Disturbing runoff pairing for Colombia

In yesterday’s presidential election in Colombia, the top two candidates were from the extremes of the political spectrum. Leading the pack is Gustavo Petro, a former guerrilla of the M-19 (which demobilized about thirty years ago and has been a political party, or component of various alliances, since). He won 40.3% of the vote. In second place is Rodolfo Hernández, with 28.2%. He is an outsider–having had only municipal political experience in a medium-sized city1–and presents as Colombia’s Trump/Bolsonaro/Bukele. Or worse, as he is on record saying he admires Hitler.

Regular readers of this blog or followers of my published research will know I have always been skeptical of two-round majority election of presidents. And this Colombian runoff pairing is a perfect demonstration of why–sometimes reducing choices to two means a choice between two brands of poison. Consider the third-place candidate: Federico Gutiérrez, who finished just under five points behind Hernández, with 23.9%. He is from an alliance of several more mainstream right-wing parties, including La U, the party originally formed to back former President Alvaro Uribe and which later backed President Juan Manuel Santos (with whom Uribe broke, but that’s another story).

Whatever one might think of any of these candidates and political tendencies, one might posit that a candidate like Gutiérrez could be a more consensus and less risky candidate for the top job. But first he had to stay ahead of Hernández and qualify for the runoff. Moreover, whatever one might think of the notion of a left-wing former guerrilla as president,2 one might posit that 40%, with a 12-point lead over the runner-up in a fragmented field, maybe should suffice. (See the double complement rule–first proposed by Shugart and Taagepera, 1994–under which this would be have been sufficient.) Instead, Colombia gets a polarizing runoff in which it is plausible that a genuine extremist outsider might rally most of the rest of the 60% on a “stop the left” plank and become Colombia’s president.

And then what? Quite apart from inexperience and ill temperament, Hernández will confront a congress in which those who backed his first round campaign have little presence. I do not know where within Colombian politics his voter support came from, but the alliance he led in the presidential election’s first round–League of Anti-Corruption Governors–did not even present a list for the Senate election. In the Chamber of Representatives, it won 2 of the 168 seats. This is one version of the Linzian nightmare scenario!

Colombia is, as I’ve written about before, one of the few presidential (or semi-presidential) systems to use an exclusively counter-honeymoon electoral cycle. The Chamber and Senate were elected 13 March, for a four-year term. Petro’s alliance–Historic Pact for Colombia–led the vote in the nationwide Senate election, but with only 14%. Its 16 seats (of 100 in the main electoral district, or 106 all told3) tie it with the old established Conservative Party. The other old establishment party, the Liberals, got 15 seats, a Green-Center alliance got 14, Democratic Center (Uribe’s other party, after the break with Santos) also 14, and the misnamed Radical Change (actually a split years ago from the Liberals) got 11, followed by 10 for the older Uribe party. What a fragmented mess! That would be hard to govern with no matter who would have been elected president. But at least either Petro or Gutiérrez would have had a base to build on. In the Chamber, the situation is broadly the same, although differing in important detail. There, the Liberals actually won the most seats, albeit only 32 (19%), with Petro’s alliance on 27 and Conservatives on 25.

In the past–including Colombia 2018 and 2010–I have suggested that counter-honeymoon elections can function as a de facto presidential primary, clarifying and narrowing the realistic choices for the upcoming presidential election. There seems to have been little of that this time, with the second runoff contender and realistic runoff winner having made no showing at all in the legislative races. Another feature–and not a desirable one–of counter-honeymoon elections is that they tend to be associated with greater fragmentation, relative to expectation from the electoral system’s seat product–than election held at other points in the period between presidential elections (Shugart and Taagepera, 2017, Ch. 12, in particular figures 12.1 and 12.3). That is certainly the case in Colombia, and specifically in this election.

The seat product for the Chamber is approximately 800 (162 seats elected in 33 districts means a mean district magnitude of 4.9, so the seat product is 162*4.9=795, but there are also various set-aside seats I am leaving out). For the Senate, it is approximately 10,000 (100*100, again leaving out the set-aside seats). That Chamber seat product would lead us to expect a largest party with around 43% of seats; the Senate’s around 32%.4 Obviously neither house is close to that. The electoral cycle is part of the reason (likely exacerbated by some parties and alliances holding actual presidential primaries concurrent with assembly elections), but certainly not the only or even most important reason. Colombia’s party system has not been “strong” by any definition since the old Liberal–Conservative duopoly began breaking down in the early 1990s.

This upcoming runoff–and the presidential–assembly relationship to follow–is deeply troubling. It seems to signify the death of the old moderate swings in Colombian policy that have typified the system up to now. With the social unrest of the past year, discrediting of the established elite was perhaps inevitable. But the institutions of Colombian democracy are about to endure a very serious stress test.

  1. Bucaramanga, the largest city (over half a million) in Santander department.
  2. I should note that Petro is also a former mayor of the capital, Bogotá, so not a total outsider.
  3. In addition, Colombia sets aside 2 senate seats for indigenous candidates who run separately from the main district, 5 for the party of the former FARC guerrilla movement (which got only 0.19% of the vote) and 1 for the eventual presidential runner up).
  4. The houses are co-equal; we lack a model of how diverging seat products in two strong chambers should affect the overall party system. Regardless, with largest parties–and different ones, at that–in each chamber under 20%, it is not the seat product’s fault!

Senegal 2017–It is all about 2019

Senegal holds assembly elections today, 30 July. Prime Minister Mahammed Boun Abdullah Dionne says: “We aren’t talking any longer about July 30, but of 2019.”

Of course. This is, after all, a counter-honeymoon election. It may not seem like it could be, because 2019 is pretty far off. But the presidential term is seven years, and the incumbent, Macky Sall, was elected in 2012.

In Votes from Seats (Chapter 12), Shugart and Taagepera (2017, in press) suggest that a counter-honeymoon election is any held three-fourths of the way through the president’s term. The months since the president was elected (first round in February, 2012) to now work out to 0.77. While there is no hard definition of when the counter-honeymoon starts–Elapsed Time between presidential elections is a continuous scale–three fourths is a reasonable approximation, and Senegal is just past that.

The main points Taagepera and I make about counter-honeymoon elections are: (1) The president’s party tends to suffer a substantial reduction in its vote share, and (2) They occur at a time when the parties are jockeying for position ahead of the next presidential election. It is the latter point that the premier is signaling is important today.

In the book, we have a formula for estimating the likely change in the president’s party’s support in a non-concurrent election (see my 23 April post from between rounds of the French presidential election). It is an empirical estimate, not a deductive model, but the pattern it captures has strong logic behind it, and it fits the data very well. It states:


where Rp is the vote share of the president’s party in the assembly election, divided by the president’s own vote share (in the first round, if two-round system), and E is the Elapsed Time (months between presidential and assembly election, divided by presidential inter-electoral period, also in months*).

In the case of today’s Senegal election, the equation returns a predicted value of Rp=0.64. Sall’s first-round vote percentage in 2012 was only 26.6%. It would be pretty bad news for him, and for governability in the final 23% of the time to the next presidential elections (19 months), if our equation’s application to today’s election turned out accurate. It would mean his party would get only 17% of the vote. Of course, the actual share could be higher. Or lower.

Attempting  to predict the votes in this case is also made more difficult by the fact that in the last assembly election, Sall’s party, the Alliance for the Republic (ARP) did not run on its own. It formed a coalition in advance of the election. That is not surprising–the electoral system is (unlike that of France) a single round, and it is quite majoritarian (like that of France). Thus there would be incentives for parties who had backed the newly elected president to join forces, and indeed they did. The last assembly election had a strong honeymoon timing, with E=0.06.

My quick search did not turn up any disaggregated votes by parties within the United in Hope coalition that was formed to support Sall in July, 2012. And I have no idea whether some of the parties that might have joined it have already jumped ship (the second point above about counter-honeymoon elections would lead me to expect a yes to that).

In the 2012 assembly election, United in Hope earned 53.06% of the vote, with runner-up Senegalese Democratic Party (PDS) way back, at 15.23%. The PDS was the party of the previous president, Abdoulaye Wade, whom Sall defeated. (Sall himself had won 65.8% in the runoff.) If that was a “normal” honeymoon, we would have expected the president’s party to have won 30.7%,  implying an additional 20% or so coming from other, post-first-round coalition partners.

The Wikipedia article on today’s election says that the coalition backing Sall includes the Socialist Party and the Alliance of the Forces of Progress. These parties’ candidates in the 2012 first round earned 11.3% and 13.2% of the votes, respectively. If we add those up we get 50.7%, which would actually imply little growth for the parties in the honeymoon election. (This is not surprising to me; the supporters of the defeated candidate would be harder to mobilize for a nonconcurent assembly election than for the presidential runoff.) However, perhaps we could use this revised figure in our Rp calculation. If we do, we get around 32% for their possible combined share, instead of 17% using only Sall’s own party. Anything much above 32% would defy the Elapsed Time expectations.

The Wikipedia article also says that the opposition failed to coordinate. The electoral system is mixed-member majoritarian. Around 40% of the seats are by PR (nationwide), with the rest in a mix of single-seat and multi-seat constituencies, using plurality (according to IPU). Thus even 32% against a divided opposition might leave the pro-Sall coalition near a majority of seats. But if his alliance is nearer 17%, governing could be difficult for the last quarter of his term.

The Senegalese system is semi-presidential, of the premier-presidential subtype. Thus the cabinet can fall if it lacks the confidence of the assembly majority, but the president has significant powers (he does not require an investiture vote to install the cabinet and he has a legislative veto). Cohabitation is a possibility, but so is a reshuffled pro-Sall cabinet.


* In the book, we actually calculate E based off the precise date, but here I have just used whole months.