Investor Sentiment and U.S. Presidential Elections*

Transcription

1 Investor Sentiment and U.S. Presidential Elections* Carlos A. Colón-De-Armas University of Puerto Rico Río Piedras Javier Rodríguez University of Puerto Rico Río Piedras Herminio Romero Univesity of Puerto Rico - Carolina * Acknowledgements: We appreciate the comments of Gustavo Grullon and Jim Brau. In addition, we appreciate the feedback of workshops participants at the 2015 Eastern Finance Meeting and University of Puerto Rico. We also thank Ricardo Marrero for excellent research assistance. Contact author: Javier Rodríguez, University of Puerto Rico, Río Piedras Campus, Graduate School of Business, P.O. Box 23332, San Juan PR 00931; telephone: Tel. (787) , ext & 87149; javier.rodriguez19@upr.edu.

2 Investor Sentiment and U.S. Presidential Elections Abstract This paper examines the shifts in investor sentiment around the last seven U.S. presidential elections (1988 through 2012). Investor sentiment is measured by changes in closed-end funds discounts, and the results are corroborated with three robustness tests, including an alternate measure of investor sentiment obtained from the survey conducted by the American Association of Individual Investors. Closed end funds discounts are significantly diminished from two weeks before a U.S. presidential election to a week before the election, and persist until the week after the election, suggesting an increase in investors optimism during that period, particularly when a Democrat is elected president. More than the particular party prevailing, however, investors appear to be more interested in avoiding the entrenchment of power since the results suggest that they become optimistic when a change in the ruling party takes place, but become pessimistic when there is power continuity in the White House. The increase in investor optimism that is observed around the time of U.S. presidential elections is not replicated during non-election years, which seems to corroborate that the elections are indeed driving the results. Keywords: investor sentiment; investment companies; closed end funds discounts; stock returns; stock prices; market efficiency; presidential-election cycle; U.S. politics JEL Classifications: G11, G12, G14, G23 2

3 1. Introduction Over the years, researchers and practitioners have examined whether the political process and the stock market are interrelated. As part of those research efforts, one of the most commonly analyzed issues is whether the results of the U.S. presidential elections have any impact on stock market returns or, alternatively, whether the market prefers a Republican or a Democrat as president. However, an issue not yet resolved is whether investor sentiment is influenced by the election of the U.S. president. This paper addresses this issue. Niederhoffer et al. (1970) and Riley and Luksetich (1980) find that U.S. stock market price increases are more often associated with the election of a Republican president, and they discover that the opposite is the case when a Democrat prevails. This apparent preference for the election of a Republican president would seem to suggest that stock market returns are higher during a Republican administration than during the years when a Democrat is in office. The evidence, however, does not appear to support that conclusion. For instance, Niederhoffer, et al. find no significant differences between stock market returns during Republican and Democratic presidential administrations, except for the third year of a president s term in office, in which they find a tendency for higher stock prices when a Democrat occupied the White House as opposed to a Republican. Huang and Schlarbaum (1982) and Huang (1985), however, find higher stock market returns during Democratic presidencies. Several authors attempt to explain this dichotomy between the apparent stock market preference for a Republican president, given the fact that returns appear higher when a Democrat is in office, from a rational economics perspective. For Hensel and Ziemba (1995) and Johnson et al. (1999), the higher stock market returns during Democratic administrations are a manifestation of the small-firm effect, since they find that only small firms have significantly higher 3

4 returns under Democratic presidencies, while the returns for large firms are not statistically different across the administrations of both major parties. Santa-Clara and Valkanov (2003) find higher returns under Democratic than Republican presidencies using both, a value-weighted and an equal-weighted index, but the results are more significant for the latter, which is more influenced by small firms. Sy and Zaman (2011), on the other hand, attribute the higher stock market returns under Democrats than Republicans to risk. Finally, Campbell and Li (2004) and Powell et al. (2007), after adjusting for methodological procedures, find no significant differences in returns between Democratic and Republican presidencies. Some researchers have taken a different approach to study the relationship between politics and the behavior of investors. For instance, Allvine and O Neill (1980), Huang and Schlarbaum (1982), Herbst and Slinkman (1984), Colón De Armas (1984), Huang (1985), Hensel and Ziemba (1995), Gärtner and Wellershoff (1995), Foerster and Schmitz (1997), Johnson et al. (1999), Booth and Booth (2003), Beyer et al. (2008), and Colón De Armas (2013) all find evidence suggesting that the returns of common stocks, at least in the United States, tend to follow a four-year cycle that seems to correlate with the presidential elections. The form of such cycle is such that common stocks provide, on average, higher returns during the last two years of a U.S. president s term in office than in the first two years. Bohl and Gottschalk (2006), however, find that the Democratic premium and the four-year election cycle are not pervasive around the world and are, at best, limited to the U.S. and only a few other countries. Others tried to find economic and political explanations to these apparent relationships between stock market returns and politics. For example, Stangl and Jacobsen (2007) could not find particular industries in which stock market returns were higher under Democratic rather than Republican presidents, or which exhibited a four-year cycle, but only find evidence of these two 4

5 effects for the market as a whole. Therefore, they suggest that the explanation for their existence must be found at the macroeconomic level. Li and Born (2006), Bialkowski et al. (2008), Boutchkova et al. (2012), and Goodell and Vähämaa (2013) find that elections may increase uncertainty and the risk of stock market investments. Pantzalis et al. (2000) analyze international data and find positive abnormal returns during the two weeks before an election, but depending on the institutional aspects of the country and on the likelihood of the incumbent being re elected. Belo et al. (2013) relate political cycles to government spending. On the contrary, Sturm (2013) argue that it is more probable that stock returns influence economic policy rather than the other way around, although he observes that the four-year cycle may be the result of tax legislation. Ferguson and Witte (2006) find that U.S. stock returns are lower, and volatility is higher, when the U.S. Congress is in session. Wisniewski (2015) offers an extensive literature review on some of these topics. In essence, this body of literature examines whether political decisions affect the fundamental factors that affect stock prices, or vice versa. A topic that has received scan attention, however, is the effect, if any, of political events on investor sentiment. Sentiment is defined by Merriam-Webster s Dictionary ( as an attitude, thought, or judgment prompted by feeling. From merely reading that definition, it seems obvious that sentiment has always been an integral part of finance in general, and investments in particular. Nevertheless, our profession did not always recognize the importance of incorporating into our analyses concepts and ideas that, like sentiment, come from other social sciences, particularly sociology and psychology. In fact, until relatively recently, the boundary between traditional finance and behavioral finance, as Wurgler (2012) indicates, was rather fuzzy. A few years earlier, however, Shiller (2003) 5

6 summarized the developments in the field and demonstrated that, at least since the 1990s, the two have become more intertwined. As part of this literature, a significant number of studies, many of which are discussed by Baker and Wurgler (2007), suggest that investor sentiment affects stock prices. Although the best measure of investor sentiment is still an open question, Lee et al. (1991) argue that closed-end funds discounts reflect individual investors sentiment, a finding criticized by Chen et al. (1993a,b), but reaffirmed by Chopra et al. (1993a,b). This debate notwithstanding, previous studies have used closed-end funds discounts from net asset values as a reliable sentiment measure and to assess changes on small investor sentiment in response to significant events. For example, Bodurtha et al. (1995) use closed-end fund s discounts to conclude that the share prices of closed-end country funds may be influenced by U.S. investor sentiments. Klibanoff et al. (1998) use discounts on closed-end funds to study the reaction to country-specific news. Burch et al. (2003) use the events around September 11, 2001 to provide more evidence in favor of the hypothesis that closed-end funds discounts reflect the sentiment of small investors, and Anderson et al. (2013) confirm it using an implied market volatility index. Chan et al. (2008), Hwang (2011), and Lutz (2015) are all more recent examples of the use of closed-end fund discounts as a metric of investor sentiment. Accordingly, we examine shifts on investor sentiment around U.S. presidential elections, as measured by changes in closed-end funds discounts, to further enhance the literature on investor sentiment, and to expand the analysis of the relationship between political events and the stock market. In doing so, we are conscious of the fact that other authors have considered the relation between investor sentiment and the U.S. presidential elections. For instance, Stovall (1992) considers sentiment indirectly using polls and find them to be somewhat useful in forecasting stock 6

7 market returns during the presidential cycle. Kräussl et al. (2014), using the Conference Board survey of consumer confidence and the University of Michigan survey of consumer confidence, find that sentiment cannot explain the four-year presidential election cycle effect. To our knowledge, however, no one has examined the relation between investor sentiment and U.S. presidential elections using closed-end fund discounts as the yardstick for sentiment. We find that closed end funds discounts are significantly diminished from two weeks before the election to a week before the election, and persist until the week after the election, suggesting an increase in investors optimism during that period. It would appear that this increase in optimism is due to the fact that the level of uncertainty regarding the election s outcome is resolved the week before the election when, perhaps, investors realize which presidential candidate is the clear favorite to win. The increase in investors optimism is stronger before and after a Democrat is elected president, which is consistent with the extensive literature documenting higher stock market returns during Democratic as opposed to Republican U.S. presidencies. When a Republican is elected, an initial increase in optimism also is observed two weeks before the election. That optimism, however, begins to disappear perhaps when it becomes more likely that a Republican will be elected, and is completely reversed when that election is confirmed. On closer examination, however, it appears that, more than the particular party prevailing, investors are more interested in avoiding the entrenchment of power since the results suggest that they become optimistic when a change in the ruling party takes place, but become pessimistic when there is power continuity in the White House. This conclusion is confirmed by the results of the 2000 election which, due to its peculiarities, is analyzed separately. 7

8 Finally, we conduct three robustness tests. First, to further corroborate that the results are indeed related to the U.S. presidential elections, we compare investor sentiment during election and non-election years. This comparison indicates that the increase in investor optimism that is observed around the time of U.S. presidential elections is not replicated during the corresponding periods of years in which presidential elections are not held. Therefore, it seems reasonable to conclude that the elections are indeed driving the results. Second, our results are confirmed using a multi-factor regression model. Third, we replicate the analysis using the survey conducted by the American Association of Individual Investors to measure sentiment and find consistent results. The rest of the paper consists of three additional sections. Section 2 describes our methodology and the sample of funds used to conduct this study. Section 3 discusses the empirical results. Finally, Section 4 offers some concluding remarks. 2. Methodology and data Closed-end funds issue a fixed number of shares that investors trade daily at marketdetermined prices that typically differ from each fund s net asset value (NAV). These differences between the prices of the funds shares and their NAVs may be due to the fact that investors may be willing to trade the funds shares at a discount or premium depending on their expectations, or how they feel, about the prospects of the fund. For that reason, and because these differences tend to move together across funds, closed-end fund discounts may serve as a measure of investor sentiment. Our analysis follows closely the methodology used by Burch et al. (2003). Accordingly, we use closed-end funds discounts from NAVs as a proxy for investor sentiment to examine the changes in investor sentiment during the last seven U.S. presidential elections. 8

9 To calculate the discounts, we let Price be the fund s closing market price per share, and NAV the fund s net asset value per share. Then, the deviation of the fund s market price from its NAV is defined as: Premium = Price NAV 1 (1) Since all deviations are defined as premiums, the sign indicates whether the deviation is a premium (positive sign) or a discount (negative sign). We use weekly (Friday-to-Friday) premiums for the two weeks before and the Friday after the election. To illustrate, consider, for example, the 1996 election of President William J. Clinton. That election was held on Tuesday, November 5, Therefore, the weekly premium for two weeks before the election corresponds to the week ended on Friday, October 25, 1996 (labeled Pre 2 on the tables below). The one week before the election ended on Friday, November 1 st, 1996 ( Pre 1 ), and the Friday after the election corresponds to Friday, November 8, 1996 ( Post 1 ). To collect our sample of closed-end funds, for each election examined, we first retrieved all the closed-end funds with daily closing price data available on the database of the Center for Research in Security Prices (CRSP). Then, for the list of funds with data on CRSP, we collected daily NAVs from Bloomberg. Following this procedure, our final sample of funds includes all the closed-end funds with closing price and NAV data for each of the three Fridays for which we compute premiums around each election date. Table 1 contains a summarized description of our sample. It includes, for the seven elections examined, the year of the election, the number of closed-end funds in the sample for that 9

10 specific election, the party of the president who won the election, and whether the election resulted in a change of the party controlling the White House. As may be observed from the distribution presented in Table 1, the number of closed-end funds in the sample goes from a minimum of 48 funds in the 1988 election, to 295 funds in Table 1 also shows that we examine four elections of a Democratic president and three of a Republican president. Finally, our analysis takes into consideration three elections that resulted in a change of ruling party. 3. Empirical results This section presents the results of our analysis of the changes in individual investors sentiment represented by closed-end fund premiums and changes in premiums around the last seven U.S. presidential elections. Since the 2000 election of President George W. Bush was such an extraordinary event, requiring even the intervention of the U.S. Supreme Court to settle the outcome, we do not include it in the aggregate results, but rather devote a separate sub-section to its analysis. 3.1 Premiums Table 2 presents individual investor sentiment levels as measured by the average premiums. Panel A includes the aggregate cross-sectional mean values across all six elections, a total of 1,195 observations (not including the 2000 election), while Panels B through G contain the results for each individual election. In Panel A, we report statistically significant discounts across all three periods ( event time in the table) we examine. The Pre 2 value, which corresponds to the average premium on the Friday two weeks before the election, is -5.0% and this value is significantly different from 10

11 zero at the one percent level. The value for Pre 1, the Friday before the election, is -4.5%, also highly significant. The average value for Post 1, which corresponds to the premium for the Friday after the election, is -4.5%, also significant at the one percent level. In the next seven panels that present the results for each election, with the exception of the 1992 election, the average premiums are all negative and statistically significantly different from zero. For the 1992 election, the average premiums, although positive for each of the three Fridays, are not significantly different from zero. These results, in general, demonstrate that our sample exhibits a general tendency for closed end funds discount. This tendency is consistent with the prevalence of discounts documented for closed end investment companies in the literature. 3.2 Changes in premiums Now, to better understand the sentiment of individual investors around the U.S. presidential elections, we compute three differences in premiums for all seven elections considered: (1) The difference between the Pre 1 premium and the Pre 2 premium; (2) the difference between the Post 1 premium and the Pre 2 premium; and (3) the difference between the Post 1 premium and the Pre 1 premium. In other words, we calculate the difference in premiums between the two Fridays before the election and the differences between the premium on the Friday after the election and the two consecutive Fridays before the election. Panel A of Table 3 presents the aggregate results of the differences in premiums. The difference between the premium of the Friday before the election and the Friday two weeks before the election (Pre 1 Pre 2) is 0.4%. This value is significant at the one percent level. We also report a significant 0.4% difference between the premium on the Friday after the election and the 11

12 Friday two weeks before the election (Post 1 Pre 2). The difference between the premium on the Friday after the election and the Friday before the election (Post 1 Pre 1), however, is not statistically different from zero. These results suggest a positive, and highly significant, increase in investors optimism from two weeks before the election to a week before the election, which persists until the week after the election. It seems that the largest level of uncertainty is present in investors sentiment two weeks before the election and this uncertainty is resolved the week before the election (Pre 1 Pre 2) when, perhaps, they realize which presidential candidate is the clear favorite to win. Panels B through G of Table 3 present the results for the difference in premiums for each election. Consistent with the aggregate results, at the individual election level we find that the Pre 1 Pre 2 difference is always statistically significant (at least at the 10 percent level). This difference, however, is positive in four cases and negative in two. In both elections with Pre 1 Pre 2 negative differences, the Post 1 Pre 2 differences also are negative and statistically significant. Perhaps more interestingly, in these two elections (1996 and 2012) a Democrat prevailed without changing the party in power. Therefore, it behooves us to consider whether these two effects have any bearing on the sentiment of investors. 3.3 Democrat versus Republican Party To consider the effect, if any, of the party of the elected president, we partition the data in two groups and aggregate the premiums by the party of the candidate who won the election. Not considering the 2000 election, we have four elections won by the candidate of the Democratic Party and two by the candidate of the Republican Party. In terms of the number of observations examined, for the elections won by Democrats we have 871 premiums and for the elections in 12

13 which a Republican prevailed we consider 324 premiums. The cross-sectional mean premiums are presented in Panel A of Table 4. Consistent with the above results, regardless of the political party of the elected president, we report significant discounts. Panel B of Table 4 contains the mean cross-sectional differences in premiums partitioned by the political party of the elected president. We find that, when the Democratic candidate is elected, all differences are positive and significantly different from zero. These results suggest that the election of a Democrat as president is accompanied with an increase in investor s optimism. The results for the case when a Republican candidate is elected president are mixed. In that case, the Pre 1 Pre 2 difference is positive and highly significant, while the other two differences are negative, although only the Post 1 Pre 1 difference is significant at the one percent level. These results seem to suggest that the usual tendency for an increase in investor optimism two weeks prior to an election begins to disappear when a Republican is more likely to be elected president (Pre 1) and is completely reversed when that election is confirmed. 3.4 Change in power Next, we consider the effect in the sentiment of investors, if any, if the election of a president also results in a change of the ruling party occupying the White House. This change in power could be the result of a Democrat being elected president when there is a Republican incumbent or when a Republican wins an election when a Democrat sits in the Oval Office. For that purpose, we separate the data in two groups. One group corresponds to the two elections in which a change in power took place (not including the change in power that resulted from the election of George W. Bush in the year 2000) and contains 395 premiums, and the other group relates to the four elections when the ruling party remained unchanged and consists of 13

14 800 premiums. As with the aggregate results, Panel A of Table 5 shows significant mean discounts in both cases, change and no change in the ruling political party. In the case of the cross-sectional differences in premiums, the results presented in Panel B of Table 5 indicate that for the elections in which no change in power takes place all differences are negative, although only Post 1 Pre 2 and Post 1 Pre 1 are significant at the one percent level. On the contrary, when a change in power results from the election of a president, all differences in premiums are positive and significant at the one percent level. These results suggest that investors become optimistic when a change in the ruling party takes place, but become pessimistic when there is power continuity. 3.5 The 2000 election Finally, we consider the election of George W. Bush in the year That election was marked by an intense drama-filled course of events that required a mandatory recount and a litigation process which ended up in the Supreme Court of the United States. As a result, even though the election was held on November 7, 2000, it was on December 12, 2000 that George W. Bush was declared, by the Supreme Court, the winner of the election over the Democratic candidate Al Gore. At the time, according to Mitchell (2000), this election was only the fourth in US history when a presidential candidate was elected without winning the popular vote. Given these extraordinary set of circumstances, we decided to analyze the 2000 election separately from the other elections considered in this study. In addition, since the election was decided by the Supreme Court on a date separate from when the election was held, we consider these two dates as separate events and analyze, not only the days around the election, but also include an analysis of the days around that Supreme Court decision. More specifically, for the 14

15 election date, we compute all the premiums as we did for all the other elections, but we also examine the Pre 1 and Post 1 premiums for the date of the Supreme Court ruling. The results appear in Table 6. Panel A of Table 6 presents the cross-sectional mean premium for each date examined. Regardless of the event, election or Supreme Court ruling, and consistent with the above results, we find significant levels of discounts. Panel B of Table 6 shows the mean differences in premiums. For the election event, we find that, consistent with the election of a Republican candidate, Pre 1 Pre 2 is positive and statistically significant. Moreover, consistent with the aggregate results, a positive Pre 1 Pre 2 difference is followed by a positive Post 1 Pre 2 difference that is also statistically significant, and a non significant Post 1 Pre 1 difference. Regarding the ruling of the Supreme Court, we find a positive, and significant, Post 1 Pre 1 difference. This result is not consistent with an election of a Republican candidate, but it is consistent with a change in ruling party, which took place when George W. Bush was first elected president in the year Non-election years As a robustness test, and to further corroborate that the above results are indeed related to the presidential elections, in this section we examine investor sentiment during election years (not including the 2000 election) in comparison with the sentiment during non-election years. The analysis follows the same approach depicted above for the computation of closed-end funds premiums and changes in premiums. We also consider the same three weekly (Friday-to-Friday) premiums. Since the election date corresponds to the Tuesday after the first Monday of November, 15

16 for non-election years, we define the event times (Pre 2, Pre 1 and Post 1) accordingly. The results appear in Table 7. Panel A of Table 7 shows aggregate average premiums. Regardless of whether it is an election year or not, and consistent with our previous results, we find significant discounts across all three event dates. In the case of Pre 2, the discount is statistically significantly more negative in election years than in non-election years. In the case of Post 1, the discount is statistically significantly less negative in election years than in non-election years. For Pre 1, however, there is no significant difference between the discounts. The most interesting results turn out to be related to the changes in premiums, presented in Panel B of Table 7. We find that the differences in premiums during election years that, in the case of Pre 1 Pre 2 and Post 1 Pre 2, are positive and statistically significant, but not significantly different from zero in the case of Post 1 Pre 1, turn out to be negative and statistically significant in non-election years except for Post 1 Pre 1, which is not statistically significant. More importantly, except for Post 1 Pre 1, the differences in premiums are significantly greater during election years versus non-election years. Given that the increase in investor optimism that is observed around the time of U.S. presidential elections is not replicated during the corresponding periods of years in which presidential elections are not held, it seems reasonable to conclude that the elections are indeed driving the results. 3.7 Multi-factor model To further confirm that the above results are indeed related to the elections, in this section we conduct a regression analysis. To this end, we develop a regression model to explain the role 16

17 of political variables in determining changes in closed-end funds premiums around U.S. presidential elections. The unbalanced panel regression we estimate has difference in premiums as the response variable. The three explanatory variables included in the model are all dummy variables representing the election year, the political party, and the event of a change in power. The results of the estimation of the model are presented in Table 8. Panel A of the table shows the results for all the elections, while Panel B excludes the election of The results of the model are partially in line with our previous findings, particularly as they relate to change in power. There is no relation between Post1 Pre 1 premium difference and the variables included in the model. This is consistent with the fact that closed-end funds discounts diminished the week before the election. On the contrary, the premium differences, Post 1 Pre 2 and Pre 1 Pre 2 are significantly related to changes in the political party in power. These results are robust to the exclusion of the 2000 election. 3.8 An alternative measure of investor sentiment As an additional robustness test, we consider an alternative measure of investor sentiment. For this purpose, we use the survey conducted by the American Association of Individual Investors (AAII). The AAII is a nonprofit investment education organization that conducts a weekly survey among its members. Through this survey, which started in July 1987, the AAII asks its members to classify themselves as bullish, bearish, or neutral on the stock market for the next six months. Fisher and Statman (2000) and Sayim et al. (2013), among others, rely on the AAII survey for their analyses of the sentiment of individual investors. Similar to other studies, we measure sentiment (AAII-premium) as the difference between the percentages of Bulls minus the percentage of Bears. Since the AAII tallies each Thursday the 17

18 responses received during that week that were dated not earlier than the previous two weeks, we can only use two (instead of the original three) moments in time to measure premiums. We use data from two Fridays before the election and the Friday just after the election. Although not reported in tables, we examine AAII-premiums around the last seven presidential elections. We find some interesting results that are consistent with our previous findings. First, sentiment is different during election years, since the average difference in AAII-premiums is 9.91% during election years and only -0.81% during non-election years. Second, the average difference in AAII-premiums is 13.73% when a Democrat is elected, versus 2.82% when a Republican is elected, which suggests that investors are more optimistic when a Democrat president is elected. Finally, and more importantly, investors become much more optimistic when there is a change in power. When a change in power takes place, the average difference in AAII-premiums is 31.25%, but when there is no change this number is -0.75%. 4. Conclusions To further enhance the literature on investor sentiment, and to expand the analysis of the relationship between political events and the stock market, we examine shifts on investor sentiment around the last seven U.S. presidential elections (1988 through 2012) as measured by changes in closed-end funds discounts. Consistent with prior findings in the literature, our sample exhibits a general tendency for the prices of the shares of closed end funds to sell at a discount from their net asset values. These discounts are significantly diminished from two weeks before the election to a week before the election, and persist until the week after the election, suggesting an increase in investors optimism during that period. 18

19 To the extent that the sentiment of individual investors is a function of political uncertainty, it would appear that this increase in optimism is due to the fact that the level of uncertainty regarding the election s outcome is resolved the week before the election when, perhaps, investors realize which presidential candidate is the clear favorite to win. In that sense, our results are consistent with Goodell and Vähämaa (2013) who find that, although presidential elections engender market anxiety, investors also form and revise expectations around these events. The increase in investor s optimism is stronger before and after a Democrat is elected president, which is consistent with the extensive literature documenting higher stock market returns during Democratic as opposed to Republican U.S. presidencies. When a Republican is elected, an initial increase in optimism also is observed two weeks before the election. That optimism, however, begins to disappear perhaps when it becomes more likely that a Republican will be elected, and is completely reversed when that election is confirmed. On closer examination, however, it appears that, more than the particular party prevailing, investors are more interested in avoiding the entrenchment of power since the results suggest that they become optimistic when a change in the ruling party takes place, but become pessimistic when there is power continuity in the White House. This conclusion is confirmed by the results of the 2000 election which, due to its peculiarities, is analyzed separately. Our results also are consistent with those of Pantzalis et al. (2000) who find positive abnormal returns during the two weeks before an election particularly if lost by an incumbent, or won by the opposition, in a not so free country. Finally, as a robustness test, and to further corroborate that the above results are indeed related to the U.S. presidential elections, we compare investor sentiment during election and non-election years. This comparison indicates that the increase in investor optimism that is 19

20 observed around the time of U.S. presidential elections is not replicated during the corresponding periods of years in which presidential elections are not held. Therefore, it seems reasonable to conclude that the elections are indeed driving the results. Finally, our findings are partially corroborated with the results from the estimation of an unbalanced panel data model and with the use of an alternate measure of investor sentiment obtained from the survey conducted by the American Association of Individual Investors. 20

21 References Allvine, F. and O Neill, D. (1980), Stock Market Returns and the Presidential Election Cycle, Financial Analysts Journal, Vol. 36 No. 5, pp Anderson, S., Beard, T., Kim, H. and Stern, L. (2013), Fear and Closed-End Fund Discounts, Applied Economics Letters, Vol. 20 No. 10, pp Baker, M and Wurgler, J. (2007), Investor Sentiment in the Stock Market, Journal of Economic Perspectives, Vol. 21 No. 2, pp Belo, F., Gala, V. and Li, J. (2013), Government Spending, Political Cycles, and the Cross Section of Stock Returns, Journal of Financial Economics, Vol. 107 No. 2, pp Beyer, S., Jensen, G. and Johnson, R. (2008), The Presidential Term, Journal of Portfolio Management, Vol. 34 No. 2, pp Bodurtha, J., Kim, D. and Lee, C. (1995), Closed-end Country Funds and U.S. Market Sentiment, Review of Financial Studies, Vol 8 No. 3, pp Bohl, M. and Gottschalk, K. (2006), International Evidence on the Democrat Premium and the Presidential Cycle Effect, North American Journal of Economics & Finance, Vol. 17 No. 2, pp Bialkowski, J., Gottschalk, K. and Wisniewski, T. (2008), Stock Market Volatility Around National Elections, Journal of Banking and Finance, Vol. 32, pp Booth, J. and Booth, L. (2003), Is Presidential Cycle in Security Returns Merely a Reflection of Business Conditions?, Review of Financial Economics, Vol. 12 No. 2, pp Boutchkova, M., Doshi, H., Durnev, A. and Molchanov, A. (2012), Precarious Politics and Return Volatility, Review of Financial Studies, Vol. 25 No. 4, pp Burch, T., Emery, D. and Fuerst, M. (2003), "What Can Nine Eleven Tell Us about Closed-end Fund Discounts and Investor Sentiment?", Financial Review, Vol. 38 No. 4, pp Campbell, S. and Li, C. (2004), Alternative Estimates of the Presidential Premium, working paper, Finance and Economics Discussion Series, Federal Reserve Board, Washington, D. C. Chan, J., Jain, R. and Xia, Y. (2008), "Market segmentation, liquidity spillover, and closed-end country fund discounts", Journal of Financial Markets, Vol. 11 No. 4, pp Chen, N., Kan, R. and Miller, M. (1993a), Are the Discounts on Closed-End Funds a Sentiment Index?, Journal of Finance, Vol. 48 No. 2, pp Chen, N., Kan, R. and Miller, M. (1993b), A Rejoinder, Journal of Finance, Vol. 48 No. 2, pp

22 Chopra, N., Lee, C., Shleifer, A. and Thaler, R. (1993a), Yes, Discounts on Closed-End Funds Are a Sentiment Index, Journal of Finance, Vol. 48 No. 2, pp Chopra, N., Lee, C., Shleifer, A. and Thaler, R. (1993b), Summing Up, Journal of Finance, Vol. 48 No. 2, pp Colón De Armas, C. (1984), The Presidential-Election Cycle in the Stock Market. Ph.D. thesis, Purdue University.. (2013), "The Presidential-Election Cycle in the Stock Market: The End of an Anomaly?", Journal of International Finance and Economics, Vol. 13 No. 1, pp Ferguson, M. and Witte, H. (2006), Congress and the Stock Market, working paper. Available at SSRN: Fisher, K. and Statman, M. (2000), "Investor sentiment and stock returns", Financial Analysts Journal, Vol. 56 No. 2, pp Foerster, S. and Schmitz, J. (1997), The Transmission of U. S. Election Cycles to International Stock Returns, Journal of International Business Studies, Vol. 28 No. 1, pp Gärtner, M. and Wellershoff, K. (1995), Is There an Election Cycle in American Stock Returns, International Review of Economics & Finance, Vol. 4 No. 4, pp Goodell, J. and Vähämaa, S. (2013), U.S. Presidential Elections and Implied Volatility: The Role of Political Uncertainty, Journal of Banking and Finance, Vol. 37, pp Hensel, C. and Ziemba, W. (1995), United States Investment Returns During Democratic and Republican Administrations, , Financial Analysts Journal, Vol. 51 No. 2, pp Herbst, A. and Slinkman, C. (1984), Political-Economic Cycles in the U.S. Stock Market, Financial Analysts Journal, Vol. 40 No. 2, pp Huang, R. (1985), Common Stock Returns and Presidential Elections, Financial Analysts Journal, Vol. 41 No. 2, pp and Schlarbaum, G. (1982), Asset Returns and Presidential Elections, working paper, Krannert Graduate School of Management, Purdue University. Hwang, B. (2011), "Country-specific sentiment and security prices", Journal of Financial Economics, Vol. 100 No. 2, pp Johnson, R., Chittenden, W. and Jensen, G. (1999), Presidential Politics, Stocks, Bonds, Bills, and Inflation, Journal of Portfolio Management, Vol. 26 No. 1, pp Klibanoff, P., Lamont, O. and Wizman, T. (1998), "Investor Reaction to Salient News in Closed End Country Funds", Journal of Finance, Vol. 53 No. 2, pp

23 Kräussl, R., Lucas, A., Rijsbergen, D., Van der Sluis, P. and Vrugt, E. (2014), Washington meets wall street: A closer examination of the presidential cycle puzzle, Journal of International Money and Finance, Vol. 43, pp Lee, C., Shleifer, A. and Thaler, R. (1991), Investor Sentiment and the Closed-End Fund Puzzle, Journal of Finance, Vol. 46 No. 1, pp Li, J. and Born, J. (2006), Presidential Election Uncertainty and Common Stock Returns in the United States, Journal of Financial Research, Vol. 29 No. 4, pp Lutz, C. (2015), "The impact of conventional and unconventional monetary policy on investor sentiment", Journal of Banking & Finance, Vol. 61, pp Mitchell, A. (December 14, 2000), The 43 rd President George Walker Bush, The New York Times, p. A25. Niederhoffer, V., Gibbs, S. and Bullock, J. (1970), Presidential Elections and the Stock Market, Financial Analysts Journal, Vol. 26 No. 2, pp Pantzalis, C., Strangeland, D. and Turtle, H. (2000), Political Elections and the Resolution of Uncertainty: The International Evidence, Journal of Banking and Finance, Vol. 24, pp Powell, J., Shi, J., Smith, T. and Whaley, R. (2007), The Persistent Presidential Dummy, Journal of Portfolio Management, Vol. 33 No. 2, pp Riley, W. and Luksetich, W. (1980), The Market Prefers Republicans: Myth or Reality, Journal of Financial and Quantitative Analysis, Vol. 15 No. 3, pp Santa-Clara, P. and Valkanov, R. (2003), The Presidential Puzzle: Political Cycles and the Stock Market, Journal of Finance, Vol. 58 No. 5, pp Sayim, M., Morris, P. and Rahman, H. (2013), "The effect of US individual investor sentiment on industry-specific stock returns and volatility", Review of Behavioural Finance, Vol. 5 No.1, pp Shiller, R. (2003), From Efficient Markets Theory to Behavioral Finance, Journal of Economic Perspectives, Vol. 17 No. 1, pp Stangl, J. and Jacobsen, B. (2007), Political Cycles in US Industry Returns, Journal of International Finance and Economics, Vol. 5, pp Stovall, R. (1992), Forecasting stock market performance via the presidential cycle, Financial Analysts Journal, Vol. 48 No. 3, pp Sturm, R. (2013), Economic policy and the presidential election cycle in stock returns, Journal of Economics and Finance, Vol. 37 No. 2, pp

24 Sy, O. and Zaman, A. (2011), Resolving the Presidential Puzzle, Financial Management, Vol. 40 No. 2, pp Wisniewski, T. (2015), Is There a Link between Politics and Stock Returns? A Literature Survey, working paper, University of Leicester School of Management. Available at SSRN: Wurgler, J. (2012), Introduction: A special issue on investor sentiment, Journal of Financial Economics, Vol. 104 No. 2, p

25 Table 1. Sample description Year No. funds Winning Party Change in power R No D Yes D No R Yes R No D Yes D No Notes: This table presents a summary of the US presidential elections we examine and the number of closed-end funds in the samples. R = Republican and D = Democrat. 25

26 Table 2. Premiums during US presidential elections No. of obs. Event Time Mean Median Std deviation t ratio Panel A: Aggregate 1195 Pre *** 1195 Pre *** 1195 Post *** Significa nce Panel B: Election year Pre *** 48 Pre ** 48 Post *** Panel C: Election year Pre Pre Post Panel D: Election year Pre *** 185 Pre *** 185 Post *** Panel E: Election year Pre *** 276 Pre *** 276 Post *** Panel F: Election year Pre *** 295 Pre *** 295 Post *** Panel G: Election year Pre ** 291 Pre *** 291 Post *** Notes: This table presents individual investor sentiment levels as the cross-sectional mean deviation (premium) of closed-end funds net asset value from market value. Panel A shows the aggregate value across all six elections. Panels B through G show the results for each individual election. t ratio is a standard t-test with the null hypothesis that the mean premium is zero. ***, **, * represent statistical significance at the 1, 5, and 10 percent level, respectively. 26

27 Table 3. Difference in premiums during US presidential elections No. of obs. Premium Change Mean Median Std deviation t ratio Panel A: Aggregate 1195 Pre 1 - Pre *** 1195 Post 1 - Pre *** 1195 Post 1 - Pre Significa nce Panel B: Election year Pre 1 - Pre ** 48 Post 1 - Pre Post 1 - Pre Panel C: Election year Pre 1 - Pre * 100 Post 1 - Pre ** 100 Post 1 - Pre Panel D: Election year Pre 1 - Pre *** 185 Post 1 - Pre *** 185 Post 1 - Pre *** Panel E: Election year Pre 1 - Pre *** 276 Post 1 - Pre Post 1 - Pre *** Panel F: Election year Pre 1 - Pre *** 295 Post 1 - Pre *** 295 Post 1 - Pre *** Panel G: Election year Pre 1 - Pre *** 291 Post 1 - Pre *** 291 Post 1 - Pre Notes: This table presents changes in individual investor sentiment levels as the differences in the deviation (premium) of closed-end funds net asset value from market value. Panel A shows the aggregate value across all six elections. Panels B through G show the results for each individual election. t ratio is a standard t-test with the null hypothesis that the mean difference in premiums is zero. ***, **, * represent statistical significance at the 1, 5, and 10 percent level, respectively. 27

28 Table 4. Political party effect Panel A: Premiums Party No. of obs. Event Time Mean Median Std deviation t ratio D 871 Pre *** Significa nce D 871 Pre *** D 871 Post *** R 324 Pre *** R 324 Pre *** R 324 Post *** Panel B: Premium Changes Party No. of obs. Premium Change Mean Median Std deviation t ratio D 871 Pre 1 - Pre *** Significa nce D 871 Post 1 - Pre *** D 871 Post 1 - Pre * R 324 Pre 1 - Pre *** R 324 Post 1 - Pre R 324 Post 1 - Pre *** Notes: This table shows the cross-sectional mean premium (Panel A) and change in premium (Panel B) around US presidential elections. The analysis considers the political party (R = Republican and D = Democrat) of the elected candidate. t ratio is a standard t-test with the null hypothesis that the mean premium (Panel A) is zero and that the mean difference in premiums (Panel B) is zero. ***, **, * represent statistical significance at the 1, 5, and 10 percent level, respectively. 28

29 Table 5. Change in ruling party effect Panel A: Premiums Change in power No. of obs. Event Time Mean Median Std deviation t ratio No 800 Pre *** Significa nce No 800 Pre *** No 800 Post *** Yes 395 Pre *** Yes 395 Pre *** Yes 395 Post *** Panel B: Premium Changes Change in power No. of obs. Premium Change Mean Median Std deviation t ratio No 800 Pre 1 - Pre Significa nce No 800 Post 1 - Pre *** No 800 Post 1 - Pre *** Yes 395 Pre 1 - Pre *** Yes 395 Post 1 - Pre *** Yes 395 Post 1 - Pre *** Notes: This table shows the cross-sectional mean premium (Panel A) and change in premium (Panel B) around US presidential elections. The analysis considers changes in party (R = Republican and D = Democrat) ruling. t ratio is a standard t-test with the null hypothesis that the mean premium (Panel A) is zero and that the mean difference in premiums (Panel B) is zero. ***, **, * represent statistical significance at the 1, 5, and 10 percent level, respectively. 29

30 Table 6. The 2000 election Panel A: Premiums Event Type No. of obs. Event Time Mean Median Std deviation t ratio Election Process 190 Pre *** Significa nce Election Process 190 Pre *** Election Process 190 Post *** Supreme Court 190 Pre *** Supreme Court 190 Post *** Panel B: Premium Changes Event Type No. of obs. Premium Change Mean Median Std deviation t ratio Election Process 190 Pre 1 - Pre *** Significa nce Election Process 190 Post 1 - Pre *** Election Process 190 Post 1 - Pre Supreme Court 190 Post 1 - Pre ** Notes: This table shows the cross-sectional mean premium (Panel A) and change in premium (Panel B) around the 2000 election of George W. Bush. The analysis considers both, the election date and the Supreme Court ruling. t ratio is a standard t-test with the null hypothesis that the mean premium (Panel A) is zero and that the mean difference in premiums (Panel B) is zero. ***, **, * represent statistical significance at the 1, 5, and 10 percent level, respectively. 30