AI And Govt Bond Yields – Onwards And Upwards?

Much of the market focus on generative Artificial Intelligence (AI) has been on potential productivity gains and equity market impact. But AI may have important implications for other markets, notably fixed income, via the level of interest rates and bond yields, which we address in this short note.

The conventional view on the impact of AI on rates, assuming faster productivity materialises from AI, and the marginal product of capital increases, is that the demand for capital increases, driving real interest rates and government bond yields higher, ceteris paribus. This would mean an increase in equilibrium real interest rates, or the Fed’s so-called R-star^[1], at which inflation is at target and the economy at full employment. Combined with a more robust, and re-capitalised financial system, this is another reason not to expect an early return to zero rates, as we have explained in previous LSEG research notes^[2].

Higher real yields and productivity were correlated in Goldilocks era

There is some evidence of upward pressure on real interest rates from higher productivity growth in Figure 1, which shows US nominal yields were around 5-6% during the faster productivity growth of the Goldilocks era, from the 1990s to 2007, but although there are parallels with Goldilocks in the post-Covid era, there are also distinct differences (see below).

Longer term studies of productivity generally show the main drivers to be (1) capital per worker, (2) technology gains, and (3) resource re-allocation. These factors drove rapid productivity gains in the period before the Global Financial Crisis (GFC), apart from in Japan, though productivity growth started to slow before the GFC in the G7, particularly in manufacturing, as the benefits from Moore’s Law^[3] and re-shoring faded. But in 2008, the GFC crushed domestic demand, financial intermediation and investment growth, ushering in secular stagnation and weak productivity growth for over a decade to 2020. Real and nominal yields adjusted quickly to zero rates, central bank QE and the weaker demand for capital once the scale of the GFC, and credit crunch became clear, and then remained at, or near historic lows until after Covid, as Figure 1 shows.

Figure 1: US Productivity Growth And 10 Yr Yields

Explanations for very weak productivity from 2007-20

The long period of very weak productivity growth, from 2007-2020, and low yields, has been explained by zero rates “zombifying” economies, pushing the cost of capital well below the marginal product of capital^[4], and distorting the allocation of capital, allowing unprofitable companies to survive. Proponents of this view appeal to Austrian and Swedish capital theory and suggest recessions and periodic bouts of “creative destruction” are required in capitalist economies to rid them of unprofitable companies, and boost productivity^[5], even if this risks deep recessions. A clear example of this phenomenon is the Japanese economy with its low productivity and GDP growth alongside zero rates since the early 1990s,^[6] and there is evidence of long productivity “tails” elsewhere in the G7 since the GFC. Figure 2 shows how US manufacturing productivity broadly flat-lined in the period since the Fed adopted zero rates in 2008, even if this does not establish causation. There are, however, very recent signs of recovery.

Figure 2: US productivity (manufacturing) and fed funds rate

AI impact on labour market: displacement versus reinstatement

Turing to the impact of AI on productivity growth and yields, the overall effect on the labour market depends on whether the displacement effect on jobs, in reducing employment, exceeds the re-instatement effect of new jobs being created, and the timing. An example from the Goldilocks era was that many displaced employees found work in new occupations like website-design. For workers in jobs partially exposed to AI, productivity may be augmented by AI. This would also increase the scalability of labour, and ease labour shortages driven by demographics and ageing labour forces.

Inter-temporal effects of AI could mean steeper yield curves…

But inter-temporal, or timing effects are important here, and there are cross-currents on interest rates. If the displacement effect of lost jobs from AI proves substantial, and unemployment initially increases, it could squeeze domestic demand and inflation lower, and force a recession, particularly if reinforcing the impact of higher short rates. That might allow inflation to fall initially, easing pressure on both nominal and real yields, even if longer term growth potential increases, putting upward pressure on rates and yields. This narrative would suggest steeper yield curves would develop in the medium term. However, flat, or inverted, yield curves continue to dominate fixed income markets, as the US Treasury curve gradient in Figure 3 shows.

Figure 3: US treasury curve gradients since 1994

…suggesting AI may drive steeper curves

Inverted, or flat, curves may be cyclical, after G7 tightening cycles, but could also reflect the view AI will drive a replay of Greenspan’s flat yield curve conundrum^[7] and low inflation rates from the Goldilocks era. Evidence of stable inflation breakevens, and the disinflation in 2023 support this view. But although there are some parallels between Goldilocks and the post-Covid eras, there are clear differences as well, notably ageing labour force demographics, much higher government debt/GDP ratios, geo-economic fragmentation and higher energy prices.

Indeed, the recent surge in demand for electricity and data centres is a consequence of the AI boom. These differences between Goldilocks and the post-Covid era are reflected in stubborn service sector inflation, which is still running at 5.2% in the US, and 5.9% in the UK, and would justify a higher term premium in longer maturities.

In summary, onwards and upwards for bond yields?

In conclusion, there are a number of structural crosscurrents in the global impact of AI on bond yields and curves, most notably improved productivity, and some initial displacement of labour, on the one hand versus higher equilibrium rates, or the Fed’s R*, on the other. Caution is also required in drawing inferences from the Goldilocks era, particularly about the shape of the yield curve, when the case for steeper curves is strengthened by higher debt/GDP ratios, and higher energy prices.

¹See ” Measuring the natural rate of interest” “, New York Federal Reserve,

²See ” The shape of things to come” “, LSEG, FTSE Russell, August 2023.

³Moore’s Law states that the number of transistors in a micro-chip doubles within 2 years, and broadly implies computers become more powerful and efficient, driving down costs and increasing output.

⁴See ” Selected papers on economic theory by Knut Wicksell”, ” pp.67-92, Erik Lindahl, 1958.

⁵See , ” Capitalism, Socialism and Democracy”, Joseph Schumpeter, 1942.

⁶See ” Productivity Trends in Japan – Reviewing Recent Facts and the Prospects for the Post-COVID-19 Era” ” – T Yagi, K Furukawa, and J Nakajima, Bank of Japan, July 2022.

⁷Alan Greenspan, Federal Reserve, Semi-annual US Congressional testimony, February 2005

Disclaimer

© 2024 London Stock Exchange Group plc and its applicable group undertakings (“LSEG”). LSEG includes (1) FTSE International Limited (“FTSE”), (2) Frank Russell Company (“Russell”), (3) FTSE Global Debt Capital Markets Inc. and FTSE Global Debt Capital Markets Limited (together, “FTSE Canada”), (4) FTSE Fixed Income Europe Limited (“FTSE FI Europe”), (5) FTSE Fixed Income LLC (“FTSE FI”), (6) FTSE (Beijing) Consulting Limited (“WOFE”) (7) Refinitiv Benchmark Services (UK) Limited (“RBSL”), (8) Refinitiv Limited (“RL”) and (9) Beyond Ratings S.A.S. (“BR”). All rights reserved.

FTSE Russell® is a trading name of FTSE, Russell, FTSE Canada, FTSE FI, FTSE FI Europe, WOFE, RBSL, RL, and BR. “FTSE®”, “Russell®”, “FTSE Russell®”, “FTSE4Good®”, “ICB®”, “Refinitiv” , “Beyond Ratings®”, “WMR™” , “FR™” and all other trademarks and service marks used herein (whether registered or unregistered) are trademarks and/or service marks owned or licensed by the applicable member of LSEG or their respective licensors and are owned, or used under licence, by FTSE, Russell, FTSE Canada, FTSE FI, FTSE FI Europe, WOFE, RBSL, RL or BR. FTSE International Limited is authorised and regulated by the Financial Conduct Authority as a benchmark administrator. Refinitiv Benchmark Services (UK) Limited is authorised and regulated by the Financial Conduct Authority as a benchmark administrator.

All information is provided for information purposes only. All information and data contained in this publication is obtained by LSEG, from sources believed by it to be accurate and reliable. Because of the possibility of human and mechanical inaccuracy as well as other factors, however, such information and data is provided “as is” without warranty of any kind. No member of LSEG nor their respective directors, officers, employees, partners or licensors make any claim, prediction, warranty or representation whatsoever, expressly or impliedly, either as to the accuracy, timeliness, completeness, merchantability of any information or LSEG Products, or of results to be obtained from the use of LSEG products, including but not limited to indices, rates, data and analytics, or the fitness or suitability of the LSEG products for any particular purpose to which they might be put. The user of the information assumes the entire risk of any use it may make or permit to be made of the information.

No responsibility or liability can be accepted by any member of LSEG nor their respective directors, officers, employees, partners or licensors for (A) any loss or damage in whole or in part caused by, resulting from, or relating to any inaccuracy (negligent or otherwise) or other circumstance involved in procuring, collecting, compiling, interpreting, analysing, editing, transcribing, transmitting, communicating or delivering any such information or data or from use of this document or links to this document or (B) any direct, indirect, special, consequential or incidental damages whatsoever, even if any member of LSEG is advised in advance of the possibility of such damages, resulting from the use of, or inability to use, such information.

No member of LSEG nor their respective directors, officers, employees, partners or licensors provide investment advice and nothing in this document should be taken as constituting financial or investment advice. No member of LSEG nor their respective directors, officers, employees, partners or licensors make any representation regarding the advisability of investing in any asset or whether such investment creates any legal or compliance risks for the investor. A decision to invest in any such asset should not be made in reliance on any information herein. Indices and rates cannot be invested in directly. Inclusion of an asset in an index or rate is not a recommendation to buy, sell or hold that asset nor confirmation that any particular investor may lawfully buy, sell or hold the asset or an index or rate containing the asset. The general information contained in this publication should not be acted upon without obtaining specific legal, tax, and investment advice from a licensed professional.

Past performance is no guarantee of future results. Charts and graphs are provided for illustrative purposes only. Index and/or rate returns shown may not represent the results of the actual trading of investable assets. Certain returns shown may reflect back-tested performance. All performance presented prior to the index or rate inception date is back-tested performance. Back-tested performance is not actual performance, but is hypothetical. The back-test calculations are based on the same methodology that was in effect when the index or rate was officially launched. However, back-tested data may reflect the application of the index or rate methodology with the benefit of hindsight, and the historic calculations of an index or rate may change from month to month based on revisions to the underlying economic data used in the calculation of the index or rate.

This document may contain forward-looking assessments. These are based upon a number of assumptions concerning future conditions that ultimately may prove to be inaccurate. Such forward-looking assessments are subject to risks and uncertainties and may be affected by various factors that may cause actual results to differ materially. No member of LSEG nor their licensors assume any duty to and do not undertake to update forward-looking assessments.

No part of this information may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the applicable member of LSEG. Use and distribution of LSEG data requires a licence from LSEG and/or its licensors.

Original Post

Editor’s Note: The summary bullets for this article were chosen by Seeking Alpha editors.