This paper uses an autocorrelation function (ACF) approach to develop a new testing procedure for international output convergence. We define convergence in terms of sample ACFs of detrended output per capita, and construct an inference set-up based on resampling and subsampling techniques for dependent data. Using per capita GDP for 15 OECD countries observed over a century, we find that the hypothesis of conditional convergence is unsupported; that, the USA apart, the linearized neoclassical growth model fails to replicate the transitional dynamics of OECD economies; and that these economies do not behave like a club.