Two series, two different stories

On the Australian Bureau of Statistics monthly indicator, all-groups consumer price growth eased to 3.8 percent over the year to June, from 4.0 percent in May. The monthly movement was a fall of 0.1 percent in both original and seasonally adjusted terms. The trimmed mean stayed at 3.6 percent and did not move at all.[1]

The divergence between these two series is not accidental. The trimmed mean is calculated by cutting away the most volatile items at either tail of the distribution. If the items dragging the headline down sit precisely in those trimmed tails, the fall in the headline says no more about the underlying process than the trimmed mean does. That is the situation in the June data.[1]

Who is carrying the bill

Housing was the largest contributor to the annual figure at 6.8 percent, above May's 6.5 percent. Within the group, residential electricity rose 22.4 percent over the year, new dwellings 5.8 percent and rents 3.6 percent. Food and non-alcoholic beverages held at 3.3 percent. Services prices rose 4.0 percent and goods prices 3.5 percent.[1]

Electricity, rent and the cost of building a dwelling are the hardest items in a household budget to substitute away from. In a composition where all three are accelerating, the increase is absorbed by cutting other spending or by borrowing, not by deferring the purchase. The channel a policy rate works through is built on deferral. A price increase with this composition therefore passes through the point where the rate has least grip.[1]

The rival explanation and a testable threshold

This reading has a rival, and not a weak one. Services at 4.0 percent sit above goods at 3.5 percent. Services prices are more sensitive to wage costs than goods prices are, and that gap is also consistent with a cost shock beginning to turn into a wage process. A single monthly release cannot separate the administered-price channel from the wage channel; that requires the wage and unit labour cost series.[1]

The threshold can be set as follows. If the trimmed mean stays at or above 3.5 percent through the September release while the housing group runs above 6 percent, the fall in the headline is a composition effect and shows no change in the underlying process. If the trimmed mean instead drops below 3.5 percent over the same period, this reading weakens and the decline must be accepted as extending beyond the trimmed tails.[1]