Note: This is a longer form of an article published at The Conversation (see also [Cape Talk interview](http://www.capetalk.co.za/podcasts/144/the-john-maytham-show/214614/sas-voter-turnout-a-mathematician-runs-the-numbers)). This version contains explanatory footnotes with further details about the sources for the data, the graphs, and the computations behind them.
It has commonly been believed in South Africa that voter turnout in national elections has been fairly high when compared with other countries (for instance, SA voter turnout better UK, US and France). However, this is not true.The unfortunate misunderstanding arises because different definitions of 'voter turnout' have being used.
In this article I will clear up some confusion and show that voter turnout in South Africa, when measured either as a percentage of the voting-eligible population as a percentage of the voting-age population, has been low by international standards for the last few elections.
For instance, in the 2016 US presidential elections, voting-age population turnout was 56%. This is the kind of figure you may have seen on CNN. Well, in the 2014 South African parliamentary elections, voting-age population turnout was 54%. And in the 2019 elections, it was about 47% - amongst the lowest in the world.
To a mathematician, voter turnout is not a very precise term. In South Africa, the IEC defines it as the number of people who voted divided by the number of registered voters. In some countries, such as the UK, France and Canada, this definition makes sense because voter registration is high (well over 90%). In these countries it is not really necessary to distinguish 'number of registered voters' from 'number of eligible voters'.
But in other countries, such as the United States and South Africa, voter registration is lower. If 'voter turnout' is supposed to be a measure of how many voters bother to cast their vote (which includes registering, if necessary), the number of people who registered should be taken into account. We need to ensure that everybody is using the same denominator.
The 'gold standard' in measuring voter turnout is therefore to take the number of votes cast and divide by the size of the voting-eligible population (VEP). The VEP is the number of citizens living in the country who are legally allowed to register and vote. [1, 2]. In South Africa, this means all citizens over the age of 18. In other countries it is slightly different; for instance in the US, it means all citizens over the age of 18 who are not felons or mentally incapacitated.
The trouble is that estimating the citizen population living in a country is difficult: counting the population is fairly easy (this is done in a census), but counting the number of citizens is harder (this is not generally done in a census [3]).
In South Africa, the IEC has sometimes provided such estimates; for instance in the 2019 elections it estimated that 74.5% of eligible voters were registered.
Since 66% of registered voters cast a vote, we conclude that the VEP turnout was 66% x 74.5% = 49%. This is the figure you may have seen in the media. See also [7].
Here is a graph [4] comparing the VEP turnout of South African parliamentary elections to the major elections in the UK, US and France, from 2000 to the present day. Hover your mouse over a data point to get further information.
From the graph we see that since the 2004 elections, the voting-eligible population turnout in South Africa has consistently been much lower than in the UK and France, and a little lower than in the US. Also, voter turnout plunged dramatically in the 2019 South African elections, from 57% to 49%.
In order to compare voter turnout in South Africa to a broader spectrum of countries, we will need to use a more readily available statistic, the voting-age population turnout. The voting-age population (VAP) is simply the number of people living within the country of voting age (that means over 18 years old in South Africa, as in most countries). This number is usually fairly accurately known from census data.
VAP turnout is not as good a measurement of voter turnout as VEP turnout, since it does not account for noncitizens and other residents of a population not eligible to vote. However, it is a more accurate reflection of voter engagement than only dividing by the number of registered voters, as it takes into account those who do not register.
The International Institute for Democracy and Electoral Assistance keeps tabs on VAP turnout data.
The VAP turnout for the 2019 South African elections is not yet in the IDEA database, but we can estimate it ourselves. The official Stats SA figures allow one to estimate that there were 37.8 million people aged 18 years and older living in South Africa in mid year 2018 [5]. Furthermore, 17.6 million votes were cast. Therefore the VAP turnout can be estimated at about 46.7%.
Here is a graph of VAP turnout in the most recent major national election for the 59 'free' or 'partly free' [6] countries in the world with reater than 1 million inhabitants. Some countries have been highlighted for the sake of comparison. Hover your mouse over a column to get further details.
We see from the graph that VAP voter turnout in the South African 2019 elections ranked amongst the lowest in the world.
Thankfully, in recent weeks, the media has started to realize the problem and the concerns of various experts are being heard.
The term 'voter-eligible population' was coined by Michael McDonald (Associate Professor of Political Science, University of Florida):
The voting-eligible population or VEP is a phrase I coined to describe the population that is eligible to vote. Counted among the voting-age population are persons who are ineligible to vote, such as non-citizens, felons (depending on state law), and mentally incapacitated persons. Not counted are persons in the military or civilians living overseas.
This term is also used by the International Institute for Democracy and Electoral Assistance (IDEA). From their Glossary:
The voting age population is based on a country's population over the age of 18. It is not intended to be an exact measure of the number of citizens entitled to vote as it does not take into account legal or systemic barriers to the exercise of the franchise or account for non-eligible members of the population, such as resident non-citizens (the voting eligible population – VEP – would capture these discrepancies but it is very hard to achieve the data required to measure VEP). It is intended as indicative only.
In the South African context the voting-eligible population is the number of South African citizens aged 18 years or older living in South Africa at the relevant time.
The term 'voter-age population' (VAP) is defined in Wikipedia as:
Voting age population (VAP) refers to the set of individuals that have reached the minimum voting age for a particular geographical or political unit. The presumption is that they are therefore generally eligible to vote, although other additional factors may cause them to be ineligible, such as lack of citizenship or a prior felony conviction. In estimating voter turnout the voting age population for a political unit is often used as the denominator for the number of individuals eligible to vote in a given election; this method has been shown to lose in accuracy when a larger percentage of the VAP is ineligible to vote. In the United States individuals become eligible to vote in political elections at age 18.
The term VAP is also defined by the IDEA (see footnote [2]) in the same way, to refer to all people of voting age living in a country.
A citizenship question was asked in the 2001 South African census, but not in the 2011 census.
For the UK and France graphs, I used the IDEA data for the VAP, and then subtracted estimates from Eurostat of the foreign citizen population for the relevant year.
For the US graph, I used the United States Electoral Project.
For the 1999, 2004, 2009 and 2014 South African elections, I used:
Schulz-Herzenberg, C, Trends in electoral participation, 1994-2014, chapter 2 of Schulz-Herzenberg, C & Southall, R. (eds), Election 2014 South Africa: The campaigns, results and future prospects, Jacana Media & Konrad Adenauer Stiftung, 2014.
For the 2019 South African elections, I computed the VEP from the IEC's January 2019 statement, as explained in the main text and footnote [7].
Some aspects of this graph worth pondering:
Eligible voter registration increased dramatically in the US between the 2004 (68%) and 2008 (88%) presidential elections.
While it is true that, on the whole, voter turnout is decreasing worldwide, it has been fairly stable in the UK, France and the US. In the case of the US specifically, the United States Electoral Project has shown that "the much-lamented decline in voter participation is an artifact of poor measurement". In other words, it is due to measuring voter-age population turnout instead of voting-eligible population turnout.
In some cases (such as for the UK), eligible voter registration actually exceeds 100%. (I have done some spot checks on the IDEA source data, as well as on US Census Bureau data which the IDEA sources its own data from, and it does seem to be a genuine artifact). I speculate this occurs because at any given point in time a fairly high number of UK citizens are abroad. Therefore they were not counted on the day of the census, although they are recorded as registered voters. However, I am open to being corrected.
The classification of countries into 'free', 'partly free' and 'not free' is taken from Freedom House, a US NGO. No doubt they have a somewhat biased viewpoint but their classification is widely used.
To estimate the VAP of South Africa relevant for the 2019 elections, I started with the official Stats SA mid year population estimate of South Africa for 2018 of 57 725 606. Then I subtracted their estimates for the populations of the 0-4, 5-9, and 10-14 age groups. Finally I subtracted 3/5 of the population of the 15-19 age group figure of 4 733 790, as an estimate of the 15-17 year old population. In summary:
VAP (mid 2018) = 57 725 606 - (5 928 951 + 5 862 081 + 5 252 485 + 3/5*4 733 790)
which gives us a VAP of 37 841 815 in mid year 2018.
This must be the same method the IDEA uses, since if you apply it to the 2014 Stats SA figures, one recovers precisely the IDEA estimate of the VAP population for 2014, namely 36 691 652.
To exactly follow the IDEA calculation, I would need to use the Stats SA mid year population figures for 2019, which are not available yet. I will not attempt to multiply by a "growth rate factor", since observations of the Stats SA population data indicate that the year-by-year changes in the population groups can vary quite a bit. So I will simply use this mid year 2018 VAP figure to estimate VAP turnout for the 2019 May elections.
Note: I am using the precise figures here, not because I think that the Stats SA population estimates are accurate down to to the last citizen (nobody claims that) but simply so that the reader can reproduce my calculations.
To estimate the VEP population for the 2019 South African elections, I started from the IEC statement released on 29 January 2019:
Based on latest voting age population estimates from Statistics South Africa, the current voters’ roll reflects a total registration by 74.5 percent of the eligible population.
Since the voters' roll at that time was 26.8 million, we conclude that the IEC estimated there were 35.9 million South African citizens over the age of 18 living in South Africa in January 2019. I emailed the IEC for clarification, and they said that this figure of 35.9 million SA citizens over 18 came from Stats SA.
On the other hand, we can use the Stats SA figures (see [3]) to estimate that there were 37.8 million people aged 18 years and older living in South Africa in mid year 2018.
The difference, around 1.9 million people, is therefore a rough estimate for the number of noncitizens living in South Africa.