Is The Law A Mere Parchment Barrier To Human
Rights Abuse?
Linda Camp Keith The University of Texas at Dallas
C. Neal Tate Vanderbilt University
Steven C. Poe (deceased) University of North Texas
The ‘‘mere parchment barriers’’ created by constitutional provisions may lead to decreases in the extent to which
nations abuse the human right not to be imprisoned, tortured, killed, or made to disappear arbitrarily or because of
your political views. A global pooled cross-national time-series analysis for a 21-year period shows that adopting
selected constitutional provisions protecting individual rights and freedoms, promoting judicial independence, and
guarding against states of emergency—and keeping the provisions in place for 10 years—has the potential to reduce
a nation’s level of state terror substantially, from one in which political imprisonment affects large numbers of the
politically active population and political murders are common, to one where the rule of law is secure, for example.
We report significant caveats about and limitations of the research. Nevertheless, we conclude that, since it may be
easier to change constitutions than to build effective democracy, to create massive amounts of new wealth, or to
avoid or defuse deeply ingrained conflicts, these findings have possible policy and scientific importance.
I
s the law an effective or merely a parchment barrier
to human rights abuse? There are few questions that
are of greater interest and importance to involved
citizens, politicians (statesmen), and scholars. If the law
is an effective barrier to human rights abuse (repression),
then those who wish to prevent it are well advised to
invest their time in establishing the kinds of law that
deter it. If the law is no more than a flimsy parchment
barrier to rights abuse, then efforts to create legal barriers
to repression may be irrelevant: those who wish to
prevent human rights abuse should probably concentrate their efforts on other prophylactic measures.
We tackle our title through an empirical analysis.
In that analysis, formal constitutional provisions
relevant to protecting rights represent the law, and
attacks on personal integrity—imprisoning or torturing people, making them disappear, or killing them
because of their political views—define human rights
abuse. Our analysis is worldwide, cross-national, and
cross-time.
Whether the law is or can be an effective barrier to
human rights abuse has been debated for centuries.
In the United States, James Madison, who helped
shape the U.S. Constitution and introduced its Bill of
Rights into Congress, was remarkably ambivalent
about the effectiveness of such ‘‘parchment’’ barriers
against abuses of power. Assessing the U.S. Constitution’s provisions for the separation of powers,
Madison wondered ‘‘Will it be sufficient to mark
with precision the boundaries of these departments in
the Constitution of the government, and to trust to
these parchment barriers against the encroaching
spirit of power?’’ (1788a). Pondering the effectiveness
of the Bill of Rights, Madison concluded ‘‘in a certain
form and to a certain extent, such a provision was
neither improper nor altogether useless.’’ (1789).
Madison was aware that constitutional provisions
and protections often did not produce the results
they were intended to produce. Yet his rhetorical
question about the separation of powers prefaces his
argument for the need to have each department of
government exercise ‘‘a constitutional controul over
the others,’’ and his chicken soup attitude toward
the Bill of Rights—‘‘it won’t hurt’’—reflected his
belief that the body of the constitution as drafted
already provided as effective a set of barriers against
government abuse of liberty as ‘‘parchment’’ could
provide.
The Journal of Politics, Vol. 71, No. 2, April 2009, Pp. 644–660
Ó 2009 Southern Political Science Association
doi:10.1017/S0022381609090513
ISSN 0022-3816
The Debate over the Efficacy of Law
644
is the law a mere parchment barrier to human rights abuse
645
This article seeks to discover whether there may
be reason for Madison’s ultimate confidence in
constitutional protections, although the unavailability of the data to test alternative hypotheses leaves
that conclusion tentative. Taken cumulatively and
allowed to work their effects over a reasonable period
of time, the mere parchment barriers of constitutional provisions associate with substantial decreases
in the extent to which nations engage in abuse of the
most fundamental of their citizens’ rights. Since it
may be easier to change constitutions than to build
effective democracy, to create massive amounts of
new wealth, or to defuse deeply ingrained conflicts,
we believe these findings have potential policy as well
as scientific importance.1
Others, like Madison the skeptic, question the
effectiveness of formal mechanisms. They know that
constitutions may provide more window dressing than
substantive protection for human rights (Howard
1991, 3). Furthermore, constitutions may also limit
rights or state that they must not harm the nation’s
interests or the public good (Ludwikowski 1996, 227).
Analysts have made few attempts to assess systematically whether Madison the constitutionalist or
Madison the skeptic is correct. While constitutional
experts have been advising newly independent states
and emerging democracies on how to provide constitutional protections to their citizens, there has been
very little empirical analysis that would support their
advice or facilitate the creation of effective constitutions.
Constitutional Change and
Human Rights Protection
What We Analyze and What
We Do Not Analyze
The fall of the Soviet Union and the spread of
constitutional democracies around the world followed upon a period of near-global acceptance of
international human rights standards that legally
bind states. By the end of the last century, constitutions had become ‘‘the only fixture of modern
government to be universally accepted’’ (Blaustein
1993, 70–71; see also Howard 1991) and more than
100 constitutions contained formal bills of rights
(Blaustein 1993).
Some observers of these changes (Ackermann
1989; Elster 1993; Finer et al. 1995), like Madison
the constitutionalist, are confident that the proliferation of formal-legal rights protections will improve
the actual state of human rights in the countries
adopting them. The institutionalization of the rule of
law and the building of an independent judiciary
were also direct goals of United States democracy
assistance (see Talbott 1998). These optimists presume that regimes will be less willing to abuse rights
that are clearly and publicly promised in a legally
binding document (Rosenthal 1990, 401). Such promises may pressure a regime to give in to demands for
fundamental protections (see Elster 1993; Finer et al.
1995). As Madison put it, ‘‘political truths declared in
that solemn manner . . . become incorporated with
the national sentiment, and counteract the impulses of
interest and passion’’ (1788b).
‘‘Is the Law a Mere Parchment Barrier to Human
Rights Abuse?’’ has been answered mostly in prescriptive or anecdotal accounts. We seek to provide
an answer based on systematic analysis of a large
quantity of relevant data. In developing that answer,
we investigate whether formal constitutional provisions promote actual protection of the human right
to personal integrity. Specifically, we examine the
impact on integrity rights of three types of constitutional provisions: (1) provisions for specific individual
freedoms; (2) provisions for an independent judiciary
(arguably necessary to fully implement the constitutional protection of rights); and (3) states of emergency provisions, which may facilitate or curb a
regime’s proclivity to abuse rights.
We want to emphasize what we are and are not
analyzing, and what inferences we are and are not
able to support. First, we stress that our analysis is
focused on a simple, but important, question: if
nations adopt constitutional provisions to protect
rights, to promote judicial independence, and regulate states of emergency, do they repress integrity rights
less? We are able to measure only whether nations have
set up Madison’s parchment barriers. We currently
lack the data to measure systematically whether citizen
rights are protected, judges are independent, and states
of emergency are regulated as the constitution says they
should be.
Given an affirmative answer to our question, we
can infer that if nations do adopt formal provisions
protecting rights, promoting judicial independence,
and regulating states of emergency, they repress their
1
Our judgment is relative. We do not imply that writing and
winning adoption of a new constitution is in fact ‘‘easy.’’
Constitutional creation and adoption is likely to be a contentious,
highly political process. Our conclusion expands this point.
646
linda camp keith, c. neal tate, and steven c. poe
citizens less. We cannot infer for certain that it is the
adoption of the constitutional provisions that causes
the reduction in repression. Perhaps governments
adopt constitutional protections when they are already predisposed to enforce them and not to abuse
personal integrity. Since we (and everyone else, so far
as we are aware) lack indicators of governmental
predispositions, we are not, ultimately, in a position
to resolve this causal conundrum. Insofar as data
allow, however, we do postulate and analyze alternative model specifications designed to eradicate the
link between constitutional provisions and respect for
human rights.
their protection by the judiciary’’ (Rosenthal 1990,
401). Presumably regimes would be less willing to
abuse rights that are clearly and publicly promised to
their citizens in a legally binding document and that
are supported by constitutional mechanisms, such as
an independent judiciary. But as we have noted in
our discussion above, this optimism is countered by
skeptics who argue that the effects of these documents ‘‘are commonly exaggerated’’ or in fact may be
‘‘shams’’ (see Epp 1998, 13) and mere window
dressing (Howard 1991, 3).
Integrity Rights Theory and
Constitutional Provisions
Our purpose is to investigate, as rigorously as possible,
the ability of constitutional rights protections, judicial
independence provisions, and state of emergency
regulations to ameliorate state abuse of personal
integrity—the right not to be imprisoned, tortured,
killed, or made to disappear either arbitrarily or
because of your political affiliations or convictions.
Constitutional Provisions
What was long regarded as an American peculiarity, a
written constitution with a bill of fundamental rights,
protected by an independent judiciary, especially one
empowered with judicial review, today is seen as a
global model of democratization or is seen as part of
the ‘‘script of modernity like presidencies and legislatures’’ as Ginsburg notes (2003, 26). Waldron
(1998) further notes that newly independent or
democratizing countries ‘‘turn almost instinctively
to some version of this constitutional arrangement’’
(335). Indeed, as Howard (1991) suggests, almost
every country in the world has adopted some document called a constitution, and as Blaustein concludes, constitutions have become ‘‘the only fixture of
modern government to be universally accepted’’
(1993, 70–71). As we noted above, for many this
trend represents substantial progress in the democratic transition away from authoritarian regimes and
engenders considerable potential for human rights
protection (e.g., Rosenthal 1990; Elster 1993 and
Finer et al. 1995; Epp 1998; Teitel 2000). This
optimism in constitutionalism may stem from the
general belief that the best way to safeguard individual
freedoms is through ‘‘the enumeration of rights and
Empirical Studies
Ultimately, the effect of formal provisions of law on
state human rights behaviors remains an empirical
question, one that political scientists have only begun
to explore in systematic analyses. Early analyses (BoliBennett 1976; Pritchard 1986) found evidence that
would support our skeptics—not only were constitutional provisions not associated with improved rights
behavior, the associations were in the opposite direction. However, these early studies relied on simple
bivariate analyses of the association between constitutions and various rights measures and tended to be
limited to a single year. The generalizability of their
results is thus limited, and, more importantly our
confidence in their conclusions is severely limited as
we know many potentially confounding factors contribute to states human rights practices.
Subsequent analyses grew in rigor and sophistication. Blasi and Cingranelli (1996) moved our attention beyond constitutional provisions that were
direct promises of freedoms and rights to institutional provisions expected to facilitate the provision
of human rights through an independent judiciary.
They constructed a constitutional index of judicial
independence that was associated with actual judicial
independence and found that, in turn, judicial independence was associated with human rights protection. While their study was somewhat restricted in the
number of countries examined, it provided the first
evidence of at least a weak effect of constitutional
provisions on respect for human rights.
Davenport (1996) moved us further along by
presenting a rigorous and comprehensive analysis
of government application of ‘‘negative sanctions’’
(‘‘limitation of the media and political restrictions
upon individual citizens and political parties’’; Taylor
and Jodice 1983) for 39 countries over a 35-year period (1948–82) and bringing additional components of
constitutionalism into his analyses. He examined constitutional provisions for and limitation of fundamental
is the law a mere parchment barrier to human rights abuse
rights as well emergency rule clauses. Only three of
Davenport’s 14 constitutional provisions had statistically significant effects. He found that the constitutional promise of freedom of press reduced, (2)
constitutional restriction of the press increased, and
3) state of emergency clauses reduced negative
sanctions. Several other provisions appeared to increase sanctions, but their coefficients did not achieve
statistical significance.
Davenport’s analysis demonstrated that constitutional provisions have some effect on rights practice
as the constitutional optimists would expect, but its
contribution to answering our title question is limited.
As Davenport notes, his sample is limited and not
representative of the world. In addition, his dependent
variable taps a quite different strain of state human
right behavior than we analyze in this article. Finally,
his time period ends in 1982, well before the development of constitutions in the Third Wave of democratization and the postcommunist era.
Keith (2002) sought to build upon the work of
Blasi and Cingranelli and Davenport by examining
across the global set of countries for a 20-year period
the impact of 10 constitutional provisions that
ranged from individual freedoms such as freedom
of speech and press to trial related provisions and the
ban against torture. She too found less of an effect
than the optimists would hope for, finding that only
provisions for public and fair trials made a difference
in states’ level of personal integrity abuse.
The theoretical and empirical literature suggests
that the constitutional provisions that are most relevant to protecting citizens against serious repression
fall into three related general categories: protections of
individual freedoms and rights, promotion of judicial
independence, and regulations of arbitrary rule during
states of emergency. To date no empirical study has
examined all three categories simultaneously and only
one category has been examined across a global set of
countries for a significant period of time.
647
use are those originally developed by Stohl and others
(Gibney and Dalton 1996; Gibney and Stohl 1988;
Poe 1992) and maintained by Mark Gibney (http: //
www.unca.edu/~mgibney/). They were expanded to
greater geographical and temporal coverage by Poe
and his collaborators (Poe and Tate 1994; Poe, Tate,
and Keith 1999). From the accounts in the annual
reports of the U.S. Department of State (SD) and of
Amnesty International (AI) that cover human rights
conditions in almost all the world’s nations, these
investigators construct indicators assessing state terror. Each measure consists of a rating scale that
ranges from 1 to 5, with 1 assigned to states with the
lowest, and 5 to states with the highest, levels of
personal integrity abuse. Raters use the following
categories:
1) Countries under a secure rule of law, people not
imprisoned for their views, torture is rare or
exceptional. . . . Political murders extremely rare.
2) A limited amount of imprisonment for nonviolent political activity. Few persons affected,
torture and beating exceptional. . . . Political
murders rare.
3) Extensive or recent history of extensive political
imprisonment. Execution or other political murders and brutality common. Unlimited detention
for political views accepted.
4) Practices of level 3 expanded to larger numbers.
Murders, disappearances are common, but terror
affects primarily those who interest themselves in
politics or ideas.
5) Terrors of level 4 expanded to the whole population. No limits on means or thoroughness with
which leaders pursue personal or ideological goals
(Gastil 1980; For examples and country reports
for each level see Gibney and Dalton 1996).
To match the availability of the constitutional provisions indicators we use, we use personal integrity
rights (or state terror) indexes for 1976–96.2
The Model
Measuring the Dependent Variable, State
Terror (Personal Integrity Abuse)
We begin the explication of our model by operationalizing the dependent variable, ‘‘state terror,’’ the
abuse of the right to personal integrity: political
imprisonment, torture, disappearances, and killings.
The indicators of state abuse of personal integrity we
2
Our dataset includes the global set of countries with a population
of 100,000 or more. Each measure contains missing data. In earlier
years, AI often did not rate countries with few human rights
problems; SD did not rate the United States or some of its close
allies. To compensate for the missing data, we followed Poe and
Tate (1994) and Poe, Tate, and Keith (1999) by substituting an SD
score for a missing AI score, and an AI score for a missing SD
score. Had we not made these substitutions, the coverage of the AI
and SD data sets would have been so different as to make their
comparison problematic.
648
linda camp keith, c. neal tate, and steven c. poe
Specifying a Model: Current Theory
The Effect of Norms
Properly assessing the impact of constitutional protections requires an adequately specified model. To
construct that model, we turn to the theory and
analysis presented in what Davenport and Armstrong
(2004) called the ‘‘standard’’ cross-national research
on the determinants of state terror or personal
integrity rights abuse (Poe and Tate 1994; Poe, Tate,
and Keith 1999), and to more recent research assessing the effects of international norms on state human
rights behavior.
Some international relations scholars argue that international norms drive state human rights behavior.
As more and more states commit to new rights,
others also feel increasing pressure to commit to
these norms (Finnemore and Sikkink 1998). A state
adopts international norms because accepts that
doing so is appropriate for legitimate statehood
(Meyer et al. 1997) or because of rewards or punishments it expects to reap or avoid by adopting such
norms (Simmons 2000).
The pressure to adhere to international norms
might represent an alternative explanation both for
why states adopt constitutional provisions intended
to protect human rights and why they refrain from
human rights abuse. If pressure from international
human rights norms drives state human rights
behaviors, evidence for the potential impact of constitutional provisions that does not take the effect of
adherence to norms into account could be spurious.
As Goodliffe and Hawkins note, ‘‘operationalizing norms is a tricky business . . . norms are difficult
to identify and their influence is hard to track’’
(2006, 361). Nevertheless, they argue that norms
lead states to commit to international rights treaties.
Relying on Simmons’ (2000) work, they create an
annual measure of the average commitment to the
United Nations Convention against Torture (UNCT)
at the level of both global and regional communities.
Their empirical examination finds evidence that
adherence to regional and global norms in support
of the UNCT increases a state’s likelihood of joining
the Convention early: adherence to existing global
and community norms affects individual national
formal commitments to refrain from torture.
To operationalize the effect of the pressure of
international norms, we follow Goodliffe and Hawkins
(2006) and Simmons (2000), measuring global and
regional commitments to a relevant international
treaty, the International Covenant on Civil and Political Rights (ICCPR).4 We operationalize the strength
of global and regional norms pressuring for commitment to the ICCPR using a 0–2 scale, where 0 is no
action, 1 is signature, and 2 is ratification or accession.
We then created an annual measure of the average
commitment scores of the global set of countries and
an annual measure of the average commitment scores
for each geographical region.
The ‘‘Standard’’ Model
Poe, Tate, and Keith (1999) identified eight independent variables that affected the extent to which a
regime abuses personal integrity. The theoretical
justification for these variables, their operationalizations, and their sources are discussed fully in their
published research (see Poe and Tate 1994; Poe, Tate,
and Keith1999). These eight measures include political democracy, economic development, civil war
experience, international war experience, population
size, British colonial experience, military control, and
state socialist regime.3
3
Since our focus is on formal-legal provisions that are intended to
protect the civil and political rights at the heart of constitutional
liberalism, we must be careful to operationalize democracy so that
it does not overlap too much with these rights. Some scholars
(Schmitter 1992; Zakaria 1997) distinguish liberalism from
democracy, arguing that the essence of democracy is rule of the
people, that constitutional liberalism ‘‘is not about the procedures
of selecting government, but . . . seeks to protect an individual’s
autonomy and dignity against coercion,’’ and pointing to the
rising number of democracies that do not practice constitutional
liberalism but rather are ‘‘illiberal democracies’’ (Zakaria 1997).
We agree with Zakaria. But even if we did not, this analysis would
be a pointless exercise if we did not take care to differentiate our
measures of democracy from our measures of constitutional
provisions. Examining measures of democracy that are available
for most of the countries and years we analyze, we conclude
that those in the Polity data sets allow us to best achieve the
conceptual separation we desire, because they focus on the
electoral process of the country’s executive. The specific measures
we use in this analysis are drawn from the Polity 98 data set
(Marshall and Jaggers 2001 for current version, Polity IV).
Statistically, it is clear that our individual constitutional protections measures and the Polity indicator of democracy are far from
identical. Simple correlations for the relationships between the
protections we discuss below and the Polity democracy scores
ranged from .12 to .33. Following Poe and Tate (1994), we
validate our results by employing a second measure of political
democracy, the Freedom House (FH) Political Rights Index that
allows us to analyze a maximum-sized set of countries. The
Political Rights Index is definitionally and empirically distinct
from our constitutional protections: the correlation coefficients
for the individual constitutional protections and political rights
ranged from .17 to .41.
4
Our analysis begins in 1976, so we cannot use the UNCT, which
did not go into effect until 1984. Simmons, with different
interests, measured commitment to the IMF treaty.
is the law a mere parchment barrier to human rights abuse
Constitutional Provisions
The sections of constitutions that are most relevant to
protecting citizens against serious repression fall into
three general categories: protections of individual
freedoms and rights, promotion of judicial independence, and regulations of arbitrary rule during states of
emergency.
Individual Freedoms and Rights
We expect two types of constitutional protections to
affect personal integrity abuse: (1) the individual
freedoms associated with constitutional liberalism:
freedom of speech, assembly, association, religion,
and press; and (2) the key due process protections
that directly relate to personal integrity abuse, the
rights to habeas corpus, to protection against arbitrary
arrest, the prohibition against torture or cruel and
unusual punishment, and the right to a fair and
public trial. Provisions for individual freedoms are
intended to allow people to openly criticize and
perhaps challenge a regime and the status quo (Poe
and Tate 1994, 123–24). Due process provisions
should protect against arbitrary government actions
against personal integrity. Observance of habeas
corpus rights would preclude disappearances and
limit political imprisonment. Providing a fair and
public trial makes it more difficult to convict people
for political reasons and gives the press and public
opinion a better chance to curb repressive behavior.
The ban on torture, when practiced, would directly
eliminate one form of personal integrity abuse and
decrease extrajudicial killings. These constitutional
protections recognize human rights modeled after the
universally accepted United Nations Declaration of
Human Rights (UNDHR) and the broadly accepted
ICCPR. We hypothesize that constitutional provisions
for both individual freedoms and due process rights will
reduce personal integrity abuse.
While providing for rights and freedoms, many
constitutions limit the very freedoms they promise to
protect, specifying that freedoms be exercised within
the boundaries of statutory law or linking bills of
rights with a list of duties that limit the rights. The
U.N. Commission on Human Rights holds that
clearly defining limitations on constitutional rights
promises is better than leaving the question of acceptable restrictions open to interpretation (Despouy
1985).
An alternative view argues that giving a regime
any opening to suspend constitutionally provided
rights and freedoms offers it a cover of legitimacy to
649
erode human rights beyond the acceptable suspensions specified in the constitution (see Frühling 1993;
Grossman 1990; International Commission of Jurists
[ICJ] 1983; and the empirical results of Davenport
1996 and Keith 1999). We are persuaded by this view
and hypothesize that specified limits on the exercise of
individual rights and freedoms will increase personal
integrity abuse.
We operationalize the extent to which constitutions provides for individual rights and freedoms by
coding the presence of 10 fundamental rights and
liberties in the text of national constitutions for each
year in the period 1976–96: freedoms of (1) speech,
(2) association, (3) assembly, (4) religion, and (5)
press, and rights to (6) strike, (7) the writ of habeas
corpus, (8) a public trial, (9) a fair trail, and (10) be
free from torture or cruel and usual punishment. We
coded each as (2) explicitly mentioned in the constitution, (1) explicitly mentioned but with (an)
exception(s) or qualification(s), such as a public interest
clause, or (0) not mentioned in the constitution.
The data sources for our constitutional variables
are Blaustein’s Constitutions of the World (1993) and
Flanz’s (1997) Constitutions of Countries of the World,
both updated annually.
An Independent Judiciary
Many commentators have stressed the potential
importance of an independent judiciary in protecting
human rights (see Ackermann 1989; Rosenthal 1990).
A truly independent judiciary should be able to
withstand incursions upon rights because (1) the
courts’ power and fiscal well-being are protected, (2)
the courts have some ability to review the actions of
other agencies of government, and (3) the judges’
jobs are constitutionally protected. Standards for
achieving an independent judiciary are offered in
United Nations (1985) and Cumaraswamy (1995).
We assessed the efforts of the world’s constitution makers to establish an independent judiciary by
coding the presence of eight indicators:
1) Guaranteed Terms. Constitution guarantees terms
of office and restricts removal of judges.
2) Decisions Final. Decisions of judges cannot be
revised except by legal appeals procedures.
3) Exclusive Authority. Courts have authority to
decide on their own competence.
4) No Exceptional or Military Courts. Courts have
jurisdiction over all issues of a judicial nature;
civilians are tried by ordinary courts, not military
or exceptional courts.
650
linda camp keith, c. neal tate, and steven c. poe
5) Fiscal Autonomy. The courts are fiscally autonomous. The salaries of judges and/or their annual
budgets are protected from reduction by the other
branches.
6) Separation of Powers. Courts are housed in a
separate branch from executive or legislature.
7) Enumerated Qualifications. Selection/advancement of judges based on merit qualifications.
8) Hierarchical system. Courts are structured in
multiple layers with the highest level court exercising final control/review of lower court decisions.
Authorities disagree as to whether judicial review
is a power that contributes to an independent
judiciary or whether an independent judiciary is a
necessary component for its real exercise (Blasi
and Cingranelli 1996 vs. Rosenthal 1990). Following Blasi and Cingranelli (1996), we include
judicial review among our judicial independence
indicators, since it has been strongly argued to be
a substantial factor in the protection of human
rights, particularly during periods of crisis or
instability (see Ackermann 1989; Chowdhury
1989; ICJ 1983), despite our intuition that it is a
consequence, rather than a component of judicial
independence:
9) Judicial Review. Courts exercise judicial/constitutional review of legislature and executive.
We hypothesize that constitutional guarantees of an
independent judiciary will increase respect for personal
integrity. We originally coded the presence in the
constitutions of each of these components promoting
judicial independence as 2 5 constitution provides for
it fully and explicitly, 1 5 constitution provides for it
somewhat or provides for it vaguely, but not fully, 0 5
constitution does not provide for it. In addition, ‘‘No
exceptional courts’’ (4 above) was coded 21 for
constitutions that specifically allowed civilians to be
tried in military courts or explicitly allowed the
formation of exceptional courts and ‘‘Judicial review’’
(8 above) was coded 21 for constitutions that gave
the power of constitutional review to another branch
of government.
Preliminary diagnostic data analyses of the properties of these variables led us to collapse codes 1 and
2 for each variable. This meant that the measures
used in this analysis for Guaranteed Terms, Decisions
Final, Exclusive Authority, Fiscal Autonomy, Separation of Powers, Enumerated Qualifications, and Hierarchical system, were coded as 0,1 dichotomies, while
No Exceptional or Military Courts and Judicial Review
were coded as –1,0,1 trichotomies.
States of Emergency
Analysts have tried hard to find appropriate legal
standards and safeguards to protect human rights
during the states of emergency that often result when
the existence or continuation in power of regimes are
(perceived to be) threatened (see Cea 1987; Chowdhury 1989; Frühling 1993; Grossman 1990). The ICJ
argues (1983) that constitutions should clearly specify
the conditions under which states of emergency may
be declared and exercised so that (1) regimes who
resort to emergency clauses in good faith will know in
advance the extent and limit of their emergency
powers, and (2) regimes not disposed to respect the
limits of their authority may be judged according to
the extent to which they live up to these specific rules.
The commission recommends that constitutions
specify that emergency measures may not affect rights
recognized as nonderogable in international law and
spell out the effects of states of emergency on the
rights of citizens and the powers of the branches of
government. The Commission recommends further
that the constitution enumerate and define situations
that justify departure from the normal legal order and
establish a procedure for declaring a state of emergency that (1) gives primary responsibility to the legislature, (2) specifies the duration of the emergency, and
(3) periodically reviews the need for its continuation
(1983, 32–34).
A quite similar set of recommendations for regulating states of emergency (International Law Association’s Paris Minimum Standards reported in
Chowdhury 1989) adds that constitutions should
require that states of emergency be extendable only
with the legislature’s approval, and that the legislature
may not be dissolved during the emergency.
Based on these international recommendations,
our indicators of constitutional regulation of states of
emergency or national crisis are:
1) Legislative Declaration. Procedure for declaring a
state of emergency gives primary responsibility to
the legislature: 2 5 legislative branch has explicit
responsibility for declaring the emergency, 1 5
executive-declared emergency must be explicitly
confirmed by legislature, 0 5 no mention of who
has the power to declare such an emergency, and
21 5 executive power to declare emergency is
explicit with no mention of a role for legislatures
or courts.
2) Cannot Dissolve Legislature. During an emergency,
2 5 the legislature may not be dissolved or ‘‘meets
by right’’; 1 5 there is a vague provision for
is the law a mere parchment barrier to human rights abuse
legislature meeting, but no prohibition against
dissolution; 0 5 there is no mention of dissolving
legislature; and 21 5 there is automatic suspension
or explicit executive power to suspend legislature.
3) Duration Limited. During an emergency, 2 5 the
duration of the emergency is limited; extension
requires legislative approval; 1 5 the duration is
limited/legislative approval of extension is required, but not both; 0 5 no mention of duration
or extension process.
4) Nonderogable Rights. For a state of emergency, 1 5
constitution lists nonderogable rights or states
that certain rights/freedoms cannot be revoked,
and 0 5 otherwise.
A logical hypothesis is that the more closely the
constitutional provisions for states of emergency follow
the international standards, the less likely the regime
will be to abuse personal integrity rights. However, just
as there have been skeptics who fear that specifying
the conditions under which individual rights and
freedoms can be derogated may promote human
rights abuse, so there are pessimists who worry that
specifying conditions regulating states of emergency
may be counterproductive. Grossman showed that,
while most Latin American constitutions have emergency clauses that supposedly apply in relatively
restricted circumstances, in periods of emergency
rule, rights were often broadly suspended; that the
constitutional emergency provisions provided a ‘‘rationalization for deprivation of human liberties’’
(1990, 188; see also Cea 1987, 665; Frühling 1993).
This research suggests a strong expectation for an
alternative hypothesis for ‘‘Duration limited’’ and
‘‘Nonderogable rights’’: that clauses intended to limit
human rights abuse during a crisis frequently may
have the opposite effect, and that a list of individual
freedoms and rights that may not be derogated
during an emergency may imply that some other
rights may justifiably be denied. Such clauses may
provide regimes with a cover of legitimacy for their
denial of human rights, even those that are nonderogable, and even at times outside of states of
emergency. Similarly, provisions that set a time limit
on and require renewals of states of emergency may
have the unintended consequence of encouraging
their extension.
Because provisions setting limits on states of emergencies and specifying nonderogable rights may be
expected either to decrease or to increase personal
integrity abuse, we use a two-tailed measure of statistical significance to test their effects. We strongly
expect the other provisions—legislative involvement
651
in the declaration of and nondissolution of the
legislature during a state of emergency—to decrease
repression and evaluate their effects using a onetailed test.
Constitutional Provisions and
State Terror: Estimating
Contemporaneous Effects
Our approach is to analyze—for as many nations as
possible across as long a time span as possible—the
relationships between our measures of formal constitutional provisions and important measures of
personal integrity rights abuse. To support the
analysis, we have built a pooled cross-sectional data
set that records (1) appropriate information about
constitutional provisions protecting rights, promoting judicial independence, and regulating states of
emergency, (2) indicators of repression of personal
integrity rights, and (3) other relevant phenomena
for each year in the period 1976–96 for 154 to 178
nation-states, depending upon the year. We use this
data set to build a statistical model to assess the
relationships implied by our research questions.
Results
We must estimate the appropriate coefficients for the
prediction of state terror (abuse of personal integrity
rights) from our sets of measures of individual
freedoms and rights, states of emergency, and judicial
independence, controlling for the variables that previous research indicates are necessary for a properly
specified equation. We use Ordinary Least Squares
regression.5 However, to cope with the unreliability
of statistical inference that can result from the heteroskedacity and autocorrelation that plague pooled
cross-sectional time series designs like ours, we follow
the recommendations of Beck and Katz (1995) and
employ panel-corrected (robust) standard errors to
correct for the former and a lagged dependent variable
5
Our dependent variable is ordinal. Why do we report OLS
analyses, rather than analyses carried out using a technique like
ordered logit or probit? We have several reasons. We discuss these
reasons based on extensive analyses in Appendix B (available on
request). Here we simply note that the similarity of ordered logit
results to OLS models, the impossibility of projecting ordered
logit coefficients over time in an interpretable fashion, an optimal
scoring transformations analysis of the ordinal state terror
measure (and analyses of the transformed dependent variables),
and especially the desirability of comparing our results with
previous comprehensive work all suggest that using OLS methods
is appropriate and proper.
652
to compensate for the latter. The inclusion of the
lagged dependent variable not only corrects a statistical
difficulty; it also has a theoretical justification because
it represents history: the tendency of regimes to use
past decisions to shape present decisions. We used
STATA 9.2 to perform the analyses.
Before including them in the equation, we
analyzed the multicollinearity among our independent variables. We found that four of the constitutional provisions were too highly correlated for all to
be included individually in the equation. To compensate for their multicollinearity, we combined the
freedoms of speech, association, assembly, and religion into an additive ‘‘Four Freedoms Index’’ (see
Appendix C, available on request, for multicollinearity details).
To enhance the robustness of our analysis and
maximize confidence in our results, we estimated our
basic prediction equation four times, once for each
combination of our two measures of state terror and
democracy.6 The coefficients produced for these four
equations are reported in Table 1. They show that
nine of the 23 constitutional provisions were associated with substantively and statistically significant
relationships in hypothesized directions in at least
one version of our prediction equation; eight were
statistically significant in two to four versions.
6
This is only one significant tactic we have used to examine the
robustness of our results. In addition to replicating the basic
model using four different specifications for operationalizations of
repression and democracy, we also conducted several other
analyses that demonstrate the robustness of our findings:
1) First, as noted above, we replicated the four basic model
specifications using ordered probit analysis, instead of OLS,
with no important differences in findings.
2) Second, we replicated the basic analyses using the composite
personal integrity abuse measure provided by Cingranelli and
Richards (2008) in their CIRI dataset. Because their measure is
only available beginning in 1980, this analysis reduced our
available N by some 20%. Nevertheless, the CIRI analysis
produced no significant differences in results.
3) Finally, we conducted several analyses replicating the basic
analysis but substituting several factor-analysis-derived composite indices for subgroups of constitutional rights/provisions. For example, we created ‘‘Four Freedoms’’ and ‘‘Five
Freedoms’’ indexes, three different ‘‘Judicial Independence’’
indexes, and two different indexes of ‘‘Emergency Rule’’
provisions in place of sets of individual independent variables
that were well-measured by the derived factors. This analysis is
itself much too elaborate and complicated to explore further
here, and, indeed, we have prepared for publication a manuscript that reports on it in another context (Tate and Keith
2008). Suffice it to say at this point, that the findings using the
composite indices give us no reason for believing that the basic
findings reported in Table 1 are anything less than robust.
linda camp keith, c. neal tate, and steven c. poe
Individual Freedoms and Rights
The Four Freedoms index produced small coefficients
(# 2.010) in the expected direction in all four models,
but none of the coefficients was statistically significant.
This lack of statistical and substantive significance
suggests that the four freedoms will be the foundation
of any set of rights adopted by a nation, so nearly
universal that they are standard symbols in the newly
created constitutions of the world, incapable of discriminating regime intentions or actions.
Two individual rights provisions, the requirements that trials be public and fair, have statistically
significant coefficients across all four equations. The
fair trial provision produces the largest coefficients:
2.067 and 2.075 in the two equations using SD State
Terror indicators and 2.108 and 2.116 in the
equation using AI measures. The maximum (2 point)
change from no constitutional protection to full
protection for fair trials produces a decrease of .14
to .24 in the 5-point state terror indexes. The
coefficients for the public trial provision are smaller
than those for fair trial, ranging from 2.042 to 2.078
across the four equations. A maximum (2 point)
change in public trial protections produces a decrease
of .08 to .16 in personal integrity abuse.
The provision for the right to strike produces
somewhat smaller coefficients (2.037 and 2.045) that
are significant in the two SD equations. The remaining
individual freedoms measures are not associated with
a statistically significant coefficient in any equation.
Some of the early analyses of the impact of constitutional provisions on various measures of human rights
abuse found it to be negative. Boli-Bennett (1976) and
Pritchard (1986) both found negative effects on rights,
and even Davenport’s analysis reports some relationships that are in the wrong direction, although, the
coefficients do not reach a sufficient level statistical
significance had he tested with no predicted direction
and a two-tailed test. This problematic finding surfaces
again with three individual rights measures, the ban
against torture, provision for the writ of habeas corpus,
and, to a lesser extent, freedom of the press. Each of
these measures produces one or more coefficients that
are signed in the unexpected direction, given our
hypotheses.7 Overall, these somewhat disappointing
results for the individual freedoms and rights parallel
7
Had we not stated our expectations and instead used two-tailed
tests of significance: freedom of press in the two AI State Terror
equations; torture in the two FH equations; habeas corpus in the SD/
Polity 98 equation, would have achieved significance (at p # .05).
is the law a mere parchment barrier to human rights abuse
T ABLE 1
653
Constitutional Provisions and Personal Integrity Abuse (State Terror): 1977–96
Regression Coefficients* (Standard Errors)
State Department State
Terror Measure
Freedom House
Democracy
Amnesty International State
Terror Measure
Polity 98
Democracy
Freedom House
Democracy
.620 (.044)a
.612 (.043)a
Lagged State Terror
.609 (.046)a
Individual Rights and Freedoms
Four Freedoms Index
2.010 (.009)
2.008 (.009)
2.004 (.007)
Freedom of Press
.029 (.021)
.021 (.021)
.063 (.022)
Right to Strike
2.037 (.025)c
2.045 (.028)b
.015 (.028)
Habeas Corpus
.027 (.014)
.037 (.015)
.033 (.022)
Public Trial
2.042 (.018)b
2.051 (.020)b
2.069 (.022)a
b
a
Fair Trial
2.067 (.021)
2.075 (.024)
2.108 (.028)a
Torture
.035 (.015)
.026 (.021)
.031 (.015)
Judicial Independence
Guaranteed Terms
.022 (.029)
.008 (.035)
.004 (.037)
Decisions Final
2.055 (.033)b
2.053 (.038)c
2.085 (.028)a
Exclusive Authority
2.055 (.037)c
2.062 (.040)c
2.020 (.035)
No Exceptional Courts
2.022 (.019)
2.025 (.017)c
2.060 (.025)a
Fiscal Autonomy
.090 (.033)
.088 (.038)
.120 (.031)
Separation of Powers
.029 (.042)
.036 (.047)
.028 (.032)
Enumerated Qualifications
.049 (.029)
.022 (.032)
.065 (.035)
Judicial Review
.036 (.026)
.021 (.031)
.076 (.027)
Hierarchical System
.018 (.035)
.020 (.032)
.013 (.027)
States of Emergency
Legislative Declaration
2.021 (.012)c
2.017 (.012)c
2.028 (.012)a
b
a
Limited Duration**
.038 (.014)
.055 (.016)
.037 (.013)a
a
a
Can’t Dissolve Legis.
2.041 (.017)
2.067 (.019)
2.023 (.020)
Non-Derogable Rights**
.098 (.037)a
.096 (.049)b
.067 (.034)
Control Variables
Civil War
.594 (.067)a
.587 (.075)a
.472 (.062)a
a
a
International War
.230 (.051)
.190 (.053)
.250 (.053)a
a
a
Democracy
2.093 (.011)
2.041 (.005)
2.092 (.011)a
b
b
Military Control
.084 (.029)
.068 (.034)
.053 (.037)b
State-Socialist Regime
2.020 (.041)
.022 (.042)
2.097 (.043)a
British Col. Exper.
2.001 (.033)
2.024 (.037)
.033 (.035)
Economic Development
2.013 (.002)a
2.013 (.002)a
2.012 (.003)a
Population
.063 (.009)a
.062 (.009)a
.069 (.010)a
Global Norms
.205 (.085)
.317 (.093)
2.052 (.077)
Regional Norms
.113 (.029)
.082 (034)
.100 (.033)
Constant**
.137 (.120)
2.401 (.121)a
.092 (.137)
R2
.78
.77
.76
Chi2 p . .00001 for each model. Freedom House N 5 2929. Polity 98 N 5 2554.
Polity 98
Democracy
.611 (.044)a
2.004
.063
.009
.037
2.078
2.116
.025
(.009)
(.024)
(.030)
(.023)
(.022)a
(.034)a
(.016)
2.028
2.084
2.008
2.071
.125
.036
.058
.076
.010
(.045)
(.028)a
(.037)
(.024)a
(.035)
(.037)
(.037)
(.026)
(.031)
2.026
.051
2.053
.067
(.011)a
(.017)a
(.026)b
(.043)
.476 (.066)a
.215 (.057)a
2.045 (.006)a
.027 (.038)
2.065 (.050)
.005 (.034)
2.011 (.003)a
.076 (.012)a
.071 (.088)
.066 (.039)
22.80 (.123)
.73
*Main entries are unstandardized OLS coefficients. Robust standard errors in parentheses.
a
p , .01, bp , .05, cp , .10 **two-tailed test of significance for these variables.
those of earlier analyses: Davenport (1996) found only
three of 14 provisions to have an impact on state
repression, and both Boli-Bennett (1976) and Pritchard (1986) found constitutional provisions to have the
opposite effect of what was predicted.
An Independent Judiciary
Three of the judicial independence provisions produce statistically significant coefficients in at least two
of the equations. The provision that makes judges’
654
linda camp keith, c. neal tate, and steven c. poe
decisions final produces statistically significant coefficients in all four models (2.055 and 2.053 in the
SD equations and 2.085 and 2.084 in the AI equations). The coefficients for the ban against exceptional
and military courts are statistically significant in both
AI equations (2.060 and 2.071) and in one SD equation (2.025). The provision that gives judges exclusive
authority to decide on their own competence produces
coefficients (2.055 and 2.062) that are statistically
significant (at p # .10) in the two SD equations. A
maximum change from a constitution that explicitly
allows exceptional courts to a constitution that specifies
the ban against these courts would decrease the predicted level of personal integrity abuse by 2.044 or
2.142, depending on the equation selected. None of
the remaining provisions for judicial independence are
statistically significant in any equation and six of the
nine are not signed in the expected direction.8
That judicial review not only fails to demonstrate
a negative impact but in fact exhibits a positive
relationship with repression in two equations is
contra-hypothesis. This ‘‘nonfinding’’ may indicate
that judicial review should not be regarded as a
component of judicial independence. It may also
support the view that judges are political actors who
possess the same values and pursue the same public
goals as other state actors. The circumstances that
lead regimes to repress may lead judges to sanction
repressive behavior, especially in issues of national
security. Even if judges do not wish to support crisis
ruler(s), the situations they face and the options
available to the rulers may make it very difficult for
them to act effectively to protect rights (Tate 1993).
Judicial review and independence are typically weakened during states of emergency, perhaps by constitutional provisions that limit their power, or perhaps,
as Chowdhury (1989, 131) suggests, by the exercise of
judicial restraint or adoption of a hands-off policy.
p # .10) in the SD equations. Banning dissolution of
the legislature during a state of emergency is significant in three of the four equations with values of
2.041, 2.067 (both at p # .01), and 2.053 (p # .05).
The reader will recall that there is controversy
over the effects of the other two state of emergency
provisions whose impacts we assess. Proponents
contend that specifying a list of rights that cannot
be derogated during the state of emergency and limits
on its duration and renewal will protect fundamental
rights. Critics argue that such provisions will only
encourage state of emergency rulers to abuse rights.
Because of this disagreement, we evaluated the
impact of the coefficients for these two variables
using a two-tailed test of significance.
Setting conditions for extending the duration of
states of emergency produces positive coefficients with
values ranging from .037 to .055 that are statistically
significant (p # .01 in three cases and p # .05 in one
cases) in all four personal integrity models. This
suggests that time period clauses encourage the regime
to renew states of emergency. Specifying a list of
nonderogable rights produces statistically significant
positive coefficients in both SD prediction equations
with values of .098 (p # .01) and .096 (p # .05). These
results support the alternative hypothesis that a constitutional provision for the derogation of any right,
may in fact lead to the abuse of other rights (including
the right to life) that may not legally be derogated.
States of Emergency
In contrast to individual rights and freedoms and
judicial independence, the coefficients for the four
constitutional provisions related to states of emergencies are statistically and substantively significant
in most of the prediction equations. Requiring
legislative declaration of states of emergency is statistically significant in all four equations with values of
2.028 and .026 (both at p # .01) in the Amnesty
equations and values of 2.021 and .017 (both at
8
Had we not stated our expectations and instead used two-tailed
tests (at p # .05), fiscal autonomy would have been significant in
all equations and judicial review in the AI equations.
The Standard Model
The ‘‘standard model’’ of state terror works as expected in our data: eight of the nine control variables
perform in our equations as they have in previous
analyses of personal integrity abuse; thus we shall not
discuss them here. The exception is British colonial
experience. When controlling for our constitutional
provisions, British colonial experience no longer exhibits any statistically significant effects. The impact it had
in previous studies was likely due to the British influence on constitution writing, as suggested by scholars
who have written about colonial heritages (see Moderne
1990, 322–27, for example). While coding the constitutional provisions, it was easy for us to observe a basic
template upon which former British colonies, as newly
independent states, based their constitutions.
Norms
Only one of the eight coefficients for the two norms
variables is signed in the right direction: global norms
in the AI/FH equation. It does not achieve an
is the law a mere parchment barrier to human rights abuse
acceptable level of statistical significance. All the
other coefficients are signed in an unexpected direction. The coefficients for regional norms would have
achieved statistical significance (at p # .05) in all four
equations had we used two-tailed tests of statistical
significance, as the global norms coefficients would
have in the SD equations. Whatever the effects of
norms in other analytical contexts, they do not
contribute to the reduction of personal integrity
abuse, controlling for the effects of the standard
model and for constitutional provisions.
Lagged Dependent Variable
Not surprisingly, the lagged dependent variable had a
powerful influence. A 1-point increase in last year’s
state terror index is associated with a .609 to .620
increase in this year’s predicted record of personal
integrity abuse, depending on the measures of state
terror and democracy included in the equations. That
repression tends to beget repression is not surprising.
We shall return to the issue of the influence of history
after summarizing the contemporaneous effects on
state terror depicted in our prediction equations.
rationale for combining the statistically significant
constitutional provisions in this analysis might be
understood through a thought experiment in which
groups of constitution makers adopt at a particular
point in time bundles of constitutional changes that,
intelligently or fortuitously, include the provisions
that we have found to have significant desirable
effects in decreasing integrity rights abuse. Through
foresight or luck they also happen not to adopt any
provisions that might increase rights abuse (see the
‘‘Caveats’’ discussion below) or else adopt other
provisions that have no effect, either positive or
negative, on rights abuse.
Combining and comparing effects for the statistically significant constitutional provisions, control
variables, and lagged dependent variable reveals that:
d
Assessing the Contemporaneous Effects
Nine of the 23 constitutional provisions we included in our equations had effects that were statistically too large to have occurred by chance. But it is not
the statistical, but the substantive, significance of these
effects that matters. The substantive questions are ‘‘just
how much can we expect a state’s personal integrity
repression score to increase or decrease in a specific
year, given one unit changes in its scores on particular
constitutional provisions?’’ The answers, given by the
regression coefficients, are not immediately obvious,
given the differences in the scales used to represent the
constitutional provisions and control variables.
Is an expected decrease of 2.067 to 2.116 as a
result of adding a qualified fair trial provision to a
constitution a little, or a lot? A change of .067 to .116
is only about a 1 to 2% change on a scale that runs 1
to 5. One might be tempted to say that it is only a
little, especially compared to the lagged dependent
variable or the major variables in the standard model.
But one must remember that the changes represented
in Table 1 are net changes that occur after all the
other variables in the equation have had their effects.
One should also remember that the constitutional
provisions are likely to be adopted in bundles, not
one at a time, so that a more informative way to
examine their effects might be to combine them. The
655
d
Focusing on the combined constitutional provisions that had a statistically significant effect in
one or more of the prediction equations, a one
unit change in the direction that would decrease
state terror on each constitutional provision
would lead to an expected contemporaneous
decrease in the state terror index of between
2.315 to 2.546, from about three-eighths to just
over one-half a point, seven to 11%. A similar
change on each of the seven variables in the
standard model would lead to an expected
contemporaneous decrease in the state terror
index of 2.823 to 21.077, or from about ninetenths to about 1 point, 18 to 21%. Finally, a
one-unit decrease in state terror last year would
lead to an expected decrease in current state
terror of 2.609 to 2.620, from about six-tenths
to five-eighths of a point, 12 to 13%.9
The combined constitutional provisions have less
important effects on the repression of personal
integrity rights than the combined independent
variables or the lagged dependent variable included in the standard model. Do they represent
useful additions to this model? For perspective,
we ask ourselves: would not most analysts and
policy makers agree that a one-year change of 7
to 11% in population, GNP per capita, or
democracy was important? We think so, and
we think the same can be said for a 7 to 11%
increase or decrease in the state terror index
associated with the constitutional provisions.
But this is not the end of our story of the effects of
constitutional provisions on repression. The process
9
These numbers are ranges of the sums of the values of the
coefficients for the variables referred to.
656
linda camp keith, c. neal tate, and steven c. poe
that our equations model is not one that runs out in a
single year. Because this year’s scores are powerfully
affected by last year’s scores, and because last year’s
scores were in turn affected by the previous year’s
constitutional provisions and standard model variables scores, the process that our equations depict is
dynamic: it can take years for the full effects of
changes that occur this year to work themselves out.
Consequently, we need now to turn to an analysis of
the long-term effects of changes in constitutional
provisions to assess their full importance.
ten-year period (Yeart to Yeart+10) of substantial
changes in each of the independent variables that
achieved statistical significance in one or more of our
prediction equations. Figure 1 plots the effects of
these substantial changes10 for the combined constitutional provisions and the combined standard
model variables that achieved statistical significance
in the AI/Polity 98 equation.11
Before tackling Figure 1, readers need to understand two assumptions that governed its construction. First, we defined the ‘‘substantial’’ changes it
models as maximum changes in the independent
variables from yeart21 to yeart, except for Economic
Development (measured as GNP per capita) and
Population. A substantial change in Civil or International War was a change from being involved in a
civil or international war, to being at peace; a change
in Democracy was a change from ‘‘least democracy’’
to ‘‘most democracy,’’ a change in Cannot Dissolve
Legislature was a change from a constitution without
a provision to prohibit the executive from dissolving
the legislature during a state of emergency, to a
constitution that has such a provision; and so on.
That such changes can occur in a single year seems
entirely plausible to us.
The same logic does not seem to apply to
changes in GNP per capita and Population. If we
restrict ourselves to the real world ranges they exhibit,
it does not seem plausible to assume that any country
could change from being the poorest to the richest, or
from being the least populous to the most populous
nation in a single year. Consequently, we assume that
a substantial change in GNP per capita is $20,000
dollars, and a substantial change in population is a
similarly large change of 10,000,000 people (see Poe
and Tate 1994; Poe, Tate, and Keith 1999).
Second, in calculating Figure 1 and to keep the
graph simple, we summed the statistically significant
effects of the combined constitutional provisions
(Public Trial, Fair Trial, Decisions Final, No Exceptional Courts, Legislative Declaration, Limited Duration, Can’t Dissolve Legislature) into a single measure,
as we did when evaluating their contemporaneous
effects. We did so because it allows us to focus on the
forest, rather than the trees; to assess the overall
impact of efforts to write constitutions that will
protect human rights and to compare them with
Constitutional Change and
State Terror in the Long Term:
Dynamic Effects
Because our prediction equations include a lagged
dependent variable, the complete effects of constitutional provisions and standard model variables on
state terror are not captured by the regression
coefficients for their contemporaneous effects. Indeed, the effects of these independent variables
actually continue for several years before they decline,
asymptotically, to zero. The long term cumulative
effect of a specific independent variable can be
calculated using:
Total cumulative effect estimate for a change in
variablei in Yeart 5 (Contemporaneous effect estimate for variablei 3 Change in variablei at Yeart) +
(Effect estimate for lagged dependent variable x Total
cumulative effect estimate for Yeart21).
For example, in the equation in the rightmost
column of Table 1 (using the AI version of the State
Terror index and the Polity measure of democracy),
d
d
the effect after one year of adopting a constitutional provision making judges’ decisions final,
when no such provision existed before (a one
unit change), is predicted to be 2.084 [the
contemporaneous effect of the variable] + (.611
[the coefficient for the lagged dependent variable]
3 2.084 [the contemporaneous effect of the
variable]) or 2.135;
the effect after two years would be 2.084 [the
contemporaneous effect of the variable] + (.611
[the coefficient for the lagged dependent variable] 3 2.135 [the effect of the variable after one
year]) or 2.167, and so on.
To illustrate the long-term potential impacts of
changes in the constitutional provisions and control
variables, we calculated the cumulative effects over a
10
We have arranged the measurement scales of the variables
plotted in Figure 1 to show the effect of each independent variable
in decreasing State Terror.
11
To save space, we report the plot for this single equation. Ten-year
effects of all the statistically significant variables and plots for all
four equations are in Appendix C. All plots show similar effects.
is the law a mere parchment barrier to human rights abuse
F IGURE 1 Overtime Effects on State Terror
(Amnesty International and Polity
Measures)
Overtime Effects on State Terror
−1
−2
−3
−4
−5
Decrease in State Terror
(Amnesty International and Polity Measures)
0
1
2
3
4
5
6
7
8
9
10
Time
Combined Cotrol Variables
Combined Constitutional Provisions
the collective effects of other important variables. The
fact is that nations typically do not adopt one or
another of these constitutional provisions in isolation
from others. They adopt them in clusters or bundles.
When we bundle them for purposes of assessing their
long term effects, we are reflecting reality. For
effective comparison, we also combined the effects
of the significant standard model variables (Civil
War, International War, Democracy, GNP pc
[1000s], logged Population).
One conclusion emerges forcefully from Figure 1:
at least in the abstract, constitutional provisions may
represent mere parchment barriers, but they are
anything but paper tigers in their long term potential
to reduce repression of state terror. After 10 years, the
projected effect of adopting all the constitutional
provisions modeled in Figure 1 would be to decrease
a nation’s repression score by 2.5 points on the 5point scale.12 Such a change could move a nation
from being one in which state terror affects the whole
population to one in which political imprisonment
affects only a few persons and torture and beatings
are rare, or from one in which political imprisonment
affects large numbers of the politically active population and political murders may be common, to
one where the rule of law is secure.
12
Our dependent variable has a discrete range of 1–5, so it is
possible for the sum of projections based on OLS results to fall
outside this range, especially when one is using maximum
independent variable scores to calculate the projections, as we
are. While out of range projections are unrealistic, the alternatives
to OLS are poor choices for this analysis, as we have already noted.
A proper response to this problem is to use the predicted scores
for purposes of illustrating and comparing the effects of the
independent variables, but not to place great confidence in precise
scores predicted for individual cases.
657
Some examples using real data may make the
meaning of the projections in Figure 1 more understandable. Uganda and Chad made major constitutional changes in 1996; Nicaragua did the same in
1987. In these years, each country was rated as a ‘‘4,’’
the next to most repressive category, on the basis of its
profile in the AI reports. Among the multiple constitutional provisions Uganda and Chad adopted in
1996 were provisions for the exclusive authority of
judges and for legislative declaration, limited duration
of states of emergency, and a list of nonderogable
rights. Uganda also adopted provisions for public and
fair trials. Chad also adopted provisions guaranteeing
freedom to strike and providing that judicial decisions
were final. In 1987, Nicaragua adopted constitutional
language providing for the right to strike, public trials,
the exclusive authority of judges, making judicial
decisions final, requiring legislative declarations of
states of emergency, and listing nonderogable rights.
Nine years after Uganda and Chad adopted the
specified new constitutional provisions (in 2005, the
last year for which we have data), the coefficients
underlying the maximum changes projections in
Figure 1 predict decreases in repression levels of
21.87 for Uganda and 21.21 for Chad, other things
being equal. In fact, Uganda, which continued to be
plagued by its twenty year war in the north, did not
see the predicted improved human rights score, but
Chad’s score decreased from ‘‘4’’ to ‘‘2,’’ a bit more
than predicted. A projection of the effect of Nicaragua’s adopted constitutional provisions across 10
years (to 1996) predicts it would decrease its state
terror level by 21.19. In fact, Nicaragua’s score
decreased from ‘‘4’’ to ‘‘3,’’ about as predicted.
Caveat: The Contra-Hypothesis
‘‘Relationships’’
Following classical practices of statistical testing, we
have, as existing theory or research allowed, postulated directional hypotheses that required the use of
one-tailed tests of statistical significance. In two cases
existing evidence could be marshaled to support
either positive or negative hypotheses, so we did
not predict direction and used two-tailed tests of
significance for the effects of duration limits on states
of emergency and specifications nonderogable rights
during states of emergency. For all our other constitutional provisions we expected that (more of) a
specific constitutional provision would decrease state
terror and thus produce negative relationships.
658
linda camp keith, c. neal tate, and steven c. poe
A necessary implication of using directional hypotheses and one-tailed tests of significance is that any
contra-hypothesis relationship cannot be statistically
significant: in the present case, testing the significance
of a relationship that is supposed to be negative
automatically assigns the positive half of the relevant
sampling distribution (the t distribution in this case)—
and thus any positive relationship, no matter how
strong—to the realm of nonsignificant findings.
Nevertheless, it can be disconcerting to find
contra-hypothesis relationships sufficiently strong
that they would have been statistically significant,
had we begun from a position of ignorance that did
not predict the expected direction of relationships for
our constitutional provisions and, consequently, used
two-tailed tests. As we have noted above, these contrahypotheses relationships are not out of sync with some
of the earlier research. Table 1 contains such contrahypothesis relationships for several provisions. Among
the Individual Rights and Freedoms indicators, freedom of press would have been positively related to
repression in both AI equations, torture in both FH
equations, and habeas corpus in the SD/Polity 98
equation, had we not posited an expected negative
direction for their relationships in advance. Even in
hindsight, it is hard to imagine why freedom of the
press, the prohibition of torture, or the guarantee of
habeas corpus might exhibit ‘‘significant’’ positive
relationships with personal integrity abuse.
Among the judicial independence indicators, the
provisions establishing fiscal autonomy for the judiciary
would have been statistically significant in all four
equations and the provision of judicial review in the
two AI equations. We have already stated our intuition
that judicial review may not an appropriate measure of
judicial independence. We speculated that judicial review might demonstrate its unhypothesized positive
relationship with state terror because it merely serves as
a device through which judges who agree with leaders
can strengthen the leader’s position by endorsing their
actions when they are challenged in courts. The relationship for fiscal autonomy fits in with the nonsignificant
but positive relationships between provisions establishing an organizationally independent judiciary and
repression, discussed in the next paragraphs.
For judicial independence, the provisions that
relate to repression in the expected negative direction
(decisions final, exclusive authority, no exceptional
courts) all protect the legal decision making authority
of regular courts against challenges from other
decision makers. In contrast, the six constitutional
measures that related in the unexpected positive
direction (guaranteed terms, fiscal autonomy, separa-
tion of powers, enumerated qualifications, judicial review, and hierarchical system), except for judicial
review, are largely concerned with the organization
of the courts and qualifications of the judges.13
Hindsight speculation might suggest that these structural features, nominally designed to promote judicial independence, serve as window dressing (truly as
parchment barriers) that allow regimes to claim they
are establishing independent judiciaries without
really doing so. Or, alternatively, perhaps the effort
to establish an organizationally professional judiciary
actually produces one that is more fully integrated with
the bureaucracy and executive, and, thereby, less likely
to challenge the actions of its executive counterparts,
even when those become repressive. These speculations,
indeed the whole set of unexpected relationships we
have encountered, bear careful examination in future
research specifically designed to test them.
Conclusion
We have sought in this research to provide sound
scientific evidence that might extend our understanding of why, in many nations, people are imprisoned, tortured, killed, or made to disappear
arbitrarily or because of their political beliefs and
actions. As good scientists, we seek empirically
grounded knowledge that will help us better understand and explain our social and political world. But
we also believe our scientific findings may have
implications for improving global protection of
human rights. Our findings about the effects of the
law—manifested in constitutional provisions designed to protect individual rights and freedoms, to
secure an independent judiciary, and to limit or
regulate executive actions in states of emergency—
may be instructive to policymakers who seek to
promote human rights globally. To put it simply:
from a practical point of view, pursuing better human
rights through constitutional law making may offer
potential for improvement in respect for human rights
that would be difficult to accomplish through attempting to affect other, less manipulable factors.
Certainly this and other research document that
state terror is a product of multiple factors, not just
of constitutional provisions. Avoiding involvement in
domestic or international war, eschewing military
government, or reducing a large population will
decrease personal integrity abuse in the long run.
So will practicing more democracy and becoming
13
With the possible exception of guaranteed terms.
is the law a mere parchment barrier to human rights abuse
wealthier. Unfortunately, as we have argued, while
policy makers might wish fervently to produce a
$20,000 one-year increase in their state’s per capita
GNP or a one-year ten-million person reduction in
its population, these are rather difficult tasks14 The
same may be true for eschewing civil or international
war and military government: avoiding these experiences may be easier said than done. Finally, the
exhortation ‘‘Practice democracy!’’ might well be
answered with the reply ‘‘Yes, but how?’’ Compared
to writing constitutional provisions into law, achieving these goals in the short term is very difficult.
Readers will recognize many limitations to this
analysis. Indeed, if we now answer Madison’s question, can nations safely ‘‘trust to these parchment
barriers against the encroaching spirit of power?’’
(Madison 1788a), our answer must be ‘‘Safely? No.’’
Nevertheless, at the very least, our research clearly has
shown that attempting to protect against abuse of
personal integrity by adopting constitutional protections is ‘‘neither improper nor altogether useless’’
(Madison 1789).
Manuscript submitted 30 January 2006
Manuscript accepted for publication 1 August 2008
References
Ackermann, L. W. H.1989. ‘‘Constitutional Protection of Human
Rights: Judicial Review.’’ Columbia Human Rights Law Review
21: 59–71.
Beck, Nathaniel, and Jonathan Katz. 1995. ‘‘What to Do (and
What Not to Do) with Time-Series-Cross-Sectional Data in
Comparative Politics.’’ American Political Science Review 89:
634–47.
Blasi, Gerard J., and David L. Cingranelli, 1996. ‘‘Do Constitutions and Institutions Help Protect Human Rights?’’ In
Human Rights and Developing Countries, ed. David L. Cingranelli. Greenwich, CT: JAI, 223–37.
Blaustein, Albert P. 1993. Constitutions of the World. Nashville,
TN: Washburn Law School.
Boli-Bennett, John. 1976. ‘‘The Expansion of National States.
1870–1970.’’ Ph.D. dissertation. Stanford University.
Cea, Jose Luis. 1987. ‘‘Chile’s Difficult Return to Constitutional
Democracy.’’ PS 20: 665–73.
Chowdhury, Subrata Roy. 1989. Rule of Law in a State of
Emergency. New York: St. Martin’s.
Cingranelli, David L., and David L. Richards. 2008. CIRI Human
Rights Data Project. http://ciri.binghamton.edu/ accessed May
18, 2008.
Cumaraswamy, Dato’ Param. 1995. Report of the Special
Rapporteur on the independence of judges and lawyers.
Pursuant to Commission on Human Rights resolution 1995/36.
14
The latter might also be accomplished through genocide, which
hardly reduces state terror.
659
Davenport, Christian, and David A. Armstrong II. 2004. ‘‘Democracy and the Violation of Human Rights: A Statistical
Analysis from 1976 to 1996.’’ American Journal of Political
Science 48 (3): 538–54.
Davenport, Christian. 1996. ‘‘‘Constitutional Promises’ and
Repressive Reality: A Cross-National Time-Series Investigation of Why Political and Civil Liberties Are Suppressed.’’
Journal of Politics 58: 627–54.
Despouy, Leandro. 1985. ‘‘The Administration of Justice and the
Human Rights of Detainees: Question of Human Rights and
States of Emergency.’’ New York: U.N. Economic and Social
Council.
Elster, Jon, 1993. ‘‘Introduction.’’ In Constitutionalism and
Democracy, ed. Jon Elster and Rune Slagstad. Cambridge:
Cambridge University Press, 1–18.
Epp, Charles R. 1998. The Rights Revolution: Lawyers, Activists,
and Supreme Courts in Comparative Perspective. Chicago:
University of Chicago Press
Finer, S. E., Vernon Bogdanor, and Bernard Rudden. 1995.
Comparing Constitutions. Oxford: Oxford University Press.
Finnemore, Martha, and Kathryn Sikkink. 1998. ‘‘International
Norms Dynamics and Political Change.’’ International Organization 52: 887–917.
Flanz, Gisbert H. 1997. Constitutions of Countries of the World.
New York: Oceana Publications.
Frühling, Hugo E. 1993. ‘‘Human Rights in Constitutional Order and
in Political Practice in Latin America.’’ In Constitutionalism and
Democracy: Transition in the Contemporary World, ed. Douglas
Greenberg et al. New York: Oxford University Press, 85–104.
Gastil, Raymond D, 1980. Freedom in the World: Political Rights
and Civil Liberties 1980. New Brunswick, NJ: Transaction Books.
Gibney, Mark, and Matthew Dalton. 1996. ‘‘The Political Terror
Scale.’’ Policy Studies and Developing Nations 4: 73–84.
Gibney, Mark, and Michael Stohl. 1988. ‘‘Human Rights and U.S.
Refugee Policy.’’ In Open Border? Closed Societies? The Ethical
and Political Issues, ed. Mark Gibney. Westport, CT:
Greenwood Press.
Gibney, Mark. 2001. ‘‘Political Terror Scale 1980–1997.’’ http://
www.unca.edu/~mgibney/.
Ginsburg, Thomas. 2003. Judicial Review in New Democracies:
Constitutional Courts in Asian Cases. Cambridge: Cambridge
University Press.
Goodliffe, Jay, and Darren G. Hawkins. 2006. ‘‘Explaining
Commitment: States and the Convention against Torture.’’
Journal of Politics 68: 358–71.
Grossman, Claudio. 1990. ‘‘States of Emergency: Latin America
and the United States.’’ In Constitutionalism and Rights: the
Influence of the United States Constitution Abroad, ed. Louis
Henkin and Albert J. Rosenthal. New York: Columbia University Press, 176–99.
Howard, A. E. Dick. 1991. ‘‘The Essence of Constitutionalism.’’
In Constitutionalism and Human Rights: America, Poland, and
France: A Bicentennial Colloquium at the Miller Center 3, ed.
Kenneth W. Thompson and Rett R. Ludwikowski Lanham,
MD: University Press of America, 18–30.
International Commission of Jurists (ICJ), 1983. States of
Emergency: Their Impact on Human Rights. Geneva: International Commission of Jurists.
Keith, Linda Camp. 1999. ‘‘The United Nations International
Covenant on Civil and Political Rights: Does It Make a
Difference in Human Rights Behavior?’’ Journal of Peace
Research 36 (1): 95–118.
660
linda camp keith, c. neal tate, and steven c. poe
Keith, Linda Camp. 2002. ‘‘Constitutional Provisions for Individual Human Rights (1976–1996): Are They More than Mere
‘Window Dressing?’’’ Political Research Quarterly 55 (March):
111–43.
Ludwikowski, Rett R. 1996. Constitution-Making in the Region of
Former Soviet Dominance. Durham, NC: Duke University Press.
Madison, James. 1788a. ‘‘James Madison Proposes the Bill of
Rights to the House of Representatives, June 8, 1789.’’ In ed.
Devin Bent. http://www.jmu.edu/madison/madprobll.htm#useless.
Madison, James. 1788b. ‘‘James Madison to Thomas Jefferson 17
October 1788.’’ In The Founder’s Constitution, ed. Philip B.
Kurland and Ralph Lerner. http://press-pubs.uchicago.edu/
founders/documents/v1ch14s47.html.
Madison, James. 1789. ‘‘Federalist No. 48.’’ In The Founder’s
Constitution, ed. Philip B. Kurland and Ralph Lerner. http: //
press-pubs.uchicago.edu/founders/documents/v1ch10s15.html.
Marshall, Monty G., and Keith Jaggers. 2001. Polity IV Project:
Political Regime Characteristics and Transitions, 1800–1999. University of Maryland. http: //www.bsos.umd.edu/cidcm/polity/.
Meyer, John W., John Boli, George M. Thomas, and Francisco O.
Ramirez. 1997. ‘‘World Society and the Nation-State.’’ American Journal of Sociology: 103: 144–81.
Moderne, Franck. 1990. ‘‘Human Rights and Postcolonial Constitutions in Sub-Saharan Africa.’’ In Constitutionalism and
Rights: The Influence of the United States Constitution Abroad,
ed. Louis Henkin and Albert J. Rosenthal. New York:
Columbia University Press, 315–48.
Poe, Steven C. 1992. ‘‘Human Rights and Economic Assistance
under Ronald Reagan and Jimmy Carter.’’ American Journal of
Political Science 36: 147–67.
Poe, Steven C., and C. Neal Tate, 1994. ‘‘Repression of Human
Rights to Personal Integrity in the 1980s: A Global Analysis.’’
American Political Science Review 88: 853–72.
Poe, Steven C., C. Neal Tate, and Linda Camp Keith, 1999.
‘‘Repression of Human Rights to Personal Integrity Revisited:
A Global Cross-National Study Covering the Years 1976–
1993,’’ International Studies Quarterly 43 (1): 291–313.
Pritchard, Kathleen. 1986. ‘‘Comparative Human Rights: An
Integrative Explanation.’’ Policy Studies Journal 15 (1): 110–28.
Rosenthal, Albert J. 1990. ‘‘Afterword.’’ In Constitutionalism and
Rights: The Influence of the United States, ed. Louis Henkin
and Albert J. Rosenthal. New York: Columbia University
Press, 397–404.
Schmitter, Philippe C. 1992. ‘‘The Irony of Modern Democracy
and Efforts to Improve Its Practice.’’ Politics and Society 20
(4): 507–12.
Simmons, Beth A. 2000. ‘‘International Law and State Behavior:
Commitment and Compliance in International Monetary
Affairs.’’ American Political Science Review 94: 819–36.
Talbott, Strobe. 1998. ‘‘Remarks and Press Q and A on 1997
Country Report on Human Rights.’’ Office of the Spokesman,
Department of State Washington, DC.
Taylor, Charles Lewis, and David Jodice. 1983. World Handbook
of Political and Social Indicators III. New Haven, CT: Yale
University Press.
Tate, C. Neal. 1993. ‘‘Courts and Crisis Regimes: A Theory Sketch
with Asian Case Studies.’’ Political Research Quarterly 46 (2):
311–38.
Tate, C. Neal, and Linda Camp Keith. 2008. ‘‘Conceptualizing
and Operationalizing Judicial Independence Globally.’’ Currently submitted for publication.
Thompson, Kenneth W. and Rett R. Ludwikowski (Eds.). 1991.
Constitutionalism and Human Rights: America, Poland, and
France. Lanham, MD: University Press of America.
Teitel, Ruti G. 2000.Transitional Justice. Oxford: Oxford University Press.
United Nations High Commissioner for Human Rights. 1985.
Basic Principles on the Independence of the Judiciary. New
York: United Nations.
Waldron, Jeremy. 1998. ‘‘Judicial Review and the Conditions of
Democracy.’’ The Journal of Political Philosophy 6 (4): 335–55.
Zakaria, Fareed. 1997. ‘‘The Rise of Illiberal Democracy.’’ Foreign
Affairs 76 (6): 22–43.
Linda Camp Keith is assistant professor of political science, The University of Texas at Dallas,
Richardson, TX 75083.
C. Neal Tate is professor of political science and
law, Vanderbilt University, Nashville, TN 37235.
Steve C. Poe (deceased) was professor of political science, University of North Texas, Denton, TX
76203.