What Credit Evidence Can—and Cannot—Tell an Allocator
Formal credit and SME performance are the subject of this series; the final question is what an allocator can responsibly infer when the average effect is positive but firms respond differently. The evidence supports a clear judgment, though not a simple prescription. Formal loans can improve performance on average, and a larger enterprise loan can produce materially different responses across eligible firms. But a mean is not a ranking, a ranking is not an explanation, and neither is an optimal policy until it survives the context in which the lender intends to use it.
One apparent contradiction, two valid summaries
The Inter-American Development Bank review reports that formal loans raise employment by about 12% on average and sales and profits by about 18% on average across its assembled impact-evaluation evidence. The Egyptian experiment reports that relatively large enterprise loans generate small average impacts. Read quickly, those claims can seem inconsistent. They are not measuring the same object.
The review is a cross-setting synthesis focused especially on developing countries and Latin America, with evidence from advanced economies used to sharpen mechanisms and external validity. Its means summarize outcomes across different programs, populations, settings, and outcome horizons. The Egyptian study is a randomized comparison of loan size among approved applicants in the Tamouh program at the Alexandria Business Association. Its average belongs to a particular lending margin: the change from a smaller comparison loan to a relatively large enterprise loan for applicants who entered that institution’s approval process.
The difference in scale and scope matters, but it is not the whole explanation. A population mean can be positive when responses within the population point in different directions. The review’s approximate percentages say that, across the evidence it assembles, the net movement in employment, sales, and profits is economically meaningful. They do not say that each borrower experiences the same movement. The Egyptian experiment makes the hidden distribution more visible: its small mean is compatible with large positive responses for a predicted high-benefit group and much weaker or negative responses for another group.
The two evidence levels therefore answer complementary questions. The review asks whether formal credit tends to improve SME performance under the range of conditions represented in the evidence. The experiment asks whether the size of an enterprise loan changes performance differently for different eligible firms inside one program. The first establishes why finance can matter at population scale. The second shows why the population average may be too coarse for deciding who should receive additional capacity.
The units should not be mixed. An approximate percentage mean from a cross-study review is not directly comparable with a profit effect reported in EGP for groups within one experiment. Nor should a small average in one program be used to reverse the review’s broader conclusion. Each result is meaningful inside its own treatment, comparator, outcome, population, and horizon. The synthesis comes from the relationship between them: a positive average can coexist with allocation-relevant variation.
The same logic explains why the title is not an argument against lending. If additional capital relaxes a binding financial constraint, a firm may be able to pursue a productive opportunity that was previously unavailable. If the larger obligation instead meets a weak opportunity or encourages decisions whose risks are not covered by returns, the same loan can generate little benefit or a downside. When those responses are combined, the mean records the net result while concealing the route by which it was produced.
The review adds a second qualification.