POLLS VARY WIDELY BUT MOST SHOW SLIGHT LEAD FOR

C O N F I D E N T I A L SAN SALVADOR 000210 SIPDIS E.O. 12958: DECL: 03/09/2019 TAGS: PGOV KPLS KDEM ES SUBJECT: POLLS VARY WIDELY BUT MOST SHOW SLIGHT LEAD FOR FMLN Classified By: Robert I. Blau, Charge d'Affaires, for reasons 1.4 (b) and (d) ¶1. (C) Summary: The majority of polls released before the 15-day ban on poll publication prior to the March 15 election show (left-wing) Farabundo Marti National Liberation Front (FMLN) presidential candidate Mauricio Funes leading (conservative, pro-U.S.) Nationalist Republican Alliance (ARENA) candidate Rodrigo Avila in a close race. Polls conducted in El Salvador are of varying quality and methodological soundness, and some firms have political affiliations that may influence their results. Poloffs met four polling firms and observed methodological differences and some unsound practices. The firms gave varied explanations for the failure of many polling firms to predict the outcome of San Salvador's January 18 municipal race, but the number of voters bused in from outside San Salvador, mentioned by several firms, may have been a factor. End Summary. -------------- MOST POLLS SHOW FUNES WITH NARROW LEAD -------------- ¶2. (C) Media outlets released a flurry of polls before El Salvador entered the 15-day period prior to the March 15 presidential election, during which, by law, polls cannot be published. Most data show FMLN presidential candidate Mauricio Funes slightly ahead in a close race. The majority of the polls reveal a narrow gap between the two candidates, and several show a statistical tie (the percentage point difference between the candidates is smaller than the margin of error). Two polls indicate that ARENA presidential candidate Rodrigo Avila is ahead, and a few outliers show Funes maintaining a double-digit lead over Avila. A sizeable amount of the electorate (approximately 12 to 20 percent) remains undecided; the withdrawal of the (center-right) Christian Democratic Party (PDC) and the (right-wing, opportunistic) National Conciliation Party (PCN) from the race appears to have given ARENA a moderate boost. Firm Dates ARENA FMLN Diff. Margin of Error Borge 2/16-22 40.9 40.0 0.9 2.8 (stat. tie) CID-Gallup 2/28-3/2 38.3 43.6 5.3 3.0 -------------- CIOPS 2/20-22 48.9 50.5 1.6 2.8 (stat.tie) COP 2/11-15 29.6 37.5 7.9 2.0-2.5 -------------- CS-Sondea 2/5-13 30.1 41.3 11.2 unknown -------------- Jabes 2/20-23 42.5 39.0 3.5 2.5 -------------- LPG Datos 2/20-23 28.0 30.9 2.9 2.2 -------------- Meba 2/20-22 41.5 44.5 3.0 2.9 -------------- Mitofsky Feb 40.1 40.8 0.7 unkn (stat. tie) UCA-IUDOP 2/1-15 31.7 49.3 17.6 1.95 -------------- Vox-Latina 2/8-13 28.5 48.6 20.1 3.7 -------------- Note: Leads outside of the margin of error are underlined. -------------- POLLS' RELIABILITY AND METHODOLOGY DIFFER -------------- ¶3. (C) The reliability and methodology used by the firms conducting polls in El Salvador varies widely, which poloffs confirmed when meeting with three universities and a newspaper that conduct polls. The wide variation in the polls for the March 15 election can be largely explained by differences in methodology and by the dates spanned by the polling. Meetings with La Prensa Grafica's LPG Datos, Francisco Gavidia's Center of Public Opinion (COP),the Technological University's Center of Public Opinion of the Salvadoran Population (CIOPS),and the Central American University's Institute of Public Opinion (IUDOP) revealed that polling firms construct their samples differently, use different wording and order in their questions, and employ different methods (some of which did not appear to be statistically sound) to reach out to respondents. In addition, the date of polls conducted in February is important; polls conducted following the withdrawal of the PDC and PCN and the decisions of the leaders of these parties to endorse ARENA show a modest increase in Avila's numbers. -------------- --- POLLING FIRMS' REPUTATIONS AND AFFILIATIONS VARY -------------- --- ¶4. (C) Many of the firms operating in El Salvador are contracted by newspapers, universities, and the political parties, which affects how their results are presented and, potentially, the firms' independence. For example, both presidential candidates claim that their internal polls show them ahead. CS-Sondea, which shows a double-digit lead for Funes, is owned by a consultant of Funes', and conducts polls for the candidate. Vox Latina, which also shows a double-digit Funes lead, is a Guatemalan firm contracted by the FMLN. Mitofsky, which shows a tie, works with Telecorporacion Salvadorena (TCS),a right-leaning entity. Borge and Asociados, which has a poor reputation regionally, and Jabes Market Research conducted their recent polls showing Avila slightly up on behalf of (right-wing) newspaper El Diario de Hoy. Double-digit leads for Funes in the university polls may be the result of poor methodology and, possibly, left-leaning slant. On the other hand, CID-Gallup, which shows Funes up by five points, is a prestigious and reputable polling firm in Latin America. CID-Gallup released poll numbers prior to El Salvador's 2004 presidential elections that were very close to the actual electoral results. -------------- - LPG DATOS' CANDID TAKE ON DIFFERENCES IN POLLS -------------- - ¶5. (C) Edwin Segura of LPG Datos, which appeared to be the most competent and responsible polling firm out of the four, was frank about the difficulties in carrying out polls in El Salvador and explained common methodological discrepancies. Segura said that the poor can be difficult to access because they live in dangerous areas, while the well-off are also hard to access because they live in walled-off residences. He lamented the difficulty in getting accurate demographic data in El Salvador to use for building polling samples and asserted that Salvadoran institutions often release incorrect statistics. Segura pointed out that several polling firms are not working from the most recent 2007 census, and that the firms use different rural-to-urban and male-to-female proportions in their samples. For example, he said that Borge and Mitofsky use the 1992 census proportion of rural-to-urban populations for their samples, while others use the 2007 census figures, and LPG Datos uses its own adjustment to the rural-to-urban proportion from the 2007 census. Differences in rural-to-urban and male-to-female proportions could cause the results to favor one party over the other; according to some pollsters, ARENA draws support from the rural poor and housewives, while FMLN voters tend to be urban and male. -------------- POLLING FIRMS DESCRIBE QUESTIONABLE PRACTICES -------------- ¶6. (C) All of the firms admitted to some practices that appeared methodologically unsound during the meetings, but the universities accounted for the worst practices. Francisco Armando Zepeda of CIOPS explained that the firm attempted to reach wealthy voters by conducting polls at places frequented by the well-off like commercial centers, supermarkets, and gas stations; this method of sampling is not random nor an accurate measure of income level. Zepeda also described how if his workers came to a house displaying party propaganda that they would throw it out of the sample, again creating a sample that is no longer random. Jeannette Aguilar of IUDOP appeared to lack confidence in IUDOP's poll and confided that, although the poll shows a double-digit lead, she thought the race would be close. -------------- -------------- EXTERIOR VOTERS MAY HAVE THROWN SAN SALVADOR POLLS -------------- -------------- ¶7. (C) Several firms attributed the failure of at least six polling firms to accurately predict the outcome of San Salvador,s mayoral race to votes from individuals residing outside of the capital. (Salvadorans vote by place of registration, not by their current place of residence). Segura said the disparity between the polls and the results suggested some type of fraud or voter manipulation, but he questioned the validity of the polling from IUDOP and CIOPS, which was off by double digits. Zepeda admitted that CIOPS probably made a mistake in their sample and that some voters changed their minds but asserted that voters from outside of San Salvador also were a factor. Elner Crespin of COP pointed out that their polling matched results in key municipal races besides San Salvador. He said that survey respondents attributed votes from outside San Salvador, the ARENA candidate,s perceived competence, and the ARENA metrobus proposal as top factors in the San Salvador outcome. ¶8. (C) Comment: Although Funes appears to have the edge in credible poll numbers (CID-Gallup, LPG Datos),the race is too close to call. Given the disparities in polling methodology and the apparent ideological bias of many of the local firms, we will only be able to judge the accuracy of the polling after the March 15 election. Voter migration within El Salvador may have thrown off the January 18 polling data; voters returning home to vote on March 15 (mostly from the U.S.) could similarly throw off the presidential election polling. End Comment. BLAU