Editor's note: Seeking Alpha is proud to welcome Alexander Halliday as a new contributor. It's easy to become a Seeking Alpha contributor and earn money for your best investment ideas. Active contributors also get free access to the SA PRO archive. Click here to find out more »

### Abstract

Closed-end funds (CEFs) present a unique opportunity to study finance in that the price of shares rarely matches the net value of the underlying holdings. This study investigates this phenomenon and how behavioral finance influences the discount. The study uses time-series regression analysis to examine the CNN Fear and Greed Index and its relationship with CEF discounts over time. Results of vector autoregression (VAR) and Granger-causality testing indicate that investor sentiment does not significantly influence the price of CEFs in relation to their net assets. Autoregressive conditional heteroskedasticity (ARCH) models show that the volatility of investor sentiment does not significantly influence the volatility of the CEF discount, but there is an autoregressive component to movements in the discount. ARIMA models show that the time series functions of investor sentiment and the CEF are cointegrated, but with no significant causation. In addition, the impact of herding on the CEF discount was evaluated using simple linear regression to show that the herding behavior of investors toward CEFs with higher distribution yields causes the fluctuation in the CEF discount. Investors may use this information to better understand the relationship between investor sentiment, herding, and the valuation inefficiencies of closed-end funds.

**Background of the Study**

It is well understood that investors act in their own best interest when allocating capital. The primary goal of investing is to maximize return. The field of finance has developed to help individuals assess the risks and returns associated with investing decisions, and to provide a set of tools to allow individuals and managers to make the best decisions possible with the information at hand. Yet, history shows us that human behavior is not always rational. Human beings respond to external forces and internal feelings and make decisions which are seemingly irrational.

The growing field of Behavioral Finance picks up where traditional financial theory ends. Behavior finance explores the realm of finance that is more heavily dependent on human feelings and patterns of thought. These human feelings influence investor behavior when assessing the value of commodities like stocks. Equity markets provide a ready-made crucible to study the elements of human behavior to gain insight into the human factors that affect investor behavior. Within the broad equity markets, the specialty market in closed-end funds (CEFs) presents an opportunity to measure the forces of irrational human behavior to determine how they influence the financial markets. CEFs are unique in that each fund can be valued by the market in two different ways. In a perfectly rational market, the two valuation methods would always yield the same price. In the CEF market, the two valuation methods almost always yield two different prices for precisely the same good. This anomaly is known as the "Closed-End Fund Puzzle" and is the subject of this study.

CEFs share some similarities with mutual funds and exchange-traded funds (ETFs) in that they are managed pools of investor cash that are invested in stocks or bonds. CEFs and mutual funds are investment vehicles with an underlying portfolio with a net asset value. CEFs differ from mutual funds in that CEF shares are traded like stocks on the New York Stock Exchange. The price of a mutual fund share is calculated after the close of the equity markets based on the value of the fund's portfolio of assets. The share price for a CEF is established by the market. The price of a share of a CEF only rarely equals the net asset value of the CEF. Many ETFs are structured to track an underlying index. ETFs have a mechanism called a redemption feature which keeps the share price aligned with the net asset value through a process of creating and eliminating shares, while CEFs have a fixed number of shares.

When an ETF manager wants to create additional shares of its fund in order to generate additional capital or meet increased demand for shares, it turns to an entity known as an authorized participant (AP). An AP is a large financial institution or a market maker. The AP acquires the individual stocks that the ETF wants to hold in its portfolio. For example, in the case of the SPDR S&P 500 ETF (NYSEARCA:SPY) which tracks the S&P 500 Index, the agency acting as AP will buy shares in all 500 of the stocks that compose the S&P 500 in the same proportions as the index, and then provide those newly purchased shares to the ETF manager. ETFs that track the S&P 500 index may do so either based on market capitalization like the SPDR S&P 500 ETF or by equal weighting of shares like the Guggenheim S&P Equal Weight ETF (NYSEARCA:RSP). In exchange for the shares of stocks, the ETF manager gives the AP a block of 50,000 ETF shares, known as a creation unit. The AP exchanges a set amount of the underlying stocks and receives the same value in ETF shares in return. These exchanges benefit both parties: The ETF provider receives the stocks it needs to fill the portfolio with the proper index, and the AP receives ETF shares which it can then resell for a profit.

The process can also function in the reverse direction. APs can also function to remove ETF shares from the market. To do this, the AP would purchase a 50,000 share tranche of an ETF and deliver those shares to the ETF fund manager in exchange for the stocks that comprise the underlying index of the fund. The AP-enabled creation and redemption process is essential for ETFs in several ways. On the one hand, it keeps ETF share prices aligned with the fund's underlying net asset value (NAV). With many APs watching most ETFs and competing for the arbitrage profit, ETF prices typically are kept close to the value of their underlying securities. This is one of the essential traits in which ETFs are different from CEFs. No one can create or redeem shares in CEFs, except under very strict circumstances. That is why closed-end fund prices can decouple from their NAV: There's no AP-driven arbitrage action functioning to balance out short-term supply and demand pressures.

Since CEFs are publicly traded, the price of the CEF fluctuates throughout the trading day. The NAV is the sum of the values of the underlying holdings, and also fluctuates throughout the trading day. The price of the CEF and the NAV are rarely equal and fluctuate with respect to each other. In the majority of cases, the share price of a CEF is lower than the NAV, resulting in a CEF discount. In some cases the share price of a CEF is higher than the underlying holdings, resulting in a premium. Analysts and academics have been debating the cause of the CEF discount for decades. The debate has focused on whether the CEF discount fluctuation is due to irrational investor behavior or whether it can be attributed solely to rational investment decisions based on factoring in the financial implications of specific CEF aspects such as management fees or arbitrage costs.

Classical Finance Theory does not include investor sentiment in market valuations; however, new studies in behavioral finance lend support to the notion that investor sentiment may play a significant role in fixing stock prices. This creates a need to investigate investor sentiment to understand further the relationship between investor sentiment and the prices of stocks. Part of the difficulty in this endeavor is with finding a suitable gauge to represent investor sentiment.

This study delves further into the issue by using a new investor sentiment indicator that has not been assessed before in the literature, but which is a common feature in the financial news. The study also evaluates the impact of an investor behavior known as herding and its role in closed-end funds. The study focuses on the effect of these two elements of investor behavior on the closed-end fund discount. Both studies draw from a data set covering the three-year period from July 2014 to July 2017.

Behavioral Finance tries to explain the behavior of the markets by bringing human psychology and irrational investor biases into the analysis. How do investor sentiment and investor bias influence the price to net asset value discount or premium in closed-end funds? This study seeks to determine the relationship between investor sentiment and herding on the CEF discount cycle. The researcher utilized a composite index made up of seven independent investor sentiment indicators known as the CNN Fear and Greed Index to assess the relationship between investor sentiment and the CEF discount cycle. The study also assessed another aspect of investor bias known as herding.

Herding behavior in investors is not clearly understood, but it has been described as a form of fear - the fear of underperforming a benchmark. This study looks at herding behavior in investors to examine the impact on the closed-end fund premium or discount. Herding refers to a phenomenon where investors simultaneously buy the same assets as other investors are buying. One study investigated the relationship between herding and investor sentiment and found that herding is influenced by investor sentiment in the U.K. equity market. This behavior is more specific than wide-spread buying brought on by greed, or wide-spread selling brought on by fear. Herding occurs under certain circumstances when investors base their decisions on the basis of the action of others, without regard to fundamental analysis. Although this sounds irrational, herding behavior may be considered rational from the point of view of investors concerned with matching the performance of a benchmark. Thus, investors may follow the investment behavior of others due to concern with being "left behind." This is most often seen among investment managers concerned with maintaining or enhancing their reputation as decision makers.

Related to this, the CFA Institute administered a survey in 2015 to select the behavioral bias that affects investment decisions the most. The survey gathered responses from 724 investment advisers from around the world. The study found that herding stood out as the most substantial bias affecting investment decision making. Some markets have more of a tendency to herd than others. The CEF market may be particularly vulnerable to the herding tendency since it represents a market where yield is of almost universal importance. Froot, Scharfstein, and Stein found that herding tends to occur when traders focus on a particular variable over the short term. This is why the researcher investigated whether investors in CEFs focus on changes in yield and if this has an impact on the discount or premium of CEFs. Also referred to a "swarming" or "behavioral convergence", this behavior may persist even in the face of adverse payoff externalities, such as a when a CEF announces a very high distribution that is unsustainable. The announcement itself initiates a cascade of investor behavior that overrides rational analysis in the short term.

The variation of the CEF price in relation to the value of the net asset value cannot be explained by rational investor theory alone. The inexplicable fluctuations in the discount have thus been deemed a "puzzle" in several papers. There are a number of studies that seek to prove that investor sentiment is a factor in the CEF discount, but there are equal number of studies that come to the opposite conclusion. As mentioned earlier, one study suggests that the CEF discount can be used to measure investor sentiment as opposed to the inverse relationship being tested here. The relationship between investor sentiment and the CEF discount cycle remains unclear. There is no general agreement so far on the selection of variables to represent the investor sentiment construct. There is also no explicit agreement as to whether the CEF discount itself is an indicator or a result of investor sentiment.

Both individual and institutional investors seek a tool to isolate investor sentiment on a sliding scale to increase stock market returns. There are seven commonly used investor sentiment indexes used to make up the CNN Fear and Greed Index, but no analysis of the effectiveness of the index has been published. In short, the study of the CEF share price discount to NAV continues to be a puzzle with no generally accepted explanation, and the Fear and Greed Index continues to be a commonly known indicator with no scientific studies to verify its usefulness. This study will determine if this well-known composite index of investor sentiment can help to predict the CEF discount cycle. Although herding has been studied with respect to large- and small-cap stocks, there are no known studies that investigate the relationship between herding behavior and the CEF discount. Thus, the researcher included this specific aspect of investor behavior in this analysis of the closed-end fund puzzle to determine if it can shed any light on this facet of investor behavior.

The Fear and Greed Index can be seen oscillating between a value of zero (extreme fear) and 100 (extreme greed) over time (see figure below). This cycle is subject to rapid fluctuations on a seemingly random basis. The seven indicators are equally weighted and provide broad support for this index as a value representing optimism versus pessimism about the future returns of the market. The cycle does not conform to any seasonal fluctuation and may reverse direction before approaching the limits of the scale. To determine the real world events corresponding to high points and low points of the historical graph, the researcher compared the troughs of the lowest levels with news corresponding to the date of the reading. The low point of the graph in the last quarter of 2014 corresponds to the racial riots that took place in Ferguson, United States. This period also saw the beginning of US involvement in airstrikes against Syria. Not all movements in the FGI correspond to memorable newsworthy events, but they are related to changes in the equity market.

This study will seek to explore the relationship of investor sentiment, investor herding, and the discount of price to net asset value in closed-end funds.

### Method

This study focused on data from the US equity markets over the three-year period from 2014 to 2017. The researcher used statistical analysis to analyze the regression equation. The effect of the Fear and Greed Index was analyzed regarding its influence on the CEF discount. Fifty (50) CEFs were selected as the sample for this study. Although there are currently 531 CEFs listed on the NYSE, this study selected 50 CEFs through a fund screening process. There have been several studies regarding the discount on country-specific CEFs to investigate the effects of arbitrage, currency fluctuations, and interest rates. This study, on the other hand, focuses on investor sentiment in the US, and its impact on CEFs traded on the NYSE. For this reason, country-specific CEFs were screened out of the sample being evaluated in this study. This study screened for the following factors: US or World equity funds, leverage less than 30 percent, and yield higher than seven percent. This screen produced 50 funds to evaluate. The analysis was repeated for each of the 50 unique CEFs to provide a broad range of data robust enough to form a conclusion.

This model used time-series to investigate a causal relationship between investor sentiment and the discount or premium of closed-end funds. The paper utilizes Vector Autoregression (VAR) with Granger-causation, Autoregressive conditional heteroskedasticity (ARCH) models, and Autoregressive Integrated Moving Average (ARIMA) models to assess the relationship between investor sentiment and the CEF discount.

The data utilized in the study was broken down into one-week intervals. The CEF discount was measured for 50 different equity funds traded on the NYSE. Country-specific funds were exempted from the study. The data represents a cross-section of different CEFs to gain an overall view of how investor sentiment impacts the discount rate regardless of the market segment.

*Source: CNN*

The raw data was extracted from the individual charts on the CNN Money and Morningstar websites and then converted to numerical data. The data was tested for normality by observing histograms for normal distribution. Weekly observations over a three-year period generated a sufficiently large same size (N=128) to establish normal distribution.

The researcher used a different statistical treatment when analyzing the influence of herding on the CEF discount. Herding has been defined as "the deviation of investor position change". In behavioral finance, herding behavior has been defined as "mimicking the actions of other investors, which constitute the market consensus". The hypothesis of this second phase of the study is whether investors herd when a CEF increases its distribution. Herding in this manner is characterized as investors buying shares in a CEF that has announced an increase in distribution, whether or not the distribution is supported by the dividends and net capital gains of the investments.

To test this hypothesis, the researcher screened the 50 CEFs to determine occurrences of distribution changes. The CEFs that changed distributions were ranked by dividend yield. The researcher tested the impact of changes in CEF dividend yield on the CEF discount using simple linear regression. The exogenous variable for this test is the direction of the change in distribution (positive or negative). The Ordinary Least Squares method was used to determine the confidence level of the correlation between CEF distribution changes and the change in the discount.

### Results and Discussions

A total of 15 different tests were run on each of the 50 CEF time series, for a total of over 750 test results. The Augmented Dickey-Fuller test showed that the CEF discount time series was only stationary in 28 of the 50 cases. Data must be stationary before using VAR and Granger. Using nonstationary data in time series modeling can lead to spurious correlation. To avoid this pitfall, data must be converted so that it is stationary. Converting a time series to stationary data may be accomplished by using the log of the data or by differencing the data. The CEF discount data is represented as a negative number in the case of a discount, and a positive number in the case of a premium. Hence, using the log of the data cannot be accomplished in all cases. The researcher used differencing to make the data stationary. After taking the first difference of the time series, the Augmented Dickey-Fuller test was again run on the data and all 50 CEF discount time series were found to be stationary.

Once the process of differencing the data was completed, the fitting of the VAR model began. The lag selection was made using the SBIC, HQIC, and the AIC lag order tests on each of the 50 CEFs in the selection. The model used the differenced Fear and Greed Index as the exogenous variable and the differenced CEF discount as the endogenous variable.

The results of the VAR model show that FGI Granger causes the change in the discount only 30% of the cases at the 95% confidence level. This does not support the hypothesis that investor sentiment is the cause of the discount in CEFs.

The effect of FGI on the performance of the S&P 500 index was tested using the same procedure. This test showed that FGI Granger-caused movement in the S&P 500 index at the 99% confidence level. This test verifies that the Fear and Greed Index has value as a gauge of sentiment, and that sentiment does affect a broad market index. It shows that investor sentiment is part of the capital allocation process when it comes to broad market indexes. It would also indicate that stock markets as a whole become less efficient during times of extreme greed and extreme fear among investors. The effect of FGI on the performance of the Philippines Investable Market Index was then tested using the same procedure, but with no indicated causation.

To further investigate the relationship between FGI and the CEF discount, the researcher used an ARCH model to determine if volatility was a factor. The data was tested for the ARCH effect using Engle's ARCH test. 88% of the CEFs showed evidence of an ARCH component. The ARCH model was then fitted for lags using Engle's Lagrange multiplier test. Tests were run using both ARCH and GARCH models to determine the better fit. The ARCH model showed that increased past volatility in FGI did cause an increase in the volatility of the discount at the 95% confidence level, but only in 18% of the cases.

These results indicate that the Fear and Greed Index is not a significant variable in influencing the volatility of the CEF discount. The results do indicate that ARCH is a significant variable in influencing the volatility of the CEF discount. This means that previous weekly volatility of the CEF discount can influence the volatility of the CEF discount.

Finally, the autoregressive nature of both price and discount of CEFs was investigated using an ARIMA model. ARIMA is used for modeling the level of the series. The researcher first used correlograms to fit the model and determined that the ARIMA (1,1,1) was the best form for the data. These numbers indicate that the model is most significant when fitted for one lag of the AR component, single differencing, and one lag of the MA component. This test showed that the Fear and Greed Index, price and the discount of CEFs was each influenced by prior performance, but with no significant interaction between variables. Although the results were negative for the multivariate test using the Fear and Greed Index and the CEF discount, there were significant findings for the univariate ARIMA model for the CEF discount. The ARIMA model was able to produce results at the 95% confidence level in 38% of the cases.

The ARCH model shows us that there is a low level of predictability in shocks in the CEF discount based on past volatility in the Fear and Greed Index and the CEF discount. The ARIMA model shows us that there is a moderate level of confidence that the time series for the data regresses on itself, and that past data for FGI and the CEF discount can be used to model future data, but that past data for FGI does not explain future data for the CEF discount. This finding could be of some use to investors seeking to determine the direction that the CEF discount is going but does not solve the CEF discount puzzle itself.

One of the elements that differentiate closed-end funds from other equity investments is high yield. Traditionally, investors have the choice between bonds and stocks for investing. Historically, the yield on bonds is lower than the dividend from stocks. The difference is the risk premium that investors are assuming by investing in relatively risky stocks. Many exchange-traded funds seek high levels of yield through different equity investment strategies. There are ETFs that invest in companies that have paid out dividends for extended periods of time. There are ETFs that invest in companies that show a history of increasing dividends. One such fund is the Vanguard Dividend Appreciation ETF (VIG), which has a yield of 2.06%. There are ETFs that invest in companies with very high dividends. One such fund is the Vanguard High Dividend Yield ETF (VYM), which has a dividend of 2.99%. Investors seeking higher yield have traditionally turned to bonds. The Bloomberg Barclays High Yield Bond ETF (JNK) has a yield of 5.72%. Investors who seek high yield in equity investments and who are unsatisfied with the dividends of stocks find higher yield opportunities in CEFs.

In utilizing CEFs, investors may invest in funds with stock holdings and still have access to yields as high as 14%. CEFs implement many strategies to produce such yields. Writing of covered calls is one such strategy. Buying stocks just before the distribution date and then selling them after they have paid the dividend is another strategy. Closed-end funds may use leverage to increase their buying power and thus increase their yield. Some funds increase their dividend by subsidizing the income from operations with the return of investor capital. These funds distribute cash above the yield of the underlying holdings to increase yield. This practice, although unsustainable in the long run, produces a higher yield that can entice investors. The researcher tested the relationship between CEF yields and the CEF discount to determine if investors were motivated to purchase shares in CEFs that increased their distribution and to sell shares in CEFs that had decreased their distribution. The researcher sought to test the hypothesis that changes in distributions caused the change in the discount or premium of a CEF. The explanation for this is that investors will buy shares in CEFs that increase their distribution, which increases the price of the CEF without increasing the price of the underlying holdings or net asset value. By increasing the price of the CEF while holding the NAV steady, the discount of the CEF will decrease.

To investigate the relationship between distributions and the discount, the researcher used the same sample of 50 CEFs during the same three-year period used in the analysis of investor sentiment. The 50 CEFs were screened to identify which funds had increased or decreased their distribution. Although some CEFs made no changes to the distribution during the three-year period, other funds made several. During the period of study, there were a total of 27 CEFs which changed their distribution with 60 instances of these funds increasing or decreasing their distribution. For each occurrence, the discount was measured prior to the announcement of the change and again 30 days later. For each occurrence, the researcher measured the change in the distribution rate as positive or negative and the change in the discount as positive or negative. The researcher used a dummy variable to represent the change in the distribution and found correlation at the 99% confidence level as shown in Table 2. This finding supports the hypothesis that an increase in distribution results in a decrease in the discount rate. This, in turn, supports the explanation that investor behavior as represented by an appetite for higher yields causes the CEF discount.

Table 1 Regression results for simple model with dummy variable for distribution change effect on CEF discount

Variable | coef | t-stat | p-value |

Dummy variable | .0583098 | 9.13 | 0.000 |

_cons | -.0226541 | -4.83 | 0.000 |

Number of obs | F | Prob > F | Adj. R-squared |

128 | 83.30 | 0.0000 | 0.3932 |

It was noted during the analysis that even CEFs that did not change their distribution had fluctuations in the discount. These fluctuations in the discount not only occurred in CEFs that made no changes to their distribution but also to CEFs between changes in their distribution. This constant variation implies that something other than the distribution was influencing the discount. Investor sentiment has already been eliminated as the cause of this change. The researcher investigated the theory that even though the CEF may hold its distribution steady, its relative yield may become more or less attractive due to changes in the distributions of other CEFs. In other words, it is not the absolute yield of the CEF that influences the discount, but rather the relative yield of the CEF.

In order to investigate this new theory, the researcher formed the new hypothesis that an increase or decrease in the relative strength of the CEF's yield causes an increase or decrease in the CEF's discount. To test this hypothesis, the researcher again looked at the previously identified 27 CEFs during the same three-year period. For each of the 60 distribution changes identified in the prior test, the researcher identified the yield of all 27 CEFs both prior to the distribution change announcement and again 30 days later. The researcher ranked each CEF by yield both before and after each of the distribution changes.

Simple linear regression was used to test the relationship between the degree of the change in the relative strength of the yield with the change in the discount. The CEFs were ranked from 1 to 27 by yield. The yield was calculated as annual distribution divided by price. The rank of each of the 27 CEFs was calculated before any of the 27 declared a change in distribution, and again 30 days later. The change in the rank of all CEFs was used as the exogenous variable. The endogenous variable was the change in the discount. This was calculated as the change in the same manner as the distribution - both prior to the distribution change announcement and again 30 days later. To simplify the model, the researcher looked at CEFs that changed rank by more than three places and less than nine places. This simplification was done in an effort to reduce the noise of very low fluctuations in relative strength and eliminate any error in calculating yield based on a one-off distribution. This screening process resulted in 128 data points for analysis. The researcher again found a correlation between the change in the relative strength of the yield and the change in the discount at the 99% confidence level as shown in Table 3.

Table 2 Regression results for model with distribution rank and CEF discount

Variable | coef | t-stat | p-value |

rank | .0036473 | 4.41 | 0.000 |

_cons | .0094054 | 2.46 | 0.015 |

Number of obs | F | Prob > F | Adj. R-squared |

128 | 19.44 | 0.0000 | 0.1268 |

To explore the herding aspect further, the CEF distributions that involved a return of capital were eliminated from the database, and the regression was repeated. The results of this model are shown in Table 4. It can be seen that the elimination of distributions that were financed with a return of capital does not significantly alter the results of the findings. This supports the hypothesis that investors chase the high distributions without analyzing the long-term financial implications for the fund. This is further evidence of herding behavior driving the CEF discount.

Table 3 Regression results for the model with distribution rank and CEF discount with the return of capital distributions eliminated

Variable | coef | t-stat | p-value |

rank | .0031825 | 3.27 | 0.001 |

_cons | .008394 | 1.85 | 0.067 |

Number of obs | F | Prob > F | Adj. R-squared |

95 | 10.72 | 0.0015 | 0.0937 |

Finally, the OLS approach was repeated as a multiple linear regression model with the previously discussed rank values and values for the Fear and Greed Index both before and after the distribution analysis. This approach was used to determine if investor sentiment as measured by the Fear and Greed Index contributed to the CEF discount simple regression model. As seen in Table 4, the effect of investor sentiment on the CEF discount is not significant, and the inclusion of this data does not contribute to the model. This finding supports the results found in the time series analysis earlier in this paper.

Table 4 Regression results for model with distribution rank, Fear and Greed Index, and CEF discount

Variable | coef | t-stat | p-value |

rank | .0036472 | 4.39 | 0.000 |

fgi | .0000259 | 0.30 | 0.766 |

_cons | .0078942 | 1.24 | 0.217 |

Number of obs | F | Prob > F | Adj. R-squared |

128 | 9.69 | 0.0001 | 0.1204 |

This finding supports the hypothesis that it is the herding aspect of investor behavior that causes the CEF discount. Herding behavior is not solely influenced by fear and greed, but by a general agreement in behavior independent of purely rational thought. As the relative strength of CEF increases, investors purchase shares without analyzing the other financial factors involved in the CEF price. Similarly, as CEF yield decreases, investors sell shares to move to other choices without analyzing the CEF fundamentals. Since there are thousands of different CEFs trading on the market at any given time and there are constant changes in the distribution rates, there exists a shifting mosaic of relative yield strength that causes investors to shift their positions, and which keeps the CEF discount in a constant state of fluctuation. The results of the multivariate model support the conclusions of the time series model that the Fear and Greed Index does not cause the change in the CEF discount.

### Conclusions

This study shows that stock markets are not entirely efficient. Investor sentiment plays a role in the movement of the broad stock market, which rules out fundamental analysis as the only source for the price of a stock. The study shows that stock markets as a whole become less efficient during times of extreme greed and extreme fear among investors. CEFs represent a unique tool for insight into the valuation of equities due to the fact that they have two different prices at the same time. Since fundamental analysis cannot explain the difference between the price of a closed-end fund and the net asset value of its holdings, further research was required to determine the cause of this phenomenon. This study sought to investigate the relationship between investor sentiment and the movement of the CEF discount.

Although the gap between the price of a closed-end fund and the net asset value of the underlying holdings should either remain constant or not exist at all, it cannot be explained by investor sentiment as measured by the Fear and Greed Index. Investor sentiment cannot be said to affect either the price or the discount of closed-end funds, as it does the general market index. Although time series analysis does show that the past price of a CEF, as well as the past discount, does help to predict future movements of the CEF discount, neither the time series analysis nor linear regression shows that the Fear and Greed Index can be used to predict the CEF discount. The data shows that increased volatility in the Fear and Greed Index leads to increased volatility in the closed-end fund discount. This study ruled out investor sentiment as measured by the seven composite indicators that make up the Fear and Greed Index as the cause of the CEF discount.

Delving further into behavioral finance, this study found a strong relationship between fund dividend distribution announcements and the CEF discount. Linear regression was used to show the direct correlation between investor behavior caused by dividend distributions. Evidence shows that an announcement of a change in distribution leads to a change in the discount.

Further, the change in the distribution yield rank of a CEF was also shown to cause a change in the CEF discount. The CEF discount was shown to change regardless of the fund's individual dividend posture, as money flows moved in the direction of the strongest relative dividend distribution. This represents herding behavior, as investors buy increased yield without due regard to the underlying net asset value of a given fund. This unique form of investor behavior explains the cycle of the CEF discount, as the relative yields of all CEFs rise and fall with respect to each other based on the constant changes in individual CEF distributions.

This paper was presented at the Infinity International Business Research Conference 2018 at Holy Angel University, Angeles City, Philippines.

**Disclosure:** I am/we are long RSP.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it. I have no business relationship with any company whose stock is mentioned in this article.

**Additional disclosure: **I have long positions in JDD, JRS, GGT, GAB, THW, PDI, and RMT. These are not specifically mentioned in the article.