Reply to: Controlling for non-independence of nations should not be the default choice in cross-cultural research

It is widely acknowledged that failing to account for non-independence in data can bias statistical estimates and that national-level data often exhibit such non-independence. In our recent paper, we show that, despite this awareness, most analyses of cross-national variation in psychological values or economic indicators do not make any attempt to control for non-independence, and for those that do, most methods deployed do little to reduce bias in simulated datasets with non-independence due to proximity or shared cultural ancestry. We further show that reanalysing a small sample of published datasets using the best performing methods from our simulations can appreciably change parameter estimates. In a commentary on our paper, Akaliyski highlights potential issues with our study design, distinguishes between different sources of confounding that can arise from non-independence, and proposes causal models where controlling for non-independence might actually bias, rather than help, inference. For these reasons, Akaliyski argues that controlling for non-independence should not be the default analytic choice in cross-national research. We appreciate this commentary on our work. It is useful to discuss these methodological issues in more detail to improve the rigour of cross-national research. That said, we feel that the commentary misinterprets our argument in places and overstates potential concerns.