Which one should be used when determining which sector contributes most to tracking error? One question I did compared the contribution to effective duration vs. the benchmark and one used the contribution to spread duration vs. the benchmark? Or does it depend on the question?