US presidential elections on volatility, stock market returns and political cycles

US presidential elections on volatility, stock market returns and

political cycles.

Bachelor Thesis Economics and Finance

University of Amsterdam

Student ID: 10796398

January 2017

Abstract

Existing literature concludes that US presidential elections have an impact on stock market returns and volatility. The studies mainly focused on the elected president and its party or the influences of political cycles. This study combines these variables and adds the effect of

the composition of Congress. The US constitution makes that congressional power should be considered, because of its prominent role of making new laws and their right to override

a presidential veto with two-thirds of both chambers. In this study over the period 1978-2016 using event study methodology and a GARCH (1,1) model, the impact of elections on

volatility and stock market returns are confirmed. Thereby, results show that variables concerning the Congress are significant and should be considered when explaining abnormal

economic behavior in times of presidential elections. Further research on significant results is done by tabling the stock market returns and comparing the outcomes. This might point out gaps in existing literature and shows what variables are interesting to look at in future

research.

Universiteit van Amsterdam Britt Molenkamp Supervisor: Philippe Versijp

Statement of Originality

This document is written by Student Britt Molenkamp 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.

Index

ELECTORAL SYSTEM IN THE UNITED STATES ... 8

METHODOLOGY AND DATA ... 9

DATA ... 10

STOCK MARKET RETURNS ... 11

VOLATILITY ... 12

POLITICAL CYCLES ... 13

RESULTS ... 13

STOCK MARKET RETURNS ... 13

VOLATILITY ... 15

POLITICAL CYCLES ... 17

CONCLUSION ... 21

BIBLIOGRAPHY ... 23

Introduction

Looking at history there is substantial evidence that events relating presidents in the US affect the stock market. An example of this is in 1959, when Eisenhower had a heart attack. As a result of this, the Dow Jones saw a drop of 6.54%. A similar event saw another drop in Dow Jones with the assassination of Kennedy. Also, more recent with the election of Donald Trump. According to some, a business man with no political knowledge or background. As indicated by figure 1, the volatility of the stock market returns on the S&P 500 around the 8th

of November was significantly high.

Figure 1. Returns S&P 500 during presidential election 2016

To also examine implied volatility figure 2 shows the daily changes of the VIX index. The VIX is often known as the ‘fear index’. The figures show that the S&P 500 and VIX correlates negatively. Literature on the relationship between both of Smales (2016) confirms this inverse and asymmetric relationship between movements in VIX and SPX returns.

Figure 2. VIX index during presidential elections 2016

The figures indicate that elections do affect both stock market and volatility but the sign of their relationship is still unsure.

Looking at volatility and market returns behavior over a period of time, consisting of several presidential elections, will show whether an ‘election effort’ exists. There have already been some studies on the impact of US presidential elections on stock market returns and separate ones on the influence of politics. For example, Santa-Clara & Valkanov (2003) focused on the

differences of having a Republican or Democratic president in office. They found that volatility during Republican presidencies is somewhat higher compared to having a Democratic president in power. They also state that there is no evidence of extremely large market returns surrounding elections dates, but that it builds up during the term.

On the contrary Riley & Luksetich (1980) concluded that returns surrounding presidential elections do significantly deviate from normal returns. For volatility, Goodell & Vähämaa (2013) and Bialkowski et al. (2007) find results of higher volatility when the probability of a candidate winning is uncertain.

This study combines different aspects and adds some new variables to give another perspective on the effects of elections and political influences. Not only the winning presidential party is included, but also its composition of Congress, more Republican or Democratic representatives. The Congress falls under the legislative branch of the American constitution. It makes new laws based on the current political guidelines and has the power to override a presidential veto with two-thirds of both chambers. Important then is its composition, because the thoughts of the majority will determine the way the Congress acts. The political view of the Congress influences the expectations of the market and their future economic behavior. This is why the variable should be considered when explaining election effects. The knowledge obtained is useful to forecast and anticipate on future volatility and stock market returns in relation to power of the ruling party. The insight into structural patterns surrounding US presidential elections gives investors and traders the ability to forecast the impact so they can anticipate on changes in stock market returns.

There are facts that have been agreed on in prior research, but there are still some contradictions and interesting parts that have not been matched by academic research. This leads to the following research question and its sub-question:

Do US presidential elections affect volatility and the US stock market?

How does the elected presidential party and Congress contribute?

Literature Review

Prior studies on the effects of presidential elections indicate that the market indeed reacts surrounding the event, but not consistently similar.

Goodell & Vähämaa (2013) state that, over the years 1992 until 2008, presidential elections increase stock market volatility. Simultaneously they found that volatility decreases when the likelihood of a certain running candidate winning the elections grows.

Bialkowski et al. (2007) came to similar conclusions. They found a significant increase in volatility over a period of 3 months prior elections, but that the event disappears when the likelihood of a particular candidate winning increases.

Santa-Clara & Valkanov (2003) found that volatility is somewhat higher during a Republican presidential cycle. They state that it is hard to determine whether political variables cause

fluctuations in stock returns because it’s a ‘chicken or the egg’ issue. Do politics influence the stock market or is it the economic cycle that determines for them?

Niederhoffer, Gibbs & Bullock (1970) looked at changes on the Dow Jones Industrial Average the day after the elections over the period 1900 and 1968. They find a significant difference in returns under a winning Republican candidate compared to a Democratic candidate. On the contrary, they do not find a systematic pattern in the returns prior the elections of the market predicting a winner.

Riley & Luksetich (1980) build on the results of Niederhoffer, Gibbs and Bullock (1970) by looking at different periods surrounding the event. They conclude that the cumulative abnormal returns are significantly different 17 weeks surrounding the presidential elections with different peaks for Democratic and Republican presidents. They also find a decline in stock prices after elections when a change of control in the White House is recorded.

The widespread believe of publican presidents having a favorable effect on stock markets has been contradicted by several, under whom Siegel (2002). His research over the period 1888 - 2001 shows that performance of the stock market under a Democratic administration result in higher returns. Siegel also states that elections results since World War II are obtuse. The small effect of elections, in some cases, have to do with continuous control of Congress by the Republicans. An example is the second time Clinton got elected as president in 1997. Booth & Booth (2003) found that stock returns during the last two years of a political term are higher compared to the first two years. They conclude that presidential elections and its outcome have a significant influence on explaining stock market returns.

Jinliang & Born (2006) and Bialkowski et al. (2006) both make use of a GARCH model to predict volatility and measure the effect of presidential elections on the markets volatility. Both studies find significant results of elections affecting the volatility of the stock market.

The convention in the literature and as described by Brown & Warner (1985) is to use 250 daily returns as a benchmark for the estimation period. This is a year of trading data on a random stock market. However, their research is based on OLS estimates and the Black Scholes model when doing an event study.

Hwang & Valls Pereira (2006) did research on small sample properties concerning GARCH estimates and persistency. They claim that the estimates of the GARCH (1,1) model are significantly negatively biased when using a small sample size. The study recommends using at least 500 trading days to establish an accurate and sufficient GARCH (1,1) model.

For formulating expectations, the Efficient Market Hypothesis and the statements above will be used. For volatility, this study will look into the likelihood of winning, a change of

composition of Congress and whether elections affect volatility positively or negativity. Interesting findings of prior literature makes that for stock market returns will be looked at what period surrounding presidential elections results in a significant difference in returns. Also, the colour of presidential party and congressional majority will be included when looking at stock market returns.

Stock market returns

The existence of the classical view is supported by Niederhoffer, Gibbs & Bullock (1970), but contradicted by Jones & Banning (2009) and Santa-Clara & Valkanov (2003). The classical view stands for the belief of Republican presidents having a better effect on stock market returns compared to Democratic ones. Also, Booth & Booth (2003) found that market returns over the last two years of a 4 year during presidential term are consistently higher than over the rest of the term. This could be because of more trust and being accustomed to the political policies of the president in power.

The constitution of the US is constructed so that power is divided. It consists of three different branches of whom the legislative vertical holds most power. In other words, it is the Congress who has the upper hand. Siegel (2002) and Riley & Luksetich (1980) both indicate that the congressional power and changes in composition influences the market. These beliefs will be tested in this study.

Hypothesis

Hypothesis 1: The null hypothesis is, the cumulative average abnormal returns during US residential election are zero. The alternative hypothesis states that the cumulative average abnormal returns during US presidential election are different from zero.

H0: (CAAR) = 0 H1: (CAAR)  0

Hypothesis 2: The null hypothesis is, electing a Republican or Democratic president does not have an impact on stock market returns. The alternative hypothesis states that under a Republican president the stock market performance is better than under Democratic power. H0: β = 0

H1: β > 0

Hypothesis 3: The null hypothesis is, having a Republican or Democratic majority in Congress does not contribute to a difference in stock market returns. The alternative hypothesis states that having a Republican or Democratic majority in Congress does contribute to a difference in stock market returns.

H0: β = 0 H1: β  0

Volatility

Research by Pantzalis et al. (2000), Nippani & Medlin (2002), Li & Born (2006) and Goodwell & Vähämaa (2013) indicate enhanced levels of volatility around the event of a presidential election. This is in line with the Efficient Market Hypothesis, which states that all available information is incorporated in stock prices (Berk & DeMarzo, 2014). Thus, when events are

uncertain as with presidential elections, it should be traceable in the volatility of the stock market returns surrounding these dates. The same holds for the effect of the likelihood of a candidate winning the elections. The more certain the market is about who the new president will be, the less volatile stock market reacts when the person indeed gets elected. This information is already incorporated in the stock prices prior election day. Other possible effects on volatility will be tested and are based on the above statement (EMH) and the knowledge provided by prior literature.

Hypothesis

Hypothesis 1: The null hypothesis is, the cumulative average abnormal volatility during US presidential elections is zero. The alternative hypothesis states that the cumulative average abnormal volatility during US presidential elections is different from zero.

H0: (CAAV) = 0 H1: (CAAV) ≠ 0

Hypothesis 2: The null hypothesis is, a higher probability of a running candidate winning the elections does not have an effect on volatility. The alternative hypothesis states that a higher probability of a running candidate winning has a negative effect on volatility.

H0: β = 0 H1: β < 0

Siegel (2002) made a statement about election results being muted after World War II, because of repeatedly controlling power in The Congress by the Republicans. It would be interesting to see whether the event of decreased uncertainty reflects itself in the beta of the regression.

Hypothesis 3: The null hypothesis is, a change of majority in Congress prior presidential elections does not have an effect on volatility. The alternative hypothesis states that a change of majority in Congress prior presidential elections has a positive effect on volatility. H0: β = 0

H1: β > 0

Electoral System in the United States

The United States Electoral College is established in the constitution as a mechanism that indirectly elects the president and vice president of the United States. Every state has a certain number of electors that is equal to the number of Senators of a state. In total consists of 538 electoral votes.

On election day, which is held on Tuesdays the day after the first Monday of the month November, citizens go to the polls to vote. The candidate that gets more than half (270) of the electoral wins the elections and will be the new president of the United States. When no

absolute majority is established, the House of Representatives can choose the new president.

For the vice-president, the Senate will pick their new leader. The last time this happened was in 1836.

Figure 3 shows the distribution of electoral votes in the United States. The number of electors determines how many members of Congress the states provides and thus how much power a certain state has to influence the election of the new president. It is a compromise between electing via popular vote and letting the Congress decide.

How the elections affect the stock market depends on people’s expectations and the manifesto of the elected party. The newly chosen president has the ability to make adjustment to current policies. Via tax proposals and additions to the use and distribution of governments revenue, new fiscal policy is set and has a specific influence on economic activity. However, any legislation has to go through both the executive and legislative branches before it gets accepted.

Besides the manifesto, the party also has a certain political campaign strategy. Elections are about winning or losing and nothing in between. That is why the strategy chosen revolves around the voters, their insights and what persuades voters to vote. Sometimes this leads to crisis management. Running candidates are exposed to unexpected events. Media or the opposite party will try to acquire information that can affect the image of the candidate, through which he or she can be seen as invalid for the job. To be able to face any situation both parties have crisis strategies mapped out to react to certain events. New information and accusations not only affect the image of the candidate in voters’ minds but also reflects itself in economic behavior. Stock markets move on markets beliefs and thus every event should be traceable in market returns.

Figure 3. Distribution of electoral votes US

Source: www.usa.gov

Methodology and Data

To answer the research question event study methodology and a GARCH (1,1) model will be used. The GARCH (1,1) model maps volatility clustering, and can be used to determine whether the abnormal volatility around election periods are consistently different from zero.

GARCH computes volatility by modelling a non-constant variance of the error term. In this study two separate regressions will be done. One will explain the volatility and the other one will be on stock market returns. In the regressions control variables, as the interest rate and the inflation rate, will be used to correct for the economic cycle. Furthermore, included are margin of victory above 50% points, the party of the elected president and the majority in Congress at that time. For both regressions stock market returns of the Dow Jones Industrial Average and the MSCI world market index will be used. Also, the returns of the Dow Jones will be listed and compared. Attention will be paid to differences in market returns under a Democratic or Republican president and the majority in the Congress at that time. This to try discover structural patterns and political cycles.

Avoiding the oil crisis's and including the last vote for Congress before new presidential elections in 1981, results in data from July 1978 until December 2016. Also, a recent period, wherein the guidelines of Democratic and Republican parties are more precise and economic behavior is more relatable, makes that results are more interesting.

Data

In this study, historical market returns of the Dow Jones Industrial Average are used. The returns for the regressions are obtained through Thomas Reuters DataStream and exist of data over the period 1978 until 2016. The advantage of the local market index is that it does not contain firm specific risk, only systematic risk that is necessary when examining volatility (Berk & DeMarzo, 2014). The market proxy, used to calculate abnormal returns and abnormal volatility, is obtained through the same source and will be the MSCI world index.

Historical information about presidential and congressional elections come from the US Senate Government website. Data on the margin of victory come from the World Data Bank. The interest rates and inflation rates are obtained through DataStream and come from the FED. All historical information has similar time constraint as the stock market returns. All historical information is shown in table 1.

TABLE 1. Historical Information Year Presidential Elections President Party Victory Margin above 50% Year Congressional Elections Majority Congress 1980 Republican 0.4089% 1978 Democrats 1980 Republicans 1984 Republican 0.4757% 1982 Republicans 1984 Republicans 1988 Republican 0.2918% 1986 Democrats 1988 Democrats 1992 Democrat 0.1877% 1990 Democrats 1992 Democrats 1996 Democrat 0.2045% 1994 Republicans 1996 Republicans 2000 Republican 0.0037% 1998 Republicans 2000 Democrats 2004 Republican 0.0316% 2002 Republicans 2004 Republicans 2008 Democrat 0.1784% 2006 Democrats 2008 Democrats 2014 Democrat 0.1171% 2010 Democrats 2012 Democrats 2016 Republican 0.0651% 2014 Republicans 2016 Republicans

Stock market returns

𝐶𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝑅𝑒𝑡𝑢𝑟𝑛𝑠𝑡(𝐶𝐴𝐴𝑅) = 𝛼 + 𝛽1∗ 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒 + 𝛽2∗ 𝐼𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒 +

𝛽3∗ 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛|𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐 𝑃𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡 + 𝛽4∗ 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛|𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐 𝐶𝑜𝑛𝑔𝑟𝑒𝑠𝑠 + 𝜀𝜄.

The control variables in the regression are the interest rate and the CPI. Both data are monthly and come from the FED. The CPI is in prices and the interest rate is used as a percentage. Thereby included are the party of the elected president and the majority of the Congress at time of presidential elections.

For the event study, the estimation period will consist of 500 trading days starting 50 days prior the event date to avoid election effects. This will be done for both the volatility as for the data on market returns. For calculating the abnormal returns, an event period of 10 days as well as one of 5 days will be used. This to make sure that it captures the whole effect and to see whether outcomes differ.

To analyze the findings of Niederhoffer, Gibbs & Bullock (1970) different event periods are tested for significance. This will show whether changing circumstances led to different outcomes when examining the period in which presidential elections affect the stock market. To determine whether US election have an effect on stock market returns an OLS regression will be done according to the event study methodology of De Jong (2007). To calculate expected returns the MSCI world index is used:

𝑅(𝐷𝑜𝑤 𝐽𝑜𝑛𝑒𝑠 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑖𝑎𝑙 𝐴𝑣𝑒𝑟𝑎𝑔𝑒),𝑡 = 𝛼 + 𝛽𝑅(𝑀𝑆𝐶𝐼 𝑊𝑜𝑟𝑙𝑑 𝐼𝑛𝑑𝑒𝑥),𝑡+ 𝜀𝑡

Abnormal returns will then be calculated using the OLS estimates with robust standard errors: 𝐴𝑅_{(𝐷𝐽),𝑡}= 𝑅_{(𝐷𝐽),𝑡}− 𝛼̂ − 𝛽̂𝑅(𝑀𝑆𝐶𝐼),𝑡

To be able to compare the periods cumulative abnormal returns of the event period of election i are calculated:

𝐶𝐴𝑅 = ∑ 𝐴𝑅𝑖,𝑡 𝑡2

𝑇=𝑡1

Then, the cross-sectional average with N=10/N=5 will be determined with corresponding standard deviations: 𝐶𝐴𝐴𝑅 =1 𝑁 ∑ 𝐴𝑅𝑖,𝑡 𝑡2 𝑇=𝑡1 𝑠 = √ 1 𝑁 − 1 ∑(𝐶𝐴𝑅𝜄− 𝐶𝐴𝐴𝑅) 2 𝑁 𝜄=1

A simple t-test will tell whether the cumulative average abnormal returns are different from zero: 𝐺 = √𝑁𝐶𝐴𝐴𝑅 𝑠 Volatility 𝐶𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝐴𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦𝑡(𝐶𝐴𝐴𝑉) = 𝛼 + 𝛽1∗ 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒 + 𝛽2∗ 𝐼𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒 + 𝛽3∗ 𝑃𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑊𝑖𝑛 + 𝛽4∗ 𝐹𝑜𝑟𝑚𝑒𝑟 𝐶𝑜𝑛𝑔𝑟𝑒𝑠𝑠 + 𝜀𝑖,.

In this regression, we also make use of the control variables interest rate and inflation. Both data are monthly data with CPI as prices and interest rate as a percentage. Other variables are the probability to win and whether there was a change of majority in Congress prior presidential elections. The probability of the elected president winning is calculated by the percentage point above 50% of the electoral votes.

To determine whether presidential elections have an effect on volatility, the normal expected volatility has to be calculated and compared to the actual realized volatility in the event period. To get an idea of normal volatility the same estimation period as for the market returns is used. The methodology applied is similar as with the market returns, only is expected volatility is calculated with a GARCH (1,1) model and its conditional variance:

𝑅 (𝐷𝑜𝑤 𝐽𝑜𝑛𝑒𝑠 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑖𝑎𝑙 𝐴𝑣𝑒𝑟𝑎𝑔𝑒)𝑖,𝑡= 𝛼 + 𝛽𝑅(𝑀𝑆𝐶𝐼 𝑊𝑜𝑟𝑙𝑑 𝐼𝑛𝑑𝑒𝑥),𝑡+ 𝑒𝑖,𝑡 ~ 𝑁(0, ℎ𝑖,𝑡)

Both equations are calculated simultaneously using the maximum likelihood method. Derived from equation 1, 𝑒_{𝑖,𝑡−1}2 is the ARCH term. Can be explained as the previous days information about US returns. ℎ𝑖,𝑡−1 is the GARCH term, which stands for the previous days residual

variance of US returns.

Before doing the actual regression, the mean model has to be checked. The conditions of volatility clustering and an existing ARCH effect have to be met to make sure the model fits. Also, other conditions as heteroscedasticity and normality have to be checked before doing the GARCH regression.

The variance during every event period is predicted by the GARCH (1,1) model. To determine the average daily normal volatility for each election, a multiple is used:

𝜇̂𝑖,𝑡 =

(∑𝑡=−50𝑡=−550(ℎ̂𝑖,𝑡))

500

To estimate abnormal volatility the expected volatility is subtracted from the actual realised daily volatility in the event period of each election. After the difference is summed to create the cumulative abnormal volatility of election i:

𝐶𝐴𝑉𝑖 = ∑ (ℎ𝑖,𝑡− 𝜇̂𝑖,𝑡) 𝑡2

𝑇=𝑡1

After the cumulative abnormal volatility is calculated the average CAR are computed. All events combined are then tested using a student t-test with robust standard errors.

Political cycles

A table with stock market returns of the Dow Jones Industrial Average over presidential periods 1978-2016 will be created. Information included are all the 𝑝𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑠𝑖 and

𝑐𝑜𝑛𝑔𝑟𝑒𝑠𝑠𝑒𝑠𝑖 with corresponding data. The first period will consist of stock market returns

one day after the elections, after it will be every year of the presidential term.

Results

Stock market returns

In table 2 the abnormal returns and cumulative average abnormal returns over an event period of 10 and 5 days are visible. T=0 are the average abnormal returns on the election days of the selected years. With these average abnormal returns the CAAR are calculated and tested.

TABLE 2. Abnormal returns and cumulative abnormal returns (-5;5)

The results in table 3 shows that the CAAR with an event period of 5 days surrounding the election date are not significant at a 5% confidence level. The null hypothesis of CAAR=0 cannot be rejected when using T= (-2;2). For the period T= (-5;5) the table shows a different result. According to statistics the null hypothesis can be rejected at a confidence level of 5%. Thereby, we can conclude that presidential elections have an effect on the stock market returns and hypothesis 1 can be accepted when using an event period of T= (-5;5). Which is in line with prior research.

The difference between the periods can be explained by more power and lasting or delayed effects of elections. The results show that taking a longer event period is necessary to capture the election effect. This is also proven by prior research of Niederhoffer, Gibbs & Bullock (1970). This study finds that abnormal returns are significantly different from zero over a period of 17 weeks surrounding the elections. Increased pressure on post-elections speeches and providing information, simultaneously with globalization which increases traveling speed of information, the period might be a little smaller than in 1970.

Table 3 shows the CAAR when using different event periods. The cumulative average abnormal returns are different from zero when using a period in-between T= 5;5) and T= (-28;28). The significant period has shifted from 17 weeks to 8 weeks surrounding election days.

TABLE 3. T-test Cumulative Average Abnormal Returns

Event Period (days) (-2;2) (-5,5) (-7,7) (-10,10) (-14,14) CAAR -0.0006 (0.0021) -0.0038** (0.0017) 0.0060*** (0.0022) (0.0025) 0.0042* 0.0081*** (0.0025) Event Period (days) (-21,21) (-28,28) (-35,35) (-43,43) (-45,45) CAAR 0.0043* (0.0024) 0.0059*** (0.0021) (0.0019) -0.0020 (0.0017) -0.0011 (0.0020) 0.0021 With *significant at p<0.10 **significant at p<0.05 and ***significant at p<0.01 Table 4 shows the regression to explain the CAAR. All the variables are significant at a 5% confidence level or lower. The control variables, interest rate and CPI, are monthly data this to avoid serial correlation.

The table shows that electing a Democratic or a Republican president does have an impact on the cumulative average abnormal returns. Electing a Republican president over a Democratic one decreases the cumulative average abnormal returns in the event period. During presidential elections, the classical view, Republican presidents having a favorable effect on the stock market, does not hold. This makes that hypothesis 2 is being rejected.

The results concerning the influence of the Congress are found to be significant. Having a Republican majority in the Congress during elections has a more positive effect on the CAAR compared to congressional control by the Democratic party. Returns are thus higher in a period of 10 days surrounding elections under when having a Republican majority in the Congress. The fact that the coefficient is significant makes that the null hypothesis is being rejected and hypothesis 3 is true.

Total Democratic power explains itself in the constant, this is due a similar CPI and interest rate regardless the party in charge. It results in a negative effect on cumulative abnormal returns. Total Republican control on the contrary has a positive impact on CAAR.

TABLE 4. Test Descriptives Regression CAAR

Test Variable Regression

Constant -0.0532***

(0.0173)

Interest Rate (monthly) 0.0017**

(0.0008) CPI (monthly) 0.0002*** (0.0001) 0/1 Democratic/Republican President -0.0126*** (0.0034) 0/1 Democratic/Republican Congress 0.0207*** (0.0031) F-statistic 13.58 𝑅2 _0.3364 Model used: 𝐶𝐴𝐴𝑅 = 𝛼 + 𝛽1 ∗ 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒 + 𝛽2 ∗ 𝐼𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒 + 𝛽3 ∗ 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛|𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐 𝑃𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡 + 𝛽4 ∗ 𝑅𝑒𝑝𝑢𝑏𝑙𝑖𝑐𝑎𝑛|𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐 𝐶𝑜𝑛𝑔𝑟𝑒𝑠𝑠 + 𝜀

With *significant at p<0.10 **significant at p<0.05 and ***significant at p<0.01

Where interest rate and CPI are monthly. 0/1 President and 0/1 Congress are dummy variables.

Volatility

Table 5 shows the test statistic of the cumulative average abnormal volatility over a period of 10 days. The table shows that the CAAV is positive and significant, what means that presidential elections do affect the markets’ volatility. Hypothesis 1 on volatility can be accepted. A period of elections results in higher volatility on the stock market.

TABLE 5. T-test Cumulative Average Abnormal Volatility

Event Period

(days) (-5,5)

CAAV 0.0022***

(0.0005)

With *significant at p<0.10 **significant at p<0.05, ***significant at p<0.01 The regression on the cumulative abnormal volatility shows that the margin of victory above 50% has an increasing effect on volatility. This is not in line with results of prior studies. Goodell & Vähämaa (2013) and Bialkowski et al. (2007) concluded that the likelihood of a candidate winning should decrease volatility. This would make sense, because it is in line with the Efficient Market Hypothesis of information already being incorporated in stock prices. A different outcome might be realized, because of a different measure method of the probability proportions. It could be that percentage point above 50% is not a good measure of the odds of a candidate winning the elections. With these measurements hypothesis 2 is false.

Table 6 shows that a change of majority in Congress prior a presidential election does have an influence on volatility. A change of power makes the market’s volatility increase. The effect could be explained by uncertainty concerning political policy and future decisions, which is in line with the EMH and makes that hypothesis 3 is true.

TABLE 6. Regression on cumulative average abnormal volatility

Test Variable Regression

Constant -0.0117***

(0.0035)

Interest Rate (monthly) -0.0001

(0.0001)

CPI (monthly) 0.0001***

(0.0000)

Margin of Victory 0.0173***

(0.0042) 0/1 Change Majority Congress 0.0028***

(0.0009) F-statistic 4.95 𝑅2 0.2679 Model used: 𝐶𝐴𝐴𝑉 = 𝛼 + 𝛽1∗ 𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑅𝑎𝑡𝑒 + 𝛽2∗ 𝐼𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒 + 𝛽3∗ 𝑃𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑊𝑖𝑛 + 𝛽4∗ 𝐹𝑜𝑟𝑚𝑒𝑟 𝐶𝑜𝑛𝑔𝑟𝑒𝑠𝑠 + 𝜀𝑖,.

With ***significant at p<0.01, ** at p<0.05 and * at p<0.10

Where interest rate and CPI prices are monthly. Margin of victory is the percentage point of electoral votes above 50% of the president winning the elections. Change majority congress is a dummy variable, 1 when a change of control took place two years prior a presidential election and 0 otherwise.

Political cycles

Looking at stock market returns over political terms of different presidents will show whether structural patterns under certain political powers exist. Figure 4 shows the prices of the Dow Jones over the period 1981 until 2016, with corresponding presidents at that time. Every horizontal line indicates the end of a presidential term and the start of a new one. The figure also shows the credit crisis with its lowest point in 2009.

Figure 4. Historical stock prices on the Dow Jones and presidential terms

In table 7 the returns of the Dow Jones over different periods are tabled. It shows the return the day after an election and the average yearly returns over the four years during the presidential term. In the year 2002 there were as many Democratic as Republican representatives in Congress, so no majority was established, indicated by (-). Also, the table shows the president and its party together with the majority in Congress at that time. To see whether a difference in returns exists under Democratic or Republican control, the averages of both over the five created periods are shown. For the first period, one day after an election, a big difference is detected. Electing a Democratic president results in a negative result of 1.68%. In order to prevent the influence of the economic-crisis on conclusions, another calculation is made by excluding the elections of 2008. Still, this results in a negative return of 0.56% compared to a positive return of 0.20% when electing a Republican president. After testing whether there is a significant difference in the mean, no conclusions can be made. Testing the returns of Democratic versus Republican presidents a day after elections results in a p-value of 0.1017. Moreover, results and explanation can be found in table 8. Looking at the whole presidential term, table 7 shows that under Democratic presidents the overall performance of the market is better compared to control by Republican ones. This is in line with the results of the regression on stock market returns rejecting the classical view.

However, after doing unpaired t-tests over the sample, with Republican and Democratic returns as separate groups, table 8 does not show any significant results over a total presidential term.

According to Booth & Booth (2003), the last two years of a president being in charge results in the highest returns over the presidential term. Table 7 does not confirm these findings. Especially the last year of a presidential term shows a more negative result compared to the other years.

To get more knowledge about congressional power in relation to the stock market, the returns on the Dow Jones according to their majority in Congress under a certain president are shown. Only, the constricted amount of presidential elections makes that the explanatory power of this study is limited, what makes it hard to find the real causes of the results. Macro-economic factors and political guidelines might play a role. Further research on the influence of Congress should be done.

Over the period 1980-2016 the stock returns under different presidential and congressional control are merged and shown. Average D-D stands for the average return under control of the presidential and congressional parties respectively. In this case it means the average returns under a Democratic president and a Democrat majority in Congress. Comparing total Democratic control to D-R and R-R, shows a big difference in results especially on the first day after the elections.

The difference between R-R and D-D the day after elections is the only significant effect that is found. Table 8 shows that the returns one day after elections are bigger under R-R compared to D-D. Having a Democratic president combined with a Democratic majority in Congress results in a lower return the day after elections take place. Furthermore, not much can be said, because of the lack of data in this period. The research period should be bigger to be able to say something about structural differences.

TABLE 7. Stock market returns in presidential terms

Name Party Date

One day after Con gre ss Year 1 Year 2 Co ng res s Year 3 Year 4

Reagan Republican

04-Nov-80 1.70% R (0.03%) 0.08% R 0.06% 0.01%

Reagan Republican

06-Nov-84 (0.88%) R 0.05% 0.12% D 0.04% 0.04%

H.Bush Republican

08-Nov-88 (0.43%) D 0.08% (0.01%) D 0.08% 0.03%

Clinton Democrat

03-Nov-92 (0.91%) D 0.04% 0.02% R 0.09% 0.09%

Clinton Democrat

05-Nov-96 1.59% R 0.10% 0.07% R 0.07% 0.02%

W.Bush Republican

07-Nov-00 (0.41%) - (0.04%) (0.03%) R 0.06% 0.01%

W.Bush Republican

02-Nov-04 1.01% R 0.02% 0.05% D 0.05% (0.12%)

Obama Democrat

04-Nov-08 (5.05%) D 0.03% 0.06% D 0.03% 0.04%

Obama Democrat

06-Nov-12 (2.36%) D 0.06% 0.05% R 0.01% 0.01% Average Democratic (1.68%) 0.06% 0.05% 0.05% 0.04% Average Democratic (excluding 2008) (0.56%) 0.07% 0.05% 0.06% 0.04% Average Republican 0.20% 0.02% 0.04% 0.06% (0.01%) Overall (0.64%) 0.04% 0.04% 0.05% 0.02% Average D-D (2.18%) ** 0.02% 0.02% 0.03% 0.04% (Excluding 2008) (1.64%) * 0.05% 0.04% - - Average D-R 1.59% 0.10% 0.07% 0.06% 0.04% Average R-R 0.36% ** 0.00% 0.06% 0.05% (0.02%) Average R-D (0.42%) 0.02% (0.02%) 0.07% 0.02%

With ***significant at p<0.01, ** at p<0.05 and * at p<0.10

Numbers surrounded by () are negative. The last column of rows indicates presidential and congressional power combined. D-D = Democratic – Democratic, D-R = Democratic – Republican, R-R = Republican – Republican and R-D = Republican – Democratic respectively.

TABLE 8. Unpaired t-test

One day

after over a term Average

p-value

Republican vs Democratic

President 0.1017 0.1529

p-value D-D vs R-R

President and Congress 0.0265** 0.1739

With ***significant at p<0.01, ** at p<0.05 and * at p<0.10

One day after: one sided test with H0: No difference in mean of returns D or R vs H1: D returns<

R returns

Average term: one sided test with H0: No difference in mean of returns D or R vs H1: R returns<

Conclusion

Based on the research that has been done, we can conclude that US presidential elections do affect stock market returns and volatility. The tests show that the days surrounding elections contain abnormal economic behavior. The average abnormal returns as well as the average abnormal volatility are different from zero when taking an event period of (-5;5). A period smaller than (-5;5) concerning stock market returns, provides different results. The influence of elections cannot be confirmed when taking a period smaller than 10 days surrounding election days.

The research of Niederhoffer, Gibbs & Bullock (1970) states that stock market returns show abnormal behavior 17 weeks surrounding the election day. Testing different event periods in this study shows that abnormal returns are different from zero from 10 days surrounding the event until 8 weeks surrounding a presidential election. A period larger than (-28,28) results in insignificant p-values. We could conclude that since 1970 information speed is higher through which stock market returns reflect election effects sooner. The election effect on stock market returns shifts from 17 weeks surrounding election day to 8 weeks surrounding the event using elections after 1980.

Regarding the stock market returns a part of the results can be explained by the regression on the cumulative average abnormal returns. According to this study we reject the ‘classical view’. Republican presidents have a more negative influence on stock market returns compared to Democratic presidents.

The impact of the Congress on stock market returns has been examined but mostly no significant evidence was found. In this study, the variable of congressional power and its majority does have an influence on the returns surrounding elections. The effect is found to be significant with a p-value lower than 0.01. The regression tells that a Republican majority has a positive effect on stock market returns, when comparing to Democratic power. Full Democratic power, having a Democratic president and majority in Congress, results in a negative effect on stock market returns. Combining these findings and the outcomes in table 7, suggests that electing a Democratic president together with having a Republican majority in Congress is the most favorable situation concerning stock market returns. However, the only significant results in table 7 are found when comparing total R-R and D-D. To be able to look into the suggestions of the regression on the CAAV, more data on stock returns under R-D is needed. Further research is necessary to find the underlying reason and to see whether the result is structural.

The hypothesizes on volatility are built upon prior research and the Efficient Market Hypothesis. The research that has been conducted concludes that a higher percentage point above 50% of electoral votes contributes to an increase in volatility. Which is not in line with EMH nor with prior research.

The affect of a change in majority in Congress prior presidential elections, result as expected in higher volatility. Changes increase uncertainty what makes that market returns are less stable.

Still, there are a few factors that cause limitations when relying on the generated results. First, the MSCI world index consists of over 50% of US firms. What makes that election effects are already highly incorporated in the returns of the market index. The high correlations between the Dow Jones Industrial Average and the MSCI world index, results in a beta close to 1. On the other hand, the influence of the US on global economy is so big that the problem is inevitable. Thereby, this study makes use of a restricted number of presidential elections causing that results are bound to a time constraint over the years 1978 until 2016. Also, the multiple to explain normal volatility is calculated by using the average of the predicted volatility, which might not be an economically realistic measure. The same holds for calculations regarding to the likelihood of winning. Polling or betting data might be a better representation of the likelihood that a certain candidate will win the coming presidential elections.

In summary, the significant results on congressional power indicate that more research on the subject should be done to fully understand, predict and forecast market behavior during elections.

Bibliography

Alesina, A. (1987). Macroeconomic policy in a two-party system as a repeated game. Quarterly Journal of Economics. Vol. 102, pp. 651-678.

Berk, J., DeMarzo, P. (2014). Corporate Finance, 3rd _{edition. Harlow: Pearson Education}

Limited

Bialkowski, J., Gottschalk, and K., Wisniewski, T. P. (2006). Stock market volatility around national elections. Journal of Banking and Finance, Vol. 32, pp. 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, Vol. 12, pp. 131-159. Brown, S., Warner, J. (1985). Using daily stock returns: The case of event studies.

Journal of Financial Economics, Vol. 14, pp. 3-31.

De Jong, F. (2007). Event studies methodology. Lecture Notes for Empirical Methods in

Finance, 1-44, Tilburg, Netherlands: Tilburg University.

Goodell, J.W., Vähämaa, S. (2013). US presidential elections and implied volatility: The role of political uncertainty. Journal of Banking and Finance. Vol. 37, pp. 1108-1117. Hwang, S., Valls Pereira, P.L., (2006). Small sample properties of GARCH estimates and

persistence. European Journal of Finance 12, 473-494.

Jinliang, L., Born, J.A. (2006). Presidential election uncertainty and common stock returns in the United States. Journal of Financial Research, Vol. 29, pp. 609-622.

Jones, S. T., Banning, K. (2009). US elections and monthly stock market returns. Journal of Economics and Finance, Vol. 33, pp. 273-287.

Niederhoffer, V., Gibbs, S., & Bullock, J. (1970). Presidential Elections and the Stock Market. Financial Analysts Journal, Vol. 26, pp. 111-113.

Nippani, S., Medlin, W.B. (2002). The 2000 Presidential Elections and the stock market. Journal of Economics and Finance. Vol 26, pp. 162-169.

Pantzalis, C., Stangeland, D.A., & Turtle H.J. (2000). Political elections and the resolution of uncertainty: The international evidence. Journal of Banking and Finance. Vol 24, pp. 1575-1604.

Riley, W., Luksetich, W (1980). The market prefers Republicans: Myth or reality. The Journal of Financial and Quantitative Analysis, Vol. 15, pp. 541-560.

Santa-Clara, P., Valkanov, R. (2003). The Presidential Puzzle: Political Cycles and the Stock Market. The Journal of Finance. Vol. 58, pp. 1814-1872.

Siegel, J.J. (2002). Stocks for the long run, 3rd _{edition. New York: McGraw-Hill}

Smales, L.A. (2016). Time-varying relationship of news sentiment, implied volatility and stock returns. Applied Economics. Vol. 48, pp. 4942-4960.

Visser, M.P. (1978). Volatility proxies and GARCH models. Amsterdam: University of Amsterdam.

Appendix

TABLE 9. Breusch-Pagan / Cook-Weisberg test for heteroscedasticity

Fitted values of DJ returns

Chi2(1) = 6834.21

Prob > chi2 = 0.0000

H0: Constant variance vs H1: Non-constant variance.

TABLE 10. White test for normality

White test

Chi2(2) = 2258.66

Prob > chi2 =0.0000

H0: Homoscedasticity vs H1: Unrestricted heteroscedasticity. Cameron & Trivedi’s decomposition of IM-test

Source Chi2 Df P

Heteroscedasticity 2258.66 2 0.0000

Skewness 0.01 1 0.9051

Kurtosis 2.17 1 0.1406

Total 2260.84 4 0.0000

TABLE 11. LM test for ARCH affect

LM test for ARCH

Lags(p) Chi2 Df Prob > chi2

1 1634.902 1 0.000

TABLE 12. The GARCH model

Return DJIA Coef. Std. _Err. z P>|z| [95% Conf. Interval] Return DJIA Return MSCI 0.9780*** 0.0041 237.48 0.00 0.9699 0.9861 Constant 0.0000 0.0001 0.14 0.889 -0.0001 0.0002 ARCH Arch L1 0.2387*** 0.0153 15.61 0.000 0.2087 0.2687 Garch L1 0.7478*** 0.0395 18.95 0.000 0.6705 0.8251

Constant 1.81e-07 1.34e-₀₆ 0.13 0.893 -2.45e-06 2.18e-06 With ***significant at p<0.01, ** at p<0.05 and * at p<0.10