Palestinian Territories 2006: Visualizing (what may or may not have been) overnomination

The election may have been eight and a half years ago, but it continues to fascinate me…

If you have not read at least the latest of the two (and more) posts on this election, you may need to do so before this one will make sense.

I am looking for ways to visualize the relationships among Fatah, Hamas, and the independent candidates in the (mostly) multi-seat districts of the nominal tier of the Palestinian 2006 election. The question is to what extent Fatah may have cost itself seats–maybe even an achievable plurality–by “overnominating”. If a party overnominates, it has more candidates than its votes allow it to elect. In this case, because the voter could cast up to M votes (where M is the magnitude of the district), a party can nominate M candidates and be OK, provided it gets its supporters to cast all their M votes for their M candidates. However, if there are independents (or candidates of other parties) who appeal to the same block of voters, a party might see attrition of its voters and fail to elect as many as it might have with fewer candidates. The problem is that there may be a “camp” of Fatah plus Fatah-rejected independents, the latter having been denied the party endorsement but deciding to run anyway.

Here are some different ways of trying to assess the question of whether the Fatah camp overnominated from the available data. (In each case, you can click the image and get to a larger version.)

The first graph shows the total vote share of independents in each district on the vertical axis. On the horizontal axis is the party’s ratio of nominal to list votes in the district. That is, the sum of all the party’s endorsed candidates, divided by the party’s list votes. The red line is the local regression (lowess), and we have separate graph panels for Hamas and Fatah.

Graph Ind Ratio

From this graph, it seems there was attrition from both parties to independents, albeit only sometimes in the same districts. (For example, in Tulkarem, where the vote share for all independents combined was around a third, Hamas had a much lower ratio than Fatah; the latter was at over .9.) It certainly is the case that both parties have a lower ratio where the district’s share of independents is greater, but there is no question that Fatah has a stronger relationship between the two variables. So, yes, it seems like Fatah may have seen more attrition from list to nominal for its candidates than did Hamas. However, I was surprised at how much a relationship there was for Hamas as well.

The second graph is a box-and-whiskers plot summarizing the distribution of a “candidate ratio” across both major parties in each district. Above each district abbreviation there are the numbers 2, for Hamas (Change and Reform) and 3 for Fatah. The candidate ratio needs a little explanation. First, I start with a magnitude-adjusted vote share for each candidate. This is their individual votes, divided by V/M, where V is the total number of nominal votes cast in the district and M is the magnitude. This way, we can normalize the shares across districts of varying magnitude. Then the ratio–what is actually being graphed–is this magnitude-adjusted candidate vote share divided by the party’s list votes. The purpose of this mathematical gymnastics is to get an idea of how much each individual candidate of a party deviated from the party list vote in the district. Thus where the box or whisker or an individual outlier point is above 1.0, it implies a candidate more popular than the party in the district, and where it is less than 1.0, the candidate ran behind his (rarely her) own party.

Graph box cand ratio

I had previously noted that the standard deviations of individual candidates’ votes were pretty small for both parties, but greater for Fatah. This is clear here, in that the “whiskers” are farther from the box for Fatah (party 3) than for Hamas in most districts. This is especially so in Jerusalem (Jsm). And all the Fatah candidates ran quite far behind their party in Bethlehem. In fact, I think the “two slates” thesis–that Fatah (or more precisely the “Fatah camp”) shot itself in the foot by nominating an extra set of candidates–really rests on these two districts. There is rather less evidence for it elsewhere. However, one can still see that, in general, Fatah candidates seem to run a little farther behind their party than do Hamas candidates, and most of the cases of the candidate ratio being over 1.0 for an entire party are Hamas.

(In cases where the data plot is just a horizontal line, we are looking at a single-seat district. Note how far ahead of his party the Fatah candidate was in Jericho. This is Saeb Erakat.)

The final graph is a little bit overwhelming, I know. But I wanted to plot all the candidates’ individual votes. Here again, I am using the magnitude-adjusted vote share. The constituencies are indicated by numeric codes (because that is the only way I could (think of to) do it in Stata), but I made sure they are in the same order as they appear in the second graph.

Graph votes_cand cst_a jitter
This version has “jittered” data points so that it is easier to see where two or more candidates were very close in votes. This is helpful, given the low intra-party deviation. The original, un-jittered, version is still on my Flickr site.

Green is for Hamas, black for Fatah, and orange for independents. Solid symbols indicate winners, open losers, though down near the bottom of the graph there are so many bunched together that they sometimes look solid. It should be noted that where there are Fatah winners around .20 or lower, these are Christian quota candidates.*

This graph allows one to visualize where there were independents in a district who were well ahead the pack of minor independents, and also where losing independents are found amidst losing Fatah candidates, or just behind the losing Fatah pack. I don’t actually see a lot of them, but I can imagine that the highest-polling losing independents in Nablus (code 9), North Gaza (10), Salfit (14), and Tubas (15), at least, may have robbed Fatah of a seat it (and its camp) could have won. It is worth noting that Salfit and Tubas are one-seat constituencies won by Hamas candidates who had under 35% of the vote. Other candidates for seat-robbing are Bethlehem and Jerusalem, where there are several independents with over 10%, and where we already saw big drop-offs in Fatah candidate votes, relative to the party list–and where I already conceded that the “two slates” thesis might have legs. But remember, from the first graph, Hamas also had substantial drop-offs in these two districts. It could be that some of these independents, albeit possibility from the Fatah camp, appealed to Hamas voters precisely because they were independent and not Fatah-branded.**

Rather than draw firm conclusions here, I just want to put the visualization of the data, by these three different but related means, out there for folks to comment on. Maybe someone will see patterns beyond those I pointed out.


* There are two of these in Bethlehem, with nearly identical vote shares, and two in Jerusalem; in addition, one of the independent winners in Gaza City is a Christian whose votes, it might be noted, are higher than any Fatah-affiliated candidate. He is the third ranked independent winner in that district.

** If so, that would have been even better reason to give the official endorsement to fewer candidates, thereby allowing some in the camp to appeal across party lines.

Key to districts, their codes (in the third graph), abbreviations, and magnitudes (in that order).

cst_n cst_a cst_abbrev mag
Bethlehem 1 Bet 4
Deir Albalah 2 Dei 3
Gaza City 3 Gaz 8
Hebron 4 Heb 9
Jericho 5 Jch 1
Jenin 6 Jen 4
Jerusalem 7 Jsm 6
Khan Younis 8 Kha 5
Nablus 9 Nab 6
North Gaza 10 Nor 5
Qalqilya 11 Qal 2
Rafah 12 Raf 3
Ramallah and Albireh 13 Ram 5
Salfit 14 Sal 1
Tubas 15 Tub 1
Tulkarem 16 Tul 3

The Fatah “over-nomination” thesis, reconsidered

Did Fatah over-nominate, and thereby cost themselves a majority in the Palestinian Legislative Council election of 2006? I have previously cast doubt on such claims, because Fatah had no more candidates in any district than the number of seats being elected. My understanding of over-nomination in a system where voters have M votes (M being the magnitude; i.e. the number elected in the district) would be a party having more than M candidates. By this standard, no, Fatah did not over-nominate.

However, this could be too narrow a view of over-nomination. We might want to include broader definitions of “camps” in our definition of the political tendencies that have potentially too many candidates.

I have ignored the independents before, because I have no information on them, and no way to know whether they were in any way affiliated with Fatah or any other party or movement in the election. A paper prepared back at the time by Jarret Blanc for IFES, “Palestinian Election Analysis: How Hamas Won the Majority“, has now come to my attention. The paper itself is no longer available on the IFES website. However, it is cited in Hamas: The Islamic Resistance Movement by Beverley Milton-Edwards and Stephen Farrell,* which I just finished reading. (I recommend it highly, by the way.) The authors cite Blanc’s finding that the number of additional legislators that Fatah could have won with greater unity was 18.

So I went back to my dataset on the election and attempted to replicate this. I made the following assumptions: (1) every defeated independent’s votes would have gone to a defeated (official) Fatah candidate, and (2) the votes would have followed the candidates’ actual ranks in votes in the district (i.e. the first losing independent’s votes would have gone to the first losing Fatah candidate, the second to the second, etc.). Note that these assumptions represent extreme leaps of faith! But without further information I can’t do much better.** If we do this, how many of the Fatah candidates now win, given their “revised” higher vote totals? I get 17. So my count essentially agrees with Blanc’s, which I assume was determined with some actual knowledge of the independents’ affiliations.

Of course, nearly all 17 (or 18) seats would have come at the expense of Change and Reform (Hamas), and the parliamentary breakdown would have been something like 62 Fatah, 57 Change and Reform. Out of 132 seats. (I have to assume the party-list votes, and hence seats, would have been the same.)

A passage in the Milton-Edwards and Farrell book further elaborates on the challenge Fatah faced:

Fatah’s campaign chief, Nabil Shaath, conceded that his biggest concern was the party’s rejected independent candidates, who stood against the official Fatah candidates out of genuine if misguided hope of winning, in order to punish the party for rejecting them, or to use the threat of candidacy as leverage to obtain some other benefit. (p. 251)

So did Fatah over-nomiante? Technically no. It nominated M official Fatah candidates per district in a system where voters could cast M votes. Moreover, as noted in the previous post on the topic, the party’s actual candidates typically did not deviate much in votes from one another, implying a fairly strong party orientation of its voters (albeit weaker than that for Hamas) around the candidates it actually endorsed.

Did it nominate badly? Maybe. But this is again down to the electoral-system design. In systems where nominal votes are decisive–as in the districts of the Palestinian electoral system–it is inherent to the system design that independents can run, that candidates can make personal appeals distinct from party label, and that some individual candidates may prove less popular than the party as a whole. The party needed to figure out a way to manage these personal factors before settling on candidates, and Fatah may have failed to do so. It does not follow necessarily that it should have run fewer Fatah-branded candidates, and there is always the possibility that had it given the party nod to some of the independents then some of its actual candidates might have run as independents instead.

None of this reconsideration of the original thesis changes the fact that Change and Reform won more votes–on the party list as well as for its candidates–than Fatah, and that the electoral system magnified the lead of the party whose candidates won the plurality. But it is possible that poor internal practices might have prevented the Fatah “camp” from converting its full pool of potential voters into seats.

Too bad they did not opt for an open-list PR system. Even SNTV would have been better for them.



* It was actually Andy Reynolds who, upon reading my earlier post, mentioned Blanc’s paper and the possibility of over-nomination. I was already reading the Milton-Edwards and Farrell book by then, and happened upon their citation to Blanc several days after Andy mentioned it.

** It is actually more likely, perhaps, that it would be the official Fatah candidates with fewer votes who would have picked up more of the independents’ votes. That is, the more popular official Fatah candidates more or less by definition already appealed to a larger bloc of the Fatah camp’s voters. The statistical assumptions involved in reassigning the votes in this way are simply too complex to bother with. (If anyone wants to try, I am happy to share the data!)


Some other points from the Milton-Edwards and Farrell book are worth quoting.

More observations on candidate-selection:

While Fatah put its strongest people on the national list, Hamas did the opposite, because it calculated, correctly, that people would vote along party lines for the party list but would be swayed by personal considerations when choosing a local MP. ‘When you vote for Hamas on the national list you are voting for the party, not for individuals. In the districts you select individuals, so if people are corrupt, not credible, people will not vote for them. Fatah wasted their strong candidates’, said Aqtash. (p. 255)

The quote within the quote is from Dr Nashat Aqtash, “a Nablus-born public relations expert”. Of course, I agree entirely, although I would again point out that the deviation of a party’s candidates’ votes shares would be expected to be higher than it was if personal factors were really a large factor. Still, the deviation was indeed higher in most districts for Fatah than for Hamas.

Aqtash is also quoted as saying:

I told [Hamas] not to run more than 50 per cent of candidates in the election because, if they actually won, they would find themselves in an impossible position. They wouldn’t be acceptable to the international community and they would be embarrassed in front of their people. But [Khaled] Meshaal insisted on running with a full list. (p. 255)

On the campaign and party label used by Hamas:

Even Hamas’s opponents conceded that its choice of electoral label – ‘Change and Reform’ – was inspired, capturing the pent-up desire among Palestinians for a new broom. The election manifesto also downplayed Hamas’s implacable external agenda, making no mention of its ultimate goal of eradicating Israel. Instead it spoke of ‘resistance to the occupation’ and ‘balanced’ relations with the West. It was a wolf in sheep’s clothing. In fact, Hamas’s new cadre of articulate spokesmen did continue to insist that the movement retained its claim to all of Palestine – Israel as well as the occupied Palestinian Territories. But they said it was still committed to Yassin’s 1997 offer of a long-term hudna based on a Palestinian state in the West Bank, East Jerusalem and Gaza. Hamas’s customary insistence on proclaiming its commitment to the armed struggle was sidelined behind management consultancy talk of priorities, competence, transparency and delivery of services. (p. 248)

The authors also offer support for something I warned about in a post at the time, warning we should not believe the exit polls: many respondents may simply have lied to exit-pollsers. They suggest this may even have been something Hamaas told its supporters to do.

The Palestinian 2006 election revisited

I wrote about the 2006 election for the Palestinian Legislative Council quite a lot at the time. For instance, as exit poll results were coming in, I warned not to believe their reported lead for Fatah, and indeed they quickly proved inaccurate. Once the results were in, I noted that the magnitude of the Hamas sweep was a product of the electoral system.

In light of recent events, I decided to go back and take an even closer look at the results of that election in January, 2006, which turned out to be not the prelude to a period of calm while Hamas figured out how to use a legislative majority and the near-majority of cabinet seats it received in a unity agreement more than a year afterwards, but a prelude to three major flare-ups of violence between Hamas and Israel. More specifically, I was prompted to go back and look by a remark in Peter Beinart’s article in Haaretz, entitled “Gaza myths and facts: what American Jewish leaders won’t tell you“. The piece itself makes some decent arguments, although I think Beinart is selective in his facts to a degree that is not entirely distinguishable from these apparently monolithic American Jewish leaders he refers to. However, let me stick to the one set of facts I do know something about: the distribution of votes and party strategy in the 2006 election.

Beinart says that one of the reasons Hamas (running under the label, Change and Reform) beat Fatah was “because Fatah carelessly and foolishly ran both its slates in too many parliamentary seats.” Underneath those words he has a link to a You Tube video of a talk by that noted specialist on comparative electoral systems, Bill Clinton (please pardon my snark). On the one hand, I am thrilled that Beinart and Clinton acknowledge that electoral systems and party strategy matter. On the other hand, they actually are perpetuating a myth here. Fatah lost because it had fewer votes, not because it split its vote. Clinton uses the analogy of southern Democrats and their factions in the past when they were dominant in the region. But the analogy is not helpful.*

Fatah did indeed run more than one candidate per district because, well, districts (most of them) elected more than one seat. Hamas did the same, and it did not hurt them. In fact, this was a system in which voters had the possibility to vote for as many candidates as there were seats in their district. Parties could not pool votes, voters could not cumulate nor could they just select a party. (I am not speaking here of the totally separate closed-list PR portion of the electoral system, and from the context I presume neither is Clinton.) A party would win all the seats in a district if it ran a candidate for every seat, and if its voters gave all their votes to candidates of the party. And, of course, if its candidates had the top-M vote totals, where M is the magnitude (number of seats being elected in the district).

It was actually Hamas that did not always run M candidates. However, the places where it ran fewer were either where some of the seats were set aside for Christians (not surprisingly, Hamas, the name of which is an Arabic acronym for Islamic Resistance Movement, did not contest the Christian seats), or where the seats it did not contest were won by independents. For instance, in Gaza City, eight were elected (M=8). Hamas ran five candidates, Fatah ran eight. Hamas elected all its five, and Fatah elected none. But the reason Fatah did not elect anyone was not that it had too many candidates. The other three seats went to independents. One of these was a Christian, for a set-aside seat. The other two were independents whose vote totals were more than 11,000 votes higher than that of the most popular Fatah candidate. These independents may have been “quiet” Hamas affiliates not bearing the label. One, Jamal Naji El-Koudary, is an academic at the Islamic University of Gaza.

Neither party suffered greatly from any failure of its voters to be willing to cast a full slate by voting for all the party’s candidates running in the district. However, Fatah did suffer a bit more. I calculated the standard deviation of each party’s candidates’ votes in each district. The closer this number is to zero, the closer the candidates were to having identical vote shares:

    Hamas, .016.
    Fatah, .026.

Not much difference, but enough to make a difference in some places. But this does not mean that Fatah’s running of too many candidates was the reason for winning fewer seats. In this type of electoral system (what I like to call Multiple Non-Transferable Vote, but others call Block Vote), a higher standard deviation can help you win more seats than you otherwise would, in cases where your party overall ranks second. That is, some of the districts where Fatah elected some candidates despite Hamas having one or more candidates with higher vote shares were precisely where it had individual candidates who were more popular than the party as a whole.

For instance in Nablus (M=6), Hamas ran five and Fatah six. The winners were all five Hamas and one Fatah. An independent (Hamas-affiliated?) was the second loser, just behind the second Fatah candidate. The one Fatah candidate who won had 39,106 votes, whereas the party’s next highest vote-earning candidate had 35,397. The five Hamas candidates had from 44,957 to 36,877. The most important point here is that Fatah could not have won more seats simply by running fewer candidates. This is not a single-vote system (like SNTV or like the US primaries Clinton referenced) where running multiple candidates can split your vote. Had Fatah run fewer candidates, its voters would have cast either fewer votes or given the remaining votes to other parties’ candidates, or to independents. It could have won only by running candidates who could beat a Hamas candidate–getting more than the 36,877 won by the district’s sixth-ranked candidate.

Fatah’s biggest problem was that it just was not as popular. Its collective vote total for candidates was higher than that of Hamas in only six districts out of sixteen. And some of those where it was in second place were the bigger districts, where even a small plurality for Hamas, combined with a low standard deviation of its candidates’ votes, would mean a Hamas sweep. Hamas won a plurality of the candidate votes in Gaza City (M=8, 32.7%-31.7%) Hebron (M=9, 51.1%-35%), Jerusalem (M=6, 33.7%-26.4%), Nablus (M=6, 38.2%-36.5%), among others.

Those vote totals show that Hamas was often well short of a majority. In fact, nationwide, its candidate votes amounted to only 40.8%, but Fatah was well behind (36.6%). That is a lower vote share than in the national list vote, where Hamas won only about 44% to Fatah’s 41%. In fact, I had never summed up the nominal (candidate) votes before. I have always reported the outcome as 56% of seats on 44% of votes, but that is list votes. Given that it was the nominal tier that was the disproportional part of the electoral system, it is actually more accurate to say that Hamas won 56% of seats on not even 41% of votes. It does not change (or reform) the outcome, but it underlines just how disproportional the system was.

A non-proportional electoral system of mostly multi-seat districts sure can turn a small plurality into a big majority. Hamas was more popular, but well short of a majority. The electoral system mattered in a big way, but Fatah’s running multiple candidates was in no sense the reason why Hamas won.

One final observation from the election results: Hamas has controlled the Gaza Strip since its rebellion against the Palestinian Authority in mid-2007, and the Strip was long an important base of operations for Hamas. However, in the election it did not show significantly greater strength in the Gaza Strip, where it won 41.3% of the nominal vote, hardly different from its West Bank support. It did dominate Gaza City’s list vote, however, with 56.7%, which was by far its best showing on the list within in any nominal-tier district. (The list was national, but results are disaggregated by district.) Interestingly, its candidates collectively won only 37.3% in Gaza City. Let’s just say that the personal vote was not how Hamas won, but the disproportionality of the nominal tier more than made up for relatively weak candidates.**

None of this helps resolve the current dreadful crisis, but it does resolve that Peter Beinart, while attempting to counter “myths” with “facts” is perpetrating a myth of his own–as is Bill Clinton–that running too many candidates was what doomed Fatah. That is simply incorrect.

* There was initially a split slate, but party registration was actually re-opened to allow Fatah to present a unified slate for the actual election.

** In most cases, both Hamas and Fatah had higher list than nominal votes. Although there were several small parties running lists that did not have candidates (at least not labelled), there were many independent candidates. Only 4 of them won, and most others were not close to winning. (Their mean ratio of votes to the district’s last winner was .127, and only one had a ratio greater than .66.) They did, however, combine for 20.7% of all nominal votes cast. I can’t rule out that some of them drained votes from Fatah candidates, although that is not Clinton’s and Beinart’s claim. They claim there were too many Fatah candidates.

Data source: Central Elections Commission – Palestine. I used Adam Carr for the district-level nominal votes, because his page was formatted in a way that was more easily transferred to spreadsheet format. I verified that it matched the Commission’s data, and corrected a few examples where it did not. I typed in the list votes per district. I generated various summary statistics and analyses in a Stata file.