Category: Artificial Intelligence

Source: European Central Bank

Keynote speech by Isabel Schnabel, Member of the Executive Board of the ECB, at the Bank of England’s 2025 BEAR Conference

London, 25 February 2025

Over the past few years, global bond investors have fundamentally reappraised the expected future course of monetary policy.

Even as inflation has receded and policy restriction has been dialled back, current market prices suggest that maintaining price stability will require higher real interest rates in the future than before the pandemic.

In my remarks today, I will argue that the shift in market expectations about the level of r* – the rate to which the economy is expected to converge in the long run once current shocks have run their course – is consistent with two sets of observations.

The first is that the era during which risks to inflation have persistently been to the downside is likely to have come to an end.

Growing geopolitical fragmentation, climate change and labour scarcity pose measurable upside risks to inflation over the medium to long term. This is especially true as the recent inflation surge may have permanently scarred consumers’ inflation expectations and may have lowered the bar for firms to pass through adverse cost-push shocks to consumer prices.

The second observation is that we are transitioning from a global “savings glut” towards a global “bond glut”.

Persistently large fiscal deficits and central bank balance sheet normalisation are gradually reducing the safety and liquidity premia that investors have long been willing to pay to hold scarce government bonds. The fall in the “convenience yield”, in turn, reverses a key factor that had contributed to the decline in real long-term interest rates, and hence r*, during the 2010s.

The implications for monetary policy are threefold.

First, a higher r* calls for careful monitoring of when monetary policy ceases to be restrictive. Second, central bank balance sheet policies may themselves affect the level of r* through the convenience yield, making them potentially less effective than previously thought. Third, because central bank reserves also offer convenience services to banks, it is optimal to provide reserves elastically on demand as quantitative tightening reduces excess liquidity.

Upward shift in r* signals lasting change in the inflation regime

Starting in 2021, long-term government bond yields rose measurably across advanced economies. Today, the ten-year yield of a German government bond is about two and a half percentage points higher than in late 2021 (Slide 2, left-hand side).

What is remarkable about the rise in nominal bond yields in the euro area over this period is that it was not driven by a change in inflation compensation. Investors’ views about future inflation prospects are broadly the same today as they were three years ago (Slide 2, right-hand side).

Rather, nominal interest rates rose because real interest rates increased. Euro area real long-term rates are now trading at a level that is substantially higher than the level prevailing during most of the post-2008 global financial crisis period (Slide 3, left-hand side).

Part of the rise in real long-term interest rates is a mechanical response to the tightening of monetary policy.

Long-term interest rates are an average of expected short-term interest rates over the lifetime of the bond, plus a term premium. So, when we raised our key policy rates in response to the surge in inflation, the average real rate expected to prevail over the next ten years increased.^[1]

What is more striking, however, is that investors also fundamentally revised the real short-term rate expected to prevail once inflation has sustainably returned to our target. This rate is typically taken as a proxy for the natural rate of interest, or r*.

The real one-year rate expected in four years (1y4y), for example, is now at the highest level since the sovereign debt crisis (Slide 3, right-hand side). Even at very distant horizons, such as in nine years, the expected real short-term rate (1y9y) has increased measurably in recent years.

To a significant extent, these developments reflect a genuine reappraisal of the real equilibrium interest rate that is consistent with our 2% inflation target. A rise in the term premium, which is the excess return investors demand for the uncertainty surrounding the future interest rate path, can explain less than half of the change in the real 1y4y rate.^[2]

These forward rates have also remained surprisingly stable since 2023, with a standard deviation of around just 15 basis points, despite the measurable decline in inflation, the protracted weakness in aggregate demand and the series of structural headwinds facing the euro area.

We are seeing a similar upward shift in model-based estimates of r*. According to estimates by ECB economists, the natural rate of interest in the euro area has increased appreciably over the past two years, and even more so than what market-based real forward rates would suggest (Slide 4).^[3]

This result is robust across many models and even holds when accounting for the significant uncertainty surrounding these estimates. In other words, for drawing conclusions about the directional change of r* from the rise in market and model-based measures, the actual rate level is largely irrelevant.

What matters is the direction of travel. And that is unambiguous: we are unlikely to return to the pre-pandemic macroeconomic environment in which central banks had to bring real rates into deeply negative territory to deliver on their price stability mandate. This suggests that the nature of the inflation process is likely to have changed lastingly.

Real interest rates are only loosely tied to trend growth

Why do markets expect such a trend reversal for real interest rates in the euro area?

One answer is that some of the forces that weighed on inflation during the 2010s are now reversing.

Globalisation is a case in point. The integration of China and other emerging market economies into the global production network and the broad-based decline in tariff and non-tariff barriers were important factors reducing price pressures in advanced economies over several decades.^[4]

Today, protectionist policies, the weaponisation of critical raw materials and geopolitical fragmentation are increasingly dismantling the foundations on which trade improved the welfare of consumers worldwide.

These forces can be expected to have first-order effects on inflation.

European gas prices, for example, are up by 65% compared with a year ago despite the significant decline over recent days. Oil prices, too, have increased since September of last year, in part reflecting the marked depreciation of the euro.

While commodity prices are inherently volatile, and may reverse quickly, other deglobalisation factors, such as reshoring and the lengthening of supply chains, are likely to increase price pressures more lastingly.

And yet, the persistent rise in real forward rates poses a conundrum in the euro area.

The reason is that increases in long-term real interest rates are typically thought of as being associated with improvements on the supply side of the economy, such as productivity growth, the labour force and the capital stock.

At present, however, these factors do not point towards an increase in r* in the euro area.

Potential growth has generally been revised lower, not higher, as many of the factors currently holding back consumption and especially investment are likely to be structural in nature, such as a rapidly ageing population and deteriorating competitiveness.

The weak link between the structural factors driving potential growth and r* is, however, not exceptional from a historical perspective.

Indeed, over time there has been little evidence of a stable relationship between real interest rates and drivers of potential growth, such as demographics and productivity.^[5] They have had the expected relationship in some subsamples but not in others.^[6]

Similarly, in the most popular framework for estimating r*, the seminal model by Laubach and Williams, potential growth has played an increasingly subordinated role in explaining why the natural rate of interest has remained at a depressed level in the United States following the global financial crisis (Slide 5, left-hand side).^[7]

Rather, the persistence in the decline in r* is explained to a large extent by a residual factor, which lacks economic interpretation.

Moreover, if growth was the main driver of r*, then one would expect all real rates in the economy to adjust in a similar way. But while real rates on safe assets have declined since the early 1990s, the return on private capital has remained relatively constant.^[8]

Decline in the convenience yield is pushing r* up

A growing body of research attempts to reconcile these puzzles. Many studies attribute a significant role to the money-like convenience services that safe and liquid assets, such as government bonds, provide to market participants.

The yield that investors are willing to forgo in equilibrium for these services is what economists call the “convenience yield”.^[9]

This yield, in turn, critically depends on the net supply of safe assets: When these are scarce, investors are willing to pay a premium to hold them, depressing the real equilibrium rate of interest. And when they are abundant, the premium falls, putting upward pressure on r*.

New research by economists at the Board of Governors of the Federal Reserve System shows how incorporating the convenience yield into the Laubach and Williams framework significantly improves the explanatory power of the model.^[10]

In fact, the convenience yield can explain most of the residual factor and is estimated to have caused a large part of the secular decline in the real natural rate in the United States (Slide 5, right-hand side).

Liquidity requirements that regulators imposed on banks in the wake of the global financial crisis, the Federal Reserve’s balance sheet policies and the integration of many large emerging market economies into the global economy have led to an unprecedented increase in the demand for safe and liquid assets, driving up their convenience yield.^[11]

These findings are in line with earlier research showing that the convenience yield has played an equally important role in depressing the real equilibrium rate in many other advanced economies, including the euro area, during the 2010s.^[12]

This process is now reversing. According to the work by the Federal Reserve economists, r* has recently increased visibly, contrary to what the model without a convenience yield would suggest.

Asset swap spreads are a good indicator of the convenience yield. Both interest rate swaps and government bonds are essentially risk-free assets, so they should in principle yield the same return.

For a long time, this has been the case: before the start of quantitative easing (QE) in the euro area in 2015, the spread between a ten-year German Bund and a swap of equivalent maturity was close to zero on average (Slide 6, left-hand side).

Over time, however, with the start of QE and the parallel fiscal consolidation by governments reducing the net supply of government bonds in the market, the premium that investors were willing to pay to secure their convenience services rose measurably. At the peak, ten-year Bunds were trading nearly 80 basis points below swap rates.

But since about mid-2022 the asset swap spread has persistently narrowed. In October of last year it turned positive for the first time in ten years, and it now stands close to the pre-QE average again.

Other measures of the convenience yield paint a similar picture. The spread between ten-year bonds issued by the Kreditanstalt für Wiederaufbau (KfW) and German Bunds has narrowed from about
-80 basis points in October 2022 to just -30 basis points today (Slide 6, right-hand side).^[13]

Furthermore, in the repo market, we have observed a steady and measurable rise in overnight rates and a convergence across collateral classes (Slide 7, left-hand side).^[14]

Over the past few years, transactions secured by German government collateral, in particular, were trading at a significant premium over others. This premium has declined considerably, reflecting a reduction in collateral scarcity.

Finally, in the United States, the spread between AAA corporate bonds and US Treasuries has declined from almost 100 basis points in 2022 to 40 basis points today (Slide 7, right-hand side). It currently stands close to its historical low.

Global savings glut has turned into a global bond glut

All this suggests that, today, market participants value the liquidity and safety services of government bonds less than they did in the past, as the net supply of government bonds has increased and continues to increase at a notable pace.

In Germany and the United States, for example, the sovereign bond free float as a share of the outstanding volume has increased by more than ten percentage points over the past three years (Slide 8, left-hand side). It is projected to steadily increase further in the coming years.

So, the global savings glut appears to have turned into a global bond glut, which reduces the marginal benefit of holding government bonds.

There are several factors contributing to the rise in the bond free float.^[15]

First, and most importantly, net borrowing by governments remains substantial. The public deficit is estimated to have been around 5% of GDP across advanced economies last year, and it is expected to decline only marginally in the coming years (Slide 8, right-hand side).

Second, rising geopolitical fragmentation is likely to be contributing to a drop in demand for government bonds in some parts of the world.

In the United States, for example, there has been a marked decline in the share of foreign official holdings of US Treasury securities since the global financial crisis (Slide 9, left-hand side). It is now at its lowest level in more than 20 years.^[16] The US Administration’s attempt to reduce the current account deficit is bound to further depress foreign holdings of US Treasuries.

Third, central banks are in the process of normalising their balance sheets (Slide 9, right-hand side). Unlike when central banks announced large-scale asset purchases, the effects of quantitative tightening (QT) on yields are likely to materialise only over time, as many central banks take a gradual approach when reducing the size of their balance sheets.

Higher r* calls for cautious approach to rate easing

These developments have three important implications for monetary policy.

One is that central banks are dialling back policy restriction in an environment in which structural factors are putting upward pressure on the real equilibrium rate. Recent analysis by the International Monetary Fund (IMF), for example, suggests that a fall in the convenience yield to pre-2000 average levels could raise natural rates by about 70 basis points.^[17]

While a significant part of these effects may have already materialised, other factors could push real rates up further over the medium term. The IMF projects that, in the coming years, overall global investment – public and private – will reach the highest share of GDP since the 1980s, also reflecting borrowing needs associated with the digital and green transitions as well as defence spending.

Recent global initiatives aimed at boosting the development and use of artificial intelligence underscore these projections. Overall, these forces may well be larger than those that continue to weigh on the real equilibrium rate, such as an ageing population.

Central banks, therefore, need to proceed cautiously. We do not fully understand how the pervasive changes to our economies are affecting the steady state, or what the path to the new steady state will look like.

In this environment, the most appropriate way to conduct monetary policy is to look at the incoming data to assess how fast, and to what extent, changes to our key policy rates are being transmitted to the economy.

For the euro area, this assessment suggests that, over the past year, the degree of policy restraint has declined appreciably – to a point where we can no longer say with confidence that our policy is restrictive.

According to the most recent bank lending survey, for example, 90% of banks say that the general level of interest rates has no impact on the demand for corporate loans, with 8% saying that it contributes to boosting credit demand (Slide 10, left-hand side). This is a marked shift from a year ago when a third of all banks reported that interest rates were weighing on credit demand.

For mortgages, the evidence is even more striking. Today almost half of the banks report that the level of interest rates supports loan demand, while a year ago more than 40% said the opposite. As a result, a net 42% of banks report an increase in the demand for mortgages, close to the historical high.

Survey evidence is gradually showing up in actual lending data. Credit to firms expanded by 1.5% in December, the highest rate in a year and a half, and credit to households for house purchases grew by 1.1% (Slide 10, right-hand side).

Strong bank balance sheets are contributing to the recovery and, given the lags in policy transmission, further easing is still in the pipeline.

Lending conditions are also relatively favourable from the perspective of borrowers. The spread between the composite cost of borrowing for households and sovereign bond yields is well below the level seen over most of the 2010s and is now close to the historical average (Slide 11).^[18]

And while some maturing loans from the period of very low interest rates will still need to be refinanced at higher rates, over time this debt has declined in real terms and interest payments as a fraction of net income are buffered by rising nominal wages.

Overall, therefore, it is becoming increasingly unlikely that current financing conditions are materially holding back consumption and investment. The fact that growth remains subdued cannot and should not be taken as evidence that policy is restrictive.

As the ECB’s most recent corporate telephone survey suggests, the continued weakness in manufacturing is increasingly viewed by firms as structural, reflecting a combination of high energy and labour costs, an overly inhibitive and uncertain regulatory environment and increased import competition, especially from China.^[19]

Such structural headwinds reduce the economy’s sensitivity to changes in monetary policy.

QE’s impact on r* is reducing its effectiveness

The second implication from the impact of the convenience yield on r* is related to the use of balance sheet policies.

If QE raises the convenience yield by reducing the net supply of government bonds, it may ultimately lower the real equilibrium interest rate. Importantly, this channel – the convenience yield channel – is distinct from the term premium channel.^[20]

So, doing QE could be like chasing a moving target.

It reduces long-run rates by compressing the term premium.^[21] But by making investors willing to pay a higher safety premium when the supply of safe assets shrinks, it may also reduce the interest rate level below which monetary policy stimulates growth and inflation.

This can also be seen by looking at how QE changes the balance of savings and investments. Fiscal deficits absorb private savings and thereby increase r*. By doing QE, central banks absorb fiscal deficits and thereby lower r*.

In other words, central bank balance sheet policies may be less effective than previously thought.^[22] This could be an additional factor explaining why large-scale asset purchases did not succeed in bringing inflation back to 2% before the pandemic.

Of course, the same logic holds true when central banks reduce their balance sheets.

If QE contributed to depressing r*, QT will raise it. Any rise in real rates may then be less consequential for growth and inflation. It would then be misguided to compensate for higher long-term interest rates resulting from QT with lower short-term rates.

This is indeed what recent research suggests: QT announcements tend to cause a significant decline in the convenience yield of safe assets.^[23]

There is one caveat, however.

QE and QT are implemented by issuing and absorbing central bank reserves, which themselves are safe assets – in fact, reserves are the economy’s ultimate safe asset because they are free of liquidity and interest rate risk.^[24]

Banks therefore highly value the convenience services of central bank reserves. So, when evaluating the effects of central bank balance sheet policies on r*, it is necessary to consider both the asset and liability side.

Research by economists from the Bank of England does exactly that.^[25] They show that the effects of QT on the real equilibrium rate depend on the relative strength of two factors.

One is the effect on the bond convenience yield, which causes r* to rise as the supply of government bonds increases.

The other is the effect on the convenience yield of reserves. That effect is highly non-linear: when reserves are scarce, banks are willing to pay a high mark-up on wholesale interest rates, as was evident in the United States in 2019 when repo rates surged strongly.

So, if QT leads to a scarcity of reserves, it may cause the overall convenience yield to rise, and hence equilibrium rates to fall.

Convenience of reserves and the ECB’s operational framework

At the ECB, we took this factor into account when we reviewed our operational framework last year.^[26] This is the third implication for monetary policy.

The new framework allows banks to demand as many reserves as they find optimal at a spread that is 15 basis points above the rate which the ECB pays to banks when they deposit their excess reserves with us. So, the opportunity cost of holding reserves is comparatively small, given the convenience services reserves provide to banks.

In addition, our framework allows banks themselves to generate an increase in safe assets – by pledging non-high quality liquid assets (non-HQLA) in our lending operations. In doing so, banks on average generate € 0.92 of net HQLA for every euro that they borrow from the Eurosystem.^[27]

Our framework therefore recognises that years of crises, more stringent regulatory requirements and the advance of new technologies – some of which increase the risk of “digital” bank runs – imply that banks may wish to hold larger liquidity buffers than they historically have done.

Supplying central bank reserves elastically will ensure that reserves will not become scarce as balance sheet normalisation proceeds. And if banks access our standard refinancing operations when they are in need of liquidity, they will also not have to adjust their lending activities in response to the decline in reserves, as is sometimes feared.^[28]

For now, the recourse to our lending operations has been limited, as there is still ample excess liquidity. But as we transition over the coming years to a world in which reserves are less abundant, banks will increasingly start borrowing reserves via our operations.

Three ideas could be explored to make this transition as smooth as possible.

First, regular testing requirements in the counterparty framework could help ensure operational readiness while also allowing counterparties to become more comfortable with participating in our operations. A lack of operational readiness was one of the factors contributing to the March 2023 turmoil in the United States.^[29]

Second, and related, obtaining central bank funding requires thorough collateral management, especially if the collateral framework is as broad as the Eurosystem’s. For non-HQLA collateral, in particular, the pricing and due diligence process can be operationally complex and time-consuming.

For this reason, central banks sometimes require counterparties to pre-position collateral to ensure that funding can be readily obtained.^[30] In the euro area, some banks already pre-position collateral voluntarily, in particular non-marketable collateral which cannot be used in private repo markets (Slide 12, left-hand side).

Banks could be further encouraged to mobilise with the central bank the collateral that is eligible but currently stays idle on their balance sheets. This would increase operational readiness, mitigate financial stability risks and reduce precautionary reserve demand as banks would have higher certainty that they can access central bank liquidity at short notice.

In the Eurosystem, given its broad collateral framework, such an approach may be more effective in helping banks adapt their liquidity management to the characteristics of a demand-driven operational framework compared with a blanket requirement to pre-position collateral.

Finally, in some jurisdictions central bank operations are fully integrated into the platforms commonly used by banks to operate in private repo markets.

This offers banks a number of advantages, including seamless access to transactions with the market and with the central bank, and – depending on the design of clearing arrangements and accounting rules – it could potentially allow banks to net out their positions, thereby freeing up valuable balance sheet space.

Offering banks the possibility to access Eurosystem refinancing operations through a centrally cleared infrastructure could contribute to making our operations more economical in an environment in which dealer balance sheets are increasingly constrained (Slide 12, right-hand side).^[31]

The design of such arrangements should preserve equal treatment across our diverse range of counterparties, regardless of their size, jurisdiction and business model, maintain the possibility to mobilise a broad range of collateral and be compatible with our risk control framework.

Further reflection is needed on these considerations, including a comprehensive assessment of the benefits and costs.

Conclusion

Let me conclude.

The shocks experienced since the pandemic led to an abrupt end of the secular downward trend in real interest rates. Whether this will be merely an interlude, or the beginning of a new era, is inherently difficult to predict.

But looking at the ongoing transformational shifts in the balance of global savings and investments, as well as at the fundamental challenges facing our societies today, higher real interest rates seem to be the most likely scenario for the future.

This has implications for our monetary policy. Central banks will need to adjust to the new environment, both to secure price stability over the medium term and to implement monetary policy efficiently.

Thank you.