_)PS- 22 3
POLICY RESEARCH WORKING PAPER   228 3
New  Tools and New  Tests                                              Some say that democracy ,s
more likely to survive uncei-
in Comparative Political                                               parliamentary governments.
Economy                                                                That result is not robus. to cf-e
use of different variables fror-
the Database of PoIiticai
The  Database  of Political Institutions                               Institutions, a large new
cross-country database :nat
may illuminate many ousher
Thorsteni Beck                                                         issues affecting and affect-(,
George Clarke                                                          by political institutions
Alberto Groff
Philip Keefer
Patrick Walsh
The World Bank
Development Research Group
Regulation and Competition Policy                                               U
February 2000



POLICY RESEARCH WORKING PAPER 2283
Summary findings
This paper introduces a large new cross-country database         Stepan and Skach, for example, find that democracy is
on political institutions: the Database on Political          more likely to survive under parliamentary governments
Institutions (DPI).                                           than presidential systems. But this result is not robust to
Beck, Clarke, Groff, Keefer, and Walsh summarize key    the use of different variables from the DPI, which raises
variables (many of them new), compare this data set with      puzzles for f'uture research.
others, and explore the range of issues for which the data       Similarly, Roubini and Sachs find that divided
should prove invaluable.                                      governments in the OECD run higher budget deficits
Among the novel variables they introduce:                   after fiscal shocks. Replication of their work using DPI
* Several measures of tenure, stability, and checks and     indicators of divided government indicates otherwise,
balances.                                                     again suggesting issues for future research.
* Identification of parties with the government               Among questions in political science and economics
coalition or the opposition.                                  that this database may illuminate: the determinants of
* Fragmentation of opposition and government                democratic consolidation, the political conditions for
parties in legislatures.                                      economic reform, the political and institutional roots of
The authors illustrate the application of DPI variables     corruption, and the elements of appropriate and
to several problems in political economy.                     institutionally sensitive design of economic policy.
This paper - a product of Regulation and Competition Policy, Development Research Group - is part of a larger effort
in the group to understand the institutional bases of poverty alleviation artd economic reform. The study was funded by
the Bank's Research Support Budget under the research project "Database on Institutions for Government Decisionmaking"
(RPO 682-79). Copies of this paper are available free from the World Bank, 1]818 H Street, NW, Washington, DC 20433.
Please contact Paulina Sintim-Aboagye, room MC3-422, telephone 202-473-8526, fax 202-522-1155, email address
psintimaboagye@worldbank.org. Policy Research Working Papers are also posted on the Web at www.worldbank.org/
research/workingpapers. The authors may be contacted at tbeck(@worldbank.org, gclarke@aworldbank.org,
pkeefer@worldbank.org, or pwalsh@Cworldbank.org. February 2000. (46 pages)
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about
development issuies. An objective of the series is toget the findings out quickly, even if the presentations are less than fully polished. The
papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this
paper are entirely thosc of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the
countries they represent.
Produced by the Policy Research Dissemination Center



New Tools and New Tests in Comparative Political
Economy: The Database of Political Institutions
Thorsten Beck, George Clarke, Alberto Groff,
Philip Keefer, and Patrick Walsh
Beck, Clarke, Keefer and Walsh: Development Research Group, The World Bank. Groff: Federal
Department of Foreign Affairs (Switzerland).
This paper's findings, interpretations, and conclusions are entirely those of the authors and do not
necessarily represent the views of the Federal Department of Foreign Affairs, or the World Bank, its
Executive Directors, or the countries they represent.



.... I I -



2
Introduction
The investigation of the institutional and political roots of government decisions has
accelerated rapidly in the last twenty or more years. However, cross-country empirical work
has been handicapped by a lack of detailed data on the political and institutional characteristics
of countries, and their change over time. Comparative political economy, therefore, has been
restricted to small-sample case studies, important in their own right, but imposing
unavoidable limits on researchers seeking to generalize their results beyond the case study
sample.
The lack of data has been a serious obstacle to better understanding of numerous areas
in political economy, including the political conditions under which governments promote
rather than retard economic development. Despite evidence that institutions underpin
economic development, we are still far away from the answers to two fundamental questions:
Which political institutions are most conducive to development and reform? And under what
conditions do such institutions emerge? Unfortunately, cross-country data that could help to
answer these questions, data that provided a disaggregated picture of a country's political
institutions and operations, have been scarce. In this article we introduce a new data set that
fills many of these data gaps, the Database on Political Institutions (DPI). It has been
compiled in the Development Research Group of the World Bank, but it rests on an
intellectual foundation laid by a large body of political economy research.
The DPI contains 113 variables (detailed in Annex 1) for 177 countries over the years
1975-95. The variables provide details about elections, electoral rules, type of political system,
party composition of the opposition and government coalitions, and the extent of military
influence on government. The DPI also contains a number of new variables, compiled from



3
the raw data, including original measures of checks and balances and political stability.
In contrast to other databases, these variables are nearly all objective and their
construction is entirely transparent. They are also disaggregated, allowing researchers to get
away from such useful but broad indicators of countries as whether they allow elections or
not, whether elections are "free", or whether the executive is "constrained." Instead, the DPI
allows researchers to use precise and concrete institutional features of countries, and to
combine them in ways appropriate to theories they are examining.
In this paper, we identify numerous and policy-relevant research directions where this
data will be helpful. We also describe the different variables and demonstrate their utility in
several applications. For example, we find that earlier results concerning the impact of
divided government on public debt (Roubini and Sachs, 1989), and the impact of
presidentialism vs. parliamentarism on democratic consolidation (Stepan and Skach, 1993),
change considerably when their methodologies are replicated using the DPI's variables on
divided government, presidentialism and parliamentarisrn. The remainder of the paper, and
the annexes, provide details on sources and collection methodology.
The need for cross-country institutional data
-A large and diverse array of inquiries into the political economy of economiQm
development and the determinants of institutional change have been hindered by the absence
of detailed cross-national objective data on the institutional and political characteristics of
countries. A few of the areas of investigation to which the DPI can contribute are outlined
here.
Political Business Cycles. Research on the existence and importance of a political business
cycle has largely focused on a select group of developed countries. One reason for the narrow



4
focus has been the absence of election data in a larger sample of countries. The Database on
Political Institutions has election dates (month and year) and election winners and losers for
more than 150 countries since 1975. Both legislative and executive elections are included, as is
the composition of governments both before and after elections.
Checks and balances: stability versus decisiveness. The contribution of political institutions
to economic policy outcomes is another important area of research. For example, divided
governments or checks and balances are thought to have conflicting influences on economic
development. One line of research suggests that they encourage budget deficits and delay
reactions to crisis. Another line of work finds that they increase the predictability and
restraint with which governments regulate firms and citizens. Research on the first topic (e.g.,
by Roubini and Sachs) focuses on a selected group of OECD countries. Research on the
second issue has relied on case studies (for example, North and Weingast, 1989, study the
introduction of institutional checks following the Glorious Revolution in England), or on
subjective data, such as the extent of "executive constraints" in a country (e.g., Keefer and
Knack, 1997). The DPI includes several objective measure of checks and balances that permit
the impact of this institutional feature of countries to be systematically analyzed, and the
tradeoffs between stability and decisiveness to be investigated.
Electoral rules, policy maker incentives, and corruption and fiscal policy. Another set of
theories in political economy predict that pork barrel spending and corruption will vary
systematically both with the number of decision makers (Rasmusen and Ramseyer, 1994) and
with the electoral rules by which decision makers are selected (Cox and McCubbins,
forthcoming). Tests of these theories are made difficult by the absence of cross country data
on both of these institutional dimensions. This data set goes some distance in remedying this



5
problem. First, it provides a better approximation of the number of key decision makers in a
country through two variables on the number of checks and balances. Second, it presents
information on voting rules in countries, including the use of plurality and/or proportional
systems, closed or open list systems, threshold or D'HondLt systems, and the relative
dominance of these systems if more than one is used.
Political stability, political tenure, and government policy. Government officials may also
make decisions that retard development because their policy horizons are limited to their time
in office. The shorter the horizon of officials, the more likely it is that they will take actions
with significant long-run costs to obtain modest short-run benefits. All existing analyses of
the impact of leader horizons (for example, Alesina, et al., 1996,and Clague, et al., 1996) rely
on executive turnover measures. However, in systems in which the prime minister or
president is not the only veto player, these turnover measures present a biased picture of the
horizons of all veto players. In particular, in countries with multiple veto players, the
replacement of a president, prime minister, or even a cabinet, does not always mean that all
veto players from the previous government have changed. Often, some of the parties that
exercised a veto role in earlier governments reappear in successor coalitions. This database
allows-more nuanced turnover indicators to be constructed, and presents two such-indicators.
Civil war, ethnic conflict and political institutions. The conditions under which ethnic
groups are in conflict and that this conflict leads to violence are a continuing and important
focus of research (see, e.g., Collier and Hoeffler, 1999). However, political institutions are
likely to have both a significant influence on whether ethnic groups encounter conflicts of
interest and on whether those conflicts precipitate violence. Institutions affect whether
different groups have a voice in the allocation of state resources; they also affect the credibility



6
of assurances provided by one ethnic or other interest group to another (Weingast, 1998). The
DPI has information bearing on both of these issues.
Electoral rules, such as voting thresholds and other choices regarding the aggregation
of votes, influence whether minority interests are more or less likely to be influential in the
formation and dissolution of governments - and hence, more or less likely to abandon formal,
institutional channels for the resolution of their grievances. Similarly, the existence of
multiple veto players - checks and balances - should make it more likely that any given ethnic
or other interest group has the ability to block state actions that are to its detriment. This is
the essence of institutional arrangements that enhance credibility.
Electoral competitiveness and policy outcomes. To the extent that the "losers" from
economic reform are politically influential - are key constituents of political decision makers
- reform is less likely. This is a truism of political economy, but it leads to a still unanswered
question: are politicians most likely to favor constituents who were supporters in past
elections but deserted them recently; those whose support is up for grabs and could swing the
next election one way or the other; or core supporters who never abandon them under any
circumstances? All of these possibilities receive some theoretical and empirical support (see
Dixit and Londregan, 1996, and Schady, 1998).
A simpler question, which is similarly unanswered, is whether greater competition in
contests for political office increases or reduces the willingness to reform. Faced with the
prospect of more competitive elections, politicians might be more sensitive to redistributive
concerns and less likely to reform. However, politicians who feel few competitive pressures
to reform may be more likely to engage in policies that benefit them and their core supporters
at the expense of the rest of society. In the latter case, greater electoral competitiveness might



7
lead to greater willingness to pursue policies that are less disproportionately beneficial to them
and their core supporters. Several variables in this datab-ase, ranging from indices of electoral
competitiveness, to the vote shares of parties and directly elected executives, permit cross-
country investigation into questions such as these.
Poverty alleviation and voice. The construction of a safety net for the poor, or of programnLs
that allow the poor to build human and financial capital to rise out of poverty, is often
distorted by political imperatives. Pritchett and Gelbach (1997), for example, have providecd
evidence and theory indicating that targeting assistance to the poorest is unsustainable over
time. In a simple majority-rule setting, they show that the less poor majority eventually
demand more expansive coverage of targeted programs as a condition for their continued
support of them. However, the structure of political institutions can mitigate - or exacerbate
- this outcome. The rules governing election to the US Congress, for example, give an
incentive to legislators representing the inner city and to those representing farming districts
to agree on the continued provision of narrowly targeted food stamps. Such incentives are less
strong in countries where legislators represent broader segments of society, most of whom
gain little or lose from targeted assistance. The DPI contains variables that help to capture
some ofthe institutional features that would affect the sustainability of targeting, including the
presence of legislatures and rules governing election to legislative chambers.
Economic tradeoffs in constitution drafting. When tremendous institutional transformations
occur, from totalitarianism to democracy, or even from one constitution to another, they
entail numerous detailed decisions from among a broad array of institutional choices. Here,
again, the empirical evidence that should inform these choices - e.g., which electoral rules?
how should the executive be selected? - has been largely unavailable. As a consequence,



8
decisions about large issues - should a system be presidential or parliamentary? - and smaller
but still important issues - how high should voting thresholds be, which rules of proportional
representation should be adopted - tend to rest on intuition and political self-interest,
uninformed by empirical evidence about the later consequences of these choices for growth,
development, or social conflict. Variables in the DPI should contribute to accelerated
empirical research on such questions.
The Variables
The variables in this dataset provide considerable scope for improving our ability to
answer these questions. The more than one hundred different variables can be divided into
seven general categories. Each category, if not each variable, is described in this section.
1. Political system and electoral competitiveness
Two fundamental characteristics of a political system are the relationship of the
executive and legislative branches, and the competitiveness of elections of the political actors
who occupy them. The DPI has information on both. With regard to the first, the variable
SYSTEM was created to capture whether countries are presidential, assembly-elected
presidential, or parliamentary. Determining whether political systems are presidential or not
depends. on answers to a number of questions. Our judgements are similar to those-of others,
but were modified to take into account our particular sources.
The first question is straightforward: is there a single executive elected by popular
vote? If both are true, the system is coded as 'presidential." However, there are countries in
which both a prime minister and a president are in place, or where the executive is elected by
the legislature but is called a president and enjoys the same degree of independence from the



9
legislature, after election, as the typical popularly-elected president. Countries with these
characteristics were coded in the following way.
For chief executives who were called presidents and who were elected by the assembly,
as in Bolivia, we ask whether the assembly can easily recall him or her. If they need a two-
thirds vote to impeach, or are required to dissolve themselves in order to force out the
executive, then the country is characterized as having an "assembly-elected president."
Otherwise, the country is characterized as "parliamentary."
Where there are both a president and a prime minister, we ask, how likely would it be
for the president's view on legislation to prevail over that of the prime minister? This led us
to place greater weight on legislative powers of the president. We distinguish whether the
system was presidential or parliamentary with a three-part test. First, can the president veto
legislation that the parliament can override only with a supermajority? If so, the country is
coded as presidential despite the presence of a prime minister. Second, in the event that
presidents lack this veto power, we ask whether they can appoint and dismiss cabinet
ministers (including the prime minister) and dissolve the assembly. If they can do both, the
system is again characterized as presidential. Finally, in those few cases where these details are
not revealed in the source material, if the source material refers consistently to the-president
rather than the prime minister as the important decision maker, the system is again classified
as presidential (as in Romania and Kyrgyzstan). The data is thoroughly annotated to identify
these rare cases.
These decision rules led us to code Portugal and France as parliamentary rather than
presidential systems. Others (Shugart and Carey (1992), Shugart (1995)) place greater weight
on non-legislative presidential powers, such as the right to appoint cabinet members, and



10
therefore label these countries as presidential. Although policy making in France under
divided executive control seems to support our interpretation, legitimate reasons exist to use
other categorization criteria. One advantage of the DPI lies in its explicit statement of the
rules we employ, allowing researchers to determine more easily the adequacy of variables to
their conceptual requirements.
The database also contains variables that characterize the competitiveness of elections
in countries. The earlier discussion outlined the potential importance of electoral
competitiveness to political economy investigations: the policy choices of politicians are likely
to be influenced by the likelihood that those choices will lead to their replacement. In
addition, in the creation of checks and balances variables, it matters a great deal whether the
legislative and executive branches of government are independent of each other. If legislative
elections as a whole are non-competitive, it is usually the case that the executive - or whoever
chooses the executive - also exercises decisive influence over the legislature. Consequently,
the legislature is less of a check on executive behavior. Finally, investigations of institutional
change, particularly of change to more democratic decision making, need to accurately
characterize the institutional environment. Knowing the formal, constitutional rules
governing countries is one way to characterize democracy; an important supplement is to
know whether these rules are applied in practice. The Indices of Electoral Competitiveness
address both of these issues.
There are two such indices in the data set, one for executive and one for legislative
elections. The core of the two indices is the number of parties that could and did compete in
the last election. Ferree and Singh (1999) present the first effort to code such variables, and our
indices are based on theirs. They classify 46 African countries on a six point scale (one to six).



11
The highest score (six) on their scale goes to countries in which multiple parties compete in
elections, even if they do not win seats. For the DPI, we collected sufficient information to
replicate this type of indicator for more than 100 additional countries. However, for wider
application of their indices we added a seventh point to the scale, in order to identify variation
among the large number of countries that exhibit elections where multiple parties run but
only one party enjoys great success. Countries receive a seven that have elections in which
multiple parties compete and the largest party receives less than 75 percent of the vote.
Conceptually, this distinction captures institutional or informal biases against political
competitors that are not revealed in our sources; empirically, this seventh category generates
much more variability among countries outside of Africa. The scale, then, is:
1     no executive/legislature
2     unelected executive/legislature
3     elected, one candidate
4     one party, multiple candidates
5     multiple parties are legal, but only one won seats (because other parties did not
exist, compete, or win seats).
6     multiple parties competes and won seats (but one party one 75 percent or more
of the seats)
7     the largest party received less than 75 percent of the seats.
As with distinctions between presidential and parliamentary systems of government,
numerous subtleties arise in assessing scores for these indices. They are thoroughly detailed in
the documentation accompanying the database. For example, many countries have executives
who are indirectly elected. In our rating scheme, they receive the same competitiveness score
as the body that elects them. Therefore, in a parliamentary system, executive elections are as
competitive as legislative elections. When there is an electoral college, elections are as
competitive as membership in the electoral college. If the executives themselves appoint the
electoral college, the systems are scored as two, on a scale of one to seven. Competitively



12
elected prime ministers get six or seven. The chief executives of Communist nations (the
chairman of the Communist Party) are assigned a three, because they are elected by the Party
Congress, an electing body which they do not appoint. Executives elected by small, appointed
juntas or by appointed electoral colleges receive a two. Countries that allow referenda on
unelected executives are scored as three.
Three other variables supplement these indices. First, elections may be competitive,
but the military may nevertheless wield significant power behind the scenes. To get at this,
we have two variables indicating whether the chief executive and the minister of defense,
respectively, are military officers. Second, our sources often mention voting irregularities or
candidate intimidation that are serious enough to affect electoral outcomes, though coverage is
far from complete. We record such references irrespective of whether claims that fraudulent
elections have occurred are based only on the allegations of the opposition or are supported by
independent international observers. Non-constitutional irregularities are also recorded as
present when important parties boycott elections or the election results.
One application of these variables is to the debate over whether presidential or
parliamentary systems are more likely to lead to democratic consolidation. Stepan and Skach
(1993),for example, argue that parliamentary systems are more likely to be associated with
long-running democracy than are presidential systems. They offer as evidence the fact that of
53 non-OECD countries, 22 were democratic for at least 10 consecutive years, and of these, 17
were parliamentary. Variables from the DPI suggest a different picture, however. Table 1
indicates a breakdown parallel to theirs using a different indicator of democracy (countries
that score at least six on the Legislative Index of Electoral Competitiveness), and the SYSTEM
variable, and applying these two variables to a large group of non-OECD countries from 1975



13
- 1995. In contrast to their findings, the sample of countries that have experienced at least ten
continuous years of democracy is evenly split between presidential and parliamentary systems
(22 countries in each), rather than weighted in favor of parliamentary systems. Table 1 does
not necessarily refute the argument in Stepan and Skach (1993). It does suggest that with more
extensive institutional databases, research can begin to accommodate new levels of hypothesis-
building in institutional analysis, especially hypotheses that can explain such different findings
as those in Table 1.
Table 1: Presidentialism, parliamentarism, and democratic consolidation among non-
OECD countries
Regime type during the democratic period
(Legislative Index of Electoral Competitiveness
greater than 5)
Presidential    Parliamentary  Indirectly-
elected
president
(DPI) or semi-
presidential
(Stepan-Skach)
Non-OECD countries  DPI Indicators    68                   40              14
that were democratic  (1975 - 95, 122
at least one year      countries)
during 1975 - 1995     Stepan-Skach       28               25              0
(1973 - 89, 53
countries)
Number of non-         DPI Indicators    22                22              5
OECD countries that  (1975 - 95, 47
were continuously      countries)
democratic for at least
ten consecutive years.  Stepan-Skach      5                17              0
(1973-89, 22
countries)
Note: This replicates Table 3, Stepan and Skach (1993), using variables from the Database on Political
Institutions (DPi). Countries are 'democratic" if they score more than 5 on the Legislative Index of Electoral
Competitiveness. Regime type scores are from the SYSTEM variable, where 0 - Presidential, 1 indirectly
elected and 2 - Parliamentary.



14
2. Preferences andparties
For most political economy questions, one is interested in the structure of political
decision making and also in the policy preferences or ideological leanings of decision makers.
Preferences are difficult to discern even in a small sample of countries. In the US literature,
the most rigorous attempts at assessing preferences have relied on voting records of legislators.
In the literature on European politics, impressive progress in comparative politics has been
made on the basis of labor-intensive efforts to review and classify the platforms of political
parties (see, for example, Laver and Shepsle). For many countries of the world, though, such
information is not available, not germane, or not a reliable indicator of decision maker
preferences.
A starting point for any analysis that requires preferences of policy makers is the party
identification of decision makers. Our data set identifies the party of the prime minister
and/or president, the three largest parties in the government coalition, and the largest party in
the opposition. In addition, our sources provide useful information for some countries about
the policy preferences of key decision makers on five different dimensions: economic,
nationalistic, regional, rural, religious. These dimensions were chosen because they do not
necessarily correlate with each other: religious or nationalistic parties adopt both left and right
wing economic policies; some religious parties are not nationalistic and some are; etc. That is,
we intended, where possible, to allow for a multi-dimensional policy space. Coverage is far
from exhaustive, unfortunately, but constitutes the most comprehensive characterization of
parties and decision makers that can be extracted from the sources upon which we relied.
Detailed information on party platforms (even on the existence of written party
platforms) is unavailable for most countries from our primary sources (Political Handbook of



15
the World and the Europa Year Book). We therefore first asked whether the orientation of a
party was immediately obvious from its name. If it was not, or if our primary sources
disagreed about orientation, we consulted a European website mantained by Agora Telematica
(http://www.agora.stm.it/elections/parties.htm) edited by Wilfried Derksen. Its coverage is
non-historical, however. Party orientation was also cross-checked against with information in
Political Parties ofAfrica and the Middle East and Politica! Parties of Eastern Europe, Russia, and
the Successor States, both published by Longman Current; Affairs series. In the rare cases where
conflict emerged among these sources, we used the web site and noted the conflict in the Excel
spreadsheet.
The first categorization refers to preferences regarding greater or less state control of
the economy - the standard left-right scale. Parties on t'he right are therefore those with the
terms "conservative" or "Christian democratic" in their names, or that are labeled right-wing
in our sources. Similarly, we classify parties as left if their names reveal them to be
communist, socialist, or social democratic or if the sources label them as left-wing. We classify
parties as centrist that are called centrist by our sources or if their position can best be
described as centrist (e.g., because the party advocates the strengthening of private enterprise
but also-supports a redistributive role for government, for example in the form of a tightly
knit safety net). Any party that cannot be classified in aLny of the three categories is
categorized as "other". These are frequently parties in non-competitive electoral systems.
The second dimension we consider is "nationalism". A party or the chief executive is
coded as nationalist if (i) a party is listed or described as nationalist in our sources, or (ii) a
primary component of the party's platform is the creation or defense of a national or ethnic
identity. With respect to the latter, parties that fought ifor independence from a colonial



16
power, such as the Indian National Congress, or advocate the restriction of rights of
minorities, such as the French National Front Party, are notable examples.
A third dimension coded in the DPI is whether a party is rural or not. A party is
classified as rural if either the platform or the constituency is mainly rural. Finland's and
Sweden's Center Parties are prominent examples of rural parties. Similarly, a party is
classified as regional if either the platform or the constituency pertains mainly to a specific
region of the country. Examples include most of the Belgian parties, since separate parties
exist in the Wallonian and the Flemish part of the country.
Finally, we classify a party as religious if government policy towards religion or the
members of a particular religion are among the party's core issues. We distinguish such parties
using, once again, the source characterizations of either the party's platform or constituency.
In predominantly Islamic countries, parties are defined as religious if the chief executive or
party leader is also a religious leader, as is the case in Iran. The database also captures, among
others, religious parties in Israel, the Buddhist Komeito party in Japan and the Hindu-based
Bharatiya Janata Party (BJP) in India. With respect to Christian-Democratic parties, if any
one of our sources indicated that religious issues were a central part of their party platform,
they aremindicated as Christian. Both the Italian and German Christian Democratic parties are
therefore coded as having religion as an important element of party orientation. Since
religious issues are arguably not the most central or salient concern of party members, this
suggests that this variable is somewhat biased towards identifying a religious orientation where
none exists. On the other hand, for social policy issues, or on questions such as immigration,
the coding is likely to distinguish these parties from others in the respective political systems
in ways that are both relevant and useful.



17
3. Tenure, turnover and popular vote share of the Chief Executive and the Chief Executive's
Party
A key hindrance to development is the tendency of many political leaders to make
opportunistic decisions that entail long run costs to society that far outweigh short-run
benefits. The expropriation, directly or via taxation or regulatory policy, of private assets is
the most obvious example of such decision making. Leaders with shorter horizons in office
are more likely to find such choices to be worthwhile, and there is evidence suggesting that
this effect is significant, at least in less democratic countries (Clague, et al., 1996; see also
Alesina, et al., 1996). All research that has documented this effect has relied on data on the
tenure and turnover of governments (cabinets or prime rninisters in parliamentary systems) or
presidents. This information is available in this data set, as well, for larger number of
countries than previously available. However, the DPI also provides opportunities to
circumvent several ambiguities that arise from reliance on existing data.
First, executives whose terms are shortened by constitutional limits on tenure might
react differently to limits on their time in office compared to other executives, with similar
horizons that are limited in other, perhaps non-constitutional ways. This data set therefore
includes variables indicating whether there are constitutional restrictions on executive terms
and whether sitting executives can be re-elected.
Second, executives with a stronger hold on power should, all else equal, have a longer
horizon than other executives. In particular, one would expect leaders with more popular
support, or weak opposition, to have a longer horizon. The strength of the political position
of leaders can be taken into account, at least partially, using variables that record the share of
votes that presidents received in their last election, and the share of votes of the prime



18
minister's or president's party.
Third, one would expect that stronger parties would rein in party members who
happen to have shorter horizons. Again, the DPI provides information that allows this to be
taken into account: the number of years that the party of the executive has been in office
(independent of the particular executive who has represented the party), and how long the
party has existed under its current name. Where the party of the executive has been in power
longer than the executive, it is more likely that the party restrains executives with short
horizons. Similarly, parties that have been in existence for a long period of time should have a
longer time horizon than parties that have not. We count the party's existence from the date
it adopted its name. Parties incur losses when they change their name, not least of which is a
loss of brand name. Frequent name changes indicate either that parties have no brand name
or are not well-established institutionally (to the extent that these two things are different).
Finally, existing investigations of the effects of expected tenure or government
turnover track only one decision maker in a country, the head of government (president,
prime minister or cabinet). Most democratic countries, however, have multiple decision
makers. A parliamentary government can fall and be replaced by government with an
entirely-new set of coalition parties, or by a new government with substantially the same set
of coalition partners. A president can be replaced with, or without, the corresponding
replacement of the majority parties in the legislature. Theory would suggest that dramatic
changes in government policy, or in the potential for expropriatory behavior by governments,
are likely only when all decision makers are replaced. Empirical measures of tenure and
turnover have not matched this, however.
The DPI records whether parties are in the government coalition or not, and whether



19
the party of the executive controls the lower house (there is no data on upper houses in this
data set). With this information, one can calculate which of the main veto gates legislature,
executive) in a system have changed hands in any year. Based on this information, the data set
contains two new variables, STABS and STABNS, on government stability that capture the
extent of turnover in any one year of a government's key decision makers. They are
calculated by dividing the number of exits between year t and t+ 1 by the total number of veto
players in year t. The variables are therefore on a 0-1 scaLe, with zero representing no exits
and one representing the exit and replacement of all veto players. Veto players are defined as
the president, largest party in the legislature, for a presidential system; and as the prime
minister and the parties in the government coalition in a parliamentary system. These
variables, are correlated at approximately .87 with simple measures of executive turnover.
While not dramatically different, they are different enough to suggest that hypothesis tests
invoking instability should be robust to both types of measures.
Along the same lines, measures of tenure length used in current research often refer
only to the executive. The DPI has several variables that provide different perspectives on
tenure, include the tenure of the veto player who has been in office the longest and the
shortest-period of time, assuming multiple veto players, as well as the tenure of the political
system - how long have executive elections been competitive (at least six on the Executive
Index of Electoral Competitiveness) or, if they have not been competitive, how long the
executive has been in office.
4. The Legislature
The role of the legislature in the political process is influenced not only by
constitutional restrictions on the scope of its initiative and veto authority, but also by its



20
composition and the relative strength of government and opposition coalitions. Currently, in
order to characterize the internal structure of legislators, researchers have had to rely on
measures of fractionalization of the legislature: the probability that any two legislators drawn
at random would belong to the same party. Greater fragmentation is taken to imply a larger
number of actors willing to act independently in the consideration of any given policy change.
This variable has proven useful, but it treats as identical highly dissimilar cases. For example,
the variable would record the same value for two legislatures: the first, in which a single
opposition party controls 49 percent of the seats, 10 government parties share 49 percent of
the seats, and a remaining government party controls 2 percent of the seats and a second, in
which a single government party controls 49 percent of the seats with a coalition partner that
controls 2 percent, while 10 opposition parties share 49 percent of the seats.
The data set introduced here allows for the calculation of the traditional measure of
legislative fragmentation. However, our sources also allowed us to identify which parties
were inside and outside of the governing coalition. For coalition governments, the sources
generally list the party affiliations of cabinet ministers. Otherwise, the country descriptions
provided this information. Where the information was not available, for a small number of
parties -nd countries, this is indicated in the DPI. With this information, variables were
created to record separately the fragmentation of the government coalition and the
opposition. The correlation of the government and opposition fragmentation measures is
only 0.267, suggesting that average fragmentation for the entire legislature is likely to be
misleading for many purposes. For example, the number of checks (parties in a government
coalition) in parliamentary systems cannot be reliably inferred from fragmentation of the
whole legislature. Similarly, in investigations of the relative influence of executives and



21
legislatures when legislatures become more fragmented, the relevant issue is the fragmentation
of government-allied parties, on which the DPI has information.
5. Electoral Rules
The incentives of political decision makers, and their ability to act independently of
each other, depends in large measure on the electoral rules under which they operate. Among
other effects, electoral rules influence whether candidates have an incentive to pursue
individual or party interests; whether small and single-interest parties have an incentive to
coalesce or to remain independent; and whether party leaders exercise significant or weak
influence on party candidates. Each of these and other, similar electoral characteristics have a
bearing on whether legislators are more or less permeable to special interests.
One electoral rule that affects responsiveness to narrow interests is the Mean District
Magnitude (MDM). The MDM is defined in the database as the average number of
representatives (rounded to the nearest whole number) elected by each electoral district in a
country.1 Higher values of MDM (combined with proportional voting rules) reduce the
incentives of parties to coalesce and allow parties representing exclusively narrow interests the
chance to gain a voice in a government coalition. MDM is coded for up to two legislative
chambers (House and Senate), if they have elected members. The MDM varies fromi one in
the United States or the United Kingdom to as many as 153 for the Congo.
A second electoral characteristic that the data set captures is whether the districts of the
upper house are states or provinces. This is particularly useful in studies of federalism: the
1 Cox (1997, p. 208), argues that the median district magnitude is superior, conceptually, but data on distribution
of seats across constituencies is not available from our sources and Cox also acknowledges that empirically, the
distinction has not proven important.



22
more closely that the constituencies of national legislators match those of sub-national
governmental jurisdictions, the more likely that those sub-national interests will be reinforced
in national-level decision making.
Voting rules are a third aspect of electoral rules. The DPI records whether legislators
are elected using majority (first past the post) or proportional representation systems. These
rules affect candidate incentives to adhere to party positions (as in proportional
representation) and to build up personal constituencies at the expense of party positions. If at
least some members of the legislature are elected by simple majority vote, the electoral system
is coded as "pluralistic." If at least some candidates are elected based on the percentage of
votes received by their party, the system is coded as proportional representation (PR).
Countries are allowed to (and do) exhibit both systems; in this case, the system that governs
the majority of candidates is indicated for each legislative chamber.
Proportional representation systems differ widely according to whether they favor
large or small parties or privilege party leaders or individual candidates (see Cox 1997 for a
thorough discussion). The data set therefore records if there is a threshold that parties have to
overcome in order to gain any seats in the legislature and whether the D'Hondt system, which
is less -fvorable to small parties, is used to calculate the number of seats that each party
receives. There is also a variable that records whether the electoral system is closed or open
list. A system is classified as closed if voters cannot express preferences within a list - that is,
if the position on a party's candidate list combined with the number of votes received by the
party is the sole determinant of whether a candidate is elected or not.
Some, albeit scanty information is available on the candidate selection process.
Especially in majoritarian settings, this is a crucial dimension along which countries vary in



23
the extent of party control of candidates. Systems that use "smoke-filled back rooms" to
identify a party's candidates offer party leaders greater influence than those that use party
primaries. Where possible, the data set therefore includes a variable that reflects whether
selection is of one of the following three types:
1     on a national level, i.e. by national executive, party leader, interest groups or party
factions;
2     on a sub-national level, i.e. by subset of constituency party members e.g. on
conventions;
3     by primary election, including party primary and primaries using all the votes of a
constituency.
The DPI, finally, records the month when presidential/parliamentary elections were
held.
6. Checks and Balances
A key element in the description of any political system is the number of decision
makers whose agreement is necessary before policies can. be changed. Numerous decision
makers with different policy preferences are likely to respond differently (more incrementally,
for example) to reform opportunities than single decision makers. On the other hand,
countries with multiple decision makers may offer greater protection from arbitrary
government action to individuals and minorities. Interests that are unrepresented by
government decision makers when there are few key decision makers may gain representation
when the number of decision makers rises.
The DPI significantly extends one existing measure of checks and balances, and
introduces new measures and the possibility of constructing others. One existing measure is
the Index of Political Cohesion, originally devised and coded by Roubini and Sachs (1989) for
a set of 14 OECD countries to study the political and economic origins of high-debt countries.



24
They record whether the same or different parties control the executive and legislature in
presidential systems, and, in parliamentary systems, whether there is a one, two, or three or
more party coalition controlling government, or a minority government. The DPI has
information on governing and opposition parties for over 150 countries, allowing the Index of
Political Cohesion to be compiled for a much larger sample of countries.
There are significant coding differences for the 14 overlapping countries, however, so
that for these common observations there is a correlation of only .53 between the two. For
example, we differ from the original coding by Roubini and Sachs of the Index in calling Japan
a minority government for particular years, in labeling Finland a parliamentary system, and in
coding the CDU and CSU in Germany as two parties rather than one (XXsee annex xx for a
complete identification of differences). This has a significant impact on their empirical
findings, at least over the ten years that the two versions overlap. Table 2 reports results from
Roubini and Sachs, replicated for the years 1975 - 1985, using their original coding of the
Index and the coding in the database.2 They argue, reasonably, that coalition governments
should have greater difficulty responding to the sudden emergence of crisis-driven budget
deficits, a common occurrence in this decade.
-They confirm this conjecture in their empirical work, replicated in Column 1, Table 2
for the eleven year period for which the two databases overlap. The IPC coefficient is large,
positive and statistically significant. However, using the coding of the Index of Political
Cohesion from DPI, we find no such relationship (see Column 2, Table 2). What explains
these divergent results? As Table 3 demonstrates, there are several reasons for the differences
2 Our data on unemployment and inflation differ slightly from theirs, but despite this, in our replication using
their own coding of IPC we find as they do that IPC has a large and positive association with public debt.



25
in their Index of Political Cohesion and the one coded for the DPI.
Table 2: Divided government and government debt
Dependent variable:   Original Index of    [PC from         CHECKSZ
change in public      Political Cohesion   Database of   from Database
debt/GDP                    (IPC)            Political      of Political
(Roubini/Sachs)    Institutions     Institutions
(1)               (2)             (3)
constant             -0.84                0.31            0.18
(0.53)              (0.54)          (0.96)
lagged change in      0.867               0.84            0.84
public debt (t-1)     (0.09)              (0.09)          (0.09)
change in             0.12                0.0)57          0.06
unemployment rate  (0.11)                 (0.12)          (0.12)
change in debt        -0.20               -0.15           -0.17
service costs        (0.18)               (0.19)          (0.18)
change in growth    -0.32                 -0.27           -0.32
rate                 (0.10)               (0.12)          (0. 1)
Japan (0-1)           2.48                0.67            0.74
(1.55)              (1.47           (1.62)
IPC or CHECKS1   0.74                     -0.053          0.02
(0.30)              (0.28)          (0.23)
N                     83                  79              83
Adjusted R2           0.54                0.5             0.5
Note: This table reproduces the specification from Table 7, column 2, Roubini and Sachs, 1989, for the
years 1975 - 85 (years of overlap with our dataset). Standard errors are in parentheses. Debt service
costs are the real interest rate less the growth rate, multiplied by lagged debt/GDP. Column 2 excludes
four observations (Italy and the US, 1975-76) for which IPC in the DPI is unavailable.
The coding differences noted in Table 2 have two effects. One is to raise the sample
mean substantially, from 1.18 to 1.51. The other is to reduce the dispersion in the DPI
measure of IPC among the fourteen countries; the IPC compiled by Roubini and Sachs
exhibits a coefficient of variation of 83.5, while that of DPI[ is 72.4. These changes are not
surprising. One would expect that fourteen wealthy, democratic OECD countries would
exhibit more significant checks on government decision makers that higher values of the IPC
imply, and that they would not exhibit large and sustained differences in such a fundamental
political variable as the extent of divided government. The effects of divided government are



26
likely to be non-linear - with the impact of a second independent decision maker being much
greater than the addition of a third, for example - providing one explanation for the
insignificant results in column 2, Table 3. Similarly, the reduced dispersion of the IPC in the
DPI would also explain the differences in the results across columns 1 and 2 of Table 2.
Table 3: Differences in Coding the Index of Political Cohesiveness:
Database of Political Institutions and Roubini and Sachs (1989)
(OECD nations, 1975-1985)
Source of Difference                 Observations Affected
DPI counts CDU and CSU as 2 parties; RS89    Federal Republic of Germany,
counts them as 1.                           1983-1985
DPI indicates 2-party government; RS89      United Kingdom, 1978
indicates a 1-party government.             Norway, 1975-1980
Sweden, 1975
DPI indicates a minority government; RS89   Belgium 1982-1985
indicates a majority government of one party    Denmark 1978, 1984-1985
or a majority coalition.                    Japan 1977-1980, 1984-1985
Norway 1985
Sweden 1983-1985
UK 1977, 1979
DPI indicates a minority government, RS89   Belgium, 1976-1977
indicates a majority government of one party
or a majority coalition. It is also possible that
language- based "wings" of parties are treated
differently - DPI often counts them as separate
parties.
For each year, DPI records the situation as of   Denmark, 1981
January 1; RS89 seem to use a different     Ireland, 1981
convention.                                 Norway, 1981-1982
Sweden, 1976, 1978-1979, 1981
Possible that language-based "wings" of parties   Netherlands, 1978-1985
are treated differently - DPI often counts them   Belgium, 1981
as multiple parties.
DPI classifies them as parliamentary; RS89  Finland, 1975-1985
classifies as presidential.                 France, 1975-1985



27
One can examine the robustness of the results in column 2 by substituting for IPC a
measure of the number of independent decision makers irn a government. The underlying
logic of the Roubini and Sachs argument suggests that such a measure, too, should be
correlated with the difficulty of responding to fiscal crisis. One such measure in the DPI is
the variable CHECKSI, a measure of checks and balances that is explained below. Results
using CHECKS1 in place of IPC are displayed in Column 3, Table 2, and again demonstrate no
significant relationship between the number of decision makers and response to fiscal crisis in
this small and relatively homogeneous sample of countries.
Because of the small and relatively homogeneous number of countries considered, the
results in columns 2 and 3 of course do not reject the Roubini/Sachs conjecture - divided
governments have greater difficulties in agreeing to a response to crisis. They only suggest
that as we advance in our attempts to make cross-country institutional characterizations of
polities, legitimate distinctions will arise among superficially similar concepts. On the one
hand, these distinctions need to be subjected to scrutiny; on the other hand, if they are
reasonable, they need to be used to test the robustness of empirical claims in political
economy. Where robustness is weak, this likely implies new opportunities for research: why
do some measures of checks and balances, following one set of criteria, lead to results that are
not replicable using other reasonable measures of checks and balances, capturing different
institutional dimensions?
The Index of Political Cohesion does not distinguish countries according to the
effectiveness of electoral checks on government decision makers. This is a particularly
relevant concern when expanding the index to a set of countries outside of the OECD. When



28
electoral checks are few, perhaps because of laws limiting political competition, executive
control of the legislative apparatus is usually strong. Legislative seats depend on party
nomination, which in turn is usually vested in the executive in such countries. The Index also
does not take into account electoral rules that influence party control over members, an issue
even in OECD countries. Where party control is weak and the same party controls both the
legislative and executive branches of a presidential government, this index would understate
the level of checks and balances by coding the country as not having a divided government.
To adjust for these circumstances, the DPI includes two new variables, CHECKSI and
CHECKS2. These variables count the number of veto players in a political system, adjusting
for whether these veto players are independent of each other, as determined by the level of
electoral competitiveness in a system, their respective party affiliations, and the electoral rules.
Both variables take into account the Legislative Index of Electoral Competitiveness. If the
index is less than four, the two measures are set equal to one since, regardless of the formal
constitutional arrangements in a system, where legislative elections are uncompetitive,
constitutional checks on officials are unlikely to be binding.
For values of the electoral competitiveness index of four or more, the variable
CHECKSI is increased by the number of veto players in the system.3 In presidential systems,
it is increased by one for the president and one for each legislative chamber. However, if
elections are conducted under closed list rules (closed list equals one) and the president's party
is the largest government party in a particular chamber, then it is assumed that the president
exercises substantial control over the chamber and it is not counted as a check. For
3 The values of CHECKS1 and CHECKS2 are nearly invariant to the use of the Executive or Legislative Index
of Electoral Competitiveness.



29
parliamentary systems, CHECKS1 is increased by one for the prime minister and the number
of parties in the government coalition, including the prime minister's own party. If the
system is closed list, however, and the prime minister's party is the largest in the government
coalition, then this sum is reduced by one. Relative to the Index of Political Cohesion, then,
CHECKS1 allows a legislative chamber to exercise veto power even if the same party controls
both the chamber and the executive branch.
Keefer and Stasavage (1999) use CHECKS1 to demonstrate that the delegation of
monetary policy to a central bank has a significant dampening effect on inflation only in the
presence of checks and balances. Keefer (1999) finds that the determinants of banking crises
are substantially different in countries that exhibit high values of CHECKS1 compared to
those countries that exhibit lower values.
The variable CHECKSI can also be used to fuel the debate about whether presidential
or parliamentary forms of government are likely to result in "decisive" decision making.
Some (see Stepan and Skach, 1993) argue that presidential governments are more likely to give
rise to gridlock, since neither the legislature nor the executive has an incentive to give way in
policy debates because they are independently selected. It: is not clear, however, which is
more likely: that an independent legislature would veto the proposals of an executive, or that
members of a government coalition in a parliamentary system would veto the same proposal
emanating from another member of the coalition. The average CHECKSI variable for all
countries that score either six or seven on the Legislative [ndex of Electoral Competitiveness is
actually much higher for parliamentary systems (3.57) than for presidential systems
(approximately 1.96). While legislative and executive interests are rarely unified in
presidential systems, governments in parliamentary systems are even less so. However, it



30
remains unclear whether bargaining among multiple coalition partners in parliamentary
systems regarding particular proposals is less costly than bargaining between only two veto
players in presidential systems. The DPI data set, however, provides one tool for advancing
research on this question.
The extent to which one political decision maker might act as a check on another
depends in part on the similarity of their policy preferences. Government decision makers are
more likely to exercise strong veto power if their policy preferences are closer to those of the
opposition and more distant from the preferences of other government decision makers. Such
a decision maker would be more willing to leave the government and join the opposition than
decision makers whose views were far from the opposition. To take into account the policy
orientation of parties, a second checks and balances variable, CHECKS2, was therefore
compiled and introduced into the data set. CHECKS2 is therefore the same as CHECKS1, but
is augmented by one for every veto player (defined as the largest government party in the
legislative chamber in presidential systems, and all coalition members in parliamentary
systems) whose orientation (left, right or center) is closer to the opposition's orientation than
to the average of the rest of the government. The opposition's orientation is given by the
orientation of the largest opposition party.
These new measures of checks and balances offer considerable additional information
about polities. Their correlation with the coding of Index of Political Cohesion in the
database is less than 0.60. This suggests that theoretical priors about the number of veto
players in government can make a significant difference for empirical work, reinforcing the
need for diverse data sets and variables.



31
7. Federalism
Sub-national political structure affects national level policy making in numerous ways.
First, sub-national units may have veto power over national level policy decisions. Second,
they may exert pressure for greater, or at least different levels of redistribution than would
otherwise be the case. Third, their existence may affect the cohesiveness of national parties,
which struggle for positions in both sub-national and national jurisdictions, although the
distribution of voter preferences in the two could be quite distinct.
There are several variables in the DPI to capture the extent of federalism in a country's
political structure. The first is whether there are contiguous autonomous regions in a
country. An autonomous region is recorded if a source explicitly mentions a region, area or
district that is autonomous or self-governing. We required that autonomous regions be
constitutionally designated as "autonomous", "independent" or "special".
A second dimension of information on sub-national governments is whether
state/provincial governments are locally elected. This is zero if neither the local executive nor
the local legislature are directly elected by the local population that they govern; one if either
is directly elected and the other is indirectly elected (e.g., by councils at subsidiary levels of
government) or appointed; and two if they are both directly and locally elected. If there are
multiple levels of sub-national government, we consider the highest level as the
"state/province" level. We also investigated, with limited success, whether municipal
governments were locally elected. This was coded the same as the state/provincial
government. If there were multiple levels of sub-national government, we considered the
lowest level as the "municipal" level.
The third dimension on which we attempted to collect information is the jurisdiction



32
of local governments. We asked whether the states or provinces had authority over taxing,
spending, or legislating. The variable is coded as one if our sources report authority over any
of these areas.
Sources and Data Collection
There were two main sources of data for this project, each used for a different time
period. For the period 1975-1984, Europa Year Book was consulted For the period 1985-1995,
the Political Handbook of the World, edited by Arthur Banks, was used. However, Europa was
also used in this second time period to plug topical gaps in coverage, mostly for electoral rule
and constitutional framework variables.
There were several reasons for dividing our study between two sources. First, prior to
1985, the Political Handbook of the World was published every 2-4 years and lacked detailed,
year-by-year observations. This factor recommended Europa for the first time period. In the
period 1985-1995, the two books provided complementary information. Europa has better
data on the number of seats held by parties, the constitutional framework and electoral rules
over this period, while the Handbook offers more detailed information on the ideological
orientation of the parties and recent political history. Therefore, both sources were used, with
an emphasis on the Handbook.
The consistency of data between these two time periods was extensively checked. The
collection system helped ensure that discrepancies would be identified and discussed, since
different people were responsible for coding different time periods. In the event of
discontinuities in the 1984 - 85 period, the coders' approaches were discussed among a wider
group, including most of the authors of this paper, to determine which was more appropriate,
and the country-variable was re-coded accordingly. Most of these discontinuities involved



33
time-sensitive variables (how long in office, in majority, etc.); electoral rule variables (closed
list, plurality/PR); and the ideological orientation of parties. They were resolved largely by
establishing and refining the conventions of data collection, such as when a ruler was credited
with taking office (in former colonies, upon internal self-government rather than
independence) or what exactly constituted an electoral rule (whatever rule governs the
majority of seats). There were, besides, extensive consultations on resolution of ambiguities.
The results of these discussions are all reflected either in the documentation sheet, in the form
of rules that were followed for coding, or in the notes embedded in the corresponding cells in
the Excel spreadsheet.
Electoral rules were more difficult to capture than other institutional variables, and
were often not available in the handbooks. For the later period (1985-1995) we took most of
our data from the website of the International Parliamentairy Union, which reports rules for
1998. When this data matched the institutions described in 1975-1984, we assumed that no
changes had been made and filled in the intervening years. In the event that a system changed
and then switched back, this introduces errors. Since this assumption was made only when
institutions from 1984 matched those in 1998, these cases are limited to very stable
democracies. In the case where IPU-1998 did not match earlier data, the IPU dat-a -was only
inputted for 1995, and blanks were left for remaining years.
There were a number of alternative sources of data that could have been used.
Prominent among these is the Statesman's Year Book edited currently by Brian Hunter. This is
a valuable source, but we rejected it in favor of the above sources for several reasons (recalling
that the addition of a single source to the compilation process implied at least four person-
months of additional work). First, although Statesman's level of detail is excellent in Europe



34
and developed nations, its coverage of developing nations is less detailed than in our preferred
sources. The number of seats won by parties is more often omitted, as is the identification of
parties as affiliated with or opposed to the government. Details on the electoral systems are
also more often absent, as are vote percentages for presidential systems. Our sources provided
information on seats, parties, and election results more consistently in developing nations.
The reporting in Statesman's also lagged events by a few years, especially in developing nations
or nations emerging from Communism. For example, election results in countries such as
Kazakhstan or Kyrgisztan are not covered until 1995, compared to 1992 in our preferred
sources. Finally, Statesman's rarely provided lists or descriptions of active parties.
Another possible source for this database is Keesing's Record Of World Events. This
source was rejected because it was primarily focused on news rather than permanent
institutions; and because it is organized by month, creating significant compilation costs -
since relevant data would have to be scoured from every month's report. Various other
sources (CiA Handbook of the Nations, OECD Political Data Handbook edited by Lane, et al.,
the Year Book of World Affairs, produced by the London Institute of World Affairs) were
rejected for reasons of coverage and focus.
--Despite the use of comprehensive sources and detailed cross-checking procedures, there
are several variables which need to be treated with some caution because of the fundamental
limitations of cross-country sources such as those upon which we relied. Party orientation is
one of these, since we were unable to rely on systematic listings of party platforms or
legislative positions. However, since our first-cut source (the party name/description in the
handbooks) and our default source (the Agora website) disagreed infrequently, we believe
these variables can be used with confidence.



35
Information scarcity hindered the coding of all of the federalism variables. Thirty-
eight percent of the observations are coded for the sub-national authority variable, and 42
percent of the municipal election variables. The remainder are blank.
The indices of electoral competitiveness were coded according to a carefully established
set of conventions, all of which are described in the docuimentation. Nevertheless, there are
many cases that defied simple categorization (all noted either with notes in the Excel
spreadsheet, or in the documentation). When executives were elected by a body other than
the overall populace, we apply the categorization to that electing body (often the legislature,
but sometimes an electoral college) and impute its score to the executive. This results in
military governments of Brazil, Thailand, and South Korea receiving unusually high scores (5-
7), since they did not "take" power, but instead followed formal institutional procedures that
involved selection by an elected legislature. Similarly, even if a powerful executive rigged the
elections (which is recorded elsewhere), the country could still be scored a six or seven. While
the formal procedures are likely to understate the level of military control of the process, and
overstate the level of electoral competitiveness, we preferred to preserve the conventions in
these cases.
-Special rules were devised to handle deviations from constitutional procedures. If an
executive or legislature illegally suspends or ignores elections, the suspending entity is scored
with a two starting in the year after the event. If unelected dictators subsequently hold a
plebiscite to confirm their claim to office, their score is raised from two to three, but only if
the plebiscite is binding. Dynastic rulers in Arab countr:ies, who inherit power, or otherwise
obtain it through obscure intra-family negotiations, are given the same "two" earned by
military officers who seize power in a coup.



36
The DPI compared to other databases
The DPI seeks to add to existing knowledge on political institutions, and to gather into
one source information that had previously been scattered. Therefore, it is reasonable to ask
how this database compares to other political databases, in terms of detail, coverage, and focus.
Comparisons to two illustrate the potential contribution of the DPI to future empirical
research in cross-country political economy. One of these is the Polity IH database compiled
by Gurr, et al. (1998), and the other are the objective assessments of government checks and
balances created by Henisz (1997).
Polity III. Gurr, Jaggers, and Moore (1998) have compiled Polity Im, covering a time
period stretching from 1800 to 1994. Polity Im covers 156 nations - all those with populations
above 500,000 - and 20 areas that were later amalgamated into a single nation-state (Saxony,
Hesse, etc. into Germany; Lombardy, Tuscany, etc. into Italy). By comparison, our database
covers all sovereign nations with populations above 100,000 - in all, 177 countries, for the
years 1975 - 1995.
Polity III includes numerous subjective indicators of the political and institutional
environment in countries. However, these tend to be highly aggregated. For example,
subjective measures of regulation, competition, and openness in executive recruitment
incorporate information about the elections or their absence; fraud; coups; scope of
participation in the selection of the executive; military involvement; the number of competing
parties; and other relevant factors. The weights assigned to these different characteristics, all
of which are appropriate to consider, are not clear, however. The Polity m variables
"regulation" and 'competitiveness" of participation, like most of the Polity III variables,
exhibit similar highly aggregated characteristics. Implicit in their codings are evaluations of



37
political violence and fraud; constituencies of parties; legality of parties; and the endurance of
parties over time. Using this aggregate level of data organization, many qualities of the
political system can be captured by a very few variables, and the net effect of formal and
informal (harder to observe and codify) influences can be summarized. The disadvantage of
this information is that those seeking specific details on the electoral system will find this
information inextricably entangled with the other measures comprising the variable.
By comparison, the DPI uses variables that reflect only one dimension. For example,
many variables are dummies, focusing on questions such as "is the closed list system used?", "is
the chief executive in the military?", or "was there electoral fraud?". Examples of non-dummy
variables include the seats, ideological orientation, and government/opposition affiliation of
political parties; details on the level of federalism; the elapsed time in office of the chief
executive; and the timing of elections. This database thus presents data at a disaggregated
level. With some exceptions, the DPI generally avoids the use of variables that wrap several
different measures into one number. The creation of such variables is left to the user. The
exceptions to this are the Indices of Electoral Competitiveness and the CHECKS variables.
The electoral competitiveness variables take into account rules governing whether parties can
run and-actual outcomes - whether parties do run and win seats. The CHECKS variables
draw on several variables within the database to construct an index of checks and balances in
the political system. While our variables therefore do not necessarily reflect the informal
constraints on electoral competition or executive authority, they allow researchers to identify
the extent of formal constitutional control on political decision makers.
In addition to methdological differences in the two databases (subjective and
aggregated, but evaluating both formal and informal conditions in a country, compared to



38
objective and disaggregated), there also differences in the types of variables included. Our
database has no information on "collective executives", such as military juntas; nor
information on the scope of government action or regulation in the polity; nor on the
creation, destruction or nature of regime change in polities. On the other hand, the DPI
permits users to track more frequent and micro-level changes in regimes, such as the
composition of the legislature, changes in executive, or changes in the electoral
competitiveness indices. It also features several dummy variables representing details of the
electoral system and much more detail on parties in the system.
Political constraints. Henisz (1997) presents another valuable source of information
on political institutions, in particular the data necessary to create a measure of checks and
balances. His series addresses some key issues, such as the number of chambers in a legislature,
the degree of federalism, Supreme Court turnover, factionalization of the legislature, and the
relative influence of different parties among the executive, legislature, and judicial branches.
From this information, he formulates a measure of political constraints that is analagous to,
but different in construction from, the CHECKS variables reported here. The Henisz data has
a longer time-series, but substantially fewer variables (e.g., with regard to government or
opposition parties, party orientation, electoral rules, the different federalism indicators, etc.).
Keefer and Stasavage (1999) report empirical results using CHECKS1, from the DPI,
political constraints from Henisz, and Executive Constraints, a subjective variable from Polity
III. The key conclusions are generally robust to which variable is used (central bank
independence matters more in the presence of checks and balances). However, these variables
are not so closely correlated that one can assume robustness.



39
Conclusion
This database is the product of an effort to bring systematic, cross-country information
on political institutions within easy reach of researchers investigating comparative political
economy. We have found no other source of information that provides as much objective
data on these issues. Others, including some of the authors of this paper, have previously used
in other work the data sources relied upon to construct this data these variables (see Clague, et
al., 1996, or Durham, 1999). However, we have found no case where all of the information
has been compiled, and we have identified many variables that have never been collected.
Examples of these are the variables identifying party orienitation, party affiliation with the
government or opposition, the military affiliation of the executive or minister of defense.
The early part of the paper outlined a subset of the questions that the Database of
Political Institutions can potentially illuminate. They stretch across political science and
economics, ranging from the determinants of democratic consolidation, the political
conditions for economic reform, and the political and institutional roots of corruption, to the
appropriate and institutionally-sensitive design of economic policy recommendations. To
enhance the availability of the DPI for these purposes, it will be placed on several websites,
particularly the political economy website of Harvard University and the website of the
World Bank's Development Economics Data Group.



40
Annex 1: List of Variables
The Database on Political Institutions (DPI) is a large new cross-country database of political
institutions that covers 177 countries over 21 years, 1975 - 1995. This database was prepared
by Thorsten Beck, George Clarke, Robert Cull, Alberto Groff, Philip Keefer, and Patrick
Walsh of the Development Research Group of the World Bank. When using this database,
please cite the current paper.
General Variables
Country Code COUNTRYC
International Financial Statistics code IFS
Year YEAR
Chief Executive Variables
Direct Presidential (0); strong president elected by assembly (1); Parliamentary (2) SYSTEM
How many years in office YRSOFFC
Is there a finite term in office? (1 if yes, 0 if no) FINIT'ITRM
Years left in curent term YRCURNT
If there are fonnal restraints on his term (NA if not), can he serve multiple terms? MULTPL?
Is Chief Executive a military officer? MILITARY
Is defense minister a military officer? DEFMIN
President got what % of votes in the 1st/only round? PERCENT1
President got what % of votes in the final round? PERCENTL
Party of chief executive has been how long in office PRTYIN
Chief Executive's Party
Name of party, if any EXECME
Right (R); Left (L); Center (C); Not applicable (0) EXECRLC
Nationalist (1 if yes) EXECNAT
Rural (1 if yes) EXECRURL
Regional (1 if yes) EXECREG
Religious (1 not specified, CH:Christian, CA: Catholic, IS: Islamic, HD: Hindu, BD: Buddhist, JW:
Jewish, 0: otherwise) EXECREL
Time since formation under this name EXECAGE
Does party of executive control all relevant houses? ALLHOUSE
In systems with both non-ceremonial PM and President, what is the party affiliation of the one not called Chief
Executive? NONCHIEF
Party Variables in the Legislature
Herfndahl Index for Govt. HERFGOV



41
Government fractionalization (chance that two random draws will produce legislators from different parties).
GOVFRAC
# of Govt. Seats NUMGOV
Largest Government Party
Name 1GOVME
Seats IGOVSEAT
Right (R); Left (L); Center (C); Not applicable (0) IGOVRL'C
Nationalist (1 if yes) 1GOVNAT
Rural (1 if yes) IGOVRURL
Regional (1 if yes) IGOVREG
Religious (1 not specified, CH:Christian, CA: Catholic, IS: Islamic, HD: Hindu, BD: Buddhist, JW:
Jewish, 0: otherwise) IGOVREL
Time since formation under this name IGOVAGE
2nd Largest Government Party
Name 2GOVME
Seats 2GOVSEAT
Right (R); Left (L); Center (C); Not applicable (0) 2GOVRLC
Nationalist (1 if yes) 2GOVNAT
Rural (1 if yes) 2GOVRURL
Regional (1 if yes) 2GOVREG
Religious (1 not specified, CH:Christian, CA: Catholic, IS: Islamic, HD: Hindu, BD: Buddhist, JW:
Jewish, 0: other-wise) 2GOVREL
Time since formation under this name 2GOVAGE
3rd Largest Government Party
Name 3GOVME
Seats 3GOVSEAT
-Right (R); Left (L); Center (C); Not applicable (0) 3GOVRI.C
Nationalist (1 if yes) 3GOVNAT
Rural (1 if yes) 3GOVRURL
Regional (1 if yes) 3GOVREG
Religious (1 not specified, CH:Christian, CA: Catholic, IS: Islamic, HD: Hindu, BD: Buddhist, JW:
Jewish, 0: otherwise) 3GOVREL
Time since formation under this name 3GOVAGE
Number of other government parties GOVOTH
Number of other government party Seats GOVOTHST
Herfindahl Index for Opposition HERFOPP



42
Opposition fractionalization (chance that two random draws will produce legislators from different parties).
OPPFRAC
# of Opp. Seats NUMOPP
Largest Opposition Party
Name Largest Opposition Party 1OPPME
Seats Largest Opposition Party 1OPPSEAT
Right (R); Left (L); Center (C); Not applicable (0) Largest Opposition Party 1OPPRLC
Nationalist (1 if yes) IOPPNAT
Rural (1 if yes) IOPPRURL
Regional (1 if yes) lOPPREG
Religious (1 not specified, CH:Christian, CA: Catholic, IS: Islamic, HD: Hindu, BD: Buddhist, JW:
Jewish, 0: otherwise) IOPPREL
Time since formation under this name lOPPAGE
2nd Largest Opposition Party
Name 2nd 2OPPME
Seats 2nd 2OPPSEAT
3rd Largest Opposition Party
Name 3rd 3OPPME
Seats 3rd 3OPPSEAT
Number of other opposition parties OPPOTH
Number of other opposition party seats OPPOTHST
Number of Parties non-aligned/allegiance unknown ULPRTY
Seats non-aligned/allegiance unknown NUMUL
Herfindahl Index Total HERFTOT
Total fractionalization (chance that two random draws will produce legislators from different parties) FRAC
Does one-opp. party have majority in House OPPMAJH
Does one opp. party have majority in Senate OPPMAJS
Details on Legislature
When were legislative elections held (number records month, 13 indicates unknown month) DATELEG
When were executive elections held (number records month, 13 indicates unknown month) DATEEXEC
Majority (number of government seats divided by total seats) MAJ
Average Age of Parties PARTYAGE
Is there a legislative election? ("1" if yes) LEGELEC
Is there an executive election? ("1" if yes) EXELEC
Is executive party special interest? EXECSPEC



43
Is 1st gov. party special interest? GOVSPEC
Are any coalition parties special interest? COALSPEC
Electoral Rules
Legislative Index of Political Competitiveness LIEC
Executive Index of Political Competitiveness EIEC
mean district magnitude (House) MDMH
mean district magnitude (Senate) MDMS
Number of senate/ (number of house + number of senate) S/S+H
Plurality? (1 if yes, 0 if no) PLURALTY
Proportional Representation? (1 if yes, 0 if no) PR
If Plurality and Proportional Representation which governs the majority/all of the House seats? (1 if Plurality,
O if Proportional) HOUSESYS
If Plurality and Proportional Representation which governs the majority/all of the Senate seats? (1 if Plurality,
O if Proportional) SENSYS
What is the vote threshold for representation? THRESH
Is the D'Hondt system used? (1 if yes, 0 if no) DHONDT
Are closed lists used? (1 if yes, 0 if no) CL
Candidate selection: national(1), sub-national(2), primary (3) SELECT
Were vote fraud or candidate intimidation serious enough to affect the outcome of elections? FRAUD
Stability and Checks &: Balances
Longest tenure of a veto player TENLONG
Shortest tenure of a veto player TENSHORT
Tenure of system of government if democratic; tenure of chief executive otherwise TENSYS
Index of Political Cohesion (based on Roubini, Sachs 1989) IPCOH
Number of veto players CHECKS I
Checks-2- Checki, + 1 for each veto player whose orientation is closer to opp than gov CHECKS2
Maximum difference of orientation among Govt. parties (0-2) POLARIZ
% veto players dropping from Govt., assume Senate changes STABS
% veto players dropping from Govt., assume Senate doesn't change STABNS
Federalism
Are there autonomous regions? AUTTON
Are the municipal governments locally elected? MUNI
Are the state/ province governments locally elected? STATE
Do Sub-national govts. have extensive taxing, spending or regulatory authority? AULIOR
Are the constituencies of the senators the states/provinces? STCONST



44
References
Agora Telematica website of political parties. Wilfried Dierksen, editor.
(www.agora.stm.it/elections/parties.htm.)
Alesina, Alberto, Sule Ozler, Nouriel Roubini, and Philip Swagel (1996). "Political Instability
and Economic Growth." Journal of Economic Growth 1 (une).
Banks, Arthur S. The Political Handbook of the World, 1986-1996. CSA Publications.
Binghampton, 1986-1996.
Clague, Christopher, Philip Keefer, Stephen Knack and Mancur Olson (1996). 'Property and
Contract Rights under Democracy and Dictatorship." The Journal of Economic Growth
1:2, 243-276 (une).
Collier, Paul and Anke Hoeffler (1999). "Justice-seeking and loot-seeking in civil war.'
Working paper, Development Research Group, The World Bank.
Cox, Gary W. (1997). Making Votes Count: Strategic Coordination in the World's Electoral
Systems. New York: Cambridge.
and Mathew D. McCubbins (forthcoming). "The Institutional Determinants of
Economic Policy Outcomes," in Structure and Policy in Presidential Democracies,
Stephan Haggard and Mathew D. McCubbins, eds.
Dixit, Avinash and John Londregan (1996). "The Determinants of Success of Special Interests
in Redistributive Politics." 7Te journal of Politics 58:4, 1132-55.
Durham, J. Benson (1999). "Economic Growth and Political Regimes." Journal of Economic
Growth 4, 81 - 111 (March).
The Europa Year Book, Vols. 1 & 2, 1977-1984, 1990-1994. Europa Publications Ltd.
London, 1977-1984,1990-1994.
Ferree, Karen E. and Smita Singh (1999). "Institutional Change and Economic Performance
in Africa, 1970-1995." Paper presented at the 1999 meetings of the American Political
- Science Association, Atlanta, Georgia.
Gurr, Ted Robert, Keith Jaggers and Will H. Moore (1998). Polity III Codebook. Center for
Comparative Politics, Department of Political Science, University of Colorado,
Boulder. Polity m available through the Inter-University Consortium for Political
and Social Research.
Heinisz, Witold (1997). "The Institutional Environment for Growth." Mimeo, The Wharton
School.



45
Hunter, Brian, current editor (issues back to 1864). The Statesman's year-book [Statistical and
historical annual of the states of the worldfor the year. J. New York: St. Martin's Press,
International Parliamentary Union, PARLINE database website.
(www.ipu.org/parline-e/parlinesearch.asp)
Keefer, Philip (1999). 'Political institutions and crisis: The effects of political checks and
balances on the dynamics of financial sector distress." Paper prepared for the meetings
of the Western Economics Association, July 1999.
and Stephen Knack (1997). "Why don't poor countries catch up? A cross-national test
of an institutional explanation.' Economic Inquiry 35:3, 590-602 July).
and David Stasavage (1999). "Bureaucratic Delegation and Political Institutions: When
are Independent Central Banks Irrelevant?" Mimeo, The World Bank (November).
Keesing's record of world events (various years). London: Longman, (weekly).
Lane, Jan-Erik, David H. McKay and Kenneth Newton (various issues). Political data
handbook: OECD countries. New York: Oxford University Press.
Laver, Michael and Kenneth Shepsle (1996). Making and Breaking Governments. New York:
Cambridge University Press.
London Institute of World Affairs (1947 - 84). The year book of world affairs. London:
Stevens.
North, Douglass and Barry Weingast (1989). "Constitutions and Commitment: The
Evolution of Institutions Governing Public Choice in Seventeenth Century England."
Journal of Economic History 49:4, 808-32 (December)
Political Parties ofAfrica and the Middle East: a Reference Guide. Longman Current Affairs
Series. Detroit, 1993.
Politica4Parties of Eastern Europe, Russia, and the Successcor States: a Reference Guide.- -Longman
Current Affairs Series. Harlow, UK, 1994.
Pritchett, Lant and Jonah Gelbach (1997). "More for the Poor is Less for the Poor: The
Politics of Targeting." The World Bank, Policy Research Working Paper, No. 1799.
Rasmusen, Eric and Mark J. Ramseyer (1994). "Cheap bribes and the corruption ban: a
coordination game among rational legislators." Public Choice 78:305-27 (March).
Roubini, Nouriel and Jeffrey Sachs (1989). "Political and Economic Determinants of Budget
Deficits in the Industrial Democracies." European Economic Review 33, 903-938.
Schady, Norbert (1998). "Seeking Votes: The Political Economy of Expenditures by the



46
Peruvian Social Fund: (FONCODES), 1991-95." Policy Research Working Paper
2166, The World Bank.
Shugart, Mathew Soberg (1995). "The Electoral Cycle and Institutional Sources of Divided
Presidential Government." American Political Science Review 89:2, 327 - 343 June).
and John Carey (1992). Presidents and Assemblies: Constitutional Design and Electoral
Dynamics. New York: Cambridge University Press.
Stepan, Alfred and Cindy Skach (1993). "Constitutional Frameworks and Democratic
Consolidation: Parliamentarism versus Presidentialism." World Politics 46, 1-22
(October).
United States Central Intelligence Agency, Directorate of Intelligence (various issues).
Handbook of the nations: a bMef/guide to the economy, government, land, demographics,
communications, and national defense establishment of each of 223 nations and other
political entities. Detroit: Gale, 1987.
Weingast, Barry (1998). 'Political Stability and Civil War: Institutions, Commitment, and
American Democracy." In Analytic Narratives, Robert Bates, Avner Greif, Margaret
Levi, Jean-Laurent Rosenthal, and Barry R. Weingast, eds. (Princeton: Princeton
University Press).






Policy Research Working Paper Series
Contact
Title                            Author                  Date              for paper
WPS2265 The Political Economy of Distress    Paola Bongini        January 2000       R. Vo
In East Asian Financial Institutions   Stijn Claessens                     33722
Giovanni Ferri
WPS2266 The Impact of Adult Deaths on     Martha Ainsworth        January 2000       S. Fallon
Children's Health in Northwestern    Innocent Semaii                       38009
Tanzania
WPS2267 Do High Interest Rates Defend     Aart Kraay              January 2000       R. Bonfield
Currencies during Speculative Attacks?                                     31248
WPS2268 The Structure of Social Disparities    Deon Filmer        January 2000       S. Fallon
In Education: Gender and Wealth                                            38009
WPS2269 Context Is Everything: Measuring    Nauro F. Campos       January 2000       J. Victor
Institutional Change in Transition                                         36549
Economies
WPS2270 The Optimal Income Tax When       Waly Wane               January 2000       H. Sladovich
Poverty Is a Public "Bad"                                                  37698
WPS2271 Corporate Risk around the World   Stiin Claessens         January 2000       R. Vo
Simeon Djankov                            33722
Tatiana Nenova
WPS2272 Ownership versus Environment:     Ann P. Bartel           January 2000       S. Fallon
Disentangling the Sources of Public  Ann E. Harrison                       38009
Sector Inefficiency
WPS2273 The Value of Preventing Malaria   Maureen L. Cropper      January 2000       T. Tourougui
In Tembien, Ethiopia             Mitiku Haile                              87431
Julian A. Lampietti
Christine Poulos
Dale Whittington
WPS2274 How Access to Urban Potable Water  Anqing Shi             January 2000       P. Sintim-Aboagye
and Sewerage Connections Affects                                           37644
Child Mortality
WPS2275 Who Gained from Vietnam's Boom    Paul Glewwe             January 2000       P. Sader
In the 1990s? An Analysis of Poverty Michele Gragnolati                    33902
An Analysis of Poverty and       Hassan Zaman
Inequality Trends
WPS2276 Evaluating the Case for Export    Arvind Panagariya       January 2000       L. Tabada
Subsidies                                                                  36896



Policy Research Working Papeir Series
Contact
Title                            Author                  Date              for paper
WPS2277 Determinants of Bulgarian Brady   Nina Budina             January 2000       N. Budina
Bond Prices: An Empirical        Tzvetan Mantchev                          82045
Assessment
WPS2278 Liquidity Constraints and Investment Nina Budina          January 2000       N. Budina
in Transition Economies: The Case    Harry Garretsen                       82045
of Bulgaria
WPS2279 Broad Roads in a Thin Country:    Andres G6mez-Lobo       January 2000       G. Chenet-Smith
Infrastructure Concessions in Chile  Sergio Hinojosa                       36370
WPS2280 Willingness to Pay for Air Quality  Hua Wang              January 2000       R. Yazigi
Improvements in Sofia, Bulgaria  Dale Whittington                          37176
WPS2281 External Sustainability: A Stock  C6sar Calder6n          January 2000       H. Vargas
Equilibrium Perspective          Norman Loayza                             38546
WPS2282 Managing Fiscal Risk in Bulgaria   Hana Polackova Brixi   January 2000       L. Zlaoui
Sergei Shatalov                           33100
Leila Zlaoui