Presidential elections with “distribution” or with ranked choice

In preparation for a grad seminar next week, I am re-reading Donald Horowitz’s 1990 article, “Comparing Democratic Systems”* in which he argues against Juan Linz’s critique of presidentialism as being too much of a “winner takes all” form of democracy.

In the piece, he speaks approvingly of alternatives to plurality and majority forms of direct executive election, and mentions the rules in Nigeria and Sri Lanka, which he says encourage the winner to appeal broadly. This has been on my mind anyway, with the surprise result in Sri Lanka’s presidential election yesterday. It is not clear to me how much these various cases of more complex rules for direct election have mattered in practice.

On Nigeria, Horowitz says:

To be elected, a president needed a plurality plus distribution. The successful candidate was required to have at least 25 percent of the vote in no fewer than two-thirds of the then-19 states. This double requirement was meant to ensure that the president had support from many ethnic groups. (p. 76)

Subsequently, a similar “distribution” requirement has been built into the presidential election methods of Kenya, where the rule requires 25% of the vote in at least five of the eight provinces. This did not prevent a winner in 1992 from having only 36% of the national votes. Nor did it prevent a serious crisis over the 2007 election.

I believe there is, or was, a distribution requirement in Indonesia, although the IFES description of the system in 2014 just says two-round majority.

The Sri Lanka system as described by IFES:

Under the contingent vote system, voters may rank up to three candidates. If no candidate wins a majority in the first round of counting, ballots whose first ranking are eliminated candidates are redistributed to the next-ranked candidates on those ballots. The winner is the candidate with the most votes after this second round of counting.

(I gather that means that third preferences come into play only for voters whose first two choices were both eliminated? As I understand it, all candidates but the top two are immediately eliminated when no one has a majority.)

Horowitz says about the adoption of this system in 1978:

It was expected that presidential candidates would build their majority on the second and third choices of voters whose preferred candidate was not among the top two. This would put ethnic minorities (especially the Sri Lankan Tamils) in a position to require compromise as the price for their second preferences.

It would seem that this motivation has never been realized; in fact, in the just concluded election, the winner did obtain support from Tamils and other minorities–but in the first preferences. Just as if it were a plurality system–or so it seems to me.

I do not think actual second or third preferences have come in to play in any Sri Lankan election, although I can’t claim to know. As for the “distribution” requirements, I wonder if any reader knows of elections in which campaign strategy was seen to be shaped by the rule. Of course, I generally believe rules shape strategy and behavior! But my limited knowledge of these cases does not lead me to be convinced that the outcomes have actually been different from what they would have been under more standard direct election rules.

* Journal of Democracy, Volume 1, Number 4, Fall 1990, pp. 73-79.

20 thoughts on “Presidential elections with “distribution” or with ranked choice

  1. Constitution of Indonesia 1945, Article 6A
    1. The President and Vice-President shall be elected as a single ticket directly by the people.
    2. Each ticket of candidates for President and Vice-President shall be proposed prior to the holding of general elections by political parties or coalitions of political parties which are participants in the general elections.
    3. Any ticket of candidates for President and Vice-President which polls a vote of more than fifty percent of the total number of votes during the general election and in addition polls at least twenty percent of the votes in more than half of the total number of provinces in Indonesia shall be declared elected as the President and Vice-President.
    4. In the event that there is no ticket of candidates for President and Vice-President elected, the two tickets which have received the first and second highest total of votes in the general election shall be submitted directly to election by the people, and the ticket which receives the highest total of votes shall be sworn in as the President and Vice-President.
    5. The procedure for the holding of the election of the President and Vice-President shall be further regulated by law.

    The trouble with regional distribution requirements (of which the US is the classic example) is what to do if the leading candidate in terms of votes does not reach the distribution requirement.

    Electoral college advocates in the US say the best solution is to elect the leading candidate by regions over the leading candidate by votes.

    All other regional distribution countries say the solution is to hold a second round where the regional distribution requirement does not apply.

    Neither rule seems particularly intuitive.

  2. Borda, Approval or some Condorcet variant may be more more effective at favouring compromise candidates with broad appeal, yet they are rarely proposed as alternatives.

  3. Even though I once did so, I would now not include the US electoral college in the same category. There is no specific rule for minimum shares of the vote in some minimum number of states. Besides, the provisions that I understand as “distribution requirements” are part of direct election rules, not indirect like the US.

  4. If I understand the Indonesia rule correctly, a candidate could be forced into a runoff despite having a nationwide majority. Presumably in the runoff, the same candidate would again win the nationwide majority. This seems odd.

    The Kenyan and Nigerian variants allow a plurality to suffice for victory in the first round, as long as the distribution is met.

    Although these rules are well intentioned, it is still not clear to me if they “work”, but then I am not sure how we would know. My idea of “work” is that they result in different strategies by campaigns than would a straight plurality or majority, which of course, requires one to figure out the counterfactual.

    JD, I am unaware of any case in which Borda, Approval, or Condorcet have been proposed in discussions of presidential elections, although I suppose it might have happened somewhere. There was, however, an academic article 25 or more years ago that promoted the idea of approval vote for the Finnish presidency, around the time that Finland was making the transition away from the electoral college.

  5. It would make more sense, if that is one’s goal, to copy the basic idea of the US Electoral College and combine the two (heads and regions) into a single tally, even if you count that tally in two or more stages (via run-offs or preferences). Eg, if you have (say) 20 regions, each is allotted two collegiate votes, allocated by PR. A ticket needs 33.34% of the vote in that region to get one of its collegiate votes and 66.67% to get both. (If the vote is so fragmented that one or even both votes go un-awarded, so be it).
    On top of that you have a set, uneven number of nationwide collegiate votes, constitutionally pegged at (say) eight times the number of regions, plus one (ie, just over 80% of the total). In this case this would mean a maximum of 161 national collegiate votes, which – combined with 40 regional collegiate votes – makes a grand total of 201. National collegiate votes are also awarded by whole (Droop) quotas. If no ticket amasses 101 collegiate votes on the first ballot, you hold a run-off between the top two. The run-off is held using the same system although, with only two competing tickets, all the collegiate votes should be awarded by whole (Droop) quotas.
    You could fine-tune it by adding turnout requirements, at least on the first round – eg “the quota per collegiate vote is [a Droop quota] of (a) the actual valid voting turnout or (b) 60% of all eligible voters, whichever number is greater”.
    You could also adapt this to a single-round preferential count. If enough votes exhaust, you might not award all collegiate votes by whole quotas, but then you could just stipulate that once only two presidential tickets remain in the count, whichever has a plurality of collegiate votes on the final (or sole) ballot is elected. If they have equal collegiate votes, more popular votes. I will not venture any recommendation about what to do if they poll equal popular votes as we’ve already covered that ground exhaustively (no pun intended).
    Changing the base allocation of equal collegiate votes per region would change the dynamics. Two per region would mean that as long as a major ticket can crack 33.3% everywhere in the country, a majority of nationwide popular votes will be decisive. Three per region, even by PR, would make it much more like the US electoral college, ie focused on the less popular regions. Four would be like two but less extreme, five like three but less extreme, and so forth. Whatever the number, I would argue strongly for keeping (a) an odd number overall, and (b) a 20% “equal per region” and an 80% “per capita nationwide” weighting.

    • The problem with the electoral college is the possibility of the reverse plurality like what happened in 2000 could happen again, fortunately it happens once in a blue moon, but unfortunately it doesn’t happen enough like twice in a row, so that it is abolished.

      NZ had two elections twice in a row where the second place party won and formed government, but it took a number of years before MMP was introduced.

  6. And I’m saying that a plurality (or final-round majority) reversal can be viewed as a feature of the system, not a bug, if (say) Party A gets 51% nationwide but carries only one-fifth of the regions, and Party B gets (say) 49% but carries four-fifths.
    Not what I’d recommend myself for Australia but there are countries where this might be warranted to prevent a civil war after the election results are declared.
    My beef with the US Electoral College is not so much that 18.7% of the voting power is distributed per statum rather than per capita, but that it’s all winner take all within each State. Five small States with four CongressReps each will collectively have as many Electors (25) as one large State with 23 CongressReps. But the large State is easier for the popular-vote winner to take all. It is virtually impossible, under present-day US electoral laws, for California, New York or Texas to split its Electoral College votes among different tickets as a result of the voters’ ballots (only as a result of “defector Electors” going rogue). A clean sweep of all five small States out of five is a tougher call, even if they are small in population. So whether the US Electoral College favours large or small States depends on whether the smaller States are deep red, deep blue, or purplish tinged. The larger States are, all else being equal, going to be closer in popular vote percentages, because of the law of large numbers.

    • “there are countries where this might be warranted to prevent a civil war after the election results are declared.”
      Considering how the electoral college brought the US to the precipice in 1860, colour me unconvinced, at least about the effectiveness of that particular model, the main problem being plurality voting as you pointed out.

    • Agreed, JD. Lincoln barely cracked 40% and won not a single Elector from the South. James Michener, in “Legacy”, has a fictitious Southern character exclaim “We didn’t vote for him! He ain’t my President!” An understandable fallacy when the official returns from the only people who legally “vote for the president” makes it look as if whole regions are unanimous.
      By the way, my maths above was (inexplicably) wrong. Those five small States in the example would have as many Electors as a biggie with 28 Representatives, not 23. You people are getting very slack. Can’t believe no one caught me out on that.

      • Lincoln was not allowed on the ballot in 10 southern states so it’s unsurprising that he had a dearth of electors in the South. It is particularly unsurprising that he did not gain any electors in South Carolina where the legislature chose the electors. The Republicans did not campaign in the South and indeed doing so would have been exceedingly dangerous for anyone foolish enough to declare themselves a Lincoln supporter.

        Nevertheless, if all the anti-Lincoln candidates had combined (an impossibility given that the other three candidates disagreed on secession) Lincoln would still have won the electoral college.

        Republican victory was due to the concentration of votes in the free states which together controlled a majority of the presidential electors.[33] The split in the Democratic party is sometimes held responsible for Lincoln’s victory,[34] but he would still have won in the Electoral College, 169 to 134, even if all anti-Lincoln voters had united behind a single candidate. In the three states where anti-Lincoln vote did combine into fusion tickets, Lincoln still won in two states and split New Jersey’s electoral college.[35]

        Douglas, the unionist Northern Democrat was the only candidate to carry states in both sections. As discussed previously, Southern unanimity for secession is vastly exaggerated and somewhat nugated by the 1 in 3 white Southerners under arms who fought in the Union army, by the rejection of secession in 4 Border South states, and by the counter-secession of West Virginia. It is very likely that East Tennessee would also have counter-seceded had the Union been able to protect them. Jones County and adjacent areas of North Mississippi were in active rebellion against both the state of Mississippi and the Confederacy for most of the war.

        It’s also worth recalling that while slaves could not vote they did effect the result because the Three Fifths Compromise vastly over-represented the South in the house of representatives and therefore in the electoral college. I am not completely sure they would have voted for secession if they had been able to vote.

        I think the electoral college a fairly silly institution, but it certainly did not drive secession or cause the Civil War.

      • Alan, if l understand correctly, this was before the time of pre-printed ballots with all candidates on them, so voters brought their own ballots usually distributed by parties. There were (don’t worry, I checked) a few slave states where a small number of ballots were cast for the Republican ticket, namely Virginia, Maryland and Kentucky.

        “it certainly did not drive secession or cause the Civil War.” Well I’d rather not get into a debate over what constitutes ‘driving’ or ‘causing’ a war or secession, but it seems clear to me that an institution that led to the election of a divisive candidate on a minority of votes which were so geographically concentrated played a big role in bringing southern states to opt for secession when they did, as opposed to, say, a two-round system where Lincoln might have lost to Douglas in the second round. What that role was exactly, or what its relative importance was, I make no claim to be able to evaluate, not being a historian or expert on the subject.

      • JD

        Minority presidents had been elected in 1824 by the famous ‘corrupt bargain’ in the house of representatives, 1844, 1848 and 1856. Without the Three Fifths Compromise Jefferson would probably have been defeated in 1800. Indeed Buchanan, the sitting president in 1860, had been elected as a sectional president on a minority of the popular vote. He carried all of the South and his own star of Pennsylvania by majorities and carried some other states in the North and West as a result of the split between Republicans and Whigs.

        What was different in 1860 was that secessionist in the Lower South were terrified by the decline of slavery, had exaggerated fears of Lincoln’s program, and (just quietly) exaggerated fears of political demands by Southern white yeomen for political and economic equality. They had beenr relaxed about minority sectional presidents, but only if they came from or supported the interests of their own section.

        Slaves were being sold out of the Border and Middle South to the Lower South in large numbers. Delaware, admittedly the smallest of the slave states, had only 1787 slaves in 1860. Maryland had over 87000, but also had 85000 free blacks who were expected to outnumber slaves in the near future. other Border and Middle South States were further behind but moving on the same trajectory. Jefferson Davis speculated publicly that the Border South would all be free states by 1870 and that Virginia would be lost to the South by 1880..

        In a two round system the leading two candidates would have been Lincoln and Douglas with Douglas facing a shortfall of 18.3% of the popular vote. I suspect the southern Constitutional Democrats would mostly have abstained. They had after all bolted the Democratic convention in Charleston because they refused to support Douglas as the presidential candidate because they regarded him as too weak on slavery. The Constitutional Unionists, at 11.8% were not in a position to elect Douglas and were in any case themselves divided over secession and slavery.

        Indeed, had Douglas been elected by some miracle under a two round system I strongly suspect he would have encountered much the same hysteria from Southern elites that Lincoln did.

        Massa’s great fear was losing control of their own states to a group described by a South Carolina senator as mudsills. Massa was alarmed by slaves but terrified of the poor whites they referred to as ‘the mob’.

        It’s worth recalling that the popular vote to elect secession conventions was usually a third smaller than the popular vote to elect the president. The minority sectional president argument is basically hysterical nonsense.

        It’s worth recalling also the unratified Thirteenth Amendment passed by the congress in 1861.

        No amendment shall be made to the Constitution which will authorize or give to Congress the power to abolish or interfere, within any State, with the domestic institutions thereof, including that of persons held to labor or service by the laws of said State.

        Thomas Jefferson called slavery the serpent under the table. In 1860 it crawled out and no electoral system could have pushed it back.

      • Alan, you make a strong case, but it’s not aided by calling me or anyone else hysterical.

        The proposed 13th amendment is of course a good piece of evidence (often forgotten, as you say) that the suppression of secession was (initially, at least) not (directly) about slavery but about maintaining the Union at all costs. I’m not sure, however, what you mean to say by mentioning it here.

      • Jd, the hysteria belongs purely to Massa and not to you.

        Many historians have noted an element of hysteria in Southern elite reactions to events in Washington. Abolitionist petitions started reaching the congress in the 1830s. A South Carolina representative proposed a roe that abolition petitions not be received. South Carolina threatened to secede if the rule were not passed immediately.

    • Perhaps keeping the electoral college, but requiring proportional representation using the STV within a state might be the best path to prevent electoral college plurality reversals, but the problem is that it would be very common for no candidate to gain an absolute majority in the electoral college, perhaps the leading candidate might win the presidency once the votes from the electoral college are counted, but then there would be extreme high political drama on who would win the election as right now most of the time when the members of the electoral college has voted; everyone knows what the result is.

      The worse thing that could happen is no candidate wins a majority in the electoral college of course that can happen under the current winner takes all system, then it has to be decided in the house and senate. Assuming the U.S house is still elected by Single Member Plurality, then it would be chaos as each state has an electoral college vote unless this part of the Constitution is changed allowing a majority vote of the House to elected the President.

      • I cannot see any prospect of an amendment that established proportional representation in the electoral college that did not also abandon each state having one vote in a contingent house election.

        I see very little prospect of any amendment to the US presidential election system that retains the electoral college. If a large blue state like Texas adopted proportional representation for choosing electors I suspect that would only drive countermeasures like adopting National Popular Vote. Uniform PR for the electoral college would itself require a constitutional amendment and it is hard to see it carrying 3/4 of the states.

        In 2000 Gore’s worst state was Idaho at 27.64%. It follows that Gore would have been the president in terms of either the Indonesia or Nigeria rules.

        Despite the frequent references in this thread to 1860, probably the worst example of electoral college reversals was 1876 where the Republicans simply stole the election and the Democrats acquiesced in return for a free hand in the South. The restoration of white supremacist Democrats to power in the South, often by actual coup d’état, condemned African-Americans to a century of informal apartheid. It cannot be called the constitution’s finest hour.

  7. By the way, “gaudiatrix” (Latin for “cheerleader”) is me. For some reason WordPress signs me in under one nom de guerre on some devices and under another on other devices.

  8. The combined unionist vote in 1860, Lincoln+Douglas+Bell/Republican+National Democrat+Constitutional Union was 73.9%. While the (northern) National Democrats carried Missouri in the Border South and part of New Jersey, the Constitutional Unionists carried the Border states of Kentucky and Tennessee, as well as Virginia in the Upper South. The (southern) Constitutional Democrats carried all of the Lower South South, the Upper South apart from Virginia, and Maryland and Delaware in the Border South. I have no idea why New Jersey split its electoral votes 4/3 between Lincoln and Douglas.

    • From Dubin (2002) via Wikipedia: “The Fusion slate consisted of 3 electors pledged to Douglas, and 2 each to Breckinridge and Bell. Nonetheless, different electors appeared in some counties for Breckinridge and Bell, resulting in lower totals for them and a split electoral outcome. The 3 Douglas electors were elected and 4 of those pledged to Lincoln. The Breckinridge and Bell electors finished behind all other candidates.”

  9. Thinking more on this, if you give every region (say) two automatic collegiate votes, they don’t need to be allocated by PR.
    If they were, then a slight change in party support within a region–from, say, 35-65% of popular votes to 32-68% — means a two-seat advantage, 2-0 instead of 1-1.
    Very crude proportionality; unavoidable when electing actual warm bodies to fill seats; but with this model we’re not really choosing live candidates to sit and vote in an electoral college, only keeping track of how the votes credited to each candidate are weighted. (It’s related to the “automatic plan” proposals, eg, James Michener in PRESIDENTIAL LOTTERY, to reform the US electoral college by getting rid of the electors but keeping the Electoral Votes)
    Instead, you could keep the same number of votes allocated to each the region but set the threshold at 40 per cent of the popular vote (or 30 per cent of total eligible voters, whichever number is higher).
    This would mean there is no single tipping point where 1-1 “suddenly” flips to 0-2. Instead at 39-61 per cent of the popular vote, a one-all split of seats becomes 1-0. It does not become to 2-0 until the popular vote margin becomes 80% to 19% or wider.
    In other words if the larger party has under 60 per cent it has only one seat if it has 60 to 79%, it has 1-0; only if it is over 80 per cent can it get a 2-0 clean sweep of all the regional unit votes for that region.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s