The short-term influence of the U.S. presidential elections
on stock markets.
Bachelor Thesis Economics and Finance Maurice Breg, 10582657
Supervisor: Ieva Sakalauskaite
Verklaring eigen werk
Hierbij verklaar ik, Maurice Breg, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan.
Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties worden genoemd.
De Faculteit Economie en Bedrijfskunde is alleen
verantwoordelijk voor de begeleiding tot het inleveren van de scriptie, niet voor de inhoud.
The question that will be researched in this paper is: ‘Do the stock indices in the United States, Japan, China (Hong Kong) and Germany have an abnormal return or abnormal volatility during an election period and is there a different return after a republican or a democratic elected president?’
This answer will be answered by calculating the cumulative abnormal return around the election day and running a regression with as dependent variable return. Also, the volatility during the election period will be compared to the normal volatility.
The cumulative abnormal return during an election period isn’t significant different from zero. However, the regression, excluding large outliers, does find a significant effect of the election day and the chosen party on the return after the election day. Only after 1990 there seems to be an abnormal volatility during an election period.
When looking at the results of the cumulative abnormal return and the regression, there is not a definite answer on the question, if the elections and the winning party do influence the daily return. The elections seem to influence the uncertainty of the stock markets only after 1990.
1. Introduction
The presidential elections in 2016 went along with the fact, that the stock markets responded on every news article, statement, speech or even tweet from the running presidents. From this fact, it seems that the presidential elections in the United States have a significant influence on the daily return and the volatility of the stock indices. After the election day, markets respond on the fact which party won the elections. But why do markets respond this way? The classical view states that markets prefer a republican president. Is there any evidence that the republican presidents did perform better on the daily return of the stock indices?
The influence of the presidential elections in the U.S. on the stock markets is an often discussed and investigated subject. In this thesis, there will be looked at three different aspects of the influence of the presidential elections in the U.S. on the stock markets. The first important aspect is if the election day has in the short run a significant influence on the daily return of the stock indices in the different countries around the world. While calculating this, the effect of the elected president will play an important role. So, the second aspect is, if there exists a preference by the market between a republican or a democratic president. This preference will be tested by comparing the daily return closely after the election day after a republican and a democratic win. The last aspect discussed in this paper is if there is an abnormal volatility closely to the election day.
All these matters can explain and show the influence of the U.S. presidential elections on the stock markets.
The influence of the presidential elections on the stock markets in the United States, Germany, Japan and Hong Kong (China) will be examined. The indices of those countries will be used, because those indices belong to the oldest indices in the world, so data from earlier in the 20th
century will be available. The countries are also in three different continents, known: North-America, Europe and Asia and play an important role in those continents. With indices in different continents, the effect of the elections on the stock markets around the world can better be tested.
The existing literature mostly focuses on the influence of the presidential or parliamentary elections on the stock markets in the same country. This paper will test as well the influence of the election on the U.S. itself. However, this paper also examines the effect of the U.S. presidential elections cross border. This will show, if the countries included in this paper also react on events like the presidential election in the U.S. Another addition of this research is the possibility of an extended period and including an extra presidential election.
2. Literature review
Earlier research did often find different conclusions. The results of testing the influence of the elections on the stock markets did often contradict each other. I will set the different results against each other to see what the contradictions are and how these contradictions arise.
About the first aspect, if the election day has in the short run a significant influence on the daily return of the market indices, are different results and conclusions. Leblang and Mukherjee (2005) did research about all three aspects discussed in this paper and used the Dow Jones Industrial Average as index, so focused on the United States only. They found that the election day did not have a significant effect on the daily return of the index in the short run. So, around the election day the Dow Jones did not have a significant abnormal return.
Evidence from Germany supports those findings. It states that there is not something like an election cycle, so there aren’t any differences between the returns, when looking how close the periods are to the election day (Döpke & Pierzdioch, 2006). They did research the effect of the elected party in Germany on the domestic stock index during the whole presidential term of 4 years. However, Pantzalis et al. (2000) did research the effect of the domestic elections on the domestic return of the stock index in 33 countries for the period 1974-1995. They did find a positive cumulative abnormal return (CAR) two weeks prior to the election day. Also, according to the paper of Allvine and O’Neill (1980), there is a significant effect of the election day on the return. Just like the paper of Pantzalis et al. (2000), they did find as well that the periods before the election day had a positive effect on the stock return. However, after the election day they did find a negative relationship. The abnormal returns arise due different components, this paper focused on the fact that when the incumbent lost, it had a positive effect on the abnormal return. Other components that weren’t investigated in this paper, but were mentioned were degree of political freedom and that the abnormal return depends on a function of the election timing.
Looking at the contradictions about the abnormal return around election day, it is not possible to give one definite answer. It is not clear if there is a positive, negative or even at least an effect of the election day on the stock return.
Other existing literature often investigated if the classical view, that the market prefers a republican president, holds. They did that, by looking if there was a significant difference between
the return of the market indices, considering a republican or a democrat president. There must be distinguished between the results that focus on the presidential cycle, so the whole 4-year period, and the ones who focus on the short-term return. The results and conclusions of the literature did often not correspond each other.
Niederhoffer, Gibbs and Bullock (1970) did find a relationship between the presidential elections and the Dow Jones Industrial Average price following the election day. According to their test results the average return of the Dow Jones rose in the following weeks after a Republican win. However, the first weeks after the election day after a Democrat win, the average Dow Jones did not rise. On average, after a democratic win, the Dow Jones did even fall. So, according to this paper, the classical view, that the market prefers a republican president, holds.
The paper of Riley Jr and Lucksetich (1980), does also support the classical view, that a republican president is preferred above a democrat one. At least in the first weeks after the election day, there was consistently a positive abnormal residual. While, when there was a democrat presidential victory, it was the opposite, a consistently negative abnormal return in the short run after the election day.
Nevertheless, existing literature about the classical view did also often not support the believe that the U.S. markets prefer a republican president. According to Huang (1985), who researched the period from 1929 till 1980, there was no evidence that the mean return under republican administration was higher. There was not a big difference between the democratic and republican administrations. In only 2 of the 6 periods measured, there was a low significant difference between the average difference of the return. In both cases the democrats had higher mean return. In this paper, the focus is on the whole presidential cycle, so not on the effect of the election day.
The research of Jones and Banning (2009) does also not support the classical view. In their test results, sample conducted of data that goes further back than the data used by Huang (1985), there isn’t a significant difference of monthly return after a republican or a democrat win. They did not focus on the short term, so the immediate effect of the election day can’t be captured. The overall performance of the different presidencies isn’t different, so there shouldn’t be an immediate reaction of the market on the winning party, because there is no evidence that one performs better for the stock market.
Santa-Clara and Valkanov (2003), did find a significant difference in excess return under Democratic and Republican presidencies. The excess returns under Democratic presidencies were higher than under Republican presidencies. However, the difference is not concentrated around election dates, so according to this paper, there is not a significant difference in the short term, after a republican or a democrat win. The difference arises throughout the whole presidential cycle.
Looking at the different test results and conclusions about the classical view it is not clear if there is an obvious difference between the returns of the stock market in the United States after a democratic or a republican win. Neither in the short run as over the whole presidential cycle.
The last theme that will be discussed in this paper is, if the uncertainty of the stock markets differs during a presidential election compared to a period without elections. About this theme there are also different conclusions after doing research in different countries.
Białkowski et all (2008), did focus on the volatilty. They examined the difference between the volatlity around election day and periods with no election in 28 OECD countries. They found that te volatiltiy does increase towards the election day. Investors do respond heavily to unexpected developments or unexpected wins.
Other researhers did find that the volatility doesn’t significantly change around election day. Leblang and Mukherjee (2005) did research the influence of the US presidential elections on the return and volatilty of the Dow Jones. They did not find significant difference between the volatility with daily data in a period around the election day and a period without one, using data from 1930 till 2000.
All these test results focusses only on the domestic response on the domestic elections. Unfortunate, there is no existing literature of high quality about the effect cross border. However, the effect on the
With all the different facts and results mentioned above, it is clear there is not a clear answer to all the questions. This paper will research if the stock markets do respond by checking if the indices do have an abnormal return or an abnormal volatilty around the presidential elections. Furthermore will be tested if there is any evidence that’s support the classical view about the preference of a republican president.
3. Methodology
To test if the election day has an immediate effect on the return of the indices, there will be looked if there is an abnormal return around the election day. This will be tested by calculating the cumulative abnormal return and by doing a panel regression. In this way, the influence of the election day on the daily return can be tested. Also, the effect of a republican or a democratic win on the daily return can be tested with these test methods. There will also be tested if the election day does influence the volatility of the daily return of the stock indices.
As indices for the countries the following ones will be used: United States: Dow Jones Industrial Average, Japan: Nikkei 225, Hong Kong: Hang Seng and Germany: DAX 30. The data that will be used for the panel regression and the calculation of the volatility begins for the Dow Jones and the Nikkei in 1960, for the DAX in 1968 and for the Hang Seng in 1972. All data ends on December 31, 2016.
3.1 Cumulative Abnormal Return
An abnormal return (Brown & Warner, 1985) can be calculated using the Capital Asset Pricing Model also CAPM, designed by Sharpe (1964). The CAPM calculates the expected return of a stock or an index. This equation has different components which should be determined first. It consists of the risk-free rate, the beta of the domestic stock index compared to a worldwide index and the daily return of that worldwide index.
𝐸(𝑅𝑖) = 𝑅𝑓+ 𝛽(𝑅𝑚− 𝑅𝑓)
𝐸(𝑅𝑖) = 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑓 𝑡ℎ𝑒 𝑖𝑛𝑑𝑒𝑥
𝑅𝑓 = 𝑡ℎ𝑒 𝑑𝑎𝑖𝑙𝑦 𝑟𝑖𝑠𝑘 𝑓𝑟𝑒𝑒 𝑟𝑎𝑡𝑒
The risk-free rate that will be used are the annual rates on 10-year treasury bonds of the United States (U.S. Department of Treasury). The annual risk free rate must be converted to a daily risk free rate.
𝑅𝑓 = (1 + 𝑎𝑛𝑛𝑢𝑎𝑙 𝑟𝑖𝑠𝑘 𝑓𝑟𝑒𝑒 𝑟𝑎𝑡𝑒) 1
𝛽 = 𝑡ℎ𝑒 𝑏𝑒𝑡𝑎 𝑜𝑓 𝑡ℎ𝑒 𝑑𝑜𝑚𝑒𝑠𝑡𝑖𝑐 𝑖𝑛𝑑𝑒𝑥 𝑐𝑜𝑚𝑝𝑎𝑟𝑒𝑑 𝑡𝑜 𝑡ℎ𝑒 𝑀𝑆𝐶𝐼 𝑊𝑜𝑟𝑙𝑑 𝑖𝑛𝑑𝑒𝑥. The beta of the indices will be determined as follows:
𝐵𝑒𝑡𝑎 =𝑐𝑜𝑣(𝑥, 𝑦)
𝑣𝑎𝑟(𝑥) , 𝑤ℎ𝑒𝑟𝑒 𝑥 𝑖𝑠 𝑡ℎ𝑒 𝑑𝑜𝑚𝑒𝑠𝑡𝑖𝑐 𝑖𝑛𝑑𝑒𝑥 𝑎𝑛𝑑 𝑦 𝑡ℎ𝑒 𝑀𝑆𝐶𝐼 𝑊𝑜𝑟𝑙𝑑 𝐼𝑛𝑑𝑒𝑥.
To calculate the covariance and the variance, first the daily return of the different indices must be determined. 𝑅𝑒𝑡𝑢𝑟𝑛 =𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝑃𝑟𝑖𝑐𝑒𝑡− 𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝑃𝑟𝑖𝑐𝑒𝑡−1 𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝑃𝑟𝑖𝑐𝑒𝑡−1 𝑣𝑎𝑟(𝑥) = 1 𝑁∑(𝑟𝑒𝑡𝑢𝑟𝑛𝑥− 𝑎𝑣𝑒𝑟𝑎𝑔𝑒𝑥) 2 𝑐𝑜𝑣(𝑥, 𝑦) = 1 𝑁∑(𝑟𝑒𝑡𝑢𝑟𝑛𝑥− 𝑎𝑣𝑒𝑟𝑎𝑔𝑒𝑥) ∗ (𝑟𝑒𝑡𝑢𝑟𝑛𝑦− 𝑎𝑣𝑒𝑟𝑎𝑔𝑒𝑦) The variances and covariance’s are the variances of the daily return of the indices.
𝑅𝑚= 𝐷𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑓 𝑡ℎ𝑒 𝑚𝑎𝑟𝑘𝑒𝑡
As market return, a global stock index will be used. This index is the MSCI World index founded and maintained by Morgan Stanley Capital International (MSCI). The index consists of 1654 stocks from 23 developed countries (31 December 2016). This index is founded in 1968 and because of that, only a limited amount of elections can be investigated, namely only 12 (the election day of 1968 can’t be used). So, with the cumulative abnormal return the results will be limited.
The total weight of the stocks per country in the MSCI World index is divided as follows:
As shown in the graph, with a weight of 60.1%, stocks from the U.S. dominates the MSCI World index and therefore the Dow Jones will probably have a very high beta. This is caused by the fact that the correlation between the MSCI World index and the Dow Jones will be very high. The same stocks influence both indices, so the covariance and the variance of the Dow Jones will be closer together. When later looking at the results this must be considered as side note. The opposite holds for China, they are not listed in the MSCI World index, because they are not seen as a developed country. However, the MSCI World can be used as a proper proxy for a worldwide stock index.
United States 60.1% Japan 8.75% United Kingdom 6.65% France 3.69% Canada 3.64% 17.17%Other
With all the calculated components of the Capital Asset Pricing Model, the abnormal return can be calculated. This can be done by comparing the actual daily return of the indexes with the expected daily returns according to the Capital Asset Pricing Model.
𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝑅𝑒𝑡𝑢𝑟𝑛 (𝐴𝑅𝑡) = 𝑎𝑐𝑡𝑢𝑎𝑙 𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 − 𝑒𝑥𝑝𝑒𝑐𝑡𝑒𝑑 𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛
= 𝑅𝑓+ 𝛽(𝑅𝑚− 𝑅𝑓) −
𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝑃𝑟𝑖𝑐𝑒𝑡− 𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝑃𝑟𝑖𝑐𝑒𝑡−1
𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝑃𝑟𝑖𝑐𝑒𝑡−1
The abnormal return in this thesis will be calculated for three different time periods. 30 days before, 30 days after and 30 days before and after the election day. To calculate the abnormal return for these periods, the cumulative abnormal return must be calculated.
𝐶𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝑅𝑒𝑡𝑢𝑟𝑛(𝐶𝐴𝑅) = ∑𝐴𝑅𝑡
The cumulative abnormal return doesn’t say much until the statistic is calculated. With the t-statistic it can be calculated if the cumulative abnormal returns are significantly different from zero for the chosen period.
𝑉𝐴𝑅(𝐶𝐴𝑅) = ∑(𝐶𝐴𝑅𝑡 − 𝐶𝐴𝑅𝑚𝑒𝑎𝑛)2
𝐻0: 𝐶𝐴𝑅 = 0
𝐻1: 𝐶𝐴𝑅 ≠ 0
𝑡 − 𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐 (𝐶𝐴𝑅) = 𝐶𝐴𝑅
√𝑉𝐴𝑅(𝐶𝐴𝑅), 𝑤ℎ𝑒𝑟𝑒 𝐶𝐴𝑅 𝑖𝑠 𝑡ℎ𝑒 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑜𝑓 𝑎𝑙𝑙 𝑣𝑎𝑙𝑢𝑒𝑠.
With these results, also the difference between the abnormal returns after a republican and a democrat win can be compared. This will be done by panel regression, by using republican as a dummy to see if the coefficient significantly differs from zero. In this case, only the cumulative abnormal return 30 days after the election is interesting, because before the election day it is not certain which president is going to win the elections.
𝐶𝐴𝑅 = 𝛼 + 𝛽1𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛
𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛 = 1 𝑖𝑓 𝑐ℎ𝑜𝑠𝑒𝑛 𝑝𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡 𝑖𝑠 𝑎 𝑟𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛, 0 𝑖𝑓 𝑐ℎ𝑜𝑠𝑒𝑛 𝑝𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡 𝑖𝑠 𝑑𝑒𝑚𝑜𝑐𝑟𝑎𝑡. 𝐻0= 𝛽1= 0
𝐻1: 𝛽1≠ 0 3.2 Panel regression
Because of the shortcomings and side notes by the calculation of the cumulative abnormal return, the effect of the election day and the party of the chosen president on the daily return will also be tested with panel regression. With panel regression, the tested period can be extended because there is no restriction with the available data of the MSCI World index. The regression will also be done for the 3 different periods, 30 days before, 30 days after and 30 days before and after the election day.
The dependent variable will be the daily return of the indices. With panel regression, the average effect of the added variables on the daily return of all indices can be calculated at once. So, first the daily returns of the indices are calculated.
𝐷𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 =𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝑃𝑟𝑖𝑐𝑒𝑡− 𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝑃𝑟𝑖𝑐𝑒𝑡−1 𝐶𝑙𝑜𝑠𝑖𝑛𝑔 𝑃𝑟𝑖𝑐𝑒𝑡−1
The effect of the election day can be calculated using a dummy variable, if the date is in the period around the election day, the dummy variable election will have the value 1. To examine the effect of the party of the chosen president on the return of the market, the dummy republican will be added. This variable will take 1 if in that period the elected president is a republican, so the variable will take one since the election day, not the inauguration day. The variable is 0 if the elected president is a democrat. 𝑅𝑒𝑡𝑢𝑟𝑛 𝑖𝑛𝑑𝑒𝑥 = 𝑡ℎ𝑒 𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑓 𝑡ℎ𝑒 𝑖𝑛𝑑𝑖𝑐𝑒𝑠 𝑅𝑒𝑡𝑢𝑟𝑛 𝑖𝑛𝑑𝑒𝑥 = 𝛼 + 𝛽1𝐸𝑙𝑒𝑐𝑡𝑖𝑜𝑛 + 𝛽2𝐸𝑙𝑒𝑐𝑡𝑖𝑜𝑛 ∗ 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛 + 𝛽3𝑅𝑒 − 𝑒𝑙𝑒𝑐𝑡𝑖𝑜𝑛 ∗ 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛 + 𝜀𝑖 𝐸𝑙𝑒𝑐𝑡𝑖𝑜𝑛 = 1 𝑖𝑓 𝑡ℎ𝑒 𝑑𝑎𝑦 𝑖𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑒𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑝𝑒𝑟𝑖𝑜𝑑 𝑜𝑓 30 𝑑𝑎𝑦𝑠 𝑏𝑒𝑓𝑜𝑟𝑒, 30 𝑑𝑎𝑦𝑠 𝑎𝑓𝑡𝑒𝑟 𝑜𝑟 30 𝑑𝑎𝑦𝑠 𝑎𝑓𝑡𝑒𝑟 𝑎𝑛𝑑 𝑏𝑒𝑓𝑜𝑟𝑒. 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛 = 1 𝑖𝑓 𝑡ℎ𝑒 𝑐ℎ𝑜𝑠𝑒𝑛 𝑝𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡 𝑖𝑠 𝑎 𝑟𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛, 0 𝑖𝑓 𝑡ℎ𝑒 𝑐ℎ𝑜𝑠𝑒𝑛 𝑝𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡 𝑖𝑠 𝑛𝑜𝑡 𝑎 𝑟𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛. 𝑅𝑒 − 𝑒𝑙𝑒𝑐𝑡𝑖𝑜𝑛 = 1 𝑖𝑠 𝑝𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡 𝑖𝑠 𝑜𝑟 𝑤𝑎𝑠 𝑟𝑢𝑛𝑛𝑖𝑛𝑔 𝑓𝑜𝑟 𝑟𝑒 − 𝑒𝑙𝑒𝑐𝑡𝑖𝑜𝑛𝑠, 0 𝑖𝑓 𝑛𝑜𝑡.
To test the short-term effect of the variable republican, the dummy republican is multiplied by the dummy election. In this way, the dummy variable republican only tests the period around the election day. The effect of republican through the whole 4-year period won’t be tested, so the variable republican is not in the regression.
The variable re-election is added, because in almost all cases, when the president was running to be re-elected, he was re-elected. This means there won’t be much changes in the regime. The expectation is that this will affect the response and the expectation of the market. However, the dummy re-elected is only interesting in the short-term, so also this dummy is multiplied with the dummy election. In this way, only the short-term effect of re-election will be calculated.
To calculate if the daily return of the indices is significant different from zero, again a t-test must be done. 𝑡 − 𝑠𝑡𝑎𝑡𝑖𝑠𝑡𝑖𝑐(𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛) =𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛 √𝑉𝐴𝑅(𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛) 𝑉𝐴𝑅(𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛) = ∑(𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛𝑡− 𝑑𝑎𝑖𝑙𝑦 𝑟𝑒𝑡𝑢𝑟𝑛𝑚𝑒𝑎𝑛)2 𝐻0: 𝛽1, 𝛽2, 𝛽3= 0 𝐻1: 𝛽1, 𝛽2, 𝛽3≠ 0 3.3 Abnormal Volatility
To test if the elections did influence the volatility of the stock indices, the abnormal volatility can be calculated. The variance of the stock index is calculated as follows:
𝑣𝑎𝑟(𝑠𝑡𝑜𝑐𝑘 𝑖𝑛𝑑𝑒𝑥) = 1
𝑁∑(𝑟𝑒𝑡𝑢𝑟𝑛𝑠𝑖𝑡− 𝑎𝑣𝑒𝑟𝑎𝑔𝑒𝑠𝑖)
2
Where si stands for stock index and t for the time.
The volatility will also be measured for the three different periods and compared to the normal volatility. 𝐻0: ( 1 𝑁∑(𝑟𝑒𝑡𝑢𝑟𝑛𝑠𝑖𝑡(𝑛1, 𝑛2) − 𝑎𝑣𝑒𝑟𝑎𝑔𝑒𝑠𝑖(𝑛1, 𝑛2)) 2 ) − (1 𝑁∑(𝑟𝑒𝑡𝑢𝑟𝑛𝑠𝑖𝑡− 𝑎𝑣𝑒𝑟𝑎𝑔𝑒𝑠𝑖) 2) = 0 𝐻1: ( 1 𝑁∑(𝑟𝑒𝑡𝑢𝑟𝑛𝑠𝑖𝑡(𝑛1, 𝑛2) − 𝑎𝑣𝑒𝑟𝑎𝑔𝑒𝑠𝑖(𝑛1, 𝑛2)) 2 ) − (1 𝑁∑(𝑟𝑒𝑡𝑢𝑟𝑛𝑠𝑖𝑡− 𝑎𝑣𝑒𝑟𝑎𝑔𝑒𝑠𝑖) 2) ≠ 0
𝑛1 𝑎𝑛𝑑 𝑛2 stands for the beginning and end day relative to the election day.
When comparing the normal volatility to the volatility in the period with election, the riskiness of the stocks during those periods can be compared.
4. Results
4.1 Results Cumulative Abnormal Return
To calculate the expected return of the index, the capital asset pricing model must be calculated first. The beta of the indices compared to the MSCI World index is as follows:
As expected, the beta of the Dow Jones is close to 1. This is caused by the fact, that stocks from the United States cover a big part of the total weight of the MSCI World index. Nevertheless, the expected return can still be calculated. The only thing that must be considered, is that the U.S. presidential elections have probably also a big influence on the expected return. The effect of the cumulative abnormal return will probably be less, than it would be if the share of the stocks from the United States was less.
The results of the cumulative abnormal return are in the table below. (𝑛1, 𝑛2) CAR Standard Error t-value P>|t|
(-30,30) 0.012171 .0097363 1.25 0.217
(-30,0) 0.007583 .0087164 0.87 0.389
(0,30) 0.004992 .0071632 0.70 0.489
In the period 30 days before the election day till 30 days after the election day, the cumulative
Country Index Beta
United States Dow Jones Industrial Average 0.981841
Japan Nikkei 225 0.62012
Hong Kong (China) Hang Seng 0.635427
abnormal daily return is on average 1.2 percent point higher than normal. However, this result is not significant different from zero at any significance level. This means, that on average there is no significant difference between the daily return in a period around the U.S. presidential election day compared to a period without an election day.
The same principle counts for the other two periods, only 30 days before and only 30 days after. For those time periods the average of the cumulative abnormal return is even lower and have a lower t-value than the period combined. So, in both periods the cumulative abnormal return isn’t significant from zero either.
To test if there is a difference between the cumulative abnormal return after a republican or a democratic win, the dummy variable republican is added to the regression, with the cumulative abnormal return as the dependent variable. The dummy republican only makes sense in the period 30 days after the election day, because before the election day it is not certain which candidate is going to win. The result of this regression is in the table below.
CAR Coefficient Standard Error t-value P>|t|
Republican .0036864 .0163585 0.23 0.823
This result is neither significant different from zero. That means that there is not a significant difference between the periods after the election day after a republican or a democratic win.
The fact that not one value of the cumulative abnormal return and the variable republican is significant different from zero can be explained by the fact that U.S. stocks have a high influence on the MSCI World index. The expected return calculated by the CAPM is also affected by the U.S. presidential elections, because of the high weight of U.S. stocks in the MSCI World index. That could be an explanation why the expected and the real daily return doesn’t differ that much. When there was a more diversified world index, there will probably be a higher chance of an average cumulative abnormal return that is significant different from zero.
4.2 Results Panel Regression
To understand which values of the daily return influence the mean and therefore the regression results, first there will be a table with descriptive statistics:
Variable Mean Max Min N N > |0.10|
Return .0003396 .1882374 -.3333046 54264 38
The table shows, that the extreme values differ a lot from the mean of all observations. Only 38 of the 54264 observations are higher than the absolute value of 0.10. Because of this there are two different panel regressions performed. The first one includes all observations, so including the values above the absolute value of 0.10. The second regression doesn’t use those 38 values, so the large outliers are deleted.
In the following table, the results of the first regression, including all observations, are represented. (𝑛1, 𝑛2) Election Republican*Election Re-election*Election
(-30,30) (.0003217) .0000314 [0.10] -.0004345 (.0005697) [-0.76] (-30,0) (.0004519) .0001064 [0.24] .0006735 (.0007855) [0.86] (0,30) (.0006223) .0006493 [1.04] -.0007686 (.0007447) [-1.03] -.0017817 (.0008132) [-2.19*]
(The normal value is the coefficient. The one between brackets, is the standard deviation. The one between the square brackets is the t-value. Only the one with the * is significant different from zero at a 5% confidence interval.)
In all periods the effect of the election period is not significant different from zero. The daily return of the indices doesn’t differ in periods with election day, compared to periods without an election day. The variable republican*elections isn’t significant from zero either. So, there is not a significant difference in periods after the election day after a republican or a democratic win. The only significant value is the variable re-election*elections in the period after the election day. The significant effect is negative, so the respond of the market after an incumbent won the elections is negative. This is in accordance with Allvine and O’Neill (1980), they did also find that when the incumbent lost it had a positive effect on the stock return.
In the next panel regression, the outliers of above a daily return of the absolute value of 0.10 are deleted. This were 38 values. Almost all values were around financial crisis, including black Monday, October 19, 1987 and around October 15, 2008, just before and after the collapse of the Lehman Brothers.
(𝑛1, 𝑛2) Election Republican*Election Re-election*Election
(-30,30) (.0003087) -.0000309 [-0.10] -.0011806 (.0005494) [-2.15*] (-30,0) -.00000510 (.0004338) [-.01] .0007944 (.0007538) [1.05] (0,30) (.0005983) .0012683 [2.12*] -.0018184 (.0007177) [-2.53**] -.0036389 (.0007895) [-4.61***]
(The normal value is the coefficient. The one between brackets, is the standard deviation. The one between the square brackets is the t-value. The ones with the * are significant different from zero at a 5% confidence value, the one with ** at a 2.5% confidence value and the one with *** at a 1% confidence value.)
In this panel regression, the results of all variables are different from the results in the previous one. The election period seems only to have a positive significant influence on the daily return of the stock indices in the period after the election day. This contradicts the findings of Allvine and O’Neill (1980) and Pantzalis et all. (2000) who did find especially positive abnormal return before the election day. Not only the variable election is significant different from zero after the election day. All three values after the election day are significant different from zero. The variable republican*elections is significant different from zero, so in the period after the election day after a democratic win, the average return is higher than after a republican win. When the current president is running for re-election, on average this has a negative influence on the daily return around the election day, just like the first panel regression.
When looking at the regression results including all observations, the elections and the winning party doesn’t have significant influence on the daily return closely to the election day. However, when deleting the extreme large outliers, the values after the election day becomes significant different from zero. It seems that those large outliers had per value a huge influence on the average.
4.3 Results Abnormal Volatility
The volatility is divided in two periods, because of the uncertainty before the election day about which president is going to win and therefore which regime and rules will apply after the election day. After the election day, the expectation is therefore that the volatility will be less, than before the election day. In the table below the volatility of the different periods are represented.
Volatility till 1990 from 1990
Normal 0,01208325 0,01382625
30 days before 0,00984375 0,0242246
30 days after 0,01068875 0,0178685
Before 1990 the volatility doesn’t seem to be affected by the election day. The average volatility is even a bit lower before and after the election day, than the normal volatility. However, after 1990, the election day does seem to have influence on the volatility of the daily return of the stock indices. In the period before the election day, the volatility is on average almost twice as high as the normal volatility. After the election day, the volatility is still higher than the normal volatility of the daily return of the stock indices.
Looking at this result, it seems that after 1990 the election day goes in line with a higher uncertainty of the stock indices. A possible explanation can be the uncertainty that arose by the different financial crises after 1987. The believe can be that the chosen president has influence on the possibility of such crises.
5. Conclusion
Considering all the test results, unfortunately there is not one definite possible answer. Looking at the cumulative abnormal return, there is not a significant influence of the election day on the daily return and no difference between the cumulative abnormal return after a republican or a
democratic win. This can be explained by the fact that the weight of U.S. stocks in the MSCI World index is disproportionally high. This gives probably a lower cumulative abnormal return than when the weight of stocks around the world was more proportionally allocated.
Looking at the results of the panel regression on the daily return, and especially the regression without the outliers, there seems to be a relationship between election period and the elected party on the daily return of the stock indices. After the election day, the influence of election period and elected party is significant different from zero, as well the variable re-election.
After 1990, it also seems that there is a higher uncertainty around the election day, however this is not tested against a significance level. The only thing that can be said is that the volatility does increase in this period, but about the fact that it is significant different can’t said much.
Because of the differences in the results it is not clear if the U.S. presidential elections have a significant influence on the stock indices of the U.S., Japan, Hong Kong and Germany.
For further research, the results of the cumulative abnormal can possibly be calculated using a different world index.
6. Bibliography
Allvine, F. C., & O'Neill, D. E. (1980). Stock Market Returns and the Presidential Election Cycle: Implications for Market Effiency. 36(5), 49-56.
Białkowski, J., Gottschalk, K., & Wisniewski, T. P. (2008). Stock Market Volatility Around National Elections. Journal of Banking & Finance, 32, 1941-1953.
Brown, S. J., & Warner, J. B. (1985). Using Daily Stock Returns: The Case of Event Studies. Journal of
Financial Economics, 14, 3-31.
Döpke, J., & Pierzdioch, C. (2006). Politics and the Stock Market: Evidence from Germany. European
Journal of Political Economy, 22, 925-943.
Huang, R. D. (1985). Commmon Stock Returns and Presidential Elections. Financial Analysts Journal,
41(2), 58-61.
Johnson, R. R., Chittenden, W. T., & Jensen, G. R. (1999). Presidential Politics, Stocks, Bonds, Bills, and Inflation. The Journal of Portfolio Management, 26(1), 27-31.
Jones, S. T., & Banning, K. (2009). US Elections and Monthly Stock Market Returns. Journal of
Economics and Finance, 33, 273-287.
Leblang, D., & Mukherjee, B. (2005). Government Partisanship, Elections, and the Stock Market: Examining American and British Stock Returns, 1930-2000. American Journal of Political
Science, 49(4), 780-802.
Niederhoffer, V., Gibbs, S., & Bullock, J. (1970). Presidential Elections and the Stock Market. Financial
Analysts Journal, 26(2), 111-113.
Pantzalis, C., Stangeland, D. A., & Turtle, H. J. (2000). Political Elections and the Resolution of Uncertainty: The International Evidence. Journal of Banking & Finance, 24, 1575-1604.
Riley, Jr., W. B., & Luksetich, W. A. (1980). The Market Prefers Republicans: Myth or Reality. The
Journal of Financial and Quantitative Analysis, 15(3), 541-560.
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. (1964). Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk.
The Journal of Finance, 19(3), 425-442.
Stovall, R. H. (1992). Forecasting Stock Market Performance via the Presidential Cycle. Financial
Analysts Journal, 48(3), 5-8.
Wong, W.-K., & McAleer, M. (2009). Mapping the Presidential Election Cycle in US stock markets.
