Small business is the hero of modern capitalism. Owners of small firms are the virtuous strivers, the job creators and the plucky entrepreneurs who drive the economy. ‘Small businesses make a huge contribution to national prosperity and supporting Australian jobs,’ states the Labor Party in Australia. And you would struggle to find a political party in any Western democracy that disagrees. A British government official made the (unverifiable) claim that firms with fewer than five employees made 95 per cent of radical innovations. Even amid the divisive politics of the United States, as the satirist John Oliver recently noted, everyone seems to agree that ‘small business is the backbone of the economy’. In a world of international conglomerates and global capital, the proverbial Main Street proprietors get a lot of love.
For all the enthusiasm, a central puzzle remains: what, really, is the role of small business in the economy? Is looking out for small business a progressive goal? Surely, the public fascination with upstarts, bootstrappers, and innovators reflects ideals of independence, improvement, and a better tomorrow. Yet history reveals another story: a distinct and powerful small business mythology at the heart of modern political life. Beginning in the late 1970s, adulation of small business acquired a new and important role in modern capitalist countries. In particular, the Reaganite and Thatcherite movements turned to celebrating small business as a stalking horse to advance the very kind of economy that handicapped upstarts and small independent proprietors, and privileged big national and multinational corporations.
Although love for small business may seem like a timeless feature of capitalism, the widespread belief that small entrepreneurs hold the keys to economic revival is relatively recent. Across the wealthy world, starting around 1980, small business emerged from the shadows of ‘Big Business’, with newfound political, intellectual, and cultural clout. In the United States, President Jimmy Carter, cast himself as the first ‘small business owner’ in the White House since Harry Truman. Carter promised to help small businesses by rolling back government regulations. Small business lobbyists also became more active. The National Federation of Independent Business (NFIB), founded in the 1940s as a mail-order survey company, reinvented itself in the 1980s as an influential lobbying group on behalf of small businesses. Intellectual attention to small business increased as well. In 1970, eight American universities offered courses on starting a new business; by 1980, 137 did. Whole magazines devoted to entrepreneurship emerged. ‘After years of neglect, those who start and manage their own businesses are viewed as popular heroes,’ one commentator raved.
A key moment in the modern myth-making around small business came in 1978. That’s when MIT economist David Birch published claims – which he repeated in testimony before Congress – that small firms had accounted for 80 per cent of all new employment opportunities between 1968 and 1976. Critics quickly pointed out that Birch’s findings were quite wrong, largely because he defined firm size according to how many employees worked in a given location (like a branch office, factory, or store), not how many the firm employed altogether. In fact, most job creation, in the 1970s and today, comes from a small number of very fast-growing firms, while most small firms either fail (killing jobs) or remain small.
Birch later admitted that the 80 per cent figure was a ‘silly number’, but the claims took firm root in popular mythology and political rhetoric by the 1980s. ‘Small businesses create eight out of every 10 new jobs,’ said Richard Lesher, president of the largest pro-business lobbying organisation, the US Chamber of Commerce.
Subscribe to Aeon’s Newsletter
Small business is among the most powerful symbols of modern capitalism. Small business owners are frequently described as virtuous, self-reliant, and independent – the same characteristics Thomas Jefferson ascribed to free farmers in pre-industrial society, or that Max Weber used to explain the Protestant work ethic which, he argued, undergirded industrial capitalism in the late 19th century. Just as important, small business, by virtue of its limited scale and scope, avoids the moral baggage often attributed to Big Business – bureaucracy, market manipulation, and good-old-boy networking, for example.
Like many powerful symbols, small business is notoriously hard to define. When creating the Small Business Administration (SBA) in 1953, the US government officially defined one as ‘independently owned and operated and … not dominant in its field of operation’. Today, to qualify for an SBA loan, US manufacturers must have fewer than 500 employees, and non-manufacturers must have annual receipts below $7.5 million (although the government reserves the right to make exceptions). More qualitative traits – like the absence of managerial hierarchies, less formalised labour relations, and closer ties to local communities – also influence how some scholars define small businesses. To make things more complicated, ‘small business’ covers a diverse range of business functions, counting everyone from the small-town dry cleaner to the wealthy software start-up. We know small business the way US Supreme Court Justice Potter Steward knew pornography: when we see it.
Historically, however, ‘small business’ did not exist in any meaningful sense until the advent of ‘Big Business’ in the late 19th century. Before the emergence of large, vertically integrated, and diversified corporations, ‘small business’ was simultaneously everywhere and nowhere, and no one spoke on its behalf. Steel, oil, sugar, and cigarette producers emerged as the first Big Businesses, and in 1890 the Sherman Act inaugurated American anti-trust policy to protect smaller competitors from their monopolistic practices.
Big corporations, with big research grants from big government agencies, worked with big universities to bring you modern life
The real boom in small business political consciousness came in the early 20th century, with the rise of the chain store model. Rooted in the anti-trust tradition, the anti-chain movement championed small retailers who faced destructive competition from mail-order houses and department stores.
In the United States, the representative Wright Patman stepped up as the face of the anti-chain movement. Patman was a doughy, balding populist and segregationist Democratic congressman from rural Texas. First elected to Congress in 1928, the son of tenant farmers made his name as an avid defender of small companies – the ‘common man’ – against the predations of eastern bankers, industrialists, and chain stores. In 1935, Patman pushed legislation that limited the discounts large retailers could offer. Hailed as the ‘Magna Carta for small business’, the Robinson-Patman Act (Senate Majority Leader Joseph Robinson (D-AR) was the co-sponsor) became law. President Franklin Roosevelt worried that the law would hamper economic recovery, but signed it anyway in a gesture toward the popularity of the cause. Patman defended the measure for its commitment to ‘fairness’ – by making the same discounts available to all buyers (whether at a chain store or a small grocer), the law struck a blow against concentrated wealth and privilege while still preserving the consumer cost advantages that mass distribution had created.
The Robinson-Patman Act marked the end, not the beginning, of a policy regime that protected small firms. By the post-Second World War years, small business was a divided and weak community. An ethic of ‘bigness’ reigned. Big corporations, with big research grants from big government agencies, worked with big universities to bring you modern life – from pharmaceuticals to aerospace, computers to communications. By the time Wright Patman died in 1976, at age 83, the popular backlash against bigness and the renewed attention to small business had not yet taken hold.
But had Patman lived into the 1980s, he would likely not have recognised the new ways politicians embraced and defended small business. Throughout the first half of the 20th century, small business advocates like Patman had claimed that small firms were inherently virtuous and worthy of special protection, even if larger companies offered lower prices or greater efficiencies. Yet by the 1980s, a decade of recession, inflation, fiscal crises and weak productivity combined to recast political culture in wealthy capitalist countries. In the United States, Western Europe, and eventually Australia, the logic for defending small business shifted entirely: rather than a virtue unto itself, smallness became the antidote to the bloat and inefficiencies of bigness; independence, the source of innovation.
The revival of small business’s symbolic political appeal in the 1980s brought another key change: activists used it not to attack Big Business, but to go after big government. Wrapping themselves in the cloak of small business mythology, those conservatives successfully redefined a hundred years of debate over economic size.
These changes did not come easily. To the frustration of small business groups and many conservative activists, the Republican Party retained its longstanding image as the party of Big Business, particularly in the early years of the Reagan administration. Many small business owners complained that Republican tax policies favoured larger firms, which took advantage of loopholes and provisions for writing off the depreciation of large assets. In addition, they charged that the growing federal budget deficit – which expanded due to a combination of Reagan’s 1981 tax cuts and the sharp recession that lasted until late 1982 – led to high interest rates that hurt the little guys the most.
Members of the Reagan administration worried about their popularity among small business owners. ‘Small business is bedrock Republican,’ as Elizabeth Dole, the director of public liaison at the White House, told George Bush, the then vice-president, in 1981. Or at least, it should have been: most small business owners were middle and upper-middle class white men, and most held economically conservative politics. But some parts of the small business community were moving away, Dole warned, because they believed ‘this administration favours Big Business and corporate America’. In 1983, White House staffer Red Cavaney warned that the Democratic National Committee planned to make overtures to the small business community. If Republicans ‘become too heavily associated with the “big” at the expense of the “small”,’ Cavaney predicted, ‘this threat could pose some serious problems.’
Republicans picked up the rhetorical mantle of small business, but instead of changing their policy ideas, they changed what it meant to speak for small business. For the better part of a century, small business activists had stressed the virtues of competition. Small businesses, they argued, demanded legal support – through punitive taxes on market dominators and the break-up of monopolies – because their very existence created a more competitive market place.
Economic conservatives in the 1980s pushed a counter-narrative. Murray Weidenbaum – first chair of Reagan’s Council of Economic Advisors – charged that economic growth, not competition, should be policymakers’ primary goal. Certain sectors of the economy, including the rapidly growing service sector, lent themselves more productively to small-scale enterprises. Industrial manufacturing, on the other hand, did well when a small number of giant operators took advantage of their size to produce more efficiently at a massive scale.
‘Entrepreneur’ today implies a growth orientation: small business owners that don’t want to remain small business owners
What mattered to Weidenbaum wasn’t size or market share per se, but rather how good businesses were at growing, because only a growing economy would create new job opportunities. The single-minded focus on small business as the creator of jobs, in other words, confused cause and effect. ‘It is not the small businesses that created the jobs,’ he concluded, ‘but the economic growth’ (emphasis mine).
By putting the focus on growth, not small business as such, conservatives subtly manipulated the mythology of small business. Most small businesses do not grow into mid-sized or large companies, and in fact the vast majority fail within five years. Earlier small business proponents understood the nearly permanent condition that small business represented and treated small business owners as a stable class. The conservative politics of the 1980s, however, focused instead on a small subset of the small business community: entrepreneurs.
Although the classical definition of an ‘entrepreneur’ simply invoked someone who started a new business (the French word means ‘someone who undertakes’), the term acquired a new connotation in the late 20th century. ‘Entrepreneur’ today implies a growth orientation; while a mere small business owner may persist in remaining small, an entrepreneur seeks to strike it rich. In short, entrepreneurs are small business owners that don’t want to remain small business owners.
The growing fetish about entrepreneurship formed an integral part of the conservative project that blurred distinctions between small and large firms. President Reagan himself perpetuated this shift. Reagan – whose pre-political private sector experiences lay in Hollywood and at General Electric, two exemplars of mid-20th century Big Businesses – positioned himself as a populist defender of the people even while promoting an economic vision rooted in the interests of concentrated wealth. Bragging about a recovering economy in 1987, he insisted that ‘small businesses fare best with stable prices, low interest rates, and steady growth’. Moreover, ‘America’s entrepreneurs are continually experimenting with new products, new technologies, and new channels of distribution.’ Small businesses, in other words, achieved their value through their innovative contributions, rather than servicing or maintaining an existing system.
Yet Reagan betrayed the bait-and-switch. ‘The great industrial and commercial centers of our nation were built by innovators like Henry Ford and Alexander Graham Bell,’ he continued, ‘whose small businesses grew to help shape a new economy.’ At a stroke, the president – perhaps unintentionally – gave up the game: small firms’ worth came not from promoting competition or preserving local values, but rather from their potential to cease to be small businesses. Left out of this formulation, of course, were the millions of nail salons, fast-food franchises, accountants, landscapers, general contractors, housekeepers, cosmetics sellers, photography studios, restaurant owners, small town lawyers, and florists who would never become the next Ford Motor Company or AT&T.
Why does all this matter?
Since the 1980s, the pace of global capitalism has quickened, and economic transactions occur at a speed and complexity unparalleled in human history. At the same time, political culture has become increasingly fragmented and atomised. From the breakdown of party authority to tribalist politics and hyper-partisanship, residential and educational re-segregation to media segmentation, fracture dominates. The bigger things have got, the more powerful the urge has been to get small.
This manic contradiction – between the scale of modern life and the powerful siren call of the atomised locality – lies at the heart of a destabilising transformation within capitalism itself.
The present historical moment is witnessing the breakdown of the so-called ‘Berle and Means’ corporation – the shareholder-owned but manager-controlled, bureaucratic, and deeply interconnected organisation first described in Adolf Berle and Gardiner Means’s book, The Modern Corporation and Private Property (1935). Since the end of the mid-20th century conglomerate wave, corporations have concentrated and streamlined. Since the 1990s, the number of publicly traded companies has declined. Liberalised trade and cross-border capital flows have accelerated the ‘Nike-fication’ of production, birthing a world where anonymous and poorly regulated sweatshops in developing countries pay paltry wages to workers who manufacture items adorned with a global brand. The internet created new opportunities for instant communication and coordination, and firms responded by outsourcing and off-shoring far more than production. Spinning off their financing, distribution, advertising, human resources, and customer service functions to the lowest bidder, many of the world’s biggest businesses are today little more than coordinators of a massive network of nodes. The dissolution of the classical corporation emerged alongside a new business focus on portfolio management and short-term valuation. Such managerial priorities reflect the rising ideological and economic clout of the ‘shareholder-value’ movement as well as a broader commitment to a neoliberal vision of value.
This breakdown of the corporation as an economic and social institution is a critical feature of capitalism today, and it deeply shapes how we value – and overvalue – small business. The disintegration of the old order, although couched in populist language of ‘shareholder democracy’, has generated uncertainty and dislocation as well as freedom and opportunity, and those ups and downs have not been distributed evenly. The well-educated with privileged access can take advantage of the new niches that open up, and become entrepreneurs. Those in the lower tiers, however, confront a deteriorating employment landscape pockmarked by wage stagnation, decreased mobility, and lower-paid and low-benefit jobs. Social safety nets are evaporating, and wealth inequality is expanding. ‘Necessity-based’ self-employment is rising in rich and poor countries alike. Self-sufficiency has always been part of the allure of opening one’s own business. In the globalised, atomised economy, it has also become an unstable lifeline.
By linking the political agenda of small business and large business, conservatives in the 1980s laid the foundation for a set of policy developments that hastened the globalising forces of late-stage capitalism and failed to mitigate its effects. By presuming that small business was uniquely or exceptionally innovative, they ignored the real world of small business owners and perpetuated a devastating myth that judged small companies by their ability to become Big Businesses. In so doing, they missed the most critical developments in global capitalism: the simultaneous fracture of the mid-century corporate world and the rise of an isolated, privileged global elite that marginalised and weakened the vast majority of small businesses.
Syndicate this Essay
EconomicsPolitics & GovernmentProgress & ModernityAll topics →
is associate professor of history at University of North Carolina, Chapel Hill, where he teaches courses in politics, business, and capitalism. He is the author of Lobbying America: The Politics of Business from Nixon to NAFTA (2014) and The Land of Enterprise: A Business History of the United States (2017).
We conducted a large randomized evaluation of a savings-led microfinance program across three countries. This evaluation provides important evidence on the impact of a popular development intervention on the lives of low-income households in rural communities by looking at its effects on usage of financial services, microenterprise activity, income, female empowerment, consumption, and the ability to cope with shocks.
Savings-led microfinance programs operate in poor rural communities in developing countries to establish groups that save and then lend out the accumulated savings to each other. Nonprofit organizations train villagers to create and lead these groups. In a clustered randomized evaluation spanning three African countries (Ghana, Malawi, and Uganda), we find that the promotion of these community-based microfinance groups leads to an improvement in household business outcomes and women’s empowerment. However, we do not find evidence of impacts on average consumption or other livelihoods.
The poor make complex financial decisions and use the limited range of financial instruments available to them to address their varying needs. The available formal and informal tools, however, are often risky and expensive or lack necessary flexibilities (1). The quest to improve access to appropriate financial services for poor households has traditionally focused on providing credit via formal alternatives to informal moneylenders. The most relevant example is the growth of the Grameen microcredit model developed by Mohammed Yunus.
However, the limited participation and geographic reach of microcredit institutions, especially among the rural poor, have shifted efforts into expanding access to savings. A growing impact literature on microcredit and microsavings shows strong welfare impacts of the latter type of programs, suggesting that they might have more transformative impacts than credit programs as currently typically implemented (2, 3).
In parallel to the development of microcredit and microsavings products for the poor, many nongovernmental organizations (NGOs) have begun promoting informal savings-led microfinance groups that emulate and improve on the model of informal associations indigenous to many societies [often called Rotating Savings and Credit Associations (ROSCAs)]. Although often implemented as a stand-alone program, these savings groups are also often a component within multifaceted programs (4). The appeal of the savings-led microfinance approach is shown by the growth of these groups, which now reach over 10 million people in more than 70 countries* after only a few years of significant expansion efforts. Many actors, including international donors, such as the Bill and Melinda Gates Foundation,† and nongovernment organizations,‡ have pushed to create and expand such groups, viewing them as a grassroots and low-cost mechanism to provide (albeit informal) financial services to the poor. Similarly, the Andhra Pradesh Government in India promoted a self-help group model (Velugu) as an alternative to formal microcredit during a crisis in 2010 (5). The model, which is being explored by many NGOs, is a type of savings group with the added feature of explicit intentions to link groups to formal sector banks for credit after they mature enough [e.g., Plan and CARE (Cooperative for Assistance and Relief Everywhere) have partnered with Barclays for this purpose] (6).
Savings-led microfinance groups vary depending on the implementing organization and the context; however, the basic features follow the Village Savings and Loan Association (VSLA) model developed in the early 1990s in Mali by CARE, one of the leading promoters of these groups and the main partner in this study. The VSLA model was designed as an improvement of the local tontine, a type of savings group where members gather at regular meetings to contribute a fixed amount of money and the total pot is assigned in full to each member in turn. VSLAs adopt the same meeting and contribution structure but introduce flexibility. At each group meeting, members can decide to contribute more than the agreed minimum and can take a loan from the group without having to wait for their preassigned turn. These loans are charged an interest rate, so that the money deposited by group members can earn interest. Savings and loan repayments are kept in a group lockbox that can only be opened at group meetings and “shared out” among members at the end of a predefined cycle. A VSLA may have an additional social or solidarity fund, which is an insurance fund managed by the group that can be accessed by members in the form of an interest-free loan or cash grant in case of an emergency. Disbursements for these purposes are assessed and determined by the group. VSLAs do not receive any capital through grants or external loans; the pots simply grow over time as individuals collectively accumulate more savings. We present the results from three randomized, controlled trials of the VSLA program implemented by CARE and its partners in Ghana, Malawi, and Uganda over a period of 22–30 mo.
First, we report evidence on the program’s impact across eight outcomes. The growth in financial intermediation resulting from the program led to improved microenterprise outcomes and women’s empowerment. The positive impact on female empowerment stands in contrast to the results of six of seven randomized trials on microcredit (3, 7⇓⇓⇓⇓⇓–13). However, when examining average treatment effects, much like the microcredit studies, we do not find impacts on typical welfare indicators, such as household income, consumption, food security, asset ownership, or community participation.
Second, we examine the program’s effect on resilience to shocks. In villages that suffered from drought (an aggregate shock to the community), we find a positive impact of the program on income (although significance does not hold up to multiple hypotheses correction). There is also some evidence (again, weak statistically) that the positive impact of VSLAs on female empowerment is reduced in villages subject to drought. We find no evidence of differential program effects for households with a self-reported bad harvest (which is a tentative indicator of idiosyncratic shocks after controlling for aggregate rainfall but subject to potential mismeasurement and selection concerns discussed below). Thus, in net, we find at best suggestive evidence that VSLAs may influence risk management capabilities, more so for aggregate shocks than for idiosyncratic shocks.
The results here can be compared with those found in two other randomized evaluations of savings-led microfinance groups in Mali (14) and Malawi (15). In Mali, researchers found positive impacts on food security and investment in cattle but found no impact on female empowerment. In Malawi, researchers found positive impacts on consumption on average (not merely during shocks), business income, and also, investment in housing structure (specifically, expansion in the number of rooms).
This study received approval from the Yale University Human Subjects Committee [Institutional Review Board (IRB) nos. 0805003819, 0904005015, and 0903004937] and the Innovations for Poverty Action Human Subjects Committee (IRB Protocols 010.08April-003, 109.09April-001, and 110.09March-001).
Implemented by CARE and 13 local partner NGOs in the study areas, the VSLA program has three main components: (i) a group-based commitment savings mechanism, (ii) a process for members to request loans from the group at any point, and (iii) a social or emergency fund financed by members with a regular contribution. In each site, after an initial community meeting to introduce the program, trained officers or agents form and guide VSLAs for an initial cycle (usually 8–12 mo) and provide oversight and support for a second cycle. Groups are comprised of 19–30 members, mostly women, who choose to come together as a result of an agent’s promotional activity in the community or after having observed other groups.
Table S1 compares the implementation strategies and characteristics of VSLAs across the three sites. In each country, the role and mandate of the trainers were slightly different. In Ghana, officers from the NGO (called field officers) were in charge of creating VSLAs in each of their villages. In Uganda and Malawi, however, implementation was designed to facilitate scaling the program in a sustainable way: Local people were trained as village agents, who then trained other groups in the village and surrounding areas for a fee. The Malawi program was designed to rely heavily on these agents at the onset. Program data from the last 3 mo of the study (2nd quarter, 2011) show that 90.3% of the groups created by study partners were created by village agents.§ In Uganda, the field officer was given the mandate to work in a group of villages and begin training agents only later on in the program’s timeline. The difference in strategy is reflected in the fact that, in the same period, only 20.4% of the groups were formed by village agents. Notably, research in Kenya finds important differences in the efficacy of savings-led microfinance groups managed by NGO-paid agents and village agents paid by group members themselves: the village-paid agent model is less expensive; generates more borrowing, savings, and enterprise investment; and likely does this by attracting more business-oriented members (16).
Take-up rates at the end of the study averaged 31.6% across the three sites, with a steady increase over time. We find evidence of the program successfully replicating VSLAs beyond the primary target village, with take up increasing over time and also reaching the control villages, which had a take-up rate of 6.2% by end line. Take-up trends and overall numbers varied slightly across sites as shown in Fig. 1.
Program uptake. Percentage of female primary respondents joining a VSLA in the study sample in treatment (solid lines) and control (dotted lines) clusters. GH, Ghana; MW, Malawi; UG, Uganda.
At the time of the end-line survey, respondents who had joined a group had been members for a median value of 14 mo, with over 61% of them having completed a full savings cycle and receiving at least one share out. Members made small weekly savings contributions, with median values ranging between purchasing power parity (PPP) 2011 US$0.66 and $0.84 across the three countries.¶ The main reported uses for savings share outs (median value of PPP 2011 US$38.5) were agricultural investments (22%), food (16%), and education costs (16%).
At end line, 68% of members had received at least one loan from the group. The median loan was PPP 2011 US$19.7, with a median interest charge of 10% flat. The main stated uses for loans were business (29%), food (13%), and education (13%).
Program design summary
We present the results from randomized, controlled trials of the VSLA program across a total of 561 clusters, 282 of which were randomly assigned to treatment and the remaining of which were randomly assigned to control. Treatment clusters received initial promotion and group formation activities by agents of the implementing organizations. Table S2 presents key features of the research design across the three research sites.
In Ghana, a cluster consisted of one community identified by the partnering NGO as a viable village for implementation of the program. In Malawi and Uganda, to measure organic replication activities of the program, a cluster consisted of two villages: one was identified as the primary village, and a second one was randomly chosen from other villages within a few kilometers of the primary village. The location of the secondary village was not disclosed to the partner NGOs to avoid targeting and, thereby, permit us to measure organic replication. Random assignment was stratified at the district level in all three countries and on other variables depending on the country, as specified in Table S2.
In each country, the village representatives were first approached to help identify adequate households for surveying. Surveyors then created a list of household heads in the village, from which they randomly selected participants. In the case of Uganda and Malawi, a majority was selected from the primary village, and a smaller number was selected from the secondary village. Overall, we collected panel survey data on 15,221 households from the three sites (survey instruments are available from the authors); 13,564 households were surveyed at baseline, 91.2% of which were resurveyed in the follow-up, and an additional 2,845 households were surveyed in Ghana at end line to mitigate concerns about lower than expected take-up rates in treatment villages and the proximity of control villages to treatment villages, which caused some of the former to adopt the VSLA program.
Four surveys were administered within each wave of data collection. First, a household survey collected information on indicators, such as agricultural production, income-generating activities, and economic shocks. Second, an adult survey collected information on the individual’s experience of gender issues and community involvement and on their savings and loans activities. For the majority of our sample (13,502 households), the adult survey was administered to an adult woman in the household.# Third, a village survey gathered information about various characteristics of the community. Fourth, a market survey was used to record the market prices of a variety of staple foods grown by respondents.
Research design summary
As is the standard practice for multisite trials, we estimate a pooled model controlling for baseline values of our outcome variables, country, and district, which was used as a stratification variable in each site. SEs are clustered at the village level (comprised of a primary–secondary village pair for Uganda and Malawi). Each column in Table 1 is a result of an independent ordinary least squares regression modeled with the following specification:where is the index k of interest for either the household or the female primary respondent i in cluster c. is an indicator variable for whether the cluster was randomly assigned to receive the VSLA program or not, is the household or female primary respondent baseline value of the outcome index k (if not available, the baseline value is replaced with −9, with a dummy variable indicating a missing value), is a vector of dummy variables for each of three countries in the study, and is a vector of dummy variables for districts.
The analysis used here is an “intent to treat” method, where we compare those who were randomly assigned to receive treatment with those who were not, irrespective of actual take up. This approach is important for interpretation and also in accordance with the program’s promotion as a community-level intervention. There are two main reasons that we do not estimate the “treatment on the treated,” in which we would use random assignment as an instrumental variable for participation in a savings-led microfinance group to estimate the effect of participation on the outcomes of interest. First, participation in a VSLA is a continuous variable, with some just having started participation and others participating for a long time. Second, such a specification requires there to be no indirect effects on others in the community, which contradicts one of the main arguments for savings-led microfinance programs: that they improve overall community wellbeing and social capital.
Because the program was multifaceted and we collected data on a large number of outcome variables, we are concerned with multiple hypothesis testing, which we address in two ways. First, we create indices for six of eight families of outcomes, which avoids interpreting one individual outcome measured with noise as indicative of a genuine change in an outcome. Second, we calculate q values across outcomes using the Benjamini–Hochberg step-up method (17) to control for false discovery rates. However, theory and evidence from elsewhere on impacts of improved financial inclusion (3) suggest that not all families of outcomes should be considered as equally likely to shift.
For the outcomes, we define six main indices and two aggregate measures: consumption and income. Each measure is formed by grouping together variables of related outcomes.
The financial inclusion index measures the extent of female primary respondents’ involvement in the informal credit and savings sector and includes savings balances, loans taken in the last 12 mo, total amount of loans received, whether a person has savings, and whether the respondent is a member of a savings group.
The food security index uses metrics on meal patterns to determine food consumption during the 12 mo prior to surveying. It includes food intake reduction, days without eating for adults and children, and a variable indicating whether the household resorted to borrowing food.
Income is an aggregate measure of household income and revenues from all common sources: agriculture, livestock, business profits, and paid labor. All income measures refer to households’ reported monthly values. Monthly household income is calculated as self-reported revenues minus expenses for all income-generating activities carried out by the household in the 12 mo before the survey. Annual microenterprise income is coded as zero for microenterprises operating less than 1 mo that year. Livestock income is the sales proceeds derived from livestock in a month minus the cost of livestock purchases. We do not deduct other input costs when calculating livestock income. Business profits are the profits earned through the households’ enterprise(s) calculated as revenues minus costs. If a business operated for less than 1 mo, its profits were not included. Finally, income from paid labor is the income earned by a household member by being a paid employee, including agricultural day laborers.
The business outcomes index measures nonmonetary performance of microenterprises run by household members over the 12 mo preceding a survey. It includes the total number of businesses operated by household members, the sum of months that these businesses were in operation in the year preceding surveying, and whether any of the household’s businesses employs labor from outside the household.
The assets index combines two standardized indices measuring household assets and productive assets held by the household. The calculation of values of assets was standardized across the three countries and time periods by expressing each asset value in relative terms: specifically as the number of bicycles needed to purchase one unit of the asset. More information on the construction of this index is provided in SI Text.
Consumption is an aggregate measure of monthly per capita expenditure on food and nonfood items. Food consumption expenditure used a 1-wk recall, and included only those items for which data were collected across the three sites (grains, tubers, nuts, and beans), meaning it presents only a partial view of the household’s consumption basket. The nonfood consumption measure includes monthly per capita household expenditure on transport, clothes, electricity, and petrol.
The women’s empowerment index captures women’s self-reported influence on household decisions, particularly in relation to food expenses for the household, education and healthcare expenses for the children, business expenses if the household operates a business, and the women’s ability to visit friends.
Finally, the community participation index captures self-reported involvement in community affairs based on whether the female primary respondent has raised an issue before a village chief, government authority, or village council; whether she has attended a community meeting in the 12 mo preceding surveying; and whether she has participated in any social group within the community (a women’s or workers’ group, for instance).
Table 1 presents the main results of all eight outcome variables, and Tables S3–S10 show their breakdown by components. The indices are constructed as , interpreted as the outcome variable k for adult respondent (or household) i in family of outcomes f within a country c. All indices and variables have been defined, such that a higher value corresponds with improved outcomes. A detailed explanation on the method followed for index construction is provided in SI Text.
Information on drought and bad harvest interactions in Tables S3–S10 examines a critical heterogeneity: whether the program had a differential impact during times of drought or among households reporting a bad harvest. We obtained rainfall data from the daily Africa Rainfall Climatology 2 dataset (18). Based on these data, we created an indicator variable for all villages that experienced a drought (defined as an annual rainfall less than 1 SD below a historic average) 12 mo before the end-line survey. In Uganda and Malawi, 34.6 and 15.4%, respectively, of villages experienced drought. Ghana is excluded from this analysis because the period of our study was exceptionally rainy; all but one village had a larger than average rainfall. In addition, we have end-line data from Uganda and Malawi on households’ self-reported experience of a “bad harvest” over the previous year (80.1 and 33.2% of households, respectively).
Impacts of the program on families of outcome variables
Index components pooled sample—financial inclusion (adult female respondents)
Index components pooled sample—food security (household)
Index components pooled sample—income and revenue (household)
Index components pooled sample—business outcomes (household)
Index components pooled sample—asset ownership (household)
Index components pooled sample—monthly household consumption (household)
Index components pooled sample—women's empowerment (adult female respondents)
Index components pooled sample—community participation (adult female respondents)
At end line, we find substantial positive impacts on financial inclusion as a direct outcome of the program. Table S3 shows that overall participation in informal savings groups (including ROSCAs and other types of groups) is 17.4% points (SE = 0.015) greater for female primary respondents in program areas. Total reported savings are also significantly higher by PPP 2011 US$13.7 (SE = 4.488), equivalent to a 34.5% increase relative to savings balances of respondents in control areas. As expected, the program improves access to credit as well: 42.1% of women obtained a loan in the year leading up to the end-line survey, an 11% point difference (SE = 0.012) from the control group. The average amount borrowed in a year is 16.2% higher or PPP 2011 US$6.6 (SE = 3.851) in program communities. Drought and bad harvest interactions in Table 1 find no evidence that the positive impact on financial inclusion is any stronger for households in treatment villages who experienced a poor harvest.
Table 1, food security index (household) shows that the program, on average, had no significant positive impacts on food security. A closer look at bad shocks (Table 1, drought and bad harvest interactions) shows that drought has a strong negative impact (−0.119 SD, SE = 0.043) on food security for households in control villages as does a poor harvest (−0.590 SD, SE = 0.036). There is no evidence that the program improves food security for treatment households reporting a poor harvest or those who experienced drought [although the point estimate for the latter is large enough to be important (0.092 SD with SE = 0.057); thus, we cannot rule out positive and important impacts on food security].
Similarly, the program does not show a positive effect on income for the overall sample. We do find, however, a positive program effect (PPP US$26.4, SE = 15.002) for treatment villages suffering from a drought event, although it is not statistically significant after correcting for multiple hypothesis testing (q value = 0.260). When we analyze components of the household income measure in Table S5, we find that the program has an overall positive impact on monthly household business profits of PPP US$5.6 (SE = 1.857) or 24.4%, but it does not affect income for other household activities. There is no evidence that the program improves income or revenues for households reporting a poor harvest. These findings suggest that the VSLA program has an overall effect on business profits and also insulates members from adverse aggregate shocks on their economic activity more broadly.
Next, we find a 0.06-SD (SE = 0.023) positive impact of the program on an index of business outcomes. Table S6 shows that VSLAs lead to a slight increase in the total number of businesses operated by the household (0.024 businesses, SE = 0.014; equivalent to a 6.3% increase). Households in treatment communities operated businesses 0.20 mo (SE = 0.098) longer than the control group, equivalent to an 8.6% increase. It should be noted that businesses operated by households in the sample are mostly short-term seasonal businesses, with an average total of 2.33 mo of business operation per household in the control group. Although it is fairly uncommon for businesses to use outside labor, we find that VSLAs lead to a 1.0% point increase (SE = 0.004)—a 26.3% increase compared with a 3.8% control group average in the number of households with at least one employee in its business(es). Thus, the VSLA program stimulates investments to extend and expand businesses operated by the households. Information on drought and bad harvest interactions in Table S6 shows no evidence that these positive effects on business activities are any different among households that experience a bad harvest or in communities that experience drought.
Table 1, total asset index (household) and Table S7 find no impact on total assets owned by households nor any differential impacts for those experiencing bad shocks.
Table 1, monthly per capita consumption ($; household) and Table S8 find no impact on consumption in both the aggregate or the component measures, nor do they find a differential impact for those experiencing drought. Here, it should be noted that the consumption measure refers only to 30 d preceding the interview and therefore, could miss consumption smoothing benefits.
Table 1, women’s empowerment index (adult female respondent) shows a significant impact of the VSLA program on the empowerment of female primary respondents. Surveyed women in program communities display a 0.06-SD (SE = 0.034) increase in their influence on household decision-making. Table S9 shows that this impact is strong across the board, with a 4.2% point (SE = 0.016) improvement in the share of women who have a high degree of control over household business decisions, a 3.7% point (SE = 0.014) increase in control over food expenses, and a 2.9% point (SE = 0.015) increase in women with influence on education expenses. The results in Table 1, drought and bad harvest interactions suggest that this increase in influence occurs only in communities that are not experiencing a drought: there is a −0.119-SD (SE = 0.065) decrease in female influence on household decision-making in treatment communities that suffer from drought.
Finally, we do not find evidence of a similar impact on female primary respondents’ community participation. However, Table S10 does show a significant 2.3% points (SE = 0.011) increase in the number of women who attended a community meeting in the last 12 mo.
Datasets S1–S7 provide some additional results of interest, such as balance and attrition analyses, and breakdowns of impact by country and components.
We find important impacts after 2–3 y of treatment beginning: VSLAs facilitate investment and empower women. We find no evidence of differential impacts on any of the outcomes after an idiosyncratic agricultural shock (experiencing a bad harvest after controlling for bad rainfall). It is possible that the VSLA treatment could affect the likelihood of a poor harvest conditional on the aggregate shock by changing agricultural investment or technology. In this case, the unobserved characteristics of farmers reporting a poor harvest conditional on bad rainfall in a VSLA community might be different from those of farmers reporting the same in a control community, making the coefficient of this interaction difficult to interpret. However, we find no evidence of a VSLA impact on farming activities (Table S5, all countries), suggesting that any such selection effect is minimal.
The positive impact on income after bad rainfall (an aggregate community shock) could suggest that VSLAs may work through a savings mechanism, not merely an insurance mechanism that builds risk-sharing within the community. Naturally, the two are not mutually exclusive channels through which the program can generate impact. Given that these results do not maintain statistical significance after correcting for multiple hypotheses, we encourage replication of these evaluations.
Beyond establishing the base evidence for the average impact of these programs, three key and related questions remain. What exactly is new? Why does the specific pattern of impacts vary across countries? What is the mechanism through which these programs generate change? A VSLA requires no external capital and no legal infrastructure to issue and collect loans and is, in fact, similar to existing informal mechanisms, such as ROSCAs. However, average impact ensued. The average household in a community offered an opportunity to form a VSLA increased its financial savings by almost 35% or $14. Average household monthly consumption expenditure is about $200, and therefore, this change in savings balances is significant, albeit unlikely to be transformative. Therefore, it is unsurprising that there is no evidence of average impacts on household income, consumption, food security, or asset ownership. We instead find changes in the realms most directly linked to the VSLA: nonfarm business operations and women’s empowerment.
However, these effects vary across populations and countries. A detailed exploration of differences in implementation across the three countries may help us understand the mechanism through which it generates the average impacts that we have documented. A savings-led microfinance group might be thought of as a new “social” technology, one for which all of the necessary components existed beforehand but that still required some external agent to promote and train groups. Indeed, researchers in Mali found that, when villages started these groups via word of mouth rather than through direct NGO promotion and training, they did not work as well (14).
In terms of mechanisms, we start with the simple: the near-zero transaction costs (i.e., no travel time required, merely time to attend a meeting) can explain its impact vis à vis formal banking. However, low costs do not explain the impact when compared with preexisting informal options, such as cash stored at home. Compared with these options, savings-led microfinance groups can be best thought of as a social commitment savings account. This logic relies on models of temptation, time-inconsistent preferences, or household bargaining inefficiencies to explain the demand for commitment. Furthermore, although some informal savings vehicles do have a form of commitment (e.g., livestock), such vehicles can be costly and not as effective as a commitment device that uses social capital to help ensure compliance.
Last, we consider the benefits relative to the costs. We find no evidence that the VSLA program changes average income or consumption, which poses complications for a benefit–cost analysis. The benefits of the program are subtle: possible improved income in the face of drought, empowerment of women within the treated communities, and improvements in a broad array of business outcomes. Asserting a monetary value for these improvements would require a series of strong modeling assumptions that would of necessity be tentative and are beyond the scope of this paper.
Despite the lack of a benefit-cost analysis, it is important to note that the program’s cost per household is low. We obtained data on program costs from Malawi and Uganda as of June of 2011, but such data were unavailable for Ghana. The most conservative estimate yields an average cost per member of $26 in Malawi and $20 in Uganda over 22 mo of program implementation (not including the opportunity cost of the participant’s time).‖ Considering the take-up rates of 22 and 36% among female primary respondents in Malawi and Uganda, respectively, and the fact that 20 and 36% of VSLA members are men in Malawi and Uganda, respectively, the most conservative estimate of the costs of the program per household in the cluster would be based on the assumptions that all of the male members have the same take-up rate as that evidenced in female members and that the former come from different households than the latter. Under these assumptions, the average costs of the program per household in Malawi and Uganda are $7 and $11, respectively.**
In 2008, we began a series of evaluations to study CARE’s savings group model and measure its impact on the lives of the people in program communities. The studies were conducted on the occasion of three scale-up efforts by CARE: the ESCAPE Program in Ghana and the SAVE UP Programs in Uganda and Malawi.
The scale-up efforts replicated the same savings group methodology in a large number of communities across the three countries. The operational models in the three sites, however, followed three different models, with implications for our study design and analysis. In Ghana, the program was implemented by NGO trainers who were in charge of creating VSLAs within the limits of their own villages. In Uganda and Malawi, CARE tested a replication model meant to more rapidly and cost-effectively form groups in target villages and surrounding regions (19). In all sites, CARE worked closely with local organizations rooted in the area to implement the VSLA program.
Village agents are VSLA members with particular leadership qualities who are selected from an initial set of VSLA groups established by the implementing organization. These agents serve as promoters of new VSLAs, supporting these groups for the first year of activity. This replication model stems from observing the natural spread of savings groups in other countries, where the simple presence of a savings group was enough to spur the spontaneous creation of new savings groups. The model is designed to maintain the quality and core elements of the VSLA methodology for these new VSLA offshoots and develop a team of trainers who can sustainably replicate the model through a fee for service arrangement, in which group members pay for the training services provided by the independent agents.
In Malawi, officers of the implementing organizations were each asked to create 12 groups in 12 different villages and follow them through their first savings cycle. From each of these groups, officers selected, on average, one member to be trained to become an agent. At the end of this initial cycle, the field officers transitioned into a supervisory role monitoring between 10 and 15 agents, each tasked with forming a minimum of 8 new groups on a fee for service basis (approximately $15/y per group) within the timeframe of the study.
In Uganda, implementing organizations used officers who resided within a target village. Each officer had the responsibility of forming and overseeing a target number of 24 groups in the village and its surrounding areas within 3 y. During the second year of activities, officers were tasked with identifying and training three agents (from existing groups) and mentoring them as they each formed two new groups.
To meet the implementation requirements of the scale-up program in Malawi and Uganda and measure the extent of the success of these replication strategies, we created clusters of villages as the randomization unit for our study. Clusters were designed in close collaboration with CARE and comprised a sufficient number of villages and households to meet the replication targets of the implementing partner organizations according to their model. In each cluster, we selected a primary village that would be targeted as the initial center of activities by the implementing partner. A secondary community was then selected within a 4- or 2.5-km radius of the primary village in Uganda and Malawi, respectively. The locations of secondary villages were not disclosed to implementing partners to avoid targeting of these villages. The data collected on households from these secondary villages allowed us to measure the extent to which the program extended beyond the primary village selected for implementation.