Small parties heading government: What are the costs and what can we learn ahead of possible Bennett-Lapid rotation in Israel?

The following is a post by Or Tuttnauer, based on a thread on Twitter. I asked Or if I could turn it into a F&V post, and he kindly agreed.


In Israel, Yair Lapid (Yesh Atid) and Naftali Bennett (Yamina) are trying now to form a cabinet, with Bennett the first prime minister in a rotation between the two. One problem (among others) – Bennett’s party commands only 6% of the parliament. Critics say he will lose even that at next election if he takes a turn as PM. Will he?  

I looked at  data of 474 PM-parties in 29 countries over 70 years (1945-2015) and how they fared in the next elections.

As the scatter plot shows, the vast majority of these parties lose votes in subsequent elections. Governing has its costs. But most PM parties are much larger than Bennett’s.

To figure out how change in vote share depends on vote share, I ran a regression with the former as DV, and the latter as well as its square value to allow for non-linearity. Turns out most parties lose votes, but not the small ones – below 30%. These are parties smaller than the average or median PM-party in the data (37% and 36%, respectively). For these smaller ruling party, the predicted gain or loss is indistinguishable from zero. Compared to the fortunes of the larger ruling parties, not losing, and not gaining votes is good news. Lucky for Bennett! 

But wait! what about ideology? Bennett’s party is also far from the centre, isn’t that a precarious position for a ruling party? Well, if we add an interaction with extremity, we see that at the very low end, extreme parties of up to 15% vote share seem to actually gain votes. This is intriguing. Perhaps (as suggested to me by Matthew Shugart), these extreme parties gain credibility after heading a government and are therefore perceived in the next elections as more moderate or mainstream than their ideology would otherwise suggest, leading to a wider electoral support. However, it may also be that there are too few cases in this range to make a meaningful inference.

So, should Bennett risk it and be PM? I say yes. First, if you follow Israeli politics, you know this is better than the alternative (if you don’t, trust me). Also, you don’t get many chances to become PM. And political narratives – like history – are written by the winner.


Below is the list of past cases of small PM parties, their extremity, and their vote share change at the next election.

CountryPartyCabinet start dateCabinet nameExtremityvote share (t)Vote share change
BELPVVVLD12-Jul-99Verhofstadt I2.005314.31.06
NORKrF17-Oct-97Bondevik I0.851613.7-1.2
ISLA28-Jun-59Jonsson II0.555612.52.7
NORKrF19-Oct-01Bondevik II0.851612.5-5.7
BELPVVVLD10-Jun-07Verhofstadt III2.005311.83-3.19
FRARPR20-Mar-86Chirac II2.499711.57.7
ITAPSI4-Aug-83Craxi I1.227811.42.9
BELCVP13-Jun-10Leterme III0.766710.850.76
NORSp12-Oct-65Borten I0.34359.8-0.8
NORSp7-Sep-69Borten II0.343592.03
NLDARP7-Jun-71Biesheuvel I0.80678.590.25

Party-switching on the way to the top

While updating, via a comment, my post about the incumbent Vice President of Panama who had become opposition leader and has now been elected President, I got to probing some of the biographical data.

I cite some data there about the propensity of vice presidents to become president, and of presidents (whether previously VP or not) to change parties prior to making it to the top job. I noted that presidents are significantly more likely to have changed parties at some point than are prime ministers in parliamentary systems: 40.7% vs. 24.6%.* This is, of course, totally consistent with the theory of “presidentialized” parties, whereby party loyalty is less important than electability when assessing candidates for the top job.

I wondered about premiers in the two sub-types of semi-presidential system.

We have data on the career-long party affiliations of 105 premiers in premier-presidential systems and 134 premiers in president-parliamentary systems. The basic distinction by subtype is in whether the formal accountability of the premier is exclusively to the legislative majority (premier-presidential) or dually to the legislature and president (president-parliamentary).

35.1% of premier-presidential premiers had switched parties before ascending to the post, whereas only 17.1% of their counterparts in the other subtype had. That is significant at p=.002.

The effect is in the direction that I expected but bigger than I expected. I figured that where the presidency is the more dominant constitutional actor, i.e. in the president-parliamentary systems, presidents would tend to appoint loyalists, who in turn are less likely to have switched parties at some point.

However, that more than a third of premiers in premier-presidential systems have switched strikes me as high. This is, allegedly, the more “parliamentary” of the two subtypes. On the other hand, that they come down right between the top executives in the two pure types makes sense. In fact, the differences between premier-presidential premiers and either parliamentary PMs or elected presidents (of any regime type) are not significant.

Thus, at least in terms of their tendency to have party-switched, premiers in premier-presidential systems mirror the genuine hybridity of their regime type, whereas their counterparts in president-parliamentary systems look like the ultimate in loyalists. That seems about right!

* That’s presidents of pure presidential systems; if we include semi-presidential presidents, it hardly changes: 41.6%. In some respects, presidents are presidents, regardless of other regime features.

Young and inexperienced–how common?

New Italian PM Matteo Renzi has never served in the national legislature or cabinet (till now), and is only 39. These are unusual characteristics. Usually parliamentary parties prefer to “vet” their executive talent for a while through having them serve in the legislature and/or cabinet before being elevated to the top job (much more so than in presidential systems, where the candidate for the top job has to be able to win a plurality or majority as an individual–see Samuels and Shugart, 2010, 2014).

How common is it for the head of government of a parliamentary democracy to be as young and inexperienced as Renzi? Some insight comes from the Executives Biographical data of Samuels and Shugart. Here I offer some lists selected with intent to compare Renzi to other PMs. Caveat: in addition to being post-WWII only, the dataset ends with 2005. I won’t be updating it any time soon, but of course I would welcome readers’ additions via comments to this entry.

The first list (Dropbox link) is of all the parliamentary PMs in the dataset who had never served in the legislature before, along with how many years they had served in the cabinet prior to becoming PM (yrscab) and their age when starting their stint as PM (agestart). The list contains only 24 names–these are all the PMs in parliamentary systems who had no prior legislative service. That’s out of 411 total. So lack of legislative experience is quite rare. Exactly one of them was younger than Renzi is now (Vasile Petru Tarlev of Moldova, 38 when he took the job in 2001). Several on this list can be explained through newness of the democratic regime itself (e.g. Mandela) or immediate post-war years. You will note the multiple appearances of a few countries* on the list, including… Italy.

A second list has all parliamentary prime ministers who assumed the position before the age of 40. It is also a short list, and it is heavily dominated by young democracies, mainly in Central-Eastern Europe. It looks like our youth champion is Pandeli Majko of Albania, 31 when he assumed the job in 1998, followed by Mart Laar of Estonia, 32 when he became PM in 1992; Laar began a second stint seven years later, when he was still about the age of Renzi now. We also see from the list that, despite their youth, some of these PMs had considerable experience already in the legislature (e.g. Felix Gaillard of France, 10 years**) and a few had cabinet service (e.g. Stanislav Gross of the Czech Republic and Aigars Kalvitis, 4 years each).

One more list of background relevant to Renzi: how common is it for a PM to have been a mayor, but not a legislator or cabinet minister before elevation to the top job? Renzi might be only the third (see caveat above), following JirĂ­ Paroubek of the Czech Republic (2005) and Jawaharlal Nehru of India (1947, the year of Indian independence). Upon assuming office, Paroubek was 53 and Nehru 58.***

So Renzi’s combination of youth and inexperience, aside from having been mayor of a major city (Florence), is indeed unusual.

Previous related post: Age of PMs and presidents upon assuming the position, in new and old democracies.

* There are five Netherlands PMs on this list, which is a bit surprising. Service in the senate, perhaps? And that makes me wonder if we counted service in the Italian senate, which we should have, given it has confidence powers over the cabinet, unlike most other second chambers.

** Ilir Meta of Albania, 15 years service as MP, is, I am sorry to say, a mistake in the data! He was born in 1969, elected to parliament in 1992, and became PM in 1999 (not 2001, as the list indicates), according to an online bio.

*** Nehru was a mayor? That is what the dataset says. According to Wikipedia, he was elected chairman of the Allahabad Municipal Board in 1923. (Yes, students, I can use Wikipedia. This is a blog post.)