AI and Technology Impact on Growth and Inflation

 The huge buzz around artificial intelligence has also raised a debate about the growth, jobs and inflation impact of this latest phase of the 4^th industrial revolution (5G, AI, Internet of Things and Big Data)? What is the likely medium-term impact?

Figure 1: Survey of Impact on Global GDP Per Annum from AI Over the Coming Decade (% of GDP) 

Source: CfM-CEPR survey (here) 

Boosting Growth, But When 

Optimists feel that the macroeconomic effects of the AI wave will lift productivity and hence long-term growth by not only making existing workers more productive but also reducing the amount of workers undertaking mundane tasks. This will occur at a rapid pace argue some economists such as Erik Brynjolfsson (here). The pessimists (e.g. Anthony Gordon (here)) worry that AI is not really a game changing general purpose technology (like the steam engine or electricity technology) and that it will not materially shift productivity. Most academic economists sit somewhere in the middle, with a recent poll by CfM-CEPR finding 64% thought that it could lift global GDP by 0-2% per annum over the next decade. This masks a difference between DM and EM economies, given that most DM economies are more able to invest in technology given their relative balance sheet strengths. 

The impact of a technology upheaval depends on how quickly it is adopted fully and then when it starts to bring improved benefits. Generative AI has caught the imagination in 2023, but AI extends well beyond such tools and industries are looking to use AI to automate or amend processes across industries. However, the wider process operates at a less rapid pace than the frenzy around chat GPT, as companies credibility means that they need to be confident that the new processes augment current production levels. Additionally, adoption speeds for new technology also diverge across an economy and between economies, which could ensure a longer AI adoption phase. Professor Nick Crafts (here), looking at previous general purpose technology waves, notes that new technologies face a number of adoption hurdles as business reorganizes and learn to use the new technology builds. Technology improvements, plus complimentary investment and innovations, can then help to bring through the full benefit over years and sometime decades. For the macroeconomist, looking at the economic cycle, this could mean that the full benefits from AI only occur towards the end of the 2020’s and into the 2030’s.

More generally, productivity has been disappointing since the mid 2000’s, due to low business investment post GFC; population aging and the lack of major new technology innovations. U.S. non-farm productivity 2008-22 is 1.5% per annum versus over 2% 1980-2007 (Figure 2). Academic economists have gone further and highlight that though lower labor or capital productivity is part of the problem that the combination and organization of all inputs (total factor productivity) has seen the steepest slowdown (Figure 3).

Figure 2: U.S. Non-Farm Productivity (Yr/Yr %)

Source: Datastream/Continuum Economics

Figure 3: Falling Total Factor Productivity More Important than Labor or Capital Productivity (%) 

Source: ECB (here)

Technology is also one positive driver of productivity in the 2020’s alongside increased climate change investment, but facing headwinds from the post COVID/Ukraine war revision of supply chains and labor market mismatches (skill shortages, illness/health and population aging). We would still forecast benefits from technology for GDP growth and productivity in the coming years, though the overall impact will likely be small to modest per annum. Firstly, AI has been part of the 4^th industrial revolution since the mid 2010’s where 5G enables AI to harness Big Data and the Internet of Things (IOT) allows widespread connectivity to devices to help spread the benefits. Some of this wave of benefit will come through the middle of this decade. Secondly, the COVID crisis prompted an openness to working from home that has not been fully reversed and shows that the labor force can use technology to work more flexibly. This could also help flexibility and working processes and hence total factor productivity. Academic debate rages on the narrow issue of how working from home impacts productivity (here), but the wider view is that hybrid working can help to lift productivity. Additionally, more flexible working from home could lift female participation and over 55 working in DM countries with aging population, which boosts labor inputs and GDP growth – while the permanent shift to more online retailing can also boost productivity (here).

One issue our economists will also watch is whether lost jobs due to AI/technology (displacement) arrive earlier than new (reinstatement) jobs in AI/technology fields. The better that government retraining programs are, the less risk of a gap exists for this key issue – though unfortunately most DM economies are not focused on this transitional retraining issue. While this is largely a microeconomic issue, it could led to some adverse macroeconomic consequences if it increases long-term unemployment and provides a drag to consumption and also accentuates fiscal problems.

Lower Inflation, But 

The flip side to higher productivity for GDP from technology innovation has been a disinflationary influence. For example the ECB estimates that information and communications technology (ICT) has knocked -0.15% per annum off EZ inflation over the last 20 years at a time when EZ inflation averaged 1.7% per annum (here). Increased productivity led to a direct effect of lower inflation. However, it has also improved the quality of these goods and services, which academic economists argue should also mean an additional quality adjustment is made to inflation. One estimate in the U.S. (Figure 4) is that this would mean average PCE inflation 0.4% per annum lower at 1.1% per annum (2008-18) due to consumer digital access services.

Figure 4: U.S. Inflation Lower Adjusted for Improved Quality of IT Services 

Source: ECB (here)

The disinflationary benefit from technology adoption should remain an influence in the coming years, as the 5G/IOT/Big Data and then AI wave’s helps productivity and GDP growth. However, if the full benefit of the AI adoption does not kick in until the late 2020’s, then a stronger disinflation force may not be seen until a later date as well.

The influence on inflation is not just a function of productivity however, but also reflects the stage of the development cycle for applications of technology. In the early stage the desire to drive adoption can lead to super competitive behavior that benefits consumers. However, once dominant players become established and barriers to entry become higher, monopolistic/oligopolistic behavior can see price increases that outstrip underlying cost trend for a technology. Perhaps this is one of the reasons why the cost of internet advertising has gone up in recent years (Figure 5), after sharp falls from 2009-19. This has prompted calls for greater regulation/taxation by government for mega technology companies to ensure that the benefits of the technology are accrued widely across an economy rather than the owners of the technology.

Figure 5: U.S. PPI Internet Advertising Sales (Dec 2009 =100)

Source: Datastream/Continuum Economics