National-Level Portfolio Allocation: Empirical Evidence

In his famous article, Gamson argues that ministerial portfolios are distributed among government parties in strict proportion to the number of seats (resources) that each party contributes to the government's legislative total.Footnote ⁷ According to his argument, regressing a party's share of ministerial portfolios on the share of seats that it contributes to the government's legislative majority should produce an intercept of zero and a slope of one. In Figure 1, we plot the share of cabinet portfolios controlled by a government party against its share of the government's legislative seats in fourteen West European countries from 1945 to 2000.Footnote ⁸ The strong positive relationship shown in Figure 1 is often taken as evidence in favour of Gamson's argument, and is what the literature refers to as Gamson's Law.Footnote ⁹

Fig. 1

Portfolio allocation in Western Europe

Notes: The gray circles indicate a government party's share of cabinet portfolios and its share of the government's legislative seats. Data are for fourteen West European countries from 1945 to 2000.⁸ The dashed line represents the predicted values from a model in which we regressed a party's share of cabinet portfolios against its share of the government's legislative seats. We report the intercept and slope coefficients; standard errors are given in parentheses. The solid line represents the predicted values that we would obtain if portfolios were allocated in a perfectly proportional manner.

Note, though, that portfolios are not allocated in a perfectly proportional manner. The solid black line in Figure 1 indicates the 1:1 relationship that would exist if portfolios were allocated proportionally. The gray circles tend to be above this line when a government party is small, but below it when a government party is large. This trend is confirmed by the dashed line in Figure 1, which represents the predicted values from a model in which we regressed a party's share of cabinet portfolios against its share of the government's legislative seats. The slope of this dashed regression line (0.79) is significantly less than one and the intercept (0.07) is significantly greater than zero, indicating that smaller parties are overcompensated when it comes to portfolio allocation while larger parties are undercompensated.Footnote ¹⁰ Significantly, country-specific regression models indicate that this pattern occurs not only on average across these fourteen countries but also within each country. But why does this empirical regularity exist?

Bargaining Models and No Confidence Votes

Theoretical accounts of the government formation process are typically based on the canonical (closed-rule) Baron-Ferejohn model, in which three parties bargain over forming a new government by making alternating offers.Footnote ¹¹ In this type of model, a proposer (formateur) is chosen in order of party size, either going in order from the largest to the smallest party, or probabilistically, where the likelihood of being selected is proportional to size.Footnote ¹² The formateur then makes a proposal, typically involving a distribution of ministerial portfolios and a government policy position, to the other parties. All parties then vote on the formateur's proposal. If the proposal receives majority support, then a new government takes office. Otherwise the exogenous selection mechanism is called upon again to designate a new proposer and the process begins once more. Prior to a new government proposal being accepted, all parties receive identical payoffs, which are often normalized to zero.

In this model, the formateur uses policy concessions and ministerial portfolios to build a legislative majority. The formateur gives out just enough portfolios to his or her proposed coalition partner to make it willing to join the government rather than continue with another round of bargaining. Bargaining itself is costly: the longer it takes to form a government, the less time cabinet parties have to enjoy the benefits of being in power. Because potential coalition partners are aware that delaying the government formation process shrinks the size of the overall pie, the formateur need only offer them the equivalent of the discounted goods that would be available in later bargaining rounds to get them to agree to his or her proposal immediately. This leaves ‘extra’ portfolios on the table that the formateur can keep. Because formateurs are likely to be large parties (due to the exogenous selection mechanisms used to choose proposers), larger parties will, on average, receive more than their fair share of portfolios, and smaller parties will receive less.

As the empirical evidence in the previous section indicates, though, it is smaller parties that receive a ‘bonus’ in portfolio allocation. One explanation for this gap between theory and empirics can be traced to the fact that almost all government formation models end once a cabinet is formed. These models implicitly assume that cabinet parties receive all of the benefits associated with being in power as soon as the bargaining is successfully concluded. In reality, though, entering office does not result in an immediate one-time payoff; instead, it results in a stream of benefits that continues as long as the government stays in power.

Importantly, the fact that a government is able to form does not mean that it will last. Conditions can, and do, change. Coalition partners who were satisfied with their share of portfolios and policy compromises when the government first formed might recalculate how large a share they are due if conditions change during the life of the government.Footnote ¹³ The likelihood that exogenous shocks will cause one or more cabinet parties to become dissatisfied with the status quo and bring the government down is greater if smaller parties have only been allocated just enough portfolios to make them indifferent between joining the government and continuing with another round of bargaining, as most government formation models predict. If smaller parties are not pushed down to this reservation value during the bargaining process, though, it will take a larger shock to break the coalition deal, and the government will last longer.

In this context, it is important to consider how formateur parties discount the future. Is it preferable to enjoy a smaller piece of the pie for a longer period of time, or a larger piece for a shorter period of time? Penn and Indriðason present models in which actors prefer to have a smaller piece of the pie but for a longer period of time.Footnote ¹⁴ From the point of view of the formateur party, the desire to enhance the government's survival prospects and thereby enjoy the benefits of office longer provides a rationale for allocating portfolios more generously than standard bargaining models predict. According to Penn,Footnote ¹⁵

The concern about government survival is due to the existence of the vote of no confidence. Governments in parliamentary and semi-presidential democracies can fall at any time if they lose their legislative majority. The possibility that coalition partners might turn against the government if they become dissatisfied with the status quo and participate in (or threaten to participate in) a vote of no confidence drives the higher-than-predicted share of portfolios allocated to smaller cabinet parties.Footnote ¹⁶ Without a vote of no confidence, formateur parties need only worry about government formation and not government survival. In effect, formateur parties would not have an incentive to provide smaller parties with more portfolios than is strictly necessary to get them to support the government proposalFootnote ¹⁷ – which is the scenario captured by bargaining models in the Baron-Ferejohn tradition. This line of reasoning generates the following hypothesis about the formateur party:

Given the nature of the exogenous formateur selection rules that favour larger parties in government formation models, the Formateur Hypothesis can be restated in terms of party size:

Note that these hypotheses predict only that formateur status and party size will be associated with a lower share of portfolios in the presence of a vote of no confidence. They do not predict the extent to which this will be the case, since that ultimately depends on how much parties discount the future.

Research Design

One way to test these hypotheses would be to compare portfolio allocation in presidential democracies (vote of no confidence absent) and parliamentary/semi-presidential democracies (vote of no confidence present). The few analyses of portfolio allocation that have been conducted in presidential democracies suggest that portfolios are generally allocated in a way that benefits formateur parties.Footnote ¹⁸ Although this is consistent with our hypotheses, it is possible that the observed differences in portfolio allocation across these regime types might be attributed to factors other than the presence or absence of the vote of no confidence. Presidential and parliamentary/semi-presidential democracies can differ in numerous ways (culture, party systems, political history, bargaining norms, etc.), many of which are unobservable, or at least difficult to measure in a reliable and systematic manner. An alternative research design would be to hold these types of things constant.

Although all parliamentary and semi-presidential democracies have the vote of no confidence at the national level, this is not necessarily the case at the sub-national level. For instance, France has a vote of no confidence for its national governments, but not its regional ones. France is unusual in this respect. The regional governments in the United Kingdom (Scotland Act 1998, Articles 45, 47), Italy (Constitution, Article 126), Germany,Footnote ¹⁹ Spain (Constitution, Article 152) and all of the other European countries that we are aware of face the vote of no confidence just like their national-level counterparts. France's institutional variation offers us the opportunity to test our theory linking portfolio allocation to the vote of no confidence across different levels of government in the same country. A benefit of this research design is that we can more plausibly claim to hold constant a variety of contextual factors that might vary in important but unobserved ways across both regime types and countries, better isolating the effect of the vote of no confidence.Footnote ²⁰

At the national level, France is a semi-presidential democracy, meaning that the government is responsible to the legislature and there is an independently elected president.Footnote ²¹ Although the president officially appoints the prime minister, it is important to note that, due to the vote of no confidence, the government (the prime minister and other cabinet members) can only be appointed with the support of a legislative majority. The government formation process and the structure of government at the regional level is basically the same as at the national level, though the names associated with the various offices are different. Following elections, the members of the regional assemblies bargain over who should be the ‘president’, the equivalent of the prime minister at the national level, and who should be the ‘vice presidents’, the equivalent of cabinet ministers at the national level. The government types observed at both the national and regional levels typically include minimal winning coalitions, surplus majority coalitions and minority coalitions.Footnote ²²

The key difference between the national and regional governments in France is that regional governments do not face the vote of no confidence. Although regional governments must be formally approved by the regional assembly through an investiture vote before taking office, this vote is final and the regional cabinet – the regional president and vice-presidents – remains in office until the next set of regional elections.Footnote ²³ Thus, in the French context, we can restate our earlier hypotheses in the following way:

In some respects, France represents a difficult case for us. Country-specific models in which we regressed a party's share of portfolios against its share of the government's legislative seats at the national level indicate that France has the highest point estimate (0.94) for the slope coefficient among the fourteen countries examined in Figure 1.Footnote ²⁴ Moreover, France is the only country with a positive and statistically significant formateur bonus at the national level. In other words, larger parties and formateur parties already do considerably better at the national level in France when it comes to portfolio allocation than their counterparts in other West European countries. Our theoretical predictions require that these parties do even better at the regional level.

Empirical Analysis

Few studies address government formation at the regional level in France. Scholars of French sub-national politics have focused on elections, public opinion and voter behaviour.Footnote ²⁵ Thus we created a new dataset containing information on the assemblies and governments that emerged in twenty-one regions following the 2004 and 2010 regional elections.Footnote ²⁶ Data on the number of seats won by the parties in each assembly were collected from the France Politique website and cross-checked against data from the Election Politique website.Footnote ²⁷ The remaining data came from the official websites of each region and regional media accounts of the various government formation processes. In sum, we collected data on forty-two governments and 129 government parties at the regional level in France. We then merged these data with the Warwick and Druckman dataset, which includes information on twenty-two governments and fifty-eight government parties at the national level in France from 1945 to 2000.Footnote ²⁸

We created several measures to test our hypotheses. Our dependent variable, Portfolioshare, is the share of ministerial portfolios controlled by party i. Our independent variables include Seatshare (the share of legislative seats that party i contributes to the total number of seats controlled by the government), Formateur (a dichotomous variable that equals 1 if party i is the formateur party and 0 otherwise)Footnote ²⁹ and Regional (a dichotomous variable that equals 1 when government formation occurs at the regional level and 0 when it occurs at the national level). We also created two interaction terms, Formateur × Regional and Seatshare × Regional, to test the conditional claims in our hypotheses that the effect of formateur status and party size depends on whether the government formed at the national or sub-national level.

We use ordinary least squares to test our hypotheses. Robust standard errors (clustered by government) deal with potential heteroskedasticity and allow us to account for the fact that observations within governments are unlikely to be independent. We also employ fixed effects to deal with various sources of unobserved heterogeneity – while Regional allows us to account for unobserved differences across the two levels of government, Regional 2004 and Regional 2010 let us deal with unobserved differences between the two regional elections themselves. The results from eight slightly different models are shown in Table 1. Models 1–4 relate to the Party Size Hypothesis, and Models 5–8 relate to the Formateur Hypothesis.

Table 1

The Effect of Party Size and Formateur Status on Portfolio Allocation at the National and Regional Levels in France

*p < 0.10; †p < 0.05; ‡p < 0.01 (two-tailed).

Note: Coefficients are shown with robust standard errors clustered by government in parentheses. ‘National’ includes data on all government parties for the twenty-two government formations that took place at the national level in France between 1945 and 2000. ‘Regional’ includes data on all government parties in the twenty-one mainland regions of France following the 2004 and 2010 regional elections. ‘Pooled’ combines the data from ‘National’ and ‘Regional’.

Consider first the Party Size Hypothesis. Models 1 and 2 indicate that the share of seats that a party contributes to a government's legislative majority is positively related to its share of portfolios at both the national (Model 1) and regional (Model 2) levels in France. As predicted, the positive effect of Seatshare on Portfolioshare is larger (1.14) at the regional level than it is at the national level (0.94). The positive and statistically significant coefficient on Seatshare × Regional in Models 3 and 4 confirm that this difference in the effect of party size across the national and regional levels is statistically significant.Footnote ³⁰

It is interesting to note the magnitude of the coefficient on Seatshare at the regional level in Model 2 (1.14). As mentioned earlier, country-specific regressions indicated that larger parties receive less than their ‘fair’ share of cabinet portfolios at the national level in all fourteen countries in our sample. Of course, national-level governments all face the vote of no confidence. If one looks at the regional governments in France, which do not face the vote of no confidence, we see that larger parties receive considerably more than their ‘fair’ share of portfolios; tests confirm that the coefficient on Seatshare in Model 2 is statistically larger than one (p < 0.001).

We now turn to the Formateur Hypothesis. The positive and statistically significant coefficient on Formateur in Models 5 and 6 indicates that, controlling for legislative size, there is a formateur bonus when it comes to portfolio allocation at both the national (Model 5) and regional (Model 6) levels in France. As predicted, though, the formateur bonus is larger at the regional level (0.35) than at the national level (0.20). The positive and statistically significant coefficient on Formateur × Regional in Models 7 and 8 confirms that the difference in the size of the formateur bonus is statistically significant across both levels.Footnote ³¹ In sum, the evidence suggests that, controlling for their size, formateur parties obtain a greater share of portfolios when there is no vote of no confidence.