Movement towards a two-party system

As long as Botswana was a largely agrarian, traditional society, the BDP dominated elections. Support for the opposition was largely limited to urban centres, areas where traditional leaders supported the opposition (e.g. Kgatleng, Southern District), and regional concentrations of politically mobilised non-Tswana groups (North East and North West). By the 1980s, however, structural changes were eroding the BDP's electoral base (Poteete Reference Poteete2009). The percentage of the population in rural areas declined from 82·3% in 1981 to 54·3% in 1991 and 45·8% in 2001 (ROB 2003: 18). Equally dramatic changes in economic activities also occurred. Most obviously, with the development of diamond mines in the 1970s, mining soon overshadowed all other sectors. The contribution of agriculture to GDP plummeted from 42·7% in 1966 to 5·6% in 1985/86 and 1·8% in 2005 (Bank of Botswana 2010; ROB 2003: 29). The late 1980s and early 1990s saw the emergence of construction (Poteete Reference Poteete2009), tourism and financial services (Bank of Botswana 2010; ROB 2003: 29) as important economic sectors.

Whatever their limitations in terms of economic development (Auty Reference Auty2001; Hillbom Reference Hillbom2008), these transformations have significant social and political implications. Education, urbanisation and a variety of legal changes decreased the influence of traditional leaders over vote choices. Demographic and socio-economic changes are gradually reducing the electoral weight of rural areas.Footnote ⁴ The influence of urbanisation extends well beyond the city limits as people move between rural and urban settings. These social and economic changes have whittled away at the BDP's electoral base, creating and expanding openings for the opposition (Poteete Reference Poteete2009; Wiseman Reference Wiseman1998).

The opposition vote consolidated behind the BNF from the 1980s into the 1990s. The BNF's share of the opposition vote climbed from 52% in 1979 to 76·5% in 1989 and 81·6% in 1994 (Wiseman Reference Wiseman1998: 247). Its overall vote share nearly doubled from 20·4% in 1984 to 37·1% in 1994, and its parliamentary representation shot up from five of thirty-four elected seats (14·7%) in 1984 to thirteen of forty elected seats (32·5%) in 1994 (see Figures 2 and 3). The trend over four electoral cycles (1979 to 1994) was consistent with the development of a two-party system at the national level.

A party system in flux

In the mid 1990s, the BNF seemed poised to gain additional seats in 1999 and unseat the BDP in the foreseeable future. Instead, in 1998, the party imploded; eleven of the BNF's thirteen MPs and sixty-eight of its councillors resigned and formed the Botswana Congress Party (BCP). The weak and divided opposition contributed to the BDP's rebound in 1999.

The party system has been in flux ever since. Opposition politicians are well aware of the costs of vote-splitting and there have been several attempts to reverse the fragmentation of the opposition (Mokopakgosi & Molomo Reference Mokopakgosi and Molomo2000). The BNF, the BPP and the Botswana Alliance Movement (BAM) formed an electoral pact and did not run candidates against each other in 2004. The BCP, however, did not participate in the electoral pact on the grounds that, as a new party, it needed to consolidate itself first. Consequently, vote-splitting increased and, although electoral support for the combined opposition climbed to 48·2% in 2004, exceeding the previous peak of 45·7% in 1994, opposition representation in the legislature recovered only partly, to 22·8% of the elected seats as compared with 32·5% in 1994. The number of constituencies won with a plurality in a race with three or more candidates rose from ten (25%) in 1999 to nineteen (33%) in 2004. The BDP won twelve of these constituencies, including six of the nine plurality contests where the margin of victory was less than 5%. The steep costs of vote-splitting prompted renewed efforts to unite the opposition within a year. The BAM and the BCP formed an electoral pact for the 2009 elections but negotiations with the other opposition parties broke down. The number of constituencies won by a plurality declined only slightly in 2009, to fifteen (26%).

Changes in the political players and in their relation to each other from one election to the next make it difficult for voters to evaluate the relative standing of the various parties. Factionalism within the BNF and the BDP exacerbates these problems. Both parties fought dramatic factional battles in the months leading up to the 2009 elections. In both cases, after rivals performed well in internal elections, the faction associated with the president of each party attempted to circumvent or reverse internal election results. Factional leaders in both parties demonstrated a willingness to hurt the electoral prospects of their own party in order to punish rivals. Unable to settle factional conflicts internally, both parties found themselves before the courts. Court rulings did little to resolve these divides. The severity and public nature of these factional wars increased the sense of unpredictability as the elections approached.

Factionalism in the BNF

Even before its implosion in 1998, divisions within the BNF had prompted defections and produced splinter parties in 1969–70, the mid 1970s, 1989, 1993 and 1994 (Makgala Reference Makgala2003, Reference Makgala2006; Mokopakgosi & Molomo Reference Mokopakgosi and Molomo2000). Ironically, many of these splits, especially in the 1980s and 1990s, have been attributed to the BNF's growth, as the rapid rise of new recruits to leadership positions generated jealousies and charges of opportunism. The 1998 split halted the BNF's growth but did not bring peace. Each subsequent leadership contest has been fought – bitterly and openly – along factional lines, and each internal election has been followed by a new round of disciplinary hearings, suspensions and expulsions, and resignations. Recurrent internal struggles repelled potential allies and thwarted the BNF's hopes of re-emerging as the main challenger to the BDP. Its share of the opposition vote has fallen steadily, from a peak of 81·6% in 1994 to 60·6% in 1999, 54% in 2004 and 46·9% in 2009.

Factionalism resurfaced in the run-up to the 2009 elections. In response to increasingly vocal internal opposition in 2006, the BNF president Otsweletse Moupo sought and received a new mandate at a special party congress in 2007. The leadership contest was highly personalised, and Moupo's challengers raised procedural questions that called the legitimacy of the elections into doubt (Modise Reference Modise2007). Rather than restoring calm, the 2007 party congress hardened factional differences. Primary elections in February 2008 presented a new challenge as Moupo was defeated and his rivals won several constituencies. Moupo's opponents called for a regular party congress, and thus another leadership contest, in 2009. The BNF central committee refused, arguing that the party constitution did not require a congress, and that a leadership contest would distract from the general election campaign. Opponents of the central committee made their grievances public, and the leadership responded with disciplinary action. Suspensions and expulsions began in October 2007 and accelerated over the next year and a half. Those affected included three incumbent MPs and three other primary election winners. In each of these constituencies, the party leadership sought to replace factional rivals with allies. Several of these disputes ended up in the courts, with some cases dragging into September 2009.

The damage from these events and their timing was multidimensional. First, the inability of the BNF to put its own house in order called into question its ability to govern. The contrast with the BAM/BCP alliance in 2009 only underlined the BNF's problems. Second, the legal disputes took time, energy and money away from the election campaign. Fundraising – a perennial problem for the opposition – was delayed and became even more difficult because of the disputes (Bashi Sengwaketse 2009 int.). Third, the court cases and suspensions delayed campaigning in some constituencies. Several parliamentary candidates were launched only two to three months before the elections. Fourth, some expelled members, including two incumbent MPs, contested the elections as independent candidates.

Factionalism in the BDP

Factional divides existed within the BDP by the 1970s (see Magang Reference Magang2008: 383 ff., 468), but were fairly fluid until the 1990s. To revitalise the party and manage internal divisions, the BDP modified its system of primary elections and altered constitutional provisions related to presidential succession. Ironically, these reforms have fuelled factional conflicts and, in the case of the constitutional amendments, galvanised the opposition.

After the 1994 elections, the BDP hired a consultant who recommended a generational change to revitalise the party and recruitment of a charismatic leader to reunite the factions (Molomo Reference Molomo2000a). Two constitutional amendments in 1998 sought to contain factional competition over the presidency and thus facilitate generational change. The first introduced a ten-year presidential term limit and was welcomed as a check on presidential authority. The second provided for the automatic succession of the vice-president to the presidency in the event that the office became vacant through death or resignation. Automatic succession was justified as a way of ensuring a smooth succession in such circumstances (Masire Reference Masire and Lewis2006).

In 1998, after these constitutional amendments had been adopted, a number of prominent BDP politicians retired, including President Masire. The provision for automatic succession was activated and Masire's vice-president, Festus Mogae, ascended to the presidency. Mogae recruited Ian Khama as his vice-president. Khama had not been actively involved in party politics but was well known and widely respected. He is the son of the first president, kgosi of the Tswana morafe with traditional authority over what is now Botswana's most populous district (the Bangwato in Central District),Footnote ⁵ and was commander of the Botswana Defence Force. Mogae and others hoped that he might reunite the BDP.

Ian Khama became president on 1 April 2008, on the expiration of Mogae's term. He declared that his presidency would be guided by four principles, referred to as the ‘four D's’: democracy, development, dignity and discipline. He also pushed for more efficient delivery of public services, and immediately after the 2009 elections designated ‘delivery’ as a fifth ‘D’. He has a reputation for instructing officials to attend immediately to issues raised during public meetings (e.g. Dube-Kelepeng Reference Dube-Kelepeng2009; Makgala & Mmekwa Reference Makgala and Mmekwa2010). He is also known for his use of directives, which he sees as enhancing efficiency: ‘if I can bring anything to this office it is the meticulous way the army runs the institution’ (Dube-Kelepeng Reference Dube-Kelepeng2009).

While Khama's efforts to improve the efficiency of the public sector are widely appreciated (Makgala & Mmekwa Reference Makgala and Mmekwa2010), his approach antagonises many people, including civil servants and fellow politicians who resent his reliance on directives and surprise announcements. Some directives – such as the introduction of a 30% tax on alcohol in 2008 and of a curfew – generated strong negative public responses. Whatever their appeal to social conservatives, these measures also harm local breweries, numerous small-scale bars and restaurants, and the increasingly important tourism industry. His fiercest critics see a pattern of behaviour that raises questions about his commitment to established legal procedures, the value of consultation, and even democracy and the rule of law (Good & Taylor Reference Good and Taylor2008; Lotshwao Reference Lotshwao2011; Molomo Reference Molomo2000b).Footnote ⁶ In the 2009 election campaign, the BCP/BAM depicted Botswana as facing a choice between ‘democracy and prosperity or dictatorship and collapse’ (BAM & BCP 2009), while the Manual Worker's Union campaigned against prominent members of the BDP, calling them ‘enemies of democracy’.Footnote ⁷

Although there is no denying the polarised responses to Khama's priorities and leadership style, the intensification of factionalism within the BDP since 1998, and the party's split in 2010, are not simply reactions to an individual leader. The growing realisation of how far the balance of power within the BDP has shifted in favour of the president represents a more fundamental source of factional conflict. In the past, BDP members had reason to believe that they could influence the selection of future presidents and vice-presidents. When the constitutional reforms of the late 1990s were debated, most BDP members failed to recognise how the new arrangement would empower the president relative to the party (Botsalo Ntuane 2010 int.). The presidential transfers in 1998 and 2008, however, showed that the interaction of the two-term limit with the provision for automatic succession enables outgoing presidents to choose the next president single-handedly.

Primary elections also contribute to factionalism. In 1984, the BDP introduced indirect primary elections in which an electoral college vetted the winners. The reversal of primary election results generated controversy and prompted some activists to leave the party. In 2002, the BDP adopted a new system of primary elections known as Bulela Ditswe (‘open the gates’ or ‘free-for-all’). Under this arrangement, local party structures vet candidates. Higher level structures hear appeals based on procedural violations. Instead of eliminating disputes, the new system intensified internal competition, and in turn encouraged the vilification of factional rivals and accusations of vote-rigging (Makgala Reference Makgala2006). In effect, Bulela Ditswe prompted the decentralisation of factional mobilisation. If factions were initially formed and maintained primarily by those interested in leadership positions, heightened internal competition now provides an incentive for extending factional mobilisation to all party structures.

By 2009, factional divisions were institutionalised, with the two sides known as the A-Team and Barata-Phathi (‘those who love the party’). In 2008, factional allies mobilised to support each other in the primary elections.Footnote ⁸ The dust from the primaries had barely settled when campaigning began for the party leadership contest in July 2009. Barata-Phathi focused their campaign on Khama, whom they characterised as repressing dissent, ignoring established rules and procedures, and threatening democracy. Khama campaigned openly and forcefully for the A-Team.Footnote ⁹

Barata-Phathi won all ten elected positions at the party congress in mid July. A few days later, Khama used his power of appointment to fill the five remaining central committee posts and all subcommittees with individuals affiliated with the A-Team (Gabathuse Reference Gabathuse2009; Toka Reference Toka2009). Over the next couple of weeks, Khama took several other decisions without consulting the new central committee. Then, on 13 August 2009, an unsigned media statement asserted that Khama had not exceeded his constitutional authority in any of his actions since the party congress (Anon 2009). The next day, the newly elected secretary-general, Gomolemo Motswaledi, explained that the central committee had requested three law firms to assess whether Khama's actions since the party congress violated the BDP constitution, but had not authorised the release of these opinions to the media. Khama, however, had authorised publication of the favourable legal opinion and, on 17 August 2009, suspended Motswaledi from the party for challenging his authority.

Motswaledi's suspension took the factional conflict to an entirely new level. The BDP halted its schedule of ‘star rallies’.Footnote ¹⁰ After several attempts to mediate the conflict failed, Motswaledi turned to the courts. He contended that his suspension was not a response to misconduct on his part but a bad-faith effort to reverse the outcome of the party's leadership contest in July (High Court 2009). Khama countered that the case should be thrown out because the Constitution of Botswana grants the president of the country immunity from prosecution for any action taken in his or her private capacity. The High Court agreed with the president and declined to evaluate the case on its merits (ibid.). Motswaledi appealed the decision on two grounds (Court of Appeal 2009). First, the provisions for presidential immunity are inconsistent with the otherwise democratic orientation of the constitution. Second, actions taken by the president of the country in his capacity as an office-holder in an organisation such as the BDP should be considered public actions. The Court of Appeal upheld the decision of the High Court. The elections took place two weeks later. The electorate had hardly any time to digest these events. As one candidate observed, ‘things are changing too fast for people. They can't keep up with the changes in the political parties, and the changes within the political parties, in the ruling party and in the opposition’ (Tawana Moremi 2009 int.).Footnote ¹¹ It is no wonder voters and political observers were ‘on tenterhooks’ on election day (Molefhe Reference Molefhe2009).

The electoral system, the party system, and strategic voting

The electoral system interacts with the party system to magnify the BDP's electoral majority in several ways. Most analyses of FPTP electoral systems focus on the effects of vote-splitting and problems of coordination in the electorate (see Chhibber & Kollman Reference Chhibber and Kollman1998; Cox Reference Cox1997). Vote-splitting definitely contributes to the legislative dominance of the BDP. Outcomes in FPTP electoral systems are also highly sensitive to the distribution of electoral support across constituencies. Opposition parties had some parliamentary representation in the 1960s and 1970s, despite relatively modest electoral support and the existence of three opposition parties, because each party had a regional stronghold. Ironically, the opposition's ability to convert votes into seats has decreased since 1998 because, as Figure 4 reveals, spatial variation in support for the BDP has declined. The BDP's aggregate vote share has fallen, but the party has actually gained support in areas where it had been weak historically. These trends have implications for vote-splitting. If a party has the support of at least 40% of the electorate, as is the case with the BDP in most parts of the country, it will generally be the largest party in a three-way match. Indeed, in 2009, the BDP won eight of the fifteen constituencies that were decided by a plurality vote.

Figure 4

BDP vote shares by district, 1969–2009

Sources: 1969–94: Republic of Botswana official election reports; 1999–2009: IEC 1999, 2004, 2009a.

In FPTP electoral systems, two main parties or candidates are more likely to emerge at the constituency level if (1) voting is strategic, (2) there is no obvious winner, (3) the relative standings of the parties are well known, and (4) few voters feel nearly indifferent about the front-runner but very strongly about the third ranked candidate or party (Cox Reference Cox1997). In Botswana, expressive voting and poor information about the relative standing of the candidates and parties hinder strategic voting and thus contribute to vote-splitting.

There are three forms of expressive voting in Botswana: expressions of loyalty to one's kgosi, party loyalty, and animosity towards specific parties. While all three favour the BDP, animosity towards specific parties has been a particularly important obstacle to reconsolidation of the opposition since 1998. There is considerable bad blood between the BCP and the BNF. Parties involved in various failed electoral alliances also point fingers at each other. Some opposition voters would prefer to let the BDP win rather than vote for another opposition party. Expressive voting benefits the BDP more than the opposition parties, but cannot account for the slight increase in electoral support for the BDP or its legislative dominance.

The BDP's resilience in 2009 is at least partly a result of information problems that interfered with the electorate's ability to track changes in a fluid party system. Opposition voters can only rally behind one party if they can assess which opposition party has the best chance of beating the BDP. Poor information about the relative standings of the various candidates and parties is a general and quite significant problem. Because there are no country-wide opinion polls, assessment of the relative standing of parties or individual candidates relies on direct exposure to campaign activities (posters, rallies, house-to-house canvassing) and media reports. Limited resources put the opposition parties at a disadvantage in terms of getting their message out. There is no public campaign financing, and many private individuals and businesses worry that donating to the opposition may threaten their access to government tenders and contracts.Footnote ¹³ The opposition parties have always had less access to financial resources than the BDP.Footnote ¹⁴ In 2009, for example, every BDP parliamentary candidate received a donated vehicle while opposition candidates provided their own transport. This situation makes media coverage all the more important.

The private media did report on a variety of constituencies across the country during the 2009 campaign, but private newspapers and radio stations are concentrated in Gaborone, their reporting on constituencies outside greater Gaborone is irregular, and newspaper circulation to rural areas is limited and delayed. Consequently, much of the population still relies upon the state-controlled media. The BDP gets more coverage in part because the state media focuses on government announcements and activities. There is also evidence of intentional bias. The BCP, for example, charged that Botswana Television (BTV) covered nearly all BDP launchings while broadcasting hardly any BCP launchings (Modise Reference Modise2009). They noted that reports on BDP activities were also longer and appeared more often during prime viewing hours. The BNF echoed these complaints. Although access to state media is a perennial problem for the opposition, some politicians felt that the degree of bias was especially severe in 2009.Footnote ¹⁵ Thus, problematic access to information contributed to the BDP's success both directly, through the positive bias in the state media, and indirectly, by making it more difficult for opposition voters to avoid vote-splitting.

Strategic responses to factionalism

In 2009, factional competition affected both partisan mobilisation and the options available for strategic voting. These dynamics had very different consequences for the BDP and the BNF. While factionalism hurt the BNF, it may have boosted the BDP vote in 2009. When the BDP secretary-general and president of the party challenged each other in the courts, a much diminished majority – or even an outright loss – seemed possible. The perceived threat of an opposition victory encouraged higher levels of voter turn-out among BDP loyalists. Since the BNF was certain to be in the opposition in any case, factionalism had more of a demobilising effect. Furthermore, in constituencies where the official BNF candidates confronted factional rivals in the form of independent candidates, disgruntled voters could express their discontent without abstaining or voting for a rival party.

Factionalism within the BDP also presented a new opportunity for strategic voting, at least in some constituencies. Faced with the choice of a Barata-Phathi candidate or an opposition candidate, some reform-minded voters probably opted for the Barata-Phathi candidate. After all, high-profile Barata-Phathi candidates had adopted many opposition positions and even echoed opposition slogans. Voters may have calculated that reform would be more effectively advanced from within the BDP (Serite Reference Serite2009). Although such strategic voting was probably most appealing to disgruntled BNF supporters in constituencies badly affected by factional conflict, BCP supporters in constituencies where their own party did not present a strong candidate may have followed a similar logic. Given the bad blood between the BCP and the BNF, supporters of these parties were more likely to vote for Barata-Phathi than another opposition party. Several politicians interviewed in 2010 argued that more unhappy BNF voters turned to the BDP than to the BCP.

The intensity of the factional conflict also raised the stakes of factional representation as opposed to party representation. Even if the court battle unsettled BDP supporters, no opposition party was strong enough to take power. On the other hand, the closely fought primaries and Barata-Phathi's sweep of the party leadership contest in July suggested that Ian Khama might not have an effective majority in the legislature. If Barata-Phathi candidates won enough seats, the faction could cooperate with the opposition to block the president's agenda, at least partially. Even in the absence of a legislative majority, the president has substantial powers of appointment and can issue directives. The cooperation of the legislature is required, however, for new legislation, approval of the budget, and the appointment of SEMPs. Likewise, Khama's options in filling cabinet posts are limited by factional representation in the parliament.

Because the BDP factions did not run rival candidates at the constituency level, the influence of faction-based voting might be expected to wash out in the aggregate if it were not for the possibility of strategic voter registration. In Botswana, voters enjoy an unusual degree of choice about where they can register. The constitution specifies that voters should register to vote in their ‘principal residence’. Because there is no legal definition of a ‘principal residence’, people may choose to register where they work, where they have agricultural lands or a cattle-post, or in their home village. The detection of fraudulent registration depends on inspections of the electoral roll by others registered to vote in the same constituency. Inspections have uncovered fraud by recognising the listing of non-residential plots as residences, or that unusually large numbers of individuals claimed the same residence (Baatweng Reference Baatweng2009). Nonetheless, the onerous nature of the inspection process and the difficulty of recognising fraud leave the system vulnerable to abuse.

This creates opportunities for strategic registration. In 2009, the Independent Electoral Commission (IEC) ran a public education campaign against what it calls ‘voter trafficking’: strategic registration solicited and organised by politicians or activists. An advertisement on BTV, for example, showed a politician convincing people at a cattle-post to register some distance away in exchange for money; these voters were disenfranchised when they showed up at a different – local – polling station on election day. The IEC declared that ‘voter trafficking had reached appalling levels’ in 2009 (Baatweng Reference Baatweng2009). Voter trafficking is illegal when it involves deception or the registration of people in locations where they have no residence. Politicians and activists are at least as likely to encourage loyal supporters to register in their constituency by taking advantage of their legal right to register either where they work, or have lands/cattle-posts, or in their home villages – and help them turn out to vote. Voters also engage in strategic registration on their own initiative.

Voters may engage in strategic registration for several reasons. First, competitiveness varies dramatically across constituencies. In 2009, the margin of victory ranged from 0·05% (Kgalagadi North) to 71·7% (Serowe North East). There is little motivation to register and vote in a non-competitive constituency, especially if one has the option of registering in a more competitive one. Second, factionalism may encourage strategic registration: as noted, factional activists prefer to register and vote in constituencies where the candidate is a factional ally. In addition, in constituencies where the incumbent is elderly or known to be in ill health (e.g. Tonota North), factional activists may register strategically in the hope of influencing an anticipated by-election. Third, voters who might otherwise abstain because they find none of the local candidates appealing could choose instead to register in another constituency with a stronger candidate.

Every politician and activist with whom I discussed strategic registration acknowledged its existence. Following a by-election in Tonota North in 2010, the winner admitted that his party had provided transport to voters residing some 500 km away (Kologwe Reference Kologwe2010). The newly elected MP asserted that these voters were legitimately registered in his constituency and that it was not the BDP's problem if the opposition lacked the resources to transport its non-resident voters to the constituency. In addition to such anecdotal evidence, a close examination of changes in the number of registered voters per constituency between 2004 and 2009, and of changes in the relative rankings of constituencies based on the 2001 population figures versus the number of registered voters, reveals discrepancies that are difficult to explain without reference to strategic registration.

Between 2004 and 2009, the number of registered voters in each constituency increased by 34% on average. That is somewhat higher than the 26·7% increase in the overall population between the 1991 and 2001 censuses, and even the 31·1% rate of urban population growth for the same period.Footnote ¹⁶ The rate of increase ranged from less than 20% in four constituencies (Selebi Phikwe West, Ghanzi North, Kgalagadi South and Ngwaketse West) to above 80% in two constituencies (Francistown South and Tonota North). In the ten constituencies with the highest rates of increase, the number of registered voters expanded much more rapidly than even the urban population.Footnote ¹⁷ There are plausible explanations for some of these changes. People moving from Orapa to Letlhakane during the closure of the Orapa mine, for example, might account for the jump in registered voters in Boteti South. And, despite sharp increases between 2004 and 2009, Gaborone West South and Mmadinare have significantly fewer registered voters than expected based on the population figures for 2001, as discussed below. There are no other contextual factors or obvious demographic changes that could account for the marked jumps in the numbers of registered voters elsewhere. At least some of these unusual increases in the number of registered voters probably reflect strategic registration.

To identify likely source constituencies, I ranked the constituencies from smallest to largest based on the population figures from the 2001 census as calculated for that year's delimitation exercise (ROB 2002). I then ranked the constituencies from smallest to largest based on the number of registered voters in 2009. A comparison of the two rankings reveals several constituencies with surprisingly low registration figures relative to population.Footnote ¹⁸ Seven of the constituencies with the most depressed figures are non-competitive BDP constituencies; the others were affected by factionalism. Non-competitive constituencies and those affected by factionalism present exactly the conditions that may encourage strategic registration.

This does not prove the existence of strategic registration or provide a precise gauge of its extent. Nonetheless, the opportunity exists, the incentives can be quite high, politicians acknowledge its prevalence, and some changes in voter registration are difficult to explain in other ways. When instigated by voters, strategic registration is uncoordinated and does not inherently favour a particular candidate or party. The fact that almost all of the likely source constituencies in 2009 were either BDP strongholds or affected by factional conflicts in the BDP suggests that the BDP benefited more than the other parties in this election. Strategic registration almost certainly contributes to the reduction in variation in the BDP's electoral support across constituencies. Although it is possible for opposition parties to benefit from strategic registration by bringing supporters together to win elections in select constituencies, this did not happen in any of the constituencies where the highest jumps in voter registration occurred in 2009.

Elections in Botswana have grown increasingly competitive in response to socio-economic changes and fragmentation of the party system. The 2009 elections were the most competitive to date. And yet, despite the erosion of its political base and increasingly competitive elections, the BDP has maintained a position of legislative dominance since the pre-independence elections in 1965. Because the BDP remains the most popular party in most of the country, it wins most constituencies with more than two serious contestants, even where it gains fewer votes than the combined opposition. These conditions have affected at least a quarter of the constituencies in every election since the BNF imploded in 1998. Consolidation of the opposition vote is hindered by party loyalties and animosities, as well as limited coverage of the opposition in the state-controlled media. In the 2009 elections, the BDP benefited from mobilisation related to factionalism, including strategic voting by reform-minded voters who might otherwise vote for the opposition or abstain, and strategic registration by activists hoping to boost their factional representation in parliament.

The story of the 2009 elections did not end with the vote on 16 October 2009 or the announcement of the results a few days later. The battle resumed when the new parliament convened. An effort by Barata-Phathi and opposition MPs to elect an alternative slate of SEMPs fell through at the last minute, but underlines both the competiveness of the elections and the severity of the divide within the BDP. Reorganisation of the party system also continues. In May 2010, the BAM and the BCP officially merged under the Botswana Congress Party label and several high-profile politicians formerly associated with Barata-Phathi launched a new party, the Botswana Movement for Democracy (BMD). By July 2010, eight MPs had joined the BMD, which displaced the BNF as the Official Opposition in parliament.

The BDP responded to the formation of the BMD with a series of ‘star rallies’ featuring President Khama and other prominent leaders, a reduction in the application of disciplinary measures against Barata-Phathi members, seminars and workshops for activists, and a revamping of its public relations strategy. In June 2010, Khama announced the suspension of internal elections for the foreseeable future and the use of compromise lists to fill positions in internal structures in the interim. By the end of July 2010, the BDP had halted its losses. A couple of MPs rejoined the BDP after a few weeks with the BMD and, shortly thereafter, the former BNF vice-president defected to the BDP. Another founding member of the BMD resigned in July 2011 and now sits as an independent MP.

Neither the split nor the BDP's responses to it, however, have eliminated factionalism within the BDP. At the time of writing, factional cracks were reappearing, exacerbated by mounting discontent with the suspension of internal elections and the government's management of a prolonged public sector strike.Footnote ¹⁹ In July 2011, Tawana Moremi resigned from the BDP at a public meeting related to the strike.Footnote ²⁰ The following month, the newly appointed BDP Secretary-General, Kentse Rammidi, resigned from the party after the government forced anti-labour legislation through parliament. By November 2011, they had joined the BMD and BNF respectively. Further changes in party affiliation cannot be ruled out.

Changes in the party systems have altered the composition of the National Assembly. At the time of writing, the BDP held thirty-nine of the fifty-seven elected seats (68·4%) and could count on the support of four specially elected MPs, the Speaker, and the president (see Figure 2). The BMD and BNF had six MPs each, the BCP had five MPs, and there was one independent. The BDP's ability to maintain a position of legislative domination depends on how much further its electoral support falls, and whether and to what extent the opposition reconsolidates. Nothing has happened to reverse the erosion of the BDP's political base. In fact, if factional competition boosted the BDP's electoral performance in 2009, the formation of the BMD may hasten the BDP's loss of electoral support.

On the other hand, the formation of the BMD further fragments the party system. The BMD has consistently voiced interest in cooperating with other opposition parties to avoid vote-splitting. At their party congresses in July 2010, the BCP and the BNF resolved to pursue cooperation with the other opposition parties. The main opposition parties rallied behind a single candidate for a parliamentary by-election in September 2010. By 2011, the opposition parties had staged several joint events, and discussions regarding cooperation for the 2014 parliamentary elections were underway. Lingering mistrust and uncertainty about the BMD's electoral support, however, represent serious obstacles to cooperation.

When the party system is fragmented and more than two viable candidates contest many constituencies, the FPTP system can generate legislative majorities – even overwhelming majorities – for parties that receive less than 50% of the vote. Such outcomes have occurred in Canada, Kenya, Lesotho, the UK, and elsewhere. But increased electoral competition can also result in highly volatile election outcomes. In 2009, more than 20% of the constituencies changed hands. If a comparable – or higher – number of constituencies changes hands in the future, sudden and dramatic shifts in representation could occur, as has happened in other countries with a FPTP electoral system. The fact that the BDP's electoral support is close to or below 40% in several constituencies where three and four way races are common increases the likelihood of sharp swings in electoral outcomes from relatively modest shifts in vote shares. Of course, reconsolidation of the opposition could result in an even more dramatic transformation of the legislature.

The analysis has important implications beyond the case of Botswana. Characterisation of a party as dominant suggests that the government faces little to no threat to its hold on power. When party dominance is defined based on legislative representation, such a conclusion can be misleading, since legislative dominance can persist despite competitive elections and serious factional conflicts. Such situations are not uncommon (Bogaards Reference Bogaards2008; Levitsky & Way Reference Levitsky and Way2002; Mozaffar & Scarritt Reference Mozaffar and Scarritt2005). While my analysis echoes many others in emphasising interactions between the electoral rules, the centralisation of state power, and the party system, Botswana's recent experience also highlights the importance of internal party dynamics, and particularly factional conflict. Indeed, factionalism features prominently in many other dominant party systems (Magaloni & Kricheli Reference Magaloni and Kricheli2010; Solinger Reference Solinger2001). Since a truly secure government is likely to behave in quite a different manner from a government with a more superficial and precarious position of dominance (see Poteete Reference Poteete2009), it is important to recognise and understand the dynamics within such systems. That means developing a better understanding of the sources and consequences of factionalism. Botswana's experience shows how political institutions can create incentives for factionalism, and influence the prominence of factional relative to partisan competition. While the dynamics of factional conflict may support persistent party dominance for some time, the analysis suggests that a tipping point may exist, beyond which the same factors – factionalism, party system fragmentation and a plurality electoral system – may make continued party dominance unsustainable.