It is shown how most methods of measuring inflation and growth have an unde
rlying spanning tree. The spanning tree whose resulting inflation (growth)
estimates are least sensitive to the choice of index number formula can he
computed using Kruskal's minimum spanning tree algorithm. Applying this alg
orithm to American, British, and Australian data sets, chaining is shown to
be the best possible way of linking annual data. For quarterly data, the o
ptimal method of linking depends on the amount of seasonality in the data.