Past as Prologue? What the parliamentary election results tell us about the September presidential election

Share

To better understand the influences that will shape the outcome of the upcoming presidential election, guest author Scott Worden (with input from Colin Cookman)* analyses recent voter registration patterns, as well as voting patterns from the last parliamentary election, in October 2018. He found large differences between provinces and regions – in security conditions, rates of voter registration and the proportion of the population that cast their votes – meaning that some areas have a much greater potential influence on the outcome of the presidential election than others. This analysis seeks to provide a useful insight into which areas of the country may have greater relative weight in the 2019 presidential election and which may not.

Summary

Comparing past registration and voting patterns provides indications as to which areas of the country will have greater or less potential to influence the outcome of the upcoming presidential election. This is the case even though the risk of fraud will dominate election concerns: the impact of fraud depends significantly on a foundation of registration and turnout. As candidates and voters consider whether there is a ‘level playing field’ in the upcoming election, voting and voter registration patterns reveal several slopes in the landscape.

• Based on parliamentary vote totals by province and region, the lowest proportion of registered voters showed up in the southeast (21%) and south (27%). Ironically, this is where registration was highest compared to population. In other words, many voters in these regions were registered, but comparatively few voted. The highest voter participation rates were in the Central Highlands and the north, where registration per capita was lowest, ie while relatively few registered, those that did were more likely to vote.
• Overall, more votes were cast in ‘urban’ than rural areas during the parliamentary election. Candidates that appeal to urban voters are likely to have a structural advantage in the presidential race. A big question is whether urban voters have different candidate preferences than rural voters – or whether other factors like ethnicity, education, or age matter more.
• More than half of the parliamentary votes were cast in seven out of 34 provinces, all of which contain large cities: Kabul province alone comprised 16.5 per cent of the total national vote, followed by Herat (9%), Nangrahar (8%), Balkh (6%), Takhar (5%), Baghlan (5%) and Kandahar (5%). Together, these provinces made up 53 per cent of the vote cast, although they contain only 47 per cent of the country’s population. Candidate preferences in these provinces may play a dominant role in choosing the next Afghan president, particularly if the contest is close.
• The 15 lowest-voting provinces (Uruzgan, Zabul, Nuristan, Logar, Farah, Paktika, Nimruz, Panjshir, Kapisa, Badghis, Laghman, Paktia, Kunduz, Jowzjan and Wardak) together made up only 15 per cent of the national vote in the 2018 parliamentary election. If these 15 provinces had voted in proportion to their population, they would have comprised 29 per cent of the vote. Most of these provinces are significantly affected by the insurgency. In that sense, the Taleban significantly disrupted the parliamentary elections and have a lot of leverage to do so again in September.
• The Central Highlands significantly over-performed in terms of votes per population in the parliamentary election. Daikundi cast 3.9 per cent of the parliamentary votes, but makes up only 1.4 per cent of the population – a ratio of 2.5 to 1. Bamyan had more than twice its share of votes cast compared to population, at a ratio of 2.1 to 1 (it has 1.9 per cent of the population, but cast 3.6 per cent of all votes). If participation rates are similar in the presidential election, the Central Highlands will represent a valuable vote block.
• While fraud undoubtedly occurred during the parliamentary elections, it appears that the scale of ballot stuffing was lower than in past elections – helped by new biometric voter verification processes and polling station voter lists. None of the 2018 provincial level vote totals are demographically implausible. Yet with dozens of candidates running for each parliamentary seat, only a few hundred votes determined winners from losers. Large-scale fraud is more necessary in presidential elections, and high voter registration in some insecure areas preserves the risk for a return of mass fraud in the upcoming vote.
• The biggest wildcards in the presidential election are the east and southeast. Voting for the parliamentary election there was anaemic, with only 569,000 votes or 16 per cent of the total votes cast (although these regions have around 28 per cent of the registered voters). The potential vote is much larger. Over 40 per cent of the population in these two regions registered to vote, equalling 23 per cent of total registration in the country and comprising a maximum potential two million votes. If turnout significantly increases between 2018 and 2019, this could be the biggest vote block in the presidential election.

The difficulty of electoral predictions

Electoral predictions are a dismal science in Afghanistan – both because of widely shifting political preferences and the murky data that underlies all planning and prediction efforts. To better understand the influences that will shape the presidential election, this report examines both where voters are registered and where votes were cast in the October 2018 parliamentary elections. This combined information offers an indication of which vote banks may be more potent in the upcoming presidential election.

Significant differences in security across the country have created an uneven landscape for voter participation and this will affect results. A new voter registration system, implemented in the run-up to the 2018 election, has placed further limits on who can vote, as will be seen below.

Pre-election polling would normally provide the means for predictive analysis based on voter populations. But there are few reliable political polls in Afghanistan and it is difficult to separate likely voters from the general population when creating a polling sample population. In 2014, the few public polls of presidential election preferences were met with accusations that the pollsters were biased and, in one case, threats against the international organisation conducting the poll.

The next best thing, therefore, is to examine geographic differences in voter participation, which can indicate where candidates may have political advantages. This is particularly relevant since ethnicity is probably the most salient feature of politics in Afghanistan right now and in many areas ethnicity correlates strongly with geography. The second round of the 2014 election showed that Abdullah won more votes in non-Pashtun areas and Ghani won the most votes in Pashtun areas, including in areas that did not overwhelmingly vote for him in the first round.

Generally speaking, Pashtun populations are more concentrated in the southern and eastern regions bordering Pakistan; Tajik populations are generally higher in the north, west and northeast; Hazara populations are generally higher in the Central Highlands; and Uzbek populations higher in a few northern provinces surrounding Jowzjan. Kabul province is probably the most ethnically diverse and has the largest share of population of any province – an estimated 16 per cent of the national total.

In addition to ethnicity, candidates and vice presidents will likely attract more supporters from provinces where they grew up or have tribal affiliations. Piecing this puzzle together for a particular presidential ticket is beyond the scope of this report. But the data here can be applied by others to estimate where vote blocs are stronger and weaker.

Methodology

The electoral data analysed in this report has been taken from the Independent Election Commission (IEC) website. The final results of the parliamentary vote were posted station by station here and laboriously downloaded by Colin Cookman (accessible as a downloadable data file here). Cookman’s data site enables comparisons of voting and registration at the polling centre level, as well as past population estimates at the district level.

Prior to the 2018 elections, the IEC published a preliminary set of voter registration figures by province here. Although never publicly released, the IEC subsequently updated its final pre-election voter registration figures after a brief clean-up process in early October 2018 to remove duplicates and other erroneous registration. That final voter registration list was obtained from an election observer and used for calculations in Cookman’s electoral dataset.

Population figures are taken from the Central Statistics Office’s (CSO) published population dataset for 2018-19. CSO population statistics are the official source used by the government and the IEC, although historically they have been lower than other estimates (such as Flowminder, iMMAP and LandScan, which are satellite-based estimates). If the CSO population estimates are indeed lower than the actual population figures, it means that an even smaller share of the population has registered and voted than cited in this report.

For the purpose of this report, the voting patterns in the 2018 parliamentary election are treated as a ‘floor’ for the presidential election, on the assumption that at least most of those who voted in the parliamentary election will do so again for the presidential vote – and more will probably do so. The voter registration patterns, whose numbers include ‘top-up’ registration this year, are considered the ‘ceiling’ for this report, since only those who have registered will be able to vote.

National parliamentary election participation trends

An analysis of the 2018 parliamentary vote shows a historically-low turnout and significant differences in turnout between the provinces. It also showed unequal voter registration.

Low turnout in the 2018 parliamentary election

The overarching story of the 2018 parliamentary elections is one of historically-low turnout according to every metric. Based on the IEC’s final results, fewer valid votes were cast than in any previous election, both in absolute terms (3.6 million vs 4.2 million in 2010) and as a percentage of the population (11% vs 17%). Fewer polling locations were open and there were fewer reported results than ever before (4,600 vs 5,200 polling centres in the 2014 presidential election). More districts saw no voting than ever before (51 out of the 421 districts and urban centres that the IEC uses for its polling centre plan).¹ Also, for the first time, an entire province had no election (Ghazni, which includes 19 of the 51 districts mentioned above). In previous elections, Ghazni cast respectively 275,000 (second round of the 2014 presidential election) and 179,000 votes (2010 parliamentary election).

It is difficult to say whether this low turnout in the parliamentary election will be repeated in the presidential election because the political dynamics are different. In parliamentary elections, local candidates and local rivalries can inspire people to vote to address local patronage demands. Presidential politics are more consequential for the nation, but can seem more remote to local constituents. This year’s presidential election will be the first not to have a local component in the form of provincial council elections since 2004. That ‘standalone’ presidential election was marked out by high levels of participation, but it took place under unique transitional circumstances and with far better, countrywide better security; it cannot be a template for what will happen this year. The lack of a local poll alongside the presidential ballot in September, some argue, means the election will attract fewer voters.

With these factors in mind, the overall low turnout in 2018 can be considered as a floor of participation that illustrates a ‘worst case scenario’ for voting participation in the presidential election.

Provincial winners and losers in the 2018 parliamentary election

There was wide variation in relative participation – votes per capita – between the provinces in 2018. Overall, the parliamentary election saw more votes per capita in the Central Highlands and fewer in the south, southeast and east. Daikundi, for example, had 132,000 votes recorded in 2018 (26 per cent of its estimated 499,000 population), while, combined, Paktia, Paktika and Khost cast 159,000 votes (just eight per cent of its estimated combined 1,955,000 population). Kandahar province cast 157,000 votes, which was only 11 per cent of its estimated population. Only 12,000 of Uruzgan’s 420,000 citizens voted in the parliamentary election – less than three per cent. Translated into a nationwide election, this would suggest that areas with high insecurity and a largely rural population would be significantly under-represented in terms of the vote.

Different levels of access to voting is significant because geographic regions have distinct political characteristics that may favour some candidates over others. Therefore, suppression of the vote in one area will tilt the playing field toward others. The constitution addresses this problem for the parliament by requiring that seats are allocated to provinces proportional to their population – thus seeking to ensure that voters in different provinces are represented in relatively equal proportions in the national legislature. In the presidential election, however, the whole nation is one constituency and all votes are weighted equally. Therefore, a province or region with more voters registered compared to its population, potentially has a greater influence on who becomes president.

To illustrate the case, in the parliamentary election, the number of votes in Uruzgan amounted to 0.3 per cent of the total number of votes cast in the country, whereas the Uruzgan population is 1.3 per cent of the national total and Uruzgan has 1.2 per cent of the seats in parliament. If this pattern were repeated in the presidential election, the influence of Uruzgan citizens in choosing the President would be roughly four times less than their population would suggest. Kunduz and Helmand are also highly insecure and also significantly under-performed on a per capita basis in the 2018 election. Together, they cast 128,000 votes, amounting to 3.8 per cent of the national total, whereas their combined population is 8.7 per cent of the national total and their parliamentary seats 6.8 per cent of the total.

While small and insecure provinces will likely have reduced influence in the presidential election, several large and more urbanized provinces will likely dominate the national presidential vote. A majority (53 per cent) of the votes cast in the 2018 parliamentary election were from seven out of 34 provinces: Kabul, Herat, Balkh, Nangrahar, Takhar, Badakhshan and Kandahar, which contain only 47 per cent of the population.

Greater urban participation

Looking more closely at the population map, one of the most salient distinctions between the 2018 election and past elections is a shift toward more ‘urban’ participation. In the 2010 parliamentary election, districts that contained a provincial capital (in most cases, the largest city in the province) contributed 30 per cent of the final vote, in line with their estimated 30 per cent share of the national population. In 2018, 38 per cent of the final vote was cast in districts containing the provincial capital, which were estimated, by then, to comprise around 32 per cent of the population.²Thus, in 2018 these urban voters had a greater influence over results than more rural areas.

At the provincial level, there is more variation and there are some extreme outliers. In Farah, 87 per cent of votes came from the capital district (21,952 of 25,200 votes for the whole province). Yet the Afghan Central Statistics Office estimates the capital district has only 23 per cent of the population (123,135 people out of a total provincial population of 543,257.) Farah, Helmand, Jowzjan and Zabul all had at least 40 per cent more of the vote coming from capital districts than the percentage of population that lives there. It is notable that each of these provinces has high levels of Taleban control outside the capital, which could explain the relative lack of votes from those districts.

Another way to measure ‘urban’ versus ‘rural’ voting influence is to look at polling centres that are closest together, based on IEC GPS data (close proximity indicates they are located in cities or villages with higher population densities) and to compare them to polling centres that have the greatest distance between them (indicating a more rural and sparse population). Overall, 41 per cent of the vote in the country was cast in the top 25 per cent most densely-concentrated polling centres. Only 16 per cent of votes were cast in the bottom 25 per cent of polling centres, that are furthest apart.

Unequal voter registration

Since the adoption of the 2018 Electoral Law, eligible citizens can only vote in the specific polling centre where they registered prior to the election. Therefore, voter registration data paints a clear picture of the maximum  vote that may occur in a given area during the presidential election.

Voter registration figures for the 2018 election indicated that only 60 per cent of estimated eligible voters (Afghan citizens age 18 or older) registered to vote (see also this AAN dispatch).³

Within that low figure, there were large disparities between provinces and regions. In seven provinces, less than 40 per cent of the estimated eligible population was registered; 43 districts had no registration at all (including all 19 districts in Ghazni, where in 2018, no election was held) and; 16 districts conducted voter registration, but no women registered to vote. The least represented provinces were Farah (6% of those over 18), Uruzgan (7%), Kunduz (8%) and Badghis (8%). Helmand, Logar and Wardak are also areas of high insecurity and had below average voter registration. Jowzjan, one of the provinces where First Vice President Dostum has ‘delivered’ the vote in previous elections to Hamed Karzai and then Ashraf Ghani, also significantly under-performed in voter registration, with only eight per cent of those of voting age registering. In neighbouring Sar-e Pul, only nine per cent registered.⁴  The fall in registration rates in some northern provinces due to the Taleban capture of large areas was so marked that the United Nations Secretary General in his 10 September 2018 report on Afghanistan to the General Assembly highlighted the “discontent” there and said that “some opposition figures accused the Government of a deliberate plot to disenfranchise northern communities.” (See also AAN reporting) on registration and voting in Faryab in 2018.)

Registration as a percentage of CSO population estimates was highest in the Central Highlands and the east. The eastern provinces of Nangrahar, Nuristan, Khost and Kunar also significantly exceeded the national voter registration average. Bamyan, Daikundi and Ghor in the Central Highlands registered above average as well. Kandahar in the south and Baghlan in the north were also in the top ten in of registration rates.

It should be noted that there were several strong indicators of fake registrations in insecure provinces such as Kandahar and Paktia (described in this AAN guest analysis). In Paktia, for instance, registration exceeded the estimated voting age population by five per cent. Even if the population estimates are inaccurate, this is such an extreme result that some form of fraud is indicated because that would mean 100% registration occurred in a province with significant security obstacles – well above national averages and rates in developed democracies.

Regardless of fraud, areas that have high registration numbers, like many southeastern provinces, have a high potential influence in the presidential election. Areas with low registration numbers, as in several southern provinces, are capped in the number of votes they are able to cast.

New voter registration for 2019

In August 2019, the IEC released new provincial voter registration figures, following a month-long ‘top-up’ exercise aimed at registering or updating voter rolls since the parliamentary elections last autumn. According to these figures, this brought the 2019 voter registration to 9.6 million, which includes 235,000 registrants in Ghazni – where no registration was conducted in 2018. Setting Ghazni aside, this roughly 565,000 increase from 2018 registration surpasses an earlier press-release that claimed only 317,395 new voters had been registered in the process, not including voters in Ghazni. Since then, IEC officials also stated that they had removed approximately 400,000 duplicate or erroneous voter registrations from the national list. The IEC has not yet reconciled these discrepancies.

Comparing the detailed 2019 consolidated voter registration numbers with the 2018 numbers shows that several provinces gained considerably more than others. Several small and/or insecure provinces, including Nuristan (35%), Faryab (16%), Badghis (16%), , Paktika (14%), Kunduz (13%), and Uruzgan (14%) saw the largest percentage increases in voter registration between 2018 and 2019 – although in some cases like Nuristan, this was from a low baseline and the absolute number of voter registrations involved is not that large at the national scale. Kabul, Nangarhar, Kandahar, Herat, and Faryab saw the largest net changes. These five provinces collectively added 207,000 votes to their 2018 voter registration. Within these provinces, the majority of that increase coming from polling centres in their capital districts, increasing the power of urban voting in the presidential election.

Voter registration as a ceiling on participation

As difficult as it is to predict where people will vote, it is easier to determine where people will not vote.

First, people will not be able to vote in areas where voter registration was unavailable Out of the known universe of roughly 7,400 polling centres from past elections, only 5,106 opened for registration, the deficit largely due to insecurity. Out of these, only 5,066 had registration data posted on the IEC website (representing 69 per cent of the ideal coverage). Overall, 43 out of 421 districts had no registered voters in the 2018 election, including the 19 districts in Ghazni. Moreover, during 2019, top-up voter registration was largely limited to district centres – although voters who made it to a district centre could choose any polling centre in the province to vote in. This meant that voters in more remote areas faced greater challenges to register due to a lack of affordable or safe transport and, for women, cultural norms that restrict them from traveling.

Second, people cannot vote if they chose not to register even where there was access. With a relatively short time to register in 2018 and almost no public outreach, participation was low even in areas that were accessible and secure. In Kabul province, for example, only 32 per cent of the estimated eligible voters registered. In very safe Bamyan province, only 30 per cent registered (although here there were also complaints that the IEC did not open enough registration sites). Choosing not to register is an individual voter’s choice. But cumulatively, it affects the size of voting blocs that candidates hope to draw on for electoral support.

Third, during the 2018 election, many poling centres that planned to open on polling day were closed – largely due to poor security but also administrative and staffing problems. Of the roughly 5,100 polling centres that had registered voters, only approximately 4,600 locations actually opened based on reports of valid results. Given that people are required to vote in the centre where they registered, closing a polling centre where registration occurred bars those citizens from voting.⁵

The consequences of uneven representation

All this means that, even if one assumes maximum participation, several provinces and regions will have more opportunity to vote than others. The lowest registration rates per population tended to occur in areas with a more active insurgency. Each of these provinces will have a significant structural disadvantage when it comes to choosing the next president.

Looking at the registration ceilings, if the same proportions of registered voters turn out in all provinces in the presidential election as they did in the parliamentary election, the east and southeast have the greatest advantage, whereas the south, west and north will have the lowest participation. This means that candidates relying on votes from the north, south and west are at a structural disadvantage when it comes to getting out the vote.

The above graph shows the difference between a province’s share of the national population and its share of national voter registration. If each province has the same voter turnout rates in the presidential election, the provinces on the left would have more influence on the choice of a president than their population would indicate. Provinces on the right would have less influence than their population would indicate.

FIGURE: GAP BETWEEN REGISTRATION AND TURNOUT BY PROVINCE

This graph illustrates the range of potential voter participation in the presidential election by province. The top line shows per-capita registration, which is effectively a ceiling of votes that may be cast. The bottom line shows the percentage of the population that actually voted in the 2018 parliamentary election. Provinces on the left side of the scale have the most uncertain vote banks. Provinces on the right already maximised their opportunities to vote in the 2018 vote.

Based on parliamentary vote totals by province and region, the lowest proportion of registered voters showed up in the southeast (21%) and south (27%) – ironically, where registration was highest. In other words, relatively-speaking, many voters in these regions registered, but comparatively few voted. The highest voter participation rates were in the Central Highlands and the north, where registration was lowest. Both the north and the west made up for their relatively low registration rates with relatively high participation, ie while relatively few registered, those that did were more likely to vote. Jowzjan had 59 per cent of registered voter turnout and Herat had 60 per cent. Badakhshan, Balkh, Parwan, Baghlan, Panjshir and Samangan were all above the national average.

There may be several reasons for the contrast between voter registration rates and voting rates. Security conditions may have changed in the six months between registration and voting. Enthusiasm for registering to vote may be greater than the appeal of voting for particular candidates, if a person does not see a favourite they like.

Fraud is another possibility. It is possible that areas with fraudulent registration amassed potential vote banks that were not used in the parliamentary election – either because the last minute introduction of biometric voter verification upset ballot stuffing plans, or because fraudulent registration was intended to be used for presidential elections and not the parliamentary elections. If this last theory is true, then extra vigilance is needed to prevent fraud in areas that have high registration and low parliamentary voting.

Ultimately, whether a particular province or region’s electorate will make the most of the possibilities provided by its level of registration is the most difficult question to answer and relies on a combination of candidate campaign mobilisation and prevailing security conditions. The turnout for the parliamentary elections may, however, provide some important clues.

* Scott Worden is Director of Afghanistan and Central Asia Programmes for the US Institute of Peace. He served as an international member of the Electoral Complaints Commission during the 2009 Afghanistan presidential elections.

(1) See attached annex listing the 51 districts where there was no registration.

(2) The 2018 figures are derived from this summary table. (specifically the national total rows for the variables “capital_district_votes_pct” and “capital_pop_pct_CSO_18_19”). The 2010 population data was calculated based on the CSO’s 2010-11 population estimate, accessible by Internet Archive here.

(3) The IEC registered 8.8 million voters for the 2018 election. Election officials have historically estimated that around 50 per cent of the population is 18 years or older, and therefore eligible to register to vote. The most recent statistics from the CSO now estimate that 54% of the population (17.4 million people) is under 18. Therefore, if the national population is approximately 32.2 million (according to current CSO data for 2019), then the estimated eligible voter population is around 14.8 million, meaning that 40 per cent of potentially eligible voters had not registered and were not be able to vote in 2019.

(4) Population / registration figures by province are included on this constituency summary table, which is also the source of the regional participation calculations in the following sections. Registration data by polling center (not rolled up into district-level in the file, but that’s where these no-registration figures came from) is here (checked against the district list here).

(5) In the 2018 election, there were also eight districts that had collectively registered 94,000 voters, but had no votes recorded on election day. The overall population of these 51 districts was 2.9 million (representing approximately 1.45 million potentially eligible voters).

Endnotes:

Revisions:

This article was last updated on 9 Mar 2020

Tags:

election parliamentary election

Authors:

Scott Warden

More from this author