Content uploaded by Tom A Moultrie
Author content
All content in this area was uploaded by Tom A Moultrie on Mar 11, 2014
Content may be subject to copyright.
Please do not cite or quote without prior permission
USED FOR ILL; USED FOR GOOD:
A CENTURY OF COLLECTING DATA ON RACE IN SOUTH AFRICA
*
Tom A Moultrie and Rob Dorrington
†
1. INTRODUCTION
South African history offers important insights into both extremes of the debate as to whether
politically and sociologically sensitive data on race, population group or ethnicity should be
collected and presented in official reports, surveys and censuses. By exploring the contours of the
uses and abuses of data on race in South Africa over the past century, this paper seeks to
contribute to that debate. As the title of this paper suggests, these data can be used for ill, such as
being used to bolster and perpetuate one of the most odious political systems of modern times.
However, these data can also be used for good as is the case more recently in South Africa, where
the self-same data (albeit with a crucially important difference, as will become clear) are
important to the process of transformation, as South African society grapples with undoing the
legacy of its racist past, although – as this paper argues – their importance may have been
overstated.
The paper is presented in three sections. The first briefly reviews the mainstream
literature on the nature of census data and its power to impose and define social relations that,
largely, serve the interests of a dominant (economic, political, social, colonial, imperial) elite. The
second contends that this conventional view can not be applied mechanistically to the South
African case: certain key features of apartheid-era legislation, notably the Population Registration
Act of 1950, had the obvious, and fundamental, effect of removing a great deal of individuals’
agency about how they chose to present and interpret their identity in relation to power. To
understand better this process, this section of the paper engages with the historical material
relating to identity in the South African polity over the course of the twentieth century. The final
part of the paper reflects on the political changes that have happened in South Africa since 1990,
and how the actively-pursued political and social agenda of transforming relationships not only
vertically between individuals and the state, but also horizontally among individuals, has been
used to justify the continued collection of data on race. In doing so, we gaze into the future, and
ask the important question of whether, indeed, the collection, capturing and interpretation of
social statistics by race is desirable in the long term, or whether more significant social and
*
Paper presented at the QICSS/INED Conference on Social Statistics and Diversity held in Montreal, Canada. 6-8 December 2007
†
Respectively, Senior Lecturer in Demography and Professor of Actuarial Science at the Centre for Actuarial Research (CARe), University of
Cape Town, South Africa. Tom.Moultrie@uct.ac.za
Please do not cite or quote without prior permission
Moultrie and Dorrington 2
economic cleavages, for example class, require that we commence a process towards social
statistics data collection along those lines instead.
2. REIFYING “RACE”: POWER AND SOCIAL STATISTICS
Over the last twenty years, awareness has increased among sociologists and philosophers that
censuses (in particular, among all the forms of data collected by a modern state) and the
categories that they impose, serve to reproduce and reify a view of social ordering and hierarchy
of a dominant elite. Some, for example the philosopher Ian Hacking (Hacking 1990, 1991, 1999),
would contend that one of the defining characteristics of a thoroughly modern state lies in its
capacity to collect and collate such data. Other authors approach the same concept slightly
differently, but all have as a common theme the power of large screeds of statistics to
simultaneously both anonymise and individualise people in a way that allows those in power to
better understand the nature of their citizenry. As Anderson (1991: 166) notes, “the fiction of the
census is that everyone is in it, and that everyone has one – and only one – extremely clear place”
therein. The urban geographer James Scott sees the same phenomena in different terms:
Officials of the modern state are, of necessity, at least one step – and
often several steps – removed from the society they are charged with
governing. They assess the life of their society by a series of typifications
that are always some distance from the full reality these abstractions are
meant to capture … State simplifications such as maps, censuses,
cadastral lists, and standard units of measurement represent techniques
for grasping a large and complex reality (Scott 1998: 76-77)
Hacking goes on to argue that Scott’s “state simplifications” require not only for the
categories themselves to be invented, but that this exercise in the “systematic collection of data
about people has affected not only the ways in which we conceive of a society, but also the ways
in which we describe our neighbour. It has profoundly transformed what we choose to do, who
we try to be, and what we think of ourselves” (Hacking 1990: 2). Simply put, the process of
categorising and labelling – of “naming into existence” as Goldberg (cited in Kertzer and Arel
(2002: 21)) puts it – is a significant component in manufacturing and reproducing identity.
Kertzer and Arel (2002) develop this line of argument further, and in the process coin the
neologism ‘statistical realism’ to describe the claims made by some scientists and demographers
that the formulation of classificatory systems in censuses are objective and are not socially
constructed. Instead, they suggest that “far from being a scientific enterprise removed from the
political fray, the census is more like a political battleground where competing notions of “real”
identities, and therefore competing names to assign to categories, battle it out” (Kertzer and Arel
2002: 20-21 emphasis in original).
There is a further dimension to these arguments, which relates to an apparent bifurcation
in the literature that draws a distinction between the use of census data to govern (or manage) a
Please do not cite or quote without prior permission
Moultrie and Dorrington 3
population benignly in a democracy, where the individuals and minorities so counted can use the
information to articulate and engage with the state to advance their own interests on the one
hand; and a more sinister notion associated with enumeration with the intention of controlling a
subjugate population on the other. The distinction is perhaps a little contrived, and assertions of
benign or malign design do not advance the debate particularly much: in large measure, the
distinction between use of data for managing or controlling a population hinges on the ability of
the people being counted to use (or subvert?) the data effectively for purposes other than those
intended by the state (for example to secure or lobby for sectoral interests).
Of all the classifications imposed on individuals in a census, it is not surprising that questions
relating to ethnicity, race and identity have been the most contentious. Kertzer and Arel (2002)
and Nobles (2002) have carefully documented the misapplication of concepts of race and
ethnicity in a wide variety of settings, ranging from the United Kingdom to Brazil.
Central to our deliberations is the need to consider what race means in the context of
social statistics. Nobles (2002) suggests that, even if it were more common before decolonisation,
only a very few countries continue to collect data on race in censuses and surveys, although it is
not regarded as particularly aberrant for the rather curious reason that questions on race are still
routinely asked in American censuses. What function does race serve in official social statistics?
Both Head (1997) and Krieger (2000) offer a useful framework for considering the matter. They
point out that there are three different conceptualisations of race and racial statistics. Two of
them simply serve to muddy the waters of the debate. They need to be dispensed with up front.
The first conceptualisation posits that race is purely biological – i.e. race as speciesism –
and that there is only one race of humans on the planet, which makes the categorisation
statistically redundant. In this view, all six billion-plus people have the same race; and hence there
can be no statistical variation by race relative to other variables.
The second is that differences in outcome variables by race are phenotypical (i.e. genetic).
While some genetic differences are (at least superficially in the case of skin colour) the most
obvious delineation of race, it is hard to develop a coherent argument that the small genetic
differences between different races directly condition demographic outcomes except in the case
of rare inherited traits such as sickle cell anaemia. Hence, works such as those by Tukufu Zuberi
(2001) and – in the South African context – Zuberi and Khalfani (1999) misdirect our attention
to old and discredited notions of racial determinism based on genetics and which hark back to
the darker side of demography’s eugenics-associated past. Certainly, the ‘statistical realists’
described by Kertzer and Arel (2002) would have to fall into this category in order to hold that
racial differences in outcome were in fact inherent to the physical race, and that the categorisation
of race was objectively justifiable.
Please do not cite or quote without prior permission
Moultrie and Dorrington 4
Zuberi, of course, also engages with the third conceptualisation of race (and the one most
frequently used by sociologists), namely that “race” is a social construct. This is the sense that has
been imparted to the debate on the collection of data in censuses and surveys, and which has
been described above.
A social constructionist approach reminds us that there is nothing inherent to an
individual’s race that is causal of social outcomes, and that in this sense, “race” is a proxy for a
gamut of other factors, such as class, place of residence, life opportunities etc.
‡
With enough of
these other factors included, racial categorisations – so the argument goes – should become
statistically irrelevant.
However, as Krieger (2000) points out, it might perhaps be naïve to strive for this. True,
race itself may not directly cause, or cause to be affected, health or other social or demographic
outcomes, but where it can be demonstrated that race may implicitly or explicitly mediate
decisions and/or opportunities over an individual’s life course, irrespective of the other variables
for which race is regarded as a proxy
§
, then race itself is important in understanding inequalities in
health and demography.
Are the examples given and situations alluded to by Krieger exceptional? We would argue
not. Vestigial racism is depressingly commonplace around the world. Thus, while we would
wholeheartedly concur with objections to regarding race as a biological determinant, the
experience of South Africa over the course of the twentieth century offers a powerful case to
understand race as a social determinant of social, health and demographic inequality, as well as
providing a useful example against which to develop a case as to why statistics on race should
continue to be collected. These issues are developed further in the next section.
3. COLLECTION OF DATA ON RACE IN SOUTH AFRICA (OR, COUNTING NON-PEOPLE)
The emergent orthodoxy that critically engages with the power of social statistics to shape
fundamentally individual realities, and its connection with the modern state, has found a voice
among some South African sociologists and philosophers. Deborah Posel, one of South Africa’s
pre-eminent sociologists and historians of the genesis of apartheid, advances the quintessentially
Foucauldian idea that the “intellectual power of statistics was closely interwoven with the
emergent power of the modern state” (Posel 2000: 117). The role of the census, she further
agues, is central: “quantitative measurement was one of the epistemological underpinnings of
power in many colonial states, where censuses and surveys were as much exercises in defining
subject populations as in measuring their various demographic characteristics” (Posel 2000: 120).
‡
It is also worth observing, in passing, that if one regards both “race” and “ethnicity” as social constructs, it is perhaps not surprising that the
boundaries between the two concepts are so ill-defined (as has been pointed out by Kertzer and Arel).
§
See, for example, some of the instances cited in Krieger (2000: 213).
Please do not cite or quote without prior permission
Moultrie and Dorrington 5
What, then, was the nature of information collected on race in the segregationist and apartheid
years?
A frequently-made error by those considering this topic in the South African context is to
presuppose that South Africa’s racist history began with the articulation of Grand Apartheid, and
the coming to power of the National Party in 1948. While 1948 represented a watershed in South
African politics, because for the first time a government was elected (by the White electorate) on
an explicit platform of racial segregation, the foundations of the apartheid state had been laid
many decades earlier.
For the purposes of this paper, it is useful to reflect in some detail on the nature of
questions on race and ethnicity in South African censuses from 1911 onwards (for interested
readers, a useful – if brief and necessarily superficial – account of census data collection in South
Africa both pre- and post-Union can be found in Christopher (2002)).
South Africa was formally constituted as a Union of four provinces in 1910. Section 34 of
the South Africa Act of 1909, which established the essential constitutional principles of the
Union, mandated that “in 1911 and every five years thereafter
**
, a census of the European
population of the Union shall be taken for the purposes of this Act” (South Africa 1911:16). The
Census Act of 1910 further allowed for censuses to be conducted at the discretion of the
Governor-General, and that it could be taken “of the population” (South Africa 1911: 116). It is
noteworthy that the neither the Census Act, nor the South Africa Act, required a census of the
entire population to be conducted at regular intervals, although in the parliamentary debates on
the Census Bill, the Minister of the Interior, Jan Smuts, was reported to have said that
In South Africa there would be the quinquennial census in 1911, and in
the South Africa Act there was a provision that every five years
thereafter there should be a census of the whole male European
population of South Africa; and the method of distribution of seats and
the representation of the Union was to follow the results of that census.
There would be a universal census every ten years, not only for the white
male adult population but for many other questions … [it] would be
possible for the government to provide for an elaborate census ,
embracing information on all matters which should be covered by a
census, and to provide every five years for a more restricted census,
which would only deal with questions of the white population necessary
for Constitutional purposes… (South Africa 1910: 61-2)
The whites-only census scheduled for 1916 was delayed until 1918 as a result of the War;
a full census was conducted on schedule in 1921. The next whites-only census was conducted in
1926, but the full enumeration that was supposed to have happened in 1931 was restricted to
whites because of the economic hardship of the Depression (South Africa 1938: vii). The 1936
**
The requirement for quinquennial censuses was dropped after 1941 by the Census Amendment Act of 1935. After 1941, censuses were required
to be held decennially.
Please do not cite or quote without prior permission
Moultrie and Dorrington 6
census was accordingly designated to enumerate all South Africans. As Jan Hofmeyr, Minister of
the Interior argued in parliament at the time of the debate on the Census Amendment Act of
1935
††
,
We have now had no non-European census since 1921 and I think we
have experienced the serious disadvantages which have followed upon
not having a full census of the population in 1931.We are fully conscious
that we could not wait until 1941 before we had another census. That is
why it was essential to hold a non-European census in 1936… (South
Africa 1935: 853-4)
The last whites-only census was conducted in 1941, with a full census being conducted in
1946. Thus, prior to the promulgation of the Population Registration Act in 1950, censuses that
sought to cover the entire population were conducted in 1911, 1921, 1936 and 1946. Posel thus
errs in her observation that “in practice, limited censuses of ‘non-whites’ were undertaken in
[only] 1911 and 1946. But only one census covering all races was conducted, in 1936” (2000:
123).
These four censuses offer a fascinating insight into the collection and categorisation of
racial data in South Africa before apartheid. A copy of the census form was not appended to the
official report on the 1911 census (South Africa 1912), but the tabulations of the population by
race reveal an idiosyncratic (not to mention quite confused) approach to racial and ethnic
classification (Table 1) that is strongly redolent of that discussed by Hirschman (1987) with
reference to Malaysia. Note particularly, the conflation of terms reflecting race, ethnicity and
nationality. This is, perhaps, not that surprising: as Christopher (2002; 2005) has noted, there was
a vigorous attempt to standardise censuses across the British Empire in the first few decades of
the twentieth century. In any event, and significant to this paper, race was (at least to a degree)
self-reported. Such latitude was certainly not given again in South African censuses until 1996.
The 1921 census set a precedent which would last for more than fifty years: four different
enumeration forms were used, based on the race of the respondent. In a move that presaged the
creation of the Bantustans along ethnic lines, questions on ethnicity were asked of Africans too.
††
Interestingly, given the widespread consternation about the racial composition of the Union that can be found at the time (see Moultrie (2005)),
the debate around the Census Amendment Act, which sought to replace the constitutional requirement for quinquennial censuses with decennial
ones from 1941, was almost entirely silent on the matter of race. MPs were far more concerned about the whether such a move was indeed
desirable, and in whose favour only decennial reapportionment of representatives would work.
Please do not cite or quote without prior permission
Moultrie and Dorrington 7
Table 1 Classification of race/ethnicity in the 1911 South Africa Census
European or White
Other than European or White Bantu Baca
Basuto, including Bapedi
Bavenda
Bechuana
Bomvana
Damara
Fingo
Hlangweni
Kaffir (unspecified)
Ndebele
Pondo
Pondomise
Swazi
Tembu
Tonga (alias Bagwamba,
including Tshangana)
Xesibe
Xosa (sic)
Zulu
Other tribes Southern Rhodesian Tribes
Northern Rhodesian Tribes
Nyasaland protectorate tribes
Portuguese East African tribes
Other
Mixed and Coloured,
other than Bantu
Hottentot Bushman
Hottentot
Koranna
Namaqua
Malay (Cape)
Mixed
Griqua
Mozambique
Chinese
Indian
Other Afghan
American Coloured
Arabian
Creole
Egyptian
Krooman
Malagasy
Mauritian
St Helena
Syrian
West Indian
Zanzibari
Other
More significant were the instructions given to enumerators in this census
‡‡
:
3. The Census embraces ALL RACES of the Population and the following distinctions
will be observed for enumeration and tabulation:
European or White Persons – This applies to all persons of European descent as
defined in the South Africa Act. Europeans are to be regarded as persons of pure
European descent. The offspring of any mixed marriage in which one of the parties
is not of pure European descent are to be regarded as Coloured persons.
Natives – The term “Native” is used to designate all pure-blooded aboriginals of the
Bantu Race, the names of the chief tribes of which will be found at the foot of form
C.4.
Asiatics – The chief Asiatic nations represented in South Africa are British Indians,
Chinese, Japanese, Syrians, Afghans and Burmese.
‡‡
Extended quotes from official census documents are shown in sans-serif typefaces.
Please do not cite or quote without prior permission
Moultrie and Dorrington 8
Coloured persons – The embraces all persons of mixed race, and the Census
classification includes, amongst others, Hottentots, Bushmen, Cape Malays,
Griquas, Korannas, Creoles, Negroes, and Cape Coloured.
4. It will often occur than an Enumerator, especially in the poorer localities, will be
asked for, say, a European form (C. 1) by persons who obviously cannot be
classified as white. In such cases, Enumerators must be instructed to refrain from
giving offence by any comment or question in the presence of the parties
concerned, but to make a private note on the completed forms against the names
of any persons he considers cannot be classed as European, and report the
circumstances to you. Thereafter the particulars in respect of the persons in
question should be transferred to the form or forms applicable to their race
(South Africa 1924: 10, emphasis in bold added)
Evidently, individuals were not given freedom to define their race or ethnicity
subjectively; but a quasi-official, but nonetheless state-sanctioned, process of racial classification
began to be imposed from above. Certainly, one’s race was now not entirely self-reported; the
final say went to the enumerators and their supervisors.
By the time of the 1936 census, fifteen years later, these ideas had shown some accretion.
Segregation was not only enforced in the real world of people and places, but also (rather
comically) in the more abstract and virtual realm of census tabulations:
4. Racial distinctions of the population. The existence in South Africa of three main
racial groups, plus a fourth of mixed origin, means that in effect four separate
censuses are taken simultaneously. As far as tabulation of the results is concerned,
four separate tabulations are actually undertaken, and the tabulation cards are
never mixed.…
Every endeavour is made during enumeration and later during audit prior to
tabulation to place the individuals enumerated to the correct ethnological groups
as defined in paragraph 4 above. The use of four different census forms although
greatly assisting the correct assignment and simplifying the work for machine
tabulation has been criticized in certain areas where the races are very mixed, such
as in the Cape Peninsula. Cases have occurred where parents who are quite willing
to use the correct form for entering their own particulars have objected to using a
separate form for their children who happened to fall within the definition of
“mixed or coloured.” These entries are accepted to avoid giving offence and
afterwards transferred to the correct form either by the Enumerator, the
Supervisor, or in the Census Office before tabulation.…
The following is an extract from the instructions to Census Enumerators:- “It is
admittedly extremely difficult to discriminate by outward appearances, especially
in the Cape Peninsula, and where doubt exists, discreet inquiries should be made
locally to ascertain whether a person of doubtful descent is looked upon locally and
received by his European neighbours as a European”. (South Africa 1938: viii,
emphasis added)
Attempts at “deception”, thus, were treated as before, with enumerators and supervisors
again having the power to reassign people’s race on the grounds of appearance and behaviour.
The 1946 enumeration followed the 1936 approach fairly closely. However, by the time
of the 1951 census, several major changes had occurred, most notably the accession to power of
the National Party in 1948, and the passing of the Population Registration Act of 1950. The Act
was one of the central supports of apartheid policy, creating the racial categories that would later
be used to determine residence, access to education, and a host of other social benefits.
Please do not cite or quote without prior permission
Moultrie and Dorrington 9
Individuals would be classified at birth in terms of the Act, and one’s population group could
only be changed by judgment of the Supreme Court.
However, in spite of the modernist, totalising, quasi-scientific bent of the apartheid state
described by Posel (1991), even by 1960 astute commentators had noted with a degree of
puzzlement the apparent anomaly that lay at the heart of much apartheid policy: that, despite “the
whole pattern of every individual’s life – from the cradle to the grave – [being] circumscribed by
his race … the absence of any uniform basis of race classification is, therefore, all the more
surprising” (Suzman 1960:339). The Population Registration Act of 1950 was supposed to rectify
this, but in his review a decade later, Suzman could still point to a bewildering panoply of racial
and ethnic definitions enshrined in national legislation other than those in the Act. In this regard,
then, West’s (1988:101) argument that the “key to the current classification system lies in the
Population Registration Act” would appear to be overdone, not least because it downplays South
Africa’s pre-apartheid segregationist history. While the Population Registration Act did indeed set
out the codification of ethnic and racial groups in the country, it was, as West himself describes
it, a “farrago of imprecision” (West 1988:103) using terms such as “race”, “class”, “tribe” without
adequate definition, and thereby presupposing that these terms were self-evidently obvious and
unambiguous. As Suzman curtly noted in 1960, the process was an attempt to “define the
indefinable” (Suzman 1960: 367).
What accounts for this imprecision? Several competing explanations can be advanced. It
could simply be the result of the fact that as a social construct, racial definition is of necessity
difficult to codify exactly. Or it could simply reflect incompetence on the part of those tasked
with categorising, or even a degree of reservation about the project. However, most likely it was a
symptom of the extreme arrogance of those in power, presuming, common with Humpty
Dumpty
§§
that the terms employed in the various pieces of legislation were common-sensically
self-evident, irrespective of the legal (in)exactitudes employed.
The 1951 census provided the seed data for the Population Register that emerged in the
wake of the Population Registration Act, and hence the categories that had been used more-or-
less unchanged since 1921 were modified to accommodate the new terminology of the Act.
Races. Throughout this report, four racial groups, namely whites, Asiatics,
Coloureds and Natives, are distinguished, but a fifth group, the Cape Malays, is
shown in tables 1 and 10 only. In all other tables the Cape Malays are included in
the group “Coloureds”. The four racial groups are as follows-
Whites – Persons who in appearance obviously are, or who are generally accepted
as white persons, but excluding persons who, although in appearance obviously
white, are generally accepted as Coloured persons.
§§
“When I use a word,” Humpty Dumpty said, in a rather scornful tone, “it means just what I choose it to mean, neither more nor less.” (Alice in
Wonderland)
Please do not cite or quote without prior permission
Moultrie and Dorrington 10
Natives – Persons who in fact are, or who are generally accepted as members of
any aboriginal race or tribe of Africa.
Asiatics – Natives of Asia and their descendants, mainly Indians and Pakistani, with
a few thousand Chinese, and small numbers of various other Asiatic nationalities.
Coloureds – All persons not included in any of the three groups referred to above.
The great majority of the persons in this group are the persons known as the Cape
Coloured, but persons of mixed white and non-white blood are also included. The
Cape Malays, when not shown separately, are also included in this group.
The last three groups, when combined, are referred to as the non-white group.
The 1951 census formed the basis of the Population Register established under the
Population Registration Act, No. 30 of 1950, and it was therefore necessary to
follow the racial definitions appearing in the Act, which differ somewhat from
those previously used for census purposes. The following aboriginal races are now
classified as Natives: Bushman, Hottentot, Koranna and Namaqua. This accounts for
some apparent inconsistencies in the figures for Natives and Coloureds for 1951 as
compared with 1946 in some magisterial districts, especially in the Cape Western
area. (South Africa 1955a: v)
Reference to explicit instructions to enumerators to “correct” racial classifications in the
1951 census cannot be found in the census volumes. Of course, this does not mean that such
activities were not pursued. However, by the time of the 1951 census, other important pieces of
apartheid legislation (notably the Group Areas Act of 1950, which determined the geographical
areas in which different population groups were allowed to work and, particularly, reside; and the
Prohibition of Mixed Marriages Act of 1949) would most probably have had the (unintended?)
effect of simplifying the process of enumeration that required different forms for different
population groups.
With only minor changes (for example, a relabelling of Native to Bantu to Black), the
categorisation from 1951 (as well as the use of four different census forms, with different
questions) endured until the end of apartheid, and is still in use. One significant addition was
made at the time of the 1970 census, namely a further question to Africans on ethnicity. This
reflected the imminent creation of Bantustans (four of which were declared independent states
between 1976 and 1981) on ethno-linguistic lines, in part a cynical ploy to maintain a fiction of
white domination by fragmenting Africans along “ethnic” lines
***
.
It can thus be surmised that in the context of apartheid South Africa, that the debate referred to
in the previous section about the power of the census to typify and impose a classifying order on
the population is rendered largely irrelevant. Covertly, in the three censuses from Union to 1948,
and overtly in the censuses that followed thereafter, individual agency in defining one’s own racial
identity was, to all intents and purposes, non-existent. This means, again, that individuals’ ability
***
It is of course not surprising to note that no equivalent process of ethnic division was applied to the White population to distinguish those with
English, Dutch, Portuguese etc. ancestry.
Please do not cite or quote without prior permission
Moultrie and Dorrington 11
to consider their race outside the parameters of a very fixed position within South African society
was dramatically curtailed. It therefore becomes exceedingly difficult to sustain an argument that
it was the census that crystallised racial identities in apartheid and segregationist South Africa;
rather than being a tool for locating self-reported identity into a small number of categories, the
census directly and indirectly, explicitly and implicitly, required of respondents to simply verify
their pre-assigned racial identity.
Perhaps paradoxically, then, it would appear that it is precisely in the situation which a
priori might have been thought to be the most striking example of a census “naming categories
into existence” that this process is weakest; not least because that process had happened
indirectly elsewhere.
4. LOOKING TO THE FUTURE: COLLECTION OF DATA ON RACE, POST-APARTHEID
The transition to democracy in South Africa in the 1990s presents some challenges to
demographers and social statisticians. In the first instance, the repeal of the Population
Registration Act in 1991 removed the legal basis for classification of the population by racial or
ethnic group. At the same time, the coding of country of birth and population group in the
national identity number that all South Africans must by law have, was dropped, and racial
classification was removed from both the birth and death certificates. In the absence of a set of
legal definitions (no matter how flawed), any classification on the basis of race must now be self-
reported. This raises important moral and legal questions as to how to enforce a racial
classification required by any corrective legislation such as that to enforce employment equity, for
example.
The question of whether social statistics in a post-apartheid South Africa should still collect
information on race was a concern during the planning for the 1991 South Africa Census, the last
one before democracy. It has not been possible to identify the exact source of the dispute as to
whether that census should collect information on population group, but – shortly before the
census went into the field – the (apartheid) state’s Bureau of Information (and, interestingly, not
the Central Statistical Services) saw it necessary to issue a communiqué entitled “Statistics on
Population Group Essential” (Zuberi and Khalfani 1999), in anticipation of the repeal of the
Population Registration Act. It would appear that the conclusion reached then, that it was
desirable to continue to collect information on population group, has, somewhat surprisingly, not
been challenged since; as we show below, there has been almost no change in the nature of the
information on race or ethnicity sought by the South African state from its citizens and residents
following democracy.
Please do not cite or quote without prior permission
Moultrie and Dorrington 12
It is interesting to track the evolution of the questions on population group over the three
censuses conducted in 1991, 1996, 2001, and the most recently conducted national survey (the
2007 Community Survey
†††
). The question asked, and the universe of possible responses, in these
surveys is as follows:
Table 2 Classification of population group in official South African censuses since 1991
Survey Question Universe
1991 Census Population group White, Coloured, Asian, Black
1996 Census How would <the person> describe him/herself? African, Coloured, Indian, White
2001 Census How would <the person> describe him/herself in terms of population group? African, Coloured, Indian, White, Other
(Specify)
2007 CS How would <the person> describe him/herself in terms of population group? African, Coloured, Indian, White
Thus, the 1991 Census had no qualms about asking about population group directly, without the
suggestion that individuals had a choice. The Population Registration Act was still in force (it was
only repealed later in the same year), and so the classificatory system employed still had legal (if
not moral) force. Evident uncertainty about the rectitude of asking the question is evident in the
1996 census, where no mention of race, or population group, is made in the question at all, but
still contrives to ask about individuals’ population group, albeit changing the order of the groups
in a clear effort to indicate that the question was different from those asked in the past. The
question was modified slightly in the 2001 Census, clearly making use of the previous
terminology “population group” to indicate what the question was aiming for, and retained in the
2007 CS.
However, in stark contrast to instructions in early apartheid censuses, enumerators in the
1996 and 2001 censuses (as well as in the 2007 CS) were under strict instructions not to challenge
responses about population groups. In the 2001 Census,
Enumerators were instructed to accept whatever response was given, even if they
did not agree with it. Enumerators were instructed to ask this question about each
person even if the population group seemed obvious. Enumerators were also
alerted to the fact that persons of different population groups could form part of
the same household. (Statistics South Africa n.d. (c. 2004): 12)
The CS enumerator’s manual for that question was similar:
POPULATION GROUP: How would (the person) describe him/herself in terms of
population group?
Ask for everybody even if the population group seems obvious. Remember that
persons of different population groups do sometimes form part of the same
household, so you cannot assume the population group of any household member.
Accept the response that is given even if you do not agree and under no
circumstances may the response be queried. (Statistics South Africa 2006: 48,
emphasis in the original)
However, of greater interest is that, while instructions to enumerators have been explicit
in their determination not to ‘reclassify’ respondents on the basis of perceptions or prejudice,
†††
The Community Survey (CS) interviewed almost a million South Africans in a sample survey conducted in lieu of a census in 2006. The next
full census is scheduled for 2011.
Please do not cite or quote without prior permission
Moultrie and Dorrington 13
there is evidently little preciousness about doing so at the data cleaning and processing stage. In
the 1996 census, no option was given to respondents to classify their population group as
anything other than one of the four apartheid-era categories:
During Census ’96 when some people whom identified as Griquas requested to be
identified as a separate population group (sic). At short notice, instructions were
changed to allow interviewers to write ‘5 Griqua’ on the questionnaire and these
responses were coded separately. However, the number of people identifying
themselves as Griqua was small and represented an under-identification of the
number of people who would have been identified as Griquas if there had been
such a category on the questionnaire to indicate that this was a valid response. In
the SuperSTAR database, people identifying as ‘Griqua’ have been included with
‘Unspecified’. Data on the Griquas identified in the Census is available separately as
a special request but users should be aware of its limitations. (Statistics South
Africa n.d. (c. 1998))
The 2001 census form made provision for a category of “Other - Specify”. However, in
the editing process, the following was observed:
The number of responses for ‘other’ was so few that results for this category are
largely unusable in tables produced at lower geographical levels. For this reason,
the category ‘Other’ was removed and re-allocated amongst the remaining valid
values during the editing process. The raw data will, however, contain the ‘Other’
category. Raw data will be available only in the product containing the sample
database (Statistics South Africa 2003: 147)
Presumably the option of classifying oneself as “Other” was dropped from the 2007 CS
as a result of these problems. However, these editing processes raise doubts about individual
agency in reporting their population group.
In 1996, 0.92 per cent of the population did not record their population group (more, of
course, might have elected consciously to mis-report their response – for example Coloured, and
even Afrikaners and liberal whites, as African). These non-responses were classified as
“Unspecified”. Presumably many of these were the Griqua referred to above. In 2001, the option
of “Other - Specify” was offered, and then removed from the formal tabulations. In this census,
however, non-response to this variable was not allowed at the editing stage. Accordingly, a
complex set of edit rules was devised to attribute one of the four population groups to people
who did not respond to the question. A similar procedure was applied in 2007. The question that
then arises is this: if a respondent is not allowed to offer a response of “not stated”, or –
preferably – “refuse to answer”, how free is the decision to report one’s population group? We
would argue that restricting one’s answer to a very narrow range of categories that draw their
meaning from apartheid legislation does not constitute a freedom to report one’s population
group according to one’s own moral position on the matter and avoids the barometer of popular
acceptance of such classification that would have been provided by the options not to state or to
specify another category. Is the situation any better than it was in 1991? Yes, but only marginally.
It does force us, however, to consider the uses to which this information is now put, and to
Please do not cite or quote without prior permission
Moultrie and Dorrington 14
understand why such a tight rein is kept on the choices offered to people of how to classify
themselves.
Should we continue to capture data on race or population group? We think the answer is – for
the foreseeable future – a qualified ‘yes’. The qualification is necessary since an absolute position
returns us to where we started; to the reification of race, and racial difference. As suggested in the
introduction, the overwhelming reason advanced for continuing to collect information on
population group is to monitor and track progress in redressing the iniquities of apartheid social
and economic policies. There is a lot of merit to this argument. Given South Africa’s past, if one
were to not collect data on race, this would be “tantamount to denying history. It would mean
denying racial oppression and its consequences … However arbitrary this definition, it is
imperative to maintain continuity, in order to measure progress in eliminating the racially defined
inequalities of the past” (Head 1997:4). However, being self-reported, the continued collection of
data on race will no doubt deviate systematically over time from its original intention of seeking
to capture information on the population’s racial composition had the apartheid definitions still
remained in force, an observation that enjoins us not to fall into the trap of again reifying race.
Under what conditions should that qualification hold? We argue that there are several.
First, if race is still significantly associated with demographic, social and economic outcomes and
capabilities, after controlling for a slew of other factors, this suggests that there is an unmeasured
effect associating race with the outcome. Second, and this may be particularly pertinent in the
case of a country with limited vital registration, official and other data, race offers an important
proxy for a host of other socio-economic differences. Were the first condition to be met, we may
seek to replace race, or population group or ethnicity, with the combination of socio-economic
variables that render race (or its equivalents) irrelevant. All too frequently this may not be the
case insofar as data relating to those other socio-economic variables may not be captured on
administrative systems. A case in point relates to data on HIV, such as prevalence, an important
determinant of mortality if not many other variables of demographic interest, one rarely has
much socio-economic information about the individual other than population group, age, years
of schooling and possibly one other variable.
Of course, for a society as racially conditioned as South Africa’s, the analysis and use of
data on population group is not restricted to the evaluation of elementary demographic
differentials in birth, death and population growth rates. Indeed, population group is usually one
dimension in bi- and multi-variate tables of social statistics. As a consequence, the use of a
population group variable has found uses, and become important, in a great many political, social,
economic and developmental debates. Thus, for example, the variable is used in attempts to
measure the impact of efforts to remedy the social iniquities of the past (by means of health
Please do not cite or quote without prior permission
Moultrie and Dorrington 15
metrics, as well as access to welfare and basic services); economically (in terms of measures of
employment, and ownership of assets); as well as for setting regulatory or developmental targets
(for example, in the case of employment equity policies).
However, in all instances mentioned (and particularly so in the case where population
group is used to define measureable goals), the great unasked question in the South African polity
is what this variable is measuring now that population group is entirely self-defined.
5. CONCLUSIONS
The South African story offers several challenges to the dominant narratives of race, power and
demography. First, there is a need to juxtapose the arbitrariness of classification systems in
South-East Asia cited by Hirschmann (with their amalgam of race, ethnicity, linguistics and
geography), with the fairly rapid accretion of state-defined identity in South Africa. While indeed
South African ethnographers were not above using race in some cases and ethnicity in others
(most obviously in their agglomeration of ‘whiteness’ as race while fracturing ‘Africanness’ along
ethnic lines to maintain the fiction of white numerical dominance), one has to question the
relevance to South Africa of the argument advanced by Appadurai (1993) and others that the act
of counting and census-taking was central to the colonial project.
The imprecision in the terminology used to define race, and its manifest inability to
accurately enumerate its population, strikes at the very heart of an enduring debate about the
nature of the apartheid state. Moultrie (2005) has traced the evolution of government rhetoric on
the racial composition of South Africa from 1900 through to the 1970s, paying particular
attention to the origins, manifestations and of notions of “swamping” and “race suicide” in
official discourse, and has shown its centrality to several key pieces of apartheid planning (notably
the findings of the Tomlinson Commission published in 1955 (South Africa 1955b)). Yet, the
inability of the apartheid state to either conclusively classify or count its subordinate populations
for most of the segregation and apartheid eras stands in stark contrast to the “mania for
measurement” described by Posel (2000) and the primacy afforded to ‘scientific’ evidence by the
“grand tradition” of Commissions of Enquiry described by Ashforth (1990).
Surely, if the actual numbers were important, greater efforts would have been made in
this regard? Instead, one is forced to conclude that it was the act of counting, not the realised
total (which was known to be wrong – but in the desired direction, i.e. underestimating the
number of “non-whites”) that was important; that – as Posel has observed – the process of
counting lent a scientific veneer to policies that were – in essence – the antithesis of science. It is,
therefore, a matter of some curiosity, given the theoretical debate about the importance of the
process of counting in shaping colonial identities, as well as the well-documented concerns about
the white population being swamped, that the counting of African South Africans was done in
Please do not cite or quote without prior permission
Moultrie and Dorrington 16
such a desultory way from 1911 to the last apartheid-era census in 1991. How does one square
the portrayal of a racist, totalising, modernist state with that of one that could not accurately
count the population it was trying to subjugate? This in a matter for further research and thought,
although we would suspect that the answer lies in one or more of the following explanations: a
shortage of administrative capacity within government; ideological battles within the ruling party
over whether idealistic or pragmatic apartheid should be pursued which undermined the utility of
a ‘scientific’ consideration of the matters at hand; a mindset that paid particular attention to
where the subordinated population should NOT be (in terms of apartheid’s spatial framework),
with a concomitant carelessness about counting the population in the areas where was allowed to
be; and – almost certainly – the difficulty of counting “non-people” including a general
unwillingness of Africans to be counted, especially if they were living on the fringes of apartheid
spatial planning.
But several other interesting questions present themselves which relate both specifically
to South Africa, as well as to the broader debate around the importance and desirability (or
otherwise) of collecting information on population group, ethnicity or race.
Having sought to answer the question of “should we collect” as we have above also
answers, in part, the question as to whether it would be preferable instead to collect sufficient
socio-economic data instead to allow some form of class-based analysis to emerge. Theoretically,
the idea has much merit; doing so would move away from the reification of a presumptive link
between a proxy for socio-economic distinction, and that distinction measured directly. But is it
really that much of an improvement? First, reducing socio-economic differences down to a
composite variable (for sake of argument, let’s call this a measure of social class) runs the risk of
being reified and fossilised in much the same way as race, ethnicity and population group have in
the past. Social class, as with most other sociological variables, is fluid over time and also has far
less permanence as a feature of an individual than does race or ethnicity. Any move to a summary
measure of social distinction has to guard against this stasis – but then, if it is constantly
changing, how does one examine trends and changes diachronically at all? In practice,
furthermore, the proposal is fraught with problems. Even in relatively wealthy societies, to collect
the range of data on socio-economic variables required to proxy accurately for race (and even
more so in South Africa) may be logistically and financially impossible.
Finally, we wonder at the relative durability of apartheid racial categories since 1991.
Perhaps, following Fanon (1967), it is not surprising that there has been relatively little
discontinuity in the data on population composition beyond that which can be explained by
emigration and differential underenumeration: concepts of race and racial distinction and
differencing are firmly embedded in the South African psyche.
Please do not cite or quote without prior permission
Moultrie and Dorrington 17
The legislative framework that supported apartheid-era classifications (in particular, the
Population Registration Act of 1950) has been repealed. And, despite the continued and common
usage of these classifications in all spheres of public life (including in censuses and surveys), there
is no longer a legal definition of what it means to be African, White, Indian or Coloured. In effect
then, population group is increasingly a self-determined identity (albeit tightly controlled by the
state), with all the attendant problems that self-reported variables pose for demographers and
social statisticians. At the same time, measurement of progress towards achieving equality in all
spheres of civic life in South Africa requires a collective use of coherent and consistent
definitions of identity as defined by the previous dispensation, the ongoing quiescence of the
population to continue to use and apply these terms to oneself is required. Such quiescence is
neither guaranteed nor enforceable without legislation. Given the country’s history of race-based
legislation, we see this as highly improbable.
REFERENCES
Anderson, Benedict. 1991. Imagined Communities: Reflections on the Origins and Spread of Nationalism.
London: Verso.
Appadurai, Arjun. 1993. "Number in the colonial imagination," in Breckenridge, Carol A. and
Peter van der Veer (eds). Orientalism and the Postcolonial Predicament. Philadelphia PA:
University of Pennsylvania Press, pp. 314-339.
Ashforth, Adam. 1990. The Politics of Official Discourse in Twentieth-Century South Africa. Oxford:
Clarendon Press.
Christopher, Anthony J. 2002. "'To define the indefinable': Population classification and the
census in South Africa", Area 34(4):401-408.
Christopher, Anthony J. 2005. "Race and the census in the commonwealth", Population, Space and
Place 11(2):103-118.
Fanon, Frantz. 1967. The Wretched of the Earth. London: Penguin.
Hacking, Ian. 1990. The Taming of Chance. Cambridge: Cambridge University Press.
Hacking, Ian. 1991. "How should we do the history of statistics?," in Burchell, Graham, Colin
Gordon and Peter Miller (eds). The Foucault Effect: Studies in Governmentality. Chicago:
Chicago University Press, pp. 181-195.
Hacking, Ian. 1999. The Social Construction of What? Cambridge MA: Harvard University Press.
Head, Judith. 1997. Collecting health statistics in the New South Africa: Which category should be used,
"race" or class? Unpublished mimeo. Cape Town: University of Cape Town.
Hirschman, Charles. 1987. "The meaning and measurement of ethnicity in Malaysia: An analysis
of census classifications", Journal of Asian Studies 46(3):555-582.
Kertzer, David I. and Dominique Arel. 2002. "Censuses, identity formation, and the struggle for
political power," in Kertzer, David I. and Dominique Arel (eds). Census and Identity: The
Politics of Race, Ethnicity, and Language in National Censuses. Cambridge: Cambridge
University Press, pp. 1-42.
Krieger, Nancy. 2000. "Refiguring "race": epidemiology, racialised biology, and biological
expressions of race relations", International Journal of Health Services 30(1):211-216.
Moultrie, Tom A. 2005. "Racism and reproduction – Population rhetoric in South Africa, 1900-
1974", African Studies 64(2):217-242.
Nobles, Melissa. 2002. "Racial categorization and censuses," in Kertzer, David I. and Dominique
Arel (eds). Census and Identity: The Politics of Race, Ethnicity, and Language in National Censuses.
Cambridge: Cambridge University Press, pp. 43-70.
Please do not cite or quote without prior permission
Moultrie and Dorrington 18
Posel, Deborah. 1991. The Making of Apartheid, 1948-1961: Conflict and Compromise. Oxford:
Clarendon Press.
Posel, Deborah. 2000. "A mania for measurement: statistics and statecraft in the transition to
apartheid," in Dubow, Saul (ed). Science and Society in Southern Africa. Manchester:
Manchester University Press, pp. 116-142.
Scott, James C. 1998. Seeing Like a State: How Certain Schemes to Improve the Human Condition Have
Failed. New Haven, CT: Yale University Press.
South Africa, Union of. 1910. House of Assembly Debates vol 1. Cape Town: Government Printing
and Stationery Office.
South Africa, Union of. 1911. Statutes of the Union of South Africa, 1935. Pretoria: Government
Printer.
South Africa, Union of. 1912. Census of the Union of South Africa, 1911. Report No. UG32-1912.
Pretoria: Government Printing and Stationery Office.
South Africa, Union of. 1924. Third Census of the Population of the Union of South Africa, Enumerated
3rd May 1921. Report No. UG37-1924. Pretoria: Government Printing and Stationery
Office.
South Africa, Union of. 1935. House of Assembly Debates vol 24. Cape Town: Government Printing
and Stationery Office.
South Africa, Union of. 1938. Sixth Census of the Population of the Union of South Africa, Enumerated
5th May 1936. Volume I. Report No. UG21-1938. Pretoria: Government Printer.
South Africa, Union of. 1955a. Census of the Population of the Union of South Africa, Enumerated 8th
May 1951. Volume I. Report No. UG42-1955. Pretoria: Government Printer.
South Africa, Union of. 1955b. Summary of the Report of the Commission for the Socio-Economic
Development of the Bantu within the Union of South Africa. Pretoria: Government Printer.
Statistics South Africa. 2003. Computer Edit Specifications for Census 2001 Report 03-02-43 (2001).
Pretoria: Statistics South Africa.
Statistics South Africa. 2006. Community Survey 2007: Enumerator’s Manual. Pretoria: Statistics South
Africa.
Statistics South Africa. n.d. (c. 1998). Metadata for the Persons File (Census 1996). Pretoria: Statistics
South Africa.
http://www.statssa.gov.za/census01/Census96/HTML/Metadata/Metadta/comm_prof
iles_dbases_toc.htm Accessed 28 October 2007
Statistics South Africa. n.d. (c. 2004). Census 2001 Metadata. Pretoria: Statistics South Africa.
http://www.statssa.gov.za/census01/html/Persons.pdf Accessed 28 October 2007.
Suzman, Arthur. 1960. " Race classification and definition in the legislation of the Union of
South Africa - 1910-1960", Acta Juridica:339-367.
West, Martin. 1988. "Confusing categories: Population groups, national states and citizenships,"
in Boonzaier, Emile and John Sharp (eds). South African Keywords: The Uses and Abuses of
Political Concepts. Cape Town: David Philip, pp. 100-110.
Zuberi, Tukufu. 2001. Thicker Than Blood: How Racial Statistics Lie. Minneapolis: University of
Minnesota Press.
Zuberi, Tukufu and Akil K. Khalfani. 1999. Racial classification and the Census in South Africa, 1911-
1996. ACAP Working Paper No. 7. Philadelphia: African Census Analysis Project.
