by Franklin Midy
Translated by Max Blanchet

On February 10, 1999, IRI published on its web site (www.iri.org/) the results of its poll whose goal "was to identify the opinions of politicized Haitians with a civic posture vis-à-vis the Haitian electoral process." Farther in the report IRI makes the point that: "in identifying these opinions, this study will define potential solutions to improve the electoral process which will result in a useful increase in the participation of citizens and political parties. This identification will also help realize one of IRI's goals, namely to correct the problem caused by the paucity of viable information on attitudes and political preferences."

These goals are undoubtedly praiseworthy since democracy is about a public space involving well thought-out debate buttressed by information available to all. But, the results fall below expectations: the "major conclusions" are not viable; they do not have a solid objective base; and the research does not meet the requirements of a scientific demarche.

These conclusions have nonetheless been received by certain Haitian leaders as being scientific. How do we prevent their being used blindly under such circumstances as pieces in support of important political decisions?

Let us hope that these critical notes will help prevent this!

The scientific flaws are detectable in all phases of the polling process: confusion as to the purpose of the research to be undertaken, scanty methodology, non-representative singular sample, deficient research tools, conflicting answers of polled individuals, inaccurate treatment and arbitrary interpretation of data, erroneous conclusions and unjustified recommendations.

1) Confusion at the level of the research itself
The goal of the research (the opinions of politicized Haitians imbued with a civic sense) and the breadth of the questionnaire (35 questions) require a complex sociological research and not just a simple poll of opinion. In fact, the category "politicized Haitians imbued with a civic sense" is not an empirical category that is readily obvious, as are categories such as men/women, single/married, peasant/city dweller. It is a sociological construct that can be identified as the result of an empirical inquiry framed by a theoretical approach. This is not the case with IRI's survey which assumes as a given, in naive scientific manner, that "politicized Haitians imbued with a civic sense" are concentrated in "popular organizations; associations of neighborhoods, young people, and women; cultural and development associations; local elected people and parliamentarians; and political parties.

2) Scanty methodology
IRI does not define the characteristics and criteria used to identify the "politicized Haitians imbued with a civic sense." It is silent as to the procedure used to select the polled sampled, on the protocol used during interviews and the language utilized in the context of different socio-linguistic situations, on the techniques relied upon to process the data and the method used to analyze the results. The methodology used, nonetheless, does not fail to impact the final results.

3) Non-representative singular sample
It is evident that the sample chosen by IRI, which is outside any proper research approach and in violation of proportional sampling techniques, is a singular sample, non-representative of the population of "politicized Haitians imbued with a civic sense" much less of the Haitian population at large. IRI itself recognizes that "the profile of the poll's participants is different of that of the population at large in different ways." But, in an illogical manner, IRI generalizes unduly singular data when it concludes that "politicized Haitians imbued with a civic sense have identified important stages to be implemented in order to improve the electoral process" and that "the implicitly negative perception of elected officials implies that the population is rather skeptical as to the commitment of such officials vis-à-vis the population."

4) Deficient research tools
The questionnaire is not structured in order to satisfy a proper research approach. It is made up of a series of simple questions which can only be answered "yes" or "no." Within this limited context, certain questions are biased, influence the answers, as they are phrased not with the purpose of collecting an individual opinion but rather to obtain approval for a pre-determined point of view (2.6, 2.7, 3.1, 3.2, 4.8, 4.9.) Others are written in rather ambiguous manner (3.4, 3.5, 4.7, 4.10.) Some are simply irrelevant (4.3, 4.4, 4.5,) for the simple reason that the choices offered are contrary to constitutional provisions or impossible, as with 2.2B, which asks the participants if they voted in the elections of 1987 which were at the outset brought to a halt in a blood bath.

5) Incoherent answers by polled individuals
54 % of interviewees answered in curious manner that they voted in the elections of 1987 which did not take place. Answers to questions 4.3, 4.4, 4.5 are incoherent for being mutually exclusive. Of the interviewees, "61 % think that CEP (Provisional Electoral Council in French) members must be chosen from organizations of civil society; 48 % deem that [this choice] must be left to the discretion of the three branches of the state(the Executive, Legislature, and Judiciary;) 26 % believe that only the political parties must set up the CEP," for a total of 135% using elementary arithmetic. But, IRI reports this rather bizarre result without soul searching. The same weirdness is found in questions 4.9 and 4.10 with mutually exclusive answers: "55 % believe that the international community should observe the elections while 64% deem that civil society should do it, " for a total of 119%.

6) Incorrect treatment and arbitrary interpretation of data
Of this, I will give only one example. To the question 2.5 "Will you vote in the next elections to be organized in Haiti?," 64% say "yes." Since IRI does not give information on non-respondents, we can only conclude that the other 36% answered "no." The next question asks of the 36% who indicated their intention not to vote to give the reasons for their decision from a list of 7 pre-chosen reasons. 58% picked the poor performance of elected officials, 55% predictable electoral fraud, 54% fear of violence, 54% corruption, 43% the exclusion of political parties, and so on -- Let us note in passing the frivolity of that last alleged reason: how can one seriously contemplate the exclusion of political parties in elections to come without delving into science-fiction? -- But let us get to the treatment and interpretation that IRI gives of this statistical data. "Small majorities have announced, however, that they will not participate in coming due to potential fraud, corruption, and the fear of violence." It is clear that IRI is royally mistaken for it does not understand that the 58%, 55%, 54%, 43% under consideration are but fractions of the 36% who will not vote and not of the 100% of the persons polled. With reference to the total population polled, we are talking about minorities: 20.9% 19.8%, 19.4%, and 15.5% respectively.

7) Erroneous conclusions
IRI's conclusion is thus erroneous when it deems that "the respondents to the poll gave contradictory answers to the question of security in the context of the coming elections. A strong majority (67%) deem that candidates will be able to campaign freely and will be safe (52%.) 54%, however, mention fear of violence as the reason for not participating in future elections. No, there is no contradiction because the 54% are but a fraction of the 36% who stated that they will not vote in the next elections; thus, they represent only 19.4% of all respondents. Contrary to IRI's conclusion, there is no contradiction between the following two results: "75% deem that the participation of all parties is a necessary condition for the coming elections and 43% declare that the exclusion of political parties is one of the reasons for not participating in the electoral process." Again, we are talking about 43% of the 36% who stated that they will not vote; thus, they represent 15.5 of the polled population. 

What is there to say to conclude these critical notes about IRI's poll. It is clear that the information resulting from this work is not viable because the survey itself was not implemented according to the rules of scientific polling. When subjected to a test as to its quality, the so-called scientific report loses all substance.

Franklin Midy, Département of Sociologie
Université du Québec à Montréal