Why Look Beyond On-Screen Liquidity

Many investors use point-in-time statistics for the most recent 30 or 90 trading days to assess the liquidity profile of an ETF that has traded on an exchange for longer than just the last 30 days. This is a problem because—with the exception of a few highly liquid ETFs—a fund’s liquidity profile can change in different market regimes. This paper explores how ETFs’ liquidity dynamics impact total cost of ownership (TCO), underscoring why investors need to look beyond a single period statistic when analyzing liquidity.

How Market and Fund Profile Changes Impact Liquidity

Many investment strategies require implementation over different periods, potentially occurring over multiple market volatility regimes and not always during periods of persistent tranquility. Volatility in the broader marketplace can impact the liquidity profile of an ETF in the secondary market, in addition to the liquidity of its underlying securities. The type of client base using an ETF can also impact its liquidity profile in the secondary market and should be considered during the due diligence process. For instance, if a fund obtains an identifiable three-and five-year track record of top-decile performance, it could improve the fund’s liquidity profile through an uptick in activity (volumes and flows) from an investor base seeking a high-performing time-tested strategy. Robust liquidity due diligence, therefore, must be tailored to account for shifts in market volatility, client demand and user base.

To illustrate the value of a regime-dependent due diligence process, we chose two ETFs focused on the same market, which we will refer to as Fund A and Fund B. Let’s compare the TCO based on expense ratio and transaction costs, with other factors (tracking error, commissions) held constant:

Fund A has the lowest expense ratio. Fund B has a slight bid-ask spread advantage over the past 90 days, but it is condensed over the past 30 days, underscoring the need to look beyond one data point. The difference between the 30-and 90-day spreads also suggests the potential for a near-term trend forming, which requires a deeper look. The lower-cost fund, therefore, depends on the lookback period used, a major flaw of period-based analytics. Perhaps Fund A has the lower total cost. Perhaps not.

Key Observations

Reviewing just one period in time may not capture all of the different trading environments as market volatility and client demand can change over time. Average spread and volume data can be skewed depending on market and client dynamics during the time period being considered. There are days within those periods where spreads can be much higher or much lower. Therefore, it’s important to take a deeper dive into the data.

Market-Based Analysis

Because an ETF’s assets and trading volume can ebb and flow, we believe it’s helpful to focus on at least two years of data when conducting both market-based and fund-specific regime analysis. Too short of a time period may not allow for multiple market environments to be analyzed. With too long a period, analysis could be based on potentially stale data that may not reflect the current market environment and fund profile.

We evaluated Funds A and B from January 2018 through June 2019, a period that covered two market corrections and ensuing rallies, as well as multiple Federal Reserve (Fed) rate hikes and other macro-based events, such as tariff-driven US-China trade tensions. Next, we extended the analysis beyond simple averages and looked at maximum, minimum and median spread levels over the time period and bucketed the distributions in a histogram to identify the frequency of days with a specific bid-ask spread range.

As illustrated in Figure 2, Fund B has the lower bid-ask spread based on both the median and recent average figures. Fund B also has consistency, trading at less than a 0.03% spread on every single day within this period. Meanwhile, Fund A has had a more normal, but less consistent, distribution —with more than 134 days (34% of observations) having a spread greater than 0.07%. Because the minimum bid-ask spread over this period is similar to the more recent 30-day average, a deeper look is required to explore a potential recent trend that would not be noticeable if an investor looked only at a single data point.


Given the variability of the spread for Fund A, the next step is to understand how the spread has changed during different volatility regimes. An increase in market volatility can lead to higher bid-ask spreads on ETFs, with the bulk of the change depending on the fund’s liquidity profile and market focus.

Figure 3 illustrates how Fund A and Fund B behaved during different market volatility regimes. The regime was identified by calculating the rolling 30-day realized standard deviation of returns for the underlying markets of focus. The realized volatility was separated into deciles, with the lowest periods of volatility represented by the first decile and the highest periods represented by the tenth decile. For each decile, we calculated the median bid-ask spread.

Key Findings

Fund A has experienced more variability based on the level of realized volatility of the underlying market. It has also shown a linear relationship between periods of high volatility coinciding with higher spreads. Interestingly, Fund B’s tenth decile, when extended out to three decimals, actually has a median spread of 0.021% —its only decile with a spread that rounds higher than 0.02% flat. This underscores how high volatility regimes can impact overall trading costs, albeit only minimally for Fund B.

Fund Profile Analysis

Dissecting spread levels into specific buckets provides context for the variability of trading costs and how spreads can change significantly based on volatility. While this process can detect certain relationships or patterns, one last step is necessary based on these findings. As documented throughout this analysis, there has been a recent compression of the 30-day bid-ask spread relative to longer time periods. A spread compression like this can occur if there has been a fundamental change in the buying behavior for a fund, whether that’s prompted by a market being in favor, strong returns relative to peers’, or idiosyncratic client-specific drivers, such as an allocation with a model portfolio used across multiple trading platforms.

A time-series analysis of spread changes overlaid on top of asset growth for the defined lookback period can provide insight into whether there has been a shift in Fund A’s demand and user base. Consistent with our preference of looking at more than one statistic, Figure 4 plots the asset growth along with the rolling 5-day average, 30-day average, and 5-day median spread.

Key Findings

While Fund A had been of substantial size prior to 2019, with over $1.5 billion in assets, there was a distinct shift at the start of the year. Assets soared to close to $3 billion, and the bid-ask spreads moved lower. This coincided with the median number of 2019’s daily trades, which is one metric that can be used to measure client demand, surging by 64% from a median daily value of 1,641 in 2018 to 2,686 trades a day in 2019. This illustrates a clear change in the liquidity dynamics for Fund A. As a result, the bid-ask spreads from a time period when the fund was smaller do not reflect the fund’s current demand, user base and profile. To reinforce this fund’s regime change, we recalculated Figure 1 with just 2019 data and found that Fund A distribution was no longer normal, as the spreads compressed and were clustered at lower levels.

Controlling for Size Outliers with Number of Trades

Notably, the dynamics for Fund A did not change entirely, as there was a slight increase in the bid-ask spread in late May and early June, right around the time realized volatility picked up (readings registered in fifth, sixth, and seventh deciles). Fund A is, therefore, not immune to volatility just because assets have increased by 162% since the start of 2018. But the spread didn’t gap as high as before, which is a positive for Fund A. Its sensitivity to volatility, therefore, can be considered to have been reduced as a result of higher assets and client user base. Even with a 0.04% bid-ask spread during the volatility spike, the TCO for Fund A (0.11% + 0.04% = 0.15%) is still below that of Fund B (0.14% + 0.02% = 0.16%).

Fund size had a direct impact on the bid-ask spread here. However, that is not always the case. In this instance, the fund size coincided with an increase in the number of trades being executed. That means it wasn’t one large investor, but many investors seeking out this exposure. With greater demand from many counterparts, trading costs were naturally lowered. Notably, a fund can also be large with one investor holding a majority of the shares outstanding. In that case, that fund has scale but not depth in client base. If that one large client does not trade, volumes are light and liquidity can dry up.

The table below depicts 2019 trading data for two smart beta ETFs that have similar assets, factor of focus and expense ratios. The difference is that Fund D has 90% of its assets owned by one investor.

Key Findings

Not having organic flow from multiple sources —even if the fund itself is large —can be detrimental to a fund’s liquidity profile, particularly in volatile markets that result in higher spreads. That’s why examining the diversity and number of investors —in addition to spreads, trading volume and assets under management —should be a consideration while conducting a fund’s liquidity analysis.

Analyzing and Trading Costs: A Multistep Process

Determining which fund is more beneficial to use depends on an investor’s typical execution strategy, trading frequency, order size and the emphasis placed on maintaining low expense ratio profiles within a broader portfolio.

After running multiple screens, the results of our analysis differ at each step:

  1. Based on the 30-day average bid-ask spread, Fund A is the fund to select.
  2. Based on the 90-day average bid-ask spread, Fund A or B could be selected.
  3. Based on volatility regime and frequency analysis, Fund B has shown more consistency.
  4. Based on using more recent data and client usage information after a change in the fund’s profile, Fund A results in the lower overall cost.

The bottom line is that getting to this answer, however fluid it is, requires more than just a simple screen of the past 30 days of trading —especially if trading volumes are light and market volatility is low.