Pré-publication, Document de travail: Since the global financial crisis of 2008, a strand of the literature has documented a threshold beyond which financial development tends to affect growth adversely. The evidence, however, rests heavily on internal instrument identification strategies, whose reliability has received surprisingly little attention so far in the finance-growth literature. Therefore, the present paper conducts a reappraisal of the non-linear conclusion twofold. First, in light of new data, second, by a thorough assessment of the identification strategy. Evidence points out that a series of unaddressed issues affecting the system-gmm setup results in spurious threshold regressions and overfitting of outliers. Simple cross-country analysis still suggests a positive association for low levels of private credit. However, adequately accounting for country heterogeneity, along with a more contained use of instruments, points to an overall damaging influence of financial development on economic growth. This association is stronger for more recent periods.