Times must be hectic at the European Council. On top of usual business, it currently has to deal with the international crisis in Ukraine, the new terrorist threat posed by ISIS, and the future of the Economic and Monetary Union (with the ‘Grexit’ time-bomb in the background). Given the importance of the matters at stake, it is crucial that its decision-making process favours the best possible decisions.
For ‘sensitive’ matters such as those mentioned above, the general principle is to ‘vote unanimously’: individual member states have the power to block any decision.1 Yet how abstentions are treated depends on the topic. In most cases, the procedure corresponds to ‘consensus decision making’. Abstentions are then treated as silent consent. If no member votes against the proposal (i.e. vetoes it), it is adopted. The Common Foreign and Security Policy (CFSP) is the exception. When presiding over the CFSP, the procedure allows for so-called ‘constructive abstention’. If more than a third of the Member States abstain, then the proposal is rejected.2
This subtle difference is not as innocuous as it may seem. Previous research has indeed shown that unanimous voting suffers from bad information aggregation properties. In this piece, we argue that this is not the case when constructive abstention is allowed. This suggests a simple welfare improving reform of the European Council decision-making process. Similar welfare improving reforms could also be undertaken in other institutions using unanimity without constructive abstention, e.g., NATO, the Mercosur, and American Juries.
Designing an optimal voting rule when agents must have veto power
In many situations, veto power is an unalterable feature of the voting rule. One reason is related to sovereignty issues and can be interpreted as a participation constraint: countries would not join the organisation if there was not a mechanism allowing them to fully protect their interests. This arguably applies to the EU, and many other international organisations.3,4
In general, one could think that comparing voting rules which award veto power to voters is of limited interest since they are all prone to gridlock. But this misses the point that members may have common objectives (e.g. promoting the peaceful resolution of the international crisis in Ukraine or finding the best solution to the economic crisis in Greece). Even when objectives diverge, some decisions could make every member better-off, provided that losers are properly compensated. Economics textbooks indeed abound with examples of potential Pareto improvements (in which an alternative allocation or proposal harms no one and benefits at least one person).
What is the problem?
When agents have no uncertainty about their preferred alternative, rules awarding veto power are equivalent – the proposal is accepted only if it is Pareto improving. Yet, agents may not know which decision is best.
Take the example of the European Council in the context of the Ukrainian conflict. Arguably, member states share the common objectives of a peaceful resolution, the protection of human rights, etc. (of course, they may also have private interests at stake, e.g. economic). However, the way to achieve these common objectives is not obvious. To determine the merits of a proposal, the expertise and information possessed by all Member States might be useful. But potential conflicts of interests often make it hard to convincingly reveal such information. It may actually even be illegal, for instance when the information is classified. Therefore, the voting process ought to elicit the information dispersed among members in order to reach the ’best’ decision.
The problem is that consensus (and unanimity without constructive abstention) feature poor information aggregation properties (Feddersen and Pesendorfer 1998). The reason is that voters cannot ’communicate’ negative information about the proposal through the voting system without blocking that proposal altogether. Perhaps surprisingly, strategic behaviour in this context results in detrimental proposals being accepted too easily (in their jury example, the innocent is relatively more likely to be convicted under unanimity without constructive abstention than under majority rules). This presents a scary prediction in the Grexit context given that there could be no way back.
One of the conclusions of Feddersen and Pesendorfer (1998) is that majority rules allow for better information aggregation. Yet these rules do not satisfy the constraint of awarding veto power to each member. The challenge is to design a rule that satisfies this constraint and enables information aggregation.
Allowing for constructive abstention
In recent research (Bouton, Llorente-Saguer and Malherbe 2015a), we argue that unanimity with constructive abstention essentially resolves the tension between information aggregation and allowing for veto power. This rule has two crucial features. First, agents have to be allowed to abstain. This means that three actions are possible: ’yes‘, ’abstention‘, and ’veto‘. Second, the proposal is adopted if nobody vetoes it and the number of abstentions is no larger than a given quorum. In other words, abstentions are counted as noes rather than yeses. The only constraint on the quorum is that it is not extreme, i.e., that is not equal to zero or the total size of the group. This is exactly what formally distinguishes the rule from consensus and unanimity without constructive abstention, respectively.
We show that unanimity with constructive abstention combines the best of both worlds – while allowing for veto power, it still features the desirable information aggregation properties of majority rule. The reason unanimity with constructive abstention better aggregates information is simple: agents can ’communicate‘ their information through the voting system in a finer way than without constructive abstention (and under consensus). In particular, it allows them to communicate information against the proposal without vetoing it.5
Potential reforms around the world
In our view, in addition to its strong theoretical properties, the simplicity of unanimity with constructive abstention makes it particularly appealing to real-world applications. Yet before advocating in favour of voting rule reforms, it is important to assess the empirical performances of unanimity with and without constructive abstention. In order to perform a clean comparison, we generated new data through a series of laboratory experiments (see Bouton et al., 2015b). Overall, we found strong support for the dominance of unanimity with constructive abstention. In our view, the path for voting-rule reform is thus well defined.
Bouton, L, A Llorente-Saguer and F Malherbe (2015a), “Get rid of unanimity: The superiority of majority with veto power”, CEPR Discussion Paper 10408, February.
Bouton, L, A Llorente-Saguer, and F Malherbe (2015b), “Veto, abstention and frames: An experimental study on the implementation of majority with veto power”, mimeo.
Feddersen, T and W Pesendorfer (1998), “Convicting the innocent: the inferiority of unanimous jury verdicts”, American Political Science Review, 92: 23-35.
Maggi, G and M Morelli (2006), “Self-enforcing voting in international organizations”, The American Economic Review, 96(4): 1137-1158.
Romme, A G L (2004), “Unanimity rule and organizational decision making: A simulation model”, Organization Science, 15(6): 704-718.
Wessel, R and R Bottner (2013), “The procedures leading to the adoption of decisions in the CFSP area - Art. 31 TEU” in HJ Blanke and S Mangiameli (eds.), The treaty on European Union (TEU): A commentary, Berlin/Heidelberg: Springer.
1 For example, see http://www.consilium.europa.eu/en/council-eu/voting-system/unanimity/
2 In 1992, the Treaty of Maastricht established the unanimity rule (without constructive abstention) as the default procedure of the European Council. In 1997, the Treaty of Amsterdam introduced the concept of constructive abstention, which allows abstentions in the European Council (when presiding over the CFSP) without blocking a unanimous decision unless more than a third of the member states abstain, or member states representing a third of the total EU population. Note that the approval quorum is binding only when member states accompany the abstention by a formal declaration. In that case, the member states who abstain are not obliged to apply the decision but must accept that it commits the Union, and must then refrain from any action that might conflict with Union action based on that decision. Wessel and Bottner (2013) argue that the latter provision essentially voids the option not to apply the decision.
3 See Romme (2004) for other applications and examples.
4 See Maggi and Morelli (2006) for a discussion of the pros and cons of simple unanimity focused on enforceability issues.
5 Interestingly, it also allows for better information aggregation than a majority rule, because it creates the possibility for agents to convey extremely negative information as well.