This commentary reviews and replicates Cheng, Ioannou, and Serafeim (2014) (CIS), a widely cited study claiming that corporate social responsibility (CSR) improves access to finance. The author identifies key methodological flaws in CIS, replicates its analysis, and uses alternative data to test the original hypothesis.

Key findings show that while a cross-sectional association exists between CSR and access to finance, there is no evidence that improvements in CSR lead to increased access within firms over time.

Review of the method used in Cheng, Ioannou, and Serafeim (2014)

CIS hypothesised that CSR affects access to finance both directly—by reducing information asymmetry—and indirectly—by enhancing profitability, which then lowers capital constraints. To test this, CIS replaced direct measures of access to finance with predicted values derived from regression-based indexes (KZ, WW, and SA), constructed from prior studies.

Unlike traditional methods that measure actual financial constraints using text or manager assessments, CIS used estimates based on accounting variables. This approach, the commentary argues, creates a tautological relationship where the outcome is mathematically determined by inputs, eliminating unexplained variance and limiting empirical inference.

Limitations to inference

The report demonstrates that CIS’s methodology prevents it from identifying either direct or indirect effects of CSR on access to finance. The commentary uses causal diagrams to illustrate that when predicted values are used as outcomes, they can only reflect known relationships and exclude any new causal mechanisms.

For example, in the SA index, size and age are the only predictors. Any claim that CSR affects access to finance via this index requires the implausible assumption that CSR significantly alters firm age or asset size. In all cases, the regression results reflect model assumptions rather than empirical discovery.

The commentary also shows that differences in data sources—CIS used international data from 2002–2009 while the indexes were built on U.S. firms from earlier periods—require assuming consistency in CSR effects across time and geographies. Literature suggests this assumption is not supported.

Replication and rectification

Replication was performed using a similar sample and the same methods as CIS. The replication reproduced CIS’s findings in pooled models (cross-sectional) but not in fixed effects models, suggesting unobserved firm characteristics influence both CSR and access to finance. For instance, the coefficient for the CSR Index predicting the KZ Index in pooled analysis was –0.754 (p < 0.01), but became statistically insignificant in the fixed effects model.

To rectify the limitations of the original study, the author used Hoberg and Maksimovic’s text-based measures of financial constraint derived from 10-K filings, including DelayConstraint, EquityDelay, DebtDelay, and PrivateDelay. These are direct, qualitative assessments of management concerns about financing.

In both the original time frame (2002–2009) and an extended panel (2002–2017), the text-based analysis found some significant cross-sectional correlations (e.g., CSR Index coefficient for Equity-Focused Delay was –0.028, p < 0.01). However, none of the firm fixed effects models produced significant results, indicating no evidence that within-firm CSR improvements lead to improved access to finance.

The likely explanation is that firm-level traits such as management quality, governance, or long-term orientation influence both CSR and access to finance, creating spurious cross-sectional correlations.

Conclusion

The commentary concludes that the methodology used in CIS cannot support its claims. When more appropriate outcome measures are applied, the hypothesised causal relationship between CSR and improved access to finance is not supported.

The analysis reinforces findings from earlier studies that cross-sectional associations do not imply causal effects. The commentary recommends researchers avoid using predicted indexes as outcome variables and instead employ direct measures, guided by causal diagrams, to evaluate CSR’s financial impacts.