On April 16, 2017, a slim majority of 51.4% of voting Turks supported a constitutional amendment aimed at empowering the president vis-à-vis other branches of government (CNN Türk 2017).Footnote 1 In effect, this implies a transition from the current parliamentary to a presidential system of governance.Footnote 2 Some 1.4 million ballots turned out to be decisive not only in directing the path of internal affairs in Turkey, but arguably the geopolitical situation of the entire region. (Note that Turkey is a member of NATO, it controls the inflow of refugees to Europe, and it is pivotal for peace in the Middle East). One could, of course, ask whether it is fair that some 25.2 million Turks ‘imposed’ their view on the huge minority of approx. 23.8 million of people, who opposed the system overhaul.

Bruno Frey proposes two institutional solutions, which may potentially mitigate the unfairness of political outcomes in narrow races. First, he offers that “the closer the outcome of a popular referendum is, the stronger should the representatives of the opposing parties engage in a constitutionally sanctioned procedure to reach a consensus which is then again put to the vote of the electorate” (proposal 6.7). This proposal arguably requires a certain political culture; first, to endogenize this institutional procedure in a constitution and, second, to de facto enforce it at the post-constitutional level (for endogenous direct democracy institutions see, for instance, Blume et al. 2014). This kind of reconciliation mechanism is not unknown in political process. In fact, one could draw an analogy between Frey’s institutional proposal and the navette system employed in several jurisdictions to reach a consensual agreement between two chambers in bicameral parliaments or the consensual decision-making process in the Council of the European Union.Footnote 3 For institutional variety of the navette mechanism see Tsebelis and Money (1995), Tsebelis and Rasch (1995). The navette system consists of sending the modified version of a bill from one of the chambers to the other. If, after several rounds, an agreement is not reached, a joint committee may be called upon to seek a compromise, as is the system in Germany.Footnote 4 Thanks to this mechanism, even relatively weak upper chambers of parliament are able to obtain significant concessions from powerful lower chambers, biasing policies toward the status quo. If applied in post-referendum circumstances, this consensual/reconciliation mechanism of reaching an agreement would arguably make political and institutional change incremental. This would prevent any abrupt changes and sharp discontinuities in policy-making, hence precluding situations such as the one occurring in Turkey.

The second solution offered by Bruno Frey is an introduction of a random process of determining policies and institutional choices, wherein random refers to mathematical probability. In fact, randomization was and still is perceived as a policy tool guaranteeing fairness when it is applied to define policy choices or select policy beneficiaries. See, for instance, a disciplinary practice to punish cowardly behavior in the Roman Army,Footnote 5 or more contemporary military drafts, jury selection methods, and allocation of judges to cases [for an overview see Dodge (2012), chapter 20]. Yet, at a higher political level, where elitist politics prevail, randomization as a decision-making procedure is largely missing.

Bruno Frey’s proposal states that “the outcome of a popular referendum can be determined by lot with the weights given by the vote shares” (proposal 6.8). In the case of Turkey, instead of a definitive ‘yes’ or ‘no’ to the proposed constitutional change, the referendum would therefore determine the ex-ante probability of the system change. Since the transition of the governance system was supported by 51.4%, this weight would be given to this policy option, while 48.6% would be assigned to retaining the existing system. To illustrate, a policy choice would be ultimately determined by a random selection from an urn with 514 green and 486 red balls, which indicate either a change to the presidential system or a lack thereof, respectively. This procedure has the obvious advantage of taking into consideration the size of the majority (and minority) in determining the likelihood of policy change.

It should also be stressed that the application of a random procedure arguably will not be neutral for the electoral turnout. One should note that in the proposed system all votes matter as they translate into higher ex-ante probabilities of given options. This, by itself, may incentivize voters to cast ballots and positively affect the turnout. As a result, the random procedure may minimize a pool of voters who do not cast ballots for the reason that their ‘votes do not count’ (their votes are not decisive).

If we further assume that abstention from voting signifies indifference toward a suggested policy change (i.e. policy choices are too alike to justify the cost of voting), one could consider extending Frey’s proposal to readily take into account the size of abstention (indifference).Footnote 6 Indifference was found to motivate significant amount of voter abstention, for instance, in the US presidential elections (Zipp 1985; Adams et al. 2006). A refined proposal is illustrated using the example of the Turkish referendum. The turnout in the constitutional referendum was roughly 85% (CNN Türk 2017). Hence, 15% of the eligible population appeared to be indifferent regarding the constitutional amendment. Under the new proposal, apart from the green and red balls indicating ‘yes’ and ‘no’, respectively, there could be a third type of ball, say white, signifying indifference. The urn’s contents would therefore be comprised of 15% white balls (capturing the indifferent population), 43.7% (51.4% of 85%) green balls (indicating a ‘yes’ vote), and 41.3% (48.6% of 85%) red balls (signifying a ‘no’ vote). If, in the first stage, a green or red ball is drawn, that would simply translate into the introduction of policy or maintenance of the status quo, respectively. However, if a white ball is drawn in the first stage, the procedure would go to the next stage where the urn contains an equal proportion of green and red balls, indicating equal probability of policy change and status quo and, hence, indifference between these two policy choices. This procedure is depicted in Fig. 1. Overall, this refined procedure leads to the situation where an ex-ante probability of introducing the presidential system in Turkey would equal 51.2% instead of 51.4%. This slight drop reflects that 15% of population is indifferent between changing the system and preserving the status quo.

Fig. 1
figure 1

Source: own figure

Two stage randomization procedure

Full size image

Accounting for the turnout or, in other words, abstention, would make disparities between vote shares less extreme. An acute example of this may be the 2015 referendum on single-member districts and public funding of political parties held in Poland.Footnote 7 In this referendum, 78.8% of voting Poles supported the introduction of a winner-takes-all system, and 17.4% were in favor of public funding for political parties (PKW 2015). At first glance, these results may be appealing, yet strikingly the turnout in this referendum equaled merely 7.8%, which indicates that 92.2% of Poles were somewhat indifferent on these policy options. If this referendum had any binding and decisive nature (one should bear in mind that this referendum had merely a consultative nature), an ex-ante probability of adopting a winner-takes-all system would be approx. 52.2%, instead of 78.8%, while the probability of maintaining public funding of political parties would be 47.5%, instead of 17.4%.

One could further consider combining Frey’s proposal 6.8 of lots with weights distributed by the vote share, with proposal 6.6 stating that “old persons can have a higher vote weight than young persons with regard to constitutional decisions on the future rules of the game”. A system could be therefore designed where votes of given communities (or municipalities) are adjusted for weights derived from their demographic structure, with a larger share of the population above 65 years or a larger share of retired population translated into larger weights. Applying this refined proposal to the constitutional referendum in Turkey would arguably lead to somewhat higher support for the presidential system of governance, as the rural and small-town regions, which were predominantly in favor of the constitutional amendment, are at the same time those regions where demographic indicators are skewed toward the older population (BBC 2017).

Yet another extension of Bruno Frey’s proposal could be the application of weighted lots to determine policy choices by parliaments. A random procedure with ex-ante probabilities as determined by vote shares of members of parliament would readily take into account the size of the opposition and, hence, should be considered as a fair policy process [it needs to be stressed that this could also be achieved by the implementation of the randomized decision rule as suggested by Holler (1985), and Berg and Holler (1986)]. The following example illustrates the context in which this procedure could be particularly valuable. Figure 2 presents a distribution of the majority size for 100 legal acts enacted by the Polish Sejm (the lower chamber of the parliament) between December 1, 2016 and April 7, 2017. The distribution is clearly bimodal with some 60% of acts enacted by a majority of 95% and roughly 30% of acts being voted in by a slim majority of 50–60%. The latter group arguably comprises those legal acts which are the most salient and involve ideologically complex issues.

Fig. 2
figure 2

Source: own figure based on information extracted from www.sejm.gov.pl. Accessed on April 18, 2017

Size of majority for legal acts enacted by the Sejm between December 1, 2016 and April 7, 2017

Full size image

Under the current parliamentary situation, the governing ‘Law and Justice’ party can in practice enact any law without considering the view from the opposition, as it holds more than 50% (235 out of 460) of the seats in the Sejm. A random procedure as defined in Frey’s proposal 6.8 could be appealing and, by some, considered fair. This is particularly appealing if one bears in mind the fact that ‘Law and Justice’ obtained 37.6% of the popular support, but due to the specificity of the electoral system (d’Hondt system and 5% electoral threshold), this support was translated into more than 50% of seats in the Sejm.Footnote 8 Hence, laws enacted by ‘Law and Justice’ do not necessarily reflect what the majority of the population wants. The application of a random procedure would result in the promulgation of almost each legal act supported by all parties (acts would enjoy close to unanimous approval), but only slightly more than 50% of acts which are supported exclusively by ‘Law and Justice’. This would arguably lead to a less controversial policy-making and less demonstrations against the current Polish government. Moreover, the application of a random procedure to the parliament’s decision-making would also be less costly than the Swiss type of direct democracy institutions.

This short comment aimed at extending Bruno Frey’s ideas on a democracy of the future as well as stimulating further discussion on this important topic. It is acknowledged that by and large the arguments presented by Bruno Frey and extensions offered here remain theoretical. One cannot overlook, however, the possibility that in the near future some local governments, particularly in federal countries, or even central governments will start experimenting with some of these ideas.