30/07/2013

Electoral Systems and the Timing of

‘Post-Conflict’ Elections

Paul MItchell

•  Part 1: Overview of the main families

of Electoral Systems.

•  Part 2: The Timing of ‘Post-Conflict’ Elections:

Are Early Elections a Good Idea?

1
 30/07/2013

Three Main Categories

•  1. Plurality / Majority Systems

•  2. Semi-Proportional Systems

•  3. Proportional (PR) Systems

Overview of Types and Trends

•  1.Plurality – Majority Systems

•  Sub-types:

•  (A) Single-Member-Plurality (SMP) often badly labelled FPTP.

–  As of 2006: 50 countries use this – representing 26% of total (world)
•  Including – Bangladesh, Canada, India, PNG, Sudan, UK, USA,

•  (B) Two-Round System (2RS) or ‘the French system’

–  21 countries (11%)
–  Bahrain, Belarus, Egypt, France, Mali, Togo

•  (C) Block Vote (SMP in multi-member districts – votes = M).

–  8 countries (4%), including Kuwait, Palestine, Samoa, Singapore.

•  (D) Alternative Vote (AV)

–  3 countries – Australia, Fiji, Nauru
–  In 2011 the UK rejected this system in a referendum. 4

2
 30/07/2013

2. Semi-Proportional

•  Sub-types:

•  (A) Mixed-Member- Parallel (PR Lists and SMDs with no compensation)

–  21 Countries (11%)
–  Albania (until 2008), Cameroon, Chad, Georgia, Hungary, Japan, Korea,
Mexico, Russia, Tunisia.

•  (B) Single Non-Transferable Vote (SNTV) – ‘the old Japanese system’.

–  Vote for a single candidate in multi-member districts.
–  3 countries: Jordan and Vanuatu,
–  and from 2005 in Afghanistan. Hamid Karzai thought it would benefit his
ethnic group – it didn’t!

3. Proportional

•  (A) List Proportional Representation (List-PR)

–  65 countries (34%)
•  Angola, Argentina, Brazil, Finland, Israel, Netherlands, Slovenia, South Africa.

•  (B) Mixed-Member Proportional (MMP) – ‘the German system’

–  6 countries – Germany, Italy (sort of !, for a while), Lesotho, New Zealand,
Venezuela, Zanzibar.

•  (C) Single Transferable Vote (STV)

–  Among political scientists probably the most popular system
–  But only Ireland and Malta use it (for the lower house that is).

3
 30/07/2013

Dimensions on which electoral systems vary

1. District magnitude One Single-member plurality: (Canada,
India, UK, USA);
(number of seats per
constituency = M) Alternative vote: (Australia);
Two-round system: (France)

More than one PR-list systems: (Israel, Spain, South

Africa, Chile, Denmark);
Mixed systems: (Germany, Japan,
New Zealand);
PR-STV: (Ireland)

•  This is one of the most important factors and has a big effect on proportionality.

•  Most obviously any system that only has SMDs (and one level of allocation)
cannot by definition be a PR system – since single seats cannot be divided up
proportinally.

•  Generally speaking (and other things being equal!) the higher the district
magnitude (M) the greater the proportionality. 7

Measuring M

•  Measuring average district magnitude is straightforward in countries

where all constituencies are of the same size: SMD systems such as
Australia, France, India, the UK and the US,
•  Or in those few other countries where all the constituencies are multi-
member and of uniform size such as Chile (2) and Malta (5).
•  In a few countries there is only one (national) constituency, so the
number of seats is the district magnitude in the Netherlands (150) and
Israel (120).
•  In some other countries although M varies, we can easily work out an
average value.
–  For example, in Spain 350 members of parliament (MPs) are returned from
52 constituencies, so average district magnitude equals 6.7, while in Ireland
there are 42 constituencies and 166 MPs and district magnitude is 4.0.

•  Things become a bit trickier when there is more than one ‘tier’ of seat
allocation (for details see PES Appendix C)
8

4
 30/07/2013

2. How many votes?

How many votes can a 2 Mixed systems (Germany, Hungary, Italy,
voter cast? Japan, New Zealand, Russia- until 2005)

1 All other systems

•  This might seem like a strange question: Since ‘one person one vote’ is a
hallmark of a democratic system, why would we encounter any variation here?

•  The reason is simple: giving people more than one vote does not violate
democratic principles provided everyone still has the same number of votes!

•  Having just one vote is very much the norm.

•  The exceptions are the ‘mixed systems’

–  For example, when voters in Germany or New Zealand go to the polling station they
are confronted with a ballot paper that invites them to cast one vote for a candidate to
represent their local single-member constituency, and another vote for a party in the
contest for seats awarded at national (or regional) level.
10

5
 30/07/2013

New Zealand (MMP)

11

3. Ballot Structure
Categorical Single-member plurality (Canada, India, UK, USA);
(also termed nominal or integral) two-round system (France);
Virtually all PR-list systems
Dividual: can ‘divide’ vote among Mixed systems (Germany, Hungary, New Zealand)
different parties
PR-list with panachage (Luxembourg, Switzerland)

Ordinal: can rank order Alternative vote (Australia);

candidates PR-STV (Ireland)

•  Categorical systems ‘channel each parcel of electoral strength into the grasp of a single
party . . .’ (Rae 1971: 18).

•  Dividual systems allows voters to divide their support among more than one party (most
obviously in mixed systems. But also possible by panachage – a few list systems in
which it is possible to cast preference votes for more than one party).

•  Ordinal voting permits voters to rank order the candidates on the ballot paper.
•  This is a central feature of both the alternative vote and PR-STV.

•  Voters may rank all (or, at least, as many as they wish, depending on the specific
12
electoral laws) in order of their choice.

6
 30/07/2013

4. How much Choice of Candidates?

No choice of candidate within party
Choice of candidate within party
Choice of candidate without regard to
party
Single-member constituency systems
(Australia, Canada, France, India)
Mixed systems (Germany, Hungary etc);
Closed-list PR systems (Israel, South
Africa, Spain)
Preferential-list PR systems (Austria,
Chile, Denmark, Finland. etc)
PR-STV (Ireland)

•  Most obviously systems that only have SMDs do not offer any choice within
parties.

•  Mixed Systems? In principle they could offer a choice within parties. For e.g.
SMP election in single districts could be combined with open-list PR in MMDs
that would offer candidate choice. (or AV could be combined with open lists).

•  But empirically it has (almost) always been the case that mixed systems have
precluded candidate choice by using closed lists. (Lithuania is the only
exception) 13

How Open are the Lists?

•  Some PR-list systems - broadly termed preferential list systems -

enable the voter to indicate a preference for one candidate (or
sometimes several candidates) on their party’s list, and these
preference votes then play a role in determining which candidates fill
the seats that the party receives.

•  But some systems are more open than others.

•  We can distinguish between:
–  Fully open lists, where the voters alone determine which candidates receive
the seats (for example if Party A wins 3 seats then these go to the 3
candidates of party A that had the highest number of preference votes).
Chile and Finland are examples.

–  Flexible lists, where the party’s initial ordering of the candidates determines
the outcome unless sufficient numbers of voters combine to overturn this.
(Belgium and the Netherlands are examples)

14

7
 30/07/2013

15

Closed Lists

•  PR list systems may have ‘closed lists’, in which the voter can choose among
parties but not among candidates within parties, and the order of candidates’
names that is decided by the party determines which of them receive its seats.

•  Underlying the choice are two different views about representation:

–  According to one concept, the purpose of elections is to enable the direct

representation of the people, and consequently preferential list systems, allowing the
people to choose their own representatives are more appropriate.

–  According to the other, representation takes place through the political parties and
the purpose of elections is to enable the parties to secure their proper share of
representation; consequently, closed lists are more appropriate than open ones
because the parties’ candidate selectors are better judges than the voters of who is
best able to realize the ideas and goals of the parties.

–  In ‘principal–agent’ terms, MPs are the agents.

–  Closed list systems seem to assume that parties are the sole principals.
–  Open list systems assume that MPs have two principals, parties and voters.

16

8
 30/07/2013

17

Choice of Candidates without regard to party

•  STV gives voters a choice not only among their party’s candidates but
also a choice across party lines

•  Voters are not constrained by party lines when deciding how to rank
order the names of all the candidates on the ballot paper.

18

9
 30/07/2013

19

5. How many Levels of Seat Allocation?

1 Single-member plurality; alternative vote ; two-round system
2 (higher tier and lower
tier)
2 (both allocations are
at the same level)
3 (lowest, middle and
highest levels)
some PR-list systems (Belgium, Chile, Finland, Israel,
Netherlands, Spain);
PR-STV (Ireland)
Compensatory mixed systems – MMP.
Some PR List systems (e.g Denmark, S. Africa)
Parallel Mixed Systems –MMM. (Japan, Russia until 2005)
Some Mixed systems (Hungary)
Some PR-List systems (Anstria)

•  In many countries there is only one level of seat allocation.

•  Each voter casts a vote in a constituency; seats in that constituency are awarded, in
accordance with the rules, to parties (and candidates); and each party’s national total of
seats is simply the sum of the seats it won in each of the constituencies.

•  There is, by definition, only one level of seat allocation in single-member constituency
systems such as Australia, Canada, France, India, the UK and the USA.
•  There is also just one level in some PR systems.
20

10
 30/07/2013

Holding Back the Tiers . . .

•  Why, then, complicate matters by having more than one level or ‘tier’ of seat
allocation?
•  There are various reasons for doing this - the most common of which is that it
gets round the problem caused by one of the most robust findings in electoral
systems research, namely that the smaller the average district magnitude, the
greater the disproportionality.
–  Small M usually = high Disprop.

•  This correlation is unfortunate, because it points to a trade-off between two

desirable properties of electoral systems: namely (1) proportionality and (2)
providing a local constituency representative for voters.

•  With just one tier, the two extremes are:

–  a single-member constituency system, which scores well on the local representation
dimension but poorly on proportionality, and
–  a PR system with just one constituency covering the whole country (as in Israel and
the Netherlands), which gives excellent proportionality but no direct representation for
localities.

•  With only one level of seat allocation, we are forced to sacrifice a bit of one
desirable property in order to get more of the other.
21

Benefits of Higher Tier Allocations

•  Having more than one level means that we might be able to have our
cake and eat it.

•  Archetypal mixed systems, such as that in Germany, illustrate the point.

–  Here, half of the MPs are elected from single-member constituencies, while
the other half are elected from party lists.

–  The list seats are awarded to parties in such a way as to ensure that the
total number of seats received by each party is proportional to its share of
the list votes.[

•  Hence, the system delivers a high degree of overall proportionality,

while at the same time each voter has a local constituency MP.

•  Mixed systems have thus been described as ‘the best of both

worlds’ (Shugart and Wattenberg 2003c: 595).
22

11
 30/07/2013

Higher Tier Compensation?

•  The higher tier is conventionally termed compensatory or corrective,

because the seats awarded at the higher tier(s) are used to
compensate the parties that were under-represented at the lower level
and to correct disproportionalities that arose there.

–  In Germany, for example, the smaller parties such as the Greens and the
FDP win few if any of the single-member seats (because the effective
threshold is high in under SMP) and are brought up to their ‘fair’ overall
share by being given the appropriate number of list seats.

•  In other cases, the two ‘tiers’ are parallel;

•  Each is on the same level and neither can be seen as higher or lower.
–  In Japan and Russia (until 2005), for example, voters have two votes just as
in Germany, but the list seats are awarded in proportion to the list votes
only, without any regard for the seats that the parties won in the single-
member section of the election, so large parties retain the seat bonus that
they usually achieve in the SMD component. 23

6. Limitations on Proportionality?
Small M DM = 1 (Australia, Canada, India, UK, USA);
DM = 2 (Chile), average 4 (Ireland);
DM is in effect small in mixed systems when
list seat allocation is separate from single-
member seat outcomes (Japan, Russia)
Significant thresholds that parties need Germany, Hungary, New Zealand,
to cross in order to get in. Russia
Malapportionment USA (Senate), Chile, Spain, Canada,
France, India

•  Proportionality is generally regarded as a ‘good thing’ – in moderation.

•  Few electoral systems go for broke on the proportionality dimension; most

have, in practice, some way of limiting it.

24

12
 30/07/2013

Thresholds

•  Virtually every PR system employs some kind of threshold that prevents

the smallest parties getting their ‘fair’ share of the seats.

–  In Germany, for example, the only parties that qualify for any list seats are
those that either win 5 per cent of the list votes, or win three single-member
constituencies.
–  In Russia, the threshold is 7 %.

•  This discrimination against small parties and their supporters is usually

justified in terms of preventing excessive fragmentation

•  and thereby making it easier to form stable governments

–  a particular concern in post-communist countries given their usually weakly

structured party systems.

25

The Mechanics of Electoral Systems

•  For details and worked out examples see Appendix A of PES

(Gallagher and Mitchell 2005).

•  SMDs
•  1. SMP – plurality winner

•  2. AV – preferential ballot, majority winner by transferring votes of

eliminated candidates.

•  3. Two Round System

26

13
 30/07/2013

The Mechanics of Electoral Systems (2)

•  MMDs
•  List System PR - uses different types of seat allocation formulae – which can
be classified as either ‘highest average’ or ‘largest remainders’ methods.
•  (A) Highest Average methods allocate seats sequentially.
–  Each seat is awarded to the party that presents the highest ‘average’—the
‘average’ denoting the number of votes it won divided by a number reflecting the
number of seats it has already been awarded.
–  Thus, while the first seat obviously goes to the largest party, its average is reduced
because of this when it comes to competing for the second seat.

–  Exactly how these formulae work out in practice depends on the sequence of
–  numbers used as divisors.
–  In principle any sequence is possible, but in practice only a few are used in real-life
electoral systems.
–  The most common is the D’Hondt sequence: 1, 2, 3, 4, 5, and so on.
–  In other words, the average of a party that has won precisely one seat is obtained by
–  dividing its vote total by 2; if it has won six seats, its vote total is divided by 7; and so
on.

27

Other Divisors

•  While D’Hondt is the most common – others are:

•  The Sainte-Lague¨ formula: the sequence of divisors runs 1, 3, 5, 7, 9,

and so on.
–  This is rarely used in its pure form (New Zealand and Latvia are among the
few examples).

•  More common than ‘pure’ Sainte-Lague¨ is a sequence known as

‘modified Sainte-Lague¨’, used in several Scandinavian countries and
in Poland, where the sequence runs 1.4, 3, 5, 7, 9, and so on.

•  (b) Largest remainders methods proceed by calculating a quota, which

is based on the numbers of votes cast and the number of seats to be
awarded.
•  Each party is then awarded as many seats as it has full quotas, and if
this leaves some seats unallocated, the remaining seats go to the
parties with the most votes left over.
28

14
 30/07/2013

Do the different formulae and quotas matter?

•  From the most favourable to larger parties to the least favourable, the
order is:

•  Imperiali highest averages

•  LR–Imperiali
•  D’Hondt
•  LR–Droop
•  Modified Sainte-Lague¨
•  LR–Hare/Sainte-Lague¨

29

Consequences

30

15
 30/07/2013

31

32

16
 30/07/2013

33

Eight Criteria for Evaluating Electoral Systems

34

17
 30/07/2013

35

Conclusions?
•  Just as with mobile phones, it is impossible to say ‘whatever you’re looking for,
this one is best’.
•  On the other hand, it is possible to say: ‘if you want something that is likely to
produce certain specified outcomes, this is the model that best fits your
prescription’.
•  Those who value stable and ejectable government, and identifiability of
alternatives, above all else, even at the expense of performance on several
other criteria, will be attracted by a single member constituency system.
•  Those for whom a high degree of proportionality and disciplined parties are the
transcendent virtues will favour a closed-list system in a nationwide
constituency.
•  Those prioritizing proportionality, a high degree of voter participation, and
personal accountability of MPs will logically gravitate towards PR-STV or open-
list PR.
•  Mixed compensatory systems score well on nearly every criterion.

•  Beyond doubt, the choice of a particular electoral system will make certain
patterns of politics more likely and make others less likely.

36

18
 30/07/2013

Electoral System Design for Divided Societies

Single Member Plurality Alternative Vote (AV) Communal Rolls

(SMP)

•  Plurality system in single •  A majoritarian electoral •  Explicit recognition of

member districts.
•  A ‘winner-takes-all’
system that can be
dangerous for divided
societies since minorities
can be permanently
excluded from power.
•  Tends to over represent the
largest party, sometimes
massively so.
•  Expected to lead towards a
small party system.
•  Not generally
recommended for divided
societies
system (usually in single
member districts)
•  with an ordinal ballot, on
which electors can rank order
their choices: 1, 2, 3 . . .etc
•  Horowitz has advocated it,
believing it contains
incentives towards
moderation
•  because parties may need to
attract second and later
preferences from other ethnic
groups.
ethnic groups in the
electoral system
•  for example via each
‘community’ having its
own electoral roll,
•  which elects only members
of ‘its’ group to parliament.
•  Provides no real incentives
for cooperation.
•  More common are ‘reserved
seats’ for certain ethnic or
minority groups.
•  Another type: pre-
determined ethnically
composed lists

UK, USA, India Australia; Papua New Guinea Fiji; India; Lebanon; Singapore
37

List PR STV

•  Proportional representation using party lists. •  Proportional representation using an

•  Fairly represents all significant groups and
can then use their representatives as the
building blocs of a power-sharing settlement.
•  With consociationalism each group can be
represented in cabinet and government in
proportion to their electoral strength.
•  Minority interests are further protected
through mutual vetoes, and community
autonomy in cultural matters.
•  Enhances leaders security
ordinal ballot in multi member
constituencies.
•  Delivers proportional results, encourages
centripetal party competition via the
desire to attract second and subsequent
preferences (hence may promote inter-
ethnic appeal).
•  Allows voters a choice of parties and
candidates within parties.

South Africa, most of Europe Ireland, Northern Ireland, Malta

38

19
 30/07/2013

Part 2

The Timing of ‘Post-Conflict’ Elections.

39

Two Views of early elections:

(1) the benefits and (2) the dangers

•  Hold Early Elections:

•  ‘Since democracies do not fight each other and tend to settle

internal disagreements peacefully, these advocates have reasoned
that democratic transitions yield peace, the sooner the better.’

•  ‘In their view, pushing autocratic war-prone regimes along the fast
track toward democracy should break the power of violent
authoritarian elites, accustom people to the habits of democratic
participation, provide legitimacy for new leaders, and, in cases of
international peacekeeping or military occupation, hasten the
withdrawal of foreign forces’.

•  Brancati and Synder (2012), p.1-2.

40

20
 30/07/2013

Delay Early Elections

•  ‘Skeptics claim that early elections reignite violence by empowering

former combatants rather than liberal, programmatic political parties.’

•  ‘Early elections, they also contend, often take place when the rule of
law is weak, making it more likely that elections will suffer from
irregularities, candidates will resort to illiberal populist appeals, and
losers will refuse to accept the results peacefully.’

•  ‘For these reasons, the skeptics recommend postponing fully

competitive elections until some progress has been made in
strengthening the institutions needed to make democracy work,
including competent state bureaucracies, independent courts,
professionalized media, and functioning market economies.’
–  Brancati and Synder (2012), p.1-2.

41

Arguments for early elections

•  Strengthens legitimacy of the peace process.

•  No elections – offers combatants no democratic means of having

influence – and hence may prioritise a return to force.

•  Elections soon after a ceasefire may occur in helpful contexts if

peacekeepers and election monitors are still present.

•  In turn the liklehood of early elections – makes outside powers more

likely to commit to the deployment of peacekeeping forces – since the
timetable for elections can form a credible part of their exit strategy.

•  Aid agencies increasingly require indicators of ‘good governance’ –

democratic elections are one such indicator.

42

21
 30/07/2013

Arguments against early elections

•  Mansfield and Synder (2005) argue that democratizing countries should

delay first elections until they have built some political administrative
capacity.

–  ‘When countries democratize in settings lacking an independent judicial

system, a competent bureaucracy, and free media, they claim, electoral
politics becomes an exercise not in civic deliberation but in coercion,
manipulation, and nationalist, sectarian, or radical appeals.’

•  In an immediate post-war environment – the former combatants are

generally stronger than pro-reform, pro-democracy forces:
–  ‘Former combatants, turned politicians, reignite warfare by rejecting the
results of unfavorable elections and returning to war in the short term, or by
governing in an arbitrary, exclusionary, and exploitative manner, which
creates new grievances and provokes renewed fighting in the long
term.’ (Brancati and Synder 2012, p.5).

43

Arguments against early elections (2)

•  When post-conflict elections are held quickly – the former combatants

are likely to have material, organizational and credibility – advantages
over still emerging pro-reform movements:
–  ‘The ability of new pro-reform groups to compete on an equal footing with
former combatants typically depends on the development of new institutions
that facilitate new political alignments that cut across old ethnic or
patronage lines (Mansfield and Snyder 2005, 2007).’

•  In some cases – early elections are held not because of a commitment

to democracy – but because either”
–  (a) armed opposition groups demand them as a method of accessing
government
–  (b) the demands of foreign countries.

•  With early elections there are greater incentives to cheat – and get
away with it – since there are fewer institutional constraints – like an 44
independent judiciary and free press.

22
 30/07/2013

Mitigating the ‘Risks’ inherent in early elections

•  While Brancati and Synder regard early elections as inherently risky – a

number of factors can mitigate these risks:

1.  Decisive military victories: the losing side in the war lacks the ability to
return to violence if it does poorly in the elections.
2.  Demobilization: of one or more sides (or integration into a new army)
also reduce the risk of a return to violence.

3.  Power Sharing:

‘Power sharing guarantees that the side that loses an election will still
retain meaningful representation in government, access to state
resources, and/or some degree of autonomy. Because power sharing
reduces the risk that elections pose to incumbents, it makes early
elections more likely.’ (p.8).

45

Mitigating the ‘Risks’ inherent in early elections (2)

•  4. ‘State Capacity’: Delay elections until strong political,

administrative, and judicial institutions are in place.
–  These help new pro-reform parties to compete on a more equal footing with
the former combatants.

–  election commissions,
–  impartial, professionalized bureaucracies
–  independent courts.
–  independent, professionalized media

•  5. Some countries enjoy a favourable institutional legacy - for

example a useable civil service or legal
system was created in the preconflict period (for example during a
colonial period).

46

23
 30/07/2013

Brancati and Synder data, 2011/12

•  Data set of civil wars from 1945-2008: 134 civil wars included.

•  Post-conflict Election Timing = the number of months elapsed from the

end of the civil war to the first election.

47

Results

•  Elections that occur later – reduce the probability of a new war.

–  In model 8 – the probability of a new war decreases by about 31% if
national elections occur 5 years after the end of a civil war rather than 1
year after (holding all other variables constant).

–  Sequencing – whether the first elections were national or subnational – had

no significant effect.

•  Military Victories – reduce the chances of war after elections

–  in Model 10 – ‘the probability of a new war decreases by 0.25 points (or
49%) if national elections are held one year after a civil war ha ended in a
victory rather than a settlement or truce . . .’ (Brancati and Synder 2012, p.
19.)

48

24
 30/07/2013

•  UN Interventions also have a big positive effect on mitigating the effects

of early elections;
–  ‘if national elections are held one year after a civil war ended after a civil
war ended in a settlement or truce, the probability of a new civil war
decreases by 0.33 points (or 60 percent) if the UN intervenes than if it does
not, holding all other variables constant at their means.’ (p.19).

•  Power Sharing – also helps, with decentralization having the biggest

effect:
–  ‘For national elections occurring one year after a civil war ended, the
probability of a new civil war decreases by 0.72 points (or 84 percent),
according to model 12, if the political system is decentralized than if it is not,
holding all other variables constant at their means.’

•  Note though that Brancati and Synder do not have a direct measure of
power-sharing (either in peace treaties, constitutions or in actual
existence).

•  Like many authors they use PR elections and decentralization as rough

proxies for power-sharing.
•  Clearly it would be better to have a more substantive power-sharing
variable. 49

18 Journal of Conflict Resolution 00(0)

Table 2. Effect of Postconflict National Election Timing (Calendar Time) on New Civil War
Onset (Prematching)
(7) (8) (9) (10) (11) (12)

Election timing –0.01* –0.01** –0.02** –0.06** –0.01* –0.01***

(0.004) (0.005) (0.01) (0.03) (0.003) (0.01)
Polity IV –0.01 –0.03 0.06 –0.002 –0.01 –0.04
(0.05) (0.06) (0.10) (0.06) (0.06) (0.06)
Post-cold war era –1.78*** –2.38* –3.08*** –1.78*** –1.77*** –2.44*
(0.66) (1.26) (1.18) (0.68) (0.67) (1.28)
Victory –0.66 –1.14 –1.05 –1.70** –0.69 –1.25
(0.69) (1.25) (0.75) (0.87) (0.69) (1.32)
UN intervention –1.56** –2.54* –2.73** –1.82** –0.79 –2.64*
(0.70) (1.44) (1.23) (0.79) (0.84) (1.55)
UN assistance 0.74 1.48 0.87 0.75 0.73 1.48
(0.69) (1.16) (1.16) (0.70) (0.70) (1.15)
Previous civil wars 0.42** 0.32 1.60** 0.48** 0.46** 0.30
(0.18) (0.23) (0.76) (0.23) (0.21) (0.23)
PR system 0.66 0.72
(0.64) (0.65)
Mixed system –0.34 –0.38
(1.03) (1.05)
Unitary executive system 0.91 0.95
(0.66) (0.66)
Decentralization –3.32*** –3.71***
(0.83) (1.03)
Timing*Victory 0.05**
(0.03)
Timing*UN intervention –0.06
(0.05)
Timing*Decentralization 0.01
(0.01)
War duration –0.02*
(0.01)
Deaths 1.02e-06
(1.09e-06)
Displaced 1.84e-06**
(8.17e-07)
GDP per capita –0.29
(0.23)
Constant 0.56 2.77 2.82*** 1.51* 0.53 3.03
(0.68) (1.85) (1.09) (0.79) (0.67) (2.00)
Log likelihood –51.02 –36.18 –18.47 –49.37 –50.33 –35.87
Observations 94 94 52 94 94 94

Note. Standard errors are in parentheses. 50

***p < .01. **p < .05. * p < .10.

results in Table 1, elections that occur later reduce the probability of a new war
breaking out. The effect, though, is not robust across models.16 The effect is much
stronger, however, if we restrict our analysis to only national elections as we do in

Downloaded from jcr.sagepub.com at London School of Economics & Political Sciences on March 6, 2013

25