+
Clear hypotheses; understandable OLS model progression; good discussion of GDP, IXP count, electricity price, and latency; recognizes the training/test performance gap; notes missing temperature, unobserved policy variables, and extreme hubs; Random Forest feature importance supports the main finding.