The European Union, the Eurozone and the equity markets during parliamentary elections

(1)

UNIVERSITY OF AMSTERDAM

MSc Business Economics, Finance track

MASTER THESIS

The European Union, the Eurozone and the equity markets during parliamentary elections.

Kristina Palovicova, 10604685 December 2014

Supervisor: Dr. Florian Peters

Abstract: Using the equity indices in the European Union countries this thesis analyzes the effect

after the parliamentary elections and Eurozone membership on the indices of the EU countries such as OMX index, DAX, and AEX. The expectations were that the countries inside of the Eurozone experience lower volatility after the elections than countries outside of the Eurozone due to the common monetary policies, tighter fiscal policies and better price stability. The volatility was not significantly higher or lower after the elections for countries inside or outside of the Eurozone, although the Eurozone membership overall lowers the volatility of the equity indices. Hence, the link between the parliamentary elections and the membership of the Eurozone was not proved.

(2)

2

The political events such as elections have an influence on the economies around the world. Niederhofer et al. (1970) first indicated that politics affect the stock markets, and Nordhaus (1975) showed that so do political cycles. These studies were done in the United States, where numerous other studies proved the abnormal responses of the equity indices around the political elections1_{. The}

Eurozone with its introduction of the common currency is becoming more synchronized in its business cycles as indicated by Furceri and Karras (2006). Whether the business cycles in the Eurozone are also affected by the political elections is yet unclear.

This study tests whether there is a reaction of the equity indices such as DAX, ATX and OMX in the EU markets after the election date. There is evidence of a reaction after the election in the U.S. (Pastor and Veronesi, 2013), and there are positive abnormal returns of the stock markets two weeks around the elections in OECD countries (Pantzalis et al., 2000). Białkowski et al. (2008) show on sample of OECD countries that the return variance can easily double around elections. Correspondingly, the research question is: Does the Election Day affect the stock markets in European countries? Is there lower volatility associated with the Eurozone membership?

The research question focuses on the effect of parliamentary elections and the Eurozone membership on the stock markets in the EU countries. My hypothesis states that Eurozone membership takes away part of the economic instability that countries without euro face through the currency risk and lack of common monetary policy. The volatility of the stock markets or the abnormal returns around the elections should be higher in non-EMU countries than in Eurozone countries with similar financial and economic situation. This hypothesis is in line with Furceri and Karras (2006), who claim that EMU lowers asymmetric shocks and increases price stability. Also, Garfinkel et al. (1999) prove that the uncertainty of elections has an effect on the volatility of exchange rates, which increases the currency risk. However, it is contrary of Bekaert et al. (2013), who find that the integration coefficient remained unchanged with the introduction of Euro.

(4)

4

The contribution of my results is that Eurozone membership lowers the overall volatility of the indices prices. However, the results indicate that the volatility of the equity indices is not higher after the elections. The Eurozone membership lowers the volatility of the equity indices overall, but there is no difference in the volatility before and after the parliamentary elections. Adding dummy variables about the election outcomes did not change the results of this study. The European Union acts as a superior over fiscal policies and the European Central Bank (ECB) sets the monetary policies for the EMU members. The member state governments still have power in many decisions such as socioeconomic regulations, development, and taxation. The research provides more detailed in sight into the reaction on the equity indices if the government also decides about monetary policy.

The European Union (EU) is a single market union that shares politico-economic visions and policies in the 28 member countries, and is based on the rule of law: the rules and regulations are stated in treaties and are approved by all member countries voluntarily. The EU originated as European Economic Community (ECC), established in 1958 to improve the economic cooperation and trading within the region. The founding countries of ECC were Belgium, Germany, France, Italy, Luxembourg and Netherlands. The European Union was created in 1993 in order to reflect the expansion of different spheres such as politics, capital and labor flow, development aid, et cetera. It operates through supranational institutions and commissions that are represented by all member countries. Economic and Monetary Union of the European Union (EMU) was established in 1993 and it is incorporated in Maastricht Treaty. In 1999, single currency (euro) was adopted by founding EU members except Denmark, Sweden and UK. The EMU regulates common economic, fiscal and monetary policies and shifts the power from local governments to supranational institutions of the European Union (Cini and Pérez-Solórzano Borragan, 2013).

However, each member country has its independent government. Li and Born (2006) show that the volatility of the equity market in the United States rises around the election date. Boutchkova et al. (2012) find that the political elections have an effect on volatility of the equity indices tested on sample of fifty countries. Whether there is a similar effect found in the EU with regards to Eurozone membership is tested in this study. I am looking at the integration between the political elections and

(5)

5

equity markets, more precisely the effect of the uncertainty created by upcoming elections on the stock market volatility on the indices prices and the EMU membership. Previous research shows that there is an effect of the political elections on stock market returns. The markets respond to the new information created by the elections, as well as other political decisions that have a possible impact on fiscal and monetary policies (Fisher, 1996). In the United States if the candidate in the elections is not dominant, the volatility of stocks rises as stated by Li and Born (2006). Also, Pastor and Veronesi (2013) show that the elections have stronger impact in economies with lower GDP, or if the political signals are stronger, for instance when the investors receive more information about the possible changes. The above mentioned papers state that there is an effect of the elections on the equity indices.

During the recent crisis in Europe, the stock markets responded with increased volatility on various political announcements, such as the cutting of Greece’s debt by half, a referendum announcement a week later, or the disagreement of the opposition with the referendum (Pastor and Veronesi, 2013). Such reactions show that even in Europe, the equity indices react to various political changes. There are empirical papers concerned with the effects of being in the Eurozone on the price stability and cost of capital. Most of the findings suggest that the EMU creates a more stable union in terms of economic indicators such as GDP, inflation, price stability, and cost of capital (Furceri and Karras, 2006; Hardouvelis et al., 2007). With the common monetary and tighter fiscal policies for the EMU members, the governments cannot affect the future policies to such an extend as the non-Eurozone countries.

The equity indices are retrieved from the Bloomberg database, with a starting date of 1990. I selected the main equity indices for countries in the EU, such as DAX, AEX and OMX. The starting date is different for some of the indices, since some of them commenced in later years (see Table 1). The later start does not significantly affect our results, since the studied sample is still large, having 126 elections in the sample. The election dates and the results and the information about the governments in all countries are hand collected from the NSD European Elections database and

(6)

6

confirmed on various databases and websites2_{. Dummy variables were created to test various}

scenarios of the elections. The membership of the Eurozone was indicated with the one of the dummy variables. Some EMU member countries were not the members from the beginning of the sample, and some elections were held during the time when they were not members. This study takes these possibilities into account. This research also tests if the volatility is higher when the elections were held early or if there was a change in the government; Pantzalis et al. (2000) indicate that the early elections create strong positive abnormal returns if the incumbent government loses. Santa Clara and Valkanov (2003) find that the right wing Democratic administration leads to higher returns than left wing administration in the US. With the last variable I indicate if the winning party is right or left wing and test if there is a higher volatility associated with the different governments.

The remainder of the study is organized as follows. Section 2 provides the literature review of studies written about political elections and the effects on the equity indices and the studies about the Eurozone membership and its economic benefits. In Section 3 the empirical methodology is introduced with the hypotheses and sample selection. Also, a robustness check is provided by changing the dependent variable of the study. Section 4 describes the data and the summary statistics and Section 5 introduces the results of this study and its limitations. The conclusion and further study recommendations are provided in the Section 6.

2. Literature Review

2.1 Elections and the Eurozone Membership

Maastricht convergence criteria for entering the EMU provide rules and indicators that have to be achieved, and are expected to lead the countries into better price stability and stronger market economies (Lipinska, 2008). The adoption of euro requires countries to comply with strict fiscal and monetary policies, and the governments make fewer changes in terms of creating policies and rules

2_{The website mainly used for election dates is NSD European Election database}

(http://www.nsd.uib.no/european_election_database/), the outcomes of the elections were collected from various news articles and different European elections databases ( http://www.theguardian.com/politics/european-elections; http://www.parties-and-elections.eu/ ), both retrieved in September, 2014.

(7)

7

that would contradict with the requirements. The hypothesis of my thesis is testing whether the EMU membership also affects the equity indices around the parliamentary elections. It is based on the predictions that the Eurozone membership creates better stability of the equity indices. The stability is created due to tighter monetary and fiscal policies and through excluding the currency risk premium on uncertainty.

Furceri and Karras (2006) find that the main benefit of entering the EMU is better price stability. This finding partially supports my hypothesis that countries outside of the EMU could have higher volatility of indices prices around the elections due to the currency risks and price instability. Hardouvelis et al. (2007) find that creation of the EMU lead to monetary and economic integration, and resulted in lowering the equity cost of capital of the Eurozone countries. The paper compared the members of Eurozone to three EU countries that did not adopt euro (UK, Denmark and Sweden) and found empirical evidence of reduced cost of capital within the EMU. Also Bris et al. (2008) find that corporate valuations in countries with previously weak currencies have increased after joining EMU due to the lower cost of equity and interest rates. My results show as well that the Eurozone membership lowers the overall volatility of the indices prices.

Comparatively, Furceri and Karras (2006) investigate the standard deviation of each country’s cyclical income. They find that the standard deviation of the cyclical output for countries that are the old EU members (countries creating EU or joining EU before 2004) declines in the last decade. The new EU members, that joined EU in 2004 and later (Bulgaria, Croatia, Czech Republic, Cyprus, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Romania, Slovakia and Slovenia), have the highest standard deviations of the EU members in cyclical income outputs. Furceri and Karras (2006) also investigated the correlation between cyclical incomes for each country compared to EMU’s cyclical income. They find on the contrary to previous research that the non-euro EU members are more synchronized than the Eurozone members.

The paper also investigates the effects of inflation rates in the EU countries. It shows that if the country entering Eurozone has higher inflation than the EMU members, it benefits from joining

(8)

8

the EMU monetary policies. Furceri and Karras (2006) prove that the inflation rates dropped after countries joined the Eurozone. On the other hand, the three old non-euro countries (UK, Denmark and Sweden) have better inflation than the EMU countries, which shows no effect of better long-term price-stability by entering the Eurozone. Morana and Beltratti (2002) find that after introduction of the euro, the volatility of less stable European equity indices decreased, specifically in Italy and Spain. They also state that euro brought stability in terms of volatility of equity indices to the EMU as a whole, due to stable strong economies like Germany imposing to the weaker economies. As stated before, the above mentioned papers are in line with my results that the Eurozone membership lowers the volatility of the equity indices.

Niederhofer et al. (1970) first examined the link between the elections and stock market by studying the market behavior around the elections in the US. They find that stock prices increased by 1.13% if a Republican candidate wins the elections, and decreased by 0.81% if a Democratic candidate wins on the 1990-1969 sample period. The effect of presidential elections in the U.S. is proved to influence the Canadian stock market by Foerster (1994). Gwilym and Buckle (1994) studied the behavior of stock prices around elections in the UK. They prove that there is a big increase in the volatility of FTSE 100 around the British elections in 1989. Most of the research done indicated that there are higher returns or bigger volatility around the political elections and that the EMU membership improves the economic situation of the countries. However, my results did not prove the higher volatility after the elections.

2.2 Change in the Government

Pastor and Veronesi (2012) look at the effects of governmental policy changes on the equity indices. The governmental policy changes mostly appear when an incumbent government does not win the elections and the new government gets in the power. If there is a change in the political ideologies, as the change of the right versus left wing parties or opposite, the change in the policies creates uncertainty. Pastor and Veronesi (2012) find that if there is an uncertainty regarding the political change, there is a significant impact on the equity indices during that change. They

(9)

9

distinguish between two types of uncertainty, the impact uncertainty; which focuses on the possible impact of the policy change on investors, and political uncertainty; which arises from the possibility of implementing new policies.

During the elections, both of these uncertainty types can be experienced. If the elections are close and the winning party is uncertain, the political uncertainty is created. Most of the parties promise changes to their voters to be elected in the government again. Even if the polls mark a certain winner, there is an impact uncertainty created, as the investors are not sure about the effect of the new policies in the future. Pastor and Veronesi (2012) find that after an announcement of a policy change, the stock prices generally fall. They also find that the volatility of equity indices increases after the policy change announcement if there is a higher uncertainty about the effect of the policy impact.

Moreover, Białkowski et al. (2008) prove that when there is a change in the government after the elections, the volatility is higher than without the governmental change. Based on the above empirical findings, my predictions were that when there is a change in the government, the volatility of the indices prices will be higher compared to the incumbent government winning the elections due to the political uncertainty, and that the EMU membership lowers the volatility when there is a government change after the elections due to the common monetary and tighter fiscal policies that the governments experience. However, the results were contrary to my predictions, as the results were not statistically significant.

2.3 Right-Wing Government versus Left-Wing Government

Botero et al. (2004) argue that left-wing parties are more inclined to redistribute taxes to the working class, compared to right-wing capitalist parties with less social benefits. This research is done on sample of 85 countries. Rueda (2005) finds similar relation on the sample of the European countries; left-wing parties protect the labor more than right-wing parties via regulations and policies. Based on the same assumptions, Boutchkova et al. (2012) indicate that left-wing governments experience higher volatility of stock markets after the elections than right-wing governments in labor-intensive industries or in countries with strict labor employment laws. They argue that this is because

(10)

10

companies fear being less profitable due to the change from right to left wing governments, as they have to spend more on employees’ benefits, and will face less favorable conditions in terms of hiring and firing labor.

Fuss and Bechtel (2008) test whether there is partisan effect in the German federal elections in 2002. Rational partisan theory is based on the suggestions that there is a better firm performance under the right-wing than under the left-wing government. Fuss and Bechtel (2008) found that stock market returns were linked positively to the probability of the right-wing party winning, and negatively to the left-wing party winning. They also found that the volatility increased if the right-wing party had higher chance to win the elections, compared to the decreased volatility if the results were uncertain. These findings contradict the results of this study and findings of Boutchkova et al (2012) that there is a higher volatility if the left-wing party wins the elections. Pierdzioch and Döpke (2006) examine the German market as well and find no significant results for partisan effect on the election cycle.

In the United States, Riley and Luksetich (1980) and Niederhofer et al. (1970) find that under Republican administration the stock market has better returns than under Democratic administration in the period 1900-1980. This is contradictory to Santa Clara and Valkanov (2003), who find that under Democratic administration stock markets performed better in the period 1927-1998. Also Leblang and Mukherjee (2004) find that there is a decrease in the volatility of the equity indices if investors anticipate the right-wing party winning the elections. Based on the previous empirical research, I expected that the volatility is lower when the right wing party wins the elections compared to the winning of the left-wing party, and the EMU membership takes away the volatility compared to the non EMU members due to the common monetary and tighter fiscal policies. However, my results were not statistically significant; there was no difference in the volatility of the equity indices when the right wing or the left wing government was elected. These results are in line with Pierdzioch and Döpke (2006).

(11)

11 2.4 Early Elections

When elections are held early, voters have less time to study in detail all the policies proposed by the parties running for the government. That creates more uncertainty since there is time pressure on the voters to make their decisions about who to vote (Pantzalis et al., 2000). Also, the early elections are not held earlier without a particular reason. They mostly happen because of some issues with the previous government, either disagreement between the coalitions, or the current government wanting to prolong its ruling. The uncertainty about the future political outcome puts pressure on the equity indices as well (Pantzalis et al., 2000; Pastor and Veronesi, 2012; Białkowski et al., 2008).

Pantzalis et al. (2000) find that the average median and mean for the cumulative abnormal returns for early elections is significantly positive at 10% level, but also the mean and median for the rest was significantly different than zero. Overall, the results are not significant; there is no difference in cumulative abnormal results between “the early elections group” and “the not early elections group”. I test if there is a difference of having the early elections on the European sample, considering that calling for the early elections mostly comes from either the incumbent party wanting to improve their position, country unforeseen political events or pressure from the parliament asking for change (Pantzalis et al., 2000). The results indicate that the volatility of the equity indices is lower for the EMU members after the elections on the (-10; 10) event window.

3. Empirical Methodology

3.1 Hypotheses

Various studies prove that there is a relationship between the political events and financial markets. Specifically, there is a significant evidence of the stock market volatility around the election dates in the U.S. (Li and Born, 2006) and in the sample of fifty countries around the world (Boutchkova et al., 2012). I would like to propose main hypothesis of this study with its subtests. The first hypothesis tests whether there is higher volatility of equity indices after the elections in the European Union.

(12)

12

H1: The elections have stronger effect on the volatility of the stock market in countries outside of the Eurozone compared to the Eurozone members.

The first hypothesis of this study tests if there is a link between countries that are in the EMU compared to those who are not in the EMU but are EU members. Recall that countries in the EMU have better price stability, common monetary policies, tighter fiscal policies, and better economic integration that results into better equity cost of capital (Furceri and Karras, 2006; Hardouvelis et al., 2007). Therefore, EMU members are expected to have lower volatility during the time of the political elections, considering the local government having less power over the monetary and fiscal policies compared to the non-EMU members.

With the first hypothesis tested, subtests are created to see if there is a higher volatility during the change of the government after the elections as proved by Białkowski et al. (2008) and Pastor and Veronesi (2012). Accompanying the change in the government, this study also tests if there is an effect of having right-wing government or left-wing government and if there is a higher volatility when the elections are held early, as proposed by Białkowski et al. (2008) and Pastor and Veronesi (2012). Both of these studies did not find the test on early elections significant. Boutchkova et al. (2012) find that the volatility is higher with left-wing government being elected and Fuss and Bechtel (2008) find that in Germany, there is higher abnormal return associated with right-wing government being elected.

The validity of this study is tested by changing the dependent variable to absolute returns over the tested event windows. This is done in order to test if the equity indices reacted to the parliamentary elections by increased or decreased returns that were not volatile. This check focuses on the absolute returns, in order to observe a pattern in either direction of the returns on the equity indices. Pantzalis et al. (2000) find higher abnormal returns associated around the election dates. As mentioned earlier, they prove that the returns are higher when the elections are called early and there is a change in the government.

(13)

13 Eq uity in dex Au str ia 02 Ja nu ary 1990 08 O cto be r 1990 ATX 8 5 3 2 7 Be lg ium 02 Ja nu ary 1991 25 N ov em be r 1 991 BE L20 7 5 1 3 7 Bu lg aria 24 O ctob er 20 00 27 Ju ne 2005 SO FIX 5 0 2 3 4 Cro atia 14 Jun e 2 002 24 N ov em be r 2 003 CR O in dex 3 0 0 2 2 Cy prus 03 Se pte m be r 2004 22 M ay 2006 CY SMM APA 2 1 0 1 1 Cz ec h R epub lic 04 Ja nu ary 1994 03 Ju ne 1996 PX 5 0 1 4 2 D en m ark 02 Ja nu ary 1996 12 M arc h 1998 O M X C op enh ag en 5 0 3 2 3 Es ton ia 03 Jun e 1 996 08 M arc h 1999 TA LSE 4 1 0 2 4 Fin la nd 02 Ja nu ary 1990 18 M arc h 1991 H EX 6 4 0 4 2 Fra nce 02 Ja nu ary 1990 22 M arc h 1993 CA C40 5 3 0 4 3 G erm any 02 Ja nu ary 1990 03 D ec em be r 1 990 D AX 7 4 1 3 5 G ree ce 02 Ja nu ary 1990 09 A pri l 1990 AS E I N D EX 8 4 4 4 4 H un ga ry dropp ed fro m th e s am ple Ire la nd 02 Ja nu ary 1990 26 N ov em be r 1 992 ISEQ 5 3 2 1 5 Italy 02 Ja nu ary 1998 14 M ay 2001 FT SE M IB 4 4 1 4 2 La tvia 03 Ja nu ary 2000 07 O cto be r 2002 VIL SE 4 0 1 4 3 Lit hu an ia dropp ed fro m th e s am ple Lu xe m bo urg 04 Ja nu ary 1999 14 Ju ne 1999 LUXXX 4 4 1 0 4 M alta 07 D ec em be r 1995 28 O cto be r 1996 M AL TEX 3 2 0 1 2 N eth erl ands 02 Ja nu ary 1990 04 M ay 1994 AEX 7 5 2 3 5 Po la nd 16 Se pte m be r 1991 20 Se pte m be r 1993 W IG20 6 0 2 5 4 Po rtu gal 01 Ja nu ary 1993 02 O cto be r 1995 PS I20 6 5 3 4 2 Ro m an ia 22 Se pte m be r 1997 27 N ov em be r 2 000 BET 2 0 0 1 2 Slo va kia 01 O ctob er 19 96 28 Se pte m be r 1998 SAX 2 0 0 0 2 Slo ve nia 01 Se pte m be r 2003 04 O cto be r 2004 SBIT OP 4 3 2 4 2 Sp ain 02 Ja nu ary 1990 07 Ju ne 1993 IBEX 6 4 2 3 3 Sw ed en 02 Ja nu ary 1990 16 Se pte m be r 1991 O M X St oc kho lm 6 0 0 3 2 U nit ed K in gdom 20 Ju ly 2001 06 M ay 2005 LS E ind ex 2 0 0 1 1 To tal 126 57 31 68 83 Co un try G ov er nm en t ch an ge Rig ht -w in g go ve rn m ent D at a a va ila ble Fir st Ele ct io ns N um be r of ele cti ons Th e E MU Ea rly e le ct io ns D at a a va ilab ility a nd sa m ple co m po sit io n Ta ble 1 The first co lu mn lis ts all EU m em ber co un trie s. The da ta av aila ble from co lu mn re pre se nts the sta rt da te of the sp ec ific da ta th at w as ob ta in ed from Bloo m be rg da ta ba se. The na me of the equ ity in dex us ed for sp eci fic co un try, is re pre se nted in the th ird co lu m n. Va ria ble N um ber of ele cti ons re pre se nts how m any ele cti ons w ere he ld in sp eci fic coun try in our sa m ple. The va ria ble Ea rly ele cti ons coun ts how m any ele cti ons w ere he ld ea rly in sp eci fic coun try. The va ria ble G ov ern m ent ch an ge aft er ele cti ons re pre se nts how m any tim es in the sp eci fic coun try w as not the in cu m be nt pa rty ree le cted or did not cre ate new go ve rn m ent (th is is the ca se if the pa rty do es not ha ve enou gh se ats to cre ate a co ali tion ). R ig ht-w in g v ari ab le re pre se nts th e nu m be r w he n t he p art y e le cte d w as o rie nte d m ore to th e r ig ht.

(14)

14 3.2 Sample

For the equity indices sample, the selection starting date of the 1st_{of January 1990 was}

chosen. However, 18 countries have different starting dates due to the indices being established later than in 1990 (See Table 1 for the full list). For example, the Czech Republic has the index starting 4th

of January 1994, since the country was only created in 1993. The different starting dates do not raise any concern over the validity of this study, as the sample is large enough. For the election dates, the data was collected from various sources, mainly from the NSD European Elections Database, or from various newspaper articles and databases.

In many countries the elections were held over the weekend, so for the purpose of this study, the first business day after the elections was considered as a first day of the election window. For example, in Finland or Portugal the elections were held on Sundays, then Monday was the first day in the election window sample. If elections were held on Wednesday (in Netherlands), then Thursday was the first day in Election window, since the results are available usually late Wednesday night or early Thursday morning. In France, two rounds of elections are held with a week apart. Due to the election window being 20-business days, the longer elections do not affect the studied window. For the United Kingdom, the index chosen is LSE index, which has the start date on 20th_{of July 2001.}

In Hungary and Lithuania, the two rounds of elections are held two weeks apart, which could have consequences on the studied window, so the countries were dropped from the selection sample. In Estonia, Italy, Latvia, Slovakia and Slovenia the elections are sometimes held two days in a row, and mostly over the weekend. In these cases, Monday is the first day of the elections window used in this study. This approach is similar to Pantzalis et al. (2000), the study uses as the starting date of the sample first observation after the elections.

There are currently ten countries that are outside of the Eurozone (Bulgaria, Croatia, Czech Republic, Denmark, Hungary, Lithuania, Poland, Romania, Sweden and United Kingdom), two of these countries were dropped from the sample. Eleven member countries of European Union created the Eurozone 1st_{of January 1999. Since then, seven new countries joined. For the purpose of this}

(15)

15

research, the countries that adopted the euro have to be adjusted accordingly by the exact date of the currency adoption. This provides precise analysis if the adoption of euro took away the currency risk and the decisions about the monetary and to some extend the fiscal policies on stock returns also around the political elections.

3.3 Dependent and Independent Variables

The dependent variables in this study are the volatility of daily prices of equity indices across Europe and the absolute return of the equity indices across Europe. Volatility is computed on daily returns of the specific indices. Volatility is averaged in the event window and compared to the volatility of the pre-event window of the same length. This approach follows the methodology of Białkowski et al. (2008), who find two to five weeks after the elections is the most volatile period. However, I use more event windows in this study. The studied event windows were 10, 20, 25, 50, 75 and 100 days.

The second dependent variable that is tested in order to prove validity of this study is absolute return computed over the tested window. For each of the event windows, the absolute return is computed as:

= | − |

Where the equals the price at the end of the event window and the is the price on a first day of the event window, which is the first business day after the elections. This variable is focused on recognizing significant pattern of the returns after the elections, in either direction.

As the independent variables this study uses the timing of the elections; if the elections were early, the variable early elections equals one, or zero otherwise. The variable government change equals one if after the election there was a change in the leading party in the government, or zero otherwise. Next dummy variable equals one if the elected party was right-wing party or zero if it was left-wing party. The main independent variable of the study is whether the country is in the Eurozone

(16)

16

or not. If the country is member of EMU, then the dummy variable Eurozone equals one, and zero if the country does not have euro as a currency.

3.4 Model Specification

To test the effect of the elections on the equity indices in the European Union countries, the Ordinary Least Square (OLS) regression is used together with the Difference-in-Difference analysis. This analysis is used to measure the effect of a treatment. Treatment can be seen as an event, that can be measured by 0 or 1, if there was a treatment or not. This study considers the election date as a treatment. The test is performed in Stata and uses the daily returns of equity indices for all countries (See Table 1) obtained from Bloomberg terminal and the databases with exact election dates and the outcomes. My model specifications also test the effect of early elections, government change and the right-wing party being elected. The specifications are:

(1) = + + + ∗ + (2.1) = + + + ℎ + ∗ + ∗ ℎ + ℎ ∗ + ∗ ∗ ℎ + (2.2) = + + + ℎ + ∗ + ∗ ℎ + ℎ ∗ + ∗ ∗ ℎ + (2.3) = + + + + ∗ + ∗ + ∗ + ∗ ∗ +

With the first hypothesis (1) I test whether there is a volatility effect of the election on equity indices in the European Union. The variable called post equals one for the specified event window starting first day after the election date. Eurozone variable equals one if country is a Eurozone member, and zero if not. The variable is the error variable. If the coefficient is positive and significant, the results will be as predicted; the volatility of stock markets is higher after the elections

(17)

17

in the European Union. The variable is expected to be negative, as the countries that are in the EMU experience less volatility after the elections compared to the countries outside of the EMU, as mentioned in Section 2.1.

In the second sub hypothesis (2.1), the government change variable is added, which equals one if there is a change in the government, and zero otherwise. The coefficient is expected to be higher than the coefficient , since the volatility is expected to be higher when the government changes and there is an uncertainty created (Pantzalis et al., 2000; Białkowski et al., 2008). The coefficient is expected to be lower than the coefficient , in order to prove that the volatility is lower for the EMU members.

The third sub hypothesis (2.2) is testing whether there is a statistically significant result if the dummy variable right wing is added. The variable equals one if the right wing party won the specific elections. The coefficient is expected to be lower than the coefficient , as showed by Boutchkova et al. (2012) that left wing parties create higher volatility on the equity indices around the elections. The coefficient is expected to be lower than the coefficient in order to align with my hypothesis that the Eurozone takes away part of the volatility.

In the fourth sub hypothesis (2.3), the early elections dummy variable is added. The early elections variable equals one if the elections were held early, and zero otherwise. The coefficient is expected to be higher than the coefficient , as the early elections are expected to create more volatility than normal elections, although the previous research did not prove the hypothesis, see Section 2.4. It is ambiguous to see if the coefficient in the hypothesis (2.3) is higher or lower than the coefficient .

3.4 Robustness Check

The previous research done is divided into two types of the tests. The papers either test the volatility around the elections (Boutchkova et al., 2012; Leblang and Mukherjee, 2004; Białkowski et al., 2008), or the abnormal returns around the elections (Pantzalis et al., 2000; Santa Clara and

(18)

18

Valkanov, 2003). In order to prove the validity of this study, the next hypothesis tests the absolute returns around the elections in the EU. The same tests on the different dependent variable are conducted as on the volatility. My model specifications are:

(3.1) = + + + ∗ + (3.2) ₌ ₊₊₊ _ℎ₊ _∗₊ ∗ ℎ + ℎ ∗ + ∗ ∗ ℎ + (3.3) = + + + ℎ + ∗ + ∗ ℎ + ℎ ∗ + ∗ ∗ ℎ + (3.4) = + + + + ∗ + ∗ + ∗ + ∗ ∗ +

As shown in Section 2, there is evidence of volatility and abnormal returns on the equity indices around the elections. With the second hypothesis this study tests whether there is a pattern in the returns in any directions after the results are announced. The expected results do not change, in the sub hypothesis (3.1) the coefficient is expected to be positive and significant to prove a pattern in the direction of equity returns in the post elections window. The coefficient is predicted to be negative, as the EMU countries are expected to have smaller abnormal returns after the elections compared to the countries outside of the Eurozone.

The sub hypotheses (3.2), (3.3) and (3.4) test whether there are different results when dummy variables: Government Change, Early Elections and Right-wing party are added. As mentioned earlier, during the change of the government, there is expected either positive or negative pattern of the returns compared to incumbent government getting elected. Sub hypothesis (3.2) tests whether there is different return if the country is in the EMU. The coefficient is expected to be lower than the coefficient , as the EMU membership is expected to lower the returns due to government having less power over monetary and fiscal policies than outside of the EMU. Sub hypothesis (3.3) tests

(19)

19

whether there is an effect when the country is on euro currency and the party who wins the elections is right-wing party. And the sub hypothesis (2.4) tests if there is an effect of country being member of the EMU and holding the elections early, compared to holding the elections early and not being member of the EMU. The expected results of these subtests are the same as when testing for the volatility as dependent variable.

4. Data and Descriptive Statistics

4.1 Elections Data

The elections data were collected from various sources. As mentioned earlier, the elections dates and results were handpicked from the NSD European Elections database, and checked on various political websites, articles and publications. The summary of the Elections data can be found in the Table 1, Panel A in Appendix. The summary provides an insight into how many elections were in the sample in each country, in a total of 126 elections. One of the variables in the Table 2, Panel A captures the number of the elections held under euro currency. This number varies across the countries, because some of the countries joined the EMU at different times. The summary also gives insights into how many elections were held early in the specific country. Particularly in Greece, four out of the eight elections were held early.

The variable Government change shows that some governments do not stay in the power for longer than one period. In Slovenia, for every election observed, there was a change in the government. In France, Czech Republic, Poland and Romania the government remained the same only one time in the observed window. On the contrary, in Luxembourg the government remained the same during the observed window of four elections. For Cyprus, Slovakia and United Kingdom the study only observes two elections, thus the data might not provide accurate estimate. The last variable observed is the political orientation of the governing party. Most of the countries have right-wing governments. In Belgium, Estonia, Ireland, Luxembourg and Slovakia only right-wing parties were in power during the observed period.

(20)

20 4.2 Summary Statistics

A summary of the statistics can be found in the Table 2, and in the Table I in Appendix. Table 2 shows the means and standard deviations of the volatility of equity indices for the individual countries. The total average mean equals 1.3428 and the total average standard deviation equals 0.8725. The statistics are measured on 40 day window around the elections. The highest mean is observed in Malta, being 3.89. The standard deviation for Malta despite such a high mean is only 0.219. This statistics shows that the index in Malta has in general high volatility. Two other highly volatile countries in the measured window are: Cyprus with value being 3.217 and Romania with 2.91.

Country Mean St. Deviation Mean St. Deviation

Austria 1,1860 0,5058 0,0869 0,0962 Belgium 1,1015 0,2236 0,0207 0,0131 Bulgaria 1,7729 0,1235 0,0820 0,1248 Croatia 1,1068 0,2201 0,0505 0,0474 Cyprus 3,2175 0,1957 0,0368 0,0377 Czech Republic 1,1356 0,5991 0,0630 0,0377 Denmark 1,3220 0,6193 0,0321 0,0227 Estonia 1,2132 0,2154 0,0608 0,0676 Finland 1,1450 0,2682 0,0419 0,0333 France 1,4989 0,8796 0,0598 0,0551 Germany 1,2130 0,6401 0,0439 0,0604 Greece 1,0359 0,3074 0,0641 0,0739 Hungary Ireland 1,4257 0,9539 0,0369 0,0111 Italy 0,9464 0,2545 0,0292 0,0318 Latvia 1,1212 0,2562 0,0118 0,0102 Lithuania Luxembourg 1,0088 0,2235 0,0225 0,0192 Malta 3,8947 0,2194 0,0178 0,0156 Netherlands 1,2062 0,2385 0,0380 0,0321 Poland 1,4207 0,7615 0,1002 0,1059 Portugal 1,3803 0,8501 0,0276 0,0196 Romania 2,9076 1,5848 0,0747 0,0803 Slovakia 0,8812 0,5362 0,0093 0,0015 Slovenia 1,2695 0,3352 0,0596 0,0642 Spain 1,0787 0,1974 0,0468 0,0335 Sweden 1,0465 0,2109 0,0634 0,0570 United Kingdom 1,7360 1,3906 0,0543 0,0700 Average 1,3428 0,8725 0,0496 0,0599

Volatility of returns Absolute returns

The first column lists all EU member countries. Two dependend variables were tested for the mean and standard deviation, the volatility of returns and Absolute returns. The Volatility of returns variables represent the mean and standard deviation of the volatility of the daily returns in the sample for each individual country. The Absolute returns varaibles represend the mean and standard deviation of the pattern of Recall that the sample consists of two month window, one month before and one month after the elections.

Table 2

(21)

21

In Romania, the standard deviation equals 1.585, which could be interpreted as the volatility changes between the elections. The lowest volatility is observed in Slovakia, with value 0.88 and standard deviation 0.536. The estimation can be misleading since as mentioned earlier, only two elections were observed. In Italy, the volatility equals 0.946 and standard deviation 0.255. Both the variables are very low, with concluding that Italy might not experience volatility around the elections. It is worth to mention the high standard deviations of France, Ireland, Romania, Portugal and United Kingdom.

Since the observed window for the summary statistics is 40 days around elections equally distributed, the higher standard deviations predict the possibility of having higher volatility after or before the elections. The summary statistics for the absolute returns of the observed windows are shown in the Table 2. The total average mean is 0.0496 with standard deviation of 0.0599. Poland has the highest mean value, being 0.1002, with standard deviation being 0.106. Both values are much higher than the total average, predicting Poland to have significant patterns in direction of the equity index around the political elections. The lowest mean was observed in Slovakia, being 0.0093 with standard deviation 0.0015. Latvia, Italy and Luxembourg experienced both low means in volatility and in absolute returns. These observations could predict lower reactions of markets on the political elections. Overall, the standard deviations for both volatility and absolute returns are high, which strengthens hypotheses of this study that elections have an effect on the equity indices.

The descriptive statistics of the independent variables can be found in the Table I in Appendix. There were 57 elections that took place in the Eurozone, which is 45.23% of the total number. Recall that the sample of this study started in 1990, 9 years before the Eurozone, so the number is reasonable. Surprisingly, 31 early elections occurred in the sample, 24,6% of the total number. This could be explained with the governments running into issues or being not so popular among the citizens. It can also be linked to the government being changed 53.97% of the total sample.

Compared to Białkowski et al. (2008) who observe only 32.09% of the total sample change in the government, in this sample occurred more changes and uncertainty. Although, the early elections

(22)

22

in the study of Białkowski et al. (2008) occurred 41.79% of the time, significantly higher number of times than in the sample of this study. Cargill and Hutchison (1991) prove that some governments in Japan call the elections early if the economic conditions are plausible in order to be reelected again, which can explain the high number of early elections in the study of Białkowski et al. (2008), as it focuses on the countries around the world, not only Europe.

The last mean shows that 65.08% the elected government is politically oriented in the right. As mentioned in the Section 2.3, left wing governments experience higher volatility after the elections (Boutchkova et al., 2012). As showed by Rueda (2005), the left-wing parties incline more to labor-protective policies, which leads markets to react more volatile. The majority of the governments elected in Europe are right wing, showing that the citizens incline more to open-market policies.

Pairwise correlation test is also performed in order to see if the tested variables are not correlated. The variables that are above 0.700 are considered as highly correlated. The results are shown in the Table II in Appendix, and there is no reason for concern about validity of this study due to the correlation. Most of the variables are correlated less than 0.100, except the right-wing party and the government change variables, where correlation equals -0.1755 and the Eurozone with early elections variables, where correlation equals to 0.1102. The values are still very low.

5. Results

5.1 Main Results

In this section I present the results of this study. These tests have as a dependent variable the volatility around the elections window. The OLS regression was used together with the Difference-in-Difference analysis. Table 3 presents the main regression of the study, comparing the volatility of the countries outside of the EMU to the countries in the EMU around the election dates. As shown in the Table 3, the event windows are the same length around the election. To see whether the elections have an effect on the equity indices, I chose different lengths of the event window. The shortest window

(23)

23

tested was 20 business days around the elections. The longest window tested was 200 days around the elections (100;100). Overall, seven different windows were tested, as shown in Table 3.

The overall squared in regression (1) is only 1.4%, and in regression (7) the overall R-squared is 3.7%. These values represent low fit of the regression model. These results are similar to Białkowski et al. (2008), who have in their study low overall R-squared values, with values from 2 to 6%. As shown in Table 3, the overall R-squared improves when the observed window is longer. The coefficient post is not statistically significant. The coefficient sign for post variable is in most of the regressions opposite to the predictions. Only in the regressions (6) and (7) the coefficient is positive. The volatility does not increase after the elections as it was predicted. The expectations of this study were to have higher volatility after the elections. These findings are in line with Białkowski et al. (2008) who find the window of two weeks before elections and four weeks after elections with higher volatility, with the 4 weeks period after the elections having the highest volatility. However, Pantzalis et al. (2012) find that the positive abnormal returns are on their peak two weeks before the elections.

Dependent variable: Volatility (1) (2) (3) (4) (5) (6) (7)

Event window (-10;10) (-15;15) (-20;20) (-25;25) (-50;50) (-75;75) (-100;100) post -0.226 -0.23 -0.235 -0.19 -0.03 0.0448 0.00796 0.144 0.146 0.148 0.132 0.109 0.108 0.0989 eurozone -0.25 -0.265* -0.301* -0.299** -0.215* -0.231** -0.230** 0.152 0.154 0.155 0.139 0.115 0.114 0.104 eurozone_post 0.312 0.264 0.185 0.159 -0.0306 -0.0644 0.00891 0.215 0.217 0.22 0.197 0.162 0.161 0.147 Constant 1.374*** 1.426*** 1.556*** 1.525*** 1.457*** 1.448*** 1.455*** 0.102 0.103 0.104 0.0936 0.0773 0.0767 0.0699 Observations 252 252 252 252 252 252 252 R-squared 0.014 0.016 0.024 0.028 0.033 0.042 0.037 Table 3

This table shows the results of OLS panel regression diff in diff model testing whether in a selected window after the elections is the volatility of equity indices in EU higher compared to the pre-election window. Post dummy variable equals one during the specified number of days in the event window after election date, and zero during the specified number of days before the elestions. Eurozone dummy variable equals one if the country was in EMU and zero otherwise. T-statistics are shown below the coefficient value and *, ** and *** indicate the statistical significance at the 10%, 5% and 1%, respectively.

(24)

24

The coefficient Eurozone is statistically significant in all regressions except regression (1). For the event window of 50, 150 and 200 days around elections symmetrically distributed, the Eurozone coefficient is significant at 5% level. The coefficients for Eurozone dummy variable are in all regressions negative, in regression (3), the coefficient is -0.301. The volatility is therefore lower in the observed sample around the elections when the elections happened in the Eurozone country. These findings are in line with Furceri and Karras (2006), who find that EMU membership leads to better price stability.

The results show no significance for the post elections window and being in the Eurozone compared to not being member of the Eurozone. Based on these results, we cannot confirm the hypothesis (1) of this study. There is no significant difference in the volatility after the elections depending on the EMU membership, although it lowers the overall volatility around the elections. These results are not as expected, although the Eurozone membership has significantly lower volatility around the elections as predicted.

It leads to conclusion that the markets are more stable in countries of the EMU during the election times. Due to the governments having less power over fiscal and monetary policies that are mostly managed from the EU, the equity indices do not react as volatile in the tested windows around the elections. These findings are in line with Morana and Beltratti (2002), who find that after the euro introduction, the volatility decreased in the less stable countries. Also, Boutchkova et al (2012) find that there is no statistical significance between the volatility measured in the election year and not election year, without adding further socio-economic variables.

5.2 Early Elections

The next regressions tested the effects of adding socio-economic variables to the hypothesis, such as early election timing, change in the government after the elections, or whether the main political party was oriented to the right or left side. Table III in Appendix shows the results of adding the early elections dummy variable. The overall R-squared is higher for the event window (-10;10), being 2.4%. However for the event window (-100;100), the R-squared of 4.1% is lower than in the

(25)

25

Table 4. The Eurozone coefficient is again statistically significant, however only at 10% level and not in all regressions. The early elections do not have any statistically significant effect, in line with the previous research (Boutchkova et al., 2012; Pantzalis et al., 2000). As shown in Table III in Appendix, there is a significant evidence of higher volatility after the elections if the country is in the Eurozone on (-10;10) day window at 10% level. This result was not significant, when tested in the previous regression, and it is opposite of what expected (having lower volatility when being the EMU member).

In the same regression (1), the coefficient for EMU membership is negative and significant at 10%, which means that the volatility is lower overall for Eurozone countries, however is higher after the elections in this event window. These results are opposite to the results in Table 3. Since these results are only significant in the shortest event window, it could be caused by unordinary elections, such as multiple elections in Greece due to failing to form a government after the first elections on 7th

of May 2012. The second elections were held on 18th_{of June 2012. As suggested by Pastor and}

Veronesi (2012), the uncertainty created due to the political change has significant impact on the equity indices. The rest of the results were not significant, which leads to not confirming the sub hypothesis (2.1) that early elections create higher volatility after the elections.

5.3 Change in the Government

The next regressions are made with regards to the change in the government after the elections. The results are shown in Table IV in Appendix. The overall R-squared was within the same range as the previous results, being the lowest for the 15;15) event window, and highest for the (-75;75) window, where it equals 4.6%. The lowest value is for the regression (2), being 2.6%. The coefficient for government change has mostly negative sign, except the regression (1). Moreover, the coefficients are not statistically significant. The volatility around the elections is lower when there is anticipated government change. The sign of the coefficients that reflect the post election window with the government change is positive, except for the regression (1).

(26)

26

These results predict that the volatility is higher after the elections if there is a government change, although these results are as well not statistically significant. For the coefficients that are taking into account the Eurozone membership, the sign orientation is negative as predicted, although the results are not statistically significant at any level. The volatility is lower, when the country is member of the Eurozone and has a change in the government compared to not being a member of the Eurozone. Overall, this study does not find higher volatility if there is a change in the government, contrary to Białkowski et al. (2008), who find higher volatility when there is a government change.

5.4 The Government Orientation

The regressions in Table V in Appendix test whether the orientation of the parties that form the governments has an effect on the volatility of the equity indices. As stated in the Section 2.3, the previous research proves that volatility is higher when the left-wing government is elected (Boutchkova et al., 2012; Leblang and Mukherjee, 2004). The overall R-squared is in line with the previous tests. The value for the (-10;10) event window is higher than in the previous tests for this regression, being 3.1%. The lowest fit is for the (-20;20) event window, equaling 3%. The coefficient right wing has a positive sign across all regressions, but there is no statistical significance. The tested coefficient of the volatility after the elections in the country with the right wing government and being in the Eurozone has negative signs across all regressions. The volatility is lower when the country is in the Eurozone and has right wing government. These results are although not statistically significant. These results are in line with Pierdzioch and Döpke (2006), who did not find significant evidence when testing for Partisan effect during the elections in Germany.

5.5 Robustness Checks

The same regressions were performed in order to test whether there is a pattern in the direction after the elections on the equity markets in the EU. The regression results are provided in the Table 4 and Tables VI-VIII in Appendix. The overall R-squared does not improve and does not have the increasing pattern when the event window increases as shown in the main results. Table VI in Appendix shows the results of the hypothesis (3.1). The coefficient sign for post election event

(27)

27

window is negative during 40, 50, 150 and 200 days around the elections, otherwise is positive. The coefficient is not statistically significant.

In this test, the Eurozone coefficients are also negative across the different event windows as with the dependent variable volatility, however they are not statistically significant. The sign of the coefficient for the Eurozone and post election event window is again not consistent, and cannot provide any details about the tests. Overall, the coefficients are not statistically significant at any event windows. These results are again in line with Pierdzioch and Döpke (2006), who did not find partisan effect on German elections or with Boutchkova et al. (2012) who did not find difference in election year and not election year without adding extra variables.

Dependent variable: Return (1) (2) (3) (4) (5) (6) (7) (8) Event window (-5;5) (-10;10) (-15;15) (-20;20) (-25;25) (-50;50) (-75;75) (-100;100) post elections 0,00566 0,00482 -0,00171 -0,0109 -0,0131 -0,0154 -0,108 -0,0983 0,00431 0,00802 0,00911 0,0114 0,0125 0,0227 0,0753 0,0719 early 0,0025 -0,00317 0,00247 0,000145 0,00699 -0,00106 -0,0813 -0,0763 0,00676 0,0126 0,0143 0,0179 0,0196 0,0357 0,122 0,115 eurozone 7,85E-05 -0,00705 -0,00586 -0,0155 -0,0153 -0,016 -0,111 -0,122 0,00469 0,00874 0,00993 0,0124 0,0136 0,0248 0,0821 0,0772 eurozone_early -0,00461 0,00823 0,01 0,0069 -0,00438 -0,0149 0,0629 0,104 0,00941 0,0175 0,0199 0,0249 0,0273 0,0496 0,169 0,159 early_post -0,000228 0,0178 0,0201 0,0384 0.0465* 0.139*** 0.313* 0,168 0,00971 0,0181 0,0205 0,0257 0,0281 0,0512 0,172 0,168 eurozone_post -0,00619 -0,0065 -0,000546 0,0105 0,0166 0,0245 0,112 0,151 0,00664 0,0124 0,014 0,0176 0,0192 0,035 0,116 0,114 early_eurozone_post 0,00723 0,00427 0,00571 -0,0288 -0,0365 -0.132* -0,313 -0,22 0,0134 0,025 0,0284 0,0355 0,0389 0,0711 0,239 0,236 Constant 0.0205*** 0.0350*** 0.0443*** 0.0561*** 0.0638*** 0.100*** 0.225*** 0.230*** 0,00305 0,00567 0,00644 0,00807 0,00883 0,0161 0,0532 0,0501 Observations 250 250 250 250 250 250 248 248 R-squared 0,017 0,038 0,045 0,03 0,038 0,065 0,028 0,015

This table shows the results of OLS panel regression diff in diff model testing whether in a selected window after the elections is the absolute value of the returns of equity indices in EU higher compared to the pre-election window. The absolute value is taken in order to observe the trend, not the direction of the trend. Post dummy variable equals one during the specified number of days in the event window after election date, and zero during the specified number of days before the elestions. Eurozone dummy variable equals one if the country was in EMU and zero otherwise. Early dummy variable equals one if the elections were held early and zero otherwise. T-statistics are shown below the coefficient value and *, ** and *** indicate the statistical significance at the 10%, 5% and 1%, respectively.

The price return trend on selected windows around elections during early elections Table 4

(28)

28

When adding the independent variable early elections, the overall R-squared improves by a small percentage, as shown in the Table 4. The early post coefficient for regression (6) equals 0.139 and it is statistically significant at 1% level. Also, the early post coefficient for regression (5) equals 0.0465, and for regression (6) equals 0.313 and is statistically significant at 10% level. The results show that there are abnormal returns after the elections on 50, 100 and 150 business days window symmetrically distributed around the elections.

These results are as proposed by Pantzalis et al. (2000) that abnormal returns are higher during times of uncertainty, however their results associated with early elections were not statistically significant. Pastor and Veronesi (2012) also found higher abnormal returns during the times of uncertainty, as discussed in the Section 2.2. If the elections are held early, in most of the cases the government could not agree on continuing working together, or other issues appeared and the political future was uncertain. The coefficient for early post elections window when the country is member of the EMU for (-50;50) event window equals -0.132 and is statistically significant at 10% level. However, for the rest of the event windows there is no significant evidence that the membership of the EMU lowers the abnormal returns when the elections were held early.

The results of the effect of the abnormal returns when there is a change in the government after the elections are shown in the Table VII in Appendix. The signs of the coefficients when there is a government change on the post election window are both positive and negative, and not statistically significant. On the (-15;15) business days window, the post election coefficient when there is a government change and the country is a member of the EMU equals 0.0431 and is statistically significant at 10% level. The coefficient is very low, and predicts that there is higher abnormal return if the country is in the EMU and has government change. This result is contrary to the predictions of this study, that the EMU creates lower abnormal returns compared to countries that are not the members. However, these findings are in line with Pantzalis et al. (2000) who prove that the response of the market is greater when there is a change in the government. Moreover, the coefficient is not statistically significant for the rest of the event windows.

(29)

29

The orientation of the government after the elections was tested and the results are shown in the Table VIII in Appendix. The overall R-squared is lower than in the previous test. The coefficient for the post election window in regression (1) equals 0.0145 and is statistically significant at 5% level. One week after the elections, the abnormal return is higher compared to the same time window before the election. This result confirms the predictions of this study that elections lead to abnormal returns of the equity indices in the EU. The coefficient is significant only in the (-5;5) event window, which leads to conclusion that the elections create short-term abnormal return and then the markets calm down.

These findings are similar to by Białkowski et al. (2008) who find the peak of the volatility within the first week after the elections. The coefficient right-wing post in the regression (1) equals to -0.0147 and is statistically significant at 10% level. This result proves that the elections create a shock on the equity indices in a short time window, as the coefficient is not significant in the longer event windows. The results show that the right-wing government creates lower abnormal results than left-wing government. Recall, that Rueda (2005) found that left-left-wing parties tend to protect the labor via regulations and policies more than the right-wing parties.

The higher absolute abnormal returns of the left-wing governments are in line with this research, as the left government might apply tighter regulations in favor of labor market, and the uncertainty is reflected on the equity indices reactions. The coefficient for the right-wing party being elected in the EMU during the post elections window is not statistically significant and does not have same sign over the various event windows. For the shortest event window, the sign of this coefficient is positive, which is not in line with the predicted findings of this study. To conclude, the membership of the EMU does not have significant effect on the abnormal returns after the elections for the studied sample.

5.6 Limitations of the Study

There are several limitations of this study. The main concern is the length of the event windows. Although the post election event window was computed as different studies suggested,

(30)

30

some studies created as a pre-event window longer periods of time. For example, Białkowski et al. (2008) used similar approach to the one used in this study for creating the event windows, Pantzalis et al. (2000) used 100-week estimation period. By creating such a long pre-event window the study incorporates the possibility of the equity indices reactions in the pre-election periods. Boutchkova et al. (2012) compared the volatility around the elections to the volatility in the non-election years in order to see the effect of elections on the equity indices. The study did not find significant evidence without adding extra socio-economic variables but is an alternative in order to improve the methodology of this study.

Another limitation could be using different indices that most of the research. Due to possibility to download the indices from the Bloomberg terminal, there was a difference compared to the studies that mostly used the Morgan Stanley Capital International (MSCI) weekly data on value-weighted indices (Białkowski et al., 2008; Pantzalis et al., 2000). The decision to use indices from different sources was driven by the availability of the indices. Only 15 EU countries were found in the MSCI database, which would not make the sample large enough. Also, the start date of the sample varied across the countries. Overall, the sample was large enough, however the same starting date could improve the results of this study. For example, some countries had only two observed elections during the sample, which could be improved in the further research.

The event window chosen was similar to Białkowski et al. (2008), who looked at the volatility after the elections. The other studied window could be before the elections, when the volatility rises. This approach was taken by Pantzalis et al. (2000), who studied 2-weeks before the elections, also Li and Born (2006) used the pre-elections window. The window chosen was proven to have significant result, although the different approach might broaden the studied topic how the EMU membership affects the stock markets around the elections.

6. Conclusion

This study examines the effect of the Eurozone membership during the elections on the equity indices in the European Union. The sample started in 1990 and the data was collected from the