African exporters combining direct and indirect modes innovate more than single-mode exporters

Firms that simultaneously pursue direct and indirect exporting show stronger innovation performance than firms using either mode alone, with the relationship significant at the 1% level for both any innovation and high innovation. Prior work focused largely on the dichotomous choice between direct and indirect exporting; the authors state they are not aware of prior theorizing or evidence on the combined strategy's impact on innovation, so this brings a neglected strategy into learning-via-exporting theory. Based on 4,254 African exporters in the World Bank Enterprise Surveys, estimated with a generalized ordered logit; the combined-strategy coefficient is 0.520 and 0.560, both significant at 1%, and marginal effects show a 0.130 drop in the probability of no innovation and a 0.088 rise in the probability of both innovation types.

The advantage of the combined strategy is framed as complementarity and synergy between the two modes rather than a simple additive effect, because excess resources and experiential knowledge generated in one mode can be redeployed at low marginal cost in the other. The authors extend Penrose's slack-induced growth logic from domestic growth to international expansion and export-mode choice, offering a new lens for international business theory. This is a theoretical argument grounded in resource indivisibility, asset specificity, risk diversification, and the transferability of cross-cultural insights, with the empirical results presented as consistent with a synergy interpretation.

Tests of the four moderators only partly match predictions: the moderating effects of the learning environment (Global Innovation Index) and knowledge stock (R&D) take signs opposite to the hypotheses, while property-rights regime and managerial experience are significant only in some models or in the instrumental-variable estimates. The authors read this mixed pattern as reflecting coexisting synergistic and antagonistic forces, consistent with the Penrose effect on managerial limits to growth, and use it to discuss the boundaries of learning-via-exporting theory. In Table 3B the R&D-by-combined-strategy interactions are negative and significant (-0.604 and -0.425), and the Global Innovation Index interaction is significant only in model 1 with a negative sign; under two-stage residual inclusion, the property-rights and managerial-experience moderations become positive and significant while the learning-environment and knowledge-stock interactions remain oppositely signed.

Only 949 of the 4,254 African exporters combine direct and indirect exporting, fewer than a quarter, which the authors read as suggesting firms and policymakers may not recognize the learning benefits of the combined strategy. The authors state that no prior study has examined the innovation outcomes of different exporting strategies among African exporters, and note that export-promotion policies seldom distinguish among export modes. The sample comprises 2,034 direct-only, 1,271 indirect-only, and 949 combined exporters; the authors suggest policy could support firms in using indirect exporting for short-term market testing while building capabilities for direct engagement.

The study speaks to researchers and policymakers concerned with innovation and internationalization strategy among African exporters, and applies to the setting of formally registered African exporting firms covered by the World Bank Enterprise Surveys between 2011 and 2020. Its findings support treating direct and indirect export channels as complementary arrangements that can transfer resources and knowledge to each other, rather than as independent or sequential stages, and suggest that export-promotion programs focused on export volume might also consider helping firms use indirect exporting for short-term market testing while building direct-engagement capabilities. The authors also note that expansion outpacing the development of internal resources and managerial capacity could erode the learning benefits of the combined strategy.

The authors note that innovation is measured by a self-reported yes/no question about new products or processes in the past three years, whose 'significant improvement' definition is subjective and may introduce systematic bias across respondents and countries; knowledge stock is proxied by a binary R&D measure and managerial experience by years working in the sector, neither of which may capture the full construct. The authors also note that the negative R&D interaction could reflect measurement limitations or reverse causality, so they treat it as suggestive. In addition, the loaded text is an incomplete version, so some tables and appendix details are not fully available and the precise reading of moderator and robustness estimates should rest on the complete published version.