The RBA Board is uncertain, including about the staff’s judgements about demand and supply. But is being data-dependent any better?
Much has been made of the number of times the word ‘uncertainty’ or its variants appeared in the minutes of the RBA Board meeting in August. Perhaps more noteworthy was the increase in the number of times the word ‘staff’ appeared: 13, compared with low single-digit numbers in most of last year’s minutes. The total was higher for meetings where new staff forecasts were discussed. Outside of forecast months, though, the word appeared not even a handful of times in each of last year’s minutes and in the September 2023 minutes, not at all.
Another word being used a lot more is ‘judge’ and its variant ‘judgement’. Until May this year, the word barely featured, appearing at most two or three times in each issue of the minutes since February last year. Since May, it is recording double-digit word counts, appearing 15 times in the August minutes. Some of this is purely stylistic: one of the drafters might have decided that ‘Members judged’ was an attractive variation on ‘Members assessed’ and some of the other options favoured by drafters of the past.
Taken together, though, this shift in word choice might suggest something deeper. Particularly noteworthy was the following passage:
“In addition, the staff’s assessment was that the gap between aggregate demand and supply would be wider than previously judged throughout the forecast period, because of both a stronger outlook for demand and a reassessment of the economy’s current spare capacity. Members noted that judgements about this gap were highly uncertain.”
Indeed, several of the uses of the word ‘uncertainty’ relate to the uncertainty about the staff’s judgement about this gap, which was mentioned more than once in the minutes. One in particular made me wonder if the Board had not been fully convinced by the arguments supporting the latest revisions to the staff forecasts.
“Based on the data and evidence over a period of time, the staff had assessed that the economy had less spare capacity than previously assumed. This reflected higher inflation outcomes over the preceding 12–18 months than could be explained by previous estimates of excess demand, and signals from labour market indicators and survey measures of capacity utilisation. Members discussed the considerable uncertainty around estimates of spare capacity and the need to reassess this judgement regularly as the data evolve.”
The final sentence in that passage suggests that the Board was not willing to base its policy decision on the staff’s judgements about the output gap.
There are a number of reasons why the Board might have been so cautious in this area. Recall our previous observation (PDF 427KB) that the net surprise on inflation over the past three quarters has been small to zero. Adjusting your assessment of your current position based on your forecast errors over the past 12–18 months is a reasonable approach. But it becomes less reasonable when most of that forecast error relates to periods that were more than a year ago.
Another issue is that we cannot truly see aggregate demand and supply; we can only infer them from data. If the indicator data being used do not map linearly to the ‘true’ levels of demand and supply, then estimates of the gap could be off-base. In addition, inflation outcomes only tell you about the gap between demand and supply. Inferring whether the surprise comes from demand or supply needs other sources of information. The Board highlighted this in the minutes, noting that there were uncertainties around the persistence of supply shocks. While some other public statements from RBA officials suggest that the RBA is assuming that pandemic-related supply shocks have fully washed through, the minutes suggest that the Board does not think this is so clear-cut.
Given these difficulties, the Board has responded by continuing to place “greater-than-usual weight on the flow of data, relative to the forecasts”.
This raises the question of which data can be relied on, given the uncertainties of measurement. As we discussed last week, hours worked data might be getting a bit less weight now than they appeared to have in the compilation of the August forecast round. US policymakers have just had an experience of this with the payrolls data as well.
Some inspiration might be gleaned from the statements of peer central banks in economies that opened earlier after the pandemic, and thus where the disinflation started earlier. For example, the Bank of Canada Governing Council decided to hold rates steady at its April meeting but signalled that it was preparing to ease; it subsequently cut rates at its June and July meetings. In its announcement after the April meeting, it highlighted four key indicators of underlying inflation pressure that it was monitoring:
- the balance of supply and demand in the economy
- corporate pricing behaviour
- inflation expectations
- wage growth relative to productivity
Three of the four would be familiar to readers of RBA communication, but the emphasis on corporate pricing behaviour is less often seen here. In addition, the role of weak productivity growth is interpreted somewhat differently in Canada. While some members publicly worried in the July statement about the strength of wages growth relative to weak productivity growth, the same statement also noted: “Wage growth was expected to moderate given the presence of labour market slack and weak labour productivity”.
In other words, in an environment of weak productivity growth, employers might be less willing to agree to large wage increases, and this would tend to reduce wages growth relative to outcomes when productivity growth was stronger, and all else equal. This potential direction of causation does not appear in RBA communication, even though wages growth has already peaked in Australia. Nor does the possibility that high labour costs might induce firms to invest more in labour-saving technologies or increase productivity in other ways.
Rather, the RBA’s analysis seems to embed a linear mapping from hours worked to productivity to unit labour costs to inflation. Arithmetically, that is how it works. In an environment of uncertainty, though, it is risky to assume that a surprise in one component (such as hours worked) maps fully to a revision in the next quantity in the chain (such as productivity), holding no implications for related variables (such as GDP). A similar example in the RBA’s August minutes is the comment that “persistence of cost pressures was a key theme reported by firms in liaison discussions”. This is no doubt true, but the implications for the next link in the chain – business pricing decisions – are less clear.
One of the risks about being data-dependent is that, in the end, there is still a model behind your judgements, even if only a mental one. In the Australian context, that model seems to have a flavour of cost-plus pricing to it. Until the staff of the RBA start to judge that some labour market slack has emerged, the RBA will likely continue to interpret the data flow from a perspective of believing that demand is still strong, and set policy accordingly.
