Economic bargaining power and financial decision-making among married and cohabitant women in South Africa

ABSTRACT

Women’s economic empowerment remains an important global policy issue and their involvement in household decision-making is critical for development. This paper aims to elucidate a key feature of bargaining models of intra-household decision-making, namely the dynamics between women’s economic bargaining power and their financial decision-making power. The paper also documents trends and transitions in financial decision-making. Using balanced panel data from South Africa’s National Income Dynamics Study (NIDS), the analysis finds that women who are married or living together remain economically disadvantaged compared to their male counterparts. Although gaining ground in terms of financial decision-making responsibility, women also remain disadvantaged compared to men in terms of financial decision-making power. Yet, employment opportunities significantly enhances women’s role in financial decision-making, thus arguing a case for continued efforts at women’s economic empowerment.

KEYWORDS:

• Gender
• economic bargaining power
• financial decision-making
• South Africa

JEL CODES:

• B54
• D13
• I39
• J16

1. Introduction

Women’s empowerment remains an important global policy issue with the United Nations’ earlier third Millennium Development Goal and the current Sustainable Development Goal (SDG) No. 5 clearly stipulating the significance of gender equality, including on the economic front. Lack of women’s economic empowerment manifests in various ways, including in the labour market. Duflo (2012:1052) for instance, states that, ‘women are less likely to work, earn less than men for similar work, and are more likely to be in poverty even when they work’. The World Bank (2016) also documents women as having limited economic opportunities, while in the African context, limitations to women’s economic participation remains a pronounced challenge (Wekwete, 2014). This is also the case in South Africa, where the feminisation of the labour force since the mid-1990s (Casale & Posel, 2002) has mainly resulted in an increase in unemployment rates among women (Casale, 2004). Furthermore, the gender wage gap, although on the decline, remains stark (Hinks, 2002; Casale, 2004; Kollamparambil & Razak, 2016). Wage discrimination persists (Hinks, 2002; Grün, 2004; Kollamparambil & Razak, 2016), with women being overrepresented in less secure and low-income jobs (Casale, 2004). Yet, access to paid work, as a pathway to empowerment, is a key resource in achieving gender equality, providing the resources to enable active agency, including participation in household decision-making (Kabeer, 2005).

This paper analyses this association between labour market outcomes and women’s financial decision-making responsibility and power in the South African context. Other studies have also examined this link between women’s employment status and their role in financial decision-making. Acharya et al. (2010) explore women’s participation in cash-based wage work as determinant of their participation in decisions on daily and large household purchases in Nepal. Boateng et al. (2014) also look into how the employment status of women in Ghana relates to their involvement in household decision-making on large and daily purchases. Antman (2014) investigates the relationship between the work status of Mexican spouses and the likelihood of their participation in decisions on large household expenditures. Pambè et al. (2014) analyse how working for cash is associated with decisions on major household purchases in Burkina Faso. This paper, however, investigates a wider range of labour market outcomes, including labour force participation and employment income, in addition to employment status.

The paper finds that women remain economically disadvantaged compared to men. Although gaining ground in terms of financial decision-making responsibility, women also remain disadvantaged compared to men in terms of financial decision-making power. Yet, employment opportunities significantly enhances women’s role in financial decision-making, thus arguing a case for continued efforts at women’s economic empowerment. The paper is structured as follows. Section 2 provides some theoretical context to the study. Sections 3 and 4 describe the data and methods. Section 5 presents the results and discussion, followed by limitations in Section 6. Section 7 concludes.

2. Theoretical perspective

In Economics, collective-type models of cooperative decision-making within households stipulate that members with greater decision-making power have the leverage to influence the household’s allocation of resources and the resultant outcomes in the direction of their preferences (Bourguignon et al. 2009; Browning et al. 2014). While there is some empirical support for these Pareto efficient models (Lundberg et al. 1997; Quisumbing & Maluccio, 2003; Bobonis, 2009; Majlesi, 2016), there are contrary evidence of inefficiencies (Angelucci & Garlik, 2015; De Rock et al. 2017), including in the African context (Udry, 1996; Baland & Ziparo, 2017). There moreover, are alternatives to these cooperative models of intra-household decision-making, such as semi-cooperative (Fletschner, 2009; Malapit, 2012; Cherchye et al. 2015) and non-cooperative models (Chiappori & Mazzocco, 2017). There is also a distinction between static and dynamic models of intra-household decision-making (Chiappori & Mazzocco, 2017). Nevertheless, static cooperative models of the collective type very much remain the current benchmark (Chiappori & Mazzocco, 2017), hence its adoption in this study as a conceptual frame for an investigation into one element of decision-making power, namely the link between economic bargaining power and financial decision-making roles. To be clear, the empirical analysis presented here is not a test of a particular theory of intra-household decision-making, but employs the static collective model with its emphasis on decision power as theoretical context.

3. Data

South Africa’s National Income Dynamics Study (NIDS) is nationally representative. The survey, with its stratified two-stage cluster sample design (Woolard et al. 2010), at baseline comprised data for 28,255 individuals from 7,305 households. This paper uses data from the first three consecutive waves of the panel (2008, 2010, and 2012). Insofar as the focus is on intra-household decision-making, the analytical sample comprises those respondents (15 years and older) who reported their marital status as ‘married’ or ‘living with partner’ (N = 17,167). The analytical sample is further restricted to the balanced panel (n = 2,737; N = 8,211), because the study aims to document trends as well as transitions in financial decision-making responsibility and power.

Table 1 describes this sample in relation to key socio-demographic characteristics. The respondents in the analytical sample are relatively old at around 50 years. As expected, the sample has aged over time, while the duration of relationships has increased correspondingly. The sample includes six women for every four men. The proportion of respondents that are household heads or are married exhibits a statistically significant upward trend. As some of those that had completed secondary school progressed to graduation, the proportion with tertiary education increased and that with grade 12 declined markedly. Household size increased post-baseline and by a statistically significant margin.

Table 1. Sample characteristics, by survey round.

	2008	2010	2012	Panel
Age (years)	46.38 (0.24)	48.81 (0.24)	50.58 (0.24)	48.59 (0.14)***
Female (%)				61.05 (0.053)
Education (%):
None	16.87 (0.71)	16.28 (0.70)	16.50 (0.70)	16.55 (0.40)***
Primary	31.50 (0.88)	31.68 (0.88)	31.31 (0.88)	31.50 (0.51)
Some secondary	31.50 (0.88)	31.46 (0.88)	31.24 (0.88)	31.40 (0.51)
Grade 12	12.18 (0.62)	10.57 (0.58)	9.03 (0.54)	10.60 (0.33)
Tertiary	7.94 (0.51)	10.02 (0.57)	11.92 (0.61)	9.96 (0.33)
Education (years)	7.16 (0.08)	7.26 (0.08)	7.32 (0.08)	7.25 (0.05)
Race (%):
African				72.96 (0.84)
Coloured				19.40 (0.75)
Asian				1.83 (0.25)
White				5.81 (0.44)
Headship (%)	44.93 (0.95)	50.40 (0.95)	52.91 (0.95)	49.42 (0.55)***
Married (%)	82.83 (0.72)	84.95 (0.68)	86.85 (0.64)	84.87 (0.39)***
Relationship duration (years)	18.71 (0.25)	20.52 (0.25)	22.43 (0.25)	20.55 (0.14)***
Household size	4.92 (0.04)	5.24 (0.05)	5.20 (0.05)	5.12 (0.02)***
Sample (n)	2,737	2,737	2,737	8,211

Notes: Results are for a balanced panel of adults who are either married or living together. Standard errors are reported in parentheses. Statistical significance ***p < 0.01; **p < 0.05; *p < 0.10.

4. Methods

The statistical analysis follows a structured approach. First, gender-based comparisons of economic bargaining power are performed (Table 2). Economic bargaining power in relation to the labour market is measured by three proxy factors. Labour force participation is a binary variable taking on a value of ‘1’ (economically active, i.e. employed or looking for work) or ‘0’ (not economically active). Employment status is a binary variable taking the value of ‘1’ if the individual is employed and ‘0’ otherwise. Employment income, a continuous variable, is measured in real South African Rand (ZAR).

Table 2 . Economic bargaining power, by gender and survey round.

	Year	Male	Female	Total	Sample (n)
Labour force participation (%)	2008	74.02	61.43	66.33	2,712***
	2010	62.35	45.08	51.82	2,717***
	2012	63.15	47.38	53.53	2,729***
	Panel	66.49	51.29	57.21	8,158***
Employed (yes/no)	2008	64.17	39.46	49.07	2,712***
	2010	54.62	32.22	40.96	2,717***
	2012	55.16	34.17	42.35	2,729***
	Panel	57.97	35.28	44.12	8,158***
Unemployment rate (%)	2008	13.31	35.75	26.01	1,799***
	2010	12.40	28.51	20.95	1,408***
	2012	12.64	27.88	20.87	1,461***
	Panel	12.81	31.20	22.87	4,668***
Employment income (ZAR) – total	2008	2,539.32	774.51	1,461.08	2,735***
	2010	2,635.30	1,046.41	1,664.54	2,735***
	2012	2,692.06	1,018.24	1,670.15	2,737***
	Panel	2,622.27	946.39	1,598.61	8,207***
Employment income (ZAR) – economically active	2008	3,420.05	1,254.08	2,193.22	1,822***
	2010	4,210.15	2,297.72	3,190.28	1,427***
	2012	4,257.77	2,140.23	3,111.79	1,469***
	Panel	3,932.18	1,833.18	2,780.80	4,718***
Employment income (ZAR) – employed	2008	3,990.90	1,978.93	3,002.30	1,331***
	2010	4,842.77	3,274.46	4,090.32	1,113***
	2012	4,888.82	2,990.30	3,954.34	1,156***
	Panel	4,544.51	2,700.21	3,644.39	3,600***

Notes: Results are for a balanced panel of adults who are either married or living together. Unemployment rates are calculated according to the broad definition. Statistical significance ***p < 0.01; **p < 0.05; *p < 0.10.

Second, trends in financial decision-making responsibility and power are compared across survey rounds (Table 3). Financial decision-making responsibility is represented by a binary variable taking the value ‘1’ if the respondent is a financial decision-maker and ‘0’ otherwise. Financial decision-making power in turn, is a categorical variable taking the value ‘3’ if the respondent is a main decision-maker, the value ‘2’ if the respondent is a joint decision-maker, and the value ‘1’ if the respondent is a none-decision-maker. These financial decision-making indicators were constructed from the combined responses to the two questions on decision-making in regards to ‘day-to-day household expenditure’ and ‘large, unusual purchases’.¹ Next, transitions in both financial decision-making responsibility and power are described in Table 4.

Table 3. Financial decision-making responsibility and power, by gender and survey round.

	Male	Female	Total
2008:
Yes	96.99	90.06	92.76***
Sample (n)	1,066	1,671	2,737
Joint	9.57	39.93	27.57***
Main	90.43	60.07	72.43
Total	100.00	100.00	100.00
Sample (n)	1,034	1,505	2,539
2010:
Yes	91.65	92.99	92.76
Sample (n)	1,066	1,671	2,737
Joint	21.29	25.93	24.14***
Main	78.71	74.07	75.86
Total	100.00	100.00	100.00
Sample (n)	977	1,554	2,531
2012:
Yes	95.77	97.00	96.52*
Sample (n)	1,066	1,671	2,737
Joint	12.63	28.87	22.60***
Main	87.37	71.13	77.40
Total	100.00	100.00	100.00
Sample (n)	1,021	1,621	2,642
Panel:
Yes	94.80	93.35	93.92***
Sample (n)	3,198	5,013	8,211
Joint	14.38	31.45	24.74***
Main	85.62	68.55	75.26
Total	100.00	100.00	100.00
Sample (n)	3,032	4,680	7,712

Notes: Results are for a balanced panel of adults who are either married or living together. Statistical significance ***p < 0.01; **p < 0.05; *p < 0.10.

Table 4. Decision-making transitions.

		No	Yes	Total	Sample (n)
(a) Transitions in financial decision-making responsibility, by gender
Male	No	12.40	87.60	100.00	121
	Yes	5.92	94.08	100.00	2,011
Female	No	13.78	86.22	100.00	283
	Yes	4.18	95.82	100.00	3,059
Total	No	13.37	86.63	100.00	404
	Yes	4.87	95.13	100.00	5,070
		None	Joint	Main	Total	Sample (n)
(b) Transitions in financial decision-making power, by gender
Male	None	12.40	19.83	67.77	100.00	121
	Joint	7.17	15.96	76.87	100.00	307
	Main	5.69	15.49	78.81	100.00	1,704
Female	None	13.78	26.50	59.72	100.00	283
	Joint	5.48	29.88	64.64	100.00	1,004
	Main	3.55	24.14	72.31	100.00	2,055
Total	None	13.37	24.50	62.13	100.00	404
	Joint	5.87	26.62	67.51	100.00	1,311
	Main	4.52	20.22	75.26	100.00	3,759
		Joint	Main	Total	Sample (n)
(c) Transitions in joint and main financial decision-making, by gender
Male	Joint	17.19	82.81	100.00	1,234
	Main	16.43	83.57	100.00	3,589
Female	Joint	31.61	68.39	100.00	285
	Main	25.03	74.97	100.00	1,607
Total	Joint	28.28	71.72	100.00	949
	Main	21.18	78.82	100.00	1,982

Note: Results are for a balanced panel of adults who are either married or living together.

Regression analysis then is used to determine the role of gender in predicting financial decision-making responsibility and power (Table 5). In this first set of regression models, financial decision-making is represented by four different binary variables, namely ‘yes’ (=1) versus ‘no’ (=0); ‘joint’ (=1) versus ‘none’ (=0); ‘main’ (=1) versus ‘none’ (=0), and ‘main’ (=1) versus ‘joint’ (=0). The general regression function in this case is represented as:(1) $\mathrm{F}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{i}\mathrm{a}\mathrm{l}\mathrm{d}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}-\mathrm{m}\mathrm{a}\mathrm{k}\mathrm{i}\mathrm{n}\mathrm{g}=\mathrm{\Omega }(\mathrm{g}\mathrm{e}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{r};\mathrm{s}\mathrm{o}\mathrm{c}\mathrm{i}\mathrm{o}-\mathrm{d}\mathrm{e}\mathrm{m}\mathrm{o}\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{p}\mathrm{h}\mathrm{i}\mathrm{c}\mathrm{s};\mathrm{s}\mathrm{u}\mathrm{r}\mathrm{v}\mathrm{e}\mathrm{y}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r})$(1)

Table 5. Gender as a predictor of financial decision-making responsibility and power.

		Decision-making power
	Decision-making responsibility	Joint vs none	Main vs none	Main vs joint
A. Pooled
Female (yes/no)	1.462** (0.214)	2.054*** (0.327)	1.315* (0.200)	0.662*** (0.048)
Sample (N)	7,534	2,099	5,774	7,195
Wald chi^2	408.27***	222.03***	432.72***	672.42***
B. Random effects (RE)
Female (yes/no)	1.526** (0.273)	2.572*** (0.584)	1.358** (0.252)	0.659*** (0.048)
Sample (N)	7,534	2,099	5,774	7,195
Wald chi^2	266.04***	107.08***	266.17***	606.73***
LR chi^2	28.22***	26.81***	17.27***	0.56

Notes: Odds ratios. Results are for a balanced panel of adults who are either married or living together. Regression models control for age, age square, years of education, race, headship, marital status, relationship duration, household size, place of residence (geo-type) and survey round. Robust standard errors (SE) for the odds ratios (OR) are reported in parentheses. Statistical significance ***p < 0.01; **p < 0.05; *p < 0.10.

The econometric analysis employs logistic regression models, with coefficients reported as odds ratios (OR). As gender is time-invariant and given the panel structure of the data, this component of the regression analysis employs random effects (RE) panel regression models. For comparative purposes, the results of the corresponding pooled regression models are reported alongside. A likelihood ratio test is employed to verify the econometric merit of the random effects model. The socio-demographic controls include the following variables: age, age square, years of education, race, headship, marital status, relationship duration, household size, and place of residence (geo-type).

Table 6 documents the differences in economic bargaining power across women who are joint and main financial decision-makers. Lastly, the association between economic bargaining power and women’s financial decision-making power is established econometrically (Table 7). The single binary dependent variable in this second set of regression models draws a distinction between joint (=0) and main (=1) financial decision-makers. The general model specification is as follows:(2) $\mathrm{F}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{i}\mathrm{a}\mathrm{l}\mathrm{d}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n}-\mathrm{m}\mathrm{a}\mathrm{k}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{p}\mathrm{o}\mathrm{w}\mathrm{e}\mathrm{r}=\mathrm{\Omega }(\mathrm{e}\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{i}\mathrm{c}\mathrm{b}\mathrm{a}\mathrm{r}\mathrm{g}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{p}\mathrm{o}\mathrm{w}\mathrm{e}\mathrm{r};\mathrm{s}\mathrm{o}\mathrm{c}\mathrm{i}\mathrm{o}-\mathrm{d}\mathrm{e}\mathrm{m}\mathrm{o}\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{p}\mathrm{h}\mathrm{i}\mathrm{c}\mathrm{s};\mathrm{s}\mathrm{u}\mathrm{r}\mathrm{v}\mathrm{e}\mathrm{y}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r})$(2) Due to concerns with multicollinearity, separate logistic regression models are estimated for each economic bargaining power factor, i.e. labour force participation, employment status, and employment income. Insofar as issues of endogeneity arise because there may be unobserved but fixed factors that are correlated with both the environments in which economic bargaining power and financial decision-making roles are determined (Antman, 2014), gender roles and norms being a case in point, the fixed effects (FE) estimator is accorded special attention. In each case, the results of the pooled, random effects (RE) and fixed effects (FE) estimators are presented for comparative purposes. The appropriateness of the random effects (RE) logistic regression model is assessed with the aid of a likelihood ratio test. The Hausman test is employed to determine whether the fixed effects (FE) logistic regression model is the appropriate estimator. The control variables are the same as in the first regression model (i).

Table 6. Labour market outcomes and women’s financial decision-making power.

	Labour force participation (%)	Employed (%)		Employment income (ZAR)
	Joint	Main	Joint	Main	Joint	Main
Total	50.10	51.52	32.85	37.55***	814.73	1,054.05**
Sample (n)	1,461	3,187	1,461	3,187	1,472	3,208
Economically active			65.57	72.89***	1,614.12	2,033.32**
Sample (n)			732	1,642	743	1,663
Employed					2,498.53	2,824.91
Sample (n)					480	1,197

Notes: Results are for F-tests applied to the pooled balanced panel of women who are either married or living together. Statistical significance ***p < 0.01; **p < 0.05; *p < 0.10.

Table 7. Labour market outcomes as predictors of women’s financial decision-making power.

	Labour force participation (yes/no)			Employed (yes/no)			Employment income (ZAR)
	Pooled	Random effects (RE)	Fixed effects (FE)	Pooled	Random effects (RE)	Fixed effects (FE)	Pooled	Random effects (RE)	Fixed effects (FE)
Total	1.208** (0.093)	1.212** (0.096)	1.155 (0.140)	1.287*** (0.097)	1.295*** (0.101)	1.254* (0.156)	1.016 (0.012)	1.016 (0.014)	1.028 (0.020)
Sample (N)	4,331	4,331	2,373	4,331	4,331	2,373	4,361	4,361	2,401
Wald chi^2 / F-test	307.64***	292.61***	10.08***	315.46***	298.55***	10.19***	304.32***	284.79***	9.84***
LR chi^2		1.19			1.24			1.21
Hausman chi^2			80.42***			79.83***			73.72***
Economically active				1.282** (0.139)	1.287** (0.142)	1.277 (0.287)	1.017 (0.014)	1.018 (0.015)	1.032 (0.023)
Sample (N)				2,272	2,272	924	2,302	2,302	936
Wald chi^2 / F-test				167.31***	153.44***	4.28***	164.15***	149.00***	4.30***
LR chi^2					0.27			0.34
Hausman chi^2						18.44			17.12
Employed							1.000 (0.000)	1.000 (0.000)	–
Sample (N)							1,617	1,617
Wald chi^2 / F-test							115.94***	100.48***
LR chi^2								0.00
Hausman chi^2

Notes: Odds ratios. Outcome is main (=1) versus joint (=0) financial decision-making status. Results are for a balanced panel of women who are either married or living together. Regression models control for age, age square, years of education, race, headship, marital status, relationship duration, household size, place of residence (geo-type) and survey round. Robust (pooled/random effects) or jack knife (fixed effects) standard errors (SE) for the odds ratios (OR) are reported in parentheses. Statistical significance ***p < 0.01; **p < 0.05; *p < 0.10.

5. Results and discussion

5.1. Economic bargaining power

Gender disparities in access to labour market opportunities and the benefits thereof remain stark and persistent (Table 2). Women are less likely than their male counterparts to participate in the labour market and less likely to be in employment. Unemployment rates among women are also significantly greater than among men. Differences in employment income are substantial regardless of whether income is compared across all respondents or across those who are economically active or the employed.

5.2. Gender and financial decision-making

Participation in financial decision-making among men and women who are married or living together is in excess of 90%. According to the results in Table 3, women’s overall financial decision-making responsibility has increased over time, from 90.0% in 2008 to 92.9% and 97.0% in 2010 and 2012, respectively. While the level of responsibility for financial decision-making in 2008 is greater for men, the opposite is true in 2010 and 2012. Yet, women hold less financial decision-making power than men. In each of the survey years and on aggregate more women are joint rather than main financial decision-makers. The difference in joint financial decision-making is particularly stark in 2008: 9.5% versus 39.9% for men and women, respectively.

Inter-temporal transitions in financial decision-making responsibility are almost identical for men and women (Table 4(a)). A majority of the few respondents who are not financial decision-makers had by follow-up taken on the role of financial decision-maker, while few decision-makers relinquished their financial decision-making role between survey rounds. The transition matrices reported in Table 4(b,c) however tell an important story. Firstly, women not making financial decisions, compared to men, are more likely to later take on the role of joint (26.5% versus 19.8%) rather than main financial decision-maker (59.7% versus 67.7%). Secondly, women (29.8%) are more likely to remain joint financial decision-makers between survey rounds compared to men (15.9%). Concomitantly, fewer women (64.6%) are promoted from a joint to main financial decision-making role compared to men (76.8%). Lastly, women are less likely than men to retain their main financial decision-making role between survey rounds (72.3% versus 78.8%). When the analysis is constrained to decision-makers, similar dynamics emerge. Women are more likely than are men to remain joint financial decision-makers (31.6% versus 17.1%). In turn, women are less likely than are men to remain main financial decision-makers (74.9% versus 83.5%). Together, these findings underscore the relatively disadvantaged position of women in regards to financial decision-making power. Pahl (1995), Dema-Moreno (2009), and Cantillon et al. (2016), likewise, have documented the inequitable gender dynamics in financial organisation and decision-making in couples.

The first three random effects logistic regression models, which all perform adequately in terms of overall fit and pass the likelihood ratio test, show that women are more likely than men to take responsibility for financial decisions. The effect is not only statistically significant (p < 0.01), but substantive in economic terms, with the likelihood of holding the financial decision-making responsibility being 52.6% greater for women than men. Further supporting this conclusion are the large, positive and statistically significant odds ratios on the outcomes for joint and main financial decision-making. Yet, a comparison of these two odds ratios also underscore the fact that women generally take on the role of joint rather than main financial decision-maker. Whereas women are 35.8% more likely than men to be main as opposed to none-decision-makers, women are more than two-and-a-half times more likely than men to be joint as opposed to none-decision-makers. The respective standard errors suggest that the difference between these two odds ratios is statistically significant. Women’s relatively lack of financial decision-making power is further exemplified by the negative and statistically significant effect of gender on the likelihood of being a main rather than a joint financial decision-maker. In this case, the likelihood ratio test (chi² = 0.56) suggests that the pooled regression model, which also passes the test for overall fit, outperforms the random effects model. According to the pooled model, women are 33.8% less likely than men to be main rather than joint financial decision-makers, which represents a substantial margin in terms of economic significance. The next question is whether improvements in women’s economic bargaining power can address this imbalance in financial decision-making power.

5.3. Women’s economic bargaining power and financial decision-making agency

According to Table 6, female labour force participation rates are slightly higher in main than joint financial decision-makers. The difference however does not meet any of the criteria for statistical significance. Employment, however, is statistically significantly more prevalent among main than joint financial decision-makers, both for women in general and among economically active women. The same is true for employment income: women who are main financial decision-makers earn statistically significantly more than women who are joint financial decision-makers. Among employed women, earned income is also greater for main than joint financial decision-makers, but not by a statistically significant margin.

The test for determining whether economic empowerment enhances women’s financial decision-making power is the logistic regression analysis reported in Table 7. The individual regression models all pass the test for overall goodness-of-fit represented by the Wald- and F-statistics. The results for the six sets of regression models are now discussed in turn. Based on the results of the pooled and random effects regression models, the likelihood of being a main rather than a joint financial decision-maker increases by approximately twenty percent when women are economically active. In the fixed effects framework, however, which is appropriate here in relation to the Hausman test statistic, labour force participation does not impact significantly on women’s financial decision-making power.

The fixed effects estimator is also the relevant econometric model for the first analysis of the effects of employment status on women’s financial decision-making power. Here, the effect is large in economic terms (i.e. being employed increases the likelihood of women being main rather than joint financial decision-makers by as much as 25.4%), but relatively weak in statistical terms (i.e. being significant at the 10% level only). The fixed effects result does not hold however when the analysis is restricted to the sub-sample of economically active women, although the odds ratio is equivalent in size (OR 1.277). In fact, the Hausman test statistic disqualifies the fixed effects estimator (chi² = 18.44), while the random effects estimator is also not acceptable on econometric grounds (chi² = 0.27). The pooled analysis, therefore, should be the focus of attention. In this model, the likelihood of a woman being a main rather than joint financial decision-maker increases by 28.2% where the said woman is employed. The result is significant at the 5% level.

The analyses for employment income, where the fixed effects estimator is appropriate for the full female sample and the pooled estimator for the sub-sample of economically active women, reveals no significant relationship with women’s financial decision-making power. The regression analysis, therefore, shows that the significant differences illustrated in the bivariate analysis (Table 6) does not translate into significant effects when allowing for the panel nature of the data and when controlling for observable covariates. (In the case of employed women, the fixed effects regression model could not be estimated due to insufficient variation across survey rounds in the study outcome. Here, the pooled estimator, which is superior to the random effects estimator, also shows that the odds ratio is not significantly different from one.) In other words, higher levels of employment income does not translate into greater financial decision-making power.

To summarise, only employment status enhances women’s financial decision-making power. Similarly, studies by Acharya et al. (2010), Antman (2014), Boateng et al. (2014) and Pambè et al. (2014) support the finding that the employment status of women positively influences women’s decision-making role in the household’s financial matters.

6. Limitations

The paper has five important limitations. As a result of adopting the balanced panel as analytical sample, for reasons explained above, the study is subject to selection bias. Table A in the annexure shows that this is the case for each of the aggregate and sub-samples. Basically, those respondents included in the study are better educated; Coloured rather than African and African rather than White in terms of race; legally or traditionally married rather than living together; in relationships of longer duration; and not residing in urban areas of the country. Furthermore, basic descriptive analysis reveals that there is also selection bias in terms of the main decision-making outcome. The balanced panel includes fewer none-decision-makers (6.0% versus 11.6%) and a larger proportion of joint (23.2% versus 21.9%) and main financial decision-makers (70.6% versus 66.4%). Important to emphasise, therefore, is that this study’s findings apply to a very distinct group of respondents and cannot be generalised to all of South Africans who are married or living together with a partner.

Decision-making questions moreover were posed with no clear distinction between high-level decisions on the allocation of resources to expenditure categories and instrumental decisions regarding the management of the allocated resources (Holden, 2011; Lauer & Yodanis, 2011; Skogrand et al. 2011). The findings of greater involvement of females in financial decision-making could reflect household members’ involvement in the instrumental management of household resources and not an overall allocative control. Financial decision-making, as measured here, cannot distinguish instrumental from allocative control, calling for further research, including nationally representative surveys with a more carefully designed module on intra-household decision-making, including measures such as Vaz et al.’s (2016) Relative Autonomy Index (RAI) and specific questions on the organisation of household finances (Vogler 2005). Recent research on household bargaining models moreover has aimed to construct more credible measures of decision power by accounting for actual control over specific households assets (Reggio, 2011) or changes in inheritance laws and rights (Calvi, 2017; Calvi et al. 2017). Other researchers in turn have investigated the bargaining dynamics in households via the random assignment of mothers’ and fathers’ eligibility to receive cash transfers, both in Africa (Benhassine et al. 2015; Akresh et al. 2016) and beyond (Gitter & Barham, 2008; Almås et al. 2018).

It is important furthermore to recognise that employment is just one pathway to gender empowerment, which Kabeer (1999) describes as a multi-faceted process. In fact, in the context of the economics of intra-household decision-making, an unemployed spouse may be more ‘empowered’ than their working partner if they are the dictator in the relationship enjoying both leisure and consumption. Positive labour market outcomes, therefore, are not necessarily synonymous with gender empowerment.

Due to the presence of reverse causality (i.e. financial decision-making power also potentially determines labour force participation and associated labour market outcomes such as employment status and employment income), the findings presented in this paper should be interpreted as associations rather than causal claims.

Lastly, it is also important to recognize that the relatively short time span of the panel (approximately four years) may make it difficult to detect significant links between economic bargaining power and financial decision-making. Some of the underlying drivers of these outcomes, such as gender roles and norms, may be slow to shift significantly over such short period. What is required, are further follow-up studies incorporating future waves of the National Income Dynamics Study (NIDS).

7. Concluding remarks

This paper investigated the association between labour market outcomes and financial decision-making roles in the South African context using the National Income Dynamics Study (NIDS). The paper finds that economic bargaining power is still a gendered phenomenon in South Africa, i.e. women are less likely to participate in the labour market and to be in employment, and earn less. Female household members married to or living with their partners therefore are less economically empowered than their male counterparts, a finding that confirms the importance of the United Nations’ Sustainable Development Goal (SDG) No. 5. Women, despite being more likely than men to be identified as financial decision-makers, lack financial decision-making power, i.e. they are joint rather than main decision-makers. According to the results, labour market opportunities, in the form of employment, plays a significant role in advancing women’s financial decision-making power. Expanding employment opportunities for women through both employment equity policies and targeted active labour market policies such as job search assistance and training programmes (Card et al. 2010) is therefore an important policy consideration in the broader development context.

In fact, development stands to benefit when women are empowered. Improvements in women’ economic bargaining power, for example, has been reported to translate into greater household expenditure on food (Hopkins et al. 1994; Hoddinott & Haddad, 1995; Schmeer, 2005; Gummerson & Schneider, 2013) and education (Quisumbing & Maluccio, 2003) and lower household expenditure on vices such as alcohol and cigarettes (Hoddinott & Haddad, 1995; Gummerson & Schneider, 2013). The empowerment of women has also impacted positively, not only on the wellbeing of women themselves, but so too on their children’s health and education. The broader empowerment of women in the developing world is positively associated with various health outcomes, including antenatal care, skilled attendance at birth, contraceptive use, child mortality, vaccination, nutritional status, and exposure to intimate partner violence (Pratley, 2016; Thorpe et al. 2016). School enrolment has also been shown to be associated with women’s decision-making autonomy in rural Mozambique (Luz & Agadjanian, 2015) and the Honduras (Hendrick & Marteleto, 2017).

Acknowledgement

The contents of this paper in no way reflect the views of the Department of Planning, Monitoring (DPME) and Evaluation and the European Union (EU).

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Sevias Guvuriro http://orcid.org/0000-0002-1365-3591

Additional information

Funding

This research was supported financially by the Programme to Support Pro-Poor Policy Development (PSPPD), a partnership programme of the Department of Planning, Monitoring and Evaluation (DPME) and the European Union (EU) [grant number PSPPD2/CfP1/2013/57].

Notes

1 For each financial decision-making sphere, the respondent has to identify the main decision-maker within the household, and where relevant, also the joint decision-maker. For the purpose of the analysis conducted in this paper, respondents were assigned their ‘highest’ recorded level of decision-making power. In other words, if the respondent identified him/herself as ‘main’ decision-maker in any one of the two financial decision-making spheres, or any other respondent identified the person as ‘main’ decision-maker in any one sphere, the person was assigned the status of ‘main’ decision-maker. Next, respondents not designated as main decision-makers were assigned the status of ‘joint’ decision-maker if they themselves or any other respondent accorded them the role of ‘additional’ decision-maker in any one of the two financial decision-making spheres. None-decision-makers are those respondents who did not identify themselves as decision-makers and was not identified as decision-makers by any other survey respondent.

Appendix

Table A1. Predictors of selection bias.

		Women
	Total	Total	Economically active	Employed
	OR (SE)	OR (SE)	OR (SE)	OR (SE)
Employed (yes/no)	1.089 (0.040)**	1.068 (0.051)	0.921 (0.063)	–
Employment income (R ‘000)	–	–	–	0.999 (0.000)
Age (years)	1.008 (0.002)***	1.004 (0.003)	1.006 (0.004)	1.002 (0.005)
Female (%)	1.581 (0.062)***	–	–	–
Education (years)	1.011 (0.004)**	1.014 (0.006)**	1.017 (0.009)*	1.019 (0.011)*
Race (comparison = African):
Coloured	1.136 (0.053)***	1.226 (0.079)***	1.213 (0.103)**	1.204 (0.119)*
Asian	0.608 (0.070)***	0.660 (0.104)***	0.752 (0.172)	0.800 (0.215)
White	0.404 (0.030)***	0.369 (0.037)***	0.310 (0.044)***	0.359 (0.055)***
Headship	1.072 (0.039)*	0.976 (0.048)	0.964 (0.067)	0.962 (0.080)
Married	2.189 (0.097)***	2.390 (0.141)***	2.068 (0.157)***	1.032 (0.005)***
Relationship duration (years)	1.018 (0.002)***	1.022 (0.003)***	1.035 (0.004)***	2.065 (0.201)***
Household size	0.988 (0.005)**	1.003 (0.008)	1.036 (0.014)**	1.073 (0.020)***
Wave (comparison = 2008):
2010	1.060 (0.045)	1.079 (0.062)	1.038 (0.083)	1.167 (0.112)
2012	0.891 (0.036)***	0.885 (0.049)**	0.742 (0.055)***	0.879 (0.079)
Geo-type (comparison = traditional):
Urban	0.842 (0.036)***	0.830 (0.047)***	0.849 (0.066)**	0.808 (0.080)**
Farms	1.172 (0.070)***	1.206 (0.096)**	1.216 (0.136)*	1.205 (0.159)
Sample (n)	15,516	8,775	4,716	3,225
Wald chi^2	1,111.45***	669.40***	410.93***	278.83***

Notes: Results are for the pooled sample of adults who are either married or living together. Outcome is being in the balanced panel (=1) or not (=0). Robust standard errors (SE) for the odds ratios (OR) are reported in parentheses. Statistical significance ***p < 0.01; **p < 0.05; *p < 0.10.