Executive Cabinets and African Politics

In this section we document the dramatic increase in the number of ministers in African cabinets over the last four decades and then draw upon the comparative literature to explain why parties are an essential part of understanding cabinets. Next, even though this study is focused on the effects of Africa’s growing cabinet size, we briefly review the literature examining its causes, including population growth, institutional reforms related to democratization, cultural norms of inclusion and post-conflict power-sharing agreements. All of these factors underscore the need to understand the consequences of Africa’s large cabinets.

In a 1989 essay written just before Africa’s era of dramatic democratization, a leading Nigerian political scientist noted ‘a phenomenal increase in the size of the modern cabinet’ (Osaghae Reference Osaghae1989: 129). Africa’s cabinets have in fact grown from an average of 16.5 ministries in 1971 to nearly 26.2 in 2006. Figure 1, which includes 1,561 observations from the Arthur Banks data set, illustrates this steady growth. Uganda provides one of many examples of this trend: when Yoweri Museveni took power in 1986 there were 33 ministers in his cabinet. Two decades later this number had grown to 69 ministers as he adapted to pressures to include a broader range of regional interests and ethnic groups (Tripp Reference Tripp2010: 49). ‘The cabinet is not only large but virtually useless’, said one Ugandan scholar. ‘The intention of the cabinet is to appease the head of state. It is based on the idea that when you have a huge cabinet, it is incapable of making serious decisions’ (Kiapi Reference Kiapi2010). The chair of Uganda’s parliamentary budget committee called the cabinet ‘too big for the economy’ (Osike and Olupot Reference Osike and Olupot2002), and shrinking the cabinet has been a top priority for the opposition. ‘The size of this cabinet is too big. We think it should be smaller’, said the head of the leading opposition party. ‘They have come to waste taxpayers' money yet we do not have the money’ (Namutebi et al. Reference Namutebi, Ariko and Nalunkuuma2006).

Figure 1 Average Cabinet Size, Number of Government Seats, in Africa since 1971

Source: Banks and Wilson (Reference Banks and Wilson2014).

Scholars often analyse the number of parties alongside the number of ministers because the two are closely related. Comparative research finds that the number of ministries tends to increase with the number of parties, at least when parties seek policy change (Laver and Schofield Reference Laver and Schofield1998). According to Lijphart’s seminal study (2012), the number of parties and ministries in a cabinet is the most important indicator of how representative the government is of society, and this is a central distinction between his two contrasting models of democracy, consensus and majoritarianism, which fundamentally differ in terms of the breadth of representative participation in the executive branch. In Africa, cabinets are especially important because the composition of ministries is a common gauge of the inclusiveness of government. For example, in Nigeria, the representative quality of the ministries is a ‘cardinal principle’ of government: ‘For practical politics and peaceful coexistence of diverse elements, the composition of the cabinet should reflect the diversities in a polity’ (Osaghae Reference Osaghae1989: 131). Posner’s study of Zambia (Reference Posner2005) claims that the ethnic background of cabinet ministers informs voter choice and politicians’ electoral strategies, reflecting an ethic of inclusion. Some cross-national research goes so far as to use cabinet size as a proxy variable for inclusiveness, since cabinet size correlates closely with level of ethnic diversity (Arriola Reference Arriola2009). Cabinet composition is also often used to estimate the political salience of ethnic groups or their relative access to power (Posner Reference Posner2004; Wimmer et al. Reference Wimmer, Cederman and Min2009).

At least four demonstrable factors have driven the growth of Africa’s cabinets. First, a demographic connection follows from a strong correlation between cabinet size and population size. Africa’s cabinets have apparently grown in part because the continent has experienced rapid population growth (Arriola Reference Arriola2009). This makes sense if one considers the division of labour in government expanding alongside increasing social complexity. Second, democratization in the 1990s influenced cabinet size because the return of multiparty politics ushered in a new wave of coalition governments (Oyugi Reference Oyugi2006). This is empirically reflected by the growth of multiparty coalition governments during this period: out of the 74 per cent of governments for which we have data, in 1971 less than 3 per cent were coalition governments and less than a quarter were civilian. By 2006, 55 per cent of the governments across 44 African countries governed through multiparty coalitions, and 98 per cent of all cases had civilian rulers.

Conflict resolution strategies are a third factor contributing to the size of African cabinets. The African Union formally recommends proportional representation for conflict mitigation (Murray Reference Murray2005), an electoral arrangement which increases the number of parties and the number of ministries in government. Pacts to protect the rights of sectarian interests in order to prevent backsliding into war have increased cabinet size through post-conflict power-sharing agreements in Sierra Leone (1996), the Democratic Republic of the Congo (2002 and 2003), Côte d’Ivoire (2003) and Sudan (2004). As part of its interim government, in 2006 Somalia formed a hugely oversized cabinet (Menkhaus Reference Menkhaus2008). In 2007, Sierra Leone and Côte d’Ivoire both formed governments of national unity designed to bring in former militants. In the year following each of these power-sharing agreements, the number of cabinet ministries increased. In nations tired of violence, demand for peace creates compelling incentives to form broadly inclusive governments, even if they impose financial burdens (Jarstad and Sisk Reference Jarstad and Sisk2008).

Norms of inclusion stand out as a fourth factor arguably shaping African cabinets. In Sundiata, the epic story telling the rise of Mali’s empire in the thirteenth century, the scheming half-brother of the story’s hero is ridiculed because he selfishly declares ‘power cannot be shared’ (Niane Reference Niane1965). Africa’s nationalists later embraced this sort of rhetoric. Tanzania’s Julius Nyerere believed his ‘inclusive, participatory’ model of governance with consensus building would both deepen democracy and facilitate economic growth (Baregu Reference Baregu1994). In pre-independence Nigeria, several constitutions enshrined political inclusion through ethnic balancing of the bureaucracy and the legislature’s top positions (Ekeh and Osaghae Reference Ekeh and Osaghae1989). For over half a century, dictators and democrats alike have implemented such practices. Senegal provides another example, where the discourse of democracy emphasizes the Wolof idea of demokaraasi. In this tradition, voters value conformity because it brings social security and they seek consensus because it brings stability. Including different opinions and ‘harmonizing’ them is therefore more important than alternance – alternating political power between parties (Gellar Reference Gellar2005; Schaffer Reference Schaffer1998).

In sum, the number of ministers and the frequency of coalition governments have grown dramatically since the 1970s for a variety of reasons, according to the existing literature, including population growth and post-conflict power-sharing agreements. There is also some evidence that large cabinets reflect African cultural traditions of inclusiveness. But many citizens value majoritarian qualities that promote alternating power rather than sharing it; African satisfaction with democracy in fact increases substantially after the electoral defeat of the incumbent party leads to alternation (Bratton Reference Bratton2008). Anecdotal evidence reinforces such cynicism about large cabinets. For example, in Zimbabwe, lawyers called the national unity government unconstitutional (Dube Reference Dube2009), while a spokesperson of a major party which joined the coalition said, ‘We have serious misgivings with the size of the cabinet particularly at a time when the economy is in such a bad state’ (Nyathi Reference Nyathi2009). As noted in the introduction, similar criticism erupted following Kenya’s formation of a unity government in 2008 (LeVan Reference LeVan2011). Moreover, the constitutional referendum in 2010 essentially rejected that power-sharing agreement, endorsing a largely presidential, majoritarian model of government.

All of this suggests that the literature would benefit from an empirical examination of the impact of cabinet size, and that African citizens are demanding answers. The next section clarifies our definitions and describes some comparative findings about the economic impact of cabinet size.

Are Africa’s Cabinets Cumbersome?

Complaints about big cabinets appear to be common in Africa. Yet we have few empirical studies which examine the economic effects of cabinet size outside developed countries. In this section we state our definitions of who constitutes a minister and what counts as a coalition government, drawing on the comparative literature. Next, we discuss research that comes to conflicting conclusions about the economic impact of large cabinets. On the one hand, large cabinets should undermine budgetary responsibility because policy logrolls become expensive in the aggregate. On the other hand, large cabinets could improve economic performance by making the policy process more specialized and creating new incentives for intragovernmental accountability. Finally, we identify common institutional features in Africa that make portfolio allocation especially important – and individual ministers unusually weak.

Who counts as a minister? This is important not only for operational purposes, but because it sets a threshold for determining who actually has the ability to affect spending decisions. One recent study suggests that relying on function to operationalize a definition can present more complications than departing from a legal, institutional understanding (Barbieri and Vercesi Reference Barbieri and Vercesi2013). A standard approach is to count only officials with full ministerial rank, thus excluding deputy ministers, secretaries of state, regional ministers or other officials who would add even more to these numbers (van de Walle Reference van de Walle2001). To the best of our knowledge, the Banks data set follows this definition closely. As for a multiparty coalition government, the classic understanding stems from parliamentary politics, where a coalition is a set of political parties who: (1) agree to pursue common goals; (2) pool their resources in order to achieve these goals; (3) communicate and form binding commitments concerning their goals; and (4) agree on the distribution of payoffs to be received after the coalition meets its objectives (Browne and Dreijmanis Reference Browne and Dreijmanis1982). Thus the size of the coalition derives from a direct measure of the number of political parties in the cabinet rather than from a proxy. Popular accommodation rather than elite bargaining drives coalition size. On average, this means that multiparty coalitions produce cabinets with more portfolios (Laver and Schofield Reference Laver and Schofield1998).

Arguing in favour of coalition governments in general, Lijphart argues that increasing representation in government does not require sacrificing economic performance. He stops short of definitely affirming that ‘consensus governments’, which he associates with multiparty cabinets (and other features such as federalism), perform better in terms of budget deficits and other measures of economic performance. Thus his influential study of 36 countries explicitly leaves open the door for additional empirical testing, though his results tend to support multiparty governments (Lijphart Reference Lijphart2012). South Africa provides one example of how the creation of ministries as tools of inclusion need not force a ‘tradeoff’, in Friedman’s words (Reference Friedman2004), between good government performance and democratic quality. Jacob Zuma created a slew of new ministries shortly after his election to the presidency in 2009, when the economy was contracting slightly; over the next three years macroeconomy improved, growing at 3 per cent on average and deficits dropped by about one percentage point. This might seem puzzling to those familiar with the towering (though arguably declining) presence of the African National Congress. But each new ministry was ‘shadowed’ by a corresponding oversight committee in a parliament that demonstrated autonomy by challenging the executive, and that gave opposition parties a disproportional share of committee seats (Barkan Reference Barkan2009).

Zimbabwe provides an example of how multiparty cabinets (rather than simply the number of ministers) might affect the economy. Though there is ample evidence that the coalition government formed in 2008 did little to advance democratization or to limit political violence in the long run (Lebas Reference Lebas2011), it did halt the country’s economic slide. Within two years, gross domestic product (GDP) went from a 12.5 per cent contraction to 3.5 per cent growth; government revenues increased and inflation stabilized (Africa Confidential 2010). By joining the cabinet, the opposition party, with strong roots in labour and civil society, successfully advanced some constructive economic policies despite the authoritarian practices of its coalition partner. In particular, the government’s most controversial policy, seizing white-owned farms, largely stopped after the opposition party joined the government.

By contrast, other studies find that large cabinets undermine economic performance. Van de Walle (Reference van de Walle2001), for example, points to Nigeria and Côte d’Ivoire, where big cabinets engaged in reckless levels of spending during the 1990s. One classic study reports that multiparty coalition governments are much more likely to run budget deficits (Alesina et al. Reference Alesina, Roubini and Choen1997). The World Bank explains that since coalition governments require so many compromises and payoffs, ‘fiscal outcomes are often worse than when majority governments are in power’ (World Bank 2002). Consistent with these findings, Persson and Tabellini (Reference Persson and Tabellini2003) report that majoritarian (that is, single-party) governments produce smaller deficits, regardless of whether the system is parliamentary or presidential. Wehner (Reference Wehner2010) more modestly reports that single-party cabinets can mitigate but not necessarily prevent deficit spending by reducing the conditions of partisan fragmentation. Other studies focused specifically on the number of ministries, rather than the number of parties, clearly link additional portfolios to larger budget deficits (Volkerink and de Haan Reference Volkerink and de Haan2001; Woo Reference Woo2003). Almost all of the cross-national research is limited to wealthy countries in the Organization for Economic Cooperation and Development (OECD).

The existing literature thus arrives at conflicting conclusions, and we know very little about the economic effects of cabinets in Africa. Cabinets could indirectly have an impact on the economy in a variety of ways, shaping the political conditions for economic growth or the ability to discipline monetary policy. But cabinets have more direct effects on public finance since they debate and shape spending priorities. For this reason it makes sense to focus on budgetary policy, particularly the impacts on revenue, spending, and the joint interaction of the two.

A causal link between cabinet size and budgetary policy could manifest itself in two ways. First, an increase in cabinet size could mean more political actors demanding resources or receiving side payments. In such situations policymaking becomes a ‘logroll’ that supplies payoffs to everyone (Strøm Reference Strøm2003). This becomes financially costly in the aggregate and appears as a strain on national budgets. Popular political parlance in Africa characterizes logrolls as sharing ‘the national cake’: politicians dividing up the spoils of the state among themselves (Gana and Egwu Reference Gana and Egwu2004; Kiapi Reference Kiapi2010). Second, large cabinets could actually limit such patronage and reduce the incentives to extract resources from citizens. ‘Public welfare is enhanced when leaders depend on a large coalition to keep them in office’, argue Bueno de Mesquita et al. ‘Under these conditions, those motivated to stay in power have no choice but to promote the public’s welfare’ (Bueno de Mesquita et al. Reference Bueno de Mesquita, Morrow, Siverson and Smith2001: 71). As in Lijphart’s studies, the composition and size of the cabinet is an important component of their ‘selectorate’ theory, which focuses on the size of the eligible pool of potential rulers. Small coalitions, they say, ‘promote corruption, black marketeering, and cronyism’ (Bueno de Mesquita et al. Reference Bueno de Mesquita, Smith, Siverson and Morrow2003: 214). Comparative institutional literature claims that such inclusive policy processes are conducive to long-term economic development (Acemoglu and Robinson Reference Acemoglu and Robinson2012).

However, the number of ministers and the number of parties in the cabinet could affect these two potential outcomes quite differently. Additional ministers could increase costly demands for patronage, leading to suboptimal policy performance, or it could enhance policy specialization and technocratic capacity, leading to improvements in budgetary policy performance. Additional parties could actually be the condition driving patronage demands, along the lines of the World Bank’s comments. After all, individual ministers have less leverage to make patronage demands, and they face collective action problems that parties can resolve by lowering the costs of information and coordination (Lupia and Strøm Reference Lupia and Strøm2008). Alternatively, additional parties governing in coalitions might increase horizontal monitoring, creating a government in which executives face multiple political actors with sufficient leverage to hold them accountable for economic policy. It is therefore important to test for the impact of both ministries and parties on budgetary policy.

Before doing so, though, it is important to point out common characteristics of the institutional environment in Africa. First, executives enjoy a variety of advantages over parties. Nearly every African executive is classified as a presidential republic in which the head of state is also the head of government, or a mixed republic with a president and a prime minister; a few countries are actually monarchies or pure military states.Footnote ¹ This is significant because, as Africa’s post-independence leaders noted, the collective responsibility of parliamentary government creates incentives for restraint (Hatchard et al. Reference Hatchard, Ndulo and Slinn2004: 69). Second, even in presidential systems, African executives often appoint sitting members of the legislature to the cabinet, giving them the ability to exercise tremendous power over individual ministers’ behaviour (Lindberg and Zhou Reference Lindberg and Zhou2009). Some constitutional scholars in fact argue that the level of executive discretion over cabinets corresponds with the number of ministries (Hatchard et al. Reference Hatchard, Ndulo and Slinn2004). Third, modern African political parties have very few ideological differences (Bleck and van de Walle Reference Bleck and van de Walle2011). This makes standard approaches from the American or European literature, such as the use of a left–right continuum and median voter models, less useful for understanding cabinet behaviour (and nearly impossible to estimate). In sum, given the absence of collective responsibility, the fusion of legislative and executive authority through the appointment of sitting members of the legislature, and weak ideological cleavages, African executives enjoy great latitude.

Operationalizations and Hypothesis Formulation

Whether one measures cabinet size in terms of parties or ministries, there are compelling reasons to expect either good or bad economic performance, and existing comparative research points to conflicting findings. We formulate two hypotheses to test the impact of both measures on budgetary policy performance. A ‘portfolio’ hypothesis tests to see if cabinets with more ministers improve budgetary policy performance. Confirmation of this hypothesis would suggest that additional ministries increase specialization, much like the narrative of Nigeria in the introduction suggested, whereas rejection would imply that the new portfolios are merely patronage payoffs that undermine the economy overall. A ‘parties’ hypothesis tests the record of coalitions. Support for this hypothesis would imply that multiparty governments improve budgetary policy through horizontal monitoring, as the cases from South Africa and Zimbabwe suggested. Alternatively, multiparty governments might simply engage in costly policy logrolls, similar to wealthy democracies (Alesina et al. Reference Alesina, Roubini and Choen1997; Persson and Tabellini Reference Persson and Tabellini2003). Using both measures of cabinet size is important since frustrations about cabinets in Uganda, Kenya and elsewhere are directed towards the number of ministries as a driver of economic wastefulness, while the broader comparative literature largely focused on the impact of the number of parties.

Our theory predicts that specialization through additional ministries contributes to good budgetary policy, but that cabinets also need horizontal accountability provided by multiparty coalitions to moderate the power of strong executives. Individual ministers face high transaction costs of coordination and few incentives to make demands of the executive. In coalition governments, parties help internally coordinate and articulate distinct interests. By doing so, they increase the opportunities for horizontal monitoring across government. All else being equal, multiparty coalitions should therefore have better budgetary performance overall compared with governments that simply increase the number of ministries. The capacity to conduct this monitoring, though, could be disrupted by strong executives or the absence of opposition parties. Gabon offers an example of this, where Omar Bongo ruled from 1967 to 2009. There was virtually no political party competition, and the country displayed the greatest variation in the number of ministers in our data set (fluctuating between 20 and 53 over 35 years), as portfolios served as patronage tools. In this context, neither parties nor portfolios should have much of an impact on accountability (which we will control for below). In sum, we argue that African ministers possess weak incentives and little individual capacity to demand patronage, while parties in coalition governments use their political leverage to increase accountability across government.

We measure budgetary policy performance with three different dependent variables. First, the variable budget is the federal budget surplus or deficit, expressed as a share of GDP in local currency units, similar to operationalizations adopted by Roubini and Sachs (Reference Roubini and Sachs1989a, Reference Roubini and Sachs1989b). Second, we separately test for the impact of cabinet size on revenue extraction by the central government with the variable revenue. This is also important because low government revenues could bias our variable measuring budget surpluses, and because the rate of taxation has been used in studies of Africa to link weak government capacity to state failure (Bates Reference Bates2008) or economic collapse (van de Walle Reference van de Walle2001). In a more conventional economic sense, revenue provides a test of the relationship between taxation and representation to see if more inclusive governments actually extract less revenue. Third, the variable expenditure measures government consumption as a share of GDP, which is common proxy for patronage in cross-national studies in the developing world (Arriola Reference Arriola2013). Directly measuring different categories of disaggregated spending (health, defence, and so on) would be ideal but such data are scarce in Africa, especially during the 1970s. Regardless, if cabinets allocate patronage to reward allies or buy off potential critics, this will appear in higher overall levels of spending. These three variables are from the World Bank’s World Development Indicators (2012).

Empirical Tests for Portfolios and Parties

We perform regression analysis to obtain OLS coefficient estimates with panel corrected standard errors for each of our dependent variables (budget, revenue and expenditure).Footnote ⁵ This is an appropriate technique since the data contain fewer years than cases cross-sectionally. Consistent results across different models is a standard econometric check for robustness (Greene Reference Greene2008). We first test our hypotheses on budget as the dependent variable as specified in Equation 1:

$$\eqalignno{ Budget_{{it}} = & a_{i} {\plus}\beta _{1} Ministries_{{it}} {\plus}\beta _{2} Coalition_{{it}} \cr & {\plus}\beta _{3} {LiberalDemocracy}{\plus}\beta _{4} Xconstrain_{{it}} \cr & {\plus}\beta _{5} Britcol_{{it}} {\plus}\beta _{6} GrowthGDP_{{it}} \cr & {\plus}\beta _{7} Oilrent_{{it}} {\plus}\beta _{8} LogPop_{{it}} {\plus}\beta _{9} EthnicF_{{it}} \cr & {\plus}\beta _{{10}} {\mathop{Re}\nolimits} ligiousF_{{it}} {\plus}\beta _{{11}} Civilian_{{it}} \cr & {\plus}\beta _{{12}} SingleParty{\plus}e_{{it}} $$

Table 1 reports regression results testing the model under different statistical specifications. The positive coefficients on the ministries variable across all models indicate that cabinets with more portfolios are more likely to run a budget surplus. One additional cabinet member implies a 0.2 to 0.3 per cent larger surplus (or a smaller deficit). The coalition variable has a significant effect on budget in one of the models, where coalition governments run deficits of about 2.2 per cent. The presence of electoral democracy in Africa reduces the budget deficit by 0.1 per cent, at a statistically significant level according to all models. Executive constraints do not have a statistically significant effect on budget. Governments in former British colonies are more likely to run a larger surplus (or smaller deficits) on average between 2 and 3.4 percentage points. The effect of overall economic conditions on the budget is positive and significant in Models 3 and 4: each percentage of GDP growth reduces the budget deficit by 0.1 per cent. In Models 2, 3 and 4, countries with oil revenue appear to run a budget surplus of about 0.1 per cent for each additional 1 per cent of oil rent per GDP increase. The significance of SingleParty suggests that party polarization has an adverse impact on budget surpluses; the same goes for ethnic but interestingly not religious fractionalization. Civilian government has no statistically significant impact. The results thus far point to statistically significant support for the portfolio hypothesis, suggesting that additional ministries enhance budgetary specialization, but they neither support nor refute the parties hypothesis, since coalition governments have a very weak effect on the budget variable.

Table 1 Dependent Variable Budget (budget surplus/deficit as share of GDP)

Note: Standard errors in parentheses. *, **, *** denote significance at the 5%, 1%, 0.1% level, respectively.

The second test of our hypotheses considers the effects of cabinet size on revenues, to see if large cabinets are more extractive. These tests also allow us to ensure that the results with the budget variable were not biased by lower revenue collection. In Equation 2, Revenue _it is the government revenue of the i^th country collected at the t^th year:

$$\openup -1pt \eqalignno{ {\mathop{Re}\nolimits} venue_{{it}} = & a_{i} {\plus}\beta _{1} Ministries_{{it}} {\plus}\beta _{2} Coalition_{{it}} \cr & {\plus}\beta _{3} {LiberalDemocracy}{\plus}\beta _{4} Xconstrain_{{it}} \cr & {\plus}\beta _{5} Britcol_{{it}} {\plus}\beta _{6} GrowthGDP_{{it}} {\plus}\beta _{7} LogPop_{{it}} \cr & {\plus}\beta _{8} Oilrent{\plus}\beta _{9} EthnicF_{{it}} {\plus}\beta _{{10}} {\mathop{Re}\nolimits} ligiousF_{{it}} \cr & {\plus}\beta _{{11}} Civilian_{{it}} {\plus}\beta _{{12}} SingleParty{\plus}e_{{it}} $$

The estimation results of Equation 2 are displayed in Table 2. The statistically insignificant values on ministries mean that cabinets with more portfolios do not extract more revenue through higher tax rates or more zealous collection. In relation to the previous tests, this means that the good performance on the budget variable is actually being underwritten by fiscal discipline, rather than extractive policies. At the same time, coalition governments systematically extract less: specifically, they collect on average 6.2 to 7.7 per cent less revenue as a share of the GDP in Models 2, 3 and 4. In terms of our controls, democracies collect on average 0.1 per cent more revenues. Stronger executive constraints are also associated with higher tax rates or better tax collection, ranging between 1 and 1.1 percentage points on revenue. These two control variables suggest an implicit relationship between democracy and taxation in Africa, whether it is measured in terms of horizontal accountability (captured by the Xconstrain variable) or electoral democracy. The control for British colonies is significant, indicating that these countries extract 1.8 to 2.6 additional percentage points more in Models 1 and 2. Countries with oil revenue appear to collect on average more budget revenues by about 0.3 and 0.4 per cent extra revenues for each additional 1 per cent of oil rent per GDP. Increased ethnic fractionalization unsurprisingly reduces the revenues collected by around 15 per cent, according to Models 3 and 4. But as in the tests with Equation 1, religious fractionalization is an insignificant factor. Civilian governments – whether dictatorships or democracies – extract on average 3.1 per cent more revenue. This time, single-party governments have no significant impact, most likely because rulers who want to avoid open multiparty competition see political risks in extracting too much from citizens. In sum, since coalition governments systematically extract less from citizens, this supports the idea that parties rather than simply portfolio promote good budgetary policy.

Table 2 Dependent Variable Revenue (government revenue, excluding grants, as a share of GDP)

Note: Standard errors in parentheses. *, **, *** denote significance at the 5%, 1%, 0.1% level, respectively.

The third test of our hypotheses considers the effects of cabinet size on government consumption as a share of GDP, which we use as a proxy for patronage. If additional ministers or parties increase the overall level of patronage, this will be visible as positive coefficients on the expenditure variable, suggesting these governments engage in policy logrolls to satisfy everyone. However, if new ministries increase specialization, or if coalition governments increase horizontal accountability as parties monitor each other, then government consumption will decline. In Equation 3, expenditure _it is government expenditure of the i^th country during the t^th year:

$$\eqalignno{ Expenditure_{{it}} = & a_{i} {\plus}\beta _{1} Ministries_{{it}} {\plus}\beta _{2} Coalition_{{it}} \cr & {\plus}\beta _{3} LiberalDemocracy{\plus}\beta _{4} Xconstrain_{{it}} \cr & {\plus}\beta _{5} Britcol_{{it}} {\plus}\beta _{6} GrowthGDP_{{it}} \cr & {\plus}\beta _{7} LogPop_{{it}} {\plus}\beta _{8} Oilrent{\plus}B_{8} EthnicF_{{it}} \cr & {\plus}\beta _{9} {\mathop{Re}\nolimits} ligiousF_{{it}} {\plus}\beta _{{10}} Civilian_{{it}} \cr & {\plus}\beta _{{12}} SingleParty{\plus}e_{{it}} $$

The estimation results of Equation 3 are displayed in Table 3. There is weak evidence in Model 3 that each additional ministry results in 0.2 percentage points more government expenditure as a percentage of GDP. The insignificant coefficients of ministries in the other models offer no support for the portfolio hypothesis. By contrast, multiparty coalition governments systematically have lower rates of government consumption at a statistically significant level. Specifically, coalition governments on average spent 5.4 to 6.2 per cent less as a share of the GDP in Models 2, 3 and 4. The level of democracy is not statistically significant in any of the models. Stronger executive constraints are associated with increases in patronage spending, ranging between 0.9 and 1.4 percentage points across all four models at a statistically significant level. Taken together, the results with these two controls suggest that it is specifically parties in government, rather than other mechanisms of horizontal accountability (or electoral democracy) that restrains patronage spending. The population control in Model 2 suggests that a 1 per cent population increase may increase the government spending by an 0.9 additional percentage points. Ethnic fractionalization appears to reduce government consumption spending between 6.9 and 8.2 per cent, as suggested by estimates from Models 3 and 4. On the face of it, this seems odd. However, it may suggest that in Africa’s highly heterogeneous nations, there are significant transaction costs involved in colluding across ethnic groups in order to distribute patronage. British colonial history, religious fractionalization, civilian and single-party governance have no statistically significant impact in any of the models. In sum, rather than engaging in policy logrolls through collusion, coalition governments reduce the level of patronage spending through horizontal monitoring, offering strong support for our parties hypothesis.

Table 3 Dependent Variable Expenditure (government expenditure, as a share of GDP)

Note: Standard errors in parentheses. *, **, *** denote significance at the 5%, 1%, 0.1% level, respectively.

These three sets of tests show that cabinet size affects budgetary policy in Africa, but that it is important to analyse parties alongside ministries. There is evidence that additional ministries increase policy specialization, along the lines of the example from Nigeria mentioned in the introduction, since these governments are associated with budget surpluses. But significantly, the absence of a strong relationship with the expenditure variable suggests that ministers possess neither the leverage to demand patronage for themselves, nor the resources to dispense it on a significant level. Coalition governments, though, extract less and spend less, suggesting that multiparty cabinets are both less predatory towards citizens and less likely to indulge in patronage spending. The results hold across a broad range of controls.

Based on these results, we argue that the impact of Africa’s cabinets on budgetary policy depends more on parties than portfolio. We have no trouble accepting the idea that cabinet appointments serve as important patronage tools for executives, as others claim (Arriola Reference Arriola2009; van de Walle Reference van de Walle2001). But our results show that ministers acting as individuals face significant collective action problems, and the mere presence of a multiparty system, measured by the SingleParty variable, does little to overcome these barriers. Africa’s executives seek an efficient coalition that affords them political latitude, and multiparty coalition governments can limit executive discretion by lowering the transaction costs of horizontal accountability across the government. In this way, they may serve as useful stepping stones – until political competition is more robust – to improved economic outcomes.