12 - COMMENT: ELECTIONS IN AGRARIAN SOCIETIES (Tuỳ Vũ)

COMMENT: ELECTIONS IN AGRARIAN SOCIETIES

Joel D. Barkan

Joel D. Barkan is professor emeritus of political science at the Univer-sity of Iowa and senior associate at the Center for Strategic and International Studies in Washington, D.C. He has written extensively on democratization and economic reform in East Africa, and on Afri-can elections. His current work in this area focuses on the application of GIS technology to problems of electoral design and is reported in the November 2006 issue of the American Journal of Political Science.

This essay originally appeared in the October 1995 issue of the Journal of Democracy.

The recent transitions to democracy in Namibia and South Africa have prompted renewed claims by advocates of proportional representation (PR) that it is the most appropriate electoral formula for plural societies in Africa and elsewhere. Their argument, as presented by Andrew Reynolds (pp. 121–34 above) and earlier by Arend Lijphart, contains four essential elements.1 First, PR is the “fairest” method of electing members of national legislatures—particularly in comparison with for-mulas that elect legislators from single-member districts (SMDs)—because it ensures that the percentage of seats won by a po-litical party is virtually the same as its percentage of the popular vote. Parties receiving a plurality or a majority of the vote are therefore not overrepresented in the legislature, while parties receiving a minority of the vote gain representation equal to their voting strength. As a result, PR also protects minority interests.

Second, PR is inclusive, because it ensures that all significant play-ers in the political system, including potential spoilers, are represented in the legislature. PR thus enhances the prospects that all players will support the constitutional order by participating in its elections and principal institutions. Such support is particularly important during the early stages of new or recently restored democratic rule, when the insti-tutionalization of the new constitutional order is incomplete. Third, PR

136 Comment: Elections in Agrarian Societies

facilitates arrangements of power-sharing or consociational democracy, enabling most political forces to participate in governance. Fourth— and this last benefit results from the first three—PR greatly enhances the prospects for democratization in plural societies, in which political cleav-ages run deep and mirror ethnic, racial, linguistic, or religious divisions.

Although the recent experience of southern Africa appears to support the argument for PR, the advantages of this system—particularly for the rest of the continent and other similar societies—are not as clear-cut as they seem. In agrarian societies, in which the overwhelming majority of the people derive their livelihood from the land, PR does not produce electoral results that are significantly “fairer” or more inclusive than plurality elections based on SMDs. PR is not, therefore, an essential feature of consociational government, but rather one method of facili-tating power-sharing within the executive branch.

PR also has serious disadvantages, one of which should be of particu-lar concern to any would-be constitutional engineer for an agrarian soci-ety. Because under the purest forms of PR legislative seats are allocated from party lists according to each party’s proportion of the total national vote, individual MPs do not identify with, nor can they be held account-able to, the residents of a specific geographic constituency. Yet in agrar-ian societies, this lack of linkage between representatives and constituents greatly reduces the prospects for the consolidation of democratic rule.

Voting Behavior Among Rural Populations

In agrarian societies, with their low levels of occupational special-ization and class identity, most people define their interests and differentiate themselves from one another on the basis of where they live, rather than what they do. They have a strong attachment to the place where they reside and affection for their neighbors. When it comes to elections, they focus on the basic needs of their local community and surrounding region—whether they have adequate water, schools, and health-care facilities, whether there is a farm-to-market road, whether the producer price for the agricultural commodity grown in the area yields a fair return to local farmers, and so on. Inhabitants of a particular rural area usually have a common set of political interests, and they vote accordingly. This explains the high geographic concentration of the vote for competing parties in the recent round of multiparty elec-tions in Africa. Except in urban areas, whose inhabitants come from many different regions and tend to have a relatively strong sense of occupational and class identity, people who live in the same place vote for the same political party.2 In agrarian societies, people evaluate par-ties and candidates in terms of their potential for, or past record of, constituency service.3

Yet PR systematically frustrates such voter expectations. Not only

Joel D. Barkan

137

are MPs not responsible for addressing the needs of specific localities, but their political careers depend primarily on satisfying their party’s leadership, which determines their rankings on the party list for the next election. In agrarian societies, and particularly in Africa, the use of PR risks the development of what Goran Hyden has termed “the sus-pended state”—a state that is disconnected from the population and eventually loses its authority and its ability to govern.4

Electoral Formulas and Democratization

Democratization hinges on the simultaneous development of two different relationships: 1) representation of citizens by their chosen leaders, a relationship characterized by dialogue and accountability; and 2) tolerance, bargaining, and compromise among rival political groups. The former links elites and nonelites who have a common po-litical interest, and constitutes the “vertical” dimension of democracy. The latter obtains mainly between leaders of opposing interests, and constitutes the “horizontal” dimension of democracy.

The relative importance of these two dimensions is different during transitions to democracy and during democratic consolidation. Because transitions involve the establishment of a new constitutional order that must be embraced by all significant factions, they are marked by in-tense bargaining among rival elites. At this stage, relatively little em-phasis is placed on the quality of representation. It is more or less taken for granted that key elites are bargaining on behalf of their respective constituencies. For example, in the negotiations over South Africa’s transition to democracy, few questioned the extent to which Nelson Mandela and Cyril Ramaphosa, on the one hand, or F.W. de Klerk and Roelf Meyer, on the other, spoke for the rank and file of the African National Congress (ANC) and the National Party (NP), respectively.

Consolidation, on the other hand, is a long-term process during which democratic practice is gradually institutionalized. Indeed, it can be said that the process of consolidation never ceases, because for democracy to survive, successive generations of citizens and elites must reaffirm the nation’s commitment to it.5 Hence consolidation requires the con-tinuous fulfillment of both the vertical and horizontal dimensions of democracy, and any electoral formula that does not sustain both under-mines the prospects for democracy over the long term. It is clear, then, that PR has both advantages and disadvantages. While it can facilitate transitions to democracy in plural societies, when these societies are also agrarian it impedes consolidation by failing to sustain democracy’s vertical component.

The experience of southern Africa offers substantial evidence of PR’s primary shortcoming, and has given rise to calls for the system’s aboli-tion or modification. PR was used in Namibia for that country’s first all-

138 Comment: Elections in Agrarian Societies

race elections in November 1989. The elections were held to establish a Constituent Assembly, which became the National Assembly (or lower house of parliament) upon Namibia’s independence from South Africa in March 1990. Seventy-two seats were allocated on the basis of the total nationwide vote of the contesting parties—that is, the party vote in a single national constituency. The same variation of PR was used for Namibia’s second parliamentary elections in December 1994. In between, Namibians elected 13 regional councils (in November 1992) from 95 SMDs. Each council, in turn, elected two of its members to an upper house of parliament, the National Council. The outcomes are instructive.

The South West African People’s Organisation (SWAPO) obtained majorities of 57.3 percent and 73.9 percent in the first two parliamen-tary elections, and has governed since independence. The principal opposition party, the Democratic Turnhalle Alliance (DTA), obtained 28.6 percent of the vote in 1989 and 20.8 percent in 1994, while the United Democratic Front (UDF) obtained 5.6 percent and 2.7 percent. The use of PR resulted in an allocation of seats to each party that closely approximated that party’s percentage of the vote.

With more than two-thirds of the seats in the National Assembly, SWAPO has the power to amend Namibia’s constitution. Thus PR by itself has failed to provide sufficient minority representation to block constitutional changes that could be injurious to minority interests. Nor has PR guaranteed the inclusion of minority representatives in the ex-ecutive branch—the essence of consociational democracy. Rather, it is the self-restraint of the majority government of President Sam Nujoma, along with international pressure, that has been responsible for the pro-tection of minority interests—particularly the property rights of Namibia’s white population.6 Self-restraint and commitment to a policy of national reconciliation have also resulted in the appointment of op-position party members to minor posts in the government and to the position of attorney general.

Most significantly, Namibia’s ethnic and racial minorities, which supported the DTA and the UDF, would be no worse off and possibly better off under an electoral system based on SMDs. Because the vote in the 1994 parliamentary elections was tabulated and reported by each of the 95 constituencies that elected the regional councils, it is possible to estimate the distribution of seats that would have occurred had the elec-tions been run under an SMD formula. The results show SWAPO obtain-ing a plurality or majority of the vote in 67.4 percent of the 95 districts, the DTA obtaining a plurality or majority in 29.5 percent, and the UDF in 3.2 percent.7 The distribution of seats is thus substantially the same under the two systems. Advocates of PR claim that SMD plurality over-represents parties that obtain a plurality or majority of the vote; yet the Namibian case shows that it can also underrepresent the majority party, and thus may provide better protection of minority interests than PR.

Joel D. Barkan

139

For example, the DTA—the party of Namibia’s white community and ethnic minorities—obtained only 20.8 percent of the vote in the 1994 election, but would have “won” 29.5 percent of the seats had each coun-cilmanic district elected a representative to the legislature. Moreover, SMD systems can be fine-tuned so as to reduce (or maintain) the associ-ated disparity between vote shares and seats won, as explained below.

The Namibian case also demonstrates the failure of PR to provide appropriate representation for the residents of agrarian societies. Inter-views conducted by myself and two colleagues with more than three dozen regional councillors in July 1994 consistently revealed that mem-bers of the National Assembly rarely make an appearance in the rural areas, because they have no constituency to which they are account-able. It is the regional councillors—precisely because they are elected from geographically defined constituencies and are physically acces-sible within their districts—to whom citizens bring their problems and those of their communities.8 This is hardly surprising. In agrarian societ-ies, in which most residents are peasant farmers who move about on foot, by bicycle, or by public transportation, the most effective representa-tion is based on face-to-face contact in the countryside or small towns.

It is also not surprising that this difference in quality of representa-tion between PR and district-based systems is becoming a political issue in Namibia and within SWAPO. Owing to the failure of PR to facilitate a continuous dialogue between MPs and the rural population, regional councillors want an end to PR for elections to the National Assembly, as well as a devolution of power to the regional councils. Some also be-lieve that a change to an SMD system would enable them to challenge incumbent MPs successfully at the next parliamentary election. For the same reasons, incumbent MPs and the SWAPO leadership oppose a change of electoral system. The disagreement over electoral approach also reflects a “generation gap”: councillors tend to be five to ten years younger than MPs. The divisions are likely to have deepened by the time of the next elections in 1999.

These shortcomings of PR resulted in similar calls for district-based voting in South Africa less than one year after that country’s first all-race elections, which were conducted under a “two-tier” version of PR. Under the South African system, 200 seats in the National Assembly are allocated on the basis of the party vote nationwide, and 200 are allo-cated on the basis of the party vote in each of nine regions.9 This version of PR has resulted in each party’s gaining a proportion of seats that is virtually identical to its percentage of the total vote, with the largest parties guaranteed participation in the government proportionate to their vote.10 Yet the absence of a representational link between indi-vidual MPs and specific geographic constituencies has been apparent from the start. Recognizing the disadvantages of this situation, the ANC announced in October 1994 that it was establishing an informal system

140 Comment: Elections in Agrarian Societies

of constituency-based representation by assigning MPs and cabinet ministers to designated areas and providing them with the financial support needed to maintain a presence there.11 Historically, the ANC has favored an SMD system, but it modified its position to support the present system when it calculated that it would do no worse under “two-tier” PR. The NP has also announced that it wants a partial return to constitu-ency-based elections in 1999.12

Leaders of South Africa’s two largest parties feel comfortable in call-ing for an end to or a modification of PR largely because they have concluded that a constituency-based formula would provide substan-tially similar electoral results, without PR’s disadvantages. Small par-ties, such as the Pan-Africanist Congress (PAC), the Democratic Party (DP), and the Freedom Front (FF), may be wiped out under an SMD system, while the fortunes of the regionally based Inkatha Freedom Party (IFP) are unlikely to change. It would also appear that the NP, which favors a continuation of power-sharing after 1999, realizes that such arrangements do not require PR. Indeed, the current power-sharing ar-rangement in South Africa is based not on PR per se, but on the number of seats each party obtains in the National Assembly. If an SMD formula yields roughly the same distribution of seats as PR, consociational de-mocracy can be maintained.

The Same Distribution of Seats

Ironically, some of the best evidence for the similarity of the distribu-tion of legislative seats under SMD plurality and PR has been assembled by Andrew Reynolds, a staunch proponent of PR. In a valuable essay in which he “re-runs” the recent elections in South Africa and Malawi, Reynolds shows that the choice of electoral formula is not nearly as significant in agrarian societies as it is in advanced industrial societies.13 Reynolds’s data, along with my own figures for the most recent parlia-mentary elections in Kenya and Namibia, are presented in the table.14

The table presents the results of four elections in four African coun-tries: South Africa (April 1994), Namibia (December 1994), Malawi (May 1994), and Kenya (December 1992). South Africa’s and Namibia’s elec-tions were run under PR, and Malawi’s and Kenya’s were run under SMD plurality. The table allows for comparison of the percentage of seats that each party actually obtained under the electoral formula used for the election in question with its percentage of the vote, as well as with the percentage of seats that the party would have obtained under alternative electoral formulas. Thus the distributions of seats obtained via PR in South Africa and Namibia can be compared with the distributions that would have resulted under SMD plurality, while the distributions of seats obtained via SMD plurality in Malawi and Kenya can be compared with the vote shares obtained by the parties—percentages that would

Joel D. Barkan						141
TABLE—DISTRIBUTIONS OF VOTES & SEATS UNDER PR & SMD FORMULAS
ELECTION	PARTY	1	PERCENTAGE	PERCENTAGE OF SEATS
			OF VOTE	NATIONAL-	EXISTING	EQUAL
				LIST PR	SMDS	SMDS2
South Africa (April 1994)	ANC		62.7	63.0	NA	70.8
	NP		20.4	20.5	NA	17.0
	IFP		10.5	10.8	NA	12.3
	FF		2.2	2.2	NA	–
	DP		1.7	1.8	NA	–
	PAC		1.3	1.3	NA	–
	ID		–	0.3	NA	6.7
Namibia (December 1994)	SWAPO		73.9	73.6	67.4	73.9
	DTA		20.8	20.8	29.5	23.9
	UDF		2.7	4.2	3.2	2.2
	Other		3.1	1.4	–	–
	ID		–	1.1	7.7	2.2
Malawi (May 1994)	UDF		46.4	NA	48.0	NA
	MCP		33.7	NA	31.6	NA
	AFORD		18.9	NA	20.3	NA
	ID		–	NA	2.1	NA
Kenya (December 1992)	KANU		29.7	NA	53.2	44.6
	FORD-A		24.2	NA	16.5	22.3
	FORD-K		20.8	NA	16.5	17.0
	DP		22.2	NA	12.2	15.4
	Other		3.1	NA	1.6	0.5
	ID		–	NA	19.1	12.0
IDs (average) for selected	Canada					11.3
established democracies	N.Zealand					10.7
with SMD plurality systems
	UK					10.5
	USA					5.4

1 South Africa: ANC, African National Congress; DP, Democratic Party; FF, Freedom Front; IFP, Inkatha Freedom Party; NP, National Party; PAC, Pan-Africanist Congress. Namibia: DTA, Demo-cratic Turnhalle Alliance; SWAPO, South West African People’s Organisation; UDF, United Demo-cratic Front. Malawi: AFORD, Alliance for Democracy; MCP, Malawi Congress Party; UDF, United Democratic Front. Kenya: DP, Democratic Party; FORD-A, Forum for Restoration of Democracy– Asili; FORD-K, Forum for Restoration of Democracy–Kenya; KANU, Kenya African National Union. 2 Data for South Africa based on election results as reported in 341 counting districts as presented by Andrew Reynolds in “Re-running the 1994 South African and Malawian Parliamentary Elections Under Alternative Electoral System Formulae,” in Timothy Sisk and Andrew Reynolds, eds., Elec-tions and Conflict Resolution in Africa (Washington, D.C.: U.S. Institute of Peace Press, forthcoming). Data for Namibia based on election results as reported in 95 councilmanic districts adjusted for population variations among districts. Data for Kenya based on election results in 188 parliamen-tary constituencies adjusted for population variations among districts. Data for the established democracies from Arend Lijphart, Electoral Systems and Party Systems (New York: Oxford Univer-sity Press, 1994), 160–62.

Note: Percentages do not always total 100 due to rounding.

have been approximated by PR. Data for the SMD system are presented in one or both of two versions—actual results (as they occurred in Malawi and Kenya) and estimated results (for South Africa, Namibia, and Kenya)

142 Comment: Elections in Agrarian Societies

had all districts contained the same number of voters. Finally, the Table presents an Index of Disproportionality (ID) for each formula for which a distribution of seats was calculated, as well as for recent elections in four advanced democracies that employ SMD plurality. The closer the ID is to zero, the closer the electoral formula comes to translating each party’s percentage of the vote into exactly the same percentage of seats. Thus the higher the ID, the greater the disproportionality of the out-come. The ID typically ranges from near zero to 20.15

The Table underscores three basic points. First, except in the case of Kenya, PR and SMD plurality yield substantially the same distribution of seats. Both formulas do a good job of translating percentage of the vote into percentage of seats. Second, PR is the superior method if achiev-ing a proportional outcome is the main purpose of an election. As argued above, however, the mere mathematical translation of votes into seats does not guarantee the establishment of a continuous process of repre-sentation after the election. When it comes to the quality of representa-tion, SMD plurality is the superior formula. The question, then, is whether the superiority of representation offered by SMD plurality is worth the cost of a modest distortion in the translation of votes into seats.

Third, the deficiencies of SMD plurality are mitigated when all dis-tricts contain the same number of voters. In Namibia, the average number of voters per district is 5,236, but the number of voters in the existing councilmanic districts varies from 1,051 to 13,592. Here, an SMD sys-tem based on districts of equal size would result in an ID nearly as low as the current PR system. Such a system would also reduce the level of disproportionality in Kenya, where the average number of registered voters per district is 25,544, but the actual number varies from 2,872 to 65,887. While the resulting ID is not as low as that for Namibia, it is near that of several advanced democracies. Most important, the creation of equal-sized districts in Kenya would reverse the most serious shortcom-ing of the present SMD system (but one that also “protects” the minority interests of Kenya’s smallest and least developed ethnic groups): a leg-islative majority for a party that obtained only 30 percent of the vote. Creating more and thus smaller and more compact districts would fur-ther reduce the level of disproportionality associated with SMD plurality.

Explaining Disproportionality

The level of disproportionality depends not so much on the choice of electoral formula as on three other variables: 1) the extent to which voters of different parties are geographically concentrated or dispersed; 2) the extent of variation in the number of voters per district; and 3) the number and compactness of electoral districts. This is why, in the con-text of agrarian societies, SMD systems yield (or can be designed to yield) distributions of seats that are close to those obtained via PR.

Joel D. Barkan

143

As explained above, voters in agrarian societies tend to vote in geo-graphic blocs that are highly homogeneous. As a result, it is not unusual for parties to obtain more than 90 percent of the vote in some areas and less than 10 or 15 percent in others. What appears from a distance to be a multiparty system is in actuality a collection of regional one-party systems. Malawi and Namibia are near-perfect examples of this ten-dency. By contrast, multiparty systems in advanced industrial societies are characterized by a geographic dispersion of the vote. Although dif-ferent regions are viewed as bastions of different parties, the degree of hegemony is much less than in developing nations, because most areas harbor voters from across the political spectrum. Newly industrialized countries (NICs) fall somewhere in between these two extremes, mani-festing a pattern of bloc voting in many rural areas combined with more heterogeneous patterns elsewhere.

The more agrarian the society, then, the higher the geographic con-centration of the vote and the more closely the distribution of seats under an SMD system will mirror the distribution of the total vote, as well as the distribution of seats that would be obtained under PR. Con-versely, the more industrialized the society, the more geographically dispersed will be the vote, and the greater the discrepancy between the distribution of seats obtained under SMD plurality and that obtained via PR. In sum, SMD systems are likely to have a low ID in agrarian societies, but a moderate to high ID in industrialized societies.16 More-over, the extent of disproportionality associated with SMD systems can be reduced by creating more districts, thus making all districts smaller and more compact. This raises the level of social homogeneity within all districts, and thus increases the extent to which residents support one party or another.

Because advocates of PR and consociational democracy such as Arend Lijphart have tended to focus their examinations of electoral systems on advanced democracies and thus on industrial and postindustrial so-cieties, it is not surprising that they have failed to appreciate the extent to which SMD plurality and PR yield similar results in developing coun-tries. Nor, for the same reason, have they been sensitive to the extent to which SMD systems in agrarian societies can be refined to reduce the extent of disproportionality. Finally, advocates of PR have not fully considered the way in which rural voters approach the electoral process. Proponents of PR place a low emphasis on the need for geographically based representation, because in societies dominated by interest-group or corporate politics, such representation is no longer required to main-tain a democratic constitutional order.

Conceding the need for geographic representation, Andrew Reynolds has argued that a modified system of PR based on multimember districts would overcome PR’s principal defect in agrarian societies. If such a system is to facilitate face-to-face contact between citizens and repre-

144 Comment: Elections in Agrarian Societies

sentatives, however, it would require districts so small in size and popu-lation as to reduce the number of representatives per district to the point that allocating seats on the principle of proportionality would be either unnecessary or unfeasible. For example, if PR is modified to allocate from three to six seats in each of a series of small multimember districts, why bother? The allocation of such a small number of seats may yield results that are less proportional than those of SMD plurality. They would certainly be less proportional than those of ordinary PR.17

Advocates of PR and defenders of SMD plurality run the risk of talk-ing past each other. In fact, each formula is particularly appropriate for one subset of political systems, but not necessarily for all systems. For agrarian societies—especially those of Africa, the world’s least devel-oped continent, where the return to democracy is at a critical stage—SMD systems may yet provide an essential ingredient of democracy, without which its consolidation is unlikely to occur: effective representation. In this context, any call for the replacement of constituency-based vot-ing with PR is likely to be premature.

NOTES

1. See Andrew Reynolds, “Constitutional Engineering in Southern Africa,” Journal of Democracy 6 (April 1995): 86–99; and Arend Lijphart, Power-Sharing in South Africa (Berkeley, Calif.: Institute of International Studies, 1985) and Electoral Systems and Party Systems (New York: Oxford University Press, 1994).

2. See Joel D. Barkan, “Kenya: Lessons from a Flawed Election,” Journal of Democracy 4 (July 1993): 85–99; and Andrew Reynolds, “Re-running the 1994 South African and Malawian Parliamentary Elections Under Alternative Electoral System Formulae,” in Timothy Sisk and Andrew Reynolds, eds., Elections and Conflict Resolution in Africa (Washington, D.C.: U.S. Institute of Peace Press, forthcoming). While ethnicity is undoubtedly an important factor in the high geographic concentration of the vote in many African countries, there is substan-tial evidence—from Kenya and elsewhere—that it is not the only one.

3. Joel D. Barkan, “Legislators, Elections and Political Linkage,” in Barkan, ed., Politics and Public Policy in Kenya and Tanzania (New York: Praeger, 1984), 71–101.

4. Goran Hyden, No Shortcuts to Progress: African Development Management in Perspective (Berkeley: University of California Press, 1981), 7.

5. As Samuel P. Huntington wrote nearly three decades ago, the institutionaliza-tion of any political practice—democratic or otherwise—requires that each element of that practice (e.g., the regular holding of elections, the legislative or judicial process, and so on) be valued by those who engage in that particular practice as well as those outside of it. See Samuel P. Huntington, Political Order in Changing Societies (New Haven: Yale University Press, 1968).

6. International donors and investors have made clear to the Namibian govern-ment that any reversal of constitutional guarantees of property rights or of multipartism would drive away both investment and aid.

7. It should be remembered that these 95 councilmanic constituencies play no

Joel D. Barkan

145

role in the present application of PR, other than to serve as counting areas for the tabulation of the national vote. Thus the number of seats in the National Assembly is not 95 but 72. To facilitate comparison of the outcomes of PR and SMD plural-ity, all actual or hypothetical distributions of seats have been expressed as percentages.

8. Joel D. Barkan, Gretchen Bauer, and Carol Lynn Martin, The Consolidation of Democracy in Namibia: Assessment and Recommendations (report prepared for the United States Agency for International Development by Associates in Rural Development, Inc., Burlington, Vermont, July 1994).

9. Seats within the nine regional legislatures are also allocated on the basis of the party vote within the regions.

10. One cabinet position is allocated for every 20 seats held by a political party.

11. Sunday Times (Johannesburg), 9 October 1994, as cited in South Africa News Update, 11–17 October 1994.

12. Citizen (Johannesburg), 20 January 1995, as cited in South Africa News Update, 17–23 January 1995.

13. Andrew Reynolds, “Re-running the 1994 South African and Malawian Parliamentary Elections.”

14. I wish to thank Andrew Reynolds for sharing his insights and data with me. I am also grateful to Judith Geist for sharing her data on the 1992 Kenya election.

15. Several methods exist for measuring the extent to which an electoral for-mula yields a distribution of seats that diverges from the distribution of the vote. Following both Andrew Reynolds and Arend Lijphart, I use the least squares index developed by Michael Gallagher. For a full discussion of the various methods of calculating disproportionality, including Gallagher’s, see Arend Lijphart, Elec-toral Systems and Party Systems, ch. 3.

16. The only exceptions to this rule are those cases where support for contesting parties is dispersed as a result of some factor, such as religion or language, that divides residents of the same rural communities. The impact of caste in India or of religion in Uganda would be examples of this phenomenon. Conversely, the high geographic concentration of the vote in South Africa, the most industrialized country on the continent, is in large part a legacy of apartheid laws (e.g., the Group Areas Act), which dictated where members of different races and ethnic groups could live.

17. See, for example, Andrew Reynolds’s own calculations for the application of this option to Malawi in “Re-running the 1994 South African and Malawian Parliamentary Elections.”