Bachelor Thesis Economics and Finance July 2014
The relation between politics and stock markets
An event study on the effect of presidential elections in the United States on stock market returns
By Wijnand van Twisk 10247084
2 Table of contents
Table of contents 2
1. Introduction 3
2. Literature review 5
3. Presidential elections in the United States 8
4. Methodology and data 10
4.1 Methodology 10
4.2 Data and sample 12
5. Results 15
5.1 Event periods (t = -5, t = +5) and (t = -10, t = +10) 15
5.2 Event period (t = -15, t = -1) 20
5.3 Average cumulative abnormal returns 22
5.4 Discussion 23
6. Conclusion 25
3 1. Introduction
The relation between economic performance and politics is discussed oftentimes in research. In many studies, the effect of politicians or government policies on macroeconomic indicators such as gross domestic product, inflation - and unemployment rates is investigated. However, stock market
performances might be another interesting resource of data in order to investigate the relation between economic performance and politics. Since stock prices reflect all currently available information, according to the Efficient Market Hypothesis (Fama, 1970), stock prices will immediately be affected if there is new information available. This hypothesis might also hold for political news created by, for instance, presidential elections in the United States. In the run up to presidential elections there arises uncertainty about future economic policy. However, when the elections are coming to an end, more information concerning future economic policy will be made public and, therefore, stock prices might be affected.
This thesis investigates the possible relation between presidential elections in the U.S. and stock market returns. Therefore, the aim of this thesis is to answer to following research question:
Do presidential elections in the United States have a significant effect on stock market returns?
The event study methodology for daily stock returns, as described by Brown and Warner (1985) and further explained by MacKinlay (1997), is used to study the possible relation between presidential elections in the U.S. and stock market returns. In an event study, the importance of an event, for instance presidential elections, can be studied by analyzing changes in daily stock market prices (Bodie et al., 2011). More specific, the influence of an event on stock market returns is measured by estimating abnormal returns in the days surrounding the day of the event. The next step is to test whether the impact of the event is significant or not.
Several studies investigated the relation between political elections and stock market returns before. Still, there is some contradiction on this topic. Riley Jr and Luksetich (1980) showed that investors prefer Republican over Democratic presidencies in the short-term. They found that the presidential elections influence stock market returns, since the stock market increases in the 8 weeks following a victory of a Republican candidate. Also, in an earlier study, the results of Niederhoffer et al. (1970) suggest that investors prefer a Republican over a Democratic presidency in the short-term. On the other hand, Jones and Banning (2009) found a little relation between presidential elections in the U.S. and the stock market, using a time period of 104 years. Only in a relatively few cases they found a significant relation between the two.
Additionally, the majority of the studies that investigated this relation are not recent.
4 thesis more recent presidential elections are taken into measure. The sample includes 10 presidential elections, covering the years 1976 to 2012. This equates to a sample covering 36 years. For the reason that presidential elections will take place in the future again, it is important to know whether these elections influence the stock market so that investors might anticipate on a possible effect.
The remainder of this thesis is structured as follows. In the following section, a literature review about the relation between political elections and stock market performances is provided. Section 3 describes how presidential elections in the U.S. work in order to get some background information before starting with the empirical research. The methodology of the event study and the clarification of the data are presented in Section 4. In Section 5, the results are described and discussed. Section 6 concludes.
5 2. Literature review
The relation between the stock market and presidential elections is discussed in several studies. In most of the cases, the literature is about the traditional Wall Street believe that the market prefers Republican over Democratic presidencies. Furthermore, the interaction between politics and stock markets is not only examined in the U.S., but also in some other countries.
Niederhoffer et al. (1970) wanted to see if the Wall Street belief that investors prefer a Republican candidate to win the presidential elections is in accord with the market movements around Election Day. They hypothesized that after the victory of a Republican candidate, the Dow Jones Industrial Average index would rise. They did not reject this hypothesis, since on eight of the nine times a Republican has won the elections, the market has risen. In case of a Democratic candidate winning the elections, only on four of the nine times the market has risen. However, this effect only holds for the short-term since Niederhoffer et al. found no systematic differences in the performance of the market during Republican and Democratic presidencies. The findings of Riley Jr and Luksetich (1980) also show that investors prefer Republican over Democratic presidencies in the short-term. They found that the stock market increases in the 8 weeks following a Republican victory. However, in the long run the stock market under Democratic presidents outperforms Republican presidencies by, on average, 5%. Santa-Clara and Valkanov (2003) studied the difference in the excess returns between Republican and Democratic presidencies using data since 1929. They found significant evidence that the excess return in the stock market is higher under Democratic than under Republican presidencies. The difference in excess returns exists because real stock market returns are higher and the real interest rates are lower if there is a Democratic president.
Jones and Banning (2009) investigated the relation between U.S. presidential elections and monthly stock market returns over a time period of 104 years. The authors found a little relation between the two, but in a relatively few cases this relation was significant. Furthermore, they found that the believe that the returns are higher in the third and the fourth year of the presidential term, turns out to be weaker than is stated by other studies. Also, a lot of different combinations of partisan control of the Senate and the White House were used to investigate if stock market returns differ in certain situations. Nonetheless, no significant evidence has been found.
Research on the effect of political uncertainty on stock markets is done by Goodell and Vähämaa (2013). They studied the relation between presidential elections in the U.S. and volatility in the stock market. They found that volatility of the S&P 500 index increases when the probability of success of the eventual winner increases. The authors believe that this result indicates that presidential elections go along with uncertainty of investors about future macroeconomic policy. Julio and Yook (2012) also studied political uncertainty in relation with economic outcomes. They found that the timing of national elections around the world is related with corporate investments. They documented
6 that firms invest, on average, 4,8% less in the period leading up to an election than in nonelection years. The authors think that the hypothesis of political uncertainty best explains the findings. Thus, firms scale down their investments until the moment that the political uncertainty surrounding an election is reduced.
The study performed by Herron et al. (1999) was about the effect of the presidential election in 1992 in the U.S. on several economic sectors. They studied this effect by regressing equity rates from the Iowa Political Stock Market on the change of the chance of a candidate winning the elections for each sector. They identified 15 sectors, of 74 examined, whose profits are influenced by the presidential elections and thus can be seen as political sensitive.
The effect of elections on stock market performance is subject to research outside the U.S. too. Jensen and Schmith (2005) investigated the effect of the presidential elections in Brazil in 2002 on the Brazilian stock market. They estimated the effect of changes in the expected probability of a presidential candidate winning the elections on Brazilian stock market. By doing regressions in several time-series, the effects on the mean and variance of the Brazilian stock market are studied. They found no relation between the increase in probability of a candidate winning the elections and the mean return of the Brazilian stock market. However, they found evidence that there was more volatility in the stock market during the elections. This was because of the uncertainty about the future policy of the likely winner and not because of an uncertain election outcome. Döpke and Pierdzioch (2006) analyzed the relation between the stock market and politics in Germany. They only found weak evidence for interaction between stock market returns and the political process. They mention that their political variables could possibly be not strictly exogenous. Also, they found no evidence that the market prefers right-wing over left-wing governments. Moreover, Döpke and Pierdzioch did not find statistically significant evidence for the existence of an election cycle in the German stock market. Füss and Bechtel (2008) also investigated the German stock market, but they did this in relation to only one election, namely the federal election in 2002. They found that if a right-wing party wins the election, it will be more attractive to invest in stocks. And, dividends will be minimal if a left-wing party wins. The results of this study show that the volatility of only the small German firms is positively correlated with the chance of a right-wing party winning the election. They believe that mid- and large-sized firms have already diversified away their political risk and therefore they were not affected by the elections. Hudson et al. (1998) studied post-war stock market performances under Tory (right-wing) and Labour (left-wing) governments in the United Kingdom. There is no statistically evidence that one of the parties performs better in terms of nominal or real returns. Yet, they note that elections affect stock market performances. Short-term analyses indicate that the market prefers a Tory government.
There are only a few studies that investigated if share prices are influenced by national elections for multiple countries. One of these studies has been performed by Pantzalis et al. (2000). They investigated the behavior of stock market indices across 33 countries surrounding political
7 elections in a time interval from 1974 to 1995. They studied the relation of the elections and the stock market returns in function of different indicators such as the country’s degree freedom with respect to economy, politics and press. Also, the timing of the election and the success of the incumbent in being re-elected is taken into account. Pantzalis et al. conclude that there are positive abnormal returns in the time period of two weeks before the election week. This effect is bigger in case of a victory of the opposition in less free countries.
8 3. Presidential elections in the United States
It is important to know how the presidential elections in the United States work before starting with the empirical research.
Presidential elections in the United States take place in every four year. In 1845, it was decided to do the elections on a uniform date. Since then, the elections are on the Tuesday after the first Monday of November. This day is called Election Day. On Election Day the electors of the Electoral College are chosen in each state on basis of popular votes. In each state, and Washington D.C. separately, the Electoral College has to give electoral votes to the presidential candidates. Every state has a certain amount of electoral votes to give. Heavily populated states, as California, have more votes than less populated states like Montana. In every state, except for Maine and Nebraska, there is a winner-takes-all system established. Thus, the candidate with the majority of the votes receives all the electoral votes in that state. The candidate with the majority of the electoral votes of the Electoral College, where the actual election takes place, wins the presidential election. Currently, at least 270 electoral votes are needed to get a majority of the votes.
In practice, since the 19th century the electors choose the presidential candidate that there are
pledged for (McLean et al., 1995). Only in the, very rare, situation of unpledged – or faithless electors
this is not the case. An unpledged elector is an elector that is not linked to a certain candidate, so this person can vote for any candidate. A faithless elector is an elector that is pledged to a certain
candidate, but votes for another candidate or does not vote at all. Considering this, the popular votes in each state, determined on Election Day, are highly important for the final result of the election. This information is useful for the following empirical research in terms of establishing the event day.
The battle for the presidency is currently dominated by only two parties, namely the
Republican Party and the Democratic Party. Since the presidential election of 1852, one of these two parties has won the elections. Other parties, for instance the Libertarian Party, receive a negligible amount of popular votes compared to the two traditional major parties. Since 1976, there have been ten presidential elections held. In five out of the ten elections there was a Republican winner and five times a Democratic one. An overview of the previous ten presidential elections is given in Table 1.
9 Table 1. United States presidential elections (1976-2012).
Election date Winner Other candidate*
1976, November 2 Jimmy Carter (Democratic) Gerald Ford 1980, November 4 Ronald Reagan (Republican) Jimmy Carter 1984, November 6 Ronald Reagan (Republican) Walter Mondale 1988, November 8 George H. W. Bush (Republican) Michael Dukakis 1992, November 3 Bill Clinton (Democratic) George H. W. Bush 1996, November 5 Bill Clinton (Democratic) Bob Dole
2000, November 7 George W. Bush (Republican) Al Gore 2004, November 2 George W. Bush (Republican) John Kerry 2008, November 4 Barack Obama (Democratic) John McCain 2012, November 6 Barack Obama (Democratic) Mitt Romney
*Candidates from other parties than the Republican and the Democratic Party are not taken into account due to the low percentage of the total popular votes (less than 1.0% of the total popular votes).
10 4. Methodology and data
4.1 Methodology
In order to investigate the possible relation between presidential elections in the U.S. and stock market returns, an event study will be performed. In their paper, Brown and Warner (1985) discuss an event study methodology for daily stock returns. Their method is also supported and explained by
MacKinlay (1997). This methodology will be used to study the effects of presidential elections in the U.S. on stock market performances.
First of all, a simple regression will be performed to estimate the Ordinary Least Squares parameters ̂ and ̂ for the estimation period.
with
( ) and ( )
In this regression, indicates the return on the stock at day t and stands for the return on the market at day t. To obtain the daily returns of the stocks which are needed in the regression [1], the following equation is used.
In this equation, represents the price of the stock at day t. After estimating the parameters ̂ and ̂ for the estimation period, the abnormal return of the Dow Jones Industrial Average index for each day in the event period can be computed. The abnormal return is the residual of the regression [1] and is calculated as follows.
̂ ̂
Here, the abnormal return for day t is presented as . Then, Brown and Warner (1985) state that the null hypothesis to be tested is that the mean excess return, ̅ , in the event period is equal to zero. However, MacKinlay (1997) state that aggregation of abnormal returns is required to test for a statistical significant effect. Aggregation of abnormal returns is defined as cumulative abnormal returns, ( ). The cumulative abnormal returns are calculated for the event period ( ).
11 ( ) ∑
Under the null hypothesis, presidential elections do not have a significant effect on stock market returns.
( ) and ( )
This sequence of actions is repeated for each election.
In order to test for an overall effect of presidential elections on stock market returns, the average of the cumulative abnormal returns using all elections can be computed and be tested for significance. Moreover, by aggregating return observations using multiple elections, tests become more useful, as reported by MacKinlay (1997).
Given the number (N) of presidential elections, ̅ is calculated as follows.
̅ ∑ with
( ̅ ) ∑ ( ̅ )
Thereafter, the average of the cumulative abnormal returns can be determined, as described by MacKinlay (1997). ̅̅̅̅( ) ∑ ̅ with ( ̅̅̅̅( )) ∑ ( ̅ )
If the null hypothesis is rejected, the presidential elections do cause abnormal returns in the stock market, meaning that the elections have an overall significant effect on stock market performances.
12 ̅̅̅̅( ) and ̅̅̅̅( )
From this point, conclusions can be drawn regarding the possibility that stock market returns are influenced by presidential elections.
According to MacKinlay (1997), the distributions of the abnormal - and cumulative abnormal return are respectively ( ( )) and ( ) ( ( )). For this reason, tests can be performed for both cumulative abnormal returns and average cumulative abnormal returns using several statistically significance levels. According to De Jong (2007), the test statistics are the following.
In the test statistic of ( ), n represents the number of trading days in the event period.
( )
( ) ( ( )) with
( ) √
The test statistic of ̅̅̅̅( ) has a student-t distribution. As stated previously, N represents the number of presidential elections. The number of degrees of freedom is equal to N-1, so for each event period there are nine degrees of freedom.
√ ̅̅̅̅( ) ̅̅̅̅( ) with ̅̅̅̅( ) √ ∑( ( ) ̅̅̅̅( )
4.2 Data and sample
To measure the impact of the event directly after the event day, it is important to use daily stock market returns, as stated by Brown and Warner (1985). Daily returns of the Dow Jones Industrial Average, or DJIA, index are used as return on the stock, indicated by . The Dow Jones Industrial Average index is a stock market index which contains 30 major U.S. companies. The average is price-weighted and it reliably indicates the basic trend of the market (Dow Jones Averages, 2014). Riley and Luksetich (1980) argue that the Standard and Poor’s Composite Index is a much broader and
13 representative index, but the daily prices of the DJIA index are available in a more consistent manner and for a longer period. To be specific, the data of the DJIA index goes back to May 26 1896. The DJIA index data is obtained from the database called Wharton Research Data Services. However, the data is only available to 2007. From 2008 to 2012, the DJIA index data is sourced from TR4DER (2014). Daily world index data is used as benchmark market index, indicated as . The data for this daily world index comes from the database named Datastream (ticker: TOTMKWD). For the reason that the world index data is only available from 1973, the first presidential election to be considered is the election of 1976. Since then, there have been ten presidential elections held. The most recent election was in 2012.
The timeline of the event study includes two parts, namely an estimation - and an event period. The estimation period captures the period prior the presidential elections in order to estimate the market trend. Secondly, there is an event period. The event period is the time interval surrounding the elections. In this period, the possible impact of the elections on the stock market returns can be investigated. The timeline for the event study is illustrated in Figure 1(MacKinlay, 1997).
Figure 1. Timeline for the event study.
Election Day is established as the event day and is defined as day 0 (t = 0). The importance of Election Day is already described in Section 3. Brown and Warner (1985) used a time interval (t = τ0,
t = τ2) that consists of only 250 trading days, but since the accuracy of the estimation increases if there
are more observations, more trading days are taken into consideration. The estimation period has not to be too close to the previous presidential election, because this might influence the estimates. Therefore, an estimation period of 500 trading days is chosen. This equates to almost two years prior Election Day, which is approximately half of a presidential term.
In this thesis, significance is tested using two event periods with different lengths surrounding Election Day. In the first case, the event period contains 11 trading days surrounding the elections (t = -5, t = +5). This time interval corresponds with the event period used by Brown and Warner (1985). Here, the estimation period starts 505 trading days before the event day and ends 5 trading days before Election Day (t = -505, t = -6). The total time interval of the event study contains 511 trading days (t = -505, t = +5).
14 In the second consideration, an event period of 21 trading days is used (t = -10, t = +10). The event period starts 5 (or 11) trading days before Election Day, because more information about possible outcomes of the election becomes available prior the elections. Therefore, there is less uncertainty about the election outcomes. According to the Efficient Market Hypothesis, stock prices fully reflect all currently available information and therefore this information will also be priced in (Fama, 1970). In the days following Election Day, investors will receive more information about the probable future policy. To record the impact of Election Day, the days before and after this day itself have to be taken into account. It is essential that there is no overlapping of the estimation - and the event period. The aim of the methodology is to capture the effect of the event by measuring the abnormal returns. In the case that the event period is included in the estimation period, the results will be biased (MacKinlay, 1997).
However, much uncertainty might be reduced in the days before Election Day, because of election polls and other predictions that are publicly available. Consequently, the results may already be priced in. For this reason, it also might be interesting to investigate the impact of presidential election on stock market returns applying an event period of 15 trading days prior Election Day (t = -15, t = -1). This is equal to a period of three weeks and is in accordance with the event period used by Pantzalis et al. (2000).They found that the positive and significant effect of elections on stock market returns is the strongest in the two weeks before the election week. In their study, they considered the time period surrounding political elections across 33 countries, so not only the United States. The timeline for the event study with an event period that only consist of days prior the event day is illustrated in Figure 2.
15 5. Results
The aim of this thesis is to answer the question whether presidential elections in the U.S. have a significant effect on stock market returns. In order to answer this question, the empirical results of the event study, as described in Section 4, are presented in this section. This section can be divided in four parts. First of all, the cumulative abnormal returns ( ) of the Dow Jones Industrial Average index for both event periods surrounding Election Day are estimated. The event periods in this part contain trading days before as well as after the event day. In the second part, the results of the analysis of the cumulative abnormal returns for the event period with trading days only prior Election Day are presented. Then, the results of the analysis concerning the average cumulative abnormal returns ( ̅̅̅̅) are presented. At last, the findings will be part of a discussion.
Before moving on to the results, two points have to be made clear. Firstly, all returns in the tables are stated in decimal form. Secondly, some information about the critical t-values is provided in Table 2. Because the test statistic of the cumulative abnormal returns is considered to have a normal distribution, as noted in Section 4.1, infinity is used as the number of degrees of freedom. For the average cumulative abnormal returns, the number of N – 1 degrees of freedom is used. Since ten presidential elections are taken into measure, the number of degrees of freedom equals nine.
Table 2. Critical values of t.
Degrees of freedom t.10 t.05 t.01
9 1.383 1.833 2.821
∞ 1.282 1.645 2.326
5.1 Event periods (t = -5, t = +5) and (t = -10, t = +10)
The number of observations in the estimation - and event period may differ for each election. This is because of the different number of closing days of the New York Stock Exchange (NYSE), the stock exchange in which the DJIA index is traded. During the presidential election of 2012, Hurricane Sandy hit New York City and therefore the stock exchange was closed for two days in the event period (t = -5 and t = -6). Also, the NYSE was closed on Election Days until the presidential election of 1980 (t = 0). This causes missing returns in the event periods, but the results give no different outcomes if the elections of 2012, 1980 and 1976 are included in the analyses. The descriptive data of stock market returns for the estimation period are given in Table 3. The description of data of the stock market returns for the event periods are presented in Table 4 and Table 5.
For the event period (t = -5, t = +5), the mean return of the DJIA index in the estimation period is 5 out of 10 times higher than the mean return of the DJIA index in the event period. In the case of the event period (t = -10, t = +10), the mean return in the estimation period is 6 out of 10 times higher compared with the event period.Comparing the mean returns in the estimation period with the
16 mean returns in the event period, the summary statistics suggests that there is no clear indication that the presidential elections affect the stock market returns.
Table 3. Summary statistics of the Dow Jones Industrial Average index returns (DJIA) and the world market index returns (WRLD), estimation periods.
With event period (t = -5, t = +5)
With event period (t = -10, t = +10) Date of
Election Day
Stock index
Obs Mean Std. Dev. Obs Mean Std. Dev. Min Max 1976, November 2 DJIA WRLD 484 500 0.0009301 0.0006605 0.0097437 0.0063722 484 500 0.0009059 0.0006745 0.0097797 0.0064129 -0.0257254 -0.0194414 0.0390783 0.0307619 1980, November 4 DJIA WRLD 485 500 0.0003154 0.0004866 0.0085379 0.0053282 484 500 0.0004001 0.0005468 0.0084902 0.005278 -0.0299195 -0.0237864 0.0404674 0.0216156 1984, November 6 DJIA WRLD 486 500 0.0004137 0.0004921 0.0091861 0.006043 485 500 0.0004475 0.000519 0.0092125 0.006056 -0.0216623 -0.0275841 0.0363265 0.0262693 1988, November 8 DJIA WRLD 485 500 0.0004047 0.0007358 0.0176471 0.0103461 484 500 0.0004333 0.0007683 0.0176564 0.0103558 -0.2261054 -0.0929566 0.1014878 0.0790478 1992, November 3 DJIA WRLD 485 500 0.0005431 0.0001076 0.0081402 0.0078825 484 500 0.0004825 0.0001074 0.0081621 0.0079002 -0.0392719 -0.0491785 0.0456772 0.0528646 1996, November 5 DJIA WRLD 485 500 0.0009871 0.0004402 0.0065172 0.004467 484 500 0.0010138 0.0004403 0.0066167 0.0044963 -0.0303526 -0.0182504 0.0202086 0.0141086 2000, November 7 DJIA WRLD 484 500 0.0004239 0.0003829 0.011573 0.008398 484 500 0.0002553 0.0003642 0.0115287 0.0083782 -0.056554 -0.0398813 0.0492715 0.0296873 2004, November 2 DJIA WRLD 481 500 0.0005423 0.0006081 0.0102074 0.0070395 481 500 0.0006738 0.0006625 0.0102867 0.0070981 -0.0360586 -0.0277115 0.0359059 0.0242269 2008, November 4 DJIA WRLD 482 500 -0.0007048 -0.0010277 0.0149578 0.0127027 481 500 -0.0004858 -0.0006647 0.0145553 0.0120638 -0.0787328 -0.0645117 0.1108033 0.0850819 2012, November 6 DJIA WRLD 483 500 0.000411 0.0001189 0.0107604 0.0105904 483 500 0.000425 0.0000983 0.0107689 0.0106233 -0.0554637 -0.0498231 0.0424079 0.0398329
17 Table 4. Summary statistics of the Dow Jones Industrial Average index returns, event period (t = -5, t = +5).
Date of Election Day
Obs Mean Std. Dev. Min Max
1976, November 2 10 -0.0007277 0.01008 -0.0180855 0.0129116 1980, November 4 10 0.0013693 0.0113487 -0.0186223 0.0170294 1984, November 6 11 0.0004291 0.0090477 -0.0103265 0.0132344 1988, November 8 11 -0.0030372 0.0085912 -0.0225376 0.0062638 1992, November 3 11 -0.0005079 0.0059008 -0.0090516 0.0111367 1996, November 5 11 0.0043781 0.0048228 -0.002294 0.0158736 2000, November 7 11 -0.001249 0.0112465 -0.021349 0.0155754 2004, November 2 11 0.0057844 0.0066402 -0.0018559 0.0175307 2008, November 4 11 0.0064606 0.0440841 -0.0504931 0.1087787 2012, November 6 10 -0.0026646 0.0101206 -0.0236266 0.0103967
Table 5. Summary statistics of the Dow Jones Industrial Average index returns, event period (t = -10, t = +10).
Date of Election Day
Obs Mean Std. Dev. Min Max
1976, November 2 20 -0.0005633 0.0082933 -0.0180855 0.0129116 1980, November 4 20 0.0019606 0.0116192 -0.0186223 0.0221391 1984, November 6 21 -0.0008432 0.0076859 -0.0151058 0.0132344 1988, November 8 21 -0.0020424 0.0082552 -0.0225376 0.0062638 1992, November 3 21 -0.0000873 0.0055492 -0.0090516 0.0111367 1996, November 5 21 0.0023564 0.0054456 -0.0072857 0.0158736 2000, November 7 21 0.0010865 0.0116276 -0.021349 0.0231478 2004, November 2 21 0.0025022 0.0070155 -0.0109419 0.0175307 2008, November 4 21 -0.0037019 0.0418439 -0.0581523 0.1087787 2012, November 6 19 -0.0021967 0.0098414 -0.0236266 0.0164955
18 In order to test the relation between presidential elections and stock market returns, the following hypothesis is tested.
( ) and ( )
If the null hypothesis is rejected, the presidential elections have a significant impact on the stock market returns. Table 6lists the regression results and the cumulative abnormal returns for the event period (t = -5, t = +5). 2 out of 10 presidential elections, in 1988 and 1996, have a significant effect on stock market returns. In 1988, a CA of -0.047019962 (t-value = -1.35) is observed. The CA in 1996 is 0.01841 (t-value = 1.74) and is significant at a 5% level. The impact of the presidential elections in the other years is not significant.
The regression results and the CAs for the event period (t = -10, t = +10) are presented in Table 7. Here, 3 out of 10 presidential elections significantly affect the stock market returns, namely the elections of 1980, 1988 and 2000. The CAs in 1980 and 2000 are statistically significant at a 10% level, and have a value of respectively 0.028522235 (t-value = 1.35) and 0.060166073 (t-value = 1.64). In 1988, the CA has a value of -0.085741953 and is significant at a 5% level (t-value = -2.03).
Since the presidential elections have a significant effect in a relatively few cases and the significant effects are not in accordance with each other using different event periods, the results suggest that there is no clear evidence observed that the elections significantly influence the returns of the stocks.
The world market index, used as Rmt, is significant at a 1% level for every presidential
election. This indicates that the word market index is a proper benchmark for the market index. Also, the adjusted R2 is relatively high for almost all presidential elections. Thus, the model of the event study methodology fits the data well.
19 Table 6.Regression results and cumulative abnormal returns, event period (t = -5, t = +5).
Date of Election Day Adj R2 Alpha 1976, November 2 0.8228 0.0000036 (-0.02) 1.366004 (47.37)*** -0.002839356 (-0.28) 1980, November 4 0.6310 -0.0003001 (-1.27) 1.256155 (28.79)*** 0.004971914 (0.28) 1984, November 6 0.6079 -0.0001705 (-0.65) 1.171803 (27.44)*** -0.018772064 (-0.96) 1988, November 8 0.3087 -0.0003086 (-0.46) 0.940368 (14.73)*** -0.047019962 (-1.35)* 1992, November 3 0.2815 0.000502 (1.60)* 0.545113 (13.81)*** -0.002677512 (-0.13) 1996, November 5 0.2446 0.000661 (2.56)*** 0.715166 (12.56)*** 0.01841 (1.74)** 2000, November 7 0.5021 0.000114 (0.31) 0.965811 (22.09)*** -0.013995836 (-0.56) 2004, November 2 0.5499 -0.0000956 (-0.30) 1.059861 (24.23)*** 0.0101706 (0.79) 2008, November 4 0.5517 0.0001773 (0.39) 0.8685438 (24.35)*** -0.026232461 (-0.31) 2012, November 6 0.7254 0.0002866 (1.12) 0.8554049 (35.70)*** -0.012064047 (-0.64) *Statically significant at a 10% level
** Statically significant at a 5% level *** Statically significant at a 1% level
Table 7. Regression results and cumulative abnormal returns, event period (t = -10, t = +10). Date of Election Day Adj R2 Alpha 1976, November 2 0.8232 -0.0000452 (-0.24) 1.362675 (47.43)*** 0.003890667 (0.28) 1980, November 4 0.6256 -0.0002925 (-1.23) 1.254391 (28.43)*** 0.028522235 (1.35)* 1984, November 6 0.6095 -0.0001744 (-0.66) 1.173077 (27.50)*** -0.010301986 (-0.40) 1988, November 8 0.3088 -0.0003032 (-0.45) 0.9392804 (14.72)*** -0.085741953 (-2.03)** 1992, November 3 0.2829 0.0004542 (1.45)* 0.5471784 (13.84)*** 0.006015379 (0.24) 1996, November 5 0.2511 0.0006767 (2.59)*** 0.7302024 (12.76)*** 0.016382779 (1.04) 2000, November 7 0.5079 -0.0000379 (-0.10) 0.9699587 (22.35)*** 0.060166073 (1.64)* 2004, November 2 0.5560 -0.0000275 (-0.09) 1.065085 (24.54)*** -0.011508479 (-0.63) 2008, November 4 0.5355 0.0000768 (0.17) 0.876955 (23.54)*** 0.030099107 (0.26) 2012, November 6 0.7259 0.0003265 (1.27) 0.8536233 (35.74)*** -0.023345191 (-1.08)
20 *Statically significant at a 10% level
** Statically significant at a 5% level *** Statically significant at a 1% level 5.2 Event period (t = -15, t = -1)
Considering the findings of Pantzalis et al. (2000), as previously mentioned, more significant evidence is expected to be found using the event period of 15 trading days before Election Day. However, the summary statistics of the DJIA index - and the world market index returns give a similar suggestion about the effect of the presidential elections on the stock market returns. In the estimation period, the mean returns of the DJIA index are in 6 out of 10 times higher compared to the mean returns in the event period. The summary statistics of stock market returns for the estimation period are given in Table 8 and the summary statistics for the event period (t = -15, t = -1) in Table 9.
Table 8. Summary statistics of the Dow Jones Industrial Average index returns (DJIA) and the world market index returns (WRLD), estimation - and event period (t = -15, t = -1).
Date of Election Day
Stock index Obs Mean Std. Dev. Min Max
1976, November 2 DJIA WRLD 484 500 0.0007426 0.0005784 0.0099246 0.0065322 -0.0350365 -0.0258835 0.0390783 0.0307619 1980, November 4 DJIA WRLD 484 500 0.0004332 0.000523 0.0084837 0.0052671 -0.0299195 -0.0237864 0.0404674 0.0216156 1984, November 6 DJIA WRLD 485 500 0.00038 0.0004986 0.0092637 0.006079 -0.0216623 -0.0275841 0.0363265 0.0262693 1988, November 8 DJIA WRLD 484 500 0.0004553 0.0007875 0.0176922 0.0103655 -0.2261054 -0.0929566 0.1014878 0.0790478 1992, November 3 DJIA WRLD 484 500 0.0004937 0.0001261 0.0081652 0.0079343 -0.0392719 -0.0491785 0.0456772 0.0528646 1996, November 5 DJIA WRLD 484 500 0.0009536 0.0004114 0.006644 0.0045077 -0.0303526 -0.0182504 0.0202086 0.0141086 2000, November 7 DJIA WRLD 484 500 0.0003303 0.0004247 0.0115282 0.0083177 -0.056554 -0.0398813 0.0492715 0.0296873 2004, November 2 DJIA WRLD 481 500 0.0007316 0.0007303 0.0103078 0.0071221 -0.0360586 -0.0277115 0.0359059 0.0242269 2008, November 4 DJIA WRLD 481 500 -0.0004386 -0.0006263 0.0137666 0.0114691 -0.0733315 -0.0642904 0.1108033 0.0850819 2012, November 6 DJIA WRLD 483 500 0.000433 0.0000768 0.0107871 0.0106464 -0.0554637 -0.0498231 0.0424079 0.0398329
21 Table 9. Summary statistics of the Dow Jones Industrial Average index returns, event period (t = -15, t = -1).
Date of Election Day
Obs Mean Std. Dev. Min Max
1976, November 2 15 -0.001872 0.0091079 -0.0130549 0.0171073 1980, November 4 15 -0.0015538 0.0093009 -0.0163435 0.0137481 1984, November 6 15 0.0014792 0.0090007 -0.0081573 0.0246595 1988, November 8 15 -0.0004588 0.0087896 -0.011307 0.0205496 1992, November 3 15 0.001835 0.0055739 -0.0065092 0.0113697 1996, November 5 15 0.0003604 0.0046211 -0.0072857 0.0063762 2000, November 7 15 0.0047191 0.0119055 -0.0145613 0.0231478 2004, November 2 15 -0.0001593 0.0070669 -0.0109419 0.0142041 2008, November 4 15 0.0004936 0.0460603 -0.0787328 0.1087787 2012, November 6 13 -0.0017735 0.0083773 -0.0182348 0.0103967
The results of regressing the DJIA index returns on the world market index returns are presented in Table 10. The cumulative abnormal returns are also included in this table. In 1976 and 2012, the CAs significantly differ from zero. The CA in 1976 is equal to 0.027352055 and is significant at a 1% level (t-value = 2.34). In 2012, the CA is -0.024403966 and is significant at a 1% level (t-value = -1.68). Again, 2 out of 10 presidential elections have a significant effect on the stock market returns. Thus, also for the event period (t = -15, t = -1) there is no clear significant evidence that the presidential elections affect the stock market returns in the 3 weeks in the run up to the presidential elections.
22 Table 10.Regression results and cumulative abnormal returns, event period (t = -15, t = -1).
Date of Election Day Adj R2 Alpha 1976, November 2 0.8277 -0.0000723 (-0.38) 1.361182 (48.17)*** 0.027352055 (2.34)*** 1980, November 4 0.6170 -0.0002249 (-0.94) 1.247391 (27.91)*** -0.012191075 (-0.81) 1984, November 6 0.6140 -0.0002205 (-0.86) 1.179453 (27.76)*** -0.020429501 (-0.81) 1988, November 8 0.3102 -0.0003023 (-0.45) 0.9424386 (14.77)*** -0.00734382 (-0.28) 1992, November 3 0.2816 0.000455 (1.45)* 0.5438159 (13.80)*** 0.023068276 (1.06) 1996, November 5 0.2518 0.0006372 (2.43)*** 0.7324904 (12.79)*** -0.000165512 (-0.01) 2000, November 7 0.5057 -0.000025 (-0.07) 0.9748696 (22.25)*** 0.032501839 (0.92) 2004, November 2 0.5578 -0.000045 (-0.14) 1.065307 (24.62)*** -0.011356422 (-0.61) 2008, November 4 0.5607 0.0000992 (0.24) 0.8938028 (24.77)*** 0.037160874 (0.31) 2012, November 6 0.7281 0.0003535 (1.38) 0.8544921 (35.94)*** -0.024403966 (-1.68)** *Statically significant at a 10% level
** Statically significant at a 5% level *** Statically significant at a 1% level 5.3 Average cumulative abnormal returns
By aggregating all the cumulative abnormal returns of each presidential election and determining the average, as described in Section 4, more useful tests can be performed since more data is included (MacKinlay, 1997). This leads to the following hypothesis to be tested.
̅̅̅̅( ) and ̅̅̅̅( )
If the average cumulative abnormal return is significant different from zero, the presidential elections have an overall significant effect on the stock markets. The results for this test are provided in Table 11. Using an event period of 11 trading days surrounding the elections, the average cumulative abnormal return over the 10 presidential elections turns out to be significant different from zero (t-value = -1.50). The effect is statistically significant at a 10% level. For the other event periods, there is no significant evidence found.
23 Table 11.Average cumulative abnormal returns over 10 presidential elections for each event period. Event period ̅̅̅̅ (t = -5, t = +5) -0.009004876 (-1.50)* (t= -10, t = +10) 0.001417863 (0.11) (t = -15, t = -1) 0.004419275 (0.60) *Statically significant at a 10% level ** Statically significant at a 5% level *** Statically significant at a 1% level 5.4 Discussion
The results in the previous parts indicate that the presidential elections in the U.S. have a significant effect on the stock market returns in a relatively few cases. This may suggest that there is not enough evidence to reject the null hypothesis. The reason for this could be that the prices are already priced in before Election Day, even before the 3-week period prior the elections. Fair (1996) state that election outcomes are quite predictable. In addition, Cutler et al. (1989) find that the stock market response on major political and world events is relatively small and that the market moves more often on days without important news announcements. However, there are significant effects observed surrounding some presidential elections, but these results differ in the multiple used event periods.
It could be interesting to study the predicted election outcomes to see whether there is a relation between the predicted election outcomes and the effect on the stock market returns. Interesting presidential elections are the ones that are wrongly predicted, since in that situation the election outcomes do not follow the expectations of the investors. Therefore, the actual election result is probably not priced in before Election Day. Gallup is a large organization that provides, inter alia, election polls. Gallup predicted the wrong election outcomes before some presidential elections. The wrongly predicted elections which are included in the sample are the presidential elections of 1976 and 2004 (Gallup, 2014). Only the event periods that cover trading days before as well as after Election Day are relevant, because the actual election result turns out to be different than the expectations of the investors before Election Day. In the findings of this thesis, the presidential elections of 1976 and 2004 do not have a significant effect on the stock market returns. This supports the suggestion that there is no clear evidence that the presidential election significantly influence the stock market returns.
It may also be interesting to look at the presidential election in which it is certain that there will be a new president. A new president has to be elected after a maximum of two presidential terms, covering eight years. It is probable that a new president will have another political policy and,
therefore, might have other effects on the economy. In the presidential election of 1976, 1988, 2000 and 2008 it was certain that a new president would be elected (Table 1). The presidential elections
24 before 1976 are not included in Table 1. Before the elections of 1976, Richard Nixon (Republican) was the president of the United States during eight years. Therefore, a new president would be elected in 1976 for sure. For both event periods in 1988, the presidential elections significantly affected the stock market returns. In 2000, a significant effect was only found in the event period that included 10 trading days before and after Election Day. In 1976 and 2008, there was no significant effect
observed.
Also, it could be interesting to study the situation in which the new president is a member of another party than the predecessor. This situation occurred in 1976, 1992, 2000 and 2008. Only during the presidential election of 2000, using an event period of 21 trading days, a significant effect was noticed. However, Snowberg et al. (2007) point out the potential criticism of the policy convergence. The candidates will converge to the same policy in order to gain as many votes as possible. That policy is the one of the median voter. This leads to candidates with almost the same policy and therefore the effect of presidential elections on the stock market will be diminished.
To summarize, there could be several circumstances that might cause a significant effect of the presidential elections on the stock market returns. But, no conclusions could be drawn from these observations (in Section 5.4) due to the small amount of data.
25 6. Conclusion
The goal of this thesis is to the answer the question whether the presidential elections in the United States have a significant effect on stock market returns. Using an event study methodology, tests are performed in order to determine whether the impact of the presidential elections on the cumulative abnormal returns and the average cumulative abnormal returns is significant. In this event study, the daily returns of the Dow Jones Industrial Average index are used as return on the stock and the world market index acts as benchmark market index. The sample includes ten presidential elections (1976-2012), covering a time interval of 36 years.
Firstly, two event periods surrounding Election Day are used in order the estimate the cumulative abnormal returns for each presidential election. In the event period which includes 11 trading days, 2 out of the 10 presidential elections (1988 and 1996) significantly affect the stock market returns. Using an event period that contains 21 trading days surrounding Election Day, in 3 out of the 10 times (1980, 1988 and 2000) a significant effect is observed. Secondly, an event period which includes 15 trading days before Election Day is used in order to estimate the cumulative abnormal returns. In this event period, 2 out of the 10 elections (1976 and 2012) have a significant effect on the stock market. Overall, only in a relatively few cases the presidential elections have a significant influence on the stock market returns. Additionally, the significant results using one event period are not in accordance with the results using another event period.
Then, the average of the cumulative abnormal returns using all presidential elections is tested for significance. More useful tests can be performed by using the data of multiple presidential
elections (MacKinlay, 1997). A significant effect is found in the event period of 11 trading days surrounding Election Day. However, using this event period to test for significant cumulative abnormal returns in each presidential election separately, only in 2 out of 10 elections the null
hypothesis is rejected, as stated previously. In the other event periods, there are no significant average cumulative abnormal returns observed.
Taking everything into account, the results suggest that there is no clear significant effect of the presidential elections in the U.S. on the stock market returns. This may indicate that there is not enough evidence to reject the null hypothesis. A reason for this might be that there is not one moment that the outcomes of the presidential election are priced in, meaning that the effect of the presidential elections on the stock market returns might be evenly distributed over a longer period.
26 References
Bodie, Z., & Kane, A., & Markus, A. J. (2011). Investments and Portfolio Management (9th edition, Global edition). Singapore: McGraw-Hill.
Brown, S., & Warner, J. (1985). Using daily stock returns: the case of event studies. Journal of
Financial Economics, 14(1), 3-32.
Cutler, D. M., & Poterba, J. M., & L. H. Summers, L. H. (1989). What moves stock prices? Journal of
Portfolio Management, 15, 4-12.
De Jong, F. (2007). Event studies methodology. University of Tilburg.
Döpke, J., & Pierdzioch, C. (2006). Politics and the stock market: evidence from Germany. European
Journal of Political Economy, 22(4), 925-943.
Fair, R.C. (1996). The effect of economic events on votes for president: 1992 update. Political
Behavior, 18(2), 119-139.
Fama, E.F. (1970). Efficient capital markets: a review of theory and empirical work. Journal of
Finance, 25(2), 383-417.
Füss, R., & Bechtel, M. (2008). Partisan politics and stock market performance: the effect of expected government partisanship on stock returns in the 2002 federal German election. Public
Choice, 135(3-4), 131-150.
Goodell, J.W., & Vähämaa, S. (2013). US presidential elections and implied volatility: the role of political uncertainty. Journal of Banking and Finance, 37(3), 1108-1117.
Hudson. R., & Keasey, K., & Dempsey, M. (1998). Share prices under Tory and Labour governments in the UK since 1945. Applied Financial Economics, 8(4), 389-400.
Jensen, N.M., & Schmith, S. (2005). Market responses to politics, the rise of Lula and the decline of the Brazilian stock market. Comparative Political Studies, 38(10), 1245-1270.
Jones, S.T., & Banning, K. (2009). US elections and monthly stock returns. Journal of Economics and
Finance, 33(3), 273-287.
Julio, B., & Yook, Y. (2012). Political uncertainty and corporate investment cycles. The Journal of
Finance, 67(1), 45-84.
27 MacKinlay, A.C. (1997). Event studies in economics and finance. Journal of Economic Literature,
35(1), 13-39.
McLean, I., & Urken, A.B., & Hewitt, F. (1995). Classics of Social Choice. Michigan: University of Michigan Press.
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 and Finance, 24(10), 1575-1604.
Riley Jr, W.B., & Luksetich, W.A. (1980). The market prefers republicans: myth or reality? Journal
of Financial and Quantitative Analysis, 15(3), 541-559.
Santa-Clara, P., & Valkanov, R. (2003). The presidential puzzle : political cycles and the stock market. Journal of Finance, 58(5), 1841-1872.
Snowberg, E., & Wolfers, J., & Zitzewitz, E. (2007). Partisan impacts on the economy: evidence from prediction markets and close elections. The Quarterly Journal of Economics, 122(2), 807-829.
Databases:
Datastream
Wharton Research Data Services
Websites:
The American Presidency Project: www.presidency.ucsb.edu Dow Jones Averages: www.djaverages.com
TR4DER: www.tr4der.com Gallup: www.gallup.com
