Three Known Unknowns Ahead of Kenya’s Presidential Election Tomorrow

Here are three factors that could potential lead to a surprise outcome in Kenya’s presidential election tomorrow.

1. Turnout: To be honest, we don’t really know what to expect with regard to turnout figures tomorrow. As Charles Hornsby notes here:

For some time I have been wrestling with an ethical problem. Reviewing the 2013 turnouts, in comparison with that from previous national elections since 2002, it became clear with the benefit of hindsight that turnouts were implausibly high not just in Luo Nyanza and Central Province, but in many other places…….. The size and scale gap between 2013 and every other election for the past 15 years is hard to explain.

In other words, there is good reason to believe that the 2013 turnout figures were artificially high.

The use of technology and a raft of reforms, including the planned announcement of results at the constituency-level, are supposed to minimize opportunities for political parties to gin up turnout in friendly areas this time round. If these safeguards work, we will all be flying blind with regard to our turnout-based predictions of the final outcome in the presidential election. All I can say for now is that Kenyatta has a structural advantage over Odinga on likely turnout, but that if Odinga matches Kenyatta on turnout (especially in Coast and Western regions) he will most likely become Kenya’s fifth president.

2. Technology: The Kenyan public’s perception of the integrity of the polls will crucially hinge on how well the IEBC’s Kenya Integrated Electoral Management System (KIEMS) is judged to work. Kenya is holding one of the most expensive elections in the world (judged by cost per vote). And the use of expensive technology has been touted as a way of minimizing human discretion and, therefore, opportunities for vote rigging. But technology only works as well as humans allow it to. And sometimes stuff hits the fan. In addition, the IEBC only partially tested its results transmission system. Which means that there will most likely be hiccups in the data transmission process from the country’s 40,883 polling stations. Widespread technological failures will tarnish the integrity of the outcome, and could be a catalyst for political instability. 

3. The accuracy of opinion polls: Throughout this cycle, all polls except one have shown Kenyatta to be ahead of Odinga. In the recent past the polls have certainly tightened, with Kenyatta leading Odinga by 3 percentage points or less. But what if the polls are wrong? In 2013 they underestimated Kenyatta’s support by about 2 percentage points, on average. The same inherent bias may be at play this year, or it could be reversed. One important factor to look out for will be the effect of local gubernatorial races on the presidential race. For instance, the outcomes of local contests in counties like Machakos, Bomet, Narok, Meru, and Bungoma will likely have non-trivial effects on the presidential race in the same counties. Yet throughout this cycle there has been very little effort to calibrate the national polls using information from the state of county contests. And so while I believe that existing opinion polls give a fairly accurate depiction of the state of the race, only the final vote counts will tell.

 

 

 

 

Was the IEBC’s distribution of BVR kits for mass voter registration fair?

In 2013 a number of pundits declared that the Kenyan election was effectively decided on the day the IEBC ended its mass voter registration exercise (It was not, turnout won the election for Kenyatta). As a result, Kenyatta’s campaign team came up with the narrative that victory was inevitable on account of the “tyranny of numbers”. This year Odinga’s campaign has adopted a similar tactic with its claim of heading of movement of “ten million” voters. The total number of registered voters in Kenya is just over 19m.

The idea that the elections can be won at the close of registration brings to mind the neutrality of the IEBC when deciding how to allocate finite resources for mass voter registration. To assess this I looked at the distribution of biometric voter registration (BVR) kits relative to a number of factors, including Kenyatta’s vote share in 2013, county area, county population, population per electoral units (wards and constituencies), and the number of electoral units.

Here are the summaries:

1.  Being an incumbent, it is conceivable that Kenyatta would have wanted to influence the allocation of BVR kits. But there is no obvious relationship between Kenyatta’s county-level vote share in 2013 and the distribution of BVR kits ahead of the 2017 election. The pro-Kenyatta counties of Nakuru, Kiambu and Meru that received a lot of kits also have large populations. If anything, it appears that pro-Odinga counties got more kits, perhaps a reflection of the fact they had relatively more unregistered potential voters after 2013.countykits2. The number electoral units (wards or constituencies) had no influence on the rate of voter registration in the 47 counties. In other words, it is not the case that counties which had a lot more electoral units (and therefore potential candidates) experienced greater rates of voter mobilization for registration.

elecunits3. However, the population per electoral unit (wards and constituencies) was negatively correlated with the registration rate. In other words, more populous wards and constituencies experienced lower registration rates relative to their less populous counterparts. This makes sense, to the extent that the IEBC was targeting a specific number of BVR kits per electoral unit per county.populations4. Bigger counties with relatively smaller populations benefited from the fact that land area was a consideration in IEBC’s allocation of BVR kits.
kits

Kajiado and Vihiga counties beg explanation.

Kajiado registered more than 100% of its projected adult population (based on the 2009 census). This may be a case of massive in-migration after the 2009 census or the deliberate importation of voters for the specific purpose of influencing the outcome of intra-county elections.

Vihiga, on the other hand, stands out for its poor registration rate. It is noteworthy that Vihiga is home to Musalia Mudavadi who came third in 2013 and is now part of the NASA coalition led by Raila Odinga. Given that the mass registration exercise ended well before it was clear that Mudavadi was not running for president, the low registration rates in Vihiga raise questions about his ability to turnout the vote come August 8th.

Finally, while it is hard to discern what happened within counties — the effort of IEBC agents is unobservable — it is fair to say that political considerations do not appear to have influenced the number of BVR kits deployed to the counties for the mass voter registration exercise.

Uhuru Kenyatta ahead of Raila Odinga in the first post-nomination poll

Ipsos just released a poll in which President Uhuru Kenyatta leads Hon. Raila Odinga 48-42% among a representative sample of voting age adults. While this is not a particularly good showing for an incumbent with a few achievements to tout, the poll confirms Kenyatta’s frontrunner status. Furthermore, a 6 percentage point lead combined with his structural advantage in the turnout game mean that if the polls do not narrow any further Kenyatta will likely win in the first round come August 8th. In 2013 most public polls consistently over-estimated Odinga’s support by about 2 percentage points by not accounting for turnout patterns.

A few things other things are worth noting from the Ipsos poll:

  1. Polls have tightened over the last few months. As Odinga consolidates the opposition, his poll numbers have converged on Kenyatta’s. This is a trend worth watching over the next few weeks.
  2. The number of undecideds, especially in Coast, Eastern and Western regions, is rather high. This should worry Odinga. Again, Kenyatta has a structural advantage in the turnout game, which means that if Odinga is to have a fighting chance he must ensure that his strongholds register both high turnout rates and give him an even bigger share of their votes than in 2013. That they remain undecided does not bode well for Odinga’s chances. In the table below, Kenyatta’s poll numbers are close to his numbers both in final poll of 2013 and the official election results. Undecideds appear to be voters who sided with Odinga in 2013.Screen Shot 2017-05-31 at 9.30.40 PM.png
  3. This poll may be over-estimating Kenyatta’s support in Western region. In 2013 Kenyatta under-performed his poll numbers in Western region by a whole 29 percentage points. And so while his 23% rating in Western region may be a sign that Deputy President William Ruto’s investments are bearing fruit, I would not take these numbers to the bank just yet.
  4. Odinga has made significant gains in the Rift Valley region since 2013. One way for Odinga to force a runoff (or eke out a squeaker of a first round win) would be to peel off enough voters in from the North Rift. He appears to be doing that. His poll numbers in the region in 2013 were spot on, making his 32% rating in the region believable for now. Recent developments also suggest that he is gaining ground in Narok, Bomet, and parts of Kajiado. That should be a source of concern for the Kenyatta team.
  5. It is still a turnout game, and Odinga is trailing. The bulk of undecideds — in Coast, Eastern, and Western regions — appear to be likely Odinga supporters. While this may mean that they are likely to break for Odinga in August, it could also mean that they will remain undecided and stay home on election day. Hassan Joho, Kalonzo Musyoka/Charity Ngilu, and Musalia Mudavadi/Moses Wetangula have their work cut out for them.

While a lot may happen between now and August 8th, it is fair to say that Kenyatta is in a strong position. Odinga has several paths to victory, but success along any of those parts is dependent on the NASA coalition running a near-perfect campaign focused on both increasing turnout and running up the score in their strongholds. To this end the lack of enthusiasm in Coast, (lower) Eastern, and Western regions is definitely not a good sign.

The one thing that should worry Kenyatta is Odinga’s apparent gains in the Rift Valley region. If Odinga gains traction in Bomet, it is conceivable that he would also be able to peel off votes in Kericho. It is not that long ago that both Deputy President William Ruto had to camp in Kericho to avoid an embarrassing loss in a by-election. His preferred candidate ended up winning with 66% of the vote. In Bomet, incumbent Governor Isaac Ruto is backing Odinga. And while he will face a tough time swaying voters to Odinga’s camp, it is not far-fetched to imagine that he could bag around 40%. If the same happens in Kericho then two of the Rift Valley’s most important vote baskets will become swing. And Kenyatta would be in serious trouble.

Note: Nearly all the polls this cycle will not take into account any “likely voter models.” I will do my best to guestimate turnout rates based on passed voting patterns and other variables. 2026 out of 5484 contacted agreed to participated in this Ipsos poll.

By the Numbers: A Look at the 2017 Presidential Election in Kenya

Kenya will hold a General Election on August 8th of this year. The national-level elections will include races for president, members of the National Assembly, and the Senate. County-level races will include those for governor (47) and members of County Assemblies (in 1450 wards).

This post kicks off the election season with a look at the presidential election. I also plan to blog about the more exciting gubernatorial races in the coming weeks.

Like in 2013, the contest will be a two-horse race between President Uhuru Kenyatta (with William Ruto as his running mate) and Hon. Raila Odinga (with Kalonzo Musyoka). The convergence on a de facto bipolar political system is a product of Kenya’s electoral law. The Constitution requires the winning presidential candidate to garner more than half of the votes cast and at least 25 percent of the votes in at least 24 counties. In 2013 Kenyatta edged out Odinga in a squeaker that was decided at the Supreme Court. Depending on how you look at it, Kenyatta crossed the constitutional threshold of 50 precent plus one required to avoid a runoff by a mere 8,632 votes (if you include spoilt ballots) or 63, 115 (if you only include valid votes cast). In affirming Kenyatta’s victory, the Supreme Court ruled in favor of the latter approach.

But despite the court’s ruling, a significant section of Kenyans still believe that Kenyatta rigged his way into office, and that Odinga should have won.

My own analysis suggests that it was a little bit more complicated. I am fairly confident that Kenyatta beat Odinga in the March 2013 election. But I do not think that he crossed the 50 percent plus one threshold required to avoid a runoff. At the same time, I am also confident that Kenyatta would have won a runoff against Odinga. All to say that I think the results in 2013 reflected the will of the people bothered to vote.

This year will be equally close, if not closer.

This is for the following reasons:

  1. Odinga has a bigger coalition: The third candidate in 2013, Musalia Mudavadi, is joining forces with Odinga this time round – as part of the National Super Alliance (NASA). Mudavadi managed to get just under 4% last time and will provide a much-needed boost to Odinga’s chances in Western Kenya and parts of the Rift Valley.
  2. Kenyatta has had a mixed record in office: The Kenyan economy has grown at more than 5% over the last four years. The same period saw massive investments in infrastructure — including a doubling of the share of the population connected to the grid and a brand new $4b railway line connecting Nairobi to the coast. However, these impressive achievements have been offset by incredible levels of corruption in government – with senior government officials caught literally carrying cash in sacks. Kenyatta is also stumbling towards August 8th plagued with bad headlines of layoffs and the ever-rising cost of living. Barely two months to the election, the country is in the middle of a food crisis occasioned by a failure to plan and a botched response that appears to have been designed to channel funds to cronies of well-connected officials.
  3. The Rift Valley: In 2013 much of the Rift Valley was a lock for Kenyatta (it is William Ruto’s political back yard). This time will be different. Parts of Ruto’s coalition in the Rift, particularly in Kericho and Bomet counties, may swing towards Odinga this year. All Odinga needs is about a third of the votes in these counties. I expect both campaigns to spend a lot of time trying to sway the small pockets of persuadable voters in these two counties.
  4. The Kenyatta Succession in 2022: In 2013 Kenyatta was elected as the head of an alliance, not a party. In 2017 he is running atop a party, the Jubilee Party (JP). It is common knowledge that JP is William Ruto’s project. Because he plans to succeed Kenyatta in 2022, he desperately needs credible commitment from Kenyatta and his allies that they will support his bid when the time comes. JP is an installment towards this goal, and is designed to allow Ruto to whip party members in line during Kenyatta’s second term (on a side note, Ruto needs to read up on the history of political parties in Kenya). But by forcing everyone into one boat, JP may actually end up suppressing turnout in key regions of the country, the last thing that Kenyatta needs in a close election.
  5. The ICC factor (or lack thereof): Because of their respective cases at the ICC, 2013 was a do or die for Kenyatta and Ruto (and their most fervent supporters). This time is different. Both politicians are no longer on trial at the ICC, and so cannot use their cases to rally voters. The lack of such a strong focal rallying point will be a test for Kenyatta’s turnout efforts.

Nearly all of the above factors sound like they favor Odinga. Yet Kenyatta is still the runaway favorite in this year’s election. And the reason for that is turnout.

As I show in the figure below, the turnout rates were uniformly high (above 80%) in nearly all of the 135/290 constituencies that Kenyatta won in 2013. Pro-Odinga constituencies had more spread, with the end result being that the candidate left a lot more votes on the table.

The same dynamics obtained at the county level (see above). In 2013 Odinga beat Kenyatta in 27 of the 47 counties. The counties that Odinga won had a total of 8,373,840 voters, compared to 5,977,056 in the 20 counties won by Kenyatta. The difference was turnout. The counties won by Odinga averaged a turnout rate of 83.3%. The comparable figure for counties won by Kenyatta was 89.7%. At the same time, where Kenyatta won, he won big — averaging 86% of the vote share. Odinga’s average vote share in the 27 counties was a mere 70%.

The same patterns may hold in 2017. The counties won by Odinga currently have 10,547,913 registered voters, compared to 7,556,609 in counties where Kenyatta prevailed. This means that Odinga still has a chance, but in order to win he will have to run up the numbers in his strongholds, while at the same time getting more of his voters to the polls. Given the 2017 registration numbers, and if the turnout and vote share patterns witnessed in 2013 were to hold this year, then Kenyatta would still win with 8,000,936 votes (51.5%) against Odinga’s 7,392,439 votes (47.6%).

The slim hypothetical margin should worry Kenyatta and his campaign team. For instance, with an 89% turnout rate and an average of 85% vote share in the 27 counties Odinga won, and holding Kenyatta’s performance constant, the NASA coalition could best Jubilee this August by garnering 8,921,050 votes (55.4%) vs 7,191,975 (44.6%).

Kenyatta is the favorite to win this August on account of incumbency and Jubilee’s turnout advantage. But it is also the case that the election will be close, and that even a small slip up — such as a 3 point swing away from Jubilee between now and August 8th — could result in an Odinga victory.

Fraud and vote patterns in Kenya’s 2013 election

Update: The video link now works. Many thanks to SAIS for fixing it and letting me know.

**********************

The video below has been making the rounds in the Kenyan online community. The Daily Nation even reported on the claims by UCSD Professor Clark Gibson and James Long, Asst. Prof. and University of Washington, that President Uhuru Kenyatta may not have crossed the 50% threshold in the March 4th election. The duo conducted an exit poll (N = 6000) on election day that showed both candidates in a statistical tie at 40.9% for Odinga and 40.6% for Kenyatta. In the presentation Clark and James make the case that exit polling is superior to PVT because it is immune to things like ballot stuffing and tallying fraud. NDI sponsored ELOG conducted a PVT that confirmed the results announced by the Kenyan EMB, the IEBC.

[youtube.com/watch?v=68a3cUrq1gI&feature=youtu.be]

I do not really know what to make of this poll finding by James and Clark at the moment. I am waiting for the actual MP and Governor elections results to be published by the IEBC so I can try and see if the results in these local races were in line with the presidential results.

Sloppy Reporting on the Kenyan Elections

Dear readers, it has been a while since I did a rant and rave post. Here is one to end the long drought. Today we look at a couple of pieces done by Al-Jazeera and the New Yorker.

First The New Yorker.

I love the New Yorker. Everyone does. Unless you are weird. Or do not like their politics.

But the New Yorker should get a better writer than James Verini on the upcoming Kenyan elections. Reporting on the first presidential debate, Verini made several unforgivable errors. I mean, I know it is hard to find information on countries that you may or may not have visited, or have only visited for a few days.

But Kenya is one of the most studied countries on the Continent. You can wikipedia or google your way to a decent article that passes a laugh test. Sadly, Verini’s does not. Here are the necessary corrections to his (original) piece (and certainly not the only ones):

  • First of all, there were two moderators, not one. Unless Verini only caught the first half of the debate. I will admit that the second half could have been better. But it wasn’t bad enough to forget that there were two moderators. Julie Gichuru moderated the second half.
  • Kenya’s first President Jomo Kenyatta was not a “Mau Mau rebel.” Also, the Mau Mau are popularly known in Kenya as freedom fighters. Dedan Kimathi, a leading light in the independence movement and Mau Mau leader, has a statue in his honor on Kimathi street in downtown Nairobi.
  • Kenya’s Prime Minister Raila Odinga is not “a human rights lawyer.” He is “an engineer by profession.” Odinga says so when he introduces himself at the beginning of the debate. What makes Verini think Odinga is a human rights lawyer? (Might be because, as he admits, he was at Njuguna’s – perhaps chasing down goat meat (nyama choma) with Tusker. The New Yorker should institute strict sobriety requirements when sourcing stories, but I digress.)
  • Nairobi is not in Central Province. Nairobi is a Province on its own. Kenya has eight provinces (now called regions) – Rift Valley, Eastern, North Eastern, Coast, Western, Central, Nyanza and Nairobi. Nairobi borders Central. But it is not in Central. I swear. You can google it.
  • The post-election violence in 2007-08 was not mainly a Kikuyu-Luo affair. Most deaths occurred in the Rift Valley in clashes between Kikuyus and Kalenjins, over land. Police brutality was number two in cause of deaths. Kikuyu-Luo clashes were horrific. But they were not the defining feature of the PEV.
  • The 2010 Constitution did not make the position of Prime Minister permanent. It abolished it. You can also google a copy of the Kenyan Constitution. There is a pdf online. I swear. Or you could just read the Wikipedia entry here.

And then Al-Jazeera:

For a news organization that claims to counter the dominance and supposed orientalist biases of CNN International and the BBC with nuanced on-the-ground reporting, this is unforgivable. Here’s is how Peter Greste opens his report on the Kenyan election:

Political science is an imprecise discipline at the best of times. But in Kenya, it feels more akin to witchcraft.

In most established democracies, astute analysts can have a reasonable stab at predicting the outcome of elections. The regular if well-spaced drum-beat of polls gives anyone who cares to look, a decent set of historical data to work with.

It’s usually possible to check the voting patterns of a particular electorate; assess the impact of demographic changes; and with the help of some intelligent opinion polling, have a good understanding of the way a country might swing.

But in Kenya, this election is stacked with so many unknown factors that a witch throwing newt’s eyes into a bubbling cauldron might have as good a chance at predicting the outcome as the political scientists.

Really Mr. Greste, really?

Witchcraft? Why that term? Why not just say that you do not have a grasp of the political reality and so don’t know how the election will turn out? Are you trying to say something about your readers (that they easily resort to witchcraft to explain things they do not comprehend) or Kenya?

I put it to you that there are three firms that have been polling the Kenyan public on their political preferences since the last election in 2007. These firms accurately predicted the outcome of the messy 2007 election (and pretty much matched the exit polls conducted by UCLA academics) and the 2010 referendum. Kenya has demographic data that politicians make very good use of. For instance, we know the ethnic composition of Nairobi, the most cosmopolitan PROVINCE (hear me, Mr. Verini) in Kenya.

Also, a few political scientists, including yours truly, have done some predictions as to the potential outcomes of the election (see blog posts below).

Why did these two do this?

To me it looks like a bad case of trying to exoticize the Kenyan elections for their audiences – what with the references to witchcraft by Mr. Greste and Mr. Verini’s over-simplification of the election to a Kikuyu-Luo tribal contest.

It is also disrespectful to Kenyans, who they seem to think will not do any fact-checking to correct their sloppiness.

Who will win the Kenyan presidential election? A look at the numbers

With elections less than a month away many in Kenya are reading the tea leaves and making predictions as to who they think will win the March 4th presidential election. Unfortunately, many of these self-styled political analysts – including the most celebrated one Mutahi Ngunyi – are merely using the raw IEBC voter registration numbers and assumptions about ethnic bloc voting.

The reality, however, is that there will be differential voter turnout in the many ethnic zones regions of the country in a manner that will have a non-trivial impact on the outcome of the election. As I highlighted in a previous post, this will be a turnout election. Everyone knows who their voters are. The swing voters will be few. And the two major contenders don’t appear to have any intention or strategy to eat into each other’s perceived strongholds.uhuru

And so just as I did last time, I ran the numbers from the latest opinion poll from Ipsos Synovate with regional breakdowns to estimate the winner of the presidential election. This time round my turnout model also includes variables on income, voter registration, and whether a region has a top presidential contender or not.

So what do the numbers say?

Well, if the polls are right Uhuru Kenyatta still leads Raila Odinga by about 740,000 votes.  I estimate that Mr. Kenyatta will get 48.87% of the votes cast to Mr. Odinga’s 41.72%, which means that a run-off is almost inevitable. I don’t expect Mr. Kenyatta to hit the 50% mark since my model is slightly biased in his favor (especially coming from the Rift Valley turnout figures from 2007 that I use as a basis of estimating turnout in 2013).  Below I show the regional tallies according to my turnout model.

A few caveats to go with my estimates: 

  • Of course the polls could be all wrong, in which case none of what I say here matters. 
  • The result of the election will hinge on the turnout in the respective candidates’ strongholds. As it is my model estimates a turnout rate in Mr. Kenyatta’s strongholds at 77.33 to Mr. Odinga’s 66.68. Mr. Odinga can easily erase Mr. Kenyatta’s lead by matching his stronghold turnout rate and having a respectable showing in Western region (by eating into Mr. Mudavadi’s vote share) on March 4th.
  • In the event of a run-off, all bets are off. Most likely the election will then centre on Mr. Kenyatta’s ethnicity (which is the same as that of the outgoing president) and his tribulations at the ICC thereby handing Mr. Odinga a lead straight off the gates.
REGION Uhuru Kenyatta Raila Odinga
Coast 64,064 410,012
North Eastern 103,978 115,531
Eastern 865,432 732,612
Central 1,623,483 113,266
Rift Valley 1,655,262 695,972
Western 58,040 444,974
Nyanza 282,626 1,299,053
Nairobi 406,099 507,624
TOTAL 5,058,984 4,319,044

For those interested in looking under the hood of Ipsos Synovate’s polls go here.