decentralized provision)• Defense of ethnic groups against each other
• Is administrative decentralization sufficient?– Technically, the central government is able to give different orders to
its alderman in different regions, and collect information from them • USSR, Persian Empire, etc.
Ekaterina Zhuravskaya ESNIE 2008 12 of 44
Administrative efficiency 3: Agency costs - theory
• Do the agency costs multiply with added tiers?– If yes, above a certain number of tiers, it may be too
expensive to pay the subordinates enough so that they do not cheat
• then, administrative centralization may not work
– Persson and Tabellini (2000) present several models which say “yes”
– The answer, however, depends on incentive structures• How could hierarchy solve agency problems?
– Career concerns (Holmstrom 1999) – Yardstick competition (Maskin, Qian, Xu 2000)
» Chinese provincial leaders vs. Russian governors (1 vs. 120 criteria)
Ekaterina Zhuravskaya ESNIE 2008 13 of 44
Administrative efficiency: Agency costs - practice
• Large bureaucracies with many tiers have extremely bad reputation– However, they occur in very autocratic regimes (i.e.,
USSR, medieval Europe, imperial China)– It might be the case that bad performance is due to
problems with accountability of the central government rather than inefficiency of hierarchy
• Counter example to both – Chinese miracle
• But, it is also true that many bureaucratic hierarchies do accumulate agency costs
Ekaterina Zhuravskaya ESNIE 2008 14 of 44
Turning to political decentralization
• Are there advantages to letting local communities choose their own political leaders and policies? – whatever the technical efficiency of different
administrative structures
• What are the costs?
Ekaterina Zhuravskaya ESNIE 2008 15 of 44
2nd (potential) advantage: Competition among governments
“Competition forces governments to avoid all sorts of taxation which would drive capital or labor elsewhere”
Hayek (1939)
1. Does competition occur under political decentralization?
2. If yes, is it beneficial?• Consider separately: competition for residents
and competition for capital
Ekaterina Zhuravskaya ESNIE 2008 16 of 44
Competing for mobile residents• Tiebout (1956): under decentralization citizens sort
themselves into local communities with efficient taxation/PG mix– Local governments’ competition for citizens has three effects:
• (1) Reveals information about preferences; (2) induces efficient sporting; and (3) enhances accountability
– Restrictive assumptions:• Mobility (in many developing countries restricted)• No land (if real estate prices reflect efficiency of PG provision,
pressure on governments may be reduced)• Constant returns in PG provision (if there are economies of scale,
there may be multiple equilibria, including inefficient ones) • Externalities (consider below)
Ekaterina Zhuravskaya ESNIE 2008 17 of 44
Competing for mobile capital• Capital is more mobile than residents
– Qian&Roland (1998), Qian&Weingast (1998): Tiebout logic can be generalized to mobile capital (local governments reduce corruption and offer better business environment to attract capital)
• Main conceptual difference with Tiebout:– Preferences of capital owners are homogenous, i.e., best risk-
adjusted return
• This generates problems with this argument:– “Race to the bottom” in PG provision, particularly, if local
governments get private benefits of investment => distortions– Large differences in initial conditions (i.e., infrastructure) =>
backward regions give up (even with t=0, capital flows elsewhere)– Externalities (consider below)
Ekaterina Zhuravskaya ESNIE 2008 18 of 44
1st (potential) disadvantage: Externalities
In order to attract residents and capital local governments pursue policies with external effects:
• “Beggar your neighbor” policies – Attract firms with high pollution – Erect inter-jurisdictional trade barriers
» China: Young (2000), Poncet (2004) » Russia: Guriev, Yakovlev and Zhuravskaya (2008)
• “Beggar your superior” policies – Compete in protection from paying federal taxes and ignoring federal
regulation » Russia: Ponomareva and Zhuravskaya (2004); Sonin (2008)
“State-corroding” federalism (Cai and Treisman 2004)
Ekaterina Zhuravskaya ESNIE 2008 19 of 44
Externalities in fiscal policy
2 fiscal policy examples of “state-corroding federalism” (a la Cai and Treisman 2004):
1. “Common pool”: Local governments may consider central budget as the “Common pool”
• Each locality bears a fraction of the cost of spending from central budget, but enjoys the full benefit – Argentina (Saliehg & Tommasi 2001)
2. “Soft Budget Constraints”: local governments may blackmail the central government if they can commit to a policy undesirable to the center (but center cannot commit)
• Russian Far North (Tanzi 1996)• Bailouts under local insolvency
Ekaterina Zhuravskaya ESNIE 2008 20 of 44
3rd (potential) advantage: Policy experimentation
Justice Brendeis called decentralized governments “local laboratories for policy experiments”– Completely wrong argument:
• As long as central government can pursue different policies in different parts of the country– Risk-aversion
• Centralized government bares less electoral risk when experimenting in one locality
– Positive information externality• Local governments do not take into account that other local
governments benefit from learning from their failures when deciding whether to experiment – a version of free-rider problem
• Central authorities can design randomization experiments to maximize useful information from experimentation
– In reality, it actually is a disadvantage (!)
Ekaterina Zhuravskaya ESNIE 2008 21 of 44
4th (potential) advantage: Effect of bundling of policy on accountability
• Besley and Coate (2003):• When both local and central public officials are elected,
their respective policy areas are the only salient issues at their respective elections
• When local public officials are appointed, in the central elections, local policy area (e.g., local roads) becomes bundled with central policy issues (e.g., foreign policy) which may be more salient during the central elections
– Appointment centralization => capture of local policies by special interests
• No informational asymmetries necessary to generate this result
Ekaterina Zhuravskaya ESNIE 2008 22 of 44
5th (potential) advantage: Checks and balances
• Weingast (1995), Tsebelis (2002): Number of veto players in forming central governments’ policy increases with constitutional decentralization
• This increases stability of local policies– But it is not clear if stability is a good thing:
• decreases likelihood of changes both for the better and for the worse
Ekaterina Zhuravskaya ESNIE 2008 23 of 44
2nd (potential) disadvantage: Fiscal coordination
Berkowitz and Li (2000): Vertical competition– Overlapping tax bases for central and local
governments may result in “vertical overgrazing,” i.e., over-taxation, as each level of government tries to extract as much and as fast as possible
– Shared responsibility over some PG may result in free riding on each other’s contribution
• Not overlapping tax bases and expenditure responsibilities
– In practice, hard, different taxes often have effect on the same tax base
Ekaterina Zhuravskaya ESNIE 2008 24 of 44
3rd (potential) disadvantage: State capture
• Federalism often leads to local tyranny in developing countries – Bardhan and Mookherjee, 2000; Sonin, 2003; Slinko, Yakovlev,
and Zhuravskaya, 2005
• Bardhan and Mookherjee (2000) write down conditions under which local governments are more captured by special interests than central governments– Intuition: local capture is higher when local markets for influence
are more concentrated while the central is not• Russia – many cities are built around one plant
• India – local land inequality is very high
– At the central, dissipation of rent make lead to breakdown of the market for influence
Ekaterina Zhuravskaya ESNIE 2008 25 of 44
Ethnic conflict, secession, and decentralization
• Example: Iraq• “It is virtually impossible to make broad generalizations
about effectiveness of federalism in multinational societies” (Simeon and Conway 2001)– Political decentralization may help to:
• satisfy moderate demands for autonomy of geographically concentrated ethnic minorities
• reduce social tension of “winner takes all” by splitting the political prizes
• to resist discriminatory central policies– But it also may:
• enable local majorities to abuse local minorities and inflame local cleavages that are salient at the local and not central level
• empower local majorities to mobilize for succession • trigger inefficient redistribution under secession threat• increase “regionalist” mentality
– Stepan and Linz (2008):• “Coming-together” vs. “Holding-together” federalism
Ekaterina Zhuravskaya ESNIE 2008 26 of 44
Pitfalls of empirical research• Different measures should have different effects in theory
– Average size of a jurisdiction– Number of tiers– Number of jurisdictions– Mode of subnational appointments– Subnational expenditure and revenue share– Constitutional arrangements
• Very scares data on decision-making decentralization• Little panel data (more being collected)• Little over time variation for political decentralization• Most of the literature focused on the subnational revenues
and expenditures (very imperfect)
Ekaterina Zhuravskaya ESNIE 2008 27 of 44
Evidence before EZ2007 was inconclusive; Treisman (2007) things it is still the case
Results vary with samples and time periods
Effect of decentralization onGrowth:
(-): Zhang and Zou, 1998 (Chinese provinces); Davoodi and Zou, 1998 (developing countries)
(0): Woller and Phillips, 1998 (developing countries);(+): Jin et al., 1999 and Lin and Liu, 2000 (Chinese provinces);
Corruption:(-): Fisman and Gatti, 2002; de Mello and Barenstein, 2001 (X-
country)(0): Treisman, 2000 (X-country)
These papers overlooked importance of political institutions
Ekaterina Zhuravskaya ESNIE 2008 28 of 44
William Riker (1964) “Federalism: Origins, Operation, and Significance”
• To sum up: – both centralization and decentralization have costs and
benefits:• Information, inter-jurisdictional competition, less policy bundling
vs. externalities and local capture
• Riker: One could get the best of both worlds by providing local politicians with national career concerns, but preserving political decentralization– Strong national political parties
• Reduce both “local capture” and “externalities” • But leave incentives to cater well to the needs of the citizens
Ekaterina Zhuravskaya ESNIE 2008 29 of 44
Appointment centralization vs. decentralization
Why strong national political parties help?– Local officials need financial and political
support of national parties during elections– Career concerns
Why administrative centralization will not?• Riker’s view: administrative centralization
undermines the benefits of decentralization in the first place – Politicians will stop caring for the preferences of local
population and only care for pleasing superior» All would depend on incentives of the superior and
superior’s informational disadvantage
Ekaterina Zhuravskaya ESNIE 2008 30 of 44
Russia vs. ChinaBlanchard and Shleifer (2001):• Fiscal decentralization occurred in both countries during
transition• Decentralization was a major growth-promoting factor in
China and an obstacle to growth in Russia– Jin, Qian, Weingast (2005) on China– Zhuravskaya (2000) on Russia
• The reason is the difference in the strength of national political parties and administrative centralization (in different periods):– China - tight administrative and political control of the communist
party– Yeltsin’s Russia - large-scale political decentralization and no
national parties
Ekaterina Zhuravskaya ESNIE 2008 31 of 44
Argentina vs. ChileEnikolopov and Zhuravskaya (2007): Both countries undertook
fiscal decentralization in the 80s and 90s– Chile - improved provision of public health and education– Argentina - macroeconomic destabilization and a large-scale economic
crisis
• Chile - strong party system with national parties and strong career concerns– Mayors of Santiago
• Argentina - national political parties are weak; provincial parties dominate political arena at the national and provincial level– Career concerns in Argentina work the other way:
• Those national politicians who managed to produce more benefits to their provincial constituencies (possibly at expense of other regions) can return back home after serving in the national government
Ekaterina Zhuravskaya ESNIE 2008 32 of 44
Enikolopov & Zhuravskaya (2007): Measures of Political institutions
1. Age of main parties– Intuition: career concerns– Control for: the age of democracy and time since
independence2. Fractionalization of governing parties
– Intuition: career concerns– Control for:
• majoritarian vs. proportional electoral rule• presidential vs. parliamentary regime
– Persson & Tabellini 2003 • secessionist tendencies
– segregation of voting patterns at the local level– presence of contiguous autonomous regions
3. Administrative centralization– Are municipal executives appointed? – Are province-level executives appointed?
Ekaterina Zhuravskaya ESNIE 2008 33 of 44
Measures of decentralization and outcomes
Fiscal Decentralization • Subnational revenue share• Subnational exp. share
Quality of government • Corruption indices • Control over corruption• Government effectiveness• Regulatory quality • Rule of law
Growth• GDP per capita growth rate,
PPP
Public goods provision• Infant mortality • DPT Immunization • Illiteracy rate• Pupil to teacher ratio
Ekaterina Zhuravskaya ESNIE 2008 34 of 44
Empirical Methodology• Standard methodology for growth and the quality of governance
regressions– Treisman (2000); La Porta et al. (1999); Barro and Sala-i-Martin (1995)
• Add fiscal decentralization measure and political institutions measures and their interaction:
• How results of fiscal decentralization are affected by political institutions?
• Sample of developing and transition countries (73 countries)– In OECD, our measures poorly reflect career concerns
• Use geographical area of countries as an instrument for decentralization
• Initial levels of political institutions as an instrument for current political institutions (corrects only measurement error)
Ekaterina Zhuravskaya ESNIE 2008 35 of 44
Standard control variables
• Fixed investments • Population • Openness • Fertility• Current level of democracy • Democratic tradition • Ethno-linguistic fractionalization• Protestants share • Latitude • Legal origin
Ekaterina Zhuravskaya ESNIE 2008 36 of 44
ALB
ARG
ARM
AZE
BLR
BOL
BRA
BGR
CHLCHNCRI
HRV
CZE
DOM
EST
GEO
HUNIND
IDN
ISR
KAZ
LVA
MYS
MUSMEX
MDA
MNG
NIC
PAKPER
POL
ROM
RUS
SVK SVN
ZAF
THATTO
UKR
-3-2
-10
1R
esid
ua
ls o
f R
ule
of
La
w in
de
x
0 10 20 30government parties fractionalization * Subnational revenues share
ALB
ARG
ARM
AZE
BLR
BOL
BRA
BGR
CHN
CRI
HRV
CZEDOM
EST
GEO
HUN
IND
IDNISR
KAZ
LVA
MYS
MUS
MEX
MDA
MNG
NIC
PAK
PER
POL
ROMRUS
SVK
SVN
ZAF
TJK
THA
TTO
UKR
-10
12
3R
esid
ua
ls o
f R
ule
of
La
w in
de
x
0 50 100 150 200Logarithm of age of the main parties * Subnational revenues share
Gov. fractionalization measure
Party age measure
Residual partial plots – party strength
80% of the developing countries have parties younger than needed for decentralization to have a positive effect on indices of government quality
Ekaterina Zhuravskaya ESNIE 2008 37 of 44
Within country relationship
ARG
ARG
ARG
ARG
AZEAZE BOL
BOL
BRA
BRABRABRABRA
BRA
BRA
BRA
BRA
BRABRA
BRABRA
BGRBGR
BGRBGR
CHNCHNCOL
COL COLCOL
COL
COL
COL
CRI
CRI
CRI
CRICRI
CRI
CRI
CRI
CRI
CRICRI
CRI
HRV
HRV
CZE
CZE
DOM
DOM
DOM
EST
FJIFJIFJI FJI
FJI
FJI
FJI GMBGMBGMBGTM
HUN
HUN
HUN
IND
IND
IND
INDIND
IND
IND
IND
INDIND
IDNIDN
IDN
IDN
IDN
IDN
IDNIDN
IDNIDN
IDNIDN
IDN
IDN
IDNIDN
IDN
IRN
IRN
IRN
IRN
IRNIRN
ISR
ISR
ISR
ISRISR
ISRISR
ISR
ISR
ISRISR
ISR
ISR
ISR
KEN
KENKEN
KEN
KEN
KEN
KENKEN
KENKENKEN
KEN
KEN
LVA
LVALTULTU
LTU
MDG
MDG
MDG
MWI
MWI
MWI
MWI
MWI
MYS MYSMYSMYSMYS
MYS
MYSMYS
MYSMYS
MYSMYS
MYS
MYS
MUSMUS
MUSMUS
MUS
MUS
MUS
MUSMUS
MEXMEX
MEXMEX
MEX
MEX
MEX
MEX
MEXMEX
MEXMEX
MEXMEX
MEX
MEX
MEX
MDA
MNG
MNG
MNG
NIC
NIC
NIC
NIC
NIC
PAN
PAN
PANPAN
PNGPNGPER
PERPHLPHLPHLPHL
PHLPHLPHL
PHL
PHLPHL
PHL
PHL
POL
ROM
ROM
ROMROM
ROMROM
SVN
SVN
SVN SVN
ZAF
ZAF
ZAF
LKA
LKALKA
LKA LKATHA
TTOTTO
TTO
TUNZMB
ZWE
ZWEZWEZWE
ZWE
ZWEZWE
ZWEZWE
ZWE
-.2
-.1
0.1
.2R
esid
ual
s o
f ln
(Pup
il to
Te
ach
er R
atio
)
-.2 -.1 0 .1 .2Age of main parties * Subnational Revenues
• Sufficient data only for PG, not QG or growth• All outliers work against the results• Immunization and pupil to teacher ratio are significant;
literacy and infant mortality are not. – Short term vs. long term effect
ARG
ARG
ARG
ARGBLR BLR
BOL
BOL
BRA
BRABRABRABRA BRA
BRA
BRA
BRA
BRABRA
BRA
BRA
BGRBGR
BGR
BGR
CHLCHL
CHL
CHL
CHNCHNCOLCOLCOLCOL
COL
COL
COL
CRI
CRI
CRI
CRI
CRI
CRICRI
CRICRI
CRI
CRI
CRI
CRI
CRICRI
CRI
HRV
HRV
CZE
CZE
DOM
DOM
DOM
ESTFJI
FJIFJIFJI
FJI
FJI
FJIGMB
GMBGMBGTM
GTM
GTMGTM
GTM
HUN
HUN
HUN
IND
IND
IND
INDIND
IND
IND
INDIND
IND
IDNIDN
IDN
IDN
IDN
IDN
IDNIDNIDN
IDN
IDN
IDNIDN
IDN
IDN
IDNIDN
IRN
IRN
IRN
IRN
IRNIRN
IRN
ISR
ISR
ISR ISRISR
ISR
ISR
ISRISR
ISR
ISR
ISR
ISR
ISR
KEN
KENKEN
KEN
KEN
KENKEN
KEN
KENKENKEN
KEN
KEN
LVA
LVALTULTU
LTU
MDG
MDG
MDG
MWI
MWI
MWI
MWI
MWI
MYSMYS
MYSMYSMYS
MYS
MYS
MYS
MYS
MYSMYSMYS
MYS
MYS
MUSMUS
MUS
MUS
MUS
MUSMUS
MUS
MUS
MUS
MUSMUS
MEX
MEX
MEXMEX
MEXMEX
MEX
MEXMEX
MEXMEXMEX
MEX
MEX
MEX
MEXMEX
MDA
MNGMNG
MNG
NIC
NIC
NICNIC
PAN
PAN
PANPANPAN
PRY
PRY
PRY
PRY
PRYPRYPRYPRY
PRY
PRY
PER
PER
PHL
PHL
PHLPHLPHL
PHLPHL
PHL
PHLPHL
PHL
PHL
POL ROM
ROM
ROM
ROM
ROMROM
SVNSVN
SVN
SVN
ZAF
ZAF
ZAF
LKA
LKALKALKA
LKA
THA
THA
TTOTTO
TTO
TUN
URY
URY
URY
ZMB
ZWE
ZWEZWEZWE
ZWE
ZWE
ZWE
ZWEZWE
ZWE
-.2
-.1
0.1
.2R
esid
ual
s o
f ln
(Pup
il to
Te
ach
er R
atio
)
-4 -2 0 2Government parties fractionalization * Subnational Revenues
Ekaterina Zhuravskaya ESNIE 2008 38 of 44
Residual partial plots – state-level appointment decentralization
ALB
ARGBRA
HRV
KORMEX
SVK
ZAF
URY
-1-.
50
.51
Re
sid
ua
ls o
f G
ove
rnm
en
t E
ffe
ctive
ne
ss in
de
x
10 20 30 40 50 60Subnational expenditures share
State executives elected
ARMBHR
BLR
BOLBGR
CHL
CHN
CRICZE
DOM
EST GEOHUN
IND
IDNISR
KAZ
KGZ
MYS
MDA
MNG
NIC
PAK
POL
ROM
TJK
THAUKR
-.5
0.5
11
.52
Re
sid
ua
ls o
f G
ove
rnm
en
t E
ffe
ctiv
en
ess
ind
ex
0 20 40 60Subnational expenditures share
State executives appointed
• Appointment centralization also seem to help fiscal decentralization
• But, not robust to controlling for FE!
• Thus, must be driven by unobserved X-country heterogeneity
Ekaterina Zhuravskaya ESNIE 2008 39 of 44
Results - Solid support Riker’s theory
1. Strong parties (low government fractionalization and high party age) make decentralization more efficient: improve quality of government, growth, public goods
2. No robust evidence for the effect of administrative centralization
– Results depend on the set of controls and particular samples
TI
- C
or
Go
v_E
ffec
t
Reg
_Qu
alit
y
Co
nt_
Co
r
Ru
le_L
aw
GD
P_g
rwth
Imm
un
Neg
_In
f_M
ort
Neg
_Ill
iter
Neg
_P_t
o_T
party_age* rev_dec 0 +* +* +** +** +*** +* +*** + +
gov_frac* rev_dec - -*** -*** -** -*** -*** -*** -*** -** -***
Ekaterina Zhuravskaya ESNIE 2008 40 of 44
Economic significance - example
A 10% in decentralization:
• at a level of party age lower than the mean by 1/2 of its SD in government quality indices of 1/2 of their SDs
• at a level of party age higher than the mean by 1/2 of its SD zero change on government quality
Ekaterina Zhuravskaya ESNIE 2008 41 of 44
Sensitivity analysis• Influential observations
– China is influential; without China results go through
• Controlled for other possible driving forces for results– Transition
• Additional controls– Initial GDP per capita squared, federation dummy,
regional dummies, colonial origin, etc.
• Results are stable
Ekaterina Zhuravskaya ESNIE 2008 42 of 44
Endogeneity of political institutionsEconomic performance can have different effect on political institutions in fiscally centralized and decentralized states1. Government fractionalization:
• In centralized countries, better performance leads to relative strengthening of the national governing parties (and lower fractionalization because the success is attributed to national policies
• In decentralized countries, voters attribute success to regional policies which leads to an increase in fractionalization due to strengthening of local political organizations
2. Administrative centralization: • In centralized countries, good performance may allow the central
authorities to get public support to switch from elected to appointed subnational governments
• With decentralization, this may be harder OLS - an upward bias in the coefficient of the interaction term– Story for party age?
Ekaterina Zhuravskaya ESNIE 2008 43 of 44
Conclusions from E&Z’s test of Riker• Key finding: political institutions play an important role in
determining the results of fiscal decentralization
– Provided we buy causality
• Solid support for Riker:
– strong national party system is an effective way of securing political accountability needed for efficient fiscal decentralization, whereas administrative control is not
– If one does not buy causality, an interesting empirical association: countries that make decentralization work also manage to have stronger national political parties
• Important finding: for most of developing countries Riker’s conditions for successful federations are not met
Ekaterina Zhuravskaya ESNIE 2008 44 of 44
Open questions• What makes political parties stronger (positive)? • How to strengthen political parties (normative)?• (Related): How can one explain Chinese miracle?
– The evidence is that administrative centralization does not guarantee good outcomes
– Why Chinese central officials seem to be growth maximizing?
– Why there is no adverse effect of absence of local electoral incentives?
• Low level of development – No need to think about social safety and public good (just yet)
• Enlightened leaders• Local capture vs. Central entrenchment
Ekaterina Zhuravskaya ESNIE 2008 45 of 44
To take away:• Very hot topic for research• A lot more data are becoming available
now (on political institutions and on decentralization)– Including time series– There is surprisingly little good empirics done
of the subject• Both at country-level and • At within country level
– See Guriev, Yakovlev, Zhuravskaya for an example» The paper (in some sense) indigenizes Riker