Transcript keefer_Database_of_political - Duke University`s Fuqua School

The Database of Political
Institutions and Empirical
Political Economy
A Description and an Application
to Banking Crises
Philip Keefer
Development Research Group
The World Bank
DPI2000
Approximately 100 variables, 150 countries, 1975-2000
including:
• Presidential or parliamentary?
• Defense minister a soldier?
• Are elections competitive?
• How many checks and balances?
• How many years a democracy/dictatorship?
• How many years have the veto players with longest/shortest
tenures been in office?
• Are the executive/coalition parties/opposition parties R-L-C?
• In what years were there pres/leg elections?
• Plurality/proportionality/district magnitude?
Other important political databases
• Henisz: POLCON – number of veto players
• Wallack, Gaviria, Panizza, Stein (IADB)
• Particularism database: electoral variables
• Persson and Tabellini: electoral/regime type variables
• Polity IV: subjective ratings of political systems
A comparative caveat
• Some of these databases have differences across
identical variables (district magnitude, regime
type, voting rule), despite overlapping sources.
• The differences matter: plurality election rules
REDUCE corruption using Persson and
Tabellini or IADB measures of voting rules.
They INCREASE corruption using DPI.
• Possible explanation: Rules for backwards
extrapolation for DPI may be stricter.
Why do we need crosscountry political databases?
Deficits
Political
institutions
Fiscal
allocations
Corruption
Trade reform
Regulation
Conflict
Studies that could use/have used
cross-country political data
Settler
mortality
Political
institutions
Insecure
property rights
Central bank
independence X
Political
institutions
Inflation
Clientelism
Low public goods
High corruption
Political
institutions
Years of continuous elections matter more
than income/capita for “governance”
0.6
Years of continuous elections
Income/capita
0.4
0.2
0
-0.2
Rule of law
-0.4
Bureaucratic
quality
Corruption
The competitiveness of elections
matters more for school enrollment
than education spending
Change in gross primary school enrollment
(percentage point change)
6
4
2
0
One standard deviation
One standard deviation
increase in education increase in competitive-ness
as a fraction of budget
of elections
Political institutions and
banking crises
Bank crises expose a perennial political
puzzle:
• Do checks and balances protect property
rights at the expense of good policy and the
timely response to crisis?
Checks and balances and
banking crisis
“Banking crises and the effect of political
checks and Balances” reviews evidence that:
• Bank crises are often the result of special
interest influence.
• Government intervention in bank crises is often
delayed by special interest influence.
And derives conditions under which
• Checks and balances make rent-seeking less
desirable for politicians.
Checks and balances and
banking crisis
The hypotheses:
• Rent-seeking falls in the number of veto
players.
• The effect of the number of veto players on
rent-seeking is lower the larger are the rents.
Checks and balances and
banking crisis
How to measure rent-seeking in banking context?
• fiscal costs of bailouts and
• lengthy forbearance after crisis is recognized.
How to measure rents?
• size of financial sector – e.g., M2/GDP.
How to measure checks and balances?
• checks from DPI
How to measure electoral pressures?
• years till next election from DPI
Results
Fiscal costs of crisis
(robust standard
errors)
-9.08
Checks
(2.8)
12.78
Checks * M2/GDP
(4.35)
Forbearance?
(marginal effects,
robust p-values)
-0.31
(0.02)
0.45
(0.05)
M2/GDP (year t – 1) -29.42
(10.38)
-1.08
(0.04)
Election year–t
1.37
(1.38)
0.17
(0.005)
R2
N
0.30
0.30
40
40