A Review of Haiti's
May 21, 2000 Senatorial Elections:
By James R. Morrell
Presented to the Roundtable "The Future
of Haiti in Light of its Past"
chaired by Prof. Robert Maguire, Trinity
Latin American Studies Association
September 8, 2001
The material distributed consists of abbreviated senatorial
tally sheets from Haiti's first-round legislative election
on May 21, 2000 and related calculations. The generally-accepted
view of the first round is that it was a landslide for
the Lavalas Family, which then went on to commit fraud
in the central counting even though it had already won.
However, a more careful review of the tally sheets finds
that the non-FL parties together actually outpolled
the FL in four of the eight departments that voted on
May 21. They did so in the Artibonite,
Central Plateau, Nord-est,
and Nord-ouest. FL outpolled
the non-FL parties in the more populous Nord,
Sud, and Sud-est
departments. (Grand' Anse did not vote.) The data are
also presented in a summary.
The review also establishes:
- In six of the eight departments, the number of votes
discarded by the fraudulent top-four method of counting
was actually greater than that received by either
- In applying the top-four method, the CEP counting
office used exactly the same 50-percent method that
it decried when suggested by the OAS.
- If exactly the same distribution of votes is projected
onto the second round, then using the OAS method (straight
50 percent), the non-FL parties stood to win eight
senate seats against nine won by FL in the first round.
(Since voters' behavior is not linear, it
is not suggested here that that would have been the
actual outcome. Any number of voters could have switched
to FL or vice-versa.)
- If an allowance is made for blank ballots (one per
five voters) and the same distribution is projected
onto the second round, the non-FL consituency stood
to win nine seats against three more for FL. (See
These findings suggest that while the overall impression
of a Lavalas victory in the first round remains correct,
this victory was more contingent than generally acknowledged.
In particular, the numbers presented Jean-Bertrand Aristide
and his Lavalas colleagues with two major uncertainties:
1. Would the blank votes be counted and so reduce first-round
FL victors from nine to approximately five?
2. Would the FL/non-FL distribution observed in the
first round hold in the second?
If either of these conditions held, the opposition
might win the second round.
The above findings suggest that Aristide's recourse
to the fraudulent top-four method may have had a greater
"pragmatic" basis than generally accepted.
Description of Material Distributed
The distributed material consists of abbreviated senatorial
tally sheets and calculations for all eight departments
that voted. The actual names of the senators are left
off in the interest of brevity. Page 9 is a summary
sheet that reviews first- and second-round scenarios.
The OAS method was simply to divide senatorial ballots
cast by 50 percent. Note that this was most favorable
to FL in that it yielded nine first-round victors. If
blank ballots are counted at the rate of one per five
voters (thus one per nine ballots cast), the denominator
increases enough to eliminate four first-round victors.
The above are the major findings of this paper. Below
we present background material already in the public
Haiti's constitution and electoral
law mandate a first-round victory to a candidate who
wins 50 percent plus one of the ballots cast. The constitution
sets a staggered schedule of senatorial elections with
nine senators, or one per department, being chosen every
two years. However, because Haiti failed to hold legislative
elections in 1998, in the 2000 elections two senators
were up for election in each department and in the Central
Plateau, three because OPL senator Yvon Toussaint had
been assassinated in March, 1999. So voters on May 21
received ballots to vote for more than one senator.
Then to establish whether a candidate received 50 percent
plus one, it was necessary to divide the total number
of votes cast by two (three in the Central Plateau)
and compare the candidate's votes to this number.
Thus if 100 voters cast 200 votes for
senatorial candidates, to exceed 50 percent of the votes
cast per senator a candidate would have to
receive 51 votes.
Instead of following this procedure,
by June 1 the CEP's directorate of operations headed
by Luciano Pharaon issued returns counting only the
top four finishers. Former president Aristide's role
has not been established, but presumably it was major.
Cutting it off at the top four reduced the field enough
so that the FL candidates, who were leading in most
races, got way over 50 percent and so could be declared
first-round victors in sixteen of seventeen races (see
the department charts).
The OAS and opposition rejected this method as fraudulent,
which it was. The president of the CEP, Léon
Manus, at first resisted the OAS but then reversed himself
and by mid-June was preparing to issue returns mandating
a second round for all but five FL senatorial candidates.
He was threatened by President Préval and Aristide
and fled into exile with U.S. government assistance.
The OAS refused to observe the second
round and withdrew its mission. The opposition boycotted
the Grand' Anse vote, the second round and the November
presidential and senatorial elections. With only FL
candidates running in these elections they ended up
monopolizing the 110-member legislature. The dispute
over the counting of the May 21 election, and the cascading
series of boycotts, remains unresolved at this writing.
In defending the top-four method, the
head of the directorate of operations, Luciano Pharaon,
wrote in June 2000 that applying the OAS method of dividing
the vote in half would have introduced phantom votes
because some voters did not vote for both senators.
Admitting these phantom votes would be contrary to the
constitution and would devalue the votes of those who
did vote for senators.
However, the tally sheets show that
the directorate of operations proceeded to use the very
same method of dividing the vote in half. The only difference
was that it divided in half only the vote received by
the top four finishers instead of all the votes in order
to attain the denominator for the percentage calculation.
On June 14, 2000 the CEP published
the returns in the proper way, counting all the votes
and dividing the senate returns by 50 percent (33 percent
in the Centre, where there were three open seats).
In determining the number of candidates
who won on the first round, it made considerable difference
whether the blank ballots were factored in. It's estimated
that one in five voters only voted for one senatorial
candidate. Then there would be one blank ballot for
every nine cast. The constitution and electoral law
mandate the counting of blank ballots. The number of
blank ballots is included in the returns from every
precinct. The blank-vote issue came to a head in the
disputed 1997 vote.
The OAS in pressing for abandonment
of the fraudulent top-four-cutoff method, and pressing
for counting all the votes, apparently left the blank-vote
issue to the side. Counting only the votes cast for
candidates, nine FL candidates won on the first round,
and this was the number the OAS gave. However, the inclusion
of blank votes at the rate of one per nine senate votes
cast increases the field (and so the denominator in
the 50-percent calculation) sufficiently to reduce the
number of FL first-round winners to four or five, which
is the number CEP president Léon Manus was preparing
The potential for the non-FL candidates
to improve in the second round is based on the fact
that in the first round the non-FL voters scattered
their votes among more than fifteen parties, while FL
voters all voted for one party. That enabled nine FL
candidates to win in the first round by the OAS method
, or five by including the blank ballots. In the second
round it would be one-on-one per senate race. Those
who had voted non-FL in the first round would have only
one non-FL candidate per race to vote for in the second.
If they all decided to continue to vote non-FL, then
from those four departments with a majority of non-FL
voters in the first round (Artibonite, Centre, Nord-Est,
Nord-Ouest) they would have the votes to elect eight
non-FLers (OAS) or nine (blank votes).
The enforced unity that a one-on-one
runoff imposes on the runner-up parties has been observed
in many countries to improve their chances. Whether
it would in Haiti we will not know until a second round
is properly held.