Resumé

The present article follows up a previous study (Anckar 1998) which showed a strong association between size and party system fragmentation. The aim of the article is to see whether the explanatory power of size can, in fact, be attributed to a federal form of government. 77 countries with free party systems constitute the research population. The dependent variable has three components: number of parties, electoral support for the leading party, and the 'effective number of parties' calculated according to the Laakso-Taagepera formula. Preliminary tests reveal that federal states have a more fragmented party system than unitary states. However, when controlling for size, electoral system, the 'effective threshold,' and presidentialism, the results clearly show that federalism is overshadowed by size and also, to a lesser extent, by the effective threshold.

Introduction

The aim of the present study is to extend some important arguments put forward in my doctoral dissertation (Anckar 1998), which focused on the determinants of party system fragmentation The results clearly showed that there was a strong relation between the size of countries and the fragmentation of the party system. When I had the opportunity to discuss these results with Professor Juan Linz at the XVIIth IPSA meeting in Seoul in August 1997, he suggested that I ought to 'try federalism' as a plausible explanatory factor for the strong relation I had detected between size and party system fragmentation. The aim of the present study is consequently to see if the explanatory power of size on party system fragmentation can be attributed to a presumed relation between size and federalism.

A Theoretical Framework

Size and Party System Fragmentation

The theoretical frames for why the party system should be affected by the
size of entities were outlined by Dahl and Tufte in their seminal work Size

---

* Carsten Anckar, Department of Business and Public Administration, Mid-Sweden University, Fabriksgatan26,S-851 70 Sundsvall, Sweden. E-mail: Carsten.Anckar@fof.mh.se

and Democracy (1973). According to Dahl & Tufte, there are two ways in which size affects the fragmentation of the party system. One can be derived from social psychological theory building whereas the other one has to do with diversity. Dahl & Tufte take as a point of departure the famous experiments on conformity conducted by S. E. Asch. With regard to these experiments, the authors stipulate that a dissenting minority in a political system must exceed a certain level in order to be able to express their differing views (Dahl & Tufte 1973, 89-91).

The Asch experiments additionally constitute the base for an assumption about political conflicts in small and large units. According to Dahl & Tufte, it is reasonable to assume that in small units there is a high risk that such conflicts turn into personal conflicts, because people know each other well and it is difficult to, as Dahl & Tufte (1973, 93) put it, 'take refuge in anonymity.' In large units, however, it is possible that the antagonists never actually meet. This means that group conflicts should occur more frequently in large units than in small ones, since individuals presumably are less likely to take part in a conflict if this means that their personal relations are affected.

Consequently, Dahl & Tufte reach the conclusion that conflict solving mechanisms tend to be institutionalized in large units, whereas in smaller units one can assume that the parts tend to solve their disputes ad hoc, e.g., with face-to-face negotiations. The common conflict solving mechanism is, of course, the political party. The arguments laid out above lead to the conclusion that large units have more political parties than small units and that in large units the electoral support of the parties is more equally distributed than in small units (ibid., 91-103).

Another reason large units should have a more fragmented party system is that largeness, by necessity, leads to organizational diversity and complexity. large unit presupposes a wide range of organizations and institutions, a small one can manage with fewer ones. Although these organizations and institutions structurally might be very similar, one can assume that a multiplicity of such organizations and institutions leads to a higher degree of specialization and complexity. In other words, increasing size leads to increasing attitudinal diversity (ibid., 31-35). The abovementioned is prevalent in the writings of other authors as well. Already in 1956, Griffith (1956, 102) argued that 'democracy is more likely to survive, other things being equal, in small states,' because small states were more 'manageable' than large ones. Similar arguments can be found in Lijphart's discussion on favorable conditions for consociational democracy: '. . . smaller countries are easier to govern than large ones because they tend to be less complex: the number and variety of groups and individuals whose interests and attitudes have to be taken into consideration are fewer' (Lijphartl977,6B).

Federalism and Party System Fragmentation

A curiosity is that while there seems to be broad consensus on which countries should be considered federal, authors at the same time generally agree that the term is problematic to conceptualize. It has, for instance, been said that federalism 'is a difficult term to define' (Paddison 1983, 98) and that 'it is not easy to pin down what federalism stands for except in a formal constitutional sense' (Lane & Ersson 1994, 164). Nevertheless, it should not be venturesome to simply state that federalism is a strong form of power devolution.

One reason for why federal states should have a more fragmented party system than unitary states is that federal states tend to be highly heterogeneous different respects. Indeed, regional differences are frequently mentioned as the exogenous factor when explaining the adoption of a federal constitution (e.g. Paddison 1983, 104-7; D. Anckar 1997, 48). Federalism is a mechanism which should make conflicts created by such diversities manageable. Regional differences can, of course, take many forms. If they are severe - and since the constitution makers have opted for a federal system, we must assume that they are severe - they can be thought to be reflected in the party system, thereby giving rise to a multiplicity of parties. The heterogeneity of federal countries is therefore likely to generate political parties with a basis of support anchored to a specific ethnic or religious or to a specific region.

Another defining characteristic of federal systems of explicit relevance for the study is the presence of an elected legislative assembly at the lower level. The division of legislative power between the federal level and the regional level should constitute a favorable condition for the creation and success of regional parties. With elections to a regional assembly with substantiallegislative one can assume that there is a special niche for regionally concentrated parties. In different regions, different interests prevail, shaping parties with different objectives and different bases of support. Since the power of the legislature at the subnational level is substantial, there is also supposedly great incentives to found political parties with the ambition to operate solely at the local level, concentrating mainly, or even exclusively, on questions concerning local matters. At the same time, it is reasonable to assume that regional parties are perceived as legitimate and relevant political actors by the electorate. This will possibly have a spill-over effect on national elections as well. When voters grow accustomed to voting for regional parties in regional elections, the threshold for voting for a regional party in national elections is lowered. Consequently,parties a regional base should have a better chance to survive in federal states than in unitary ones. True, elections to subnational representative institutions do take place in unitary states as well. However.

in a federal system the price of gaining representation is generally much more awarding. By gaining representation at the regional level, a party can, at least to a certain extent, carry some weight when it comes to shaping national policy outcomes. In other words, while there is only one carrot to win in unitary states, parties in federal states compete for two carrots.

One can also find other arguments that link federalism to party system fragmentation. Federalism tends to go hand in hand with bicameralism. In fact, in my research population (to which I shall return shortly) there is only one exception from this rule, The Federated States of Micronesia. Combining based on population and representation based on territory means that at the national level there are elections to two representative bodies instead of only one. Therefore, it is not unlikely that vote splitting might occur. A voter who favors a regional party, but is uncertain whether the party in question will gain representation or carry any weight in national politics, might as a compromise choose to split his vote between a large, national party, which presumably will carry some political weight, and a regional party which politically attracts him the most.

The theoretical arguments laid out above suggest that ethnic, religious or regional differences and, in particular, an elected legislative assembly at the lower level are of importance for party system fragmentation. In this study, a state is consequently regarded as federal if and only if it a) formally regards itself as federal (thereby explicitly expressing that there are serious differences different parts of the country) and b) has an elected regional legislature. This means that I am operating with a qualitative and not a quantitative variable. The strategy can of course be criticized on the ground that I disregard the fact that there are significant differences within the categories. True, Spain is a more decentralized state than France. However, since the operationalization of decentralization is problematic, and for the validity of the study also dangerous, I choose to operate with only two categories which can be easily and naturally separated from each other.

The Research Population

Since the number of countries that fulfil my two criteria for a federal system is fairly low, it is natural to include as many countries as possible in the study. Some restrictions do, however, apply. Needless to say, it is important that only countries in which the formation of political parties is free can be included in the study. As a departure point, I use the meritorious study Democracy and Development by Hadenius (1992). The author measures the level of democracy according to two dimensions: elections and political freedoms. Each dimension covers a field of subdimensions. Here, I use only those subdimensions that are of explicit relevance to the formation of

parties, namely universal suffrage, meaningful elections, and organizational freedoms. For each dimension, states are given points on various scales. On the elections dimension, the scale ranges from 0 to 12 points. I have chosen to include countries obtaining the value 6 or more. On the organizational freedoms dimension, the scale ranges from 0 to 8. Countries obtaining the value 5 or more are included. Concerning universal suffrage, Hadenius measures it according to the following formula: the proportion of the population that is enfranchised times the proportion of seats in parliament that are filled by public elections. I choose a slightly different strategy, excluding all countries that impose restrictions on the right to vote. In addition, I demand that at least 90 percent of the representatives are elected by general elections.

Since Hadenius' study does not reflect the present situation (his data are from 1988), and does not cover all the countries in the world (OECD countries and Eastern European countries are excluded), I must also use other sources. I have therefore chosen to make use of Freedom House's renowned compilations of political rights and civil liberties. Both dimensions are measured on a scale from 1-7, where values close to 1 tell us that the degree of rights or liberties is high, and values approximating 7 denote a lack of such rights and liberties. For the present study, the political rights dimension is more relevant than the civil liberties dimension. I therefore focus on this dimension only, using the value 5 as a cutting point, and only countries scoring below 5 are included in the study. Since the study covers a time span of approximately ten years, I demand that countries meet the criteria derived from Hadenius' study and from Freedom House's surveys on two different occasions namely the 1994/1995 issue and the latest issue (available on the Internet, http://www.freedomhouse.org/ political/fr table I .htm). For countries not included in Hadenius' study, I use the 1987/1988 issue of Freedom House's survey as a point of reference. Since a party system might be unstable at the initial stage of the democratization I also demand that there has been at least three consecutive elections carried out under democratic rule (before June 1996). Altogether, 77 countries meet the criteria laid out above. The countries are listed in Appendices 1 and 2.¹

Three Measures of Party System Fragmentation

A natural, albeit rough, estimate of party system fragmentation is the number of political parties in a country. Of course, defining a political partyis an easy task. According to Downs (1957, 27), a political party is 'a team of men seeking to control the governing apparatus by gaining office in a duly constituted election.' Lapalombara & Weiner (1966, 6) put forward four criteria for a political party:

(1) Continuity in organization - that is, an organization whose expected life spån is not dependent on the life spån of current leaders; (2) manifest and presumably permanent organization at the local level, with regularized Communications and other relationships between local and national units; (3) self-conscious determination of leaders at both national and local levels to capture and to hold decision-making power alone or in coalition with others, not simply to influence the exercise of power; and (4) a concern on the part of the organization for seeking followers at the polis in sorae manner striving for popular support.

Epstein (1967, 9) gave a loose definition, stating that 'any group, however loosely organized, seeking to elect governmental officeholders under a given label . . . rather than an organization is the crucial element.' Sjöblom (1968, 21) defines a party as 'an organization that appoints candidates at general elections to the system's representative assembly and also to other political positions,' whereas Sartori's (1976, 64) 'minimal' definition states that 'a party is any political group that presents at elections, and is capable of placing through elections, candidates for public office.'

Although there are some differences between the quotations above, there seems to be broad consensus that political parties strive for power by placing candidates for election. This is the crucial point when drawing a line between parties and interest organizations. When operating with a large number of countries, it is important that the same criteria apply for each country. Fortunately, there is a useful source, namely the handbook-series published by Longman Current Affairs (see Appendix 2). The source uses self-definition as a defining criterion, although 'editorial judgments have had to be made on case-by-case basis' (Lewis & Sagar 1992, vii).

Another measure of party system fragmentation is the support for the leading party in parliamentary elections. The larger the support, the more homogeneous the country in terms of political attitudes. However, since this measure does not take into account the overall degree of fragmentation - the vote shares of all other parties than the largest one are disregarded - it is necessary to complement this measure with one that takes into account the vote shares of all parties obtaining votes in general elections. The most widely used measure is 'the effective number of parties,' created by Laakso & Taagepera (1979) and calculated according to the following formula:

where Vt is the vote proportion of the i-th party.

I consequently operate with three dependent variables: number of registered parties in a country, electoral support for the leading party, and effective number of parties. The results will show if and how these three measures of party system fragmentation are affected by variations in the independent variables. Electoral data have been collected from the two

latest elections. The mean value constitutes the electoral support for the leading party and the effective number of parties. The latest data are from May 1996. For countries with two-chamber parliaments, I have regarded the elections to the lower chamber only. Concerning the number of parties, I have for each region used the latest version of the Handbook series publishedby Current Affairs.

Empirical Evidence

We find that of the 77 countries with free party systems, 15 are federal, while 62 are unitary. To test whether or not federalism is linked to party system fragmentation, I shall first conduct a comparison of means test, using Tstatistics. concerning electoral support for the leading party and the effective number of parties are not available for Kiribati and the Solomon Islands. In addition, Papua New Guinea is excluded from all analyses concerning effective number of parties. In Papua New Guinea, a very high number of candidates usually run as independents. This means that the effective number of parties will be very high indeed (the calculations have been done presuming that all independents do not constitute one homogeneous but instead are 'independent' from each other as well as other parties). Papua New Guinea would therefore be an extreme 'outlier' in the data.

Let us first establish how size, federalism, and party system fragmentation interrelated. We already know that there is a strong link between size and party system fragmentation (Anckar 1998). Accordingly, we must establish whether a corresponding association between federalism and party system fragmentation exists and whether size and federalism are interrelated. If this is not the case, the conclusion is that federalism does not account for the explanatory power of size on party system fragmentation. Tables 1 and 2 report empirical findings on these relationships. Since the two size dimensions as well as the number of parties are distributed lognormally, I use the logaritmized versions of population and area as well as of the number of parties.

As can be seen, size is, no doubt, related to federalism in the sense that federal states tend to be much larger than unitary ones. This holds true concerning both population and area. The results furthermore indicate evidence of a link between federalism and party system fragmentation. There is a strong association between state structure and number of parties. The number of parties is much higher in federal states than in unitary ones. Electoral support for the leading party is not related to state structure. As concerns the effective number of parties, the result is not statistically significant. it does point in the presumed direction.

Contesting Variables

So far, I have dwelled upon the presumed relation between size, federalism, and party system fragmentation. In addition to size and federalism, a number of factors are frequently cited as relevant for party system formation. study of the determinants of party system fragmentation is not complete without the inclusion of these factors. Consequently, I shall introduce the following contesting variables: electoral system, effective threshold, and presidentialism.

The Electoral System

The effects of electoral systems on party systems have received much
attention in the literature. Normally, Hermens (1941) and Duverger (1951)
are mentioned as the pioneers in the field, although Droop (cited in Riker

1986, 22-23) preceded them both. Duverger's biggest merit was that he was the first to present clear theoretical arguments for why and how different electoral systems affect party systems. In short, his argument was that a plurality electoral system leads to a two-party system, whereas a proportionalelectoral leads to the formation of a multi-party system.

Duverger gave two reasons for why the formation of a two-party system in a plurality electoral system is inevitable. First, there is the mechanical effect, which is associated with the rules that regulate the transformation of shares of votes to shares of seats. In plurality electoral systems, these rules favor, to a large extent, the leading party in each constituency, as the party with the largest share of the votes gets 100 percent of the seats (e.g., the only seat in a single-member constituency). At the same time, its share of votes can be the lowest possible as long as it is higher than the share of votes of any other party. Second, there is also a psychological effect. When voters are aware of how the mechanical effect operates, they tend to concentrate their votes to the large parties to avoid wasting the vote, i.e., give the vote to a party that lacks realistic possibilities of getting its representative elected (Duverger 1964, 206-55). In this study, the electoral system is treated as a dummy variable, countries with a plurality electoral system are given the value 0, and countries with a proportional electoral system obtain the value I.²

The Effective Threshold

There is strong evidence supporting a link between district magnitude and party system fragmentation. The argument is that the larger the number of persons elected from an electoral district, the higher the degree of proportionality the better the possibilities for representatives of small parties of getting elected (e.g., Rae 1971; Taagepera & Shugart 1989, 112-25; Gallagher 1991; Jones 1993, 64-66; Lijphart 1994). The legal thresholds used in some countries to make it more difficult for small parties to gain representation have the same presumed effect on party system fragmentation small electoral districts. Legal thresholds can be expressed as a certain percentage of the votes or, alternatively, a fixed number of votes or some other criteria. In most cases, the threshold operates at the national level, but in some countries the threshold operates on the district or regional level (Lijphart 1994, 11-12).

The effects of district magnitude and legal thresholds cannot be assessed separately. Often, countries with large district magnitudes use legal thresholdsas countermeasure to the proportionality obtained by the large district magnitudes. Some countries even combine low district magnitudes with high legal thresholds (e.g., Turkey). We are in both cases unable to state which factor carries more explanatory power (Anckar 1997, 507-8).

District magnitudes and legal thresholds are therefore combined into one single measurement, the effective threshold. This measurement was first suggested by Taagepera & Shugart (1989, 273-75) and later on modified by Lijphart (1994, 25-30). In short, the difference between the two versions is that while Taagepera & Shugart's measurement converts legal thresholds into district magnitudes, Lijphart's measurement does the opposite. In this study, I use Lijphart's measurement, which is calculated according to the following formula:

where M stands for district magnitude.

For countries that employ legal thresholds, the value obtained by Lijphart's formula is compared to the legal threshold. The higher value constitutes the effective threshold. Calculation of the effective threshold is problematic in countries with multi-tier districting, i.e., where seat allocation takes place in two (or, as in Greece, three) levels. Things get even more complicated in countries with multi-tier districting and legal thresholds calculated in more complex ways. Space prohibits a long discussion of how thresholds should be calculated in these cases. Let me therefore only say that I have chosen the same strategy as Lijphart (1994, 30-39). I have only made one slight alteration. Lijphart calculates the effective threshold based on the electoral district where a party most easily can gain a seat. I choose to make the assumption that all districts are equally large, and that the votes are distributed equally between the parties throughout all electoral districts. The motivation for this strategy is that the threshold is relevant for small parties, which are often regionally concentrated and not necessarily to the largest electoral districts (i.e., mostly the capital district). In fact, it is not unreasonable to assume that these parties have their base of support in distant parts of the country. In some countries, the district magnitude or legal thresholds have changed between the two elections from which data have been collected. In these cases, I operate with the average value.

Establishing the effective threshold is more problematic for countries using plurality electoral systems. Based on a reasonable assumption that four to five parties put forward candidates in each constituency, Lijphart (1994, 28-29) a rough estimate of 35 percent for plurality electoral systems. In the present study, this value will constitute the effective threshold for countries with plurality electoral systems.

Presidentialism

There is some evidence of a relation between a presidential form of governmentand
system fragmentation. According to Shugart and Carey

(Shugart 1988; Shugart & Carey 1992, 206-58), a parliamentary system is likely to have a more fragmented party system than a presidential one. Their results have been confirmed by other authors (Stepan & Skach 1994, 121; Mainwaring 1993, 210-14). The argument is that since a presidential election often is a race between two candidates only, one can assume that this might have the same consequences on voting behavior as a plurality electoral system. The presidential election is thereafter thought to have a 'spill-over' effect on the parliamentary election, resulting in a concentration of the votes to a few parties. The empirical evidence shows that this is the case only if a) the presidential election is conducted under a plurality voting system; b) the election of the legislature coincides with the election of the president; and c) the president is seen as a relevant political actor (Lijphart 1994, 130-34). To determine when a president should be seen as a relevant political actor, I use Sartori's definition of a presidential system. According to Sartori, a political system is presidential if 'the head of state (president) i) results from a popular election, ii) during his or her pre-established tenure cannot be discharged by a parliamentary vote, and iii) heads or otherwise directs the governments that he or she appoints' (Sartori 1994, 84).

The form of government is treated as a dummy variable, countries which meet all the above-mentioned criteria, and where the president is elected by plurality vote and the presidential elections coincide with parliamentary elections are regarded as presidential and given the value 1. Other countries are given the value 0. Some countries use mid-term elections, which means that one election from which data have been gathered has coincided with the presidential election, whereas the opposite is true for the other election. In these cases, I have given the countries the value 0.5 on the scale. The question of how to treat the so-called semi-presidential systems does not have to be tackled, since none of the countries with a semi-presidential form of government (Finland, France, and possibly Sri Lanka) meet all the other requirements for a presidential system.

Empirical Evidence

Table 3 shows the levels of association between the independent and the dependent variables for all the countries included in the study. In these regression models, state structure is treated as a dummy variable. Unitary states obtain the value 0, federal states the value 1. Since the two size dimensions are highly interrelated (Pearson's r .87), I use separate regression models for population and area. The same applies for the electoral system and the effective threshold. To avoid too many tables, I only show, for each dependent variable, the result of the regression analysis with the highest explanatory value (R²).

The results clearly show that the impact of federalism on party system fragmentation is negligible. The strong association between federalism and the number of parties vanishes when we control for population size. The explanatory power of the independent variables is approximately the same for all three dependent variables. In addition to size, the effective threshold is a powerful determinant of the effective number of parties. Also, we note that a presidential form of government reduces the effective number of parties to a certain extent. In addition to area, the electoral system affects the electoral support for the leading party.

The results obtained in Table 3 indicate that size is a far more important determinant of party system fragmentation than state structure. This is particularly the case concerning the number of parties. Concerning the other dependent variables, it is less clear whether the explanatory power of federalism can be attributed to size, since there are several other independent with substantial explanatory value. It is therefore necessary to establish whether size accounts for the explanatory power of federalism. The reader is reminded that the internal association between federalism and the electoral support of the leading party was very weak. Let us therefore run regression analyses where only size and federalism constitute the independent variables. The results are shown in Table 4 and leave little or no room for interpretation. Nearly all the explanatory power of federalism disappears when controlling for size. This holds true concerning all dependent

Conclusion

The findings leave little or no room for doubt. The explanatory value of
federalism is clearly overshadowed by size, which leads to the conclusion

that physical characteristics play a much more important role than institutionalfactors party system fragmentation. Of course, the reasons for this are hard to pinpoint. However, one plausible line of reasoning is that federal systems are likely to generate a special kind of parties, namely regional parties. Although these parties in some cases are relevant and even powerful actors on the political arena, for instance in Canada and India, voters still tend to regard ideological cleavages as more important than cleavages based on territorial or ethnic differences (e.g. Lewis-Beck 1986; McDonough et al. 1988, 204, 218; Dix 1989, 24; Rokkan & Urwin 1983, 154-56; Keating 1985). At the same time it should also be stressed that regionally concentrated parties do not exist exclusively in federal countries. Let me here only point at the important role the secessionist Lega Nord is playing in Italian politics.

---

NOTES

1. I make the following exceptions: Singapore has continuously received the value 4 on the political rights dimension. In the 1993/1994 issue of Freedom in the World Ibis value was raised to 5. However, in the latest issue, it was again lowered to 4. Electoral data for Singapore has, in any case, been collected prior to the raise. Turkey's values have been oscillating between inclusion and exclusion. In the 1992/1993 survey, Turkey received the value 2. The oppression of the Kurdish population led to a sharp rise and in the 1995 survey, Turkey received the value 5. In the latest ranking, however, Turkey is again barely qualified for inclusion, obtaining the value 4. I therefore choose to include Turkey in the study.

2. Countries which use majority electoral systems, i.e., either the second ballot system (France, Kiribati, Monaco) or the alternative vote system (Australia, Nauru), are, perhaps rather unconventionally, included among the countries with proportional electoral systems. This is because the possible implications of these electoral systems for party system fragmentation are closer to those of proportional systems than those of plurality systems (e.g., Duverger 1964, 240; Anckar 1998, 65-68). Countries using the single transferable vote system (Ireland and Malta) and the single non-transferable vote system (Japan until 1996, Taiwan, and Vanuatu) are also included among countries with proportional electoral systems. Some countries use mixed systems, where a part of the legislature is elected by proportional representation and the other part by means of plurality. In those cases where the voters have different ballots for the different elections, I have only regarded the proportional election. For the strategies adapted in those countries where voters have only one ballot, the reader is referred to Appendix 1.

REFERENCES

Anckar, C. 1997. 'Determinants of Disproportionality and Wasted Votes,' Electoral Studies
16,501-15.

Anckar, C. 1998. Storlek och partisystem. En studie av 77 stater. Åbo: Åbo Akademi
University Press.

Anckar, D. 1997. 'Federal and Bicameral Microstates: Gigantic Miniatures?,' in Anckar, D.
& Nilsson, L., eds., Politics and Geography. Contributions to an Interface. Sundsvall: Mid-
Sweden University Press.

Burdick, A. 1988. 'The Constitution of the Federated States of Micronesia,' in Ghai, Y.,
ed., Law, Politics and Government in the Pacific Island States. Suva: Institute of Pacific
Studies, University of the South Pacific.

Dahl, R. & Tufte, E. 1973. Size and Democracy. Stanford: Stanford University Press.

Downs, A. 1957. An Economic Theory of Democracy. New York: Harper & Row.

Duverger, M. 1964. Political Parties. Their Organization and Activity in the Modern State.
London: Methuen.

Epstein, L. 1967. Political Parties in Western Democracies. London: Pall Mali.

Freedom in the World: Political Rights and Civil Liberties. Various issues. New York: Freedom
House.

Gallagher, M. 1991. 'Proportionality, Disproportionality and Electoral Systems,' Electoral
Studies 10,33-51.

Griffith, E. 1956. 'Cultural Prerequisites to a Successfully Functioning Democracy: A
Symposium Preface,' American Political Science Review 50, 101-15.

Hadenius, A. 1992. Democracy and Development. Cambridge: Cambridge University Press.

Internet: http:/ /www.freedomhouse.org/political/frtablel .htm

Johnson, G. 1988. 'Marshall Islands: Politics in the Marshall Islands,' in Ghai, Y., ed., Law,
Politics and Government in the Pacific Island States. Suva: Institute of Pacific Studies,
University of the South Pacific.

Jones, M. 1993. 'Electoral Laws in Latin America and the Caribbean,' Electoral Studies 12,
59-75.

Jones, M. 1995. 'A Guide to the Electoral Systems of the Americas,' Electoral Studies 14,
5-21.

Keating, M. 1985. 'The Rise and Decline of Micronationalism in Mainland France,' Political
Studies 33, 1-18.

Laakso, M. & Taagepera, R. 1979. '"Effective" Number of Parties: A Measure with
Application to West Europe,' Comparative Political Studies 12, 3-27.

Lane, J.-E. & Ersson, S. 1994. Comparative Politics: An Introduction and New Approach.
Cambridge, MA: Polity Press.

Lapalombara, J. & Weiner, M. 1966. 'The Origin and Development of Political Parties,' in
Lapalombara, J. & Weiner, M., eds., Political Parties and Political Development. Princeton:
Princeton University Press.

Lewis, D. S & Sågar, D. J., eds. 1992. Political Parties of Asia and the Pacific. Harlow:
Longman Current Affairs.

Lewis-Beck, M. S. 1986. 'Comparative Economic Voting: Britain, France, Germany, Italy,'
American Journal of Political Science 30, 315-44.

Lijphart, A. 1977. Democracy in Plural Societies: A Comparative Exploration. New Haven:
Yale University Press.

Lijphart, A. 1994. Electoral Systems and Party Systems: A Study of Twenty-Seven
Democracies, 1945-1990. New York: Oxford University Press.

Mainwaring, S. 1993. 'Presidentialism, Multipartism, and Democracy. The Difficult Combination,'
Political Studies 26, 198-228.

McDonough, P., Barnes, S. H. & Lopez-Pina, A. 1988. 'Social Identity and Mass Politics in
Spain,' Comparative Political Studies 21, 200-30.

Paddison, R. 1983. The Fragmented State. The Political Geography of Power. Oxford: Basil
Blackwell.

Rae, D. 1971. The Political Consequences of Electoral Laws (2nd ed.). New Haven: Yale
University Press.

Riker, W. 1986. 'Duverger's Law Revisited,' in Grofman, B. & Lijphart, A., eds., Electoral
Laws and their Political Consequences. New York: Agathon Press.

Rokkan, S. & Urwin, D. 1983. Economy, Territory. Identity: Politics of West European
Peripheries. London: Sage.

Sartori, G. 1976. Parties and Party Systems. A Framework for Analysis. Cambridge:
Cambridge University Press.

Sartori, G. 1994. Comparative Constitutional Engineering: An Inquiry into Structures, Incentives
andOutcomes. London: Macmillan.

Shugart, M. 1988. Duverger's Rule, District Magnitude, and Presidentialism. Ph.d.diss. Irvine:
University of California.

Shugart, M. & Carey, J. 1992. Presidents and Assemblies. Constitutional Design and Electoral
Dynamics. Cambridge: Cambridge University Press.

Sjoblom, G. 1968. Party Strategies in a Multiparty System. Lund: Studentlitteratur.

Stepan, A. & Skach, C. 1994. 'Presidentialism and Parliamentansm in Comparative
Perspective,' in Linz, J. & Valenzuela, A., eds., The Failure of Presidential Democracy,
Vol. 1, Comparative Perspectives. Baltimore: The Johns Hopkins University Press.

Taagepera, R. & Shugart, M. 1989. Seats and Votes: The Effects and Determinants of Electoral
Systems. New Haven: Yale University Press.