Bachelor thesis Economics and Finance University of Amsterdam
Spillover effects of the U.S.
presidential elections on the
Dutch stock market
Author: L.E. JansenStudent ID: 10755861 Supervisor: Dr P.J.P.M. Versijp
Date: June 2018
Abstract
Past research has provided evidence for a positive relationship between political uncertainty and the stock market, in terms of return and volatility. This study aims to examine the spillovers of the
U.S. presidential elections on the Dutch stock market, contributing to past research by using a GARCH(1,1) framework and by doing so including volatility in the study. This was done for the period from1984 to 2018. No significant evidence in the return of the AEX was found for spillovers of the U.S. presidential election cycle. However, the effects of the election of a Democratic president instead of a Republican and the event of a reelection, were evident in the return of the Dutch stock market. Furthermore, in all 4-year periods, significant abnormal volatility was measured. Although, further examination of the cumulative abnormal volatility shows that it is not related to the U.S. presidential election. This raises the question of what the origin of this volatility is, which results in
Statement of Originality
This document is written by Student Lucie Jansen, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of content
1. Introduction 3
2. Literature review 5
2.1 Politics and uncertainty 5
2.2 Politics and the stock market 5
2.3 Presidential election cycle 5
3. Data and methodology 7
3.1. Data description 7
3.2 Framework & hypotheses 8
4. Results 12
5. Conclusion 15
6. Discussion 16
1. Introduction
According to Bialkowski, Gottschalk and Wisniewski (2008), financial economists are left with a big challenge explaining the fluctuations in the stock market because standard valuation models fail to do so. These standard models only consider dividend and earnings as drivers of volatility, which is insufficient (Shiller, 1981). This means that other causes needed to be identified. According to Pantzalis, Stangeland and Turtle (2000) political events are one of these causes. They state that since these events can lead to changes in monetary and fiscal policy, they can result in big changes in the financial market. Since investors are aware of this, they tend to reconsider their expectations when a political event approaches. In this way, the market responds to new information concerning political decisions. According to Pantzalis et al. (2000), presidential elections in particular, are important events to consider because they allow citizens to (partly) influence economic policies. Additionally, because of the big media attention towards elections, information regarding the strategy of the incumbents is widely spread which enables investors to compose a probability distribution for the possible outcomes of this election.
An extension of the examination of the effect of political events on the stock market is the identification of the U.S. presidential election cycle. This cycle implies that in the four years of a presidency, a pattern in the stock market can be recognized. This election cycle, with regard to stock price in the United States, was examined by Wong and McAleer (2009). They showed that the findings of Stovall (1992), who discovered the cycle, are still valid and are as follows. In the election year the stock market flourishes. This is a consequence of political election strategies in which announcements of favorable tax and spending regulation policies are often included. In the first half of the presidency, stock prices decrease and will hit their bottom for this cycle. In the third and fourth year, they will increase and reach a peak as a consequence of politicians acting with allowing for upcoming elections. This means that in the election year, the stock market flourishes.
In this thesis, the focus lays on the stock market in the Netherlands, the so-called AEX, and its relationship with the U.S. presidential elections. Earlier studies have shown that the AEX is affected by the national elections (Brunner, 2009). However, the big trading relationship between the U.S. and the Netherlands, raises the question if we should account for spillovers from the U.S. in the AEX as well. Namely, the partnership between the two countries is one of the oldest continuous relationships the U.S. has. Since there are no significant trading barriers, they have been big
partners in investing, as well as in export and import. The Netherlands is the 8th American export market of goods and 23rd import partner. The relation is even bigger when looking at it from the Dutch perspective. In 2016, the U.S. was the 5th biggest export country of goods, and the 4th when it comes to export of services. Besides this is the U.S. the biggest foreign direct investor in The Netherlands (“European Union, Trade in goods with USA”, 2018).
Foerster and Schmitz (1997) already provided statistical evidence for a significant effect of spillovers of the U.S. presidential elections on the Netherlands. More precisely, they demonstrated the presence of the U.S. presidential election cycle in the returns of the AEX. Their results exhibit that the return in the Netherlands (and 13 other countries) in year two of the cycle are significantly lower than the combination of returns for the years one, three and four. Because of this, they argue
that the U.S. presidential election cycle is an important factor in predicting not only U.S.- but also international stock returns.
In this thesis, I want to extend the past research by making use of a GARCH(1,1) framework, and by doing so include volatility in the analysis. My aim is to map out the effects of the U.S.
elections on the Dutch stock market. To the best of my knowledge, the effect of the U.S. presidential elections on the AEX has never been tested before with this method. Therefore, the following research question was formed:
“What are the effects of the United States presidential elections on the AEX, in terms of
volatility and returns?”
Among other things, there will be looked at the existence of a U.S. election cycle in the AEX, the presence of a significantly different effect of an elected Republican or Democrat and the effect of a president being re-elected.
The rest of this thesis is constructed as follows. In the literature review section, past literature will be discussed in order to understand and form the hypothesis. In section three, the methodology and the use of data will be explained, wherein section four the results will be presented. Following, in section five, a conclusion for these results will be drawn and finally, a discussion and recommendation for further research will be presented.
2. Literature review
In the following section, past literature will be discussed.
2.1 Politics and uncertainty
The studies of Pantzalis et al. (2000), Li and Born (2006) and Bialkowski et al. (2008) all three acknowledge that presidential elections provoke uncertainty in financial markets. According to the Uncertain Information Hypothesis (UIH), uncertainty can be seen as imperfect information (Brown et al., 1988). Assuming investors are rational, they use all available information in decision making. Not only for decision making in the present but also when forming expectations about the future. With these expectations, uncertainty comes along. The UIH explains the reaction of rational, risk-averse investors when uncertainty arises because unanticipated information is announced. The theory predicts that this response causes both risk and return to increase. Consistent with the UIH, Ozoguz (2009) demonstrates the existence of a negative relationship between uncertainty and asset valuation. In her research, she states that when people get a hold of new information, old beliefs are revised. As a consequence, the volatility of returns changes because a new time-variation in investment opportunities arises.
2.2 Politics and the stock market
Li and Born (2006) examined the effect of political uncertainty on the stock market in the United States. This was done with the use of public opinion polls to capture whether the outcome of the elections was a surprise or not. They state that in most of the past elections, up until the voting itself, the outcome was uncertain. Being uncertain about the winning incumbent results in suspense about regime changes. This causes the stock market volatility and average return to move upwards. An exception arises when there is certainty about the current president being re-elected. In this case, the returns and volatility in the election period do not differ from the nonelection period.
Bialkowski et al. (2007) examined the same relationship in a sample of 27 OECD countries. Using an event-study framework they found that, whilst lots of effort on predicting election
outcomes is made, investors are still surprised by the results. By using a GARCH framework, they concluded that since the surprise element results in uncertainty, peaks in stock volatility take place in the week around the election event. This is a rise in the country-specific component of the volatility. Overall, they concluded a positive relationship between voters’ uncertainty and the volatility.
2.3 Presidential election cycle
As stated before, using an EGARCH intervention analysis, Wong and McAleer (2009) found evidence for the presidential election cycle in the United States. When examining stock prices of the U.S. stock market they found that in the first half of the presidency, the prices tend to fall and finally reach a nadir in the second year. In the third year, the prices start to rise again and reach a peak in the fourth-, or the so-called election year. This pattern is found in the period 1965 - 2003 and is especially evident for cases in which the incumbent is Republican. In his Political Business Cycle
theory, Nordhaus (1975) argues that the peak in the election year is a cause of opportunistic behavior of politicians with the goal of winning voters. Altogether, this suggests that a Republican incumbent actively engages in policy manipulation in order to win a reelection. This Political Business Cycle theory was the groundwork for the research of Booth and Booth (2003), which supports the findings of Wong and McAleer (2009). They state that returns on the U.S. stock market are significantly lower in the first two years than in the last years of the presidency.
3. Data and methodology
The content of the following section was constructed as follows. Firstly, the data used in the analysis is described. The choice for the utilized framework is explained afterward. And finally, the hypothesis and test statistics are displayed.
3.1. Data description
In order to conduct this empirical analysis, the AEX which is the price index for the Netherlands, was used. This was obtained using the Thomson Reuters financial datastream. The AEX is the most widely applied measure of the Dutch stock market and is made up of the 25 largest shares with the highest trading volume listed on the Euronext Amsterdam. In this way, it serves as a reflection of their performance (AEX index [indices.euronext], 2018).Prices were retrieved for the period 1984-2018, on a daily basis.
As stated before, the period 1984-2018 was examined. The following table summarizes information about the U.S. elections which have taken place in this time span.
Table 1
Sample description Election
number Date of election Elected president Main opponent Reelected Party president
Electoral vote winner
1 6-11-1984 Reagan Mondale Yes Republican 525
2 9-11-1988 Bush Dukakis No Republican 426
3 3-11-1992 Clinton Bush No Democrat 370
4 5-11-1996 Clinton Dole Yes Democrat 379
5 7-11-2000 Bush jr. Gore No Republican 271
6 2-11-2004 Bush jr. Kerry Yes Republican 286
7 4-11-2008 Obama McCain No Democrat 365
8 6-11-2012 Obama Romney Yes Democrat 332
9 8-11-2016 Trump Clinton No Republican 304
Note.
Information retrieved from the U.S. Electoral College in the online archive (Historical Election Results [archives.gov]).
In order to further estimate the spillover, the daily return of the index was calculated with the following formula:
,
Rt= P t−1 (P −Pt t−1)
whereas the variance of the return outside the event-window was computed as:
, σ
︿
2=n−1 R−R
where R is the return on trading day n, is the mean of the return of the trading days in election i,R
and n is the number of trading days in this period minus the total of 50 days in the event window.
3.2 Framework & hypotheses
Graph 1 shows the clustering of the volatility of the AEX. This can be modeled with the GARCH (1,1) framework. The underlying idea of this model is that current volatility can be explained by past volatility (Lamoureux & Lastrapes, 1990). Additionally, GARCH(1,1) does not assume a constant variance, which is very unlikely for financial time series. In fact, it describes how the variance of the error develops. Because of this, it’s a very suitable framework for this kind of data. The method of this thesis was based on the approach of Wong and McAleer (2009) and Bialkowski et al. (2008).
In this thesis, both the return and volatility were examined. The hypotheses are presented in the following section, separating the hypotheses testing for an effect on the return and the hypothesis testing for the effect on volatility.
Return
The first three hypotheses will test for spillover effects of the U.S. presidential elections on the return of the AEX.
Hypothesis 1:
The first hypothesis will test for the existence of a U.S. presidential election cycle in the AEX. This hypothesis reads as follows:
:
H0 α1= α2= α3= α4= 0
: or or or
H1 α1=/ 0 α2=/ 0 α3= / 0 α4=/ 0
Where α1− α4 are the effects of the dummy variable for the years one to four in the GARCH(1,1) framework, respectively.
The following framework will be used:
I I I I , ε (0, )
Rt= α1 1t+ α2 2t+ α3 3t+ α4 4t+ εt t~ N σt2
σ ε
σt2= γ
0+ γ1 t−12+ γ2 2t−1,
whereRtis the return on the AEX. I1trough I4 represent dummy variables indicating the first to the fourth year of the presidency by taking on the value one for the corresponding year and zero otherwise. σt−12 is the conditional variance or the garch term, and ε is the error term or the
t−12
arch term. Finally, the constant in the regression was omitted to avoid collinearity problems because of the used dummy variables. The framework will be applied on all eight elections separately.
The previously mentioned study of Foerster and Schmitz (1997) supports the existence of the U.S. presidential election cycle in the AEX. Following this, one would expect
- α1 α4 to be
non-zero. So, by examining these coefficients, there can be concluded whether the stock returns follow a random walk or, as discussed before, follow the U.S. presidential election cycle.
Hypothesis 2
Hypothesis two will test for a significant effect of the election of a Democratic incumbent on the AEX.
: =0
H0 α5 : >0
H1 α5
Where α5 is the effect of the dummy variable for the election of a Democratic president in the GARCH(1,1) framework.
Santa-Clara and Valkanov (2003) and Johnson, Chittenden and Jensen (1999) conclude that the returns in the U.S. stock market were higher during the presidency of a Democrat, than during the presidency of a Republican. Therefore, one would expect
α5 to be above zero. In case this
coefficient can be inferredas significantly higher than zero, there can be concluded that the election of a Democratic president has a positive effect on the return of the Dutch stock market.
Hypothesis 3
Finally, hypothesis three will test whether the reelection of a president has a significant effect on the return.
: =0 H0 α6
: <0
H1 α6
Where α6 is the effect of the dummy variable for reelection in the GARCH(1,1) framework.
According to Li and Born (2006) the reelection of a president reduces uncertainty in the financial market. The smaller the uncertainty, the smaller the upward movement of the return in the U.S. stock market. Consequently, one would expect the
α6 to be below zero when spillover
effects of the reelection of a U.S. president on the AEX exist.
In order to test hypotheses two and three, the following GARCH(1,1) framework will be utilized: I I I I D R , ε (0, ) Rt= α1 1t+ α2 2t+ α3 3t+ α4 4t+ α5 + α6 + εt t~ N σt2 σ ε σt2= γ 0+ γ1 t−12+ γ2 2t−1,
where all the variables explained for hypothesis one, stay the same. However, two dummy variables are added. The variable D takes on the value one when the elected president is a
Democrat, and zero when it’s a Republican; the variable R equals one when the president was reelected, and zero when he was not. This framework will be applied on the period from 1984 until 2018, examiningthe whole period together.
Volatility
The last hypothesis was formed in order to test the spillover effects on the volatility. Hypothesis 4:
This hypothesis will test whether the cumulative abnormal volatility is bigger than-, or equal to zero.
CAV H0: = 0
CAV H1: > 0
Where CAV is the cumulative abnormal volatility.
As stated before, both Li and Born (2006) and Bialkowski et al. (2007) concluded a peak in volatility because of the elections. Li and Born (2006) argue that the volatility rises after elections, where the exception is made for when voters are already certain of the president being reelected. This is in alignment with Bialkowski et al. (2007) who state that uncertainty and volatility have a positive relationship. So, when spillovers to the AEX exist, we would see a rise in volatility meaning a cumulative abnormal volatility bigger than zero.
In order to test hypotheses two and three, the following GARCH(1,1) framework was utilized: I I I I , ε (0, ) Rt= α1 1t+ α2 2t+ α3 3t+ α4 4t+ εt t~ N σt2 (1) σ ε σt2= γ 0+ γ1 t−12+ γ2 2t−1, (2)
where all the variables used are the same as for hypothesis one and explained above.
In order to test this last hypothesis, the abnormal volatility needs to be identified. The GARCH model is used to present the volatility in case the election had not occurred. This value serves as the benchmark. For every U.S. presidential election in the period from 1984 until 2018, the GARCH model will be implemented. This means that (1) and (2) are jointly estimated using maximum likelihood estimation in STATA. A symmetric event window of 25 days surrounding the elections was chosen since this was the event window resulting in the most significant outcomes in the research of Bialkowski et al. (2008). The estimation period equals all trading days in the 4-year period the elected president reigns. Following the study of Bialkowski et al. (2008), who use a period of 500 trading days, this 4-year period could be seen as a too large sample. However, since this study focuses on the spillover of the entire election cycle, including only 500 trading days would be insufficient.
The variance obtained with the GARCH framework will be compared with the variance measured in the returns of the stock prices over the four years of reigning, excluding the returns of the event window. The difference between these two is considered to be the abnormal volatility of election i
and is calculated as follows:
-AVi = σt2 σ
t2
︿
Whereas the cumulative abnormal volatility can be calculated as the sum of these abnormal volatilities:
AVC i= ∑8
i=1 AVi
In order to test the significance of the impact of the U.S. presidential elections on the volatility, a test statistic for the CAV is included. In other words, the equality of the distribution of the CAV and the returns in the event window will be examined with a Chi-square test.
4. Results
In this section, the results of this research are displayed and explained.
Table two summarizes the outcomes of the GARCH(1,1) framework, testing for the election cycle.
Table 2
Effect of election cycle in terms of returns
Election number (1) (2) (3) (4) (5) (6) (7) (8) Year1 0.00111 0.000789* 0.00129*** 0.00249*** -0.000968 0.000926 0.00140** 0.000711 (0.000887) (0.000473) (0.000421) (0.000845) (0.00100) (0.000635) (0.000715) (0.000679) Year2 0.000275 -0.000109 0.000231 0.00157** -0.000179 0.00139** 0.000485 0.000470 (0.000543) (0.000452) (0.000351) (0.000710) (0.000869) (0.000629) (0.000809) (0.000777) Year3 -0.000736 0.000641 0.000396 0.00119 -2.03e-05 0.000642 -0.000294 0.000731 (0.000731) (0.000526) (0.000479) (0.000812) (0.000852) (0.000651) (0.000861) (0.000556) Year4 0.00173*** 0.000183 0.00106** 0.000626 0.000399 -0.00150*** 0.000800 -0.000114 (0.000615) (0.000500) (0.000429) (0.000800) (0.00119) (0.000436) (0.000824) (0.000549) Observations 1,047 1,039 1,046 1,046 1,041 1,046 1,046 1,046 Note.
The model used was:
I I I I , ε (0, )
Rt= α1 1t+ α2 2t+ α3 3t+ α4 4t+ εt t~ N σt2
σ ε
σt2= γ0+ γ1 t−12+ γ2 2t−1,
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
The dummy Year1
was found significant in four of the eight elections, whereas the dummy
Year2
was only found significant in two of the eight elections. These dummy variables were
simultaneous significant for solely one election. Additionally, Year3
was not significant for any of
the elections incorporated in this study. Year4
is significant for three elections, overlapping with
one election where Year1
is significant, and one election where Year2 is significant. As a result, no
evident pattern can be recognized for the dummy variables representing the presidential election cycle (Year1 - Year4
). Therefore, one could argue that no evidence is found for either a negative or
positive spillover effect of the presidential cycle in the AEX. This finding is in contrast with the discussed study of Foerster and Schmitz (1997).
Subsequently, the results concerning the spillover effects of the election of a Democratic president and the event of a reelection are summarized in the following table.
Table 3
Effect of the election of a Democrat and reelection in terms of returns
Return AEX Year1 0.000639** (0.000270) Year2 0.000135 (0.000257) Year3 -0.000132 (0.000266) Year4 0.000106 (0.000266) Democrat 0.000509** (0.000222) Reelected 0.000472** (0.000229) Observations 8,650 Note.
The model used was:
I I I I D R , ε (0, )
Rt= α1 1t+ α2 2t+ α3 3t+ α4 4t+ α5 + α6 + εt t~ N σt2
σ ε
σt2= γ0+ γ1 t−12+ γ2 2t−1,
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
As can be seen, the dummy variables Reelected
and Democrat are both significant on a 5%
level. To start off, the coefficient of Democrat has a positive value. This suggests that when the elected president is a Democrat, the return of the AEX is higher than in case a Republican president is elected. As stated before, Santa-Clara and Valkanov (2003) and Johnson et al. (1999) concluded that, when a Democratic president was elected, returns in the U.S. stock market were higher. Following the findings of these studies, the spillovers of this effect are evident in the Dutch stock market.
Furthermore, the value of the coefficient of Reelected
is positive as well. Consequently, this
analysis suggests that when a U.S. president is reelected, the return of the AEX is higher than when a new president is chosen to govern. Li and Born (2006) argued that a reelection causes less uncertainty on the financial market. Consequently, the upward effect on the return is smaller than when a new president is elected. Therefore, as a consequent of the spillover of this lower return in the U.S, the return of the AEX was expected to be lower as well. In this study, the opposite turns out to be the case.
Finally, table four shows the effects of the presidential election on the variance of the Dutch stock market.
Table 4
Effects in terms of volatility
Election
number (1) (2) (3) (4) (5) (6) (7) (8)
Conditional variance
0.000599*** 0.000375*** 0.000745*** 0.00147*** -0.000183*** 0.000363*** 0.000599*** 0.000449***
(2.86e-05) (1.11e-05) (1.37e-05) (2.10e-05) (1.56e-05) (3.43e-05) (1.89e-05) (1.06e-05)
Variance 0.000214 8.72035e-05 4.93802e-05 0.000205 0.000342 0.0003546 0.0006054 0.0004489
P-value variance comparison
<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
AV 0.000385 0.000287 0.000696 0.001265 -0.000525 8.4e-06 -6.4e-06 1e-07
CAV 0.0021105
P-value 0.243
Note
. The model used was:
I I I I , ε (0, )
Rt= α1 1t+ α2 2t+ α3 3t+ α4 4t+ εt t~ N σt2
σ ε
σt2= γ0+ γ1 t−12+ γ2 2t−1,
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
As can be seen in this table, all conditional variances are significant on a 1% level and for every election, an abnormal variance is evident. On the basis of the p-value of the variance comparison test, the null hypothesis for similarity of the conditional variances and the variances measured outside the event-window can be discarded. The lowest abnormal variances are seen for election five and eight. Li and Born (2006) and Bialkowski et al. (2007) found that volatility rises when uncertainty about the outcome of the election is bigger. One could argue that the higher the number of electoral votes the winner obtained, the lower the uncertainty about the outcome of the elections. This implies that the elections with the lowest electoral votes for the winner should have the highest variance. Looking at the results, election five has the lowest amount of votes for the winning incumbent and yet has the lowest volatility. Election eight has an average number of electoral votes, but still a relative low abnormal variance. For this reason, no pattern in the abnormal variance and uncertainty can be recognized which is not in line with the findings stated before.
Furthermore, the summation of abnormal variances resulted in a positive cumulative abnormal volatility. However, the test statistic suggests independence of the CAV and the event of presidential elections. Just as the findings mentioned above, this is not in line with the findings of Li and Born (2006) and Bialkowski et al. (2008).
5. Conclusion
In this section, a brief summary is given and conclusions are presented.
The aim of this thesis was to examine the existence of spillovers effects of the United States presidential elections to the AEX, in terms of return and volatility. This was done by examining the period from 1984 until 2018, including eight elections. In order to test this effect, a GARCH(1,1) framework was used where different dummy variables were incorporated. This was done to test if movements in the return of the AEX could be explained by spillovers of the U.S. presidential election cycle, the election of a Democrat instead of a Republican, and the occurrence of a reelection. Furthermore, the abnormal volatility of all eight elections was analyzed.
To start off, no significant evidence was found for spillovers of the U.S. presidential election cycle. This is in contrast with the study of Foerster and Schmitz (1997). Their findings show that the returns of the AEX in year 2 of the cycle, are lower than the mean of the return of the years one, two and three.
Second, a positive effect of the election of a Democratic president on the return of the AEX was found. This suggests the existence of spillovers to the AEX of the findings of Santa-Clara and Valkanov (2003) and Johnson et al.(1999) since their findings state that the election of a
Democratic president causes a higher return in the U.S. stock market. However, this is not consistent with Bialkowski et al. (2007) who report that no significant statistical evidence was found for higher returns under a Democratic regime.
Furthermore, a positive effect of the event of a reelection in the U.S. on the return of the AEX was found. Although, the findings of this study suggest other spillovers than were expected. The expectation was based on the following theory. Li and Born (2006) argued that a reelection causes less uncertainty on the financial market. Consequently, the effect on the return is smaller than when a new president is elected. Because of this, the expectations that the same effect would be visible in the return of the AEX were formed. But, the opposite turns out to be the case since the findings of this study suggest a bigger return of the AEX, in case of a reelection in the United States.
For all elections included, a significant difference between the conditional variance obtained by the GARCH(1,1) framework and the variance measured is evident. This suggests a spillover effect of the election in the United States on the volatility of the Dutch stock market. However, based on the test statistic this can be questioned. That is to say, the test statistic of the cumulative abnormal variance shows the independence of this variance and the event of elections. As a matter of fact, no spillovers of the U.S presidential elections on the volatility of the AEX can be concluded.
6. Discussion
In the following section, limitations of this research and suggestion for further research are presented.
The main suggestion for further research has arisen from the following argument. This study shows the existence of a (cumulative) abnormal volatility. Though, since a relationship between the volatility and the U.S. presidential elections is discarded by the test statistics, the origin of this volatility is not proven. Further research could study where this abnormal volatility is coming from, and how investors could take this into account when investing or constructing a well-diversified portfolio.
Also, some flaws in the used framework should be acknowledged. Table five shows the arch and garch effects for each election incorporated in this study.
Table 5
Arch and garch effects
Election number (1) (2) (3) (4) (5) (6) (7) (8) L.arch 0.161*** 0.0685*** 0.0459* 0.208*** 0.144*** 0.368*** 0.0934*** 0.172*** (0.0274) (0.0157) (0.0245) (0.0400) (0.0274) (0.0410) (0.0238) (0.0389) L.garch 0.844*** 1.343*** 1.012*** 0.866*** 1.175*** 0.708*** 1.253*** 0.925*** (0.0901) (0.139) (0.391) (0.125) (0.120) (0.0653) (0.156) (0.136)
Constant -1.61e-05 -3.81e-05*** -3.59e-06 -1.47e-05 -0.000110*** -1.14e-05* -8.78e-05*** -1.01e-05
(1.28e-05) (1.04e-05) (1.77e-05) (1.93e-05) (3.15e-05) (6.61e-06) (3.04e-05) (1.25e-05)
Observations 1,047 1,039 1,046 1,046 1,041 1,046 1,046 1,046
Note
. The model used was:
I I I I , ε (0, )
Rt= α1 1t+ α2 2t+ α3 3t+ α4 4t+ εt t~ N σt2
σ ε
σt2= γ0+ γ1 t−12+ γ2 2t−1,
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
First, when looking at table five the following can be concluded. For every election, the sum of the arch and the garch effect is more than one. This implies that the variance increases over time and that the model is non-stationary. Moreover, the number of trading days used to estimate the GARCH(1,1) model was around 1000 days where Bialkowski et al. (2007) discuss that using 500 days as a sample would be optimal. Additionally, the number of electoral votes may be an
insufficient indication for the uncertainty of voters, since this is only established after the elections have taken place. Besides this, the usage of the dummies Reelection
and Democrat simultaneously
canresult in a bias. This is due to the fact that the event of a reelection causes the president to be either a Democrat or Republican again. As a result of these shortcomings, the validity of this research could have been affected.
7. Bibliography
Allvine, F. C., & O'Neill, D. E. (1980). Stock market returns and the presidential election cycle: Implications for market efficiency. Financial Analysts Journal
, 36(5), 49-56.
Bialkowski, J., Gottschalk, K., & Wisniewski, T. P. (2007). Political orientation of government and stock market returns. Applied Financial Economics Letters
, 3(4), 269-273.
Białkowski, J., Gottschalk, K., & Wisniewski, T. P. (2008). Stock market volatility around national elections. Journal of Banking & Finance
, 32(9), 1941-1953.
Booth, J. R., & Booth, L. C. (2003). Is presidential cycle in security returns merely a reflection of business conditions?. Review of Financial Economics
, 12(2), 131-159.
Brennan, M. J. (1989). Capital asset pricing model. In Finance (pp. 91-102). Palgrave Macmillan, London.
Brown, K. C., Harlow, W. V., & Tinic, S. M. (1988). Risk aversion, uncertain information, and market efficiency. Journal of Financial Economics
, 22(2), 355-385.
Brunner, M. (2009). Does politics matter? The influence of elections and government formation in the Netherlands on the Amsterdam Exchange Index. Acta Politica
, 44(2), 150-170.
European Union, Trade in goods with USA. (2018, 14 april). Retrieved from
http://ec.europa.eu/trade/policy/countries-and-regions/countries/united-states/ Engle, R. (2001). GARCH 101: The use of ARCH/GARCH models in applied econometrics.
Journal of economic perspectives
, 15(4), 157-168.
Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. The
journal of Finance
, 25(2), 383-417.
Freeman, J.R., Hays, J.C. and Stix, H. (2000) Democracy and markets: The case of exchange rates. American Journal of Political Science 44(3): 449–468.
Foerster, S. R., & Schmitz, J. J. (1997). The transmission of US election cycles to international stock returns. Journal of International Business Studies
Füss, R., & Bechtel, M. M. (2008). Partisan politics and stock market performance: The effect of expected government partisanship on stock returns in the 2002 German federal election. Public Choice
, 135(3-4), 131-150.
Goodell, J. W., , S. (2013). US presidential elections and implied volatility: The role of political uncertainty. Journal of Banking & Finance
, 37(3), 1108-1117.
Herron, M. (2000) Estimating the economic impact of political party competition in the 1992 British election. American Journal of Political Science 44(2): 320–331.
Huang, R. D. (1985). Common stock returns and presidential elections. Financial Analysts
Journal
, 41(2), 58-61.
Johnson, R. R., Chittenden, W., & Jensen, G. (1999). Presidential politics, stocks, bonds, bills, and inflation. Journal of Portfolio Management
, 26, 27-32.
Lamoureux, C. G., & Lastrapes, W. D. (1990). Persistence in variance, structural change, and the GARCH model. Journal of Business & Economic Statistics
, 8(2), 225-234.
Li, J., & Born, J. A. (2006). Presidential election uncertainty and common stock returns in the United States. Journal of Financial Research
, 29(4), 609-622.
Pantzalis, C., Stangeland, D. A., & Turtle, H. J. (2000). Political elections and the resolution of uncertainty: the international evidence. Journal of banking & finance
, 24(10),
1575-1604.
Roll, R., & Ross, S. A. (1984). The arbitrage pricing theory approach to strategic portfolio planning. Financial analysts journal
, 40(3), 14-26.
Santa-Clara, P., & Valkanov, R. (2003). The presidential puzzle: Political cycles and the stock market. The Journal of Finance
, 58(5), 1841-1872.
Sharpe, W. F. (1977). The capital asset pricing model: a “multi-beta” interpretation. In Financial
Dec Making Under Uncertainty
(pp. 127-135).
Shiller, R. (1981). The Use of Volatility Measures in Assessing Market Efficiency. The Journal of
Finance,
36(2), 291-304. doi:10.2307/2327010
Stovall, R. H. (1992). Forecasting stock market performance via the presidential cycle. Financial
Analysts Journal
Wong, W. K., & McAleer, M. (2009). Mapping the Presidential Election Cycle in US stock markets. Mathematics and Computers in Simulation
