Development and gender: Longitudinal entrepreneurial gender effects of the inner-city Johannesburg street-trading context

Abstract

An empirical investigation was undertaken into entrepreneurial gender effects within the inner-city street-trading context of Johannesburg, a large South African city. A cross-sectional non-parametric quantitative research design was applied in each of three consecutive years, 2008, 2009 and 2010, and a longitudinal investigation was enabled. Differences in earnings, rental stand operation, and the effects of specific and general human capital by gender were tested using non-parametric Kruskal–Wallis methods. Further testing of the non-parametric associations of each factor by gender was undertaken using non-parametric Spearman rho and Kendall tau measures. Male traders are found to earn more. However, a return on specific and general human capital is found for female traders. Security threats in this context might have a disproportionate effect on female street traders, and most specifically on female street traders of foreign origin.

Keywords:

• gender
• street traders
• entrepreneurial development

1. Introduction

According to the tenets of human capital theory, human capital endowments – or the investments in learning of individuals – are typically associated with increased productivity (Becker, 1975). Such increases in productivity are expected to be associated with higher earnings, irrespective of arbitrary differences such as gender (Becker, 1975). According to ascription theory: ‘[i]n every culture women and men are assigned disparate functions and patterns of behaviour well beyond those dictated by clear psychological differences’ (Marwell, 1975:445). However, according to Marwell (1975), gender differences that are not functional to the societal group will effectively reduce in a context where skills are scarce. According to both of these theorists, work-related and earnings-related gender differences in working environments characterised by the scarcity of skills and human capital will diminish over time.

However, a tension exists between the predictions of this body of theory and the theory and empirical findings which support the notion that discrimination and gender disadvantages continue to plague female entrepreneurs and employees in enterprise contexts (Morrison et al., 1987; Powell & Butterfield, 1994). Nevertheless, across time, and in different contexts, changes toward gender parity are evident in the literature (Igbaria & Baroudi, 1995; Powell & Butterfield, 2002). Moreover, a further change is also evident from a review of the education literature: the increasing dominance of female students in higher education uptake (US Department of Education, 2007; Nhantsi, 2010) and educational attainment in certain international (Demie, 2001) and local contexts (Nhantsi, 2010). The extent to which gender-related change in the structure of earnings and human capital in the informal sector context reflects changes in the broader societal context is therefore the focus of this research.

2. The problem addressed by this research

The objective of this research is to test theory that relates differences in the associations of human capital and earnings to gender. By doing so, this research aims to extend gender and human capital theory (Becker, 1975) into the South African informal street-trading context for testing. It is argued that knowledge of gender differences in this context can offer insights to policy-makers and stakeholders that may contribute to upliftment in the sector.

Despite a growing body of literature that relates the gender dimension of human capital-related change to certain international and South African contexts, absent from the literature is research that specifically investigates this change at the nexus of entrepreneurial development and the fringes of informal enterprise: the informal street-trading sector. To the extent that individuals with almost no resources can access the informal sector, street trading can represent an entrepreneurial training ground, or a development platform that can enable enterprise development; a structural path to potential formal ventures (De Soto, 1989). This sector might therefore have a societal developmental role to play. However, if this sector is nested within a broader societal, and international, context, then gender changes in the overarching context are expected to transmit, to some extent, into this context. In the absence of literature that relates the human capital-related interrelationships of this sector to differences by gender, knowledge of how these effects might contribute to development is scant. If ascription is declining on a global basis (Marwell, 1975), what is not clear is the position of female traders within the South African informal street-trading context. What are also not clear are the implications of such gender changes, or the implications of the absence of such changes, in this context. This research therefore aims to address the following question: to what extent are changes in the relationships around gender manifested in the current informal entrepreneurial context? And, more specifically; what are the changes associated with endowments of human capital by gender in this context? Hence, an investigation is undertaken into: gender-related changes in earnings; access to resources such as rental stands; the effects of specific human capital and general human capital; and changes in the underlying human capital factor structure of the sector over time. The study applies an exploratory quantitative cross-sectional correlation design in three consecutive years. Longitudinal insight is drawn from the process.

3. Theory and derivation of hypotheses

If entrepreneurship is fundamentally the pursuit of opportunity, then such opportunity is a necessary condition for entrepreneurship; the individual is key to the operationalisation of opportunity (Shane & Venkatataman, 2000). However, the environment or context is also critically important as an individual engages with opportunity (Zahra & Dess, 2001). Stevenson & Jarillo (1990) argue that entrepreneurship is fundamentally associated with learning; learning is at the heart of the pursuit of opportunity. General human capital, for Becker (1975), represents the cumulative investments of learning in an individual that can lead to productivity enhancements across different work contexts. In contrast, specific human capital represents cumulative investments of learning in an individual that are context specific; that cannot contribute to increased productivity in contexts other than a specific context (Becker, 1975). General human capital, in the form of formal education, is also expected to facilitate further specific human capital learning (Becker, 1975) that contributes to an entrepreneur's learned adaptation to a specific context. Entrepreneurial behaviours are typically associated with increased earnings or other entrepreneurial outcomes to the extent that they match the specific context in which an entrepreneur finds himself or herself (Lumpkin & Dess, 1996). This research therefore focuses on potential constraints to earnings and differences in the associations around human capital in this context according to differences in gender. Further theory and research literature is now considered and hypotheses are derived.

Situated at the nexus of transport nodes and high-density residential areas, the informal street-trading sector is a space also characterised by a high density of traders. Traders are vulnerable to theft and other security-related constraints to their development (ESSET, 2007).

The Johannesburg inner-city street-trading population is expected to exist not in isolation, but within a larger societal and global context, in which gender roles might possibly be changing (Hausmann et al., 2009). Entrepreneurship behaviour is a function of gender roles, which in turn may be a function of culture (Mueller, 2008). Gender differences in the effects of culture have been found at both the societal and the organisational levels (House et al., 2004). According to Emrich et al. (2004), the economic competition perspective of the relationship between gender and economic growth stresses that, despite gains made in female economic participation, gender-based clustering still occurs in certain areas of the economy such as service and communication. Therefore economic growth might reflect some dimension of stereotypical gender roles if ‘pink collar’ ghettos persist (Emrich et al., 2004). It may therefore be relevant to explore the extent to which the informal sector might be a space in which such clustering occurs, and the economic relationships potentially associated with gender inequality in this space.

Gender differences have been found to be marginally associated with differences in entrepreneurial orientation (Chow, 2006). Gender differences have also, in certain contexts, been found to be associated with differences in job commitment (De Clerq & Ruis, 2007) and with differences in the dimensions along which entrepreneurial persistence is associated with entrepreneurial choice (Gatewood et al., 1995). Gender differences have also been found to be reflected in differences in enterprise gender-role orientation, but not entrepreneurial self-efficacy (Mueller, 2008). Other research indicates gender differences in non-pecuniary motivation in enterprise contexts (Burke et al., 2002). However, certain differences between the genders in terms of enterprise effects may be due to the effects of culture (Mueller, 2008) and these effects may be context and culture specific. The implication of this is that, if certain aspects of female enterprise behaviour are constrained by culture, and if the influence of such a culture decreases over time or contexts, then these constraints will be expected to reduce.

For whatever reason, women face higher business failure rates, according to Robb (2002). On the basis of the potential relative vulnerability of female traders to crime or security issues in the informal context, and, further, on the basis of other challenges faced by female entrepreneurs that have across other contexts been associated with higher business failure rates or lower earnings, it is hypothesised that levels of earnings differ significantly according to gender (Hypothesis 1). Female traders are expected to earn less, and to be more relatively dissatisfied with continuing in street trading if this was to be found to be a discriminatory context. If, at lower levels of access to resources, female traders are exposed to a discriminatory context, reflected in lower levels of access to financial resources, female traders would be expected to have less access to financial capital, and might subsequently be less likely to be able to afford rental stands. It is therefore hypothesised that differences in rental stand operation rates will differ according to gender (Hypothesis 2). However, if enterprise behaviour is considered to be ‘nested’ within the broader societal milieu, then it might be expected that the broader society might have an influence upon the enterprise behaviour within it. Changes in the broader society might eventually be expected to be transmitted to contexts within it; including the informal street-trading context. Evidence has been found of the transmission of societal gender-related cultural practices and values to the organisational level (House et al., 2004). Therefore potential changes in gender effects at the level of society are deemed to potentially be relevant to the gender effects of the informal street-trading context. Two areas in which an overarching trend in gender change seems to be increasingly evident is in societal gender gaps, and in learning, or educational, effects.

A measure of changes in society along the dimension of societal gender gaps has been developed; the Global Gender Gap Index (Hausmann et al., 2009). This measure indicates the size and extent of gender-based disparities or gender gaps in economic, political, or educational dimensions, which are published as statistics (Hausmann et al., 2009). According to the World Economic Forum Global Gender Gap Index, South Africa had improved to sixth spot in the world rankings by 2009 (Hausman et al., 2009). In 2010 South African ranked 12th on this index (World Economic Forum, 2011). The index measures the extent to which economic participation and opportunity, educational attainment, health, survival and political empowerment is associated with gender equality (World Economic Forum, 2011).

According to Becker's (1975) prediction, over time it is expected that exit rates of the informal sector toward the formal sector would be related to general human capital endowments, and not to arbitrary characteristics such as gender. However, women are still under-represented in formal work in the South African context (Nhantsi, 2010); this might reflect unequal opportunities in entering the formal sector. It might be possible that gender inequality is reflected in the opportunities that exist for exit from the sector. Measures of changes over time in the sector may indicate overall trends in this regard. Therefore it is hypothesised that time spent in the sector will differ according to gender (Hypothesis 3). Male traders may be more able to exit the sector if a discriminatory context does indeed exist. However, the broader societal context and the informal context within it are expected to become less discriminatory over time, according to Becker's (1975) argument. It is important to understand the extent to which gender-related changes on the societal level are potentially transmitted, or correspond, to gender inequality in the street-trading context.

On a global basis, in certain countries, changes in education-related effects seem to indicate that dominance by males has given way to dominance by females. In the context of the United Kingdom, Demie (2001) found that females outperform males in school-based education, at all key stages in some, yet not all, school subjects. Demie (2001) found that these effects manifested irrespective of ethnicity. Before 1989, males dominated bachelor degree and master degree qualifications graduation cohorts; yet after this period, female attainment has dominated (US Department of Education, 2007). This might be taken to represent a ‘tipping point’ of sorts, whereby societal change may have been reflected in this measure of human capital attainment.

Since this ‘tipping point’, the female share of bachelor degree graduates has risen from about 53% in 1989/90 to about 57% for 1999/2000, and to approximately 57.5% for the period 2005/6 (US Department of Education, 2007). The same trend is evident in the attainment of master degrees, with a rise in female attainment from 52.6% for 1989/90 to 55% for 1999/2000, and to 59.95% for 2005/6 (US Department of Education, 2007). Changes have also been evident in the South African context. However, in the South African labour market, women are still not found to be equally represented in higher skills jobs or in jobs that are relatively more highly paid (Nhantsi, 2010). Yet in terms of educational attainment, the general trend seems to be towards female dominance. South African higher education institutions report increases in female enrolment from 44% in 1993 to 51% in 1999, and to 54% in 2001 (Nhantsi, 2010). On the basis of these changes evident in the literature, it is hypothesised that endowments of higher education will differ by gender (Hypothesis 4). Over time, female traders are expected to be becoming increasingly more educated, if the structure of relationships within the informal sector does reflect the trends in the broader society. The changes in gender effects over time predicted in the literature are therefore expected to be reflected in demographic changes in the sector. Deriving from the above discussion, an implication arises: that the structure of interrelationships around gender that exist in the inner-city informal street-trading context might be changing to reflect the broader societal context of gender-related interrelationships. On the basis of this conception, it is hypothesised that the underlying associative factor structure of the interrelationships has changed over time in the sector (Hypothesis 5). If this was occurring in reality, factor analysis processes would be expected to provide tentative evidence of the changes in such gender associations in the sector over time. The methodology applied in this study is considered as follows.

4. Research methodology

This study utilises an exploratory quantitative cross-sectional design. However, three studies provide three sets of data. This allows for a longitudinal perspective of the inter-relationships around gender in this context. Three samples were drawn from the population of inner-city Johannesburg street traders, one each for the years 2008, 2009 and 2010. An area of approximately 224 city blocks was identified to capture the inner-city context, separate from the adjacent areas that were comprised of residential, industrial and transport-oriented blocks. Random sampling was used in order to select about 10% of these blocks: 23 blocks. The street traders operating on the sides of these blocks were counted. On the basis of this number, a broad estimate of the inner-city street-trading population was obtained: of about 5200 traders. A sample size calculation was used to ensure that the sample sizes were sufficient in terms of the principles of inferential statistics.

The samples comprised 339, 309, and 303 traders for the years 2008, 2009 and 2010 respectively. Piloting was undertaken in order to ensure validity and reliability in the design of the items of the instrument. Attempts at further random sampling of individual traders were confounded by refusal rates, which due to ethical principles were unconditionally accepted. Therefore, claims are made on the basis of convenience sampling. However, it is argued that with almost 6% of the population sampled each time, and with a longitudinal anchor, the results are generalisable to similar segments of street-trading populations to the extent that these other populations are similar to this context. Scales were developed. The research design was built around tests considered appropriate for the testing of the hypotheses. The first choice of methods was parametric testing. However, non-parametric alternatives were planned in case the univariate analysis failed to demonstrate that the assumptions of the parametric tests could be met (Tabachnik & Fidel, 2007).

On the basis of the specific types of data, and the results of the univariate tests, non-parametric tests of difference were applied. One-way non-parametric analysis of variance was utilised for hypotheses testing. The Kruskal–Wallis results from this process of testing are reported. Hypotheses 1 to 4 were tested though Kruskal–Wallis tests. Hypothesis 5 was tested though an analysis of factor analysis outputs.

For the 2008 and the 2009 data, the responses for earnings are capped at 690 rand per day. In the 2010 data the responses are capped at 890 rand per day. Experience is measured as years spent in street trading (Tables 1 and 2). Total education is measured as years of completed education. Initial investment is measured as the amount of cash invested in the enterprise at start-up. Hours worked per day and days worked per week are measured as such. Rental stand is a measure of whether a trader operates a City Council-provided rental stand. Age is measured in years. The results of the process of statistical testing are reported, and also discussed, as follows.

Table 1: Univariate analysis measures of location/central tendency and dispersion

Variable	Mean	Median	Mode	Variance	Standard deviation	Range	Interquartile range
Age	34.14	32	27	105.39	10.27	56	13
Years in Johannesburg	10.79	8	3	94.62	9.73	52.8	13
Hours worked per day	10.70	11	11	3.82	1.95	14	2.5
Days worked per week	6.36	6	6	0.566	0.752	7	1
Initial investment	1 179.41	650	350	1 651 019	1 285	6 999	1 200
Total education	9.94	11	12	12.16	3.49	17	4
Experience	6.44	5	1	21.71	4.66	14	8
Continuance satisfaction	6.77	8	9	6.58	2.56	12	5
Earnings	351.6	230	900	76 090	275.84	870	370

Note: 2010 data, n = 303. Mean median and mode tested using SAS. Skewness, kurtosis and Shapiro–Wilk measures are SPSS output.

Table 2: Univariate analysis measures of normality

Variable	Skewness	Kurtosis	Shapiro–Wilk W
Age	0.65	–0.01	0.962***
Years in Johannesburg	1.41	1.85	0.855***
Hours worked per day	0.14	0.63	0.987**
Days worked per week	–1.75	7.93	0.724***
Initial investment	1.93	3.32	0.742***
Total education	–1.01	0.89	0.899***
Experience	0.60	–0.90	0.888***
Continuance satisfaction	–0.39	–1.18	0.770***
Earnings	0.95	–0.45	0.0001***

Note: 2010 data. Measures tested using SAS.

5. Results and discussion

5.1 Null Hypothesis 1: There is no significant difference in earnings by gender

According to the Kruskal–Wallis tests, at the 5% level of significance there is a significant difference in earnings between the sexes for the years 2008 and 2010 (Table 3). The 2009 data also reveal a difference, yet just outside the 5% level of significance. From these findings the null hypothesis is rejected, and the alternative hypothesis is accepted. In 2008 male traders are found to earn more than female traders; almost 16% more. In 2009 the difference is barely significant. However, in 2010 male traders are found to earn about 38% more than female traders. This might possibly indicate that gender inequality is a long-term feature of the inner-city street-trading context.

Table 3: Gender Kruskal–Wallis test results

Variable	2008	2009	2010
Earnings	4.93/1/p < 0.0264	0.057/1/p < 0.0811	12.45/1/p < 0.0004
Experience	0.979/1/p < 0.3224	23.62/1/p < 0.0001	19.08/1/p < 0.0001
Total education	5.422/1/p < 0.0199	13.23/1/p < 0.0003	10.50/1/p < 0.0012
Initial investment	10.830/1/p < 0.001	25.90/1/p < 0.0001	10.51/1/p < 0.0012
Hours worked per day	1.0392/1/p < 0.308	4.064/1/p < 0.0438	4.150/1/p < 0.0416
Days worked per week	3.7575/1/p < 0.0526	0.1364/1/0.7118	3.05/1/p < 0.0808
Rental stand	3.7474/1/p < 0.0529	0.079/1/p < 0.7786	1.22/1/p < 0.2688
Age	3.139/1/p < 0.0764	28.9/1/p < 0.0001	39.29/1/p < 0.0001

Notes: Data presented as chi-square statistic/degrees of freedom/significance: Significant statistics are in bold.

Another candidate for an explanation of this phenomenon might be the influx of males of foreign origin into the sector, who according to the findings of this study typically invest more into their enterprises at start-up and might therefore earn more because of their higher capital investments (Table 4). Besides this cluster, the sector, however, might be considered to be segmented into a category of more survivalist traders, and a category of more entrepreneurial traders (Callaghan & Venter, 2011). To the extent that female traders are found to earn less in this sector, female traders might be more likely to be exposed to a context that has a different structure of relationships associated with lower levels of earnings.

Table 4: Means by year: female/male

Variable	2008	2009	2010
Earnings	148.22/171.26*	264.95/260.68	285.26/392.73***^a
Experience	4.54/4.03	7.92/4.88***	8.01/5.48***
Total education	3.05/3.39*	9.48/10.52***	9.08/10.47**
Initial investment^b	355/480**	296/450***	926/1336**^a
Hours worked per day	10/10.26	10.17/10.62*	10.41/10.87*
Days worked per week	6.13/6.24	6.38/6.35	6.28/6.40
Age	33.9/31.8	35.2/29.54***	38.7/31.32***

Notes: ^aDifferent upper limit to measure therefore not comparable with other datasets. ^bUpper limit to measure. Significance, reflecting Kruskal-Wallis tests: *p < 0.05; **p < 0.01; ***p < 0.001.

The finding that females earn less than males may also support previous research that indicates constraints to female enterprise performance (Robb, 2002). Placing the gender changes in the sector within the broader societal trajectories of gender change, earnings-related gender differences seem to be widening in this specific sector. The predictions of Becker (1975) and Marwell (1975) are not found to be supported in terms of the gender differences associated with gross earnings in this sector. A gender-based ‘glass ceiling’ (Morrison et al., 1987), in the form of earnings-related gender inequality (Powell & Butterfield, 1994), is possibly also present in this context. The entrepreneurial development of female entrepreneurs requires the active participation of the individual (Shane & Venkataraman, 2000) that is not independent of some kind of interaction with the specific entrepreneurial context (Lumpkin & Dess, 1996; Zahra & Dess, 2001). If the context of street trading is associated with factors that constrain entrepreneurial performance that are not gender neutral, this context might be discriminatory. However, if females earn less due to a relative effect, such as an influx of male traders of foreign origin with greater resources, then further research might offer further insight into the relative strengths of these effects. For the purposes of gaining a further understanding of these differences, the significant associations of earnings that differ according to gender are reported and discussed as follows.

Male traders of foreign origin are found to earn significantly more than male traders of local origin in 2008 and 2010. A notable exception to this trend is the case of 2009, where male foreign traders are found to earn significantly less than their local counterparts. On the basis of these results, it is argued that the xenophobic events which occurred in the city centre in 2008 might have influenced the relative earnings of traders of foreign origin versus traders of local origin. Such an effect might also explain the case of 2009 where female traders of local origin are found to earn significantly more than females of foreign origin. Notwithstanding the fact that tests of statistical associations cannot ascribe causality, it is argued that these findings suggest that the xenophobic violence of 2008 might have had an effect lasting until about 2009, or about a year after these events. Alternatively, this evidence suggests that some other structural change might have occurred between traders of local and foreign origin over this period.

In 2008 female traders are found to comprise about 30% of traders of foreign origin yet about 60% of traders of local origin. However, after the xenophobic events of 2008, in 2009 females are found to make up only 25% of traders of foreign origin and about 45% of traders of local origin. In 2010 female traders again make up about 30% of traders of foreign origin, yet make up about 51% of local traders. If the xenophobic violence was to have imposed a cost upon traders in this sector, the incidence of such a cost might have been borne disproportionately by female traders.

Female traders that work more hours a day earn more in all of the sampled years. However, male traders that work more hours a day only earn more in the case of 2009. Female traders typically work fewer hours and days a week. This is one specific dimension of a constraint that might be seen to represent a specific form of ‘glass ceiling’ to the earnings of female street traders (Igbaria & Baroudi, 1995). The inherent vulnerability of women in this sector to physical threat, stock theft, or intimidation by other traders on the basis of physical threats (ESSET, 2007) might expose women disproportionately to certain harmful effects in this context.

Improved security on the part of city authorities might shield women traders from certain of these effects, which might be construed as effective discrimination on the part of the informal context itself. The larger coefficients found for the Spearman rho associations between initial investment and earnings for male traders might suggest that males obtain a higher return on their capital investments on entering the sector. If these higher returns accrue to male traders more than female traders because they are more resistant to theft, which is prevalent in this specific sector (ESSET, 2007), then this would be another source of gender inequality in the sector.

The trend toward an increasing engagement with education, or education-related outcomes, for females (Demie, 2001; Nhantsi, 2010) is possibly found to be reflected in the 2010 data for the first time; unlike their male counterparts, females with more years of education are found to earn more and to have a significant association between tertiary-level education and increased earnings.

In 2009 the attendance of training since entry into the sector is found to be significantly associated with higher earnings. In the same context, attendance by street traders at street-trader skills' training has been found to be associated with higher levels of earnings, with a stronger effect for female traders (Callaghan & Venter, 2010); perhaps due to a benefit from the trading itself, or alternatively reflecting that more resourced traders may be more likely to have attended this training. Furthermore, the 2008 data weakly indicate that female traders that have attended training earn more (p < 0.10). These empirical findings suggest tentative support for the notion that the trend toward female dominance of higher education achievement in graduations or enrolments globally and locally (US Department of Education, 2007; Nhantsi, 2010) might possibly be reflected in the relationships around female education in this sector. It is suggested that further research should investigate this effect. If the gender gap in earnings in this sector was to be reducing, then this might, in the future, possibly be due to the learning effects of entrepreneurial development (Stevenson & Jarillo, 1990); which may be driven by education effects and human capital.

Experience, or time spent in the sector, is also found to be significantly associated with higher earnings only for female traders in 2008 and in 2010. However, only male traders that have been in the sector longer are found to earn more in 2009. If the 2009 effect is atypical, and the 2008 and 2010 effects do represent the basic underlying trend in the sector, then it might be possible that female traders reflect a different operationalisation of specific human capital. According to Becker (1975), human capital can be differentiated into general human capital, which is transferable across contexts, and into specific human capital, which is less transferable and is typically associated with learning specific to a certain job or context. In this case, these empirical findings seem to indicate that the productivity gains (Becker, 1975) from learning associated with specific human capital, or experience in this sector, may accrue to female traders only. If this was so, then this would support the overarching trend that seems to be evident in the increasing educational attainment of females (US Department of Education, 2007; Nhantsi, 2010). Further research is suggested with regard to investigating causal the mechanisms that might underpin this gender effect.

According to Tissington (2009), many traders in this context make their decision of what to sell on the basis of their ability to escape stock seizure by the authorities. Operating a rental stand is significantly associated with higher levels of earnings for both male and female traders in 2009, but not in 2008 and 2010. This might possibly indicate that after the xenophobic violence, increased policing or other consequences of these events might have decreased the earnings of traders without rental stands relative to those with rental stands. Further research is also recommended into the influence of childcare or responsibility for dependent children on street-trading earnings. It is acknowledged that family-to-work spillovers might also account for certain of the variance associated with earnings differences by gender in this context.

5.2 Null Hypothesis 2: There is no significant difference in rental stand operation by gender

On average, male traders are found to operate relatively more rental stands than female traders. A limitation inherent in the measurement of rental stand data might relate to the differences between individuals that hire the stands directly from the city (and operate the stands themselves for profit) versus those that are hired by them (and who earn a wage) to run the stand. It is acknowledged that gender differences might also be reflected in these relationships. Further research is recommended into these potential differences in this context. The Kruskal–Wallis tests, however, indicate that differences in rental stand operation by gender are not significant in any of the years sampled. The null hypothesis is not rejected.

5.3 Null Hypothesis 3: There is no significant difference in time spent in the sector (experience) by gender

According to the Kruskal–Wallis testing, experience or years spent in the street-trading sector is found to differ significantly according to gender in the 2009 and the 2010 data. On the basis of this trend in the data, the null hypothesis is rejected, and the alternative hypothesis is accepted. Female traders are found to report spending more time in the sector than male traders in all the three years. Age, years in Johannesburg and South African origin are all positively associated with time spent in the sector in all three years for both male and female traders. Males that are of Johannesburg origin are found to be in the sector longer in 2009. However, this effect is not significant for 2008 or for 2010. This might possibly suggest that relatively more traders not of Johannesburg origin might have temporarily left the sector after the xenophobic disruptions. These findings may suggest that the sector is being used by younger males as a more temporary means of entering Johannesburg and accessing the opportunities associated with the city. An interesting finding is that for female street traders there is a negative association between initial investment and time spent in the sector (albeit at just below the 10% level of significance), suggesting that female traders that could perhaps afford to invest more in their enterprises might have been more likely to have left the sector. The alternative explanation is that a relatively larger and relatively more recent influx of female traders with higher levels of initial investment may have occurred. For the male sample, the possible explanations are the reverse of these, as the sign of the variable is positive. However, the association of the male sample is found to be significant to within the 5% level of significance.

5.4 Null Hypothesis 4: There is no significant difference between levels of total education by gender

Total education is found to differ significantly by gender in all three years. The most significant difference is found in the case of the 2009 data. The null hypothesis is rejected, and the alternative hypothesis is accepted. Males are found to have higher endowments of education over this three-year period. This finding confounds the trend that has emerged in different educational contexts whereby females are beginning to dominate aspects of educational attainment (Demie, 2001; Nhantsi, 2010). Age, years in Johannesburg and South African origin are found to be negatively associated with total education for female street traders for the 2008, 2009 and 2010 samples. These findings might suggest that for the female sample, younger traders are more educated. Female traders of foreign origin and female traders that have entered Johannesburg relatively more recently may have relatively higher endowments of education. It is argued that the shorter tenure of more educated female traders, which has a stronger tested effect than that for male traders, indicates that higher levels of education might be enabling exit from the sector. It might be possible that the formal sector is ‘pulling’ more educated traders from the sector. It is possible that this effect is favouring women. However, such a ‘pull factor’ might favour women if most of the male traders are of foreign origin and face certification and legal barriers to formal employment.

5.5 Null Hypothesis 5: There is no substantive difference in factor structure of the associations of the sector with gender over time

The similar pattern of loadings of Factor 1 seems to reflect an underlying commonality in the data across the years 2008 and 2010 (Table 5). Factor 2 also reflects a commonality across 2008 and 2010, with schooling and tertiary education loading in 2008 and 2010, yet with initial investment and earnings also loading in 2010. In terms of the differences in the structure of relationships around earnings, for 2008 female traders' earnings loads as Factor 3 together with hours worked, initial investment and continuance satisfaction. Similarly, the male sample displays the same basic structure, with earnings loading with initial investment and continuance satisfaction, also as Factor 3. In 2009 earnings loads as Factor 2 with initial investment and days worked for both female and male samples. For the female sample, hours worked also loads on this factor, as does continuance satisfaction for the male sample. In the 2010 data, earnings load with initial investment and the educational factors as Factor 2. However, for the male sample, earnings load as Factor 3 with initial investment and continuance satisfaction. On the basis of these changes, the structure of the relationships around earnings for female street traders seems to have changed. This change, toward more of an association between earnings and the education factors, seems to reflect the changes in certain aspects of educational effects in the broader international and local societal contexts. On the basis of these findings, Hypothesis 5 is rejected. Conclusions and recommendations for further research are offered as follows.

Table 5: Factor analysis test results – factor structure differences by gender

Variable	Factor 1	Factor 2	Factor 3	Factor 4
2008
Female sample	Age	Schooling	Hours worked	Hours worked
	Experience	Tertiary education	Initial investment	Days worked
	Years in Johannesburg	Earnings
	Continuance satisfaction
Male sample	Schooling	Age	Initial investment	Hours worked
	Tertiary education	Experience	Earnings	Days worked
	Total education	Continuance satisfaction
2009
Female sample	Age	Hours worked	Initial investment
	Years in Johannesburg	Days worked	–Continuance satisfaction
	Experience	Initial investment
	–Total education	Earnings
Male sample	Age	Days worked	Hours worked
	Years in Johannesburg	Initial investment	–Total education
	Experience	Earnings
	Continuance satisfaction
2010
Female sample	Age	Initial investment	Hours worked
	Years in Johannesburg	Schooling	Days worked
	Experience	Tertiary education	Initial investment
	Continuance satisfaction	Total education
	Earnings
Male sample	Schooling	Age	Initial investment	Hours worked
	Tertiary education	Experience	Earnings	Days worked
	Total education	Years in Johannesburg	Continuance satisfaction
	–Continuance satisfaction

Notes: Minus sign represents negative loading.

6. Conclusions and recommendations for further research

This study attempted to explore the structure of relationships around gender of the informal street-trading sector of a large South African city. Theory relating to earnings and gender was first tested. The predictions of human capital theory (Becker, 1975) and theorists such as Marwell (1975) were not found to be supported with regard to earnings of street traders. The difference between gender outcomes in this sector supports research that has found gender differences in entrepreneurial and organisational contexts to still exist (Burke et al., 2002; Mueller, 2008; Robb, 2002). The overarching reduction in the gender gap indicated by certain literature (Hausmann et al., 2009), which suggests that gender inequality is reducing, was therefore not found to be supported in terms of earnings in this sector. In fact, males were found to earn significantly more than females in this sector. This gap is found to be increasing.

The effects around general human capital and gender were then investigated. Female traders were found to demonstrate a significant association between earnings and proxy measures of both specific and general human capital. On the basis of the findings, it is argued that the overarching trend of some kind of gender difference in favour of female traders associated with learning or education might also be present in this context (Demie, 2001; US Department of Education, 2007; Hausmann et al., 2009; Nhantsi, 2010; World Economic Forum, 2011). The significance of this association is taken to reflect important gender-related changes in the sector. Further research might offer insight into the mechanisms that underlie this gender effect.

These findings might suggest that the trajectory of gender differences associated with lower earnings of female traders may in time be off-set by the more effective or efficient operationalisation of specific and general human capital by female traders. This evidence is broadly taken to support the hypothesis of human capital theory (Becker, 1975); that learning-related investments of human capital which increase productivity are associated with higher earnings, which will over time make arbitrary differences such as gender less important.

It is further argued that two primary gender-related clusters exist in the sector akin to the predictions of the economic competition perspective of gender development (Emrich et al., 2004). Younger and more educated male traders of foreign origin that stay in the sector for shorter periods were found to comprise one core cluster type. Relatively older female traders of local origin that stay longer in the sector were found to comprise another. It is concluded that the former cluster might account for a certain amount of the difference in gender earnings in the sector, yet that physical security concerns might dominate the character of the sector. The latter effect might pose a fundamental challenge to female traders that might be more vulnerable to these contextual effects than male traders.

7. Recommendations for practice

On the basis of the findings, certain recommendations are suggested. Increased security in the city centre might reduce the unequal exposure of female traders to physical threats in this context. This would enable female traders to have more equal choices regarding hours and days worked. The xenophobic effects of 2008 might have had a disproportionate effect upon female traders. Policy-makers and political leaders should proactively create a space of tolerance in the broader society. Any societal violence might disproportionally affect females that are relatively more vulnerable to such effects. If female traders were to be disproportionately exposed to the earnings impact of by-law infringement on the basis of not being able to afford rental stands, then policy-maker interventions might be justified. More trading stalls, provided on a gender-equitable basis, might lessen potential inequality in this regard. If female traders have a more substantial return on specific and general human capital, the increased provision of specific skills training might enable a reduction in the gender gap in the sector. If females, globally and locally, seem to be more actively pursuing their human capital development, this might be a focus for policy-makers. Such a focus might provide a productivity mechanism whereby the gender gap could possibly be reversed across contexts. These measures might enable female traders to avoid the survivalist experience of street trading and to access the entrepreneurial space of informal development.