With less than one month until the general election it was to be expected that at least one further round of national polls would have been conducted/released since my previous (and third) piece for The Elephant, leaving me with two more pieces to offer: one before and one after August 9. The five main developments and issues that are covered here are: a summary/comparison of the results of the three surveys relating to the presidential contest; consideration of respondents claiming to be “undecided” about their presidential voting intentions; some challenges in weighing the impact of deputy presidential running mates as well as of “the gender factor”; issues related to the interpretation of county-level data and media performance in this regard; a comparison of the sub-national regions/zones now used by the three mainstream pollsters, as well as the implication of such areal categories for consideration of the “Elephant in the room”: the ethnic factor in Kenya’s electoral politics.
Comparison of the three polls: third horse entry?
Each of the three “mainstream pollsters” that this series has been tracking – TIFA Research, Infotrak and Radio Africa – released results during the week of 10 July. Though TIFA had conducted its survey during the last week of June, it delayed release of its data to give a client, the Standard newspaper, enough time to publish findings from the issue-based questions that it had loaded onto the questionnaire. These questions constituted about one-third of the survey’s total content, thus in declaring “the sponsor” of the poll as required by the Publication of Electoral Polls Act of 2012, TIFA was not obliged to identify The Standard Group, although the Standard’s own article reporting and analysing the findings clearly stated that it had engaged TIFA for the purpose.
As the table below shows, the main change since the previous round of surveys is the emergence of a third candidate – George Wajackoyah – whose initial ratings suggest the profound impact he could have on the “two-horse” race by denying both of them the required “50 per cent + 1” to avoid a second round run-off contest.
Before offering a few points about the figures – and without casting any personal aspersions on Wajackoyah – the various and non-mutually exclusive possible motivations for such “minor candidate” ballot appearances for all elective positions in Kenya, and in no particular ranking order of importance or actual occurrence, may be suggested: to build one’s public profile for a possible future run or for shorter term business or professional benefits, or even just for social purposes, or with the aim of obtaining an appointed position by the eventual winner or winning political party (at either the county or national level); to draw votes away from one or more of the leading candidates, whether self-motivated or instigated/supported by others; and to promote a particular policy agenda, whether individual or on behalf of some issue-based party or lobby group (e. g., environmental protection).
Clearly, in a contest such as Kenya’s presidential one where a second round, run-off, contest must be held if no candidate initially garners enough votes spread out over enough of the counties to achieve a win, the bargaining power of any candidate who can take “credit” for this outcome, increases significantly. (Although an alternative outcome is for any such candidate(s) to be “enticed” to stand down – at least through a public declaration if it is too late to have any names removed from the ballot paper – for whatever motivation or benefit.)
|Sample Size/Margin of Error
|Data Collection Dates (2022)
|1,308 / +/-2.7%
|3,000 / +/-1.8%
|3 Survey Average
* The TIFA survey sample was 1,533, but after removing those who declared that they are not registered voters, as well as those who said that they were but would “definitely not vote” on August 9, the sample decreased to this figure.
** Information on the sample size, the (correct) margin of error(not +/-0.8 per cent as reported by the Star), and the mode of data collection (not indicated by the Star, and not by SMS “invite” as in Radio Africa’s three previous surveys but by “ordinary” CATI) and dates (not 7-11 July as reported by the Star) was obtained from a senior Radio Africa editor, but who also (incorrectly) stated that the margin of error is +/-1 per cent). Radio Africa also announced (in the Star) that they will be conducting “weekly polls” between now and the election.
A few comments about these three polls should be added. First, looking at the error margins of the TIFA and Infotrak results provides a useful lesson about random national surveys: that even massively increasing sample size (with the accompanying cost) adds little value in terms of the (national only) results. Specifically, even with a sample size of nearly nine times that of TIFA’s, Infotrak’s results fall within the margins of error of the two surveys (as shown, +/-2.7 per cent for TIFA’s and only +/- 1 per cent for Infotrak – and this is so even if there was a full week (and more) difference in the data collection dates).
Another point is that even if all three surveys were conducted by CATI, the Radio Africa poll once again reports a far lower proportion failing to answer the “which presidential candidate will you vote for” question. Why this survey has no figure for “no response” is puzzling; so, too, is Radio Africa’s reversion to a CATI methodology given that its last several surveys have been based on SMS “participation invitations”; the accompanying story offered no explanation for this. It seems improbable that this contrast (i.e., only 3 per cent “undecided” vs. 10 per cent in TIFA’s and 16 per cent in Infotrak’s) is a consequence of Radio Africa’s slightly later data collection date, although this absence of a “No Response” figure explains Radio Africa’s significantly higher figures for both of the main candidates. (I have sought an explanation from Radio Africa about this – for example, do their interviewers put any “pressure” on their respondents to “just name the candidate you think you might vote for”? I await a reply. In this regard, it is also unclear why the Daily Nation writers of the story on the Infotrak survey, in noting TIFA’s “delay” in releasing their results (on 12 July) suggested that this “could make the [TIFA] numbers obsolete in a fluid political situation”, yet – as noted above – they are statistically identical with those of Infotrak.
If Wajackoyah can maintain the level of his current popularity, he has the potential of forcing a run-off.
Whatever the case, the overall conclusions from these three recent surveys are first, that while Odinga maintains his lead over Ruto, he has not increased it over the last month – if anything, it has decreased slightly – and second, that if Wajackoyah can maintain his current level of popularity, he has the potential of forcing a run-off. This is evident if all those respondents who stated that they were as yet “undecided”, together with those who declined to answer this question at all (for whom, as noted, Radio Africa reported no figure), are removed from the calculation. Nevertheless, this reality exists even if the figures from one firm (Infotrak) put Odinga barely over the required “50 per cent + 1” threshold. (Interpreting these same figures, James Mbaka of the Star is thus in error when, after recently reporting the results shown in the above table, he asserted that “Three recent opinion polls by credible firms projected that neither Raila nor Ruto would manage to win . . . in the First Round. . .” since he evidently failed to do this adjusted arithmetic. (Whether the Treasury can afford a run-off contest is another matter.)
One example of the failure to do such basic re-calculations was provided by Brian Otieno of the Standard in suggesting that “Ruto . . . would need the entire undecided vote to swing in his favour and also some two per cent from his opponent’s – Raila or George Wajackoyah, at four per cent baskets.” Again, he failed to do the required (and quite simple) calculation.
Regardless of the likelihood of a runoff, such ratings for Wajackoyah raise the question as to just who his would-be voters are. As shown by TIFA in its media Release (of 11 July), they are most numerous (in proportionate terms) in the South Rift (8 per cent), Lower Eastern (7 per cent) and Mt. Kenya (6 per cent).
Further, and perhaps more significant, among probable voters, more than three times of those declaring an intention to vote for him are among the youngest age cohort (i.e., 18-24) as among the oldest (above 35): 7 per cent vs. 2 per cent. And recall here that such voting-intention questions were asked only to those who claimed to be registered voters, excluding those who said that they “definitely” would not vote.
More generally, and just on the basis of speculation, four (again, non-mutually exclusive) motives may constitute the basis of the support for Wajackoyah that these surveys have captured. At least among those who, on the basis of such polls (or other information), realize that he has absolutely no chance of winning, it could be an unhappy “protest” vote against the main ballot-choice of Ruto and Odinga (for whatever reasons), the “fun” of voting for an extremely “non-conformist” candidate based on whatever combination of his character and advertised policies (e.g., the legalization/promotion of the growing and marketing of marijuana, the execution of those convicted of corruption, etc.), the hope that his vote total will force a run-off contest in which he may be able to “sell” his overt support in exchange for some personal or policy presence in the next government, and/or the hope that he will be encouraged to participate in some future election (for whatever position) with a better chance of winning.
Regardless of the likelihood of a runoff, such ratings for Wajackoyah raise the question as to just who his would-be voters are.
At the same time, it is possible that such figures will not be reflected in the official results after the votes are counted, based on the fact that such survey responses were either not sincere when they were given, or that at least a significant proportion of such people will decide that votes cast for him will be ‘wasted’, and therefore force themselves to choose between the two viable candidates on August 9, especially if the polls continue to show the Odinga-Ruto race as ‘too close to call’. Time (and further survey research) will tell.
The undecideds: Who’s who and why?
Again, based on these survey figures, we have seen, as expected, that the proportion of all respondents who were unable or declined to mention a preferred presidential candidate has continued to decrease since the beginning of the year. For example, according to TIFA, it has dropped by about half, from 30 per cent in January to just 14 per cent in late June. At the same time, it cannot be assumed that all such respondents have not, in fact, made up their minds, but may be too shy to reveal their voting intentions, for one reason or another. Indeed, only with the benefit of credible official results will it be known if at least some of those declining to reveal their voting intentions have actually concealed them – similar to the significant proportions of respondents in the surveys that were “wrong” with regard to Donald Trump’s victory in the 2016 US election and in the UK’s “Brexit” vote the same year. Assuming that is the case, who between Ruto and Odinga will benefit most when the real votes are counted?
While an answer to this question must wait a bit longer, it is clear that the proportion of respondents who have declined to name a preferred presidential candidate in this electoral season remains larger than it was in the period leading up to any of the last three elections. The relevant figures taken from surveys conducted about one month before them are as follows: in 2007, 1 per cent; in 2013, 5 per cent; and in 2017, 9 per cent. It should also be noted that the surveys which yielded these figures were all conducted face-to-face at respondents’ households, in contrast to the three at issue here. It may be assumed that in the former setting, where interviewers and respondents are able to establish a more “personal” relationship, it would be more “awkward” for a respondent to avoid answering this question.
Aside from any differential impact of methodology, however, it may also be suggested that the choice of the main presidential candidates in this election is rather more complex or challenging than in any previous (multi-party) contest, in that the leading contenders have largely exchanged their political “clothing”. For his part, the deputy president is largely campaigning against his president – and thus the status quo – even if throughout their first term, there appeared to be not an iota of daylight between them. By contrast, the former prime minister finds (or has put) himself in the somewhat awkward position of trying to sit in two chairs at the same time: competing with the DP in offering credible “change” improvement for the vast majority of the electorate currently suffering a plethora of economic (among other) woes, while largely unable to attack the outgoing president to whom he owes whatever advantage the latter’s support provides. Indeed, in a TIFA survey of June 2021, fifty per cent more respondents identified Ruto rather than Odinga as “the political leader most active in terms of criticizing the Jubilee government and trying to hold it to account”, and in TIFA’s April 2022 survey, some three-quarters of respondents identified Odinga as “Uhuru’s preferred successor”. Such a situation makes it largely impossible for Odinga to assume the anti-government posture he has assumed in the last five elections, notwithstanding his short-lived absorption into Moi’s KANU government and party in 2001.
The proportion of all respondents who were unable (or declined) to mention a preferred presidential candidate has continued to decrease since the beginning of the year
Whatever reasons might be offered about such higher figures, the question remains as to just who these “undecideds” are (as well as those who simply declined to answer the question – coded as “No Response”). A cursory look at the data offers some indication. First, dividing all respondents who claimed to be registered voters into those who did vs. those who did not name any candidate, rather more of the former answered the question about their likelihood of voting by saying they “will definitely vote” (71 per cent vs. 63 percent), suggesting a somewhat greater interest among them in the election altogether. An even greater contrast is found in terms of gender, with almost three times as many women as men not naming a preferred candidate (21 per cent vs. 8 per cent). Again, whether this is due to shyness, a lower level of interest in elections, or taking longer to make this ballot choice, perhaps due to Kenya’s largely patriarchal culture in which women receive “instructions” various issues from husbands, fathers, etc., cannot be discerned. Such a contrast is nearly equally apparent in terms of education levels, as significantly fewer of those without any or just a primary education named any candidate as compared with those with secondary or higher education. Further, in ethnic terms, while at least nine out of ten Luo and Kalenjin named a candidate, the figures for nearly all other (major) groups is about 10 per cent lower, aside from those in the Mt. Kenya grouping, who are in an intermediate position, evidently based on the presence of a fellow ethnic running mate candidate on both sides of the main partisan divide.
On the other hand, no contrasts in the proportions of those who did vs. those who did not name a candidate are found in terms of political party/coalition alignment, age, and employment status. (Actually determining the relevant salience of each factor would require a complicated regression analysis that goes beyond the confines of this piece!)
In sum, it seems clear that the most frequent response of the “undecideds” in TIFA’s April survey as to what would most enable them to decide whom to vote for – “more information about policies/manifestos” – is not the whole story.
The running mate (and gender) factor
Precisely measuring the impact of running mates on campaigns is always a challenging task, in large part because many respondents may be unclear about this in their own minds, or unwilling to acknowledge it even if they are. When the respective supporters of the two main presidential candidates were asked a more general question in TIFA’s last survey, “How satisfied are you with Raila’s/Ruto’s choice of Martha Karua/Rigathi Gachagua as his deputy president running-mate? Are you…?”, there was a marked contrast in their responses, with considerably more of Odinga’s than Ruto’s supporters stating that they were “very satisfied” (90 per cent vs. 67 per cent), yet Ruto’s overall support rating rose slightly more than Odinga’s (4 per cent vs. 3 per cent compared to the previous survey, although this 1 per cent difference is within the survey’s margin of error).
(One other point: When the Standard reported these results – based on several questions they had sponsored in TIFA’s June survey, as noted above – the story’s caption was: “Poll: Karua will net more votes for Raila than Rigathi for Ruto”, yet as Nzau Musau explained in his first sentence, this conclusion was derived from a perception question, not an analysis of the actual candidates’ ratings/change of fortunes since TIFA’s previous survey. That is, whereas 49 per cent stated that Karua will add to Raila’s vote total, only 30 per cent felt likewise about Ruto’s choice of Gachagua, with another 21 per cent not certain as to which running-mate will bring along most votes.)
In ethnic terms, while at least nine out of ten Luo and Kalenjin named a candidate, the figures for all other (major) groups are about 10 per cent lower.
Moreover, with specific regard to the Karua/gender factor – and again, notwithstanding the perception that she is considerably more useful in terms of adding votes to Odinga than Gachagua is to Ruto – the rise in their respective ratings is (as noted) statistically identical. Further, Odinga suffers from a significantly greater “gender gap” (i.e., male vs. female) than does Ruto (47-37 per cent for the former but only 40-38 percent for the latter). Indeed, this 10 per cent gap for Odinga is exactly what it was at the end of April (37-27 per cent) before running mates were announced. At the same, it may be the case that her presence on the Azimio ticket will encourage higher voter turnout among women on August 9 (whether to vote for her and Odinga, or across the board), but the main point is that whatever Karua is contributing to Odinga’s electoral prospects, there does not appear to be any “gender” advantage – so far, at least.
County-level data and ‘battlegrounds’
Reporting the latest Infotrak poll, Daily Nation writers Collins Omulo and Onyango K’Onyango began by referring to “ten crucial counties with a total of 3.6 million votes” that “recent opinion polls have classified as battlegrounds, where the vote could go either way.”
It should first be noted that only Infotrak (once again) used this term in releasing its results, without, however, giving it any numerical definition. (In June the number of respondents from each county was reported, but not for July; since the sample size in both surveys was identical, it can be assumed that the county numbers are the same.) For example, these writers stated that: “However, Mombasa with 641,913 voters and Tana River with 143,096 voters are now battlegrounds with Mr Odinga’s popularity in Mombasa at 46 per cent, Dr Ruto at 27 per cent and 20 per cent being undecided voters.” In other words, in Mombasa Odinga enjoys a 19 per cent margin. Yet with some 270 respondents drawn from this county, the margin of error is +/-6 per cent, equal to a 12 per cent spread – giving Odinga a clear lead of (at least) 7 per cent. So how much larger would Odinga’s lead have to be for Infotrak to classify Mombasa as among his “strongholds”? We have no idea.
In any case, such a statement is misleading in three senses. First, since Kenya does not have a US-type electoral college system, counties are not electoral units that are “won” or “lost”; the only thing that matters is how the votes are distributed between the candidates across the entire country in their efforts to attain the “50 per cent + 1” threshold. (This statement assumes that neither of the two main candidates will have any difficulty in obtaining at least 25 per cent of the vote in at least 24 of the 47 counties – which all recent polls suggest is certain to be the case.) In the Kenyan context, therefore, “winning” a sparsely populated county (in terms of registered voters who actually turn out to vote on election day) such as Lamu or Marsabit by netting a few more votes than one’s main opponent is not nearly as critical as “losing” a highly populated county such as Nairobi or Nakuru by simply increasing one’s share of the vote there by a few per cent.
Second, and more egregiously (as suggested above), the Daily Nation’s writers fail to interrogate the statistical basis of Infotrak’s lists of counties in the “grip” of either Odinga or Ruto: 21 in that of the former and 16 in that of the latter. Specifically, there is no reference to any definition of this term; presumably, some stated margin between the two candidates’ ratings in each county. To repeat the point from my previous Elephant piece, before accepting Infotrak’s “stronghold” lists, it is necessary to calculate the margin of error for each of these counties. For example, Garissa, with a Ruto-Odinga gap of 22 per cent (based on figures of 50 per cent vs. 28, respectively), is included among Ruto’s “strongholds”. Yet with a registered voter population of about 165,000 and an allocated sample of about 85, the resultant margin of error is +/-11 per cent, equal to a 22 per cent spread – exactly the difference between them. Should that earn Garissa the “stronghold” label in the Ruto list?
In other words, while a national sample of 9,000 looks impressive, when divided (proportionally) into 47 counties, the resultant margins of error require attention.
This Nation piece further reports the Infotrak CEO as stating that: “We have seen a complete flip in Lamu and Kwale”, referring to an increase for Odinga from June to July of 26 per cent in the former county and of 19 per cent in the latter. Yet the margins of error for these counties are +/-17 per cent in the former (for a 34 per cent spread) and nearly +/-8.5 per cent in the latter (for a 17 per cent spread), meaning that the change in the figure for Lamu falls within this county’s margin of error and that of the latter only just outside its margin of error (i.e., 2 per cent), perhaps not qualifying for the description of “a complete flip”.
While it does not employ any vote-support categories (such as “battlegrounds”), TIFA likewise could be more explicit about the error margins of the nine zones for which it provides sub-national results. The most extreme case is that of South Rift (which, as shown, is comprised of just two counties: Kajiado and Narok (see below). Constituting only 5 per cent of TIFA’s total sample (in this most recent survey of 1,533 – but as shown, only 1,442, having removed those who stated that they are not registered voters, and then leaving only 1,308, having also removed those who state that they “will definitely not vote”), this amounts to only (“likely to vote”) 65 respondents. Based on a total registered voter population of about 1,270,000, the margin of error is +/-12 per cent, equal to a 24 per cent spread. Keeping this in mind, neither the increase for Odinga-Karua by 17 per cent, nor even the decrease for Ruto-Gachagua by 33 per cent looks quite so dramatic (since the former’s gain could really be just 5 per cent, although the latter’s loss remains a hefty 21 per cent).
Similarly (in the same Nation article), Infotrak reports that over the last month, Ruto’s popularity (i.e., the expressed intention to vote for him) “jumped” from 52 per cent to 55 per cent in Mt Kenya while “Mr Odinga’s approval [sic] currently stands at 24 per cent from 27 per cent in June.” In other words, even if Odinga gained so much in the (sparsely populated) South Rift so as to overtake Ruto by 10 per cent there, the 3 per cent gain in Mt. Kenya, combined with Odinga’s decline of the same amount, gives the DP a far more (potential) vote boost, given the vastly greater population of registered voters in the latter zone.
Since Kenya does not have a US-type electoral college system, counties are not electoral units that are “won” or “lost”.
As for Radio Africa, the report of their most recent survey offered correlations of preferred presidential candidate with (reported, presumably monthly) income. In doing so, the second category shown (after “no income”) is Shs1-30,000/-, which surely must include at least half of the sample. Yet they then use five additional more affluent categories, the highest being “above Shs150,000/-“ which, based on data from the last few years of TIFA surveys, could not have included more than a handful of respondents, if that. For example, in TIFA’s most recent poll, only 4 per cent of respondents reported earning more than Shs50,000/- per month, yet Radio Africa presents results for four high income categories beginning with Shs50,000/- to Shs70,000. Based on a sample of 3,000, that would be equivalent to about 120 respondents, for whom the margin of error (if all those with reported monthly earnings above Shs50,000/- were lumped together) is +/- 9 per cent – equal to an 18 per cent spread. In other words, even if Radio Africa did not display the margin of error for each income category, they should have shared the number of respondents in each one with their readers and let them judge what, if any, statistical integrity such correlations have.
Perhaps the overall point is that even if there is no agreed minimum number of respondents among even ‘credible’ survey firms for which such sub-total results should be presented, whether such categories are income, regions, or any other variable, there should be more transparency about such sub-national error-margins.
Zone comparisons and the (other) elephant in the room
Radio Africa has now jumped on the TIFA “bandwagon” by adjusting its previous sub-national categories as Infotrak began to do, starting with its most recent previous survey (although they did not include a chart for these as they have done in the past; TIFA always includes a list of its nine zones, listing the counties in each one). All three firms have thus now moved away from using the eight pre-2010 provinces for this purpose. The table below shows the sub-national units each one used in releasing their most recent survey data reported above.
|Regions / Zones (TIFA)
*These regions/zones were provinces in the pre-2010 Constitution era.
**TIFA includes only Kajiado and Narok in this zone while for Infotrak it also includes Kericho and Bomet. While this makes geographic sense, TIFA prefers to place all the main Kalenjin areas in Central Rift.
Although, as shown, the regions (or in TIFA’s terminology, “zones”) differ slightly across the three firms, they nevertheless allow for some comparisons at this sub-national level, even if none of them includes the margin of error for each one, an omission which helps to explain a certain amount of erroneous interpretation by journalists in asserting that one candidate or another has “gained” or “lost” votes in a particular region when the change actually falls within that region’s margin of error, which is by necessity much greater than for the national sample as a whole, as also discussed above. At least Infotrak and TIFA always show the percentage of the total sample that was drawn from each region/zone, so that, knowing the total sample size, it is possible to take a margin of error table and a calculator and do the “math” to ascertain these.
Presenting survey results at this sub-national level raises a question rarely asked by local journalists (or others), even if it seems that many are thinking about: To what extent can these units be considered as “substitutes” for at least the main ethnic group resident within each one?
This question arises simply because no survey firm releases results with ethnic correlations, for the (perhaps obvious) reason that none of them (nor any media house) would want to be accused of “dividing Kenyans”, let alone “threatening national unity”, even when – as is certainly the case in this pre-election season – the data reality shows that Kenyans are much less polarized along ethnic lines than many assume. (Let me also note here that several attempts over recent years to obtain public policy “guidance” on this issue from the National Cohesion and Integration Commission yielded no “edible” fruit, notwithstanding the apparent interest they displayed in the figures that were shared with them.)
For example, it was found (in a June 2021 TIFA survey) that only 40 per cent of Kenyans answered the question, “Is there anyone who you consider to be the main leader of your ethnic community?” in the affirmative. True, this national figure rose as this year’s election approached (in TIFA’s June 2022 survey) to 54 per cent – clear evidence that like the proverbial “hangman’s noose”, elections tend to concentrate communal minds, but this seems far below what most people consider to be the case. And this defiance of “common knowledge” holds true even if the specific figures are as high as two-thirds among the Luo, Kalenjin and Kamba and below 50 per cent for the Kikuyu and Gusii. Also significantly, among those who believe their community has such a leader, there is far from unanimity as to who that leader is, even for the two communities with “serious” presidential candidates: the Luo and the Kalenjin. (The lower figures for the Kikuyu and Gusii are clearly in part a reflection of the fact that neither has a serious presidential candidate in this election, while the former has two deputy presidential candidates and a president about to retire.)
While thoughtful people may reasonably disagree about what the impact of releasing such figures would be, given such widespread assumptions about their salience in electoral choices, it is clear that much analytical capacity – and thus public understanding – is lost by “hiding” them (even when it is clear that the major campaign teams make considerable use of such data in crafting and implementing their vote-hunting and turnout strategies).
In the absence of such ethnic correlations in publicly released findings, the public is left with the regional correlations that the main survey firms almost always include.
The following table (based on TIFA’s June survey data) shows the largest (and where included, also the second largest) ethnic group in each region.
|Predominant Ethnic Group(s)
|Kikuyu / Meru
|60 / 20
|Somali / Turkana
|35 / 20
|Kalenjin / Kikuyu
|65 / 20
|Luo / Gusii
|50 / 25
It is clear, therefore, that while one or another ethnic group predominates in most of these zones, there remains considerable heterogeneity in most of them.
Moving from ‘what’ to ‘why’, and other Issues
Given the reality (described above) that not a single ethnic group is homogenous in terms of its presidential voting intentions, the question arises as to what accounts for these intra-ethnic divisions. For example, within a (largely) ethnically homogenous area such as Mukurwe-ini in Nyeri or Kilungu in Makueni, what factors explain why some people will vote for Ruto and others for Odinga? At this stage, what should be clear is that even beginning to answer this question requires not assumed generalizations but detailed research, and of a nature that would best include and also go beyond quantitative surveys.
Another issue not considered here is the so-called “bandwagon” effect: that candidates or parties shown to be leading in polls will thereby attract more votes, based on the assumption that many people want to be on “the winning side”. For now, it is enough to say that it is widely assumed to exist, and at a significant level. If not, why would we see candidates and other partisans so vociferously bashing results that do not show them leading, as well as sponsoring “fake” polls – sometimes by “unknown” survey firms, and at other times attributing results to credible firms that had nothing to do with them. The non-profit research organization, Code for Africa, recently reported that it has been identifying six to seven “fake” polls per week over the recent past – which they define as attributing survey results to firms that did not conduct them. What is clear is that candidates find it difficult to remain silent when a credible survey firm shows them trailing, or even just decreasing in popularity. Just how the impact of such “fake” – as well as genuine – polls might be measured will be taken up in my next piece.
In the meantime, with less than three weeks remaining before the 5-day embargo period prescribed in the Publication of Electoral Polls Act kicks in, and with all the mainstream pollsters either having begun or about to launch their final (or nearly final) round of surveys, there is certain to be plenty more material to present and discuss before “D-Day” on August 9.