Author Details

Andy Sparks

Andy Sparks

Managing Director, MSCI Research

Juan Sampieri

Juan Sampieri

Vice President, MSCI Research

Chris Fenske

Chris Fenske

S&P Global Market Intelligence

Social Sharing

Extended Viewer

Bond-Index Replication While Navigating Volatility

  • Market volatility poses major challenges to investors trying to track bond indexes while also keeping transaction costs low.
  • We use a “smart turnover” replication approach to construct hypothetical transaction-cost-aware bond portfolios.
  • Our findings highlighted how this approach may have significantly reduced transaction costs during the COVID-19 crisis.

The costs of transacting in the bond market can be high, especially during volatile periods. Portfolio managers aiming to track a benchmark might find that trying to buy every bond in proportion to its weight in the benchmark could result in excessively high transaction costs.

In the most extreme case, it simply may not be possible to buy a specific bond — for example, if the portfolio manager is seeking a bond held in an insurance company’s buy-hold portfolio, which allows only limited trading. This could create a mismatch in portfolio and benchmark weights that, combined with spread and rate volatility, will contribute to tracking error.1

An alternative is to buy fewer bonds in the portfolio than in the benchmark and allow portfolio weights to differ from the benchmark’s. But what is a potential way to do so that balances transaction costs and tracking error?


Using ‘smart turnover’

We employed optimization techniques to construct transaction-cost-aware portfolios that track the MSCI USD Investment Grade Climate Change Corporate Bond Index. In our methodology, turnover times bid-ask generates transaction costs.2 We used a “smart turnover” approach designed to lower transaction costs by controlling purchases and sales and keeping portfolio risk close to the benchmark’s.

Liquidity is multidimensional, of course. Using bid-ask only is an imperfect measure of liquidity, as it doesn’t assess the tradability of a bond. To address this, we incorporated S&P Global Market Intelligence’s parsed dealer-quote-depth data in the portfolio-construction process. Specifically, we allowed only buys and sells of bonds that had at least six unique dealers quoting the bond during the prior month.3

Market makers are an integral part of a functioning corporate-bond market, and the quotes they send to their clients may provide a gauge of a bond’s tradability. Dealers’ willingness to provide quotes often depends on their dual roles as principal and agent. As a principal, their desired inventory partly reflects their ability to hedge and views on relative value. As an agent, their willingness to serve client flows with bid-ask depends on their ability to source buys and sells.

We constructed three hypothetical portfolios, each subject to a different penalty for allowing tracking-error volatility (TEV).4 We backtested these portfolios between December 2018 and February 2022, using monthly rebalancing.5 The number of bonds in the benchmark ranged from 2,100 to 3,300 over the period of our analysis. The number of bonds in the portfolios was constrained to be close to 500.

Portfolio 1 had the lowest TEV relative to the investment-grade benchmark, but also the highest cumulative transaction costs. Nevertheless, it was still 4 basis points (bps) below the benchmark’s calculated transaction costs. Portfolio 2 resulted in transaction costs that were 9 bps lower than the benchmark’s, but at a “cost” of greater TEV. Portfolio 3 came in almost 20 bps below the benchmark’s transaction costs, but had the greatest TEV. All portfolios had approximately the same effective duration, spread duration and option-adjusted spread as the benchmark. The portfolios also had less than 200 issuers each, compared with an average of 360 issuers in the benchmark.


The trade-off between transaction costs and tracking error


Returns were adjusted to include transaction costs. Tracking-error volatility was computed using exponential time decay with weekly return intervals and a half-life of one year.

The portfolios consistently had lower bid-ask than the benchmark. In the months after the onset of the COVID-19 pandemic, there was significant index turnover, primarily triggered by high levels of bond-market issuance. The portfolios also experienced spikes in turnover, but less than that of the benchmark.


Focusing on transaction costs



Transaction-cost-aware portfolios

Index rules generate turnover with the addition of new bonds or the removal of existing ones. In contrast, transaction-cost-aware index-replication approaches may be able to reduce transaction costs by having reduced turnover (“smart turnover”) with bonds that have tighter bid-ask spreads. These approaches may have even greater effect on transaction costs in volatile periods with sharp changes in index composition. In addition, such approaches may help investors navigate the rapid integration of climate and ESG considerations into the investment process by reducing the cost of benchmark conversions.



1Tracking-error volatility (also known as active portfolio risk) is the projected annualized volatility of the difference in returns between the portfolio and the benchmark.

2Published returns on the benchmark do not include transaction costs. In our analysis, we adjusted index and portfolio returns to include transaction costs using the same methodology for both.

3Aggregated dealer-quote depth by month that includes round-lot (>=USD 1 million size) dealer quotes (bids, offers, two-markets) from the top 12 U.S. corporate-bond broker-dealers. The data counts the number of unique dealers that quoted a bond during a month.

4Specifically, we used MSCI’s Barra® Open Optimizer, the short-term version of the MSCI Multi-Asset Class Factor Model, MSCI’s Bond Liquidity Measure of bid-ask and S&P Global Market Intelligence’s parsed dealer-quote-depth data. For each portfolio, the optimization goal was to minimize transaction costs subject to a different penalty for allowing tracking-error volatility. All portfolios were also constrained to have approximately 500 bonds; weights within each sector to be within 10% of the benchmark’s; duration times spread within each sector to be within 20% of the benchmark’s; key-rate durations within 20% of the benchmark’s; a maximum 3% weight by issuer; and a maximum .4% weight and a minimum .01% weight by bond. All bonds in the portfolios were required to belong to the benchmark. Buys and sells were allowed only for bonds that had at least six unique dealers quoting the bond during the prior month.

5There are frequently material differences between backtested or simulated performance results and actual results subsequently achieved by any investment strategy. Past performance — whether actual, backtested or simulated — is no indication or guarantee of future performance.



Further Reading

Surging Corporate-Bond Supply: Reason to Worry?

Did Bonds Deliver? Leveraging Fixed Income During the COVID Crisis

How easy is it to track a bond market index?