es2

Allen

/Costs

of Economic

Sanctions

925

and

unable

to

spread

their

message

domestically,

as

the Milosevic

government

restrictedaccess

to valuable

supplies such as newsprint (Lopez and Cortright 1995;

Licht

1995).

As

a

result

of

increased

repression

and

decreased

resources

for opposi

tionactors, political activitywill

likelydecrease

under sanctions in these states.

In

autocratic

states, economic

sanctions

can

have

the

perverse

consequence

of

strengthening the regime in power,

increasing

its ability

to limit the activities of

opposition

forces.

In

these

societies,

there

is often

a

strong

relationship

between

those with political

power

and

those with

economic

power. Control

of scarce

resources coming

inand out of the statewill

give

those inpower

leverage over their

rivals

(Woodward

1995; Gibbons

1999, Cortright and Lopez

2000). The govern

ment

and

its supporters are also

likely to profit from black-market

activities

that

springup

in thewake

of sanctions (Niblock 2001 ).7

Beyond

strengthening

the regime,

sanctions

that

harm

the

country

may

not

hurt

the

leaders

in

power.

Under

certain

circumstances,

it may

be

that

sanctions

are

costly for states but not for leaders.8 Sanctioned

leaders and elites

inpolitical

sys

temswith a limitednumber of veto players may

be able

to consolidate

power under

sanctions.

In Yugoslavia

in the early 1990s, militia

groups

used

their control of

security

checkpoints

and

transportation

routes

to

enrich

themselves

monetarily

while also increasing theirpolitical influence.

Given

the likelihood

thatrepression will

increase

thepolitical

(and perhaps

also

economic)

costs

of

political

action

under

sanctions

(when

economic

costs

are

Autocratic

Opportunity Hypothesis:

The

difference

between

the occurrence

of antigov

ernment

activity

when sanctions

are

in place and

in their

absence will be

smallest for

states with strong

authoritarian

institutions.

Range

of Opportunities

inDemocratic

States

States with high levels of societal controls are

likely to use

repression, but not

all

leaders can use

repression effectively (Gelpi

1997). Democratic

principles and

institutions are both

thought to deter elites

from using violence

against theirciti

zenry (Gurr 1986). These

realities affect thedecisions

thatgovernmentsmake

con

cerning

sanctions

response

and

may

result in concession

by states

that

cannot

use

repression.

Democratic

leaders,

who

are

held

accountable

for

their actions

in regular

elections,

are

unlikely

tomake

the

unpopular

choice

to

impose

repressive

measures

in

response

to

antigovernment

behavior.

Democratic

societies,

however,

have

created

a

range

of

"appropriate" methods of expressing political discontent.By

institutionalizingpublic

political

expression

through

regular

elections,

democracies

have

reduced

the

incen

tives

for

irregular antigovernment

activity.

The

norms

of

democratic

societies,

such

as

furtherdecrease

thebenefits for irregularantigovernmentactivity.

Tilly

(1978)

described

political violence

as the outcome

of conflictbetween

the

ruler and

the ruled as

the government responds

to protest by repression, which

in

turn

sparks

more

violent

opposition.

In

democracies,

not

only

are the means

of

expressing discontent institutionalized, so too are themeans

of

response

by

the

government.

Democratic

leaders

are unlikely

to

resort

to repression

because

of

the

political costs thatwould be associated with

such a move. Without

the repression,

theneed

for a violent

response

isminimized.

Sanctions do not greatly change

the

costs

associated

with

irregular

antigovernment

activity

in a

democracy.

Leaders

in sanctioned democratic countries are forced tomake

difficultdecisions.

If thedecision

ismade

to resist sanctions, how can

scarce resources be distributed to

keep

enough supportershappy? Leaders

in these states are constantlyendeavoring to

maintain

or increase public

support,and sanctions put this support in jeopardy. This

fact

helps

explain

why

democracies

are more

likely

to

concede

to sanctions pressure

and do

so more

quickly

than do other states (Allen

2005,

2008). Antigovernment

activity

may

not

be necessary

in

these

states

to bring

an

end

to

sanctions.

Falling

between

democracies

and

autocracies

in

terms

of

costs

and

benefits

are

in antigovernment activity in these states face

mixed

regimes. Individuals engaging

neither thehigh costs of repression, as

in autocracies,

nor the low benefits of such

action

in democracies. Leaders

inmixed

regimes

lack the power

and

influence to

repress that autocrats have

at theirdisposal. At

the same

time, they also

lack

the

societal mechanisms

for orderly public

expression

as well as

the sense of

legiti

macy

that their democratic

counterparts enjoy.

In

these

states, the likelihood

of

antigovernment

behavior

is greatest,

a fact

that is enhanced

by economic

sanctions.

Feelings

of

deprivation

increase

under

sanctions,

but

the costs

associated

with

anti

government activity do not. This

leads to a final political-opportunity hypothesis.

Mixed

Opportunity

Hypothesis:

The

difference

in

the occurrence

of

antigovernment

activity

when

sanctions

are

in place

and

in their absence

will

be

greatest

formixed

regimes.

Data

andMethods

Methodology

To examine

thepublic political

response

to sanctions, patterns of political

action

are observed.

Do

more political activities,

such as

demonstrations

and riots,

occur

in thepresence

of sanctions? Unfortunately, it is impossible from thedata available

to discern

whether antigovernment

activity

under

sanctions

is a response

to

the

sanctions themselves. Some or perhaps all of the antigovernment activitymay be

in

Allen /Costs of Economic

Sanctions

927

level,

sanctioners

anticipate

that

their

policy

may

cause

antigovernment

activity

in

direct and indirectways. For

this reason,

I believe

it is appropriate

to consider

all

antigovernment activity and

the differences in rates of activity between

sanctioned

and

unsanctioned

periods.

The

nature of theoutcome variables being examined

(the number of thepolitical

events

thatoccurred)

limits the types of analysis

that are appropriate. The

values

that

the

dependent

variables

can

take

on

are

bounded

in

that

there

cannot

be

nega

tive

numbers of

events. These

values

are

also constrained

to

be

integers.

Because

of these characteristics,

the data

are

heteroskedastic

necessarily

(Greene

1997),

thus requiring special methodological

consideration.

Event-count

data

such

as

these

can

be

modeled

in

several

ways,

depending

on

to

overdispersion.

Since

the

rate

of

occurrence

(X)

of

these events

is not

constant,

the

basic

Poisson

model

is

unsuitable.

Overdispersed

count

data

have

a

variance

greater

than their

mean,

which

also

violates

one

of

the basic

assumptions

of the

Poisson model. To

avoid

inefficientestimates

(which result from using

thePoisson

model

to estimate

relationships

in overdispersed data), an analysis

of the negative

binomial event count is appropriate (Liao 1994; King

1989).

To perform

the analysis,

measures

of

political

activity

are

needed.

In

an attempt

to get

the broadest

data

coverage,

I use

measures

from

the

Cross-National

Time

Series Archive

(CNTSA),

1948-1997

(Banks 2003). Although flawed, these data

provide me with

thegreatest breadth of coverage

(both in termsof time and space).

Mass

political

conflict is popular mobilization

fornoninstitutionalized collective

action. For thepurposes of thisanalysis, I look at two distinctmeasures

of domestic

1997).

Antigovernment

demonstrations

are

typical

protest

events.

Protest

events are

the more

likely

form of

collective

action

against

the government

inmore

open

states

(Gurr 1989). The CNTSA

codes demonstrations events thatmeet the following cri

teria: "Any peaceful public gathering of at least 100 people

for theprimarypurpose

of displaying or voicing

theiropposition

togovernment policies

or authority,exclud

ingdemonstrations of a distinctlyanti-foreignnature"

(Banks 2003).

In contrast,

rebellion

events

are more

violent,

as

participants

aim

to change

more

thanpolicies,

looking to alter the system itself.Riots

typifyrebellion events. These

events

are more likely

forms of

expression

in closed

societies

(Gurr

1989).

By

including

these events, I hope

to discern whether distinct patterns of action

928

Journal of Conflict

Resolution

under

sanctions

emerge?both

as

society

type

varies

and

as

events

intensify.

Riot

events

are

coded

under the following

definition: "Any

violent

demonstration or

clash

of more

than 1000

citizens

involving

the use

of physical

force"

(Banks

2003).

IndependentVariables

The

primary independent variable of interest is theoccurrence of sanctions from

1948 to 1999. These

data are drawn primarily from theHSE

data

set (1990),

aug

mented byMarinov

(2005). This

indicator variable

is scored as

1 forevery year that

the sanctions

were

in place

and

0

otherwise,

for

a

total

of

1,303

sanctioned

years.

A

line of theorizing complementary

to thatpresented above

is the idea thatdif

ferent types of sanctions affect leaders and populations

in differentpolitical sys

tems indifferentways

(Brooks 2002). To

this end, indicators of sanctions type are

also

included

in the analysis. These

data

come

fromHSE

with

forward coding by

this

author.

Financial

sanctions,

which

directly

target

the assets

and movement

of

elites, are believed

tohave

the greatest effect on autocratic

leaders

(Brooks 2002).

If the public

is not directly affected by

the sanctions, antigovernment activity is

unlikely

to

increase.

sanctions tend to create stakeholders in the targetstatewho benefit as long

Export

as sanctions remain inplace

(Seiden

1999). This

suggests that the impact of export

sanctions on

increasing political

action

should be weaker

than the impact of other

types

of sanctions.

On

the

other

hand,

import

sanctions

do

not

provide

the

same

opportunity for domestic

substitutions and

thus should be

associated with higher

levels of political action, or a "fifthcolumn effect" in theparlance of Seiden

(1999).

Ideally,

I would

also be

able

to include a dynamic measure

of the costs of the

sanctions. Unfortunately,

thisposes

a difficult challenge

over

the full time period

of interest (1915

to 1990). Estimates forGNP

and GDP

are not consistently avail

able

for most

states

before

1950,

so

an

alternative

measure

must

be

used.

The

HSE

data

include an estimated

annual change

inGNP

because

of

sanctions, but their

measure

is a

single

average

estimate

for

each

sanctions

episode

rather

than

a mea

sure thatvaries over timeduring the sanctions period.Marinov

(2005) updated this

measure,

and

I use

these data

tomeasure

the cost

of

sanctions.

To

capture the domestic

politics of

the target state, I use

Polity

IV

(2003)

democracy-autocracy

scores

(Marshall

and Jaggers 2006). This

variable ranges

in

value

from-10

to 10. In addition to the simple scores, thepolitical-opportunity

lit

erature suggests a curvilinear relationship,which

can be picked

up by

including a

squared Polity term.Furthermore,when

thepolitical-opportunity idea is applied

to

sanctioned

states,

the

expectation

is

that

the impact

of

sanctions

should

vary

across

regime

types,

so

an

interaction

term

is also

included.

Allen /Costs of Economic Sanctions

929

Table 1

Summary

Statistics for Independent

Variables

Variable

Mean

Std. Dev.

MinMax. .

(N)

Presence

of sanctions (9,346)

0.1390.346

0

1

Polity

x

sanctions

(5,641)

-0.443

3.010

-109

Sanctions

costs

(6,765)

0.1309.546

0

14.7

Polity

score

(7,462)

-0.425

-710.607

Polity

score

squared (7,462)

58.039

33.227

0

100

ln(population)

(9,617)

8.721.794

2.773

14.061

ln(energy consumption)

(9,123)

7.995

3.003

0

14.946

Regime

durability

(9,053)

20.088

28.940

0

193

I have also

included the regime durabilitymeasure

fromPolity IV. Stable

regimes

are

thought to have

less

irregularpolitical

action

than transitional states (Snyder

2000). This

variable

ismeasured as time since the last regime transitionor firstyear

of independence. Summary statistics for these characteristics of the targeted regime

variables

(aswell as theother independent variables)

can be found inTable

1.

The

success

or

failure

of

previous

political

action,

such

as

demonstrations

and

riots, is likely to influence the decision

to use

thesemethods

in the future.The

political-violence

literature

also

suggests

that

violence

often

breeds

violence

(Lichbach

and Gurr 1981). For this reason, a

lagged dependent variable

is included

to control

for

nonindependence

across

time.

Standard

errors

are

also clustered

by

country

to

further

account

for dependence

within

country-units.10

In

addition

to

these

sanctions and

political

variables,

I

look to the political

violence

literatureto control forother factors thatare likely to influence thenumber

of political

events occurring. Jenkins and Schock

(1992)

strongly recommend

the

need

to control forpopulation

size, so an

indicatorof population (logged)

is included

in theanalysis. Not doing so ignores thepossibility of the simple hazard of collective

action's occurring

as an

explanation.

Economic

conditions

are also

important

for the

prediction of political

violence,

almost as much so

as

political

conditions

(Gurr

1989). Because

of thedifficultyof getting comparable

data across all thenecessary

log of thousands

of coal-ton equivalents.

Both of these

variables

are drawn from the

Correlates ofWar

Capabilities data (Singer and Small

1996).

The results of thenegative binomial

regression analysis forpolitical demonstra

tions appear inTable 2. The firstmodel

is for all sanctions, while the other three

Table

2

Negative

Binomial

Analysis:

Political

Demonstrations,

1948-1999

Variable

(Std. Err.)

All Sanctions

Financial

Sanctions Import Sanctions

Export

Sanctions

Lagged

dependent

variable

0.244***

0.241***

0.243***

0.244***

(0.035)

(0.035)

(0.035)

(0.035)

Polity

score

0.031***

0.034***

0.031***

0.030***

(0.010)

(0.010)

(0.009)

(0.009)

Polity

squared

-0.010***

-0.010***

-0.011***

-0.011***

(0.002)

(0.002)

(0.002)

(0.002)

Polity

x

sanctions

0.043**

0.046*

0.071***

0.073***

(0.020)

(0.023)

(0.027)

(0.026)

Presence

of sanctions

0.577***

0.674***

0.473*

0.406

(0.146)

(0.172)

(0.254)

(0.249)

Sanction

costs

-0.025

-0.057*

-0.028

-0.024

(0.033)

(0.030)

(0.037)

(0.037)

Regime

durability

-0.003

-0.003

-0.003

-0.003

(0.003)

(0.003)

(0.003)

(0.003)

ln(energy consumption)

0.115**

0.120***

0.127***

0.132***

(0.047)

(0.047)

(0.046)

(0.047)

ln(population)

0.304***

0.0308***

0.318***

0.310***

(0.067)

(0.068)

(0.067)

(0.068)

Intercept

_4.465***

?4.524***

-4.536***

-4.520***

(0.455)

(0.470)

(0.4564)

(0.459)

Disperson

Parameter

Zn(oc)

1.081***

1.087***

1.090***

1.092***

(0.109)

(0.109)

(0.111)

(0.109)

N

5,109

5,109

5,109

5,109

Wald

x2

474.510

456.470

520.460

514.920

Prob >

0.000

0.000

0.000

0.000

x2

Log-likelihood

-4075.440

-4087.570

-4093.220

-4089.830

Note:

Significance

levels: * <

10%;

:<

1%.

The

results

across

the

four models

are

largely

consistent. First,

previous

political

action does

appear

to lead to furtherpolitical

actions, as

thepositive coefficients on

the lagged

dependent variables

indicate. Publics

that have

engaged

in political

action

in the past aremore likely to do

so again,

following

the earlier findings of

Hibbs

(1973). This

result, along with

the statistical significance of thepositive dis

persion

parameters,

also

suggests

that

the data

are

overdispersed

and

that the

nega

tivebinomial model

is the correct specification.

The

findings concerning regime type are a bitmore

nuanced. The

Polity

term

and

thePolity Squared

term combine

to predict one nonlinear

impact on

thepre

dicted numbers of events. The

negative

sign on

the squared

term indicates that the

curve

opens

downward.

The

shape

of

this

relationship

can

be more

easily

under

stood graphically inFigure 2.

Allen /Costs of Economic Sanctions

931

on the squared term).11This

shape is generally in linewith

the "invertedU"

hypoth

esis of political

opportunity that suggests thatpolitical

action should be most

likely

when political

constraints are at a moderate

level?neither

high, as in autocracies,

nor low, as indemocracies. The

predicted number of demonstrations indemocracies

isnot as low as

it is inautocracies,

likelyevidence of the idea thatnonviolent protest

is themethod of choice forpolitical

action in free societies

(Gurr 1989).

sanc

Most

significantly for thepurposes of thisproject, the effect of economic

tionsmust be evaluated. Because

the sanctions variable is used to create

the inter

action

term to test the conditional

hypothesis

that the impact of sanctions varies by

regime

type, it is necessary

to determine themarginal

effects of sanctions alone.

The coefficient offered inTable

2

is instructiveonly for stateswith a Polity

score

of zero. Following the procedures

laid out inBrambor, Clark, and Golder

(2006),

I calculated and graphed the independentmarginal effect for sanctions across

the

Polity scale (Figure 3).12

scores greater than

From Figure 3, it is possible to see thatfor stateswith Polity

0, sanctions have a statistically significant positive impact on

the likelihood

of

as states become

less autocratic provides some support for theAutocratic Opportu

nity Hypothesis.

Figure 3 does not offer consistent support for theDeprivation

Hypothesis.

Political

actors

in all

states do not demonstrate an increased willing

ness

to take action against

theirgovernments; this is only

true for actors inmore

open

societies, which

is again more

in linewith the idea

thatopportunity ismore

influential thandeprivation.

size of coefficients is roughly the same for all

Interestingly, the direction and

types of sanctions. Antigovernment

activity appears to increase whenever sanctions

are in place.

Seiden

(1999)

anticipates that import sanctions are the sanctionsmost

likely to encourage mass

political action, but the coefficient estimate for the import

sanctions variable does

not stand out

in comparison to theother typesof sanctions.

Allen /Costs of Economic Sanctions

933

of demonstrations predicted

is higher when

sanctions are in place

than in their

absence. That difference between

thepredictions is smallest for autocratic states and

largest for anocratic states,which

supports theMixed

Opportunity Hypothesis. For

ease of interpretation,I only

included the confidence

interval (CI)

for thenumber of

events under sanctions.When

both sets of confidence bounds are

included, the two

intervalsdo not overlap

for stateswith Polity

scores greater than 2,

suggesting that

antigovernment activity

inmost

states with

and without sanctions

is statistically

indistinguishable. If sanctions

senders are hoping

to inspire acts of discontent

to

threatenleaders, particularly autocrats, theyare likely tobe disappointed.

The

cost of sanctions does

not obtain

statistical significance

in themajority

of

themodel

runs (the exception

being

for financial

sanctions), mirroring the results

of Jing,Kaempfer, and Lowenberg

(2003). Using

a simultaneous-equations model

to capture

how

the choice

of

coercive

tool

(military action,

trade sanctions,

or

financial sanctions) affects theoutcome of sanctions episodes,

Jing,Kaempfer, and

Lowenberg

find the cost to the target to be an insignificantpredictor of outcome.

Marinov

(2005)

also found that the costs of sanctions were not a significantpredic

torof leadership change.

The

relative-deprivation school of thought suggests thatas

the cost of sanctions

increases, feelings of deprivation

are likely to increase, which

in turn should lead

to an increase

in antigovernment activity. The

lack of support for the cost variable

may

be caused

by

the lack of a

dynamic

measure

of cost, or

it

be

caused

by

may

the fact thatthedeprivation

idea

is an oversimplification. One

of thekey criticisms

of the relative-deprivation explanation

for political

violence

generally

is a

lack of

clear understanding of when

individual

feelings will

result inmass

action.14 The

same can be

truefor thisexplanation

forhow

sanctions should affect thedomestic

politics of the target.Another possibility

is that antigovernment behavior

is trig

gered bymore

than just feelings of deprivation. Sanctions may

also

signal interna

tional

support to domestic

political-opposition

actors, increasing

the benefits for

antigovernment

behavior.

the control variables

for population

size and energy consumption

Additionally,

(as a proxy for level of development)

also

attain statistical significance. Both

have

a positive

relationship with

the

likelihood

that demonstrations will

occur. The

population

finding supports Jenkins and

Schock's

(1992)

idea

that increasing the

population

size increases

the

baseline

hazard

rate

for

incidence

of

demonstrations.

More

people

introduce

more

possible

causes

for antigovernment

activity.

The

posi

tive

coefficient

for the energy-consumption

variable

is a

bit

more

surprising,

sug

gesting

that

as

states

become

more

developed,

demonstrations

are

more

likely

to

occur. This

resultmay

be driven by

the fact thatpopulations

in the poorest

states

are unable

to engage

inpolitical

action, as

theyare forced todevote

all theirenergy

to subsistence

(Niblock

2001).

To

clarify

the

substantive

impact

of all

the

variables,

incidence-rate

ratios

have

been generated forboth types of action. These

ratios are presented

inTable

3. Like

odds

ratios for

models,

these ratios

represent

the relative

change

in the inci

logit

in a

dence

rate for a

1-unit change

particular variable

(Zorn

forthcoming). For

example,

the incidence

rate

for demonstrations

under

sanctions

is

1.745

times

what

in the absence

of sanctions.15A

1-unit increase inPolity

score leads to

itwould

be

an

increase

in the incidence

rate for both types of events of about

1.03.Moving

from-10

to 10 on thePolity

scale

(or a 20-unit change) corresponds

to an

increase

in

the

incidence-rate

ratio of

1.63.

This

means

that holding

all

other

variables

con

stant, a

democratic

state with

a

Polity

score of

10 will

have

1.63

times more

demonstrations

than

an

autocratic

state

at

the

other

extreme.

Now

turningto the analysis of riots, thefindings are similar (see Table 4). In line

with earlier scholarship and theabove

results,we see thatprevious violent acts (riots)

ability variable)

also decreases the likelihood of violent rebellion. This

result,which

ismuch more stronglysupported for themore violent act of riots than itwas

in the

demonstrations

model, has

bearing

on the scholarship

on democratization

and

pro

pensity forwar

(such asMansfield

and Snyder 1995). States inwhich

the political

system is influx are unable

toquell

violence eitherby repression or by creating law