South Africa Distance Learning Project





Measuring Differences by Ethnicity in Optimism towards the New Government in Post-Apartheid South Africa, 1993

by Chester Choi



Overview of Apartheid
Prior Research & Possible Outcomes
Statistical Methods



This module examines whether there are significant differences in the optimism in whether a new government elected in the 1994 elections would make life better among various ethnic groups. This question is motivated by the history of apartheid and the attempts of the National Party government to implement "divide and rule" tactics to maintain control over South Africa. Apartheid policies may have directly or indirectly created stronger or weaker ethnic identities that affected how an individual viewed the post-apartheid political environment relative to other ethnic groups. I will first provide a brief (and incomplete) history of the system of Apartheid and the policies that accentuated the divisions between African ethnic groups, white ethnic groups, and between Africans, Coloureds, and Indians. Based on the historical background and past research, I will briefly discuss what the possible outcomes we might expect to find. I will then describe the statistical analysis performed using the results from the SALDRU Household Survey and provide an explanation of the results. The survey was undertaken during the nine months prior to the first democratic elections in April 1994 in which the African National Congress (ANC), led by Nelson Mandela, won in a landslide victory.


Overview of Apartheid

The policy of apartheid – simply meaning apartness – has its roots in Afrikaner nationalism dating back to the 17th century, when Dutch settlers first established a refreshment station at the Cape of Good Hope for the Dutch East India Company in 1652. The Afrikaner identify was strengthened during the era of British rule, as Afrikaners resented foreign interference and desired to rule the country on their own terms. For many Afrikaners, ethnic identity was more important than occupational and class differences. Afrikaners were worried about the state of race relations and nearly all believed that the state should do more to maintain white supremacy and the "purity of the white race"(Thompson p. 185). As Afrikaner nationalism gained momentum, politicians and intellectuals began formulating a policy that would secure the interests of the Afrikaner people. In the 1930's, Afrikaner intellectuals coined the term apartheid as a label for these policies.

The system of apartheid was supported by four main ideas (Thompson p. 190):

1. The population of South Africa was comprised of four main racial groups – White, Coloured, Indian, and African - with their own distinct culture
2. Whites were entitled to have absolute control over the state
3. White interests should always prevail over non-white interests and the state was not obliged to provide equal facilities to each racial group
4. The white racial group formed a single nation, with Afrikaans and English-speaking components, while Africans belong to their own distinct nations based on their ethnicity.

The next few sections discuss how the apartheid system created ethnic divisions within the racial groups and discusses the implications for the results of the module.

Ethnic Divisions between Whites

Afrikaners and English speaking whites share a tumultuous history in South Africa. Afrikaners have maintained a distinct identity throughout the period of British rule and always desired to rule South Africa on their own terms. The defeat of the Afrikaner republics in the Second Boer War (1899-1902) led to the formation of the Union of South Africa four years later, which was part of the British Commonwealth. During this period, the "racial question" centered around the ethnic cleavage between Afrikaners and English speaking whites. While the racial question was posed in ethnic terms, it also had a class element as well. In the early twentieth century, people of British origin virtually monopolized the entrepreneurial, managerial, and skilled positions in every sector of the economy except agriculture, whereas many Afrikaners were impoverished (Thompson p. 155). In towns, Afrikaners were concentrated in the lowest-level white occupations: unskilled laborers, miners and factory workers, teachers, and junior civil servants. The average English-speaking white was twice as wealthy as the average Afrikaner. In terms of wage income, Afrikaners made only 47 percent as much as English-speaking whites.

After the National Party took power in South Africa in 1948, the new government Afrikanerized every institution in the country. By 1977, about 540,000 Whites were employed in the public sector and Afrikaners occupied more than 90 percent of the top positions (Thompson p. 199). The vast majority of the white bureaucrats were ardent supporters of apartheid. The government also assisted Afrikaners to close the wage gap between themselves and English-speaking white South Africans in the private sector. The government directed official business to Afrikaner banks and allotted valuable state contracts to Afrikaners. Afrikaner business people channeled Afrikaner capital into ethnic banks, investment houses, insurance companies, and publishing houses. By 1976, the wage gap closed with Afrikaners making 71 percent as much as English-speaking whites and continued to decrease as time went on (Thompson p. 188). In addition, the National Party achieved a major ethnic objective in 1961 by passing a referendum that made South Africa a republic, separating itself from British control. Thus, while apartheid policies defined the white South African nation as inclusive of English speaking whites, it ensured that Afrikaners became level with English-speaking whites on economic, social, and political terms.

During the era of Apartheid, white opposition to apartheid was divided along ethnic lines, with opposition coming mainly from the English speaking population. For instance, leaders of all the white South African churches, except the Dutch Reformed churches issued statements criticizing apartheid shortly after the National Party¡¯s ascension to power in 1948. The English speaking universities, particularly the Universities of Cape Town and the Witwatersrand, were the foci of opposition to apartheid. The National Union of South African Students (NUSAS) organized a series of demonstrations in 1959 against the closures of the established black universities and 1966 arranged a visit by Senator Robert Kennedy, who denounced apartheid. NUSAS was responsible for introducing generations of predominately English speaking white students to critical thinking about South African politics and society (Thompson p. 205).

There was no powerful white economic interest that openly opposed apartheid before the late 1970's (Thompson p. 206). It was not until the economy fell into a long period of recession during the 1980's that business leaders supported reform. Foreign pressure to end apartheid and the general instability from black resistance led many multinational corporations to pull out their investments from South Africa and the poor education system for Africans, Coloureds, and Indians left the country with a shortage of skilled labor to jump-start the economy.

Ethnic Divisions between Black Africans

Prior to the implementation of apartheid, black Africans had already suffered from laws that promoted segregation and discrimination in favor of whites. Two of the more significant laws were The Natives Land Act, No 27 of 1913 and The Natives (Urban Areas) Act of 1923. The former made it illegal for blacks to purchase or lease land from whites. This restricted black occupancy to less than eight per cent of South Africa's land. The latter laid the foundations for residential segregation in urban areas. Blacks also suffered from poor wages relative to whites, were not allowed to organize strikes against employers, and were barred from voting and representation in the national and provincial assemblies.

After the National Party took over the leadership of the country in 1948, segregation and discrimination became more intense and systematic. Afrikaners were the clear minority in South Africa, and thus created a "divide and rule" policy, which was aimed at ensuring white survival and hegemony by dividing the nonwhite population along racial and ethnic lines (Henrard p. 37). Thus, the racial majority was divided into a host of minority groups, which could no longer pose a threat to the white minority.

The primary tool of creating ethnic divisions among black South Africans was the creation of “homelands.? Homelands were meant to be independent African nations administered under white tutelage by a set of Bantu authorities, consisting mainly of hereditary chiefs. In its Homeland, an African nation was to “develop along its own lines?and its citizens were granted all the rights that they were denied in the rest of South Africa. Government propaganda compared the creation of the Homelands to the decolonization of the European empires in tropical Africa (Thompson p. 191).

In 1971, the Bantu Homeland Constitution Act enabled the government to grant independence to any Homeland. Transkei was first to become officially independent in 1976, followed by Bophuthatswana in 1977, Venda in 1979, and Ciskei in 1981 (Thompson p.191). A full list of the Homelands and their representative ethnic group is provided below.

• Bophuthatswana (Tswana) - declared independent on 6 December 1977
• Transkei (Xhosa) - declared independent on 26 October 1976
• Ciskei (also Xhosa) - declared independent on 4 December 1981
• Venda (Venda) - declared independent 13 September 1979
• Gazankulu (Tsonga)
• KaNgwane (Swazi)
• KwaNdebele (Ndebele)
• KwaZulu (Zulu)
• Lebowa (Northern Sotho)
• QwaQwa (Southern Sotho)

South Africa

In order to speed the process of gentrification, the government tried to forcibly move as many black Africans as possible, except for those whom white employers needed as laborers. To limit the influx of black Africans into towns, pass laws were created that prohibited them from visiting urban areas for more than seventy-two hours without a special permit and by authorizing the arrest of any black African who could not produce the required documents. Every year, more than 100,000 people were arrested under the pass laws (Thompson p. 193). The government also removed black African squatters from unauthorized camps near the cities, placing those who were employed in segregated townships and sending the rest to the Homelands or to farms where white farmers needed labor. Lastly, the government began to eliminate "black spots" in rural areas, forcibly taking away land owned or occupied by Africans in white areas (Thompson p.193). African farmers were often left with no choice but to move to the Homelands since they were not allowed to enter urban areas.

At its peak, over 50 percent (12 million) of the black African population resided in the Homelands. While the goal of the government was to completely eliminate the black African population from the white areas, white capitalists and farmers needed black African labor. Under the apartheid system, the condition of the Homelands deteriorated and they were completely dependent on subsidies from the central government. As conditions in the Homelands became worse, it created economic incentives for black Africans to leave South Africa entirely, either as migrant laborers or permanently.

Besides their common victimization from apartheid, black Africans had varied experiences. In urban areas, African communities resisted the government's attempts to create divisions along racial and ethnic lines. For example, they ignored the government's attempt to carve up the townships into ethnic divisions; they married across ethnic lines; and members of the younger generation identified themselves as Africans rather than as Xhosa, Zulu, Sotho, Pedi, or Tswana (Thompson p. 201).

After the 1994 election, most of the Homelands were peacefully reincorporated into South Africa. However, several homelands had leaders that challenged the ANC's leadership of South Africa and demanded independence following the end of apartheid. There was some resistance from the local elites, who stood to lose out on the opportunities for corruption provided by the homelands. The dismantling of the homelands of Bophuthatswana and Ciskei was particularly difficult, with South African security forces having to intervene in the latter in March 1994 to defuse a political crisis. There was also a strong ethnic movement among the Zulus under the Inkatha Freedom Party led by Chief Mangosuthu Buthlezi. The ANC made concessions prior to the election and in the new constitution to help appease the concerns among the tribal groups.

Coloureds and Indians

Indians and Coloureds also suffered from discrimination during the apartheid period. The government accentuated racial differences, favoring Coloured and Indians over Africans. However, they still were not granted equal rights to whites. In the cities, the government transferred large numbers of Coloureds and Indians from land they had previously occupied to new segregated townships. Under the Group Areas Act (1950), the government divided urban areas into zones where members of one specified race along could live and work (Thompson p. 194). In the constitutional reforms made in 1983, the government allowed the Coloured and Indian minorities limited participation in separate and subordinate Houses in a tri-cameral Parliament (a development which enjoyed limited support). Constitutional reform also created separate and unequal services for Coloureds and Indians, creating separate departments for education, health and other public services. While Coloureds and Indians played an important role in the resistance against apartheid, both groups tended to vote for the white-led National Party in the 1994 election.


Prior Research and Possible Outcomes

Based on the history of the apartheid polices, one would expect to see different levels of optimism in the new government's ability to make life better between the different ethnic groups. Among the white ethnic groups, Afrikaners should be less optimistic about the new government since they are the ethnic group that benefited most from apartheid. Among the black ethnicities, there may be significant differences within ethnic groups that initially resisted incorporation into the new South Africa. On the other hand, the collective discrimination faced by all African ethnicities may have led to the deterioration of ethnic identities, leading to little or no variation among African ethnic groups. For Coloureds and Indians, their optimism towards a new government should be higher than white ethnic groups and lower than African ethnic groups based on their voting patterns.

Some of these predictions are supported by academic research regarding the political environment in South Africa around 1994. Academics were questioning whether post-apartheid South Africa was truly sustainable. Some predicted that ethnic conflict would replace racial divisions in post-apartheid South Africa. Analysts thought that once in power, the ANC would splinter along ethnic lines, similar to what happened to the Zambian Movement for Multiparty Democracy in the 1980s (Piombo, p. 450).

There has also been empirical research performed that analyzed differences along ethnicity lines in how effective the Truth and Reconciliation Commission (TRC) has been in bringing forth the truth of events during apartheid and bringing about reconciliation. Afrikaners found the TRC to be less effective than English-speaking White participants and much less than Xhosa participants (Vora and Vora p. 447). While this study is addressing a different question and only analyzes three ethnic groups, it does shed light on differences of how different ethnic groups view post-apartheid society. Thus, the quantitative analysis performed in this module could provide researchers with valuable insights as to how optimistic different ethnic groups were in the ANC making significant changes in South African society in 1994.


Statistical Methods

To conduct the statistical analysis, I will use the SALDRU household survey data found at The commands that I will be performing correspond to the statistical software STATA.

The first step is to identify the variables that will be used in the analysis. The dependent variables of interest are "Do you think the situation of your household will get better, stay the same, or get worse" if a new government is elected. An easy way to identify these variables is to use the "lookfor" command in STATA. Using "lookfor government" will identify the variable "new_govt."

The next step is to identify the independent variable of interest: ethnicity. When searching for ethnicity with the "lookfor" command, one will realize that there is no ethnicity variable. This variable will have to be constructed. A fairly good proxy for ethnicity is the survey participant's primary language and his/her race. By creating this variable using race and language, we do lose some observations. For example, "Other". I have decided to drop these observations. Using the "gen var" command, we can create a new variable for ethnicity using the "lang_cod" and "race" variables. We can see how the sample is broken down by ethnicity by using the "tab" function.

tab ethnicity

Ethnicity Freq. Percent Cum
---------- ---------- ---------- ----------
Aricaner | 2,488 5.84 5.84
English | 1,417 3.33 9.17
Other White | 50 0.12 9.28
Xhosa | 7,743 18.18 27.46
Zulu | 10,347 24.29 51.75
Tswana | 3,742 8.78 60.53
North Sotho | 5,277 12.39 72.92
South Sotho | 2,950 6.92 79.84
Venda | 762 1.79 81.63
Tsonga | 1,756 4.12 85.75
Swazi | 1,097 2.57 88.33
Ndebele | 541 1.27 89.60
Coloured | 3,308 7.76 97.36
Indian | 1,124 2.64 100.00
---------- ---------- ---------- ----------
Total | 42,602 100.00

Now that the dependent and independent variables of interest have been identified, there are a couple of additional steps before we can run the first regression. To simplify the analysis, I will be using a linear probability model to measure differences in optimism about the new government by ethnicity. To do this, we need to re-code the dependent variable into a binary variable, 0 or 1 to create a response probability. By doing this, every ethnicity's probability of being either optimistic or pessimistic will add to 1. To see how the "new_govt" is currently coded, we can use the "tab" function in STATA.

tab new_govt

8:effect |
of new |
government | Freq. Percent Cum
---------- ---------- ---------- ----------
-4 | 3,240 7.37 7.37
-3 | 94 0.21 7.59
-1 | 120 0.27 7.86
01-better | 24,510 55.77 63.63
02-same | 6,658 15.15 78.78
03-worse | 9.327 21.22 100
---------- ---------- ---------- ----------
Total | 43,949 100.00

We can re-code the ¡°new_govt ¡° into a binary variable so that they are coded as follows:

0 – life will stay the same or get worse under the new government
1 – life will be better under the new government

We also need to deal with the responses for the "-4", "-3" and "-1" responses for "new_govt." After investigating the survey, one cannot make a clear deduction as to whether they should be coded as "better" or "worse." I decided to replace these responses as missing. To re-code the variable, use the "replace" function on STATA. The new binary variable is called "lpnew_govt." When the new "lpnew_govt" variable is created, we see that 60.5% of the entire country believes that life will improve under a new government. This is not surprising since black Africans are the racial majority in the country.

tab lpnew_govt

| Freq. Percent Cum
---------- ---------- ---------- ----------
worse or same | 15,985 39.47 39.47
better | 24,510 60.53 100
---------- ---------- ---------- ----------
Total | 40,495 100.00

The last step is to whether these variables are household or individual level variables. To do this, we can "list" the variables by household ID. First, we need to sort the data by "ethnicity."

Sort hhid
List hhid ethnicity new_govt

When we do this, it is clear that lpnew_govt is a household level variable. This means that there is one response per household, no matter how many people live in that household. Thus, we must make an adjustment to only use one response per household. If we did not make this adjustment, we would be biasing our results by putting a heavier weight on the responses from households that have more members. We will make this adjustment when running the regression.

Bi-variable regression

We can now run a bi-variable regression to analyze differences in the probability of being optimistic about a new government by ethnicity. We will use the "reg" command in STATA to do this. When running a regression, the dependent variable comes first, and the independent variable is second: reg "dependent variable" "independent variable." As mentioned previously, the dependent variables is a household level variable. Thus, to limit the observations to one per household, I am only going to look at the head of the household. The reason I am looking at the head of household is that later, I will be running a multi-variable regression to control for other potential confounders, such as education level. Since education is an individual level variable, I want to account for the education level of the primary income earner in the household. This usually corresponds to the head of household in the survey. The "rel_head" variable represents the relationship to the head of household in the survey. Using the "if" command, we can limit the responses to those of the head of household "1" or the absent head of household "2".

. xi: reg lpnew_govt i.ethnicity if rel_head==1|rel_head==2

Source SS df MS Number of obs = 7383
------ ----- ----- ----- F(13, 7369) = 171.91
Model 417.797881 13 32.1382985 Prob>F = 0.0000
Residual 1377.63514 7369 .18695008 R-squared = 0.2327
------ ----- ----- ----- Adj R-squared = 0.2313
Total 1795.43302 7382 .243217695 Root MSE = .43238
lpnew_govt Coef. Std.Err. t P>t 95% Conf Interval
English | .0607781 .0283371 2.14 0.032 .0052292 .116327
Other White |.0645363 .1062109 0.61 0.543 -.1436674 .27274
Xhosa |.6743 .020631 32.68 0.000 .6338573 .7147427
Zulu | .4924723 .0203914 24.15 0.000 .4524994 .5324452
Twsana | .7002768 .0231734 30.22 0.000 .6548502 .7457033
North Sotho | .7014554 .0222958 31.46 0.000 .6577494 .7451615
South Sotho | .6680788 .0251233 26.59 0.000 .61883 .7173276
Venda | .7855989 .0423243 18.56 0.000 .7026311 .8685667
Tsonga | .7588997 .0277814 27.32 0.000 .7044401 .8133592
Swazi | .7053638 .0328001 21.50 0.000 .6410662 .7696614
Ndebele | .577324 .0481223 12.00 0.000 .4829906 .6716574
Coloured | .3378845 .0269573 12.53 0.000 .2850404 .3907285
Indian | .2915494 .0345139 8.45 0.000 .2238922 .3592066
Constant | .0531108 .016843 3.15 0.002 .0200936 .0861279

Since we are using a linear probability model, the regression coefficients in the first column represent the probability of being optimistic about a new government in improving living conditions. The constant term represents the probability of Afrikaners in believing things will be better under a new government (5.3%) and the coefficients of the other ethnic groups represents the probability of being satisfied relative to Afrikaners. For example, English speaking whites have a 11.4% (5.3% + 6.1%) probability of being optimisitc. The difference is statistically significant at the 95% level since the t-statistic is 2.14 (in general, a t-statistic around 2 verifies statistical significance at the 95% level). Within African ethnic groups, there is also some variation in the probability of believing things will get better under a new government. The Zulus have the lowest probability among African ethnic groups at 55%, with Ndebele also having a lower probability relative to other groups at 63%. The other African ethnic groups have a probability of at least 70% in believing the government will make their lives better. Coloureds and Indians both have lower probabilities than African ethnic groups, with a 39% and 34% probability in believing their lives will improve.

Multi-variable Regression

Based on the results of the bi-variable regressions, we see some statistically significant differences in the probability of believing that the new government will improve their lives between different ethnic groups. However, there are a number of confounding factors that may affect how optimistic an individual is about the new government. Possible omitted variables that I am going to account for include:

• Income – An individual's economic circumstance can have an impact how optimistic or pessimistic about change from the newly elected government. For wealthy Afrikaners and English whites, they may be more pessimistic about the new African-led government, especially if they are expecting pro-African economic policies to passed. The reverse might be said for richer Africans who would stand to benefit from pro-African economic policies passed by the newly elected government.
• How you feel about your situation compared to your parents – While individual may be poor, if that individual is significantly better off than his/her parents, it may be that they are more optimistic about a new government.
• Educational attainment – Education can have a varying affects. The more education an individual has, the more likely that person is able to recognize the benefits of a new government elected in a fair and open election. Or it could be that a person with more education is more conscious of his/her environment, leading him/her to be less optimistic about the new government improving life substantially after 40 plus years of systematic discrimination.
• Whether you live in an urban or rural area– As mentioned above, Africans in urban areas often ignored government attempts to create ethnic divisions and identified themselves as Africans regardless of ethnicity. Also, South Africa's economic policy focused on industrialization and capital intensive industries during the apartheid area, leading to more wage growth in industrialized areas relative to rural areas. Thus, we need to control for the geographical area in which an individual lived.
• Victim of a crime – If you are a victim of a crime, you may be more or less optimistic about whether a new government could make things better.
• Gender – Female headed households may face greater difficulties than male headed households regardless of race or ethnicities due to limited job opportunities, wage inequalities, and general discrimination against females.
• Exposure to apartheid policies – Africans who were forced to live in the Homelands may have a stronger ethnic identity and a desire for an independent state. On the other hand, due to squalid living conditions and corruption among the leaders, living in the Homelands may have lead to a weaker ethnic identity and a stronger motivation to support a pan-African leadership. Also, urban Indians, Coloureds and Africans who were forced to live in townships may have a stronger desire to see a change in government and be more optimistic about the new government creating positive change in their lives.

I will next run a multi-variable regression controlling for all of these potential confounders. I have selected the following variables to control for the potential confounding factors above.

• "logtotminc" – This is a constructed variable that measures the log of total monthly household income. To do this, use the command "gen logtotminc" = "og(totminc)".
• "parents_" – Question asking how your life is relative to your parents: better, same, or worse
• "educ_new" – A recoded variable of education which is a measure of how many years of schooling one has. To see the explanation of the recoded education variable, visit
• "metro" – Question asking individual if he is living in a metro, urban, or rural setting. This variable also partly captures some of the affects of apartheid policies since those categorized as "metro" are those who live in the townships of Cape Town, Johannesburg, Durban, and Port Elizabeth.
• "gender_n" – Binary variable that is "2" if female, "3" if male
• "crime_q" – Binary variable that is "1" if victim of a crime and "2" if not a victim.
• "homeland" – This is a constructed variable that try and measure the affect of apartheid policies on non-white groups. The "homeland" variable is "0" if your last place of residence was NOT in a Homeland, "1" if it was in a Homeland. I constructed this variable using "last_rel", which asks where your last place of residence was. This may not be a very good way to capture all of the individuals that lived in the Homelands, especially if South African families tend to move often. In fact, when I tab homeland, there are only 206 households whose last place of residence was in a Homeland. I will include this variable later to see if it makes an impact.
• "township" – The township variable is another constructd variable using the "metro" and "race" variables. If the individual's race was African, Coloured, or Indian and the metro variable equaled "3", then it was assumed they lived in a township. If the metro variable equaled "1" or "2" and non-white, then they were coded as not living in a township. Again, this variable may not capture all of the Africans, Indians, and Coloureds who experienced life in the townships.

We are now ready to run the multi-variable regression. Here it makes more sense to use the "if rel_head==1|rel_head==2" adjustment since we will be getting the head of households educational attainment and his/her perspective of how they are doing relative to their parents. Also, it is likely that the head of household was the individual who responded to the survey.

xi: reg lpnew_govt i.ethnicity logtotminc gender_n i.metro i.parents_ i.educ_new i.crime_q if rel_head==1|rel_head==2

Source SS df MS Number of obs = 6939
------ ----- ----- ----- F(32, 6906) = 64.90
Model 390.468438 32 12.2021387 Prob>F = 0.0000
Residual 1298.35848 6906 .188004414 R-squared = 0.2312
------ ----- ----- ----- Adj R-squared = 0.2312
Total 1688.82692 6938 .243416967 Root MSE = .43359
lpnew_govt Coef. Std.Err. t P>t 95% Conf Interval
English | .0733599 .0306858 2.39 0.017 .0132063 .1335134
Other White | .0348965 .1135972 0.31 0.759 -.1877889 .2575819
Xhosa | .6361623 .0265062 24.00 0.000 .584202 .6881226
Zulu | .4615817 .0257443 17.93 0.000 .411115 .5120484
Tswana | .6499641 .0280568 23.17 0.000 .594964 .7049641
North Sotho | .6618194 .0280977 23.55 0.000 .6067392 .7168995
South Sotho | .6313577 .0299973 21.05 0.000 .5725538 .6901616
Venda | .6313577 .0299973 21.05 0.000 .5725538 .6901616
Tsonga | .7178568 .0328749 21.84 0.000 .6534119 .7823016
Swazi | .6572092 .0367854 17.87 0.000 .5850985 .7293199
Ndebele | .5372211 .0511651 10.50 0.000 .4369218 .6375205
Coloured | .3199589 .0302838 10.57 0.000 .2605934 .3793245
Indian | .2651857 .0364477 7.28 0.000 .1937369 .3366344
Log Totmine | -.0015394 .0056359 -0.27 0.785 -.0125876 .0095088
Male | .0101752 .0119532 0.85 0.395 -.0132567 .0336071
Urban | .006677 .0152491 0.44 0.662 -.023216 .0365699
Metro | -.0676998 .0155651 -4.35 0.000 -.0982122 -.0371874
Same as parents | -.0612649 .0155267 -3.95 0.000 -.091702 -.0308278
Worse than par | -.0297039 .0134964 -2.20 0.028 -.056161 -.0032469
_Ieduc_new_1 | -.0319237 .0232249 -1.37 0.169 -.0774515 .0136042
_Ieduc_new_2 | .0164635 .0242632 0.68 0.497 -.0310999 .0640269
_Ieduc_new_3 | .010858 .0257153 0.42 0.673 -.0395519 .061268
_Ieduc_new_4 | -.015072 .0230822 -0.65 0.514 -.0603203 .0301763
_Ieduc_new_5 | .0598858 .0208805 2.87 0.004 .0189536 .100818
_Ieduc_new_6 | .0154384 .0196254 0.79 0.432 -.0230334 .0539102
_Ieduc_new_7 | .0396093 .0286658 1.38 0.167 -.0165846 .0958031
_Ieduc_new_8 | -.0021483 .0224648 -0.10 0.924 -.0461863 .0418897
_Ieduc_new_9 | -.0337736 .0291925 -1.16 0.247 -.0909999 .0234527
_Ieduc_new_10 | .0073615 .0238021 0.31 0.757 -.0392979 .0540209
_Ieduc_new_12 | -.0137709 .0290655 -0.47 0.636 -.0707483 .0432065
_Ieduc_new_16 | -.0105202 .039328 -0.27 0.789 -.0876152 .0665748
Crime | .0038832 .0178855 0.22 0.828 -.0311778 .0389443
Constant | .100286 .0595181 1.68 0.092 -.0163879 .2169599

After controlling for income, how participants view their life relative to their parents, educational attainment, victimization of a crime, gender, and geographical area, we still see variation in the probability in believing the new government will make life better. The constant term, representing the probability for Afrikaners, is 10.0%. English speaking whites have a statistically significant higher probability of 17.3%. Among African ethnic groups, the Zulu and Ndebele groups still have a lower probability relative to the other African ethnic groups with probabilities of 56.1% and 63.7% respectively. The Venda and Tsonga ethnic groups have the highest probabilities, 84.5% and 81.8% respectively. We see again that Coloureds and Indians both have lower probabilities than African ethnic groups and higher probabilities than white groups, with a 42.0% and 36.5% probability in believing their lives will improve under a new government.

The results for the potential confounders provide some interesting results:

• Income – The results show that income does not affect the probability of an individual believing that the new government will make his/her life better.
• Gender – TGender had little or no affect on the probability of believing that life will improve under the new government.
• Geographical Area – The probability of those who live in a metro area is 6.7 percentage points less than those who live in rural areas in believing that their lives will improve under a new government.
• Parents – The probability of individuals who feel they are just as well off as their parents is 6.1 percentage points less than those who feel they are better off than their parents. For those who feel worse off than their parents, their probability is 3.0 percentage points less. Both results are statistically significant.
• Crime – Being a victim of a crime had little or no affect on the probability of believing the new government will make their life better.
• Education – We see statistically significant differences after five years of education. For individuals who received five years of education, their probability is 6.0 percentage points higher more than those with no education.

The last regression that I am going to run is to see whether the difference in the probabilities is statistically significant between African ethnic groups. I also want to add two additional variables, "homeland" and "township" to account for exposure to certain apartheid policies. I have included interaction terms to see how these variables interact with the different ethnicities. Because Indians and Coloureds also experienced life in the townships, I am including them in this last regression.

xi: reg lpnew_govt i.nonwhite*homeland i.nonwhite*township crime_q i.parents_ i.educ_new logtotminc i.gender_n if rel_head==1|rel_head==2

Source SS df MS Number of obs = 6003
------ ----- ----- ----- F(48, 5954) = 12.41
Model 122.432759 48 2.55068247 Prob>F = 0.0000
Residual 1223.67252 5954 .205521082 R-squared = 0.0910
------ ----- ----- ----- Adj R-squared = 0.0836
Total 1246.10528 6002 .224276121 Root MSE = .45334
lpnew_govt Coef. Std.Err. t P>t 95% Conf Interval
Zulu | -.19337 .0207134 -9.34 0.000 -.2339758 -.1527642
Tswana | .0181003 .0247896 0.73 0.465 -.0304964 .0666969
North Sotho | .0173419 .0230422 0.75 0.452 -.0278291 .062513
South Sotho | -.0232909 .0274581 -0.85 0.396 -.0771187 .0305369
Venda | .1121131 .0471147 2.38 0.017 .0197511 .204475
Tsonga | .0816381 .029738 2.75 0.006 .0233408 .1399354
Swazi | .0301989 .0345799 0.87 0.383 -.0375903 .0979881
Ndebele | -.0944496 .0556219 -1.70 0.090 -.2034886 .0145895
Coloured | -.2495089 .0376592 -6.63 0.000 -.3233346 -.1756831
Indian | -.320996 .0467032 -6.87 0.000 -.4125511 -.2294409
Homeland | -.0909094 .0534841 -1.70 0.089 -.1957576 .0139387
ZuluXhome | -.0082687 .1474952 -0.06 0.955 -.2974127 .2808753
TswanaXhome | .12284 .0883333 1.39 0.164 -.0503253 .2960053
N.SothoXhome | -.1966445 .1812243 -1.09 0.278 -.5519097 .1586208
S.SothoXhome | .0863995 .1812074 0.48 0.634 -.2688328 .4416317
VendaXhome | -.0585149 .2755082 -0.21 0.832 -.5986109 .481581
TsongaXhome | .1327509 .0959197 1.38 0.166 -.0552866 .3207883
SwaziXhome | -.0289319 .269172 -0.11 0.914 -.5566066 .4987428
NdebeleXhome | .408002 .2726892 1.50 0.135 -.1265676 .9425716
ColouredXhome | .542258 .4581683 1.18 0.237 -.355918 1.440434
IndianXhome | (dropped)
Township | -.0655951 .0322928 -2.03 0.042 -.1289006 -.0022895
ZuluXtown | .0656342 .0448734 1.46 0.144 -.022334 .1536024
TswanaXtown | -.1307734 .0634387 -2.06 0.039 -.2551362 -.0064106
N.SothoXtown | .0252682 .0596068 0.42 0.672 -.0915827 .142119
S.SothoXtown | .066179 .0594215 1.11 0.265 -.0503086 .1826666
VendaXtown | -.100366 .1628517 -0.62 0.538 -.4196143 .2188824
TsongaXtown | -.2291031 .111625 -2.05 0.040 -.4479287 -.0102776
SwaziXtown | -.3395913 .1439647 -2.36 0.018 -.6218142 -.0573684
NdebeleXtown | -.1483389 .1364007 -1.09 0.277 -.4157337 .1190559
ColouredXtown | -.145537 .0563478 -2.58 0.010 -.2559992 -.0350749
IndianXtown | -.158121 .0737353 -2.14 0.032 -.3026689 -.0135731
Crime Victim | .0019929 .0214398 0.09 0.926 -.0400369 .0440227
Same as parents | -.0693683 .018024 -3.85 0.000 -.1047019 -.0340348
Worse than par | -.0368817 .0155833 -2.37 0.018 -.0674306 -.0063328
_Ieduc_new_1 | -.0282584 .0244014 -1.16 0.247 -.0760939 .0195772
_Ieduc_new_2 | .0176488 .0255037 0.69 0.489 -.0323476 .0676452
_Ieduc_new_3 | .0063192 .0270038 0.23 0.815 -.046618 .0592565
_Ieduc_new_4 | -.0185514 .0243533 -0.76 0.446 -.0662927 .02919
_Ieduc_new_5 | .0597 .0219629 2.72 0.007 .0166448 .1027551
_Ieduc_new_6 | .0157836 .0210253 0.75 0.453 -.0254337 .0570009
_Ieduc_new_7 | .0334987 .0308546 1.09 0.278 -.0269875 .0939848
_Ieduc_new_8 | .0080793 .0255126 0.32 0.751 -.0419346 .0580932
_Ieduc_new_9 | -.0254834 .0331541 -0.77 0.442 -.0904775 .0395106
_Ieduc_new_10 | .0228707 .0295168 0.77 0.438 -.034993 .0807344
_Ieduc_new_12 | -.0257304 .042949 -0.60 0.549 -.109926 .0584653
_Ieduc_new_16 | .0107943 .0842845 0.13 0.898 -.1544338 .1760224
Log_totmine | -.0000515 .0062227 -0.01 0.993 -.0122503 .0121473
Male | .0156531 .0130992 1.19 0.232 -.0100261 .0413323
Constant | .7560405 .0617567 12.24 0.000 .6349749 .877106

The results show that some of the differences in probabilities between African ethnic groups are statistically significant after controlling for the possible confounding factors described above. In this regression, the constant represents the probability of individuals in the Xhosa ethnic group (75.6%). Zulu and Ndebele ethnic groups have the lowest probability (56.3% and 66.2% respectively) and Venda and Tsonga have the highest (86.8% and 83.8% respectively). The rest of the African ethnic groups have probabilities within 1 to 3 percentage points of the Xhosa ethnic group and the differences are not statistically significant. Indians and Coloured have significantly lower probabilities than all the African ethnic groups (35.6% and 50.6% respectively).

Both the "Homeland" and "Township" variables show statistically significant differences in the probability of believing that the new government will make life better. For instance, the probability that an individual who's last place of residence was in a Homeland believes that his/her life will get better under the new government is 9.1 percentage points less than someone who¡¯s last place of residence was not in a Homeland. For an individual that resides in a township, his/her probability is 6.1 percentage points less than an individual who does not live in township. The interaction term "ethnicityXhome" show us the affect of the "homeland" variable by ethnicity and the interaction term "ethnicityXtown" shows us the affect of the "township" variable by ethnicity.



Due to the limitations in the data, it is impossible to say whether apartheid policies made ethnic identities any stronger or weaker than they already were since we are not able to observe the counterfactual (i.e., life in 1993 without the prior 40 plus years of apartheid). Even though the "Homeland" and "Township" variables were significant, we should be skeptical of the results since these variables are not able to account for many of survey participants that experienced life in the Homelands and Townships. Overall, the only conclusion that we can make is that there are some statistically significant differences in the probability of believing that a new government will improve their lives between different ethnic groups in 1993. English speaking whites are 7 percentage points more likely than Afrikaner whites to believe that life will be better under a new government. Among African ethnic groups, the Zulu and Ndebele people have lower probabilities (19.3 and 9.4 percentage points less respectively) of believing that life will improve under a new government than the Xhosa, whereas Venda and Tsonga have higher probabilities (11.2 and 8.1 percentage points higher respectively) than the Xhosa. Indians and Coloureds have lower probabilities than the African ethnic groups, but higher probabilities than the white ethnic groups.

AUTHOR'S NOTE -This module was written with the sole purpose of sparking interest in social science research using the statistical software package STATA. If you have any questions about the STATA commands, the creation of certain variables, etc., please do not hesitate to contact me at [email protected].




Henrard, Kristin, "Post-Apartheid South Africa: Transformation and Reconciliation," World Affairs. Washington: Summer 2003. Vol.166, Iss. 1; pg. 37.

Piombo, Jessica, "Political Parties, Social Demographics, and the Decline of Ethnic Mobilization in South Africa, 1994-1999," Party Politics. Thousand Oaks: Vol. 11, No. 4; pp. 447-470.

Thompson, Leonard. A History of South Africa, Third Edition. New Haven and London: Yale University Press, 2001.

Vora, Jay A. and Erika Vora, "The Effectiveness of South Africa's Truth and Reconciliation Commission: Perceptions of Xhosa, Afrikaner, and English South Africans," Journal of Black Studies. Thousand Oaks: Jan 2004. Vol.34, Iss. 3; pg. 301.

Wooldridge, Jeffrey M. Introductory Econometrics: A Modern Approach. Mason, Ohio: Thomson South-Western, 2003.

Hosted by

1 1