• Edward Ballsdon



(There is an outline of yield curve and “steepen/flatten” terminology in the Appendix).

In the last 30 years, many have discussed the validity of the yield curve as a good predictor of an impending recession. As the curve flattened in 2006, many warned about a forthcoming recession. This was unsurprising as the previous incidents of curve flattening in 1989 and 2000 had predicted the subsequent recessions. The Fed published a paper in 2006, which concluded:

“Our analysis suggests a number of practical guidelines for the use of the yield curve to predict recessions in real time:

  • Treasury rates are most likely to produce accurate forecasts.

  • The best maturity combination may be three months and ten years.

  • Levels of the spread are more informative than changes.

Below is the chart they produced of the 3m-10y yield spread vs recessions (shaded).

In 2018, many analysts started highlighting the flattening curve and the risk of an impending recession. But unlike previous occasions, and before Covid19 struck, the US economy was not showing signs of tipping into a classic end of cycle recession. Indeed, many leading indicators were fairly robust, pointing to a decent US GDP ahead. Below shows the 2s10s Treasury curve spread vs the ISM – the Services was relatively robust, and the Manufacturing had turned upwards. This raises the question of whether the most recent curve flattening in 2018 and 2019 gave a false signal?


The chart below shows 2y (Red) and 10y (Green) US Treasury yields and the resulting 2s10s spread between the two (Black). Note:

  1. In general, as the curve has flattened AND steepened over the last 30 years, 2yr and 10yr yields have trended lower towards zero.

  2. At every trough of 2s10s spread (black vertical) just before a recession, the 2 year yield peak was lower than the previous cyclical 2yr yield peak (pink arrows).

  3. Each peak in 2s10s curve steepness coincides with the cyclical trough in 2yr yields. However each subsequent 2yr yield trough is at a lower yield to the previous trough (blue arrows).

  4. The 2yr is far more volatile than the 10yr, despite the latter’s higher duration (risk).

A simple regression analysis confirms that over the last 20 years, the 2y10y spread has had a significantly higher correlation to the 2yr compared to the 10 year ((R2 = 0.62 and 0.22 respectively). This leads to the hugely unsurprising confirmation that if yield curve flattening warns of recessions, then it’s the raising of short end rates that probably helps cause that recession, and the cutting of rates eases the recession towards GDP growth.

Of course the drivers of a recession are far more complex than this, but the point of the charts is to show that there is now an issue – a new trough below the previous one cannot occur unless rates go negative. This time looks different.


Once the Fed cut rates to combat the consequences of Covid19 on the US economy, the whole yield curve shifted downwards in a steepening fashion (i.e. short end yields dropped more than long end yields), as 1m Bills to 2yr Treasuries almost reached 0%.

One of the consequences of low interest rates is that the determinant of the curve shape suddenly changes from the short end to the long end, as front end yields get anchored at 0%. Indeed, since the 2yr yield collapsed in March, the shape of the 2s10s curve has now become highly correlated to changes in the 10 yr yield.

An exactly similar relationship change occurred in Japan in 1995, and more recently in Europe in 2012, when 2yr yields dropped to 30bp. The shape of the European curve, like the Japanese curve, is now determined by the long end (chart below). If the argument stated earlier is correct, i.e. that it’s the rise in short end rates that warns about an impending recession, then following a relationship flip, the curve is no longer a reliable signal for impending recessions.

More importantly, it could be argued that the curve now gives a completely different and more ominous signal – that a debt trap is now in place, signed, sealed and delivered.


The events in Q418 now seem even more significant in retrospect. The Fed’s 100bp rate hikes in 2018, together with market expectations of further rate hikes in the future, pushed short end yields higher throughout 2018, thereby flattening the curve. The cost of debt servicing rose.

Throughout 2018, the Fed ALSO reduced its holdings of US Treasuries, just as Trump was adding $1trn to the deficit. The net result was that the net supply of Treasuries to the market was $1.4trn. Because foreigners did not change their holdings, this resulted in a $1.4trn drain of liquidity from the domestic private sector (blue columns below). It’s hardly surprising that risk assets corrected in the 4Q when the economy was doing ok.

This risk asset correction resulted in an immediate about turn in Fed policy, who not only promised to no longer hike rates and subsequently cut them in 2019, despite core inflation at 2% and very low unemployment, but also led to a changing of their balance sheet policy, increasing its UST holding from Sep19 onwards. In effect by changing tack so quickly, the Fed “front ran” a possible recession that might have come if the asset price correction that started in 4Q had continued in a more meaningful fashion (which was more than possible given the over leveraged corporate sector, rising rates and contracting liquidity due to huge UST financing requirements).


Assuming no move to deep negative rates, the Fed has now fully utilised its interest rate tool, and can now only implement its balance sheet to buy USTs and MBSs for its own account, or to buy Corporate Bonds on behalf of the US Government. It could be argued therefore that Central Bank Monetary Policy is now spent as the Fed can no longer increase private sector credit growth – it is now Fiscal Policy that will determine private sector credit growth, even if the Fed’s balance sheet is the transmission mechanism.

The problem is that unlike certain governments whose finances escaped a marked deterioration post GFC (e.g. Australia, Canada, Sweden), US government debt is no longer 41% of GDP as it was pre GFC, or 102% as it was at the end of 2019 coming into the Covid19 pandemic. It is now a large 117% (using 31Dec19 GDP).

Outstanding US Marketable and Non-Marketable Federal debt now stands at $25.4trn, compared to $9.1trn in Dec07, resulting in the servicing of the debt stock being almost 3 times as sensitive to changes to interest rates. Given the lack of traditional monetary policy tools, the US Government can, and most likely will issue more debt to stimulate the economy, especially now that Central Banks can finance that debt. But as the debt increases, that sensitivity will increase further, necessitating a continuation of anchoring of interest rates at very low levels. This, after all, is the Japanese (and Italian and Greek) scenario.


An outcome of further debt issuance and anchoring of short term interest rates will make the curve a redundant indicator of future recessions, just as it is now for Japan and Europe. PMIs and other such indicators will be far more useful. This scenario, which is not entirely unlikely given current fiscal policy announcements and enacted Central Bank Balance Sheet actions, leads to the question of long term interest rate volatility. Consider the two charts below.

The left shows the 10yr interest rate in 10 years time (the 10y10y) for the US (1.14%), UK (0.47%), EUR (0.25%) and JPY (0.33%). The market is now expecting that long maturity yields will be low, not just in the short term, but in 10 years time. This is “Japanification” – there are further details in this Blog Post.

The right hand chart shows the implied volatility for an at the money 10y10y swaption (i.e. how volatile the market expects the 10yr to be for the next 10 years). Despite the convergence of long dated forward interest rates to those of Japan, and the European 10y10y rate trading BELOW the Japanese 10y10y rate, implied vols have only declined mildly and remain substantially higher (more than double!) than the JPY implied vol level.

Long dated volatility (vega) used to be a good hedge against risk asset declines. Implied vol would spike as equity declined/credit spreads widened, and because the implied vol curve was negative, you could be long vega as a risk asset hedge and have a positive carry. But then vega prices were almost completely unchanged in the 4Q18 equity sell off, and there was no sizeable change in 2020, despite an increase in realised 10y10y interest rate volatility (charts below, use 63day realised vol).

This lack of convergence suggests that the interest rate volatility market has still not come round to the Japanification theme. This is not entirely surprising - it goes against the grain and mainstream economic thinking. As can be seen from the charts above, it took a long time for Japanese implied vol to decline after Japanese long dated forward rates declined. For sophisticated investors who know their options inside out, and believe in the Japanification argument, this offers the opportunity to sell out of the money payer options on long dated interest rates (i.e. sell the probability that rates rise in the future) and receive a premium as a financing leg against other asset class options.


“This time is different” is a very dangerous phrase to utter in financial markets, as those many market participants (and they were in very many) found out when they said the curve was giving false signals in 2007. But the value of the curve as a predictor of recessions only holds when a state of “over-indebtedness” does not exist and interest rates have room to rise or decline. It does not hold when front end interest rates are anchored due to excess debt.

When the interest rate and debt dynamics change significantly, then it’s fair to say “this time is different”, because it is not different - it’s actually just a progression of the credit cycle. It is natural that curve dynamics change when there is anchoring of short term interest rates at very low levels, as seen first in Japan and then in Europe, Sweden and Switzerland and now in the UK and US. After identifying the change in the curve relationship (the “Japanese flip”), it becomes important to identify that the curve is now signalling something new and far more important. Because of the much enhanced sensitivity of public (and private) debt to interest rates, the whole interest curve is likely to now be capped, including long dated forward rates. This leads to a reduction in realised interest rate volatility, thereby bringing a decline in implied volatility – until that happens, there is an opportunity to sell upside rate vol to finance cross asset options.

If this is correct, then there are extremely important implications for equity market valuations, whose richening has depended on declining long dated yields. More on that another time!


A “Yield Curve” shows the different yields that an investor obtains by investing across different maturites of US Government Bonds. E.g. a 2yr yields 0.16% whilst a 10yr yields 0.64%.

The “Yield Curve Spread” is the interest rate differential between two different maturities. E.g. the 2s10s yield spread is currently 0.48% or 48 basis points (bp).

If the Spread reduces, the yield curve is deemed to “flatten”. The chart below shows the yield curve on the 1st Jan 2017 and the current yield curve. Over time the curve has moved lower, but longer dated maturities have declined more than shorter dated maturities. Thus the 2s10s has declined (by 58bp) and the 2s10s curve has “flattened”. If the opposite occurs, and the 2s10s spread increases, then the 2s10s curve is deemed to have “steepened”.