Yours, mine, and ours: gender, intra-household dynamics, and financing solar home systems in Tanzania

The financing of off-grid solar is a crucial component for the expansion of electricity access, particularly across sub-Saharan Africa. Emerging literature in energy access research analyzes the role of gender and inequalities in access and subsequent outcomes; however, beyond gendered outcomes, the field has yet to interrogate the intra-household dynamics of obtaining access. The literatures of the intra-household dynamics of budgeting and energy access have remained distinct. Bridging these literatures, we present the first analysis of if and how intra-household dynamics relate to how individuals obtain energy access through our specific case of financing solar home systems (SHSs). Drawing on 113 interviews across four districts in Tanzania, we employ qualitative thematic analysis and quantitative generalized linear model-based prevalence ratio estimates. We find five categories of how SHS adopters describe gender and intra-household dynamics relating to the financing of their SHS. We find it inappropriate to treat the household as a single entity as in half our sample, women are either solely or jointly responsible for financing the SHS. Female headed households have a higher prevalence of relying on debt to finance their SHS, while households with female financial decision-making power have a higher prevalence of relying on savings. The unitary model of a household limits the sector’s ability to efficiently support multi-adult households as they navigate individual needs and preferences in the financing and ultimate acquisition of the SHS. Researchers, policy makers, and the private sector could further acknowledge and explore intra-household dynamics and consider shifting the focus away from debt-based financing towards energy-dedicated savings accounts for female customers and improving borrowing conditions for female headed households.


Introduction
Approximately, 568 million people in sub-Saharan Africa (SSA) lack access to electricity [1]. Offgrid solar has and will continue to contribute to Sustainable Development Goal 7's (SDG 7) call for universal energy access [1], providing an interim level of electricity access [2,3]. Although less expensive, and touted as more affordable, than centralized grid connection, solar home systems (SHSs) remain out of reach for low-income individuals [4][5][6]. Multiple financing 3 strategies, such as rent-to-own models and energy-as-a-service, have expanded but still represent a minority of the market [7]. Little is known about the specific financing instruments that the majority of SHS adopters use to obtain their systems [8,9], and the intra-household dynamics surrounding that purchase.
The majority of research on energy assumes the household as the unit of analysis, in which a household finances, purchases, or affords the modern energy technology (e.g. [5,6,10]). Households, however, are composed of multiple individuals with different spending priorities [11,12]. Understanding how individuals within households allocate budgets and finance SHS could inform policy to expand financing of electricity access. Thus, we ask: (1) How do SHS adopters discuss gender and/or intra-household dynamics when describing the financing mechanisms used to obtain their SHS? (2) What is the relationship between how SHS adopters discuss gender and/or intra-household dynamics and (a) the financial mechanisms they use to obtain an SHS and (b) where they would consider borrowing from to obtain an SHS?

Literature on intra-household dynamics
Neoclassical economics assumes the household is a utopian commune, maximizing joint welfare [20]. The development of bargaining models and questioning of assumptions regarding perfect information and optimization, led to a broader collective model of the household [21].
Within collective models, there are two main categories of intra-household behavior: cooperative and non-cooperative. Cooperation implies that pooled resources achieve optimal outcomes, while non-cooperation indicates inefficient outcomes due to asymmetrical information, limited commitment, and/or lack of formal enforcement. In a cooperative model, household resource allocation would have no effect on each members' consumption. Both observational and experimental research has found this to not be the case.
Observational studies rely on (1) ethnography that intimately follows individuals' income and expenditure flows or (2) individual budget surveys that track labor supply, earnings, and nonlabor income. The experimental studies conducted games in which separated spouses decide what to do with their day's earnings under different levels of transparency and distribution scenarios (e.g. [22][23][24]). These methods investigate the bargaining behind and outcomes of intra-household dynamics.
Guyer conducted ethnography in Southern Cameroon finding that there was no single household expenditure plan, but rather women control their own money and use it mostly to supplement household food supply. Analysis of budget data in Brazil and India found that income under women's control increased investment and positive outcomes related to household services, health, and education [15,[25][26][27].
Field experiments in low-income countries, have found that individuals within households attempt to maximize individual welfare, resulting ultimately in losses for the household. In Uganda, Iversen et al found that spouses frequently do not behave in a cooperative manner, and overall household outcomes are larger when the women were in charge of the common pool [23]. A field experiment in India found that 25% of individuals opted to hide a portion of their money from their spouse, even when it resulted in a loss of 24% of average maximum potential earnings [22]. In rural Kenya, researchers found that spousal differences in saving preferences resulted in inefficient savings behavior [24].
An evolution of the discourse on 'women in development' followed. As the evidence emerged that allocating resources to women led to better 'development' outcomes for households (e.g. [15,27]), the depictions of women as victims and then 'Rotten Wives' shifted to 'Good Mothers' [21]. This led to pushes to focus development programming on women [28][29][30]. Others argued that inequality could be optimal decision making [31], that these interventions may potentially disrupt advantageous bargaining arrangements [30], or that women will not always behave in a societally favorable manner [21]. These critiques require policy to be thoughtfully tailored to intra-household dynamics, but not ignored.

Literature on gender and energy access
The field of energy access has evaluated gender differentiation in the burdens from energy poverty and the distribution and benefits of access. Women's disproportionate burden of energy poverty in terms of health, time, and physical burden is well established [18,32,33]. The sector thus focused energy access provision on women. However, research has interrogated the assumption that energy access would automatically benefit women. Research has found limited evidence that access meaningfully or equally benefits women [19,32,33].
A few studies investigated the intersection of gender and energy access. A study in India conducted 31 household diaries and ethnographic observation on appliance use stratified by gender, paired with a six-state household survey. They found that women are neither the sole nor primary beneficiary of electricity access [34]. Specifically, regarding SHS, a study in Kenya relied on semi-structured interviews to compare the benefits across genders under different provision systems (e.g. national grid vs. SHS). They found that husbands decided the location of light bulbs [35]. Other studies in Kenya, Uganda and Ghana did not have an explicit investigations into gender, but rather noted the absence of appliances that would benefit women (e.g. lightbulbs in the cooking area or electric cookers [36,37]). Other studies on pay-as-you-go financing plans for solar systems found that women were five percentage points more likely to work outside the home, had more autonomy over their time, and were able to start new businesses [38].
The studies that evaluate gender, and nod to intra-household dynamics, largely assess the outcomes of intra-household dynamics for women, after some level of access is achieved. While this is critical work, additional investigation is required to reveal bargaining power or patterns of decision making between household members regarding the provision of and outcomes from energy access. Put simply, the sector would benefit from understanding how modern energy is obtained and who is involved in its financing.
Our study fills this critical gap in the literature, positioned at the nexus of the intra-household dynamics of budgeting and energy access. Our work adds to the literature in three important ways. It is the first to interrogate the role of gender and intra-household dynamics in the financing SHSs. Our work builds on the budding literature on intrahousehold dynamics in energy access at large, moving beyond gendered outcomes. Finally, we provide concrete policy recommendations to facilitate uptake of modern energy systems and directions for future research.
These questions are particularly salient in our study site. In Tanzania, 77% lack access to electricity and 26.5% use off-grid solar as their main lighting source [39]. The private and public sector has outlined gender based solar energy policies (e.g. Tanzania's Rural Electrification Agency's 'Gender Strategy and Action Plan, Solar Sister, an NGO focused on women distributing SHS, and the Tanzania Gender and Sustainability Energy Network [40,41]). Beyond Tanzania, numerous solar energy initiatives have been expanding and either focus or ignore gender [3,42,43].
The remainder of this paper is structured as follows: section 3 research design, setting, and methods provides and overview of our study's approach and methodology; section 4 results and discussion describes the substance and importance of our findings; Finally, section 5 conclusions and policy recommendations provides recommendations based off our findings.
The interviews probed the financing mechanisms used to purchase respondent's system and the SHS adopters' conceptions of debt. Four Tanzanian research assistants, trained in ethnographic methods, conducted the interviews between February and June of 2022 in the respective local languages.
In two accompanying papers [8,9], we found that adopters used 12 financial tools to obtain their SHS, the most common of which were savings, intermittent income from sale of farm produce or small business, consistent salary, rent-to-own agreements, and a one-time SHS purchase. Respondents demonstrated a large aversion to debt-based financing and opted to borrow from individuals with whom they have strong ties.
We included questions that investigated household dynamics regarding the purchase and use of the SHS. We addressed the first question on how SHS adopters discuss gender and household dynamics surrounding the SHS purchase through thematic analysis. This qualitative analysis formed the basis of our generalized linear model (GLM) to investigate a relationship between how respondents discussed gender, the financial mechanisms used, and from whom the respondent would consider borrowing from to obtain the SHS (Q2a and Q2b). We triangulated our findings to address whether, how, and why gender and financial mechanisms interact when purchasing the SHS.
The research assistants selected participants through the snowball method, but subsequently chose individuals to reflect different socio-economic statuses, ethnicities, and religions in each area, based on local knowledge and observations. They gauged socio-economic status by noting building materials, compound sizes, and visible assets.

Thematic analysis
Following a grounded theory approach [44,45], the first author open coded the transcripts into meaningful categories, concepts and themes. She subsequently grouped codes into families, recording their frequency. Interpretation required returning to the data and reorganizing it by code or code family. The first author hand coded, creating physical mapping of codes to easily splice and link different categories [46]. With the research assistants, the authors looked for patterns, throughlines, and themes, noted irregularities [47], and ultimately generalized [48].

i. Independent variables
Our thematic analysis informed the construction of nine independent variables for our quantitative analysis. We translated the five major intrahousehold dynamic categories found into binary independent variables. We did the same for de jure [49,50] and self-reported female-headed households [50] regarding the SHS purchase and use. We coded two binary independent variables for respondents reporting joint decisions on the SHS purchase and uses (table 1).

ii. Dependent variables
Our dependent variables were six binary variables indicating whether a respondent used a specific financial tool to obtain their SHS and from where a respondent would consider borrowing (detailed in [9]). This resulted in ten dependent variables, six indicating the use of savings, intermittent income from sale of farm produce or small business, consistent salary, rent-toown agreements, a one-time purchase, any debt financing, and four indicating the consideration of relatives, friends, neighbors, or a formal bank to borrow (table 1).

iii. Control variables
We used a set of variables, derived from the literature, that are likely to impact the financial tools used by households, which include age, tribe, religion, occupation, education, setting, and weekly household expenditure (table 1). We also included district-level fixed effects to account for variation in unspecified district-level factors that could affect financial decisions. Allen et al 2016 found globally that older, wealthier, more educated, urban, employed, and married or separated individuals had a higher probability of having and saving in a formal financial account [51]. They also found that older, more educated, richer, married men were more likely to formally borrow. Fungácová and Weill 2015's results supported these results, finding additionally that income and education affect the choice between formal and informal financial tools [52]. Demirgüc-Kunt et al 2013 found that religion can affect financial inclusion [53]. They found that Muslims are less likely to own or save in a formal account, particularly in SSA. Zins and Weill's also found the same associations for informal banking [54]. Finally, recent evidence from India suggested that tribe also affects financial inclusion [55]. These potential confounders were the exact reason that we pursued multivariable analysis, which we will describe next. Table 1 outlines the descriptive statistics (percentage or mean) of all variables. We used a GLM with the Poisson family and log-link to estimate prevalence ratios (PRs). We pursued this approach given our cross-sectional data, particularly with outcomes with low prevalence [56,57]. We used the following model to characterize the distribution of our outcomes and estimate PRs (equations (1) and (2))

Statistical analysis: prevalence ratios
where Y i is the binary dependent variable of interest (e.g. using savings to finance a SHS) of an individual SHS adopter, x i is the specific independent variable (e.g. describing a unified household model), c ji represents all the control variables (e.g. expenditure), and PR is the prevalence ratio. We calculated robust standard errors, clustered by district. We note that our methods allow us to investigate the correlative, not causal, relationship between our variables. We only report results statistically significant at the 5% level in the main text; all analyses, conducted in Stata 16.1, are in supplemental materials. Figures were designed in Python and R.

Sample demographics
Our respondents were balanced across gender and location. The average age and household size were 40.9 years [min:19, max:79] and 5.0 [min:1, max:12] respectively. Most respondents were married, Christian, and completed primary school. Our sample was 14% de jure female-headed households (i.e. single, widowed, or divorced women). The respondents varied widely regarding tribe and occupation, often participating in multiple income generating activities. The majority used intermittent income to obtain their SHS, followed by debtfinancing (44%), savings (35%), and consistent salary (15%). Forty percent used a rent-to-own financing scheme, while 50% purchased the entire system at one time. Across our respondents, 68%, 50%, 34%, and 14% considered borrowing from a friend, a relative, a neighbor, and a bank respectively (table 1).

Describing household income streams to purchase SHS
Our research revealed that respondents 4 described income flows used to purchase their SHS across five distinct categories. Figure 2 depicts the proportion of respondents that flowed from our entire sample into each category. The majority (51%) of respondents only mentioned a male family member's income, typically the husband or father. We will refer to these respondents as 'male-financed' households. A male respondent in Arusha explained that he 'plowed onions and after harvesting onions, [he] sold and got money to purchase [his] SHS' (respondent CD9). A female respondent from Arusha explained that purchasing the SHS was her husband's 'responsibility because he is the head of [the] family' (CD5). Another female respondent explained that the SHS purchase was not a burden on their household because, 'of the good budget, my husband prepared for it effectively' (DL11).
Fourteen percent of respondents mentioned only a female family member's income (i.e. Figure 2. Sankey diagrams to describe how we excluded de jure female headed households and then investigated how respondents described purchasing their SHS, the decision to purchase, and the decision regarding how to use the SHS once obtained. The width of each link indicates the proportion of the initial specific category to flow into the subsequent category based on our semi-structured interviews. For example, 86% of the entire sample constitute the remaining sample (after taking out de jure female-headed households). From that remaining sample, 51% reported only a male household members' income.
'female-financing households'). Thirteen percent of respondents mentioned one individual's contribution, while noting the non-monetary contribution of a partner of a different gender 5 5 We acknowledge that gender is a spectrum; however, in the context of the districts in Tanzania we study, gender is a binary, (i.e. 'consultation' households (table 1 and figure 2). A male respondent from Tabora stated, 'my wife advised me. She did not want to involve this purchase with the debt from anywhere' (DL22), and only heterosexual relationships between men and women are legal and openly discussed. PRs and 95% confidence intervals from our investigation into how gender and intra-household dynamics correlate with the financing tools SHS adopters use to obtain SHS. The PRs for a respondent reporting that a female decided the SHS purchase and only mentioned a female's financial contribution were much larger (>10) than the scale of the other PRs. We only report results that were statistically significant at the 5% level. All results reported in supplemental materials.
while another 20 year-old male explained that 'my father listened to the advice from mother and we children' (DL10).
A similar percentage of respondents (14%) mentioned a male and female financial contribution, describing these income streams separately (i.e. 'jointly-financed' households) (table 1 and figure 2). A female respondent in Arusha stated, 'I have a farm of onions and sometimes my husband is also adding some money' (CD21). A male respondent distinguished income streams, explaining, 'when I do not have money my wife helps me pay' (CD18). These separate income streams were even delineated over time for different purchases. A respondent explained that he covered the upfront cost for the rent-to-own SHS, but his wife was responsible for the daily recurring cost.
Only eight percent described a unified household income, (i.e. 'unified households') describing the money came 'from the family' (SO10) (table 1 and figure 2). One male respondent only used the term 'we' when describing the SHS purchase, stating, 'we normally use our own money' (SO12).

Decisions on SHS purchase and uses
Twenty-three percent and seventeen percent of respondents claimed that a female family member decided to purchase and uses of their SHS, respectively. Twenty percent and thirteen percent of respondents claimed that the purchase decision and use respectively were decided jointly between typically the husband and wife (table 1 and figure 2).
Male respondents were most often responsible for SHS purchase and use decisions. One respondent explained, 'my family listens much to the regulation from my husband on how to use the solar appropriately' (DL28). Our results suggest that women are rarely solely involved in financing the system (14%) or the decision to purchase (23%) and use (17%) the SHS of the 'household.' Respondents claiming solely female or joint decision making were roughly equal to the percentage of respondents reporting sole male decision makers for the decision to purchase (57%) and finance (51%) the SHS.

Intra-household dynamics and financial tools to obtain an SHS
We next explore the PRs between the distinct categories of household dynamics to obtain the SHS and the financial mechanisms pursued.

i. Savings
Respondents from 'jointly-financed' households had a lower prevalence 6 of using savings than those respondents indicating that the purchase was not a joint decision. Contrastingly, respondents claiming a woman decided the SHS purchase and uses had a higher prevalence of relying on savings ( figure 3). This suggests that women with decision making power are more likely to rely on savings. Most financial tools meant to increase modern energy access have focused on debt-based financing. Savings accounts have been found to benefit women [29], and have been suggested to increase modern cooking energy access [58,59]. ii. Intermittent income Respondents from 'consultation' households had a lower prevalence of using intermittent income to finance their SHS ( figure 3). Contrastingly, respondents from 'femalefinanced' households had a higher prevalence of relying on intermittent income ( figure 3).
Households with joint financial communication may be better positioned to plan and not need to rely on intermittent sources. However, female income is often much lower and less consistent, thus increasing the importance of intermittent income. iii.

Consistent salary
Respondents from 'female-financed' households and those with female decision-making power over the SHS purchase and uses all had a much higher prevalence of relying on a consistent salary ( figure 3). Contrastingly, respondents from 'consultation' households had a much lower prevalence of relying consistent salary to finance the SHS. This was also true for respondents from 'unified' and 'jointly-financed' households. Finally, respondents claiming that the SHS purchase, and uses were a joint decision had a lower prevalence of relying on consistent salary ( figure 3). Overall, respondents having a sole female decision maker had a higher prevalence of using consistent salary, while those reporting joint incomes, decisions, or consultations had a lower prevalence. Interesting, we find a higher prevalence of both reliance on intermittent income and consistent salary for respondents from 'female financed' households. SHSs may be out of reach for female-headed households without consistent incomes. Women experience lower pay and fewer opportunities in the formal employment sector [60,61]. Women seem to prefer consistent income to obtain an SHS if available. iv. All at once vs. debt-based financing Respondents claiming that a woman decided the SHS uses had a higher prevalence of purchasing the entire SHS at one time ( figure 3). Female-headed households and those from 'jointly-financed' households had a higher prevalence of using debt-financing (figure 3). We contextualized this result with our findings that respondents reporting female decisionmaking power had a higher prevalence for using savings, that there is a universal strong aversion to debt [8], and that female-headed households face disproportionate barriers [62]. Further, previous literature found that debt schemes burden female and low-income SHS adopters [6,19,63]. v.

Rent-to-own
Respondents from 'jointly-financed' households had a higher prevalence of SHS rent-to-own schemes. However, respondents stating that a woman decided the SHS uses had a lower prevalence of SHS rent-to-own schemes ( figure 3).
Most rent-to-own models were initially structured to accommodate low, intermittent income. However, research reveals that the off-grid market has been systematically exploiting, and ultimately excluding rural, low-income customers [63,64]. Households with two contributing individuals may be more comfortable signing strict rent-to-own-agreement.

Intra-household dynamics and borrowing to obtain an SHS
Next, we investigated the association of gender and intra-household dynamics and from whom respondents were comfortable borrowing. Respondents with a woman deciding the SHS purchase had a lower prevalence of considering borrowing from a bank ( figure 4). Women have less information about and access to formal financing 7 [65], and mobile money preferentially assists low-income women [66]. Respondents claiming that a female decided SHS uses had a higher prevalence of considering borrowing from neighbors (figure 4). Respondents stating that the SHS uses are a joint decision and those from 'consultation' households had a slightly higher prevalence of considering borrowing from friends (figure 4).
Respondents stating that the SHS uses are jointly decided or those from 'jointly-financed' households had a lower prevalence of considering borrowing from relatives. Contrastingly, respondents from 'consultation' households had a higher prevalence of considering borrowing from relatives (figure 4). Our findings imply that households with female-decision making power, although not financial, are more comfortable translating that intra-household power into their social networks to obtain financial assets. Respondents with spousal consultation may be more comfortable approaching other family members on financial issues.

Conclusions and policy recommendations
Our study provides the first investigation into the role of gender and intra-household dynamics in financing modern energy access. Bridging the fields of intrahousehold dynamics of budgeting and energy access, our work investigates the financial relationships and tools used to finance an SHS. Confirming a collective model of the household, we find five distinct categories of intra-household dynamics. We find that (1) female decision makers have a higher prevalence for using savings, (2) respondents within households with female decision making power had higher prevalence for using both intermittent income and consistent salary, (3) female-headed households have a higher prevalence of relying on debt-based financing, and (4) women with SHS decision power have a lower prevalence for considering borrowing from a bank, turning to their social networks. We provide recommendations for policy and future research to facilitate just SHS adoption.
Modern energy access stakeholders may look beyond the unitary household model to support adoption. Energy surveys could delineate by gender [19], but also could ask about intra-household dynamics, to investigate outcomes and underpinning dynamics.
Tanzania's evolving energy policy focuses on female participation and gender integration into institutional capacity and productive uses. More specifically, our results suggest that the solar energy sector could shift the focus away from debt schemes. The Tanzanian government could partner with banks and solar energy companies to pilot female focused saving programs for SHS, with more favorable interest rates. This could increase financial and energy access in parallel.
Debt-based SHS financing schemes could consider intermittent income and consistent salaries, particularly for female customers. The platforms could even nudge users at a chosen frequency, for one-click mobile payments. Debt-based schemes could be conscious that female-headed households have a higher prevalence of relying on them, and thus could proactively offer better conditions, potentially subsidized by the government. Finally, modern energy awareness campaigns could encourage intra-household discussions.
Our work is based on a single cross-sectional data set, prohibiting any discussion of causality. We attempt to limit sampling, interviewer, and social desirability bias through thorough testing the survey 8 , working with local leaders to ensure coverage, explaining that there were no incorrect answers, and creating rapport with respondents.
Further work could investigate the impact of female focused savings interventions for energy access, or tailored debt-based financing programs for female-headed households. We only evaluate households that did obtain the SHS. Future studies could test whether intra-household dynamics relate to the ultimate ability or decision to obtain modern energy. Experimental work could track SHS adoption after randomly selecting spouses, without SHSs, to receive cash transfers (of equal amount to a SHS) and SHS informational sessions. Individual income data could be regressed against investment in SHS. Ethnography could further interrogate the differences within households that do and do not obtain SHSs. The unitary household model limits the solar energy sector's ability to efficiently to facilitate the financing of SHS, and further efforts toward SDG 7.

Data availability statements
The data cannot be made publicly available upon publication because they contain sensitive personal information. The data that support the findings of this study are available upon reasonable request from the authors.