Theoretical perspectives on the allocation of resources within households

Early economic models of household resource allocation followed the notion of a household with a single decision-maker. That is, traditionally household behaviour was analysed in the same way as the behaviour of an individual with a unique utility function and subject to a single budget constraint (Reference BeckerBecker, 1981). This approach implied that household resources are ‘pooled’ such that they would be allocated in a way that was proportional to the needs of different household members, regardless of who contributed towards their provision.

More sophisticated and realistic models have emerged in recent decades to account for differences in the preferences of individuals within the household. Bargaining models, and their later generalisation to collective models, describe households as being comprised of individuals with their own separate utility functions, and most often income pooling is not assumed (see for example, Reference Apps and ReesApps and Rees, 1988; Reference ChiapporiChiappori, 1988). In bargaining models, household decision-making outcomes are determined by the man’s and the woman’s ‘threat position’ – this is the utility each would receive should the partnership breakdown. In collective models, household decision-making is described as being focused on the maximisation of household utility, with this comprised of the weighted sum of each partner’s utility function. The achieved outcomes of each partner depend, in part, on their bargaining power within the household as this would affect his or her threat position in a bargaining model, or the Pareto weights in a collective model. In turn, bargaining power is, described as being linked to ‘distribution factors’, such as the relative income of each partner (Reference Friedberg and WebbFriedberg and Webb, 2006), their relative wage rates (Reference PollakPollak, 2005), or their relative labour market status (Reference Lundberg, Pollak and WalesLundberg et al., 1997; Reference ThomasThomas, 1990; Reference TiefenthalerTiefenthaler, 1999).

In feminist economics, key works on intra-household resource allocation have been informed by Sen’s cooperative conflict model (see for example, Reference AgarwalAgarwal, 1997; Reference KabeerKabeer, 1997; Reference Himmelweit, Himmelweit, Simonetti and TriggHimmelweit, 2001; Reference IversenIversen, 2003; Reference PurkayasthaPurkayastha, 2003; Reference Nyman, Dema, Stocks, Diaz and HallerodNyman and Dema, 2007; Reference Sung and BennettSung and Bennett, 2007). In this model, households are sites for potential gains from cooperation, but they are also sites of potential conflict between partners who might have different preferences about the use of household resources. As with other models, the cooperative conflict model predicts that the intra-household distribution of wellbeing will be affected by the bargaining power of the partners and, thus, their fallback positions. However, the model is unique in its inclusion of distribution factors that derive from gender norms. Reference Sen and TinkerSen (1990) highlights that an individual’s perceptions of the value of their own and their partner’s contributions to household resources can affect their sense of whether they (or their partner), is entitled to a share of household resources. Under the influence of gender norms, perception biases may ignore or understate the value of non-market activities (such as cooking and caring for children) within the household, and only consider paid employment and other market roles as ‘deserving’ contributions. With women taking up the majority, of domestic tasks, norms that either devalue unpaid work or link entitlements over financial resources to paid work are potentially an important source of unequal resource allocation and a gendered distribution of wellbeing within households. For example, within couples that adhere to male breadwinner ideologies the man’s power is naturalised in a way which reduces the value attached to the woman’s financial contributions and her claim over household resources (Reference BennettBennett, 2013). However, perceptions matter not only to the valuation of contributions but also to how households internalise and reproduce gender inequalities.

Empirical studies of the distribution of wellbeing in couple households

One of the main challenges in understanding the processes of distribution of resources within households is its unobservability. In an attempt, to understand factors which capture each partner’s relative access to household resources, recent studies have utilised data based on subjective assessments. ^{Footnote 1} A number of these studies have observed the links between men’s and women’s relative income shares and their satisfaction with household finances. They have indirectly tested for income pooling by assessing whether changes in relative income shift the partners’ opportunities to benefit from household resources, with these opportunities proxied by their relative financial satisfaction. If couples pool their income, changes in relative income shares, should have no impact on the distribution of financial satisfaction within the household, ceteris paribus.

The findings from these studies often rejected the assumption of income pooling. For instance, Reference Alessie, Crossley and HildebrandAlessie et al. (2006) found that on average in Denmark, France, Greece, Italy, Portugal, Spain and the United Kingdom, an increase in the woman’s income share results in an increase in her own financial satisfaction but a lower level of financial satisfaction for her partner. Reference BonkeBonke (2008) found mostly similar patterns in his study using the same data. Using a Danish sample, Reference Bonke and BrowningBonke and Browning (2009) and Reference Ahn, Ateca-amestoy and UgidosAhn et al. (2014) also found that an increase in the woman’s income share results in an increase in her own financial satisfaction but a lower level of financial satisfaction for her partner. Reference Ahn, Ateca-amestoy and UgidosAhn et al. (2014) additionally found that for a Spanish sample, an increase in the woman’s income share contribution lowers her own financial satisfaction as well as her partner’s financial satisfaction.

Reference MysíkováMysíková (2016) found that in the Czech Republic, for couples without children, an increase in the woman’s income share increases her own financial satisfaction but lowers the man’s financial satisfaction; for couples with children, income share has no significant effects on either the man’s or the woman’s financial satisfaction. In a study using German data, Reference ElsasElsas (2016) ran regressions which account for various working arrangements between couples. Within the samples where both partners work full-time, on average, an increase in the woman’s income share lowers her partner’s financial satisfaction but has no effect on her own financial satisfaction. In the sample of male breadwinner couples (where the man works full-time and the woman works part-time or does not work), on average, an increase in the woman’s income share has insignificant effects on her partner’s financial satisfaction, and only positive effects on her own financial satisfaction if she works part-time.

In sum, each of the reviewed studies have found income share to be an important distribution factor. However, none of these studies have systematically confronted the intra-household effects of employment on financial satisfaction. This is a potentially important gap because related studies have found that work roles are important in defining individuals’ bargaining position within their household (see, for example, Reference AgarwalAgarwal, 1997; Reference Friedemann-SánchezFriedemann-Sánchez, 2008; Reference Vogler and PahlVogler and Pahl, 1993; Reference NoonanNoonan, 2001). Independent of income, employment status might matter for one’s bargaining position if notions of effort (and the associated claims over a share of household resources) are tied to paid work hours. If part-time work is not associated with the long-term financial security of the household, it might also fail to generate a claim over household resources, regardless of current income flows. Furthermore, if, for example, a woman is not earning any income, and is economically inactive so that she can provide more domestic and caring contributions, her bargaining position may be different compared to if she was not earning an income because of unemployment. Analysis of the links between employment patterns and the intra-household distribution of financial satisfaction can cast light on these possibilities and is, thus, an important alternate approach to the study of intra-household resource allocation. Reference De Henau and HimmelweitDe Henau and Himmelweit’s (2013) was the first – and, to date, most important investigation adopting this approach.

Findings from Reference De Henau and HimmelweitDe Henau and Himmelweit (2013)

Reference De Henau and HimmelweitDe Henau and Himmelweit’s (2013) study used BHPS panel data (from 1996 to 2007) to examine how changes in the employment status of men and women living in couple households affected their own and their partner’s satisfaction with household income. They found that both men’s and women’s satisfaction with household income was generally highest when they were engaged in full-time work, indicating that paid work contributions were generally perceived to have more value than unpaid contributions. The authors identified this as a gendered effect since most often women, take on the majority of the domestic roles and are therefore more likely to be the ones in less than full-time employment. Furthermore, on average, men’s own contributions from employment were more influential in determining their satisfaction with household income than their partner’s contributions from employment. However, for women, on average, their level of satisfaction with household income was influenced more by their partner’s employment than by their own. Reference De Henau and HimmelweitDe Henau and Himmelweit (2013) thus concluded that both gender asymmetry in contributions, as well as gender norms affecting the valuation of his versus her contributions, affected the way in which gender inequalities are reproduced within the household.

De Henau and Himmelweit’s study and its findings are important to the current project for a number of reasons. It developed an empirical methodology to examine key ideas in Sen’s model of cooperative conflict. While other studies had been limited in their focus, tending to focus solely on how income shares affect the intra-household distribution of financial satisfaction, De Henau and Himmelweit address the gendered perceptions of partners’ contributions to household resources, proxying this by a different distribution factor – employment status. Their results, as noted, suggest that such factors are important in determining the intra-household distribution of satisfaction with household income. However, their findings have yet to be assessed in other country contexts and time periods. The current study attempts to address this research gap by conducting a similar analysis of the links between employment and the intra-household distribution of financial satisfaction using Australian data from a longer and more recent time period. The results of this study contribute important evidence on the broader (and current) relevance of the findings presented by De Henau and Himmelweit. In doing so, the study will help inform the direction of future theoretical and empirical work on the allocation of resources and the implications of this for the distribution of financial satisfaction within households.

Data and measures

This study utilises data from the HILDA survey. Household, Income and Labour Dynamics in Australia is nationally representative and provides extensive information on the lives of Australians on topics such as employment, health and wellbeing, income, life events, childcare, values and attitudes and relationships (Reference Summerfield, Bright and HahnSummerfield et al., 2019). It is the only such survey of its kind available in Australia and is comparable to the BHPS used by Reference De Henau and HimmelweitDe Henau and Himmelweit (2013). This study uses 18 waves of HILDA data, from interviews conducted each year from 2001 to 2018. An important feature is that data is collected at a household level, and therefore contains the information of both partners for couple households. This feature makes it a survey well-suited for studying intra-household elements.

Using all 18 waves of data, the study sample includes cohabitating men and women in couple relationships. Couples are either married or in a de-facto relationship, with or without children. Similar to Reference De Henau and HimmelweitDe Henau and Himmelweit (2013), the sample is limited to working-age men and women (18–65 years). After imposing these restrictions, an unbalanced panel of 57,092 couple observations from 8478 distinct couples is obtained.

The dependent variable in this study is a measure of each partner’s satisfaction with financial situation (SWFS) score. The study utilises responses to a HILDA survey question which, each year, asks respondents: ‘How satisfied are you with your financial situation?’ The responses to these questions are organised on a scale from 0 to 10, with a score of 0 indicating completely dissatisfied, and a score of 10 indicating complete satisfaction. As seen in Table 1, men and women, within the sample, reported, on average, similar SWFS scores, at 6.66 and 6.64 points, respectively.

Table 1. Descriptive statistics: HILDA, mixed-sex couple households, 2001–2018.

Note: This sample consists of men and women in mixed-sex couples, between the ages of 18–65 years old.

SWFS: satisfaction with financial situation; HILDA: household, income and labour dynamics in Australia.

The measure used in this study differs slightly from the one used in Reference De Henau and HimmelweitDe Henau and Himmelweit’s (2013) model, ‘satisfaction with household income’, because an equivalent measure is not available in HILDA. This is a potential limitation of the current study because when respondents answer questions on their satisfaction with household income, they are likely to be reflecting on their household’s (rather than just their own) financial resources. With the HILDA question, the respondents might be reflecting only on their own resources. To assess the importance of this measurement issue, in an initial step, the correlation between the partners’ responses to questions on financial satisfaction were compared to those on more individual elements (their pay). The results show the correlation in SWFS of partners within household tends to be relatively high, at 0.49. In contrast, when comparing measures of satisfaction with truly individual measures of financial standing – pay – the correlation between partners is only 0.12. This provides some confidence that the key HILDA question on SWFS does, in fact, capture the respondents’ evaluation of their benefits from household, and not only individual, financial resources.

The main independent variables are similar to those used by Reference De Henau and HimmelweitDe Henau and Himmelweit (2013), in that they capture each partner’s employment status. Employment status is divided into four categories, including employed full-time, employed part-time, unemployed, and economically inactive. The last three categories may suggest a renegotiation of the level of non-financial contribution to the household by each partner. For example, partners who are employed full-time may be seen as largely making monetary contributions while those who are in part-time employment or are not in the labour market are likely to be seen as contributing more towards care and domestic duties.

Referring to Table 1, which contains descriptive statistics for the sample person-year observations, the level of full-time work is more than twice as high for men as it is for women. A total of 84% of the employment observations for men are for full-time work, as compared to only 39% for women. Women in part-time work made up 36% of the observations, while only 9% of the observations for men are in part-time work. The sample distributions in Table 2 show that 34% of the observations capture dual earner couples. However, couples where the man is the primary breadwinner still predominate, with 19% of observations representing couples where only the man works and 31% of observations representing couples where the man is the primary earner and the woman works part-time.

Table 2. Distribution of household types: Household, income and labour dynamics in Australia, mixed-sex couple households, 2001–2018.

Note: N (person-year observations) = 57,092. This sample consists of men and women in mixed-sex couples, between the ages of 18–65 years old.

Other variables comprising of equivalised household monthly income (in 2018 Australian dollars), number of dependent children, age, health and education are included in this study to account for factors that may have independent effects on SWFS. ^{Footnote 2} The presence of dependent children may result in additional demands on household finances and may intensify traditional labour division within households. Age can be related to changing financial pressures as couples negotiate the life course. Education levels may signal career aspirations and different earning expectations too. A person’s level of education might also impact the importance attached to equity within the household, and this may affect their level of satisfaction with the distribution of financial resources (Reference BonkeBonke, 2008).

Empirical strategy

The focus of this study is on how changes in either partner’s contributions towards household resources, as measured through changes in employment status, shift both partner’s SWFS. An appropriate method for such an investigation is a fixed effects regression analysis. Following Reference De Henau and HimmelweitDe Henau and Himmelweit (2013), we use a collective approach, with each partner’s SWFS modelled as a linear function of independent variables relating to their own and their partner’s employment status

(1) $s_{jt}^m={\beta }_{1m}{E}_{\mathrm{jt}}^m+{\beta }_{1w}{E}_{\mathrm{jt}}^w+{\gamma }_{1\text{m}}{C}_{\mathrm{jt}}^O+{\gamma }_{2\text{m}}{C}_{\mathrm{jt}}^P+t_1+{\mu }_{\mathrm{jt}}^m+{\epsilon }_{1jt}$

(2) $s_{jt}^w={\beta }_{2m}{E}_{\mathrm{jt}}^m+{\beta }_{2w}{E}_{\mathrm{jt}}^w+{\gamma }_{1\text{w}}{C}_{\mathrm{jt}}^O+{\gamma }_{2\text{w}}{C}_{\mathrm{jt}}^P+t_2+{\mu }_{\mathrm{jt}}^w+{\epsilon }_{2jt}$

The variables $s_{jt}^m$ and $s_{jt}^w$ denote the SWFS of the man and woman in the jth household at time t, respectively. The vectors $E_{\mathrm{jt}}^m$ and $E_{\mathrm{jt}}^w$ capture the employment status of the man and woman in the jth household with full-time employment as the reference category. The vector $C_{\mathrm{jt}}^O$ consists of a set of controls which capture the individual’s own characteristics that may have independent impacts on levels of SWFS, including, age (with a squared term to capture possible non-linear effects), health and years in education. The vector also includes other relevant controls for number and ages of children, and equivalised monthly household income. Characteristics of the individual’s partner are captured by $C_{\mathrm{jt}}^P$ , and these include their age (with a squared term to capture possible non-linear effects), health and years in education. $t_1$ and $t_2$ consist of year fixed effects, while ${\mu }_{\mathrm{jt}}^m$ and ${\mu }_{\mathrm{jt}}^w$ denote individual fixed effects that control for time invariant characteristics of the man and woman, respectively. Finally, ${\epsilon }_{1jt}$ and ${\epsilon }_{2jt}$ are randomly distributed error terms with a mean of zero.

As an extension, an alternate specification which analyses a breadwinner typology of households is also included (Table 4). This alternate model allows more specificity in accounting for gender role arrangements amongst couples. The models for both men and women are identical to (1) and (2) except that employment status is replaced with different employment combinations with a sole male breadwinner household (where the man works full-time and the woman is not employed) being the reference category.

An important feature of the estimation strategy is the use of fixed effects. By exploiting the panel nature of the data, within-individual comparisons are made such that the same individual is analysed at different points in time, thereby eliminating time invariant unobserved heterogeneity. This identification is therefore more comprehensive in testing the intra-household effects of contributions on each partner’s SWFS compared to many previous studies which make generalised between-individual comparisons (such as Reference BonkeBonke, 2008; Reference Bonke and BrowningBonke and Browning, 2009; Reference MysíkováMysíková, 2016).

The choice of a linear fixed effects approach for studies using subjective satisfaction measures has been debated, given the categorical nature of such variables (see Reference KristoffersenKristoffersen, 2010, for a summary). Nonetheless, a number of studies have compared results from linear and non-linear model estimations (see for example, Reference Blanchflower and OswaldBlanchflower and Oswald, 2004, Reference Blanchflower and Oswald2005; Reference Ferrer-i-Carbonell and FrijtersFerrer-i-Carbonell and Frijters, 2004; Reference Gardner and OswaldGardner and Oswald, 2001; Reference Headey and WoodenHeadey and Wooden, 2004). These studies have concluded that both approaches produce similar results in terms of effect signs and significance. As such, given the ease of interpreting ordinary least squares (OLS) estimates, equations (1) and (2) are estimated by employing linear fixed effects models. Nonetheless, as a robustness check, estimates on the baseline model using a non-linear estimation strategy are also included in a Supplementary file.

Interpreting the research questions

Estimates for the coefficients in equations (1) and (2) provide insight into the key issues of interest in this study. In relation to RQ1, ‘For mixed-sex couples, how does each partner’s contribution towards household resources impact their own and their partner’s level of financial satisfaction?’ if the coefficients $E_{\mathrm{jt}}^m$ and $E_{\mathrm{jt}}^w$ on less than full-time employment statuses are negative, this is interpreted as contributions of the partner in less than full-time employment are perceived as less likely to sustain one’s financial position in comparison to contributions through full-time employment. If this is the case, then an immediate gender inequality will be acknowledged as cross-sectional patterns in Tables 1 and2 indicate that within the sample, women are more often engaged in less than full-time employment compared to men.

Comparing the differences in coefficient magnitudes between $E_{\mathrm{jt}}^m$ and provide insight into RQ2, ‘For mixed-sex couples, does the influence of a type of contribution on financial satisfaction, in either or both partner’s assessment, depend on the gender of the contributor?’ Differences might indicate that the value attached to similar contributions to household financial resources depends on the gender of the partner making the contributions. Within households it is possible, for example, that the man’s employment is perceived to be more valuable that the woman’s. The inclusion of controls for household income, strengthen this analysis because they allow us to assess whether, when household income stays constant, each partner’s financial satisfaction is still affected by a changing distribution of paid and unpaid contribution. Furthermore, if an individual’s SWFS changes in a way that is dependent on the gender of the partner making the contributions, then there is little evidence of income pooling.