Beyond Duration and Credit Spreads: Are We Measuring the Right Sources of Return in Fixed Income?

For decades, fixed-income performance attribution has been built around a familiar set of drivers: duration, yield-curve positioning, credit spreads, currency exposure, and security selection.

These frameworks have served the industry remarkably well.

But as I look at the rapid emergence of tokenized assets, digital cash, and AI-driven investment processes, I find myself wondering: Are we still measuring the right sources of return?

The Assumption We Rarely Question

Traditional attribution models were developed in a financial system where market infrastructure was largely invisible.

Whether a bond settled through one system or another made little difference to investors. Infrastructure was simply a utility layer sitting beneath the market.

Today, that assumption is being challenged.

Imagine two bonds with identical issuers, maturities, coupons, and credit ratings. One trades through traditional market infrastructure. The other exists as a tokenized security with near-instant settlement, broader accessibility, and enhanced collateral mobility.

If investors consistently value one more highly than the other, where does that excess return come from?

It is not duration.It is not credit.It is not security selection.It may be something else entirely.

The Rise of "Infrastructure Return"

Historically, we viewed infrastructure as a cost centre. Increasingly, it may become a return driver.

Settlement efficiency, liquidity access, collateral flexibility, and network participation can all influence asset prices and investor behaviour.

This raises an intriguing possibility:

What if part of future portfolio performance is generated not by the asset itself, but by the infrastructure through which it exists? If so, traditional attribution models may be missing an increasingly important piece of the puzzle.

Digital Cash Is Not Just Cash Anymore

Another assumption being challenged is the idea that cash is homogeneous.

For generations, cash was simply cash.

Now we have:

• Traditional bank deposits

• Money market funds

• Stablecoins

• Tokenized deposits

• Tokenized Treasury products

Each offers different characteristics around settlement, liquidity, interoperability, and accessibility.

As these ecosystems mature, investors may need new ways to distinguish returns generated by yield from returns generated by transactional utility.

The concept of "cash alpha" may sound unusual today, but so did "liquidity alpha" not long ago.

The AI Attribution Problem

Artificial intelligence introduces another layer of complexity.

Traditionally, excess returns from security selection were attributed to portfolio managers and investment teams.

But what happens when AI systems increasingly influence:

• Trade ideas

• Portfolio construction

• Risk management

• Relative-value identification

  
  

If an AI model identifies an opportunity that generates alpha, who or what deserves the attribution?

This may sound philosophical, but it is quickly becoming a practical challenge for asset managers seeking to understand the true drivers of investment performance.

A New Attribution Framework?

Perhaps future attribution models will need to move beyond market risk alone.

Instead of asking only: "What market factors drove returns?"

We may also need to ask:

"What infrastructure generated value?" And "What intelligence generated value?"

A future framework might include four layers:

Market Layer

• Duration

• Yield curve

• Credit spreads

• Currency

Security Layer

• Issuer selection

• Relative value

• Credit research

Infrastructure Layer

• Settlement efficiency

• Tokenization effects

• Liquidity enhancement

• Network participation

Intelligence Layer

• Human decision-making

• Quantitative models

• AI systems

• Human-AI collaboration
  

The Bigger Question

The most interesting financial innovation of this decade may not be tokenization, stablecoins, or AI individually.

It may be the convergence of all three.

As assets become programmable, cash becomes digital, and investment decisions become increasingly automated, the sources of return themselves may evolve.

If that happens, performance attribution will need to evolve too.

Because before we can improve performance, we first need to understand where it is actually coming from.

What do you think?

Will traditional fixed-income attribution remain sufficient, or are we approaching a point where infrastructure and AI need to be treated as distinct sources of investment return?