Macroeconomics tends to treat the economy as a single thing. But most shocks and every new technology skews in some way towards particular industries or activities.
One of the limitations of macroeconomics is that we tend to assume that everything is the same across the whole economy – it’s a reflection of the ‘macro’ in the name. We talk about a single inflation rate, even though the CPI is comprised of the prices of a wide array of goods and services. We talk about a single unemployment rate, even though different people will experience that unemployment rate differently, depending on whether they are one of the unemployed. We treat each economy-wide phenomenon as a ‘thing’ and start our analysis from the models we learned at university that assume some amorphous ‘representative agent’.
Better data availability has improved the situation and encouraged a more sectoral approach. And there are newer models that allow some variation between people or firms, or ‘heterogeneous agents’ in the economics jargon. But too often, people assume that company-level or industry-level shocks wash out in the overall outcome, even though the conditions for that to be true do not actually hold. Most of the time, this simplifying assumption doesn’t matter – until it does.
As we have previously argued, when it comes to setting monetary policy, it is appropriate to focus on overall inflation and not pick and choose the bits of inflation you will focus on. But that is not the same as pretending that the divergences in the total do not exist. It is all very well to be alert to sectoral or regional differences. But if the models of inflation used to forecast implicitly assume a single-good economy, there is a risk your forecasts will go astray.
This issue is particularly salient because the shocks that have hit the Australian economy over the past quarter-century have, in essence, been sectoral shocks. First was the dot-com boom and bust, which rather obviously hit the tech sector hardest, though the longer-run boost to productivity was more widespread. The GFC was a crisis in the finance sector. The mining investment boom and bust was even more consequential for the Australian economy than the dot-com boom, and even more skewed to a few sectors and regions.
The pandemic was the quintessential sectoral shock. In-person services were shut down to reduce the spread of COVID-19. Demand for certain goods spiked as people set themselves up to work from home, at the same time as supply chains for these goods were disrupted by pandemic-related restrictions. Goods inflation and services inflation had completely different trajectories – though to be far, that is true most of the time.
One could reasonably object that in a well-functioning market, labour and other resources will move and we can treat the economy as a single goods market, or the labour market as one market. But in the real world, it takes time – and changes in relative prices – to get resources to shift.
In fact, a key lesson from the mining investment boom and bust was that, when a sector that had previously dominated growth and squeezed out all the other sectors eventually turns around and retrenches, it is hard for the rest of the economy to bounce back quickly and fill the gap. The other sectors just do not adjust as fast as the models assume, especially if that means people moving states as well as jobs.
This is why we are a little nervous about what happens when the incredible ramp-up in the care economy finally matures. The non-market sector (health & social care, education and public administration & defence) accounted for about 85% of all the jobs growth over 2023 and 2024 (we will have Q4 data for 2024 next week to confirm). This was despite this sector accounting for just 27% of total hours worked. It is probably a large part of the reason why the vacancy rate and business survey measures of labour market tightness remain elevated. While the market sector is not increasing employment much, firms in that sector are still having to replace workers who have left to take care economy jobs, and they are finding that difficult.
This outsized growth will eventually end, as we have previously flagged. Will the market sector bounce back quickly enough to fill the gap? Currently our forecasts assume that the handover is reasonably smooth and the overall labour market weakens only modestly. But we are mindful that a shaky handover between public demand and private demand, between non-market and market sector employment, is a risk scenario.
The expansion in the care economy has also led to some discussion about the implications for measured productivity growth. (Though possibly not actual productivity growth, as anyone who has benefited from keyhole surgery or moonboots can attest.) But there is another layer to the issue of productivity and sectoral differences.
Productivity growth is not an economy-wide phenomenon. It is the product of many individual decisions about how we use our time (recall that labour productivity is just ‘Stuff’ Divided by Time). In particular, it is partly the result of decisions about whether and how to innovate – in particular by adopting new technologies. Productivity growth is not some trend that is bestowed from the heavens – it is kicked along by waves of new technologies.
The point is that every new technology – even the so-called ‘General Purpose Technologies’ – is tuned to different activities and thus industries. Steam power and electricity had general application, but they were more transformative to energy-intensive activities in manufacturing. Computers and the internet revolutionised information-intensive activities like finance far more than, say, the hospitality industry.
So it will also be for the AI revolution. Generative AI and related innovations leverage our ability to produce content or transform information. In the real ‘meatspace’ world of physical action, it has less to offer (though probably not nothing, to the extent that improved robotics are also part of the story). This means the impact will be greater, and faster, in some industries than others.
In particular, that very ‘meatspace’ physical industry of construction is likely to be less impacted. AI could improve scheduling and procurement for projects, and could simplify and speed up regulatory processes (something Vancouver is already doing). But there is a big difference between generating a realistic picture of a bricklayer and actually laying bricks.
As noted above, we do not regard the overall trend in (non-mining) market sector productivity growth as particularly alarming. But again, this is an area where it pays to dig below the surface total and examine the detail. In the construction industry, the story is far less benign. The
Productivity Commission has recently released a report showing the long-run decline in construction productivity. The building part of the sector has seen an outright trend decline in the level of productivity for the past two decades. This is an international phenomenon, not Australia-specific. The volume and nature of regulation matters here, but so does the structure, culture and composition of the industry.
Even with population growth now rolling over, the imperative to house a growing population becomes all-the-more acute. Rising productivity will be a key enabler, and that means leveraging technological innovation. The problem is that the innovations we are currently in early-stage adoption of might not be the ones best tuned to that task.
