Personal Power in Africa: Legislative Networks and Executive Appointments in Ghana, Togo and Gabon

Abstract

Personal relations and networks have long been argued to dominate African politics. Since personal power is difficult to measure, much of the literature has remained either anecdotal or has used ethnicity to approximate power distributions. This article is proposing a social network approach to the analysis of personal power in legislatures and cabinets in three cases: Ghana, Togo and Gabon. We combine survey data on parliamentary discussion networks with a new data set on cabinet appointments. We find that power accumulation in one institution correlates with power accumulation in the other in all three countries, irrespective of the level of democracy: individuals build up a unique power base to advance their careers. We also find differences between the modes of power accumulation and elite integration across our cases. Our findings could stimulate new debates on personal power, regime survival and elite reproduction across different regimes.

Leaders do not rule alone but rely on beneficial political alliances. This truism has been applied to a multitude of contexts and regions, from African autocracies and anocracies (Kroeger 2020) to renaissance rulers (Mesquita et al. 2005), and to Western parliamentary democracies (Indridason and Kam 2008). Effective coalition building enables leaders to insulate themselves from a variety of threats, including the political opposition, internal political rivals or social and economic shocks. Effective elite coalition building and networking have been used as explanations for rulers being able to hold onto power in spite of poor governance, economic stagnation or lack of civil liberties (Mesquita et al. 2005).

In the African context especially, the literature has placed emphasis on informal relations and personal networks as means of governance. Frequently these networks are examined in relation to the distribution of state resources and patronage, and are seen as an impediment to the formal rules of governance (Chabal and Daloz 1999; Jackson and Rosberg 1982). Newer studies present more nuanced arguments on the causes and effects of electoral clientelism in multiparty systems but still highlight the importance of personal relations (Kelsall and Booth 2010). Personal networks create webs of mutual obligation, and leaders elevate elites in return for their political support. These informal relations operate within the framework of the modern state and permeate formal institutions (Erdmann and Engel 2007). They thus become visible in formal positions of power such as cabinet appointments. We theorize that power ‘travels’ with individuals, and that power accumulation in one institution correlates with power accumulation in another. Looking at cabinets and legislatures, the article answers the following research question: how does personal power accumulation correlate across cabinets and legislatures in different African countries?

In spite of the supposed importance of personalism and clientelism in African politics, these relationships are rarely studied empirically. The main reason for this is the fact that data on personal relations are not easily available. Using Social Network Analysis (SNA), this study analyses personal power in parliaments and the executive in Ghana, Togo and Gabon.

Our contribution is threefold: (1) we propose a novel way to measure personal power across institutions; (2) we empirically demonstrate how personal networks cut across two different institutions, namely parliaments and governments; and (3) we provide comparative evidence on differences in personal power networks in three countries.

The article is structured as follows: we next provide an overview of the literature on personal relations and politics in Africa. The subsequent section discusses our theoretical framework and deduces hypotheses. We then introduce SNA as a tool to study power relations, and follow this with an explanation of our research design. The final two sections present our empirical analysis and discuss the findings.

Personal relations and politics in Africa

Our theoretical discussion draws on two overlapping but insufficiently linked research strands: the literature on authoritarian regime survival and the more regionalist debate on clientelist politics in Africa. Barbara Geddes (1999) sees personalism as a distinct type of authoritarian rule, while Michael Wahman et al. (2013) argue that it can be present in any given regime – albeit to varying degrees. The African studies literature has for a long time placed an emphasis on informal relations as the mainstay of African modes of governance, often leading to an overgeneralization of all regimes as ‘neopatrimonial’ (Erdmann and Engel 2007; Kelsall and Booth 2010). Outside of Africa, the examination of the political importance of elite accommodation and informal relationships has been applied to more institutionalized countries such as Japan and the US (Helmke and Levitsky 2004) but is also discussed in the comparative literature on electoral authoritarianism and democratization. More recent work is calling for comparative research on the varieties of clientelism across and within regime types (Berenschot and Aspinall 2020).

We believe that progress in the study of informal personalist politics can only be made on the basis of more empirical insights. We demonstrate that personal relations permeate formal institutions in all three African countries under investigation. This is in line with the more recent interest in the interplay of institutions and personal politics. We begin with a short review of the literature on personal power, before we look more closely at the institutions of cabinets and parliaments.

Personal power can be studied from two complementary perspectives:

1. 1. From the perspective of the ruler who seeks to include selected powerful individuals into his ruling coalition to stabilize and strengthen his regime.

2. 2. From the perspective of the political entrepreneur who seeks to accumulate personal power in order to gain influence and individual benefits.¹

From the perspective of the ruler, intra-elite relationships are perceived as crucially important in the construction of stable political bargains. Incorporating potential rivals is key for maintaining political stability and control (Carboni and Raleigh 2021). Similarly, political phenomena such as protests, political party formation and intra-party defections are portrayed as part of a negotiation process in which political elites demonstrate to the leader the necessity of their inclusion and try to elevate their position within the political hierarchy (Andrews and Honig 2019). In this way, opposition leaders also seek their incorporation into government, often leading to the further fragmentation of the opposition (Arriola et al. 2021) and to weakly institutionalized party systems. Alternatively, highly institutionalized political parties can themselves be patronage networks that negotiate the integration of the individual into the top hierarchy (Driscoll 2020 on Ghana).

From the perspective of the individual, a large personal network grants bargaining power. Political entrepreneurs make use of economic and symbolic resources to bind their followers (Compagnon 2012; Nugent 2007). Having a dominant position within the patchwork of elite relationships makes one a valuable asset to a leader. Therefore, personal connections also have an important impact on individual political fortunes and influence.

Thus, the relationship goes both ways: leaders are interested in drawing powerful people close to the regime, either simply to broaden the regime's support base or to neutralize potential competitors by co-opting them into government. At the same time, individuals seek to build personal power by nurturing close relations with other people of power. These political entrepreneurs often use this leverage for their own career advancement. In this way, personal power begets political power.

Institutions and power

The importance of ministerial positions

The literature on authoritarianism and regime survival highlights the role that institutions such as parliaments or executives play in the distribution of spoils and the management of elite coalitions (Gandhi and Lust-Okar 2009).

Both cabinets and parliaments are arenas of elite accommodation, but they differ in their level of importance for rulers and political entrepreneurs alike. A number of existing studies have used the composition of the highest level of executive government – the ministers – to approximate leaders’ strategies of sharing political power (Francois et al. 2015; Kroeger 2020; Lindemann 2011). Ministerial positions provide elites with high salaries, policy influence and the opportunity to steer the allocation of public resources (Kroeger 2020). This article refers to ‘ministerial positions’, the ‘executive’ and ‘cabinet’ interchangeably.

Ministers can also hold important influence over policies. Multiple studies have demonstrated that leaders, and senior-level ministers, engage in favouritism in the allocation of public resources for development to improve outcomes for their own constituencies (Franck and Rainer 2012; Kramon and Posner 2016).

Consequently, the executive is an important tool for building one's coalition of elite supporters. Though many democratic institutions were abolished or severely weakened in the post-independence era, the sharing of executive power remained an important tactic for integrating varying interests and creating an elite consensus in a heterogeneous environment (Arriola 2009; Carboni and Raleigh 2021). To political entrepreneurs, these positions are extremely attractive. Ministers can enrich themselves beyond their salaries by acting as gatekeepers to the reservoir of public resources, making them indispensable to business elites, multinational companies and domestic clients (Szeftel 2000).

The importance of parliaments

Parliaments in Africa are presumed to be comparatively weak, and in some cases irrelevant – ‘rubber stamps’ which exist to legitimize the decisions of the executive (Francois et al. 2015). On the other hand, they are a potential arena of elite dissent and can place limits on the power of the executive and leader (Barkan 2009; Collord 2021). For example, Bakili Muluzi of Malawi, Olusegun Obasanjo of Nigeria and Frederick Chiluba of Zambia all attempted to amend the constitution to extend their tenure but were blocked by parliament (Posner and Young 2007). Even Guinean autocrat Lasana Conte was forced to negotiate with protesting trade unions and end a repressive state of emergency in 2007 by the national assembly (McGovern 2007).

In contrast, control over the legislature grants the executive the power to set the legislative agenda and pass laws that enhance the regime's power. In Cameroon, President Biya engaged in a concerted programme of gerrymandering after he nearly lost the 1992 election (Opalo 2012). The ruling party's increased seat share and dominance over the legislature enabled Biya to scrap constitutional term limits in 2008 (Opalo 2012). Given the value of compliant parliaments to leaders, we would then expect leaders to select ministers who are influential within the legislature and who will enhance the executive's control over parliament.

Parliaments are also attractive for individual entrepreneurs. A seat in the national assembly comes with a number of personal gains: access to personal discretionary budgets, benefits like cars, drivers or immunity from prosecution (Lust-Okar 2009). Moreover, parliament is an arena where ‘big men’ meet to build and nurture useful connections. In countries where ministers are more or less regularly recruited among Members of Parliament (MPs), a legislative career is often a stepping stone for further career advancement, but also an opportunity to support private and business activities.

Consequently, we believe that the simultaneous interests of rulers and political entrepreneurs will result in two complementary patterns. Rulers who are interested in building effective ruling coalitions need to include powerful individuals in their ruling coalition and project power onto parliament. Therefore, they are more likely to appoint ministers who occupy an influential position within the parliamentary networks.²

By the same logic, former ministers who then become MPs should also be highly popular as patrons. Given that they have been relatively close to the centre of state power at least for a while, they are assumed to possess insider knowledge, authority and superior bargaining power. For ‘smaller’ or ‘rising’ political entrepreneurs in the parliament it will therefore be a rational strategy to seek connections and personal relations with them.

We can thus formulate our first two hypotheses:

Hypothesis 1: Former government ministers who then become MPs have higher levels of personal power in the parliament than average MPs.

Hypothesis 2: MPs with high levels of personal power are more likely to be appointed to cabinet than MPs with low levels of personal power.

Previous studies have shown that the configurations of elite networks are influenced by the culture and institutional set-up of the political landscape they inhabit (Kroeger 2020). Little is known, however, about the relative importance of personal relations in different regime types. Dominant autocratic regimes are often stereotyped as broad coalitions to bind elites in a heterogeneous political environment (Lindemann 2011; Roessler 2011). In more autocratic regimes, political relevance without significant allies within the ruling party or clique is impossible. Autocratic regimes also aim to dominate the legislature to stifle any opposition (Lust-Okar 2009). Thus, the personal power of elites in the legislative and executive branches of government is likely to be highly correlated.

While traditionally the advent of multiparty democracy was meant to reduce the clientelistic state of politics in Africa, research has shown that personalized politics persist (Berenschot and Aspinall 2020; Szeftel 2000). The key difference, however, is that elites in competitive multiparty democracies can maintain political or personal relevance outside of the regime as a member of an opposition party. We therefore argue that we should see differences in how power travels between the legislature and executive based on regime type:

Hypothesis 3: The correlation of personal power in parliament and presence in the cabinet is weaker in democracies than in authoritarian regimes.

The next section introduces SNA as a tool to study personal power.

Networks, influence and power

Power is an inherently relational concept. In recent times, relational analyses have become more prominent as a way to study politics (Marks and Stys 2019). While much of the literature has used the term ‘network’ metaphorically, few researchers have so far empirically studied patterns of elite interaction (Keller 2016; Osei 2018). Previous approaches have used the ethnic composition of the state to approximate personal networks of patronage (Francois et al. 2015; Posner 2004). Other measures of approximation have included the size and composition of senior government positions (Arriola 2009; Francois et al. 2015). These sometimes-crude approximations of personal networks were partly caused by limitations in data availability due to the opaque nature of many governments on the continent. Researchers have, however, found ways to collect data that lend themselves to relational analysis, be it from publicly available sources such as politicians' biographies (Keller 2016), data from cabinet reshuffles (Woldense 2018) or co-sponsoring of bills and co-voting (Desmarais et al. 2015). A few studies use surveys that record certain aspects of elite interrelations (Higley et al. 1991). Most of these approaches conceptualize power as network centrality or – differently said – the probability that a tie is formed in connection to a particular node. A high degree of connectedness within social networks has been found to be important in other, more recent, political contexts. Franziska Keller (2016) finds that the connectivity of Chinese Communist Party members in the Central Committee influences their chances of promotion to the most senior political body, the Politburo.

In this article, the approach combines survey data on the interaction structure of MPs in three countries with data on cabinet reshuffles. We want to see whether there is a correlation between being a minister (i.e. a powerful formal position in the government) and being a powerful MP in the parliament (i.e. a more informal conception of personal power). We look at the network positions of people who had been ministers in the past, who were ministers simultaneously to their mandate in the legislature, and those who had been appointed ministers after completing their mandate. The contribution is therefore threefold:

1. 1. We examine how power ‘travels’ with its holder from one institution to another.

2. 2. We show how formal and informal power are linked and how they can further individual careers.

3. 3. We present comparative insights into co-optation, elite integration and networks of power in three African countries.

There are two caveats, one concerning the timing of data collection, and the other the scope of the data that are available. First, the surveys in the parliaments are of relatively recent timing (2014–2019). Data on appointments that were made in the aftermath of the data collection are therefore only available for Ghana. Moreover, we have data only on ministers who had been MPs during the timing of the parliament surveys. Thus, we can say little about the general logic of appointments, but we can compare the power of MPs who did become ministers with those who did not.

Research design

Case selection

A purposeful theoretically guided selection of cases is usually the best strategy, but it is only possible under the assumption that the same quality of data is accessible for all countries. In this article, case selection is restricted by the availability of data, with the empirical resources coming from two developing data sources: the African Cabinet and Political Elite Data Project (ACPED)³ and the ongoing project Do Legislatures Enhance Democracy in Africa? (DLEDA).⁴

Intersecting these two data sets, we find three countries to analyse: Ghana, Togo and Gabon. These countries are of course not representative of the full universe of cases, but they constitute a convenience sample. The selected countries differ in terms of level of democracy, elite configurations and political histories.

The case studies: political history, elite networks and democracy

Ghana has a well-institutionalized two-party system in which the major parties, the New Patriotic Party (NPP) and National Democratic Congress (NDC), alternate in power (Gyimah-Boadi 2009). Both parties have ethno-regional strongholds; the Akan-populated regions for the NPP – most notably the Ashanti Region – and the Volta Region and the north of Ghana for the NDC. At the time of the survey used in this article (2013), the NDC had a majority in the national assembly, but it lost the 2016 elections to the NPP. In Ghana, it is possible to be an MP and minister at the same time. Article 78 of the constitution even requires that the president recruit at least half of the ministers from parliament.

Togo has been ruled by the Gnassingbe family since 1967. Political power has been strongly centralized and backed by the military. The ruling party, the Union pour la République (UNIR), has its regional stronghold in the north of the country, but most notably among the president's ethnic group, the Kabyé. Due to a combination of repression, patronage and various incumbency advantages, the UNIR has won all elections that followed the reintroduction of multiparty politics in the 1990s (Attisso 2012; Toulabor 1986). At the time of the survey, the UNIR had a majority of 62 of 91 MPs in the national assembly. In Togo, it is not allowed for someone to be a minister and an MP at the same time. The party system belongs to the dominant type with a strong ruling party surrounded by an instable opposition. Government–opposition relations were often hostile in the past (Osei 2018).

Gabon has been ruled by the Bongo family since 1965. Compared to Togo, elite politics are more accommodative and less repressive (Mouity 2011; Nzamba 2011). The Parti Démocratique Gabonais (PDG) regime invested the country's oil revenues into strategic patronage arrangements that incorporated crucial elites from all ethnic groups and regions. Just as in Togo, the opposition is splintered into various small parties. At the time of the survey (2019) the PDG held the majority of parliamentary seats. Of the other parties represented in the national assembly, most have joined the presidential majority (mouvance présidentielle), leaving only a few parties in the opposition camp. The party system is dominant, as in Togo, but Gabon has a long history of accommodation politics and consensus building between the ruling party and the opposition (Rossatanga-Rignault 2000). In Gabon, it is very common to appoint MPs to the executive but it is not legal to hold both functions at the same time. Each MP therefore has a seconder (suppléant) who will take over the mandate if the elected MP is appointed as minister. Gabon has a bicameral parliament, but this article deals only with the lower chamber, the Assemblée Nationale.

The countries differ in the level of democracy: V-Dem ranks Ghana in 2020 with 0.72, Togo with 0.38 and Gabon with 0.41 (Coppedge et al. 2020). The indicator for electoral democracy that is used here runs from 0 (least democratic) to 1 (most democratic). In addition, they also differ in their institutional set-up. Only in Ghana can a person be an MP and minister simultaneously. In Gabon and Togo, all ministers covered in the parliamentary census are former ministers and so these countries are used for testing H1. In Ghana, there was a change in ruling party and executive after the data on MPs was collected, allowing us to test H2, in addition to H1. A comparison between the three cases is used to assess the validity of H3, yet we acknowledge that issues such as institutional variation may limit the ability to attribute differences in power relations between parliament and executive purely to differences in democratization.

To summarize, our research design is theory-testing in respect of the relationship described in H1 to H3, but also hypothesis-generating in respect of our discussion of the findings and tentative conclusions about possible reasons for the observed cross-country variation. We are aware of the fact that using a convenience sample limits the possibility of generalization, but it is a useful starting point to test whether the hypothesized relationship exists at all across countries for which data are available. There are a number of other variables of interest – colonial background or ethnic group distribution – that are not systematically incorporated into the research design but are still used in the discussion of our findings.

Data collection

Data on parliamentary networks was collected between 2013 and 2019. All three case studies involve a full population survey in the parliament of the respective country. The surveys collect basic biographical data as well as relational data. The biographies contain, among other things, information about the career patterns of MPs. They were asked to name any important position that they held in the past – among these are positions as ministers, which we use in this article. Second, the surveys collected social network data. For this, a name-generator question was used: ‘Looking back over the past six months, who are the people in the parliament of [country] with whom you have discussed political issues? Please give me their names.’ The aim of the survey is to estimate which individuals are influential nodes of political decision-making and coalition formation, by identifying the individuals that are most sought out. A full network of parliamentary discussions was then constructed from the personal relations of each respondent. The response rates are at 92% for Ghana, 79% for Togo and 83% for Gabon.

Second, we use the ACPED data set of ministers and positions by month across African states. This includes all fully fledged (i.e. not deputy) ministers, regardless of whether they are in ‘the cabinet’. The unit of analysis is the minister by month. Using a mixture of locally based consultants and archival research, the data set provides the following information for each minister each month: gender, political affiliation, ethnicity and regional background.

Methodological approach

The Members of Parliament data set contains 273 observations for Ghana, 91 observations for Togo and 111 observations for Gabon. In addition to the relational and biographical information that we have from the survey, we construct the following variables from the ACPED data set:

• • Minister is a count variable of the ministerial positions that an MP held at any point in time.

• • Minister before MP is a count variable that contains only the ministerial positions that a person held before being elected to parliament.

• • Minister during MP exists only for Ghana; it is a binary variable measuring whether a person was a minister simultaneously to being an MP.

• • Minister after MP is a binary variable measuring whether a person was nominated minister after completing the term as MP; the variable exists for Ghana and Togo, but for Togo the number of observations is too small for meaningful analysis.

• • Inner circle is a binary variable that singles out the most important posts in the cabinet.⁵

The first part of the analysis will look at the correlation of the Minister variable with a number of network centrality measures (see Table 2). Since the Minister variable is skewed (the majority of parliamentarians in the sample have not been ministers at any point in their career) we use Spearman's rank correlation. As explained earlier in the article, centrality is often used to measure the power or prominence of nodes. There are numbers of different concepts, each with a different understanding of what a prominent position in a network signifies. We use the following concepts:

• • Degree centrality: The most basic concept is simply looking at the number of ties that a node has. Since our networks are directed, each node has incoming and outgoing ties that can provide important information on the popularity (prestige) and activity of a node. In SNA, the number of incoming ties is referred to as indegree, and the number of outgoing ties as outdegree. Nodes that receive a large number of ties can be seen as central in the network. Having a large number of in-ties mean a node is popular, as it is chosen by many other nodes as a point of contact. Outgoing ties rather describe the activity of that node itself – that is, the number of ties that the person is building up him/herself.

• • Closeness: Closeness centrality describes the nearness of a node to all other nodes. Nodes with high scores on closeness can easily reach everyone in the network and spread information quickly to many people.

• • Betweenness: Betweenness centrality denotes the number of times that a node lies on the shortest path between two other nodes. Nodes with high betweenness connect people who otherwise have no connection to each other or are very distant – betweenness can therefore describe a type of brokerage position.

• • Eigenvector: Eigenvector centrality is another measure of ‘power’ in a network. It takes on high values when a node is connected to other nodes that are well connected (Bonacich 1987). For our purpose, eigenvector is especially interesting because it allows inferences about the type of interrelations between people of power. As Phillip Bonacich argues, power is a positive function of the power that one has over others – if the ego's contacts are powerful, this makes the ego even more powerful. Yet, as he further argues, in bargaining situations ‘it is advantageous to be connected to those who have few options; power comes from being connected to those who are powerless’. Powerful people avoid other powerful people as allies ‘because each actor wants to be as powerful as possible’ (Bonacich 1987: 1171). In this sense, eigenvector as a measure is useful to map patron–client relations: a number of powerful patrons compete for (less powerful) clients but do not form alliances with each other. They draw support from vertical, not from horizontal relations. For rulers, this situation is nearly ideal because competitors will not form a united counter-elite and can be played out against each other.

A deeper understanding of tie formation in a network is possible with exponential random graph models (ERGMs). These models assume that relational ties are shaped by the presence or absence of other ties. By taking into account the complex dependencies within relational data, these models allow us to predict the joint probability that a set of edges exists on nodes in a network (Handcock et al. 2008). The central idea is that network ties depend on each other and are at the same time influenced by actor attributes and other exogenous factors (Lusher et al. 2012). ERGMs compare an observed network to a large number of random networks by modelling the effects of interest, and thus allow us to model both exogenous effects (covariates) and structural effects in the network simultaneously (Cranmer and Desmarais 2011). The models are fitted via Markov chain Monte Carlo maximum likelihood estimation. The dependent variable is the log odds of establishing a network tie. Coefficients are interpreted as log-odds ratios conditional on the rest of the network. For all ERGMs the statnet package for R is used (Handcock et al. 2008).

Empirical analysis and results

Descriptive statistics and country comparisons

Table 1 shows that the three countries differ in major ways. First, Ghana has a much larger parliament but also over 20% of parliamentarians are/have been/will be in the cabinet. In contrast, in Gabon and Togo the figures are 14% and 12%, respectively. In Gabon, the most experienced minister held nine positions in the government, compared to six in Ghana and only two in Togo.

Table 1. Descriptive Statistics of Parliament Size and Number of Ministers

Table 1 shows that Ghana, as a consolidated two-party democracy, has much more exchange between the legislature and executive. It must be borne in mind that the majority of ministers must be chosen from the parliament. Both Gabon and Togo have less movement between the two branches, perhaps reflecting a less open political environment, but elites have different trajectories within these systems. In Gabon there is a lot of volatility with ministers occupying multiple positions, while in Togo the system is very stable with little movement. The table also gives scores for indegree centralization. Centralization describes the extent to which a network is organized around particular nodes. Although centralization is difficult to compare across networks, it is mentioned here because it shows a dynamic that provides some key information to the understanding of the differences between the countries: Togo stands out as the most centralized network, followed by Ghana. Gabon has the lowest score for indegree centralization, meaning that the network is decentralized and power is relatively evenly distributed. This tells us that in Togo, by comparison, there are a few individuals who are extremely popular.

Table 2 shows the correlation of the Minister variable with a number of network centrality measures. In all cases, association with the cabinet leads to more indegree ties but not more outdegree ties, showing that parliamentarians who are former/current ministers are sought out by others. Closeness centrality is significant in all cases: former or current ministers can more easily reach and influence a larger number of people in parliament.

Table 2. Comparing Minister against Non-Minister Parliamentarians – Correlations and t-test

Note: *** p < 0.001, ** p < 0.01, * p < 0.05, ⁺ p < 0.1.

Eigenvectors are negative in all cases, but only significant in Ghana and Gabon. This provides partial support for Bonacich's argument that powerful elites in fact build power through a network of less powerful and less well-connected clients. The fact that this is only the case in Gabon and Ghana reflects the volatility of these two political systems compared to Togo.

In Gabon, the volatility comes in the form of the various satellite parties in the mouvance présidentielle. Ministers in the Gabonese cabinet have the lowest median tenure in terms of months, and this is due to the rapid turnover of ministers from satellite parties (Table 3). Therefore there appears to be a stable core of important PDG elites who compete over an unstable cohort of non-PDG politicians.

Table 3. Ministerial Tenures by Country

To summarize these results, in Ghana, parliamentarians with ministerial experience play an integrative function, they connect people and are relatively close to everyone in the network. This gives them the chance to moderate and spread information. At the same time, they are in competition with each other, as the negative correlation with eigenvector shows. In Togo, only indegree centrality and closeness are significant – ministers have less bridging power than in Ghana but are also in a position of being in easy and quick communication with many nodes in the network. Gabonese ministers seem to build their power from weakly connected nodes and are probably more competitive, as they do not have the high betweenness scores that the Ghanaians have. Thus, the structure of the Gabon network is coming closer to the idea of competing patrons and their clients than is the case with the other two countries. This finding must be seen in the context of the frequent reshuffles: obviously, MPs build up some power from their connections to nodes of lesser importance. They are not connected, however, to other important nodes and do not seem to be able to build up a counter-elite.

Although Togo and Gabon might at first glance appear to fall into the same regime typology, their elite building strategies are very different. In Gabon there is rapid change that does not allow would-be barons to become powerful enough to present a real challenge. In Togo, as we shall see in the next part of the article, the powerful ministers are more strongly connected to the president's ethnic group, which gives them the chance of building comparatively more power without challenging the regime.

Next, a series of ERGMs is run. All models contain the term ‘edges’, which is the baseline propensity of any tie. The main effect that we are interested in is always the influence of a ministerial position on tie formation in the network. We tried to keep the variables in the models as comparable as possible. For each country, we include controls for gender, ruling party and network effects. For Ghana, we also control for Ashanti ministers as this ethnic group is regarded as especially powerful in the NPP. For Togo, we add the ethnic group of the current president, the Kabyé, to the controls. In Gabon, coalition building has been ethnically inclusive, so we do not test for single ethnic groups. Instead, we look at the mouvance présidentielle, a non-institutionalized coalition of parties that have joined the presidential majority in parliament. To control for the effect of political experience, we include a count variable of the number of legislative periods that a politician has spent in parliament.

We also include terms that control for network effects. The most important network effect that needs to be controlled for is transitivity – the likelihood that friends of a friend also become friends. Including these terms not only helps to understand the network structure better but is also important to exclude the possibility that our social selection effect – ministers have more ties than by chance – is merely caused by the underlying network structure; that is, that MPs are not selected as discussion partners based on their actor attributes, but merely as a consequence of triadic closure. We therefore add the geometrically weighted edgewise shared partner distribution (gwesp) and the geometrically weighted dyadwise shared partner distribution (gwdsp) to our models. Positive values for gwesp indicate that MPs who have one discussion partner in common are also likely to become discussion partners themselves; positive values for gwdsp indicate that any unconnected pair of MPs that has one partner in common is also more likely than by chance to have a second shared partner. Popularity spread (gwidegree) and activity spread (gwodegree) measure the tendency to which the whole network is centralized on indegree or outegree.

Model fit is assessed using the AIC (Akaike information criterion) and BIC (Bayesian information criterion), where lower values indicate better fits. We provide information on the model fit in the Online Appendix.

Ghana: dispersed power in a two-party democracy

Table 4 shows a number of bivariate models for Ghana. For the Minister variable (Model 1), the bivariate model shows a significant positive result: the number of ties increases with the number of ministerial positions. When the variable is disaggregated, we still find the same result for people who became ministers after they finished their term as MPs (Model 2), and for those who have been ministers before coming to parliament (Model 3), and for ministers in the inner circle (Model 5). Only people who were ministers and MPs at the same time do not have a higher tie probability than by chance (Model 4).

Table 4. Exponential Random Graph Model – Ghana: Ties against Ministerial Status

Note: *** p < 0.001, ** p < 0.01, * p < 0.05, ⁺ p < 0.1.

In Table 5, the Minister variable is split into indegree and outdegree (Models 6 and 7). As expected from the correlational analysis, we find a strong positive effect only for indegrees. Obviously, ministers receive more ties than they send out. Model 8 adds some controls – gender (0 for female, 1 for male), party (0 for NDC, 1 for NPP) – and the number of legislatures and four network structural effects: the geometrically weighted edgewise and dyadwise shared partner distributions (gwesp and gwdsp), as well as popularity spread (gwidegree) and activity spread (gwodegree).⁶

Table 5. Exponential Random Graph Model – Ghana 2

Note: *** p < 0.001, ** p < 0.01, * p < 0.05, ⁺ p < 0.1. For better visualization, the key variables are in bold.

We find a generally higher probability of a tie when one of the nodes belongs to the NPP, which tells us that the NPP is in sum better connected than the NDC. At the time of the survey, the NPP was in opposition while the NDC was in government. This shows that Ghana is unique among our cases in that dense networks of power can reside outside of the regime. Moreover, the number of legislative periods that an MP has served in the parliament also increases the likelihood of a tie. There is no effect for gender, which tells us that men are not more popular in the network than women.⁷

With regard to the network structural terms, we find a positive effect for edgewise and a negative one for dyadwise shared partners: MPs that share a discussion partner are more likely to become connected, but unconnected MPs that share a discussion partner are not more likely than by chance to share another discussion partner. Negative values for both activity and popularity spread tell us that the network is not highly centralized, either on indegree or on outdegree; MPs tend to have similar levels of popularity and activity. What is most important about the network terms is that the effect for the Minister variable remains significant – we can thus have more confidence in our finding on the social selection effect.

The Ghana data give us the opportunity to look specifically at those people who were appointed to ministerial positions after they had been MPs. Model 9 repeats the full model with the variable Minister after MP. There is not much difference to Model 8, but the coefficient for Minister after MP is higher. Obviously MPs who later became ministers had already built a network of dense relations in the parliament. This position of power and influence might have influenced their nomination.

To further check for the robustness of our results, we test a number of interaction effects: how does the party affiliation, gender or ethnicity interact with the Minister variable?⁸ Since the NPP took over power after the parliament network data were collected, we expect that most of these ministers belong to the NPP and that they are male. We furthermore expect a strong positive effect for Ashanti,⁹ which would tell us that the NPP has a preference to appoint MPs from their regional support base to ministerial positions. We fit separate models because the variables are highly correlated. Model 10 in Table 6 first tells us that there is a huge gender bias. Male ministers have a much higher tie probability than expected by chance, while the main effect for gender remains non-significant. Contacts among ‘ordinary’ MPs do not tend to be gendered, but male ministers tend to have more contacts with other male ministers than with female ministers. Model 11 shows a high coefficient for Ashanti: obviously, this ethnic group does play a huge role, and ministers of this ethnic group belong to a circle of power. The NPP government does – not surprisingly – appoint NPP MPs to government, as indicated by Model 12.

Table 6. Exponential Random Graph Model – Ghana: Interaction Effects

Note: *** p < 0.001, ** p < 0.01, * p < 0.05, ⁺ p < 0.1. For better visualization the key variables are in bold.

Though gender and ethnic favouritism play a role, they do not explain everything about personal power. In all models the Minister variable remains robust.

Togo: centralized power around an ethno-political dynasty

We repeat the same tests on Togo. However, the Minister value only refers to people who were previously ministers before the survey. Simultaneously holding a cabinet position and legislative seat is not allowed, and we have only one record of an MP who later became a minister (Table 7).

Table 7. Exponential Random Graph Model – Togo

Note: *** p < 0.001, ** p < 0.01, * p < 0.05, ⁺ p < 0.1.

For Togo, the results look quite similar to Ghana's: the Minister variable is positive and significant in almost all bivariate models, except for outdegrees. Just as in Ghana, Togolese ministers receive significantly more ties than they would by chance, but do not send out more ties. This finding is robust even when the controls are added (Model 17). The ruling party UNIR has a negative coefficient, which indicates that opposition MPs are more active in the network. We add Kabyé to control for the president's ethnic group, which holds considerable power in the country. As expected, being a Kabyé increases the likelihood of a tie being formed. Apart from that, neither gender nor the number of legislative periods served has an effect on tie formation in Model 17.

The network structural effects are similar to those for Ghana. The only difference is the positive coefficient for popularity spread (gwodegree), which is significant at the 10% level. This can be taken as an indication of the fact that the network is more strongly centralized on indegree, meaning that there is a small number of highly popular people.

In Table 8, we test for the interaction between having been minister with gender, with the ruling party UNIR, and with the president's ethnic group, the Kabyé. Across all models, the Minister variable remains significant, indicating that none of the interaction effects fully explains the popularity of certain actors. Nevertheless, we find a positive effect for all three interaction effects, with a high coefficient for Kabyé ministers. Ministers who belong to the president's ethnic group have considerable power, although belonging to UNIR and being a minister also increases the number of ties. This suggests that even within the ruling party, there is a strong ‘ethnic core’ that holds real influence. Power in Togo is thus ethnicized and highly concentrated. Another piece of evidence for the high concentration of power is the positive and significant coefficient for popularity spread, which indicates a high centralization on indegree.

Table 8. Exponential Random Graph Model – Togo: Interaction Effects

Note: *** p < 0.001, ** p < 0.01, * p < 0.05, ⁺ p < 0.1. For better visualization, the key variables are in bold.

Gabon: dispersed power around a volatile coalition

In Gabon, we find a similar picture: the probability of forming a tie increases significantly when one of the nodes has been a minister (Table 9). Outdegrees are negative. Here too, the ruling party PDG is less active than the opposition. The number of legislative periods has no influence on tie formation, but tie formation processes appear to be more gendered: males have an increased tie probability. While the coefficient for Minister is relatively low, it increases for the Inner circle variable. This pattern has been observed for the other countries as well, but for Gabon the effect size differs between the two variables. Obviously, there are more and less important ministers. Together with the fact that government reshuffles are more frequent, this suggests that power is more dispersed: there are on the one hand ministers with relatively moderate power, as well as those who hold important portfolios and are closer to the centre of decision-making. Interestingly, the network is still not highly centralized, as indicated by the insignificant effect for popularity spread.

Table 9. Exponential Random Graph Model – Gabon

Note: *** p < 0.001, ** p < 0.01, * p < 0.05, ⁺ p < 0.1.

In Gabon, we test for the interaction with gender, the ruling party PDG and mouvance présidentielle, the coalition of parties that joined the presidential majority in the parliament (Table 10). Being a minister and at the same time member of the PDG or the mouvance présidentielle significantly increases the likelihood of a tie. The coefficient for PGD is higher, which tells us that the distribution of power is of course highly influenced by party politics. The ruling party is an important locus of power, which does not mean, however, that ministers from other parties are not powerful. The main effect for Minister remains significant across the models. In other words, the power of a PDG minister is not necessarily higher than that of a non-PDG minister, and the same is true for the presidential majority and for gender.

Table 10. Exponential Random Graph Model – Gabon: Interaction Effects

Note: *** p < 0.001, ** p < 0.01, * p < 0.05, ⁺ p < 0.1. For better visualization, the key variables are in bold.

Discussion and conclusion

The most interesting observation of the article is the robustness of the association between having been a minister and being a central discussion partner in the parliament network. Ministers are more popular but not more active than other MPs. They receive their status from incoming connections – that is, they are sought after by other people. These results hold for all three cases. In all countries, the coefficient is higher for the Inner circle variable. This is another indication for the correlation between high formal power and strong personal networks. Even when interaction effects are controlled for, the positive and significant effect of the Minister variable remains robust. Thus, there is very clear evidence for H1.

We also find strong support for H2, which could be tested only for Ghana. Not only does ministerial status increase the likelihood of parliamentarians receiving more incoming ties, but a better network position increases the likelihood of a parliamentarian being elevated to the executive. This means that the significant positive relationship between ties and ministerial status cannot be interpreted solely as a function of current access to the executive.

The results suggest a circular relationship between executive and legislative power: parliamentarians with a high degree of political clout are more likely to enter the executive, while former or current ministers who enter parliament are sought after as powerbrokers by other parliamentarians. In other words, personal power ‘travels’ with individuals across institutions. This finding carries a number of implications for the current literature on executive–legislature relationships in Africa. The consistent positive significance of indegree ties for current and former ministers reflects that parliamentarians perceive executive elites as vitally important powerbrokers within the government. Ministers often control slush funds that are protected from judicial or parliamentary oversight (McKie and van de Walle 2010). Existing research has shown that parliamentarians perceive their responsibility to deliver private and public goods to their constituents as more pressing than issues of executive oversight (Lindberg 2010). Thus, it can be more rational for MPs to build personal relations to people with access to executive power than to work towards parliamentary oversight and the separation of powers. At the same time, nomination patterns and institutional settings (the suppléant in Gabon or the simultaneous holding of parliamentary and government positions in Ghana) give governments an option to control the parliament, reward loyal supporters and control or inhibit the formation of alternative power centres.

There is, however, little support for H3. Given the limited sample of three countries, support for H3 would never be robust. Furthermore, the potential interference of other variables such as institutional variation means that we would need a much wider sample to prove H3 robustly. However, the results suggest that more autocratic states do not necessarily lead to a stronger correlation between personal power in parliament and presence in the cabinet.

Personal relations do not disappear with higher levels of democracy, and, by the same token, countries that seem to belong to the same regime type can exhibit very different underlying dynamics. While there is a close relationship between formal power and informal/personal networks in all the examined case studies, different political environments manifest different models of elite integration. Even in the purportedly similar cases of Gabon and Togo, interaction patterns between elites differ. In Togo, power is concentrated in the hands of a small, ethnically defined group. Ministers have a comparatively longer tenure and become powerful individuals closely tied to the regime. In Gabon, elite rotation is more rapid. Under the condition of a multi-ethnic elite, everyone can theoretically get a share of state power. Due to frequent reshuffles, however, individuals hardly get powerful enough to build strong alliances that could give rise to alternative centres of power. The outcome is a parliamentary network of rather dispersed power: a number of competing ‘patrons’ build relations to less powerful MPs but do not cooperate among each other. In Ghana the correlation between a strong personal network in parliament and ministerial status (past and future) appears to be particularly strong. The literature on political competition in Ghana has argued that institutional and historical conditions created a clientelist democracy (Driscoll 2020; Paller 2014). Although there is competition between two strongly institutionalized parties – leading to favourable outcomes such as regular turnover – both parties are internally permeated by clientelistic networks (Driscoll 2020). It is therefore reasonable to assume that upward mobility within the political elite continues to rest on personal contacts.

Our findings point to the need for future research. While the importance of elite accommodation has been empirically demonstrated by authors such as Leonard Arriola (2009) and Philip Roessler (2011), we add nuance to this literature. Future research needs to address how different models of elite accommodation are shaped by party systems and their historical legacies (see Sanches 2018: 3–10). Second, we have shown that formal and informal institutions interact with each other. In this way, our findings support the more recent agenda of comparing different types of patronage democracies (Berenschot and Aspinall 2020) rather than simply equating democracy with formal institutions and depersonalized politics.