The Connection Between Current Discounts and Future Returns
Please note: past performance is not a guarantee of future results. The study covers the time period of 1/29/1988 to 9/25/2009, and has not been updated through current day. It’s possible that more recent market conditions and events would impact both the University of Oregon study and our hypothetical back-tested study, if the data was expanded to current day. However, we don’t believe it would create any material changes or lead us to draw significantly different conclusions.
We started managing separate accounts during 2000-2005 that focused on discounted CEFs, and we soon realized it was difficult to find discount-focused research on the CEF market as a whole. We could find plenty of research on individual CEF issues, but we were looking for broader research on discounts in the CEF market, with the goal of uncovering any potential themes or trends that could then be implemented into our portfolio management and overall investment strategy.
In 2008, we initiated our own research study to investigate whether any relationship between CEF discounts and subsequent total returns had historically persisted over time. We could find no evidence that anyone had engaged in such an investigation before. The only CEF index we found at the time was a 30-fund index calculated by Herzfeld Advisors in Miami, and provided weekly in Barron’s.
The goal of our study was to answer the simple question:
Do discounts matter?
So in conjunction with the Cameron Center for Finance and Securities Analysis (Lundquist College of Business; University of Oregon), we set about constructing our own indexes of CEFs. We quickly found that nearly all the databases with information about CEFs threw out the data when funds went out of existence (either through liquidation, merger, or conversion into an open-end fund). And so we decided to do the best we could by taking all 600+ CEFs in existence in 2008, and collecting all relevant data on these funds back as far as we could. For us, relevant data included month-end prices, month-end NAVs, all distribution history, and any capital change information (for each fund dating back to its inception).
The end result was a comprehensive dataset that covered the time period of 1/29/1988 to 9/25/2009. Once all the relevant data was collected, we began slicing and analyzing the dataset in every meaningful way that we could think of.
Study Results Summary
Historically, for U.S. CEFs, discounts to NAV have been highly correlated with subsequent total returns.
Results are statistically significant across time, and across different CEF types.
Important Disclosures About This Study
We are aware that our study has potential limitations, the principal one being survivorship bias. In addition, it would not be possible to know precisely ahead of time what a fund’s discount at the end of a trading day would be, and so one would not know precisely what quintile it would fall into. Further, there are liquidity constraints, and closing prices may not be available the following day. Lastly, the study has not been updated and may not account for more recent market conditions and events. We’ve attempted to address some of these issues within the context of our data.
First, regarding survivorship bias, we have been running the quintiles “live” since 2005, therefore retaining the results of funds that go out of existence. The results have been similar to the pre-2005 results from a statistical perspective. Further, we note that survivorship bias ought to be less of an issue with closed-end funds than it is with open-end funds, since closed-end fund assets are much more effectively “captive”. A poor-performing fund manager can much more easily continue to run his fund forever, rather than being forced to close it due to asset flight (and we’ve seen many examples of this). In fact, we believe survivorship bias may work the opposite way in closed-end funds, since many funds go out of existence by being liquidated or converted at NAV. To the extent such funds would have been in our widest-discounted quintile index, they might have contributed positively to overall index results, given the large positive contribution to total return that would have come from the elimination of the discount to NAV. Even if a 1-2% survivorship bias penalty were applied (2% is the largest effect we’ve ever seen in a study of open-end funds), the back-tested results would still be attractive, in our opinion.
Second, regarding the uncertainty about which quintile a fund would fall into at the end of a trading day, we note that the results are so broad (the widest quintile in our study has contained over 100 funds since 2004, and over 50 funds since 1994) that they are unlikely to be impacted by a handful of funds on the border between quintiles.
Third, regarding liquidity, we are well aware that this is a concern for our study. There is in fact a (small) correlation between market cap and size of discount, with smaller funds trading at wider discounts. Even for the medium-size and larger funds, one would be hard-pressed to construct and trade a portfolio of closed-end funds with over $1 billion. However, our index evidence suggests that the alpha being captured is overwhelmingly “discount alpha” rather than “liquidity alpha”. We removed the smallest 20% of funds and re-split the remaining funds into quintiles monthly back to 1988, and found that the alpha and other characteristics were similar to the original index data.
Fourth, the study has not been updated through current day. It’s possible that more recent market conditions and events would impact both the University of Oregon study and our hypothetical back-tested study, if the data period was expanded to current day. However, we don’t believe it would create any material changes or lead us to draw significantly different conclusions.