Category: Blogs

“All financial responsibilities will rest on her shoulders, and she will not be able to save any money,” responded Grace, when asked “What will happen if a woman has control over family decisions?” This sentiment was a recurring theme in our WEE-DiFine-funded study, “Adapting and Validating WEE Indicators in an Experimental Study of Savings” conducted in Uganda.

The goal of our study was to identify specific measures of economic empowerment related to women’s savings behavior using two complementary modes of validation. In the first phase, content validation, we applied qualitative methods to derive a comprehensive list of WEE indicators relevant to the everyday savings practices of Ugandan women. In the second phase, construct validation, we substantiated these indicators through their statistical correspondence with women’s observed savings behaviors, including their ability to set meaningful goals, to mobilize savings toward those goals, and to use the saved money as intended. We found several factors related to WEE measurement that were extremely local and not widely discussed in existing tools.  They were, however, very prominent in our survey responses and repeatedly selected by feature selection algorithms. Our takeaway is that the local context is absolutely vital for rigorously measuring women’s economic empowerment.   

Content Validation Insights: Bringing Hidden WEE Indicators to the Forefront

The content validation phase identified over a hundred WEE indicators. Here we focus on those that appear most frequently in our survey data,and that are rarely addressed in existing WEE measurement frameworks. 

Balancing Power and Responsibility

Decision-making power is normally viewed as an indicator of empowerment, but it brings financial consequences that can undermine women’s ability to save. This is especially true in Uganda, where the communal nature of households continually introduces new responsibilities and roles for women. Many women, like Grace, discussed the burdens that come with decision-making power, recognizing that there is a point beyond which the associated responsibilities become too heavy. Decision-making power cannot be assessed as a WEE measure without an understanding of what it entails and who bears the financial costs.

Fear as a WEE Factor

Fear significantly impacts women’s sense of agency in deciding how much to save for the future.  These fears include divorce, illness, and, most prominently, a fear of aging. Many women expressed worries regarding whether relatives or children would still be available to care for them as they grew older. This uncertainty prompted them to adopt savings strategies aimed at securing their financial stability in later life, especially considering their lack of retirement funds.

Mental Accounting and the Role of Trusted Knowledge

Women can find it challenging to organize and manage their financial resources—a behavioral aspect strongly tied to mental accounting. Navigating an overwhelming and often unreliable influx of information complicates their decision-making, hindering their ability to stay focused on their financial goals. Training on savings practices, local business issues, budgeting, and related topics are highly valued in helping them achieve realistic financial objectives. Additionally, women emphasized the need for clear and straightforward information from financial providers to facilitate comparisons of products and services, enabling them to make better financial choices. 

Construct Validation: Identifying the Most Relevant WEE Indicators for Savings

Using our rich data set of over a hundred WEE variables and account-level data, we applied ML feature selection algorithms to identify the WEE indicators that resonate most with our participants throughout their savings journeys. 

The results (Table 1) clearly confirm that Grace’s observation about the burden of financial responsibilities and its relevance to savings behavior was not just a personal anecdote, but a vital issue for women in Uganda. Most new indicators discussed during the content validation phase were consistently selected across all ML feature selection algorithms and ranked among the top 5 of the 25 selected WEE constructs.

Financial responsibilities, ranked first, highlight the importance of incorporating these roles into WEE measurement within the context of financial inclusion. Attendance at training (ranked third) and ability to compare financial products (ranked fifth) emphasize the need for a comprehensive approach to assessing financial knowledge within WEE tools. These indicators also underscore the need for clear, accessible information and practical financial knowledge to help women navigate mental accounting challenges. The alignment between the fears expressed during the content validation phase and the high ranking of financial strategies for old age (ranked fourth) further reinforces the importance of including this issue in WEE measurement. As women plan for the future, their outlook—closely tied to their savings behavior (ranked second)—is shaped by concerns about financial security in later years.

Importance of Contextual Lens

Our dual validation method revealed themes underexplored in existing frameworks, emphasizing that WEE indicators must align with specific outcomes and local realities. This approach not only reflects the true experiences of women like Grace, but also honors the dignity of Ugandan women navigating their financial challenges with determination. Contextually grounded WEE measurement tools enhance our ability to understand and support their aspirations.

[1] Such as Measuring Women’s Economic Empowerment: A Compendium of Selected Tools or Measuring Women’s Economic Empowerment in Financial Inclusion.
[2] One key insight we uncovered during our qualitative interviews was the significant role of the extended family in shaping women’s agency. In Uganda’s dynamic household structure, focusing solely on the roles of husbands or immediate family risks overlooking the critical influence of the extended family and the diverse circumstances they bring to women’s lives.
[3] For example, a recently developed tool by industry experts includes decision-making power as a key WEE measure, but does not address the financial burden that accompanies such responsibilities.
[4] We employ LASSO Stability Selection, VSURF for Interpretation, VSURF for Prediction, and Boruta. By comparing the features selected by more than one ML approach we identified the features with the most stability and consistency across models.
[5] Table 1 displays the top 25 WEE indicators from over a hundred tested. Those not shown were either not selected or selected with minimal frequency. ‘Percent Times Selected’ indicates how often the construct was selected in all best performing models combined (LASSO Stability Selection, BORUTA, and VSURF Interpretation and Prediction). 

Agricultural incomes are notoriously seasonal. The bulk of outflows occur at the start of a growing season, and inflows occur at the end of the season. A large body of evidence documents the effects of seasonal liquidity on many aspects of farmer welfare, including nutrition, productive investments, agricultural input use, and labour market participation (Devereux et al. 2013). In areas where formal and informal financial markets are not well developed, farmers find it difficult to smooth consumption over time. As a result, farmers often experience periodic food insecurity in the months most distant from the previous harvest. Fluctuations in average body weight of up to 4 kilograms in between harvests have been long documented among women in Africa and Asia (Ferro-Luzzi and Branca 1993). This period, known as the lean or hunger season, is further intensified during periods of adverse or unpredictable weather and crop failure. In Kenya, where a majority of the rural poor depend on rainfed subsistence farming, 2.8 million report experiencing food insecurity in the lean season. This estimate is expected to increase further with climate change (WFP 2023). 

Crop insurance is a risk management tool that can protect vulnerable farmers from the negative consequences of harvest failure and income uncertainty. However, insurance typically provides compensation for damaged crops at the end of the agricultural season, after harvest has occurred. Farmers may find traditional compensation schedules ineffective for a number of reasons – payments are too far in the future for farmers who discount time heavily; later payments could be less useful for farmers who need to make expenses, such as buying inputs or smoothing consumption, in the lean months before harvest; and payment itself is uncertain, which is problematic for farmers who prefer to plan their expenses ahead. Through our study funded by the WEE-DiFine Initiative, we pilot an insurance feature, enabled by recent advancements in crop loss detection and financial technology, that provides farmers with customized compensation timelines. Using models that predict damage at each major growth stage of the crop, and availing mobile money to disburse payouts, this innovation allows farmers to receive timely mid-season payouts in tranches on chosen dates. We in turn study the effects of this innovation on insurance adoption.

We conducted a randomized controlled trial with 1,765 farmers in Kenya in 2023. We offered eligible farmers the opportunity to purchase a standard weather-index based crop insurance policy that disbursed compensation in one payment after harvest, or a modified policy in which farmers could receive compensation in stages at times of their choosing, referred to as “timely-pay insurance.” We measure compensation preferences and willingness-to-pay for the standard and timely-pay policies for all farmers. To express their compensation preferences, farmers were asked to allocate four payout stages, representing losses estimated at the end of four crop growth stages, across six months, as detailed in an earlier blog.[1] Farmers could thus customize when they wanted to receive a payment, and how large (or small) they wanted each payment to be. Additionally, we recorded farmers’ self-reported experience of food insecurity during the same six months. We then randomized the product offered to measure uptake of insurance. Our sample included both men and women, allowing us to observe differences in compensation preferences by gender. 

Our data supports past evidence that farmers experience seasonal food insecurity that increases over the length of time from the preceding harvest (Figure 1). The proportion of farmers reporting that their household lacked access to sufficient quantity or quality of food rose from 18.7% in October at the start of the long rains growing season, to 55.47% in March at the end of the season. Overall, 65% of respondents reported experiencing food insecurity at some point during the cropping season. Reporting did not differ systematically between men and women farmers in any month. The vast majority of farmers (86%) wanted to receive insurance compensation in multiple stages rather than through a single transfer. Nearly half (45%) chose to receive compensation in 4 separate transfers, the maximum allowed.

Men were significantly more likely to request smaller and earlier transfers, especially in the first 4 months pre-harvest (Figure 1). Our financial empowerment measures indicated that women were relatively less-empowered and had sole agency over smaller sums of money than men, which we had anticipated would drive them to prefer smaller transfers. However, women were more likely to request later and fewer transfers. Women were also significantly more likely to request for transfers in the last two months of the season (February and March), nearer to the typical harvest period, but also the time of greatest reported food insecurity. In our preliminary results, we find that being less-empowered and experiencing food insecurity increases the likelihood that a farmer will choose to receive fewer and larger transfers and receive payment in the lean months, with the effect stronger amongst women. We also find that willingness-to-pay for the timely-pay insurance is significantly higher than for traditional insurance on average, and particularly among the less-financially empowered. Additionally, the uptake gap between men and women decreases when farmers are offered this novel insurance feature. 

Our early findings indicate the potential for mid-season or staggered compensation to improve the impacts of crop insurance among those most vulnerable to climate change, such as women and financially less-empowered smallholders. We look forward to sharing our final results in the near future.

References

Devereux, S., Sabates-Wheeler, R., & Longhurst, R. (Eds.). (2013). Seasonality, Rural Livelihoods and Development. Routledge.

Ferro-Luzzi, A., and F. Branca. 1993. “Nutritional Seasonality: The Dimensions of the Problem.” In Seasonality and Human Ecology, edited by S. J. Ulijaszek and S. S. Strickland, 149–165. Cambridge, UK: Cambridge University Press

WFP (2023) Kenya Annual Country Report 2023. World Food Programme URL: https://www.wfp.org/operations/annual-country-report?operation_id=KE02&year=2023#/26567, accessed September 2024 

[1] These six months corresponded to the short rains agricultural season from October to March of 2023. Some  constraints applied to farmers’ compensation preferences.

The structure and dynamics of Ugandan households presented a unique context for our study. Unlike the traditional nuclear household structure, which typically consists of a couple and their children, Ugandan households are communal and dynamic, encompassing a broad and evolving network of relatives and community members. These households frequently change in size and structure due to various economic and cultural factors, and sometimes even without clear reasons. This fluidity has implications for the measurement of women’s economic empowerment (WEE), as observed while conducting our WEE-DiFine-funded study, “Adapting and Validating WEE Indicators in an Experimental Study of Savings.” The effects of these dynamics had a profound impact on our research, influencing everything from the wording of our questionnaire to the final WEE outcomes. We believe that our insights into the fluidity of household sizes and its impact on WEE measurement will be of particular value to other researchers navigating similar contexts.

Measuring Household Size     

As we embarked on our pilot phase of endline data collection, one of our regular quality check protocols involved comparing the baseline reported household size with the endline reported size. We presumed that minor changes in household size would occur, such as a member joining or leaving in some households, and only in rare cases, given that the endline survey was conducted just 12 months after the baseline survey. Larger changes would likely indicate potential data quality issues. 

The results of the initial checks performed on the first 613 surveys revealed an almost 60% mismatch[1] of household size between baseline and endline, and a discrepancy of more than 2 members amongst 20% of these households (Figure 1). This trend persisted throughout the survey, with approximately 50% mismatch on the full sample.

Our immediate response was to conduct a quick survey audit to verify the validity of the responses.[2] The audit revealed insightful reasons behind the differences, surfacing interesting implications on women’s daily lives, which influence nearly all aspects of their economic empowerment.

Many household size mismatches were due to variations in the number of children reported between the baseline and endline. In Uganda, children often move between relatives’ or neighbors’ homes for communal upbringing.[3] Additionally, economic hardships frequently lead to children being adopted or cared for by others, particularly in rural areas. Finally, boarding schools, which are common in Uganda, cause seasonal or temporary relocations of children. These changes directly impact women’s daily lives. As primary caregivers, they must constantly adapt to new situations, making it hard to plan ahead or commit to long-term opportunities.[4]     

Other mismatches came from frequent relocations of a broader array of household members. These relocations significantly influence household dynamics and the individuals who have a say in the household. Poor housing conditions or weather-related problems can lead to children or the elderly being sent to live with other relatives. Moves for specialized medical care, as well as inheritance and property disputes, are also common. Our women respondents mentioned moving, usually with children, to be closer to a strong support network during difficult times or significant life changes. Finally, in polygynous households, children and spouses frequently move between the households of a father’s wives. These relocations significantly influence household dynamics and the individuals who have a say in the household. Women are forced to take on new roles and responsibilities, significantly impacting their decision-making power, financial stability, economic autonomy, and overall empowerment.

Revealing the Neglected Complexity of Household Dynamics in WEE Research 

The ‘Who Else’ and ‘What Else’ Factors

The insights gained from our audits perfectly aligned with and confirmed our findings from the qualitative interviews we previously conducted to validate our empowerment constructs and design survey questions. During these sessions, our women participants emphasized that due to a constantly changing household structure, focusing solely on husbands’ roles in women’s agency and decision-making would overlook the vital contributions of other household members, extended family, neighbors, friends, the broader community, and the diverse circumstances they bring.

For instance, during our cognitive interviews testing women’s agency questions from academic publications on Uganda (see Table 1), participants noted that focusing solely on husbands’ roles in WEE is too narrow and misrepresents the complex realities women live in.

The women further elaborated that restrictions on their agency and decisions are imposed by many other family members that directly beget from the fluidity of household structure. 

Instead of limiting WEE questions to spouses’ role or impact, our participants suggested offering choices that include siblings, parents, children, in-laws, and even friends or neighbors. The women also emphasized the necessity of having multiple choice options, as they are often required to navigate restrictions on their agency and autonomy imposed by multiple individuals simultaneously!

And in fact, our endline survey data shows that it is crucial to capture not only the wide spectrum of ‘who’ permits or restricts, but also ‘who else’—the combination of people involved in influencing the agency of women. To illustrate, one WEE question suggested by our participants, which we included in our endline survey, asks about individuals who forcibly take and use their money (Table 2).

Had we only asked about spouses using money against women’s will, we would have seen just a 5-6% occurrence. This might have led us to believe that the issue is not widespread and has limited relevance to WEE measurement. However, when we allowed our respondents to consider all individuals who take money from them forcefully, the frequency more than doubled to 12%. This highlights a significant problem where various people, sometimes in combination, exploit these women, undermining their financial autonomy.

Additionally, our participants noted that the ever-changing household dynamics introduce additional factors into the equation. It’s not just about ‘who’ or ‘who else’ interferes with their control, but also ‘what’ and ‘what else’. Extra childcare, housework, and a range of constantly changing obligations lead to fatigue and health issues, further diminishing their agency and decision-making power.

For constructs regarding women’s ‘control over time,’ our participants urged us to include a comprehensive list of potential sources, both individuals and circumstances. They emphasized that the restrictions on their time control are imposed not only by different people, but also by the circumstances those people create.

The endline survey in fact reveals a complex and multifaceted landscape of empowerment restrictions that Ugandan women face. Fewer than 5% of women are restricted by only one person or family member – the vast majority experience time control restrictions from multiple people and factors simultaneously (Table 3).

Once again, if we had tailored the question solely to ask if spouses control how these women spend their time, we would have observed a low occurrence of the issue and mistakenly concluded that this construct has limited relevance to WEE measurement. However, by allowing women to list all the people and factors that limit their time agency, we uncovered the full extent and complexity of the challenges they face.

Measuring WEE is no easy feat—it’s a complex task that hinges on our precise understanding of the communities and people we study. To get it right, we must immerse ourselves in these communities, truly grasp their daily lives and unique challenges, and incorporate these insights into our study designs. Like the dynamics of household structure discussed above, each unique factor within the communities might be crucial to getting the full and correct picture. By moving beyond our assumptions and genuinely listening to the individuals we study, we can craft questions that resonate with them, revealing powerful insights that might otherwise remain hidden.

[1] Discrepancy of at least one member in the reported household size between the baseline and endline surveys.
[2] The data quality audits involved telephonic checks to verify household member counts and gathered nuanced insights into the discrepancies through in-depth interviews.
[3] Our observations indicate that the communal nature of Ugandan households is deeply rooted in cultural traditions that emphasize strong social bonds and collective responsibility. This support network also serves as a practical strategy to cope with economic challenges.
[4] For instance, one woman initially reported four children, but by the endline, she had eight, as her sister, who went to work abroad, left her four additional children, forcing her to work longer hours at night to provide for them and take care of the children during the day. Another case involved a woman who, after ending her marriage and having a child, returned to live with her many siblings, significantly altering her priorities and life direction, as her autonomy diminished, and decisions were now made by her relatives instead of her. In another instance, a grandmother’s daily life was impacted when her grandchildren were taken by their mother, who found permanent work in Kampala.

Digital connectivity has become crucial for women’s social and economic empowerment in Bangladesh, offering unprecedented access to information, opportunities, and engagement. However, recent political upheaval, marked by the government’s collapse in August 2024, presents both risks and opportunities for advancing digital inclusion. This piece explores the importance of digital connectivity for women, the challenges posed by the current political situation, and how Bangladesh can leverage this moment to further women’s digital empowerment.

Digital Connectivity: A Catalyst for Women’s Empowerment

Digital platforms offer Bangladeshi women access to education, healthcare, and economic opportunities that were previously out of reach. The rise of digital entrepreneurship, for instance, has enabled women to start businesses from their homes, breaking traditional barriers. Yet, barriers remain, including socio-cultural norms, digital illiteracy, economic constraints, and privacy concerns.

Despite the potential of digital connectivity, several barriers impede women’s access to and use of digital technologies in Bangladesh. Socio-cultural norms often restrict women’s access to mobile phones and the internet. A study supported by the WEE-DiFine Initiative at the BIGD illustrated that smartphones are considered taboo for women in some contexts, limiting their access to digital financial services. Furthermore, the Global Findex survey found that over 60% of unbanked adults in Bangladesh need assistance to use a financial account, with women being 12% more likely than men to report this.

Economic constraints also affect digital access. For many women, especially from low-income households, the cost of devices and internet services remains prohibitive. According to the white paper of BIGD’s WEE-Connect Initiative, while smartphone penetration rates have increased significantly in low- and middle-income countries (LMICs), gender disparities in meaningful connectivity—where women not only access but effectively use internet-enabled services—persist. Women face additional challenges such as limited autonomy over household spending decisions, restricted mobility, and lower levels of digital literacy, which further widen the gender gap. The WEE-DiFine Initiative’s white paper explores how digital financial services (DFS) can mitigate some of these barriers by improving women’s access to financial services through increased privacy and control over their finances. For example, digital accounts allow women to access loans and participate in market activities while adhering to cultural constraints on mobility and safety concerns. These trends naturally transition to a consideration of the larger external forces that can shape women’s digital inclusion, such as the political landscape.

The Political Landscape: A Moment of Risk and Opportunity

The political instability following the government’s collapse poses a threat to ongoing efforts aimed at promoting digital inclusion. Disruptions in funding for infrastructure and literacy programs may hinder progress. For instance, key initiatives aimed at expanding internet coverage in rural areas could be delayed or shelved due to shifts in government priorities. This would particularly affect women in low-income or remote areas, where access to technology and education on digital literacy is already limited. Without stable funding, the necessary expansion of broadband infrastructure or the continuation of mobile literacy training programs could be at risk, widening the digital divide.

Additionally, international donors and organizations that provide financial and technical support for these initiatives may hesitate to continue investments in an unstable political climate, further impeding progress. Any delays in policy implementation or shifts in government focus could disrupt long-term plans for ensuring equitable digital access for women. However, this period of transition also presents an opportunity to introduce gender-sensitive reforms that could advance women’s access to technology.

Interim Chief Adviser Muhammad Yunus has pledged to stabilize the situation, and with political will, this could become a turning point for empowering women through digital initiatives. Effective collaboration between the government, private sector, and civil society will be key to scaling these efforts. Programs like WEE-DiFine and WEE-Connect offer a model for fostering partnerships between researchers, policymakers, and practitioners to promote gender-inclusive digital policies.

The Data Gap: A Critical Challenge

A significant barrier to promoting women’s digital inclusion is the lack of gender-disaggregated data. Without this data, policymakers struggle to create interventions that address the unique challenges women face in accessing and using digital technologies. This gap limits the potential for informed, evidence-based policies and risks reinforcing existing inequalities.

The call to action is clear: we need more investment in gender-disaggregated data collection and analysis in the digital sphere.

Initiatives like WEE-DiFine and WEE-Connect are working to fill this gap. WEE-DiFine supports research on the impact of digital financial services on women’s economic empowerment. Meanwhile, WEE-Connect expands this framework to include broader digital connectivity, generating data to inform policies that can survive political transitions and adapt to the rapidly evolving technology landscape.

Success Stories: Digital Platforms Empowering Women

Despite these challenges of unequal access and persistent gender inequities, several digital initiatives in Bangladesh and beyond demonstrate that technology, when informed by thoughtful research, can transform women’s lives. The collaboration between digital interventions and research is pivotal in two key ways: this approach helps tailor solutions to women’s unique needs by context, and can subsequently demonstrate the effectiveness of these interventions to policymakers, setting the stage to scale successful solutions. Here are two examples of how BIGD’s WEE Initiatives have supported research on digital platforms that contribute to empowering women:

One local success story is HelloTask, a gig economy platform that connects female domestic workers in Dhaka with short-term job opportunities, aiming to help women increase their earnings and gain greater control over their work schedules. However, despite their integration into the cash economy through wage earnings, very few of these female domestic workers robustly use digital financial services.

A study conducted by WEE-DiFine in collaboration with HelloTask revealed that, although most workers in the baseline sample had access to mobile phones and had used mobile money accounts, fewer than 5% had ever deposited money into their accounts. In response, the research team designed an intervention that included in-person mobile financial services (MFS) training. When many women struggled to attend this training due to travel difficulties and competing household responsibilities, the team adapted their approach. They introduced in-home training and automated refresher phone calls, offering a more inclusive solution that navigated barriers specific to these Dhaka-based domestic workers. Preliminary analysis reveals that the MFS training significantly increased women’s knowledge of mobile money, savings via formal channels, and women’s labor time and earnings, relative to the control group.[1]

Another example comes from a Kenya-based study in partnership with ACRE Africa. This case exemplifies how academics, armed with gender and economic theory, and knowledge of existing empirical evidence, can partner with implementers to develop and test solutions that work for women.

From the project’s inception, the research team worked closely with ACRE Africa to design an agricultural insurance product that addresses women’s unique needs. Recognizing that women often face distinct financial constraints and responsibilities, the team created an innovative policy, known as “timely pay insurance”, that allowed smallholder farmers to receive insurance payouts in installments during the growing season, or as a lump sum at harvest time. Additionally, the team employed a hands-on approach to explain payout options and gather farmers’ preferences for different payment schedules. Preliminary analysis hints at the merits of combining academic expertise with on-the-ground implementation knowledge – women were significantly more likely to adopt “timely-pay” insurance than traditional crop insurance.[2]

Thoughtful collaborations of this nature have the potential to lead to more effective and inclusive digital interventions that can improve women’s economic empowerment in the long run. Initiatives like WEE-DiFine and WEE-Connect are crucial springboards for these dynamic partnerships. BIGD’s WEE Initiatives support the design of interventions tailored to women’s specific needs, evaluate the effectiveness of these interventions, and ultimately generate the evidence needed to scale successful programs. By providing this empirical foundation, these initiatives enable policymakers to create inclusive, sustainable solutions based on evidence.

A Path Forward for Women’s Digital Empowerment

To fully unlock the potential of digital connectivity for women’s empowerment in Bangladesh, several key steps must be taken. Some are self-evident—expanding digital literacy programs, particularly in rural areas, and reducing economic barriers to devices and internet access are crucial steps. Additionally, ensuring online safety through stronger privacy protections will foster a secure digital environment for women. Finally, as argued above, prioritizing the collection of gender-disaggregated data on these interventions, both to inform their design and to generate valuable evidence, is essential for creating policies that address the unique needs of women. This evidence base is crucial, especially during a time of significant political change, as it can directly influence policy decisions and ensure that women’s voices are represented.

Collaboration between the government, academia, private sector, and civil society will be vital in scaling these initiatives. By focusing on these priorities and leveraging evidence to guide policy, Bangladesh can transform current political uncertainty into an opportunity for transformative change, ensuring that women are at the forefront of the country’s digital future.

[1] Rahanaz, M. and Wahhaj, Z. (2024). WEE-DiFine Initiative Final Report. Internal report submitted to BIGD: unpublished.
[2] Cecchi, F. and Kannan, S. (2024). WEE-DiFine Initiative Final Report. Internal report submitted to BIGD: unpublished.