Development organizations have spent billions trying to help small businesses in poor countries grow. The logic seems impeccable: train entrepreneurs in business skills, give them access to capital, connect them to markets, and watch poverty decline as enterprises flourish. Yet after two decades of rigorous experimental evaluation, the evidence base for private sector development interventions remains stubbornly weak.

This isn't a story of failed programs or incompetent implementation. It's something more troubling for the field. Randomized controlled trials consistently show that our most popular interventions—business training, microfinance, market linkage programs—produce effects that are small, temporary, or difficult to detect at all. The gap between theoretical promise and measured impact has forced a fundamental rethinking of how we approach enterprise development in poor economies.

The persistence of weak evidence deserves serious examination. Either we're measuring the wrong things, intervening at the wrong points in firm development, or operating with flawed assumptions about how small businesses actually grow. Understanding which explanation holds—and the answer likely involves all three—matters enormously for the allocation of development resources and the design of future programs. The stakes involve not just effectiveness but the opportunity cost of funds diverted from interventions with stronger evidence bases.

Business training programs represent perhaps the most widely implemented private sector development intervention. The premise is straightforward: entrepreneurs in developing countries lack formal business education, so providing training in accounting, marketing, inventory management, and financial planning should improve firm performance. Hundreds of such programs operate globally, often as flagship initiatives for development agencies.

The experimental evidence tells a more sobering story. Meta-analyses of randomized evaluations consistently find average effects on profits and revenues that are statistically significant but economically modest—typically in the range of 5-10% improvements. More troubling, these effects often fade within a year or two of program completion. The knowledge transfer occurs, but translation into sustained business practice proves elusive.

Several mechanisms explain the gap between learning and application. Training programs typically deliver standardized curricula to heterogeneous populations of entrepreneurs, many of whom face binding constraints unrelated to skills—unstable demand, infrastructure failures, regulatory harassment. Knowledge alone cannot overcome these structural barriers. Additionally, the opportunity cost of attending training sessions may exceed the value of marginal business improvements for subsistence entrepreneurs.

The evidence is somewhat stronger for training programs that target specific skills to specific populations. Personalized business consulting shows larger effects than classroom training. Programs focusing on psychological and behavioral aspects of entrepreneurship—goal-setting, mindset, personal initiative—have produced more durable results than technical skills training. This pattern suggests that binding constraints may be internal as much as external, involving motivation and aspiration rather than knowledge per se.

Recent research has shifted toward identifying which entrepreneurs benefit most from training interventions. Pre-existing business size, owner education, and initial profitability predict larger treatment effects. This heterogeneity analysis points toward better targeting but also raises uncomfortable questions about whether training programs serve those who need them most or merely accelerate success for entrepreneurs already positioned to grow.

Takeaway

Training programs transfer knowledge effectively but rarely transform businesses, suggesting that skills deficits may be less binding than the structural and psychological constraints entrepreneurs face.

The theoretical case for microfinance and small business credit appeared airtight. Poor entrepreneurs face credit constraints that prevent profitable investments; relaxing these constraints should unlock growth. Capital markets fail in developing economies due to information asymmetries and weak contract enforcement, creating a clear role for development finance institutions. Yet the experimental evidence has consistently disappointed.

The landmark randomized evaluations of microcredit—in India, Morocco, Bosnia, Mongolia, Ethiopia, and Mexico—produced remarkably consistent findings across diverse contexts. Expanded credit access increased borrowing and sometimes business investment, but effects on profits, revenues, and household consumption were small and often statistically insignificant. The transformative potential of credit access failed to materialize at the scale theory predicted.

Understanding why requires examining the structure of microenterprise economies. Most small businesses in poor countries operate in highly competitive, low-margin sectors with limited scope for differentiation or scale economies. Marginal returns to capital may be high at very low investment levels but decline rapidly. The profitable investment opportunities that would justify borrowing at market rates simply may not exist for most enterprises.

Additionally, credit products designed for the poor often impose costs that offset potential benefits. High interest rates, inflexible repayment schedules, and group liability requirements constrain how borrowed capital can be deployed. Entrepreneurs may use credit for consumption smoothing rather than productive investment, a perfectly rational response to their actual circumstances even if it disappoints program designers.

More recent evidence from graduation programs—comprehensive interventions combining asset transfers, training, and ongoing support—shows larger and more sustained effects than credit access alone. This suggests that capital constraints interact with other binding constraints, and addressing finance in isolation may be necessary but insufficient. The bundling insight has shifted program design toward more intensive, multi-component interventions, though at substantially higher cost per beneficiary.

Takeaway

Credit constraints are real, but relaxing them rarely produces transformative effects because most microenterprises lack the profitable investment opportunities that would make borrowed capital valuable.

Interpreting the weak evidence base for private sector development requires grappling with fundamental measurement problems. Business outcomes are notoriously difficult to measure reliably, and the challenges are compounded in informal economies where administrative records don't exist and survey responses may be strategically biased.

Self-reported profits and revenues are the standard outcome measures in enterprise evaluations, yet their reliability is questionable. Entrepreneurs may not maintain accurate records, may not understand the distinction between revenue and profit, and may have incentives to misreport. Measurement error attenuates estimated treatment effects toward zero, potentially masking real impacts. Studies using administrative data or direct observation often find larger effects than self-report surveys from similar programs.

The timing of measurement matters enormously. Business outcomes are volatile, subject to seasonal variation, demand shocks, and lifecycle effects. A snapshot measurement at any particular follow-up point may miss the true average effect. Moreover, many business investments have delayed returns—the impact of a new supplier relationship or marketing strategy may take years to fully materialize. Standard evaluation timelines of 12-24 months may be too short to capture genuine program effects.

Perhaps most fundamentally, we may be measuring the wrong outcomes. Profit and revenue growth assume that firm expansion is the pathway to poverty reduction. But successful entrepreneurship might equally manifest as business exit—entrepreneurs who gain skills and capital moving to wage employment that offers higher, more stable returns than self-employment. Treating exit as program failure when it represents economic advancement distorts our understanding of intervention effects.

The measurement challenges don't excuse weak findings—they complicate their interpretation. Better measurement might reveal larger effects, but it might equally reveal even smaller ones. What's clear is that the evidence base contains substantial noise, and confidence intervals around estimated impacts are wider than point estimates suggest. This uncertainty should inform how we weight private sector development interventions against alternatives with cleaner measurement.

Takeaway

Before concluding that business support programs don't work, we must reckon with the possibility that our measurements are too noisy, too short-term, and too narrowly focused to capture the effects that matter.

The weak evidence base for private sector development interventions doesn't necessarily mean these programs should be abandoned. It means we face a genuine epistemic problem: we cannot confidently distinguish between programs that don't work and programs whose effects we cannot reliably measure. This uncertainty should induce humility about both advocacy and criticism.

What the evidence does support is a shift away from standardized, large-scale interventions toward more targeted, intensive approaches. The entrepreneurs most likely to benefit from business support are not the poorest or most constrained but those with existing capacity and opportunity. Accepting this reality has implications for how we think about poverty reduction through enterprise development.

The broader lesson concerns the limits of private sector development as a poverty reduction strategy. If even well-designed interventions produce modest effects, perhaps the emphasis on entrepreneurship as an exit from poverty deserves reconsideration. For many poor people, wage employment or social protection may offer more reliable pathways than the uncertain returns of microenterprise.