Do exchange-traded funds (ETF) exhibit unique calendar-based anomalies? In their April 2018 paper entitled “Evidence of Idiosyncratic Seasonality in ETFs Performance”, flagged by a subscriber, Carlos Francisco Alves and Duarte André de Castro Reis investigate calendar-based patterns of risk-adjusted returns and tracking errors for U.S. equity ETFs and compare findings to those of underlying indexes. They aggregate returns of their ETF sample via equal weighting. They consider returns calculated based on either market price or Net Asset Value (NAV). For risk adjustment, they consider alpha from either 1-factor (market) or 4-factor (market, size, book-to-market, momentum) risk models of stock returns. They look for raw return or alpha patterns in calendar months, calendar quarters, months of calendar quarters, calendar half-years, days before holidays (New Year’s Day, Martin Luther King Jr. Day, George Washington’s Birthday, Good Friday, Memorial Day, Independence Day, Labor Day, Thanksgiving and Christmas), days of the week and turn-of-the-month (last trading day of a month through three trading days of the next month). Using daily prices and NAVs for 148 index-tracking U.S. equity ETFs and associated indexes, and contemporaneous equity factor model returns, during December 2004 through December 2015 (11 years), they find that:
- Using raw returns, there are no statistically significant calendar-based anomalies, except for weak indication that NAV-based returns are strong in April.
- Using 1-factor and/or 4-factor alphas:
- The first (second) half of the year is strong (weak) for NAV-based returns. More granularly, the second (fourth) quarter is strong (weak) for NAV-based returns.
- The first (third) month of a quarter is strong (weak) for both price-based and NAV-based returns.
- January and perhaps May are strong using NAV-based returns, and April is exceptionally strong using either price-based or NAV-based returns.
- June, September and especially December are weak using NAV-based returns.
- There is no turn-of-the-month effect.
- There are weak indications that Wednesday is strong using price-based returns and Friday is weak using NAV-based returns.
- Benchmark tracking errors are relatively low during April, the second quarter and the first month of each quarter.
- Anomalies detected for ETFs are generally (except for April effects) not present in benchmark index returns, indicating that identified patterns are largely unique to ETFs.
In summary, evidence suggests that there are few, if any, exploitable calendar-based anomalies unique to U.S. equity ETFs. The most promising indication is for the month of April.
Cautions regarding findings include:
- The 2005-2015 sample period is short in terms of variety of market conditions. The 2008-2009 crash may be decisive to findings.
- Since aggregation of ETF returns is via equal weighting, behaviors of small funds may drive findings.
- The study focuses on statistical, not economic, significance. Magnitudes of findings are generally modest even when statistically significant.
- Most anomalies found are for unexploitable NAV-based returns.
- Exploitation of findings for ETF timing would involve trading frictions debited from any pattern-based edges.
- Searching the same data for multiple anomalies based on two ways of calculating returns (price and NAV) and using both raw returns and two alphas introduces material data snooping bias, thereby overstating significance of findings.
For other perspectives, see “Sector Performance by Calendar Month”, “Does the Turn-of-the-Month Effect Work for Sectors?” and “Does the Turn-of-the-Month Effect Work for Asset Classes?”.
