Monitoring Perceptions of the Causes of Poverty in South Africa

Note that Julius Malema was suspended at the time this publication was finalized for publication.

In South Africa there are four broad settlement types: (1) formal urban areas, (2) informal urban areas, (3) commercial farms, and (4) tribal areas and rural informal settlements. According to Statistics South Africa a ‘rural area’ is ‘any area that is not classified urban. Rural areas are subdivided into tribal areas and commercial farms’ (i.e. relying on geographic type classification), while urban areas were classified ‘based on dominant settlement type and land use.’ Cities, towns, townships, suburbs, etc., are typical urban settlements. Within the urban areas Statistics South Africa distinguish between formal and informal settlement areas. The informal settlement areas are those areas defined as unplanned settlement on land which has not been surveyed or proclaimed as residential, consisting mainly of informal dwellings (shacks) (Statistics South Africa 2003).

The living standard measure is a unique means of segmenting the South African market. It cuts across race and other outmoded techniques of categorising people, and instead groups people according to their living standards using criteria such as degree of urbanisation and ownership of cars and major appliances. A total of 29 variables are used. Each variable carries a different weight, some positive, others negative, and the respondent’s position on the SAARF LSM scale is arrived at by adding together the weights of the variables that she/he possesses. A constant is also added to the total score to remove negative total scores (http://​www.​saarf.​co.​za).

Since the development of the perception of the causes of poverty scale (PCPS) by Feagin it has been applied in various countries under different circumstances. Sun (2001), for example, administered a revised version of the PCPS to determine how American social work students (SWS) and non-social work students (NSWS) perceived the causes of poverty. The study concluded that overall American SWS believe that poverty is more a cause of structural factors within the environment than individualistic factors. This result is contradictory to the general opinion of ordinary Americans. Another study described the Chinese perception of the causes of poverty scale (CPCPS). The CPCPS was developed to measure perceived causes of poverty in the Chinese culture (Shek 2002). The primary focus of the study was on the psychometric properties of the CPCPS. It was found that the scale was reliable and valid and measured the underlying poverty dimensions. CPCPS covered four categories of explanations: personal problems of the poor, lack of opportunities to escape from the poverty cycle, exploitation of poor people, and bad fate. A follow-up study by Shek (2004) investigated the beliefs about the causes of poverty in Chinese parents and adolescents experiencing economic disadvantages showed that the same four factors were stable across time and across different samples. A study by Weiss and Gal (2007) used measures similar to those found on the Feagin Scale and argued that the internal consistency of the revised questionnaire increased since some items were removed, some rephrased and some added. In addition, a panel of three social work researchers reviewed the questionnaire items for face validity and found that the items adequately measured the attitudes to the causes of poverty. In a much earlier study Marshall et al. (1999) investigated beliefs about inequality in thirteen established Western-democratic and newly post-communist industrial nations. The first 30 attitudinal question items used on their questionnaire, seek evidence about people’s perceptions of inequality. These items are also categorized into individualistic, structural and fatalistic dimensions. For example, the following reasons are given as to why people are poor: ‘lack of equal opportunity’ (structural), ‘lack of effort by poor themselves’ (individualistic), and ‘bad luck’ (fatalistic).

Please refer to the methodology for the actual question wording of the various question items measuring perceptions of the causes of poverty.

It should be note that this item measured structural perceptions of the causes of poverty.

The results of the possible statistical assumption violations, for missing values and outliers revealed that there were very few missing values reported for all the variables. The data was also normally distributed with all the P–P plots looking reasonably normal since the data points are all close to or on the diagonal lines. In addition, the skewness and kurtosis values were almost all within the acceptable range of −1 to +1 thus indicating a normal distribution. Next the collinearity among the independent variables is assessed. A review of the tolerance statistics of the three multiple regressions reveal that not a single tolerance value for any of the variables in all the multiple regressions were found to be less than or equal to 0.01. The results of the VIF also revealed that all the values for all three multiple regressions were less than 10 which imply no multicollinearity among the independent variables. In addition, the conditional index for each of the three multiple regressions showed that none of the independent variables is equal or greater than 30.

Dummy variables were entered for gender, race, geographical location, and employment status. More specifically, for race the dummy variables white, coloured and Indian were entered (with black being the implicit reference group). Dummy variables employed and not working were entered for employment status with unemployed being the implicit reference group. Geographic location had dummy variables for urban informal, traditional/tribal areas and rural formal except for urban formal (which acted as the implicit reference group), while gender had a dummy variable for male (with female being the implicit reference group).

It should be noted that the explanatory power of this linear regression model is quite weak (as indicated by Adjusted R²) and should therefore be interpreted with caution.

Dummy variables were entered for gender, race, geographical location, and employment status. These dummy variables are the same variables that were entered for the first regression with the structural index as the dependent variable.

Again it is important to note that the explanatory power of this linear regression model is quite weak (as indicated by Adjusted R²) and should therefore be interpreted with caution.

Dummy variables were entered for gender, race, geographical location and employment status. These dummy variables are the same variables that were entered for the first two regressions in the previous sections.

Also note that the explanatory power of this linear regression model is quite weak (as indicated by Adjusted R²) and should therefore be interpreted with caution.