Democratic Transitions

David Epstein, Robert Bates, Jack Goldstone,

Ida Kristensen, Sharyn OHalloran
In the News
Building democracy is getting new
levels of scrutiny in the popular press
Afghanistan
Built off existing institutions (loya jirga)
Still, no real national government (warlords)
Iraq
One justification for war was to establish a
democratic government in the Middle East
But a long interim govt may be needed first
Big Questions
How do you establish democracy?
What keeps democratic governments
from sliding into autocracy/chaos?
Home-grown vs. Imported democracy
How do you get non-violent transitions?
If democracy=elections, is this even a
good thing for developing polities?
Zakaria Illiberal Democracies
Institutions vs. institutions
We have certain outcomes that we want
a political system to deliver
Policy that reflects public opinion
Public officials responsive to their constituents
Protection for minorities, esp. ethnic
Rule of Law via a fair, efficient court system
Uncorrupt bureaucracy
These are Institutions
Really, objectives of the system
Institutions vs. institutions
We know little about what institutions
can provide these in different settings
Parliamentary vs. Presidential systems
Federalism and/or corporatism
Electoral rules and campaign finance laws
Like economic institutions in China
We have some examples of what fully
developed, stable democracies look like
But how to reach this destination?
This Paper
Try to look at how countries have
democratized in the past
Review existing theories
Present some new data and methods
Look at the results and their implications
Speculate about how these trends are
changing in the 21st Century
What, if anything, can the international
community do to help?
Quick Peek at the Results
Results
Modernization theory holds up well
Positive income effects at all levels of
democratization
Also some support for inequality effects
Casts spotlight on partial democracies
Numerous
Most volatile
Hardest to predict/understand
 Modernization Theory
Lipset (1959) Money matters
Higher income per capita is associated with
democratic governance
Supported by evidence until
Przeworski, et. al. (PACL, 2000)
Positive results were due to over-
aggregation of data
Separate into democracies and all others
Nothing predicts transitions up
Once up, per capita income of over $4,000
helps keep things stable
 Other Theories
Acemoglu & Robinson (2003)
Transitions are economically based, as they
redistribute income from rich to poor
Nonlinear impact of inequality on transitions
Rosendorff (2001)
Inequality as well, plus increases in workforce and
capital stock make revolutions less likely
Haggard & Kaufman (1995)
Inflation & low growth: transitions out of autocracy
Schmitter, ODonnell Consolidation
Factors promoting transition could make it fragile
 Data: Partial Democracies
PACL employ a dichotomous regime type
Democracy if:
Chief executive elected
Legislature exists and is elected
More than one political party
At least one successful transition of power
Autocracy otherwise
We employ a trichotomous measure,
including partial democracy
 Distribution of Polity Scores

.2
Fraction

.1

0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Polity Score
 Distribution of Polity Scores

.2

Autocracy
Fraction

.1

0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Polity Score
 Distribution of Polity Scores

Full
.2
Democ.

Autocracy
Fraction

.1

0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Polity Score
 Distribution of Polity Scores

Full
.2
Democ.
Partial
Democ.
Autocracy
Fraction

.1

0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
Polity Score
 World Democratization Trends, 1955-2000

100%

90%

80%

70%

60%

50%

40%

30%
Democracy
20% Partial
10% Autocracy

0%
1955

1957

1959

1961

1963

1965

1967

1969

1971

1973

1975

1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999
Year
 Regional Democratization Trends, 1955-2000

Africa Asia

Australia Europe

North & Central America South America

 Transition Frequencies
1-Year

A P D

A 97.47 1.89 0.63

P 5.12 91.97 2.90

D
1.00 0.87 98.13
 Transition Frequencies
1-Year 5-Year

A P D A P D

A 97.47 1.89 0.63 A 89.64 7.13 3.24

P 5.12 91.97 2.90 P 20.43 65.76 13.81

D D 3.84 4.13 92.03

1.00 0.87 98.13
 Transition Frequencies by Type and Year
8 8

6
66 obs. 6
22 obs.
Autocracy
4 4

2 2

0 0
1956 1960 1970 1980 1990 2000 1956 1960 1970 1980 1990 2000
year year

AP AD
8 8

6
44 obs. 6
25 obs.
Partial Dem.
4 4

2 2

0 0
1956 1960 1970 1980 1990 2000 1956 1960 1970 1980 1990 2000

PA
year year

PD
8 8

6
16 obs. 6
14 obs.
Democracy
4 4

2 2

0 0
1956 1960 1970 1980 1990 2000 1956 1960 1970 1980 1990 2000
year year

DA DP
 Data: Previous Transitions
Previous attempts at transition may
affect success of current attempt
Is it like:
Trying to quite smoking (+); or
A sign of structural instability (-)
Hard to capture this systematically
We use Sumdown, the sum of previous
downward Polity movements
Example: Turkey, 1955-2000
 Illustration of Sumdown Variable for Turkey, 1955-2000

Polity Score Sumdown

30
26 27
25
20
Polity Scale

11
9
10 8
4 9 9 8
7 7

0 1
0
-2
-5
-10
1955 1965 1975 1985 1995 2000
Year
 Data: Inequality
Plays a big role in many stories,
including Acemoglu and Robinson
Previous measures are rather incomplete
Even Denninger and Squire agree with this
We create a new measure
Idea: infant mortality is caused by both
health conditions & inequality
So regress IM on health correlates and take
the residual as a measure of inequality
Infant Mortality Residual vs. Gini Coefficient
 Econometric Approaches
Tobit
Traditional regression approach, using full data range
But with limits at 10 and 10
Markov
Breaks data into three categories: autocracy, partial
democracy, full democracy
Looks at impact of variables on transition probabilities
Duration Analysis
Analyzes length of time until a given transition occurs
Predictions
Modernization: GDP/capita (+)
Acemoglu & Robinson: Inequality (-),
Inequality2 (-)
Rosendorff: Inequality (-),
Workforce (-), CapitalStock (-)
Haggard & Kaufmann: GDP growth (+),
inflation(-)
Consolidation: non-monotone hazard
 Markov Analysis
Autocracy Partial Democracy
1

1
GDP/cap
.5

.5

.5
0

0
4.338934 10.69568 4.597264 9.659036 5.209844 10.73535
1

1
Inequality
.5

.5

.5
0

0
-1.45599 1.74149 -1.641473 1.25747 -1.379472 1.452754
1

1
Previous
.5

.5

.5
Attempts
0

0
0 43 0 43 0 27
 Duration Analysis
Question: What prevents newly
democratized countries from backsliding?
Different from the question of what gets them
democratic in the first place
Can ask this for full or partial democracies
I.e., most newly democratized nations will
start off as partial democracies
Tricky estimation, since we have repeated
events (can democratize more than once)
 1.00
0.80
0 1
Survival Probability
0.60
0.40
0.20

0 50 0 50
analysis time
Observed: highgdp = 0 Observed: highgdp = 1
Predicted: highgdp = 0 Predicted: highgdp = 1
Graphs by highgdp
 Smoothed Hazard Estimate, Adjusted for Per Capita GDP
.6
.5
.4
.3
.2

10 20 30 40 50
analysis time
 Conclusion
Modernization theory holds up well
With 3-way dependent variable, income level
does predict transitions out of autocracy
Helps keep partials from backsliding
Acemoglu and Robinson get some support
Focus debate on partial democracies
Increasingly common
Most volatile
Hardest to predict
 Markov Analysis

Autocracy
Probability

Partial Dem.
Full Democracy

Variable