Why Good Evaluations Still Disagree
Reliable evaluation does not guarantee a single answer about entrepreneurship support. This chapter diagnoses why credible studies can diverge by separating five possible causes: the policy may be misguided, delivery may change the design, evaluation may omit time or place, firms and entrepreneurs may differ, and support may create unintended effects. The practical value of this diagnosis is to replace the reflexive question “Did the programme work?” with a more useful one: which part of the support system produced the observed result, under which conditions, and for whom?
A mixed result is therefore a symptom, not a cause. It tells us that the evidence varies across measures, populations, or circumstances. It does not tell us which explanation is correct. Treating every mixed result as proof of bad policy is as weak as treating every positive result as proof of good design.
Five different causes can produce the same mixed result
The first possibility is a problem in the policy’s rationale. Government intervention may be justified when a market failure or social inequality prevents valuable entrepreneurial activity from occurring. But a broad desire to increase entrepreneurship does not, by itself, establish that a particular subsidy, training course, or accelerator addresses a real constraint. If the support is aimed at a problem that does not require public intervention, even excellent delivery may produce little additional impact.
The second possibility is delivery. A policy can change as it moves through legislation, administration, intermediaries, eligibility rules, and local implementation. The formal objective may remain unchanged while the practical service becomes easier for one population to use, more attractive to a different type of firm, or more focused on a visible short-term result. A programme that appears ineffective may therefore be a sound idea delivered in a way that does not reach the intended problem. Conversely, a programme can acquire positive effects through local judgment even when its original design was vague.
The third possibility is an incomplete evaluation. A study may use a strong comparison and still miss the time period, macroeconomic conditions, regional differences, or wider ecosystem that shape the result. A programme can look effective during a boom and weak during a recession, or produce one result in a region with dense networks and another where talent and finance are scarce. Evaluating one programme in isolation can also miss effects produced through interaction with regulation, markets, culture, or other public services.
The fourth possibility is population diversity. SMEs and entrepreneurs differ in age, sector, resources, growth intention, management capability, location, and likelihood of survival. A policy designed for a growth-oriented technology venture may be irrelevant to an owner seeking stable self-employment. A training course that helps a firm with an identified managerial problem may be generic and low-value for a firm whose main constraint is finance or demand. A group average can conceal both strong benefits for one subgroup and no benefit for another.
The fifth possibility is unintended consequence. Support can change the behavior of competitors, non-recipients, administrators, and recipients in ways that were not in the original objective. A subsidized entrant may displace another firm. A threshold may encourage a business to remain eligible rather than grow beyond it. A programme can also produce positive spillovers that its narrow evaluation fails to capture.