# Statistical arbitrage in closed-end funds

Sometimes, it is cheaper to buy a basket of assets than it is to buy the assets in the basket. In this post, we discuss closed-end funds and why they often trade at a discount to their net asset value. Furthermore, we explore whether this could be the basis for an algorithmic trading strategy.

## Closed-End Funds vs ETFs

Most of us are probably familiar with exchange-traded funds (ETFs) – baskets of securities where the basket itself is tradable. The ETF industry has grown rapidly over the past few decades because ETFs provide a low-cost way to get exposure to a certain market, sector, or strategy.

In this post, we will instead be discussing closed-end funds (CEFs). On the surface, these may seem similar to ETFs: CEFs are also baskets of securities, and they can also be listed and traded on an exchange. A pedant might point out that by definition, a CEF is an ETF since it is a fund that trades on an exchange – this is unfortunately not correct, as there is a true conceptual distinction.

We often like to think of ETFs as upscaled equivalents of a personal trading portfolio. One gives money to a manager, who in return issues a share of the ETF to represent the ownership and invests that money in a basket of assets. This mental model is, in fact, a much better description of CEFs. ETFs are a little bit more complicated. For both ETFs and CEFs, an important point is that the share price is not “naturally” guaranteed to be equal to the value of all of the securities that the fund owns, i.e the net asset value (NAV). This is because the share price of the fund is determined by supply and demand – if everyone in the world suddenly wants to buy the shares of a fund, the share price can shoot up independently of the value of the underlying assets. However, ETFs have an interesting mechanism for ensuring that these deviations between the price and NAV are very short-lived.

ETF fund managers are seldom the people who buy and sell the securities. They instead turn to authorised participants (APs), typically investment banks, who transact in the underlying securities. The fund manager then issues shares of the ETF to the AP, in return for the securities that were bought. This process is known as creation because shares of the ETF are being issued. The AP is then free to hold these shares, or more often, trade them on a stock exchange. However, the AP’s role goes far beyond the initial purchase of the securities to set up the ETF. They play a critical role in ensuring that the ETF share price never strays too far from the NAV. Any momentary supply-demand imbalances can be arbitraged away – for example, if there is suddenly less demand for the ETF, and it trades at a discount to the NAV, the AP can buy ETF shares on the open market and redeem them from the ETF fund manager. This locks in an arbitrage profit and drives the ETF price back towards the NAV.

This is why ETFs are open-end funds: the fund manager can issue/redeem however many shares they want. Conversely, CEFs issue a fixed number of shares when they first IPO, and there is no subsequent creation or redemption. As a result, we might expect there to be a larger spread between the CEF price and the NAV (compared with an ETF’s price and its NAV).

Although CEFs don’t have the same creation and redemption mechanism as ETFs, there is still a fundamental link between the NAV and the price.

## Conclusion

There is clearly a lot more exploration that needs to be done – we have only looked at the spread for a single fund, and have chosen a simplistic information coefficient analysis rather than conducting a full backtest. Nevertheless, the results thus far have been quite encouraging. In addition to the reasonably high information coefficient, we have a clear economic hypothesis about why the discount to NAV may be a persistent predictive factor. In a recent MacroVoices podcast, Eric Peters shared a great insight (emphasis is mine):

Our approach to anything that we ever do in markets, is to first ask the question: why do you get paid to do something?

Without this economic understanding, it’s much more likely that any pattern you’ve discovered is a statistical spectre that haunts the historical data but will disappear in the daylight of future data.

To answer the question in the introduction, I do think that this could form the basis of a trading strategy and hope that over the next few weeks I will have a chance to investigate further.