Reader survey

Are you ever curious about who reads F&V, how often, and what the readers like? I am. At the suggestion of JD Mussel and Henry Schlechta (whose names many readers will recognize due to their frequent posts here), there is now a Readership Survey on line. It is anonymous and will not take much time to complete. So please consider completing it.

(This post is a “sticky” and will remain up front for a while; please scroll for new content.)

Advertisements

How not to federalise

Since his election in 2015, Filipino President Rodrigo Duterte has done little that supporters of liberal democracy should admire. His program of extrajudicial killings and contemptuous attitude to the judicial institutions of the Philippines demonstrates that he is in many ways a dangerous figure for the country. However, it’s true that a stopped clock is right twice a day, and one of Duterte’s more positive ideas for the country has been a consistently expressed willingness to change the country’s political system, from the existing system of unitary presidentialism to a semi-presidential federal system.

In keeping with this principle, a group of Senators have presented a specific proposal for a dramatically revised new federal constitution. Having only given it a cursory examination, I responded positively-however, a more careful examination demonstrates several substantial issues with the draft.

The proposal would divide the Philippines into eleven states, with Manila City becoming a federally administered region. Only half of Duterte’s proposal would be implemented, though, as presidentialism is kept-the most substantial change to the structure of the now federal government would be a change to the Senate, which would go from being a twenty-four member body elected using MNTV in one nationwide district for six-year terms (with staggered elections to half of the chamber every three years) to a seventy-five member body elected with states acting as districts, nine senators per state, and three senators elected per state every two years.

In general, the Philippines has not been a positive example of presidentialism. Presidents have regularly been elected in fragmented races (Fidel Ramos was elected in 1992 with 24% of the vote), and the party system for the legislature is if anything more weak and fragmented, with fluid allegiances (best demonstrated by the numbers of President Duterte’s PDP-LABAN party going from two at the election to a comfortable majority afterwards).

The federal aspect of the constitution also leaves something to be desired. Governors are elected directly, with first-past-the-post for four-year terms. Initially, and somewhat unusually, state legislators are elected by local councils. Three are elected for each province, a truly astonishing degree of malapportionment. For example, in Central Visayas State, the province of Siquijor (pop. 96,000) will have the same three seats as Cebu (pop. 4.6 million). Three further members are elected: one for farming, one for fisheries, and one for senior citizens.

Indirect election by local councils becomes variable by state legislatures after the first election. It is left unclear as to whether states are able to amend the composition of their legislatures, or whether local councils are still able to recall members of state legislatures.

Weirdly enough, states can create ‘autonomous regions’ within their own territories, the powers of which are only vaguely defined. States also have exclusive power, strangely enough, over “trade…tourism…weights and measures” as well as “pilgrimages to places outside the Republic”, which, as JD pointed out to me, could allow corrupt state officials to be spirited away from federal police on fairly spurious grounds.

None of this is to say, of course, that the principles of federalism don’t make sense for the Philippines. As a large country with clear political diversity, it makes sense to devolve power from a potentially unrepresentative core. Nonetheless, the proposal put forward by the Senators risks creating equally unrepresentative state-level governments with a somewhat esoteric mix of powers. More work is needed, and presumably the plan will be looked at more carefully through the later process of the reform process.

Parties and personal-vote earning attributes in OLPR

In the forthcoming Votes from Seats, Rein Taagepera and I build on the earlier argument of Bergman, Shugart, and Watt (Electoral Studies, 2013) about incentives of political parties to “manage” the competition among their candidates under various intra-party allocation rules. The short version of the story is that parties under open-list PR should be willing to tolerate “laissez faire” competition, because no excess in the number of candidates nor imbalance in the candidates’ votes can affect the party’s ability to convert its collective vote total into a proportional share of the seats (within the limits of the district magnitude and inter-party allocation formula).

The claim about laissez faire competition under OLPR rests on the assumption that parties are only interested in seat-maximization, and not in the precise set of candidates who win. It also rests on the assumption that “party” and “list” are the same thing.

The second assumption is already relaxed in Votes from Seats, where we devote almost an entire chapter to the topic of how alliance lists work, focusing on the cases of Brazil, Chile, and Finland. In these systems (and some others) many lists contain candidates of two or more parties. In that case, the parties on the list are in direct competition with one another for a share of the seats won by the list as a whole. Thus parties would need to manage their vote–i.e., concern themselves with the distribution of votes across their candidates.

The first assumption–regarding parties’ indifference about their personnel–is not something we actually believe is true in practice. Science involves making simplifications, and we show in the book that using this simplifying assumption is quite powerful in predicting, via deductive logic, the average patterns in the preference vote shares of candidates (i.e., candidate votes divided by total list votes in a district). So, for the purposes of the book (and the earlier article), the strict assumption of indifference worked to get us a step farther down the road to understanding how electoral systems shape candidate vote shares.

In earlier drafts of the book, we worked on attempts to analyze how parties might affect the election of specific candidates, even though they lack ranking control, through nominations. We took these sections out because we were unable to come up with a deductive model of the process–a key methodological criterion around which the book is based. In the remainder of this entry, I will post and discuss two graphs that we took out of the book but that demonstrate the (still underdeveloped) idea of parties’ engaging in forms of intraparty management–even under OLPR.

The immediate reason for returning to think about this now was the recent American Political Science Association annual meeting, at which I presented a paper with Åsa von Schoultz that incorporates both the logical models of preference-vote distribution and the personal vote-earning attributes (PVEAs) of the candidates themselves. On the same panel was a fascinating paper by José Antonio Cheibub and Gisela Sin, which (among other things) analyzed the discontinuity in ratios of one candidate’s votes to the next candidate’s when they are sorted in descending order by preference votes. They find that, in Brazil, there is a tendency for these ratios to be greater at “last winner to first loser” and at “first loser to second loser” than among winners higher up or losers lower down.

A pattern like that found by Cheibub and Sin would not be found if there were not some “coordination” going on. Such coordination could be done by voters or by interest groups or others with a desire to see certain candidates elected over others. Or it could be done by parties. If by parties, it would be a form of intra-party management. For instance, parties could achieve a desired concentration of votes on the eventually elected candidates by ensuring a mix of candidates appealing to different groups of voters, or through allocating campaign resources, or some mix of these and other tactics.

One way to manage the vote would be through exploiting the party’s knowledge of the relative appeals of specific candidates or types of candidates. If the party had perfect information, it could renominate just the right number of incumbents and nominate the right number of local council members, or other politicians with popular appeal and whom the party sees as promising future legislative personnel. In other words, through nominating candidates with given PVEAs it could structure the balance of different traits and constituencies represented within its delegation.

The following data plots from Brazil and Finland point towards how such PVEA management might work.

 

The plots show the share of candidates at any given relative rank (list position/seats won) who have a given PVEA: incumbent assembly (national) member or local council. The local regression (lowess) curves plot the pattern, and in the case of Brazil, I also plot a lowess for the state assembly members running on the deputies list (but not the data points, because of the clutter). The incumbent MP curves for the two countries are nearly identical, with relatively few MPs losing and more near the top preference vote totals.

The local candidates’ curves also have a similar shape—rising near the bottom of the electable ranks on the list and then still rising among the top losers, before plummeting. The obvious difference is that there are a lot more locals in Finland than in Brazil. The curve for Brazilian state legislators running for federal deputy looks like a much-flattened version of the incumbent deputies’ curve.

These plots may be showing that parties are indeed managing the distribution of votes across candidates. They are doing so by whom they nominate. They probably have pretty good information about the vote-earning potential of various candidates, and they can “clear a path” for the candidates they consider sufficiently valuable by not putting too many similarly strong candidates on the list against them. Obviously, what I have shown here does not prove that point, but it is suggestive of how parties might “coordinate” on the intraparty dimension, through managing the types of candidates they select.

A possible objection is that parties could not possibly know the votes that a candidate could bring to the list. After all, these lists–especially in high magnitude districts–are so competitive! Another graph suggests it might not be so hard for parties after all, at least when nominating candidates who have run before for some office.

 

Do parties have good information about the votes a candidate will obtain? Evidently so. This graph compares 2002 votes of Brazilian candidates to 1998 votes, whether their 1998 campaign was as a deputy candidate or a state legislative candidate. The diagonal is the equality line; a regression is not much different from it. In other words, a candidate’s votes in the prior election are a pretty strong predictor of the candidate’s votes in the current election (at least in Brazil, 1998-2002). This is generally true for those who won their contest the previous election and those who had small vote totals. And it applies to state legislative candidates even though they are running in different-magnitude districts. (The legislators of a given state are elected in a single statewide district, just as the national deputies are elected in state districts, but the magnitudes are greater for the state legislature.)

Parties amaze sometimes at how good they are with this stuff!

Certainly, when I see things like this I realize that all the old ideas about chaotic competition in OLPR or parties lacking control just do not stand up.

So, yes, parties can tolerate laissez faire competition among the candidates on their list–provided they are interested only in maximizing the list’s seat total. And assuming that this is all they care about allows us to understand average patterns of vote distribution. A key goal is to introduce other variables–notably PVEAs–to understand how individual candidates deviate from these logically predicted (and empirically confirmed) averages. That was the point of the von Schoultz-Shugart APSA paper, focused on Finland. The next step is to try to understand PVEAs and prior vote totals as a window on how parties manage the vote, even under OLPR.

Yes, all is well

It has been a long time since a post, so I thought I should confirm that I am indeed still here. And all is well.

I blame the inactivity on the fact that APSA is coming up. Plus a trip to see the total eclipse!

Regarding the eclipse, I don’t have photos of the totality itself–that really takes skills I don’t have–but I may post some of the period just before and after, which is itself interesting. The whole experience was amazing in one of those “no words” kind of ways. We were able to view the total eclipse from Prineville, Oregon. And we also were able to avoid crowds! (Some heavy traffic on the way back, but not too bad, and none getting into the area.)

I hope to be back with things to say in September. In the meantime, enjoy the rest of August!

If there’s anything current going on in the Fruits & Votes field, feel free to use this as an open thread.

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:

Rp=1.20–0.725E,

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.

The Polish president’s veto

Polish President Andrzej Duda has exercised his veto against two of the government’s bills to change the appointment of judges. Media accounts are treating this action as a “surprise”, but it really should not be seen as such. Both the government (premier and cabinet) and the president (directly elected) are of the same party, Law and Justice (PiS), yet the institutional dynamics make the action unsurprising.

First of all, such a possibility was signaled by the president on 18 July. Secondly, the theory of presidentialized parties (Samuels and Shugart, 2010) should make presidential action against their own governing party’s controversial decisions the default expectation. That is, the institutions “predict” such actions, and it is politics that sometimes intervenes and prevents the president from acting on his or her independent mandate. In this case, the institutions prevailed.

Duda is not the leader of the party in the sense of who holds the formal organizational title; that would be Jarosław Kaczyński, a former premier. But that fact only makes today’s veto by the Polish president an even better case for demonstrating the power of the argument. The fundamental point of my book with Samuels is that parties face a “moral hazard” problem under presidentailism, due to separate survival (fixed terms). Once a presidential candidate has won the election (separate origin), he is no longer an agent of the party. He now has (potential) incentive to respond to his wider constituency and head-of-state obligations, even when those might conflict with the party that nominated him. He is secure in his fixed term against efforts by the party to restrain him, unlike a prime minister (who, by definition, can be dismissed by the assembly majority or by the party).

One manifestation of presidentialization of parties can be the reversal of the principal-agent relationship, such that the president redirects his own party, making it his agent, rather than the other way around. But–we point out in the book–this is only one possibility. Another is that the president and the party face conflict–an intraparty separation of powers.

The latter seems to be the case here. I heard a BBC interview earlier this morning with an official of the PiS party. He slammed the president for going against the party’s manifesto, and said that the government (which is a single party, with a majority) should be allowed to implement its promises, including judicial reform. The statement overlooks a key institutional fact: Poland is not parliamentary, but semi-presidential. The president is freed from the manifesto by virtue of his separation of origin and survival–that is, his independent election and fixed term.

Duda received significantly more votes in winning the presidency (in a two-round election) than the PiS earned in winning its parliamentary majority. The governing party’s majority is manufactured by the electoral system (details in my earlier entry), whereas Duda’s was earned in the two-candidate runoff. In addition to the electoral system, the PiS surely owes its being in the position to gain that assembly majority from Duda–or, specifically to its own election in the “honeymoon” of Duda (again, see the earlier entry for details).

I do not know if the issue of judicial reform was raised in his own election campaign, but the wider point is that it hardly matters. He has the institutional capacity to act independently, and he did. He has the constitutional power of veto, and he exercised it.

For better or worse, this is how presidentialism (including semi-) is supposed to work. As a head of state (who must deal with other governments unhappy with Polish plans) and with a wider popular constituency (with many Poles in the streets over this issue), and protected by the fixed term, the president can act against the will of the government and its own parliamentary majority.

The veto takes a three-fifths vote of parliament to override, and the PiS is far short of this threshold. Unless it can bring other parties around to this bill, the president’s veto will stand. What this might mean for Duda and his relationship with his party is not for me to predict. But, based on the theory of presidentialized parties, what has happened today is far from surprising; it is predictable.

Poland’s protests–and institutions

Protests in Poland have been making international news. The object of the protests is the government’s plans to undermine the independence of the judiciary. The case illustrates several important points about how political institutions affect policy (and can have potential international repercussions). The escalation of the crisis over these plans in recent days makes the decision some months ago to lead with Poland’s 2015 elections and their aftermath in the forthcoming Votes from Seats look like a good one.

We (Rein Taagepera and I) use the case to demonstrate how seemingly small institutional details can tip outcomes in favor of a party with a strong agenda that might undermine democracy itself. (Yes, the very first example mentioned in the book is the US presidential election of 2016.)

The background is that the Polish ruling party, Law and Justice (PiS) has a majority of seats in the first chamber of the assembly (the Sejm). PiS also holds the presidency. It has a manufactured majority, despite the use of a proportional representation system. In fact, the party won only 37.6% of the votes. This would place the assembly majority on the short list of lowest vote percentages ever turned into a majority by a “proportional” electoral system. Moreover, it was a “honeymoon” election–the assembly was elected about six months after the president.

The manufactured majority was possible due to several factors. One is the use of D’Hondt divisors, which tend to boost the largest party’s overrepresentation, especially when it has a strong lead in votes over the runner up (in this case, Citizen’s Platform, with 24.1%). It is also due to the legal thresholds: 5% for a party and 8% for a pre-electoral alliance. This resulted in substantial wasted votes, particularly with the Democratic Left running as an alliance and getting 7.6% of the vote. Had it become a “party” it would have been represented, and the PiS surely would not have won a majority of seats. Or, obviously, had the alliance won just over 0.4% more votes, it would have won seats.

However, the 37.6% itself did not come in a vacuum. As noted, this was a honeymoon election: it was six months after the election of the president to a five year term. Based on a formula (purely empirical, but based on theory) in Votes from Seats, we expect an election with 10% of the presidential inter-electoral period elapsed to result in a modest boost for the newly elected president’s party. The formula (shown in an earlier post on the French 2017 elections) suggests what we can expect for the party’s assembly votes, expressed as a ratio to the president’s own votes (in the first round if a runoff system). We call that ratio the “presidential ratio” or RP. For elapsed time of 10%, we expect RP=1.13. Given that the president, Andrzej Duda, had earned 24.8% of the vote, we’d expect the party to get 39.3%. This is barely over what it actually got (37.6%). So, yes, honeymoon elections matter. (By comparison, PiS had 29.9% in the previous assembly election, in 2011.)

It was not even a sure thing that it would be the PiS that would benefit from the honeymoon boost. First it had to win the runoff, and it did so quite narrowly (51.8% to 48.4%).

So, the current crisis could have been averted, most likely, if any one of the following had been true, thereby preventing the PiS majority government:

(1) Duda had lost the runoff;

(2) The assembly election had not been scheduled so soon after the presidential;

(3) The electoral formula was some common proportional method other than D’Hondt;

(4) There was not a high threshold for alliances, or the Left had made itself into one party (or gotten just a small increase in its votes).

So, yes, institutions do matter!

And institutions may yet matter one other way to this story, and in a way relevant to a couple of my earlier co-authored books. The president has a veto that takes 3/5 to override, and Duda has indicated opposition to his own party’s proposals on the judiciary.  He may veto it. Presidents, after all, may help their parties gain power, but they are not beholden to them once in office. Under Poland’s semi-presidential system, the president and prime minister may disagree, even when from the same party. So the story may have at least one other act yet to come.