Macroeconomic Factors and Stock Returns: Evidence from Emerging Markets accepted in the European Union

Macroeconomic Factors and Stock Returns: Evidence from

Emerging Markets accepted in the European Union

University of Groningen

Faculty of Economics and Business

MSc Finance

Master’s Thesis

June 2019

Author: Ivan Ivanov

Student number: 3068501

Supervisor: Dr. J. V. Tinang Nzesseu

Abstract

2

1. Introduction

Financial markets are the cornerstone of the current capitalistic era as they ensure capital gathering and liquidity for businesses and investors. Buyers and sellers meet on financial markets to trade assets like stocks, bonds, currencies and different types of derivatives. The stock market specifically, allows investors to own shares of public corporations and therefore to participate in their management. They can get return on their investment by receiving dividends or by the appreciation of share prices. Because the stock market can be a suitable investment vehicle for investors with risk appetites ranging from, low-risk-low return, to high-risk-high return, it has been widely researched throughout the years, and various factors have been employed to study stock returns. This study’s objective is to contribute to this research, by examining the effects that macroeconomic factors have on Emerging market stock returns, for countries before and after they join a political and economic union (in this case the European Union). By doing the study in such manner this paper will not only try to acknowledge any effect macroeconomic factors may have on stock returns, but will also investigate if such an effect changes when an integration process such as joining the European Union (EU) occurs.

3

activity and stock returns (Fama, 1981; 1990). Inflation measures the increase in prices of goods and services. Higher inflation causes input prices to become higher and reduces the purchasing power of consumers. These are reasons to believe that it has negative relation with stock returns as documented by some researches (Bodie ,1976; Fama, 1981). Interest rates are seen as another important determinant of stock prices. Higher interest rates increase the borrowing cost for companies, which translates into lower earnings. Therefore, interest rates must have negative relation with stock returns, as researched by some studies (Fama, 1981; Campbell, 1987).

This paper analyses the effect macroeconomic factors have on stock returns for countries that join a political and economic union, namely the European Union (EU), and if this effect changes after joining. The study is conducted for countries with Emerging equity markets (Stock markets in transition). The selected countries are Bulgaria, Romania and Croatia since they are the last ones to have been accepted and have relatively new stock exchanges. They are classified as Emerging markets and developing economies by the International Monetary Fund (World Economic Outlook, 2015). By examining if economic factors have different power of influence on stock markets after countries join the EU, compared to before, this paper might become a good guideline about how financial markets of the new future members with economies and stock markets similar to the ones researched will behave once they join the union. This can turn out to be vital for investors as well as for governments since the stock market is an important part of the economy.

The enlargement of the EU is a key political and economic process that leads to collective identity and policies for countries (Schimmelfenning and Sedelmeier, 2005). The EU has been rapidly expanding (10 members joined at the same time on May 1, 2004) in the last two decades with more and more countries joining with the purpose of European integration. The goal is to achieve high degree of collaboration between the members, and to create better environment for trade, investment and employment. The member countries are free to trade with each other with no additional costs. Movement is free and open to all citizens. Some of the countries also use the euro which reduces the currency risk in the union. There are currently 28 countries in the EU which makes it one of the biggest economies in the world. According to Eurostat (Eurostat TGM) the estimated GDP of the EU for 2018 was $18.8 trillion, which represented 22% of global economy. EU’s official currency, the euro, is the second largest reserve currency in the world after the US dollar.

This economic and political union allows for capital to flow freely where it can generate the highest possible return. This might lead to more economic growth for the countries and development of their financial markets, especially for the ones that have recently joined. It also means that this potential growth will rely on risk sharing benefits between countries and that there will be less opportunities for diversification (Pagano, 1993).

4

an event. This study measures the effect that the following macroeconomic factors: economic activity, inflation, interest rates have on stock returns and then compares this effect for the two periods: before and after joining the EU. The major stock indices of Bulgaria, Romania and Croatia (SOFIX, BET and CROBEX, respectively) are used as a measure of stock performance. The decision to use the main indices lies with the fact that Emerging markets are usually characterized by having low liquidity and thin trading (Antoniou et al. 1997). Therefore, indices that have high listing standards and contain the most traded companies can address this issue. Additionally, STOXX Europe 600 Index is used as a market factor. The macroeconomic variables cannot be expected to capture all the information available to markets. Because of the smoothing and averaging characteristics of most macroeconomic time series, market indices are added as an additional factor (Chen et al., 1986). Since most of Emerging markets can be qualified as being between the extremes of integration and segmentation, it is reasonable to include both global and local risk factors (Pajuste et al., 2000; Fifield, Power and Sinclair, 2002). More information about the different country indices and the market index, such as listing standards and maximum stock weights, can be found in Appendix A.

It should be noted here that the three countries do not join at the same time. Bulgaria and Romania joined on 1 January, 2007 to finish the fifth enlargement (that started in 2004) of the EU. Croatia joined the EU later, on 1 July, 2013.

The research question of the study is: Does the effect of economic forces in European Emerging stock markets change after countries join the EU? That effect will be measured by using the local stock index for each country and explaining the returns using the local macroeconomic variables and the market factor. As will be discussed in the literature review, there is an effect of financial integration after a country joins the EU. That means, that there are potentially new forces added to the pool of factors affecting stock markets. In that case, it also means that the effect of previous factors is likely to change. If macroeconomic variables can be significant in explaining stock returns, then this raises the question if their power changes after a country joins the EU because of the integration. Therefore, the following hypotheses are defined:

H1: The economic factors are significant in explaining stock returns.

H2: The explanatory power of the economic variables changes after a country with Emerging stock market is accepted in the EU.

The same logic and expectations can be applied to the market factor. As the indices of Croatia, Bulgaria and Romania can be considered to be part of a large portfolio set of European stocks, a market factor based on developed countries in Europe (as STOXX Europe 600 is), will likely have different explanatory power for markets of countries, before and after they join the EU. Because of that this study defines two more hypotheses:

H3: The market factor is significant in explaining stock returns.

5

The rest of the paper is organized as follows. The next section is the theoretical framework for this study. It starts by examining the financial integration phenomenon and the different types of market development. Then it presents relevant literature with regards to the ability of the macroeconomic variables to explain stock returns. Finally, it shows some alternative uses of macroeconomic factors and provides a summary. Section 3 gives the methodology review and the regression models used for the study. Then it describes the dataset. Section 4 presents the results of the study. It first shows the descriptive and correlation statistics of the variables and then gives the results of the study. Finally, it further discusses and summarizes the main results. Section 5 concludes. Section 6 presents the bibliography used for the paper. Section 7 shows the appendices.

2. Literature review

2.1 Financial integration and the European Union

Financial integration can be defined as a process that defines the degree to which a certain economy does not restrict cross-border transactions, facilitates risk sharing, enhances economic growth and functioning of its financial systems (Obstfeld, 1994; Levine, 2001; Edison et al., 2002). It has been an increasing process since financial markets have become an important component in local and global economic systems. Although the effect of international financial integration must be interpreted cautiously it is positively associated with important indicators such as: GDP, education, banking sector development, law and government integrity (Edison et al., 2002). The integration helps financial markets to develop and become more and more efficient as they gain the ability to attract huge capitals (both domestic and foreign) and become an effective intermediary that provides diversification and potential for returns. Although international financial integration is sometimes limited due to different country characteristics it has been increasing substantially throughout time even during periods of market distress (Claessens and Schmukler, 2007; Tong, Chen and Buckle, 2018).

6

EU membership (not Euro adoption), increases financial and economic integration between European countries. This is an important finding because it suggests that EU members are able to integrate their economies and financial markets even without being part of the Economic and Monetary Union (EMU).

Being part of the EU, entities like governments, banks and corporations operate and form their policies and strategies on local as well as on international level. The same applies for the focus of investors. As the EU is expanding, so are the potential opportunities for investors to seek higher returns. If a country wants to join the union, it needs to fulfill certain economic and political conditions. Once it joins, its economy and markets are connected with the EU’s. This leads to a phenomena where the integration between the new EU countries and the old ones increases over time (Cappielo et al., 2006). Christiansen and Ranaldo (2008) is a study that measures financial market integration by observing how often extreme returns on different markets occur simultaneously. Using coexceedance methodology and multivariate logit model the paper finds that new EU markets have become more integrated with the old ones after joining the EU and therefore are more exposed to adverse comovements and volatility. Another study conducted for the period 2001 – 2007 (Aslanidis and Savva, 2008) shows substantially increasing correlations between Polish and Czech stock markets with those of the Euro-zone. They conclude that this increase is not a worldwide phenomenon but is rather European market specific. The study also shows that the high correlation of the Hungarian market remained stable and that there are low correlations for Slovenia and Slovakia, which shows that the effect can vary from country to country.

2.2 Market development and stock market factors.

It is obvious but important to notice that countries are different in size and development. And so are their respective financial markets. A developed and global market would be operating in an advanced economy that is characterized by high income, capital flows and efficiency of its institutions. The biggest of companies are listed on such markets and there is a lot of activity taking place on them. Alternatively, there are Emerging markets, which are developing markets. They are attributed to rapidly growing economies that are less developed. They usually have higher market risk which is sometimes seen as a reason for higher potential returns. Even though the described countries and their markets are naturally different there is a degree of integration documented between Emerging and developed stock markets (Nasser and Hajilee, 2015). The mentioned paper examines 5 Emerging stock markets (Brazil, China, Mexico, Russia and Turkey) and developed markets (US, UK, Germany).

7

news, performance etc.), or can be oriented for the exact industry the studied companies occupy. Factors like firm’s size, value, profitability and investment patterns have been found to be able to explain stock returns (Fama and French, 1993, 2015). They can also be ones that measure investor sentiment (Baker and Wurgler, 2006) or they can be economic factors that measure the overall wellbeing of the economy as shown by Chen et al., (1986) that used the economic factors and a market index as proxies for the state variables in the Arbitrage Pricing Model (APM), which is a multi-factor asset pricing model that is used to predict returns using the linear relationship between expected returns and variables that capture systematic risk (Ross, 1976). As financial markets are highly connected with their economies, it has been of great interest how economic factors affect them. Interest rates, inflation, human capital, economic activity and policies are some of the economic factors, that are commonly believed to be related to financial markets as they are vital for the business environment in any country.

2.3 Developed markets and macroeconomic factors

Most of the studies that try to explain stock returns with economy-wide factors have been focusing on more developed markets. Chen et al., (1986) explores various economic factors as systematic influences on stock market returns in the US. The paper employs 5 macroeconomic variables [industrial production, inflation (CPI), risk premium, term structure, oil prices] and a market index as proxies for the state variables in the Arbitrage Pricing Model (APM). The authors chose these factors with the assumption that they influence either future cash flows or the risk-adjusted discount rate, which are the two key variables to look at when pricing stocks by the expected present value of future cash flows. Using size based portfolios of stocks, the paper finds that industrial production, unanticipated change of the risk premium, unanticipated inflation and the unanticipated change in the term structure are the most important factors affecting expected stock returns. Also the change in expected inflation and unanticipated inflation proved to be highly significant in a period of highly volatile inflation. Additionally, oil price is found not to be separately rewarded in the stock market. Since the finding of this study, that macro variables have a systematic effect on stock returns, there has been a big interest in the subject because of the implication that APM has higher predicting power than the CAPM, which is widely used in finance to price securities using the risk of the assets and cost of capital.

8

open to criticism. They carry a similar set of tests with UK data and find that variables similar to those of Chen et al., (1986) do not affect UK stocks in the same manner as described. The authors suggest that there could be other macroeconomic variables at play as well as the methodology might be inefficient to capture the pricing relationship. In the discussion part they raise the question about the size based portfolio returns being explained by the macroeconomic factors as suggested by Chen et al. and the fact that if the size of the firm is important feature influenced by the macroeconomic factors in the US market, it does not necessarily mean that this is applicable in other areas of the World since legislation, taxes etc. are usually different from country to country.

If macroeconomic factors are to be significant in explaining stock returns, there should be sufficient results coming from various countries with different economies that possess different entities, have different activities, politics and organization. Cheng (1996) is a study that attempts to provide that by specifying the economic variables that best correspond to stock market factors in the US and UK. The study suggests that the multiple regression analysis that Chen et al., (1986) used is sensitive to the number of independent variables included and that there is multicollinearity among the economic variables. The paper uses canonical correlation analysis as a procedure to explain the relation between the factor scores of security returns and the factor scores of economic indicators. It uses a number of factors that represent the following categories: stock market, money supply, industrial production, labour market and international trade. The results show that UK and US returns are affected by a combination of UK and US economic factors. Another notable study is Asprem (1988). It uses major stock indices and studies their relation with macroeconomic variables using quarterly data. Due to data availability the main countries that are covered are: France, Germany, Italy, Switzerland and UK. The paper uses five economic activity measures: changes in industrial production, real gross national product, gross capital formation, employment and exports. Regressions are done by the Ordinary Least Squares (OLS) method. The findings show that employment, imports, inflation and interest rates are negatively correlated to stock prices. It is discussed that changes in imports might be regarded as proxy for changes in consumption so the relation presented supports the Consumption Capital Asset Pricing Model. Changes in the stock indices show positive correlation with future industrial production and exports. The yield curve in the US also has positive correlation when regressed to the changes in stock indices. The results in the paper are not equally strong for all countries tested as some individual countries show results contrary the expected. The strongest connection between market returns and macroeconomic variables was found in France, Germany, the Netherlands, Switzerland and the UK. Additionally, most of the countries in the study show positive correlation between the S&P400 (S&P’s industrial index) and the stock indices. A constructed portfolio of European indices showed even stronger correlation with country indices.

9

know under what circumstances certain factors are affecting stock returns. Flannery and Protopapadakis (2002) starts by discussing the controversial findings of the relevant literature on the subject of macroeconomic events influencing stock returns. They discuss the successfully documented negative relation of stock returns to inflation and to money growth (Bodie ,1976; Fama, 1981; Geske and Roll ,1983; Pearce and Roley ,1983) but they also say that the impact of real macro variables on stock returns has shown to be unconvincing and hard to establish [(Chan, Karceski and Lakonishok ,1998) deny the relevance of macroeconomic factors]. The factors they consider (macro announcement series) either affect returns or increase the volatility of the market in the US. The paper uses 17 macro series announcements over the period 1980-1996 and finds that two inflation variables (CPI and PPI) affect the market portfolio’s returns, three real factors affect only the returns’ conditional volatility (Balance of Trade, Employment/Unemployment, Housing Starts). Monetary Aggregate affects returns and volatility. Interestingly the paper finds that the macro announcements of two of the popular macroeconomic variables (Real GNP and Industrial Production) are associated with lower return volatility.

Human capital is an interesting measure that addresses the skill, capacity and attributes of labor and explains how it interacts with the economy. It is a good forecaster of future stock returns at short and intermediate horizons and can be proxied by the growth rate per capita labor income and consumption-aggregate wealth ratio (Campbell, 1993; Jagannathan and Wang, 1996; Lettau and Ludvigson, 2001).

2.4 Emerging markets and macroeconomic factors

As discussed earlier, developed and Emerging markets are different in nature. The term Emerging market can unify both very big and very small in size countries. It is more the level of progress that is the determinant. A key feature of an Emerging market economy is the increase in local and foreign investment. Because Emerging markets generally offer higher potential returns their importance increases, and as they are usually very volatile it is important to know how they relate to developed markets (Harvey, 1995). The mentioned paper engages a discussion about the correlations Emerging equity markets have with developed ones and explores the reasons of the high expected returns in Emerging markets. Harvey uses data on more than 800 equities from Latin American, Asian, European and African markets. He finds that Emerging markets are more predictable compared to developed markets and that local information variables have more power (compared to global ones) in predicting Emerging market returns. He found that local factors accounted for more than half of the predictable variance of Emerging market returns.

10

most relevant domestic and worldwide economic factors. Then a regression analysis is conducted to explain index returns. Following the previous literature, the study selects inflation (CPI), foreign exchange rates, short-term interest rates, GDP, money supply and trade balance for domestic variables. The world market return, world inflation, commodity prices, world industrial production, oil prices and US interest rates were used as proxies for worldwide economic factors. The findings of the paper show that GDP, inflation, the money supply, interest rates, world industrial production and world inflation are able to explain returns from ESMs. Similar to other studies the explanatory power of those factors varies from country to country and in some countries they do not have any effect. To check how macroeconomic variables, affect stock returns in Emerging markets Mateev and Videv (2008) follows the Chen et al. approach using a sample of stocks traded on the Bulgarian stock exchange. For the period 1995 – 2005 they find that macroeconomic variables like trade deficit (trade balance), inflation (CPI) and country risk premium play significant role in explaining stock movements in Emerging markets.

2.5 Alternative uses of macroeconomic variables

Macroeconomic variables have been also used in studies that take different approaches compared to the discussed literature until now. For example, Chen (2009) uses interest rates, inflation rates, activity, unemployment etc. to test if recessions in the US stock market can be predicted by macroeconomic variables. It finds the yield curve spreads and inflation rates as the most useful predictors. Birz and Lott (2011) see newspaper stories as a measure of macroeconomic news since they are related to the sentiment of investors. The results show that news about GDP and unemployment affect stock returns. Macroeconomic factors have been also used to explain risk premiums in bond markets. Joslin, Priebsch and Singleton (2014) find that shocks to unspanned real economic activity and inflation have large effects on term premiums in the US market.

2.6 Summary

Table 1 summarizes the types of macroeconomic variables used throughout the studies described, and defines the potential proxies that can measure them.

Table 1

11 Economic activity

 Industrial Production

 Gross Domestic Product (GDP)

 Employment/Unemployment

 Trade balance

 Imports

 Exports Human capital

 Per capita labor income

 Consumption Interest rates

 Short-term interest rates

 Long-term interest rates

 Corporate and Government bond spread Inflation

 Consumer Price Index (CPI)

 Country credit rating

 Money supply

 Reserve money

Chen et al., (1986), Fifield, Power and Sinclair (2002), Asprem (1988), Mateev and Videv (2008)

Campbell (1993), Jagannathan and Wang (1996), Lettau and Ludvigson (2001)

Chen et al., (1986), Fifield, Power and Sinclair (2002), Asprem (1988), Mateev and Videv (2008)

Chen et al., (1986), Fifield, Power and Sinclair (2002), Asprem (1988), Mateev and Videv (2008), Flannery and Protopapadakis (2002)

Notes: Economic activity is measured by factors connected with the overall business climate and future earnings of companies. Human capital is proxied by human capital risk factors. Interest rates measured by interest rate risk and default risk. For inflation factors that influence business activity and the business climate are selected.

3. Methodology and data

3.1 Methodology review

12

question of this study as previous researches will give an accurate guideline for specific factors selection. Similar to Asprem, (1988) the Ordinary Least Squares (OLS) method is used in this paper. It is a statistical method that estimates the relationship between dependent and independent variables by minimizing the sum of the squares in the difference between observed and predicted values of the dependent variable.

As it is widely discussed that stock returns can be summarized by expected return part which is already assimilated by investors, and unexpected return which occurs due to uncertainty and unexpected events in the future, it is assumed by the APT that economic forces are accountable for a large part of the unexpected events. The APT has the following assumptions (Ross, 1976): it is a multi-factor asset pricing model relying on the fact that returns can be predicted and explained using linear relationship between expected returns and macroeconomic factors, there are enough securities to sufficiently diversify away unsystematic risk and well-functioning markets do not allow for the persistence of arbitrage opportunities. A basic form of the APT equation can be presented as follows:

𝑅 = 𝛼 + 𝛽₁𝑋₁ + 𝛽₂𝑋₂+ ⋯ + 𝛽_𝑖𝑋_𝑖+ 𝜀 (1)

Where:

𝑅 is the expected return on a security 𝛽 is the response to systematic risk 𝑋 is systematic risk

𝜀 is unsystematic risk

Macroeconomic factors and sometimes market factors are used as inputs in the APT.

3.2 Regression models

The economic variables for this study are the relevant proxies for economic activity, inflation and interest rates based on the discussed literature and availability, plus a market factor. Log returns will be used to transform the data where appropriate.

The regression can be written as:

𝑅_𝑡 = 𝛼 + 𝛽₁𝑒𝑐𝑜_𝑡+ 𝛽₂𝑖𝑛𝑓_𝑡+ 𝛽₃𝑖𝑛𝑡_𝑡+ 𝛽₄𝑚𝑘𝑓_𝑡+ 𝜀_𝑡 (2)

Where:

𝑅 represents the return of the separate country indices. The main country indices are chosen for the study due to the fact that they contain the biggest and most liquid companies from various sectors which makes them a good proxy for the overall economy’s health.

13

availability) and after conducting heteroscedasticity and autocorrelation tests, highly correlated variables are removed from the model. The inputs are matched on a monthly basis at time 𝑡.

The regression is performed twice for each separate country for the periods before and after joining the EU.

Additionally, a regression is done using dummy explanatory variable (D), which will take the value of 0 and the value of 1, respectively before and after a country joins the EU. In this model the cross-terms of the dummy variable and risk factors/macroeconomic variables capture the change in the explanatory power of the variables due to a specific event. By doing that the model measures how stock returns are explained by the same variables after EU accession. The regression equation has the following form:

𝑅_𝑡 = 𝛼 + 𝛼𝐷 + 𝛽₁𝑒𝑐𝑜_𝑡+ 𝛽₁𝑒𝑐𝑜_𝑡𝐷 + 𝛽₂𝑖𝑛𝑓_𝑡+ 𝛽₂𝑖𝑛𝑓_𝑡𝐷 + 𝛽₃𝑖𝑛𝑡_𝑡+

𝛽₃𝑖𝑛𝑡_𝑡𝐷 + 𝛽₄𝑚𝑘𝑓_𝑡+ 𝛽₄𝑚𝑘𝑓_𝑡𝐷 + 𝜀_𝑡 (3)

Where in addition to the variables from equation (2) we add the interaction variables:

𝛼𝐷, 𝑒𝑐𝑜_𝑡𝐷, 𝑖𝑛𝑓_𝑡𝐷, 𝑖𝑛𝑡_𝑡𝐷, 𝑚𝑘𝑓_𝑡𝐷, which begin being measured at the point in time, when each separate country have joined the EU.

Finally, a panel regression (it has individual dimension 𝑖, and time dimension 𝑡) with fixed effects is performed for Croatia, Bulgaria and Romania. An additional regression is done for Bulgaria and Croatia only, because they joined at the same time and a wider period of time can be covered. The purpose of this approach is to correct for possible endogeneity problem that occurs when there is omitted variable correlated with some of the regressors and the dependent variable, there is one or more of the independent variables that is determined simultaneously in a system or there is a measurement error. The error term (𝜀) in equation 2 can be rewritten as:

𝜀_𝑖,𝑡 = 𝜇_𝑖+ 𝑣_𝑖,𝑡 Where:

𝜇𝑖 is a time-invariant individual effect

𝑣_𝑖,𝑡 is the remainder disturbance

Therefore, equation 2 can be written as a fix-effect model:

𝑅_𝑖,𝑡 = 𝛼 + 𝛽₁𝑒𝑐𝑜_𝑖,𝑡+ 𝛽₂𝑖𝑛𝑓_𝑖,𝑡 + 𝛽₃𝑖𝑛𝑡_𝑖,𝑡+ 𝛽₄𝑚𝑘𝑓_𝑖,𝑡+ 𝜇_𝑖 + 𝑣_𝑖,𝑡 (4)

3.3 Data

14

picked to represent Economic activity due to being frequently used, also relevant for imports and exports (Asprem, 1988; Fifield, Power and Sinclair, 2002; Mateev and Videv, 2008) and available on a monthly basis. Arguably the most used proxy, Industrial production, used in Chen et al., (1986), is disregarded for this study due to insufficient data for Romania. TB is also preferred over Employment/Unemployment since the latter was not found to be significant in explaining stock returns in Emerging markets (Fifield, Power and Sinclair ,2002; Mattev and Videv, 2008). Finally, the Gross Domestic Product is only available on a quarterly basis, which is insufficient for this paper. Although an interesting measure, Human capital was not included in the model since it is best used to describe stock returns in shorter periods of time (Campbell ,1993; Jagannathan and Wang, 1996; Lettau and Ludvigson, 2001). Due to availability and to the fact that investors’ opportunity cost in stock markets is represented by short-term interest rate (Asprem, 1988), the National banks’ Monetary policy rates (BR) for the three countries is obtained and used as a proxy for interest rates. Finally, the Consumer Price Index (CPI), being one of the most common measures of inflation (Chen et al., 1986, Mateev and Videv, 2008; Flannery and Protopapadakis, 2002) is retrieved as a relevant proxy for this study.

To conduct this study, the above mentioned proxies, the country indices and the market index are obtained and matched together on a monthly interval, using Datastream. It is important to make a comparison between Price Return Index (PRI) and Total Return Index (TRI). The first one captures only the capital gain or loss and not the dividend received for securities. The second one captures both. For an identical set of securities, the return of a TRI will always be greater, because of the additional dividend payouts. This study uses the PRI, because it was available to cover the whole period of the research. The TRIs for the markets used, were not calculated until 2012. First, country indices are obtained beginning from their initiation and availability dates and then converted into first differences. CROBEX, the Croatian Zagreb Stock Exchange index is obtained for the period 2/1/1997 – 01/01/2019 with a total of 263 observations, SOFIX the stock index of Bulgarian Stock Exchange for the period 11/1/2000 – 01/01/2019, 218 observations and BET the Romanian Bucharest Stock Exchange index is retrieved for the 11/1/1997 – 01/01/2019 interval with 257 observations. The first differences of natural logarithms are used:

𝑅_𝑡= 𝑙𝑜𝑔_𝑒𝐼𝑁𝐷𝐸𝑋(𝑡) − 𝑙𝑜𝑔_𝑒𝐼𝑁𝐷𝐸𝑋(𝑡 − 1)

Next, TB is retrieved from Datastream. For Croatia and Bulgaria, the statistic is available for the period 1/15/2000 – 12/15/2018 (228 observations) and 1/15/2000 – 1/15/2019 (229 observations). Trade balance of Romania is obtained for the period 1/15/1999 – 1/15/2019 (241 observations). TB is included in the model as the change in country’s trade balance, expressed as a difference between imports and exports:

15

CPI is obtained as the proxy for inflation used as the monthly first difference in the logarithm of the index. Croatia CPI is available for 1/15/1998 – 2/15/2019 (254 observations), Bulgaria CPI for 1/15/1999 – 2/15/2019 (242 observations) and Romania CPI for 11/15/1990 – 2/15/2019 (340 observations). It is defined as follows:

𝑖𝑛𝑓_𝑡 = 𝑙𝑜𝑔_𝑒𝐶𝑃𝐼(𝑡) − 𝑙𝑜𝑔_𝑒𝐶𝑃𝐼(𝑡 − 1)

Finally, the change in National banks’ Monetary policy rates (BR) is included in the model to represent interest rates. Croatia’s bank rate is obtained for the period 1/31/1994 – 2/28/2019 with a total of 302 observations. Bulgaria’s and Romania’s bank rates are obtained respectively for 1/31/1999 – 2/28/2019 and 1/15/1996 – 2/15/2019 for 242 and 278 observations. The statistic is defined as:

𝑖𝑛𝑡_𝑡 = 𝐵𝑅(𝑡) − 𝐵𝑅(𝑡 − 1)

Monthly change in STOXX Europe 600 is used as a final explanatory variable for the model. The market factor (mkf) is available to cover the periods of all other variables and it is defined for this study as:

𝑚𝑘𝑓𝑡 = 𝑙𝑜𝑔𝑒𝑆𝑇𝑂𝑋𝑋(𝑡) − 𝑙𝑜𝑔𝑒𝑆𝑇𝑂𝑋𝑋(𝑡 − 1)

4. Results

The next sub-section presents the descriptive statistics results for the variables and indices for the whole period of the study. In the second sub-section the correlation between the different macroeconomic variables and indices is presented. Thereafter, in the third sub-section the regression results of the study are given.

4.1 Descriptive statistics

Table 2 presents the descriptive statistics for the change of macroeconomic variables (TB, CPI and BR), the change of the country indices (CROBEX, SOFIX, BET) and the market index (STOXX) for the whole period of the study (01/2000 – 01/2019).

Table 2

Descriptive statistics of macroeconomic variables and indices

16 Croatia TB CPI BR CROBEX Bulgaria TB CPI BR SOFIX Romania TB CPI BR BET STOXX -0.03434 0.000737 -0.04585 0.00208 -0.0761 -0.0152 -0.0193 0.00326 -0.0950 -0.00003 -0.14192 0.0051 -0.000041 -0.01144 0.000868 0 0.00132 -0.00276 0.001113 0 0.00072 -0.03354 -0.000043 0 0.00634 0.00233 0.5464 0.00703 1.5 0.15421 31.504 0.0132 1.24 0.12416 0.7638 0.0126 1 0.15733 0.04858 -1.1236 -0.00473 -2.75 -0.17178 -7.495 -3.8435 -1.39 -0.22117 -4.365 -0.01504 -4.94 -0.16415 -0.06752 0.2369 0.00218 0.362094 0.03151 2.5426 0.253005 0.244669 0.03555 0.507622 0.003161 0.551618 0.0378 0.01998 Notes: This table presents the descriptive statistics of the macroeconomic variables on a monthly basis. Using the change in Trade balance (TB), Consumer Price Index (CPI), Bank rate (BR), the statistics are shown for the three countries as indicated in the table. Additionally, the table shows the descriptive statistics of the country indices change on a monthly basis. CROBEX is the official index of the Zagreb Stock Exchange, Croatia. SOFIX is the official index of the Bulgarian Stock Exchange and BET is the main index of the Bucharest Stock Exchange, Romania. STOXX Europe 600 represents the market index.

The first two columns show the mean and median statistics for the macroeconomic variables and the indices, followed by the third and fourth columns showing the maximum (Max) and minimum (Min) observations throughout the whole period. The variability in the data is measured in the fifth column with the standard deviation (Std.Dev.) statistic. It can be seen that the mean and variability indicators for the three macroeconomic factors differ across countries as well as their extreme values.

It is shown that the mean statistic differs in size between the indices. For the country indices it is positive and higher whereas for the market index it is negative and lower. Additionally, it can be seen that both of the extreme values (Max. and Min) have higher absolute values which is in line with the hypotheses that Emerging markets offer higher return with more risk (Harvey, 1995). The variability in the data is similar for the country indices and lower for the marked index.

The descriptive statistics results for macroeconomic variables and indices, applying for the periods: before and after joining EU, can be found in Appendix B (Tables 8, 9).

4.2 Correlations

17

Table 3

Correlations between macroeconomic variables and country indices

Croatia TB CPI BR CROBEX

TB CPI BR CROBEX 1 -0.12478* -0.01349 -0.09175 1 0.027786 0.1907*** 1 -0.0128 1

Bulgaria TB CPI BR SOFIX

TB CPI BR SOFIX 1 -0.0155 -0.01469 -0.00391 1 0.15* 0.107 1 0.0462 1

Romania TB CPI BR BET

TB CPI BR BET 1 0.09722 0.032527 0.063301 1 0.007761 -0.174*** 1 -0.1625*** 1

Notes: The table shows the correlation coefficients between the macroeconomic variables and each country index. Change in Trade balance (TB), Consumer Price Index (CPI), Bank rate (BR), Crobex index (CROBEX), Sofix index (SOFIX), BET index (BET) were used and the statistics are shown for the three countries as indicated in the table. The sample for Croatia has 227 observations, Bulgaria 218 observations and Romania 229 observations. Statistical significance at the 10%, 5%, and 1% levels is indicated with *, **, ***, respectively.

It can be seen from Table 3 that in addition to difference in their scale the correlations of the macroeconomic variables and indices have different signs across countries. We observe low correlations between all variables and significance for the correlation between TB and CPI for Croatia and between CPI and BR for Bulgaria. Additionally, CROBEX and CPI show significance in the output for Croatia and BET is found to have significant correlations with CPI and BR for Romania. The highest positive output is between CROBEX and CPI with a value of 0.1907. The highest negative output is between BET and CPI, -0.174.

The correlations between the macroeconomic variables for the periods before and after EU accession can be found in Appendix C (Tables 10 and 11).

Table 4 presents the correlation statistics for the three country indices and the market index, for the whole period of the study (01/2000 – 01/2019), as well as the correlations of the market index with the separate country indices for the 2 periods (before and after EU accession).

Table 4

Correlations between country and market indices returns Full

Sample CROBEX SOFIX BET STOXX

18 CROBEX SOFIX BET STOXX 1 0.513044*** 0.44447*** 0.511368*** 1 0.50078*** 0.40114*** 1 0.51*** 1 STOXX CROBEX SOFIX BET 1 0.563*** 0.1704 -0.0676 1 0.129 0.534*** 0.699*** Notes: The first part of this table displays the correlation coefficients between country and market indices returns for the full sample (01/2000 – 01/2019) as indicated. The second part of the table presents the correlations of the market index with the country indices for the periods before and after EU joining for the separate countries. CROBEX is the official index of the Zagreb Stock Exchange, Croatia. SOFIX is the official index of the Bulgarian Stock Exchange and BET is the main index of the Bucharest Stock Exchange, Romania. STOXX represents the market index. Statistical significance at the 10%, 5%, and 1% levels is indicated with *, **, ***, respectively.

The first part of the table, using the full sample, shows highly significant results. It can be seen that the 4 indices used are positively correlated which can possibly be explained by the overall effect of financial integration (Claessens and Schmukler, 2007; Tong, Chen and Buckle, 2018).

To further examine if there is a degree of EU specific integration the second part of the table shows the correlations of the three country indices with STOXX Europe 600 for the periods before and after joining the EU for each country (each country’s correlation is consistent with its own EU accession).

We can see that the correlations of SOFIX (Bulgaria) and BET (Romania) with the STOXX Europe 600 index are low and insignificant before EU accession (0.1704 and -0.0676, respectively). After joining the EU, these statistics increase substantially to 0.534 and 0.699 and become significant at the 1% confidence level. This result can possibly be explained with EU membership integration (Bekaert et al. 2013) but such interpretations should be considered with caution since correlation is not necessarily a consequence of causation.

On the other hand, we can see different results for the correlation between CROBEX (Croatia) and STOXX Europe 600. Before joining the EU, the two indexes exhibit a correlation of 0.563 and after joining, 0.129. Only the correlation before joining is found to be significant (1%). Again with caution with regards to correlation and causation this paper offers the following possible explanations. The first one is the low number of observations, because of its late EU accession. It has 66 observations compared to 145 for SOFIX (Bulgaria) and BET (Romania). The second one comes from more intuitive point of view. Croatia first applied to join the EU in 2003 and the European Commission recommended making it official candidate and started negotiations in 2005. The accession process was prolonged due to political issues between Croatia and Slovenia (Eurostat). Nevertheless, being recognized as an official candidate in 2005, which is 8 years before the country actually joins the union, Croatia is possible to have undergone a process of development and integration to the EU prior to its accession.

19

Table 5 shows the results of the OLS regression equation: 𝑅_𝑡= 𝛼 + 𝛽₁𝑒𝑐𝑜_𝑡+ 𝛽₂𝑖𝑛𝑓_𝑡+ 𝛽₃𝑖𝑛𝑡_𝑡+ 𝛽₄𝑚𝑘𝑓_𝑡+ 𝜀_𝑡 Table 5 OLS Regressions Croatia Before After Bulgaria Before After Romania Before After C TB CPI BR STOXX Adj. R² 0.0006 (0.0029) -0.0116 (0.0098) 2.0497 (1.3271) 0.00013 (0.0092) 0.9059*** (0.2100) 0.03303 -0.0036 (0.0017) 0.0008 (0.0072) 0.7338 (0.5935) 0.0034** (0.0015) 0.1099 (0.0782) 0.0465 0.0119** (0.0046) -0.0031 (0.0097) 0.9097 (0.7447) -0.0056 (0.0088) 0.2657 (0.2017) 0.0432 -0.0023 (0.0031) -0.0002 (0.0005) 0.3511 (1.2808) -0.0026 (0.0204) 0.9588*** (0.3083) 0.2888 0.0139*** (0.0033) 0.0139*** (0.0033) -3.1992** (1.4122) -0.0054 (0.0039) -0.0842 (0.1882) 0.1487 -0.0008 (0.0018) 0.0041 (0.0047) -0.3516 (0.7589) -0.0229*** (0.0071) 1.2424*** (0.1616) 0.5082 Notes: The table presents the OLS regression outputs for Croatia, Bulgaria and Romania, where the dependent variables are respectively CROBEX, SOFIX and BET. The standard errors are reported in parentheses. The dataset is based on monthly observations with 2 observation periods for each country. The period before joining the EU for Croatia, 02/2000 – 07/2013 covers 162 observations. The period after joining the EU, 07/2013 – 12/2018 covers 66 observations. For Bulgaria before EU, 12/2000 – 01/2007, 74 observations and after EU, 01/2007 – 01/2019, 145 observations and for Romania, 01/2000 – 01/2007, 85 observations and 01/2007 – 01/2019, 145 observations. The constant term (C), Trade balance (TB), Consumer Price Index (CPI), Bank rate (BR) and STOXX Europe 600 (STOXX) statistics are shown for the three countries as indicated in the table. The last row shows the Adjusted R² (Adj. R²). Regressions are done with heteroskedasticity and autocorrelation consistent (HAC), Newey-West estimation. Statistical significance at the 10%, 5%, and 1% levels is indicated with *, **, ***, respectively.

Because the OLS model assumes that the residuals are independent and normally distributed with constant variance, a Newey-West estimator that uses Heteroskedasticity and Autocorrelation Corrected (HAC) standard errors, is applied (Newey and West, 1987). This is done using EViews, Newey-West automatic lag and bandwidth method, by which the sample size defines the number of lags.

20

The output for Croatia, does not show any macroeconomic variable to be significant except for the BR after the country joined the EU. It is significant at the 5% level and has a coefficient of 0.0034. The market factor STOXX is shown to be highly significant at the 1% level for the period before the country joined the EU. The value of the coefficient is 0.9059.

For Bulgaria the results show no significant macroeconomic variable for any of the two periods, but we can see that the market factor is highly significant at the 1% level with a coefficient of 0.9588 after EU accession.

The results for Romania before joining the EU show significance at the 1% level for TB and at the 5% level for CPI with coefficients of respectively 0.0139 and -3.1992. After joining we have a significance level of 1% for both BR and STOXX with coefficients of -0.0229 and 1.2424.

If we try to consider the results for the separate countries together, we can see that TB and CPI, are significant before EU accession (Romania) and that BR is significant after EU accession (Croatia and Romania). These particular results are consistent with such studies that find economic activity, inflation and interest rates to be able to explain stock returns (Chen et al., 1986; Asprem, 1988; Fifield, Power and Sinclair, 2002; Mateev and Videv, 2008). In the rest of the cases the macroeconomic variables exhibit insignificant results. Based on the literature discussed, these findings can potentially be accounted to the inability of this exact macroeconomic variables to explain stock returns or the inefficiency of the methodology to capture the pricing relationship (Poon and Taylor, 1991). It should also be mentioned that macroeconomic variables are found to have different power of significance in different countries and in some cases they do not exhibit any explanatory power (Fifield, Power and Sinclair, 2002).

An interesting observation arises when we observe the significance of the market index. We can see that for Bulgaria and Romania, from insignificant before joining EU, STOXX becomes highly significant (with high coefficients as shown above) after joining. Because the STOXX Europe 600 represents companies from 17 developed European

countries, its ability to explain stock returns after EU accession can possibly be accounted to the effect of overall and financial integration that can be global and EU specific (Claessens and Schmukler, 2007; Tong, Chen and Buckle, 2018; Bekaert et al. 2013).

21

and after it became an official candidate. The results can be found in Appendix D and will be discussed here.

The first period we observe looking at Table 12 in Appendix D is 02/2000 – 02/2005. It can be seen that TB is significant at 5% with a coefficient of -0.0207, BR is significant at 10% with a coefficient of -0.0144 and STOXX is significant at the 1 % confidence level with a coefficient of 0.3993. The second period covers 02/2005 – 12/2018. We can see that CPI is significant at 5% with a coefficient of 2.05 and that STOXX is significant at 1% with a coefficient of 1.0112. Unlike our main result, (using EU accession) in this case STOXX is highly significant, with high coefficient in the second period. It is also highly significant, although with a lower coefficient for the first period.

Other possible reasons for the different results of Croatia might be that the country differs from Bulgaria and Romania with regards to its financial markets and economy or simply has different characteristics in such manner that different forces affect its market. This goes beyond the scope of this study and will not be further discussed.

To test if using two separate regressions, representing the periods before and after EU accession for each country, is the best fit, a Chow test (Chow, 1960) is done. Its purpose is to indicate if there is a significant difference in the estimated equations. A significant difference would indicate a structural change in the relationship. The null hypothesis of this test is that there is no break point (meaning that it is better the data set to be represented by only 1 equation). The test is done for each country separately and the null hypothesis is rejected in all cases. The results can be observed in Appendix D, Table 13, and show significance at 5% for Croatia and at 1% for Bulgaria and Romania. Therefore, using two separate regressions is more effective.

22

activity. Finally, the same importance applies to the market factors’ ability to explain stock returns in Emerging stock markets as it is found to vary across the periods before and after joining the EU.

Table 6 shows the OLS regression results using with dummy variables included: 𝑅_𝑡= 𝛼 + 𝛼𝐷 + 𝛽₁𝑒𝑐𝑜_𝑡+ 𝛽₁𝑒𝑐𝑜_𝑡𝐷 + 𝛽₂𝑖𝑛𝑓_𝑡+ 𝛽₂𝑖𝑛𝑓_𝑡𝐷 + 𝛽₃𝑖𝑛𝑡_𝑡+ 𝛽₃𝑖𝑛𝑡_𝑡𝐷 + 𝛽₄𝑚𝑘𝑓_𝑡+ 𝛽₄𝑚𝑘𝑓_𝑡𝐷 + 𝜀_𝑡

Table 6

OLS regressions using dummy variables

Croatia Bulgaria Romania

C C Dummy TB TB Dummy CPI CPI Dummy BR BR Dummy STOXX STOXX Dummy Adj. R² 0.0038 (0.003) -0.0007 (0.0034) -0.012 (0.0099) 0.0128 (0.0121) 2.1292 (1.3785) -1.3954 (1.4928) 0.000001 (0.00924) 0.0033 (0.00934) 0.9075*** (0.2117) -0.7976*** (0.2255) 0.3190 0.01196*** (0.0045) -0.0143*** (0.0055) -0.0031 (0.0092) 0.0029 (0.0093) 0.9057 (0.7238) -0.5546 (1.4766) -0.0057 (0.0085) 0.0031 (0.0222) 0.2651 (0.1890) 0.6936** (0.3631) 0.2467 0.0142*** (0.0031) -0.0150*** (0.0036) 0.0142*** (0.0039) -0.0101* (0.0061) -3.2521*** (1.3291) 2.9009** (1.5306) -0.0053 (0.0036) -0.0176** (0.008) -0.071 (0.1862) 1.3134*** (0.2474) 0.3813

23

As it was done for the first regression in Table 5, Newey-West estimator is applied in order to fix for potential heteroscedasticity (variance of the residuals is not consistent) and autocorrelation (the residuals are not independent).

The coefficients of the four variables and their dummy variables like in Table 5, have the highest absolute values for the Consumer Price Index (CPI) and STOXX Europe 600 (STOXX).

The output for Croatia does not show any of the microeconomic variables to be significant in either period, but the market index is found to be significant in both of the cases. First the table shows a coefficient of 0.9075 for STOXX, before Croatia joined the EU. The result is significant at the 1% level. Then we have a coefficient of -0.7976 for the Dummy of STOXX, significant at the 1% level, which addresses the period after EU accession. That means that the effect after joining has a coefficient of about 0.1099 (0.9075 – 0.7976) and therefore, joining the EU has made CROBEX less sensitive to STOXX.

The results for Bulgaria as well do not find any of the macroeconomic variables significant for the two periods of the study, however we do see STOXX Dummy, which accounts for the market index after EU accession to be significant at the 5% level with coefficient of 0.6936. In the dummy model that makes the coefficient after EU accession to be 0.9587 (0.2651 + 0.6936).

For Romania the results show TB to be significant at the 1% with a coefficient of 0.0142 before joining EU and TB Dummy to be significant at 10% with a coefficient of -0.0101, making the effect of TB after joining to be 0.0041. CPI is significant at 1% with a coefficient of -3.2521 and CPI Dummy is significant at 5% with a coefficient of 2.9009, so the effect of CPI after EU accession is about -0.3512 (-3.2521 + 2.9009). It can be seen from these results that, both TB and CPI had stronger effect on the stock returns before EU accession. BR Dummy is found to be significant at 5% (-0.0176), making the effect of BR after EU accession -0.0229 and STOXX Dummy is significant at 1% (1.3134), which means that the effect of STOXX after joining the EU was about 1.2424.

In the dummy regression we only observe the index of Romania, BET to have its stock returns significantly explained by the macroeconomic variables (TB, CPI and BR). Those findings are in line with the studies mentioned in the first OLS model (Chen et al., 1986; Asprem, 1988; Fifield, Power and Sinclair, 2002; Mateev and Videv, 2008). We should again mention potential reason for the portion unexplained by the macroeconomic variables like the inability of this exact macroeconomic variables to explain stock returns or the inefficiency of the methodology to capture the pricing relationship (Poon and Taylor, 1991) as well as the fact that different countries may exhibit different results (Fifield, Power and Sinclair, 2002).

24

the possibility of financial integration that occurs both in time and after EU accession (Claessens and Schmukler, 2007; Tong, Chen and Buckle, 2018; Bekaert et al. 2013).

It would also be appropriate at this moment to introduce the idea that the CAPM (discussed in section 3.1 Methodology review) may have higher explanatory power after a country joins the EU. The market index in our study STOXX matches the factor of the CAPM: the risk of the overall market (Sharpe, 1964; Lintner, 1965), since the country indices can be looked upon as part of the European portfolio set. As it was shown in both models so far, STOXX was found to be significant with high coefficients after EU accession for Bulgaria and Romania.

In the case of Croatia this study again finds different results regarding the ability of STOXX to explain the returns of CROBEX. The coefficient effect goes from 0.9075 before EU accession, to 0.1099 after EU accession (both times significant at 1%). To further investigate this finding, as it was done for the first model, a new regression is performed using the date of becoming an official candidate for the EU to separate the two period.

The results can be found in Appendix D (Table 14). The first period is 02/2000 – 02/2005, and the second period covers 02/2005 – 12/2018. TB for the first period is significant at 10% with a coefficient of -0.02099, while TB Dummy is significant at 10% with a coefficient of 0.02282. That means the effect of TB for the second period is 0.00183. BR for the first period is significant at 10% with a coefficient of -0.01437 and BR Dummy is significant at 5% with a coefficient of 0.02282. That makes the effect of BR in the second period 0.00845. Finally, we can see that STOXX and STOXX Dummy are significant at the 1% with coefficients of 0.3858 and 0.6254 respectively, meaning that the effect STOXX has on CROBEX has increased for the second period from 0.3858 to 1.0112.

Similarly, to the first regression, we can engage an economic discussion about the results that can be observed in Table 6. Again the fact that the explanatory power of economic activity, inflation, interest rates and the market factor can be significant for some countries and not existent for others is an important observation for investors that look at these indicators, when making investment decisions for Emerging stock markets. Furthermore, Table 6 allows the reader to observe the differences that occur in the explanatory power of the 4 factors. In the output for Romania, economic activity and inflation measures are shown to be more important before EU accession, whereas interest rates better explain stock returns, after EU accession. These results can be important for investors that base their strategies on such measures, and for governments of future EU members since the stock market is a big part of the economy and they need to be aware of potential changes in factors that affect it. Additionally, the market index also exhibits different explanatory power in the different periods, which is important for investment and diversification strategies.

25

As discussed earlier this regression is done with the intention to correct for possible endogeneity problem that occurs when there is omitted variable correlated with some of the regressors, the dependent variable and one or more of the independent variables are determined simultaneously in a system or there is a measurement error.

Table 7

Panel Least Squares regressions CRO BUL ROM

Before After BUL ROM Before After C TB CPI BR STOXX Adj. R² 0.0120*** (0.0023) 0.005 (0.0045) -0.6027 (0.6378) -0.0065* (0.0038) 0.2140 (0.1063) 0.017 0.0011 (0.0011) 0.0002 (0.0004) -0.2016 (0.4617) 0.0025 (0.0052) 0.2579*** (0.0724) 0.041 0.015*** (0.0032) 0.0089* (0.0053) -0.9656 (0.7422) -0.0054 (0.0046) 0.0901 (0.1448) 0.007 -0.0015 (0.0017) -0.0002 (0.0007) 0.003 (0.6207) -0.0138* (0.008) 1.0959*** (0.0842) 0.3766

Notes: This table presents Panel Least Squares regressions. The standard errors are reported in parentheses. First the table shows the results for Croatia, Bulgaria, Romania for the periods before joining the EU, 01/2000 – 01/2007 with 85 periods and after joining the EU, 07/2013 – 01/2019 with 67 periods. Then it shows the results using only Bulgaria and Romania for the period before joining, 01/2000 – 01/2007 with 85 periods and the period after joining, 01/2007 – 01/2019 with 145 periods included. The constant term (C), Trade balance (TB), Consumer Price Index (CPI), Bank rate (BR) and STOXX Europe 600 (STOXX) statistics are shown for the three countries as indicated in the table. The last row shows the Adjusted R² (Adj. R²) statistic. Statistical significance at the 10%, 5%, and 1% levels is indicated with *, **, ***, respectively.

Looking at the output for Croatia, Bulgaria and Romania we can see that BR is significant at 10% (-0.0065) before EU accession and that STOXX is significant at 1% (0.2579) after accession.

The sample including only Bulgaria and Romania shows TB to be significant at 10% with a coefficient of 0.0089 before joining the EU. After joining we observe that BR is significant at 10% with a coefficient of -0.0138 and STOXX is significant at 1% with a coefficient of 1.0959.

26

It is important to note that in both scenarios that are described in the panel data model, from being insignificant in explaining stock returns, STOXX becomes highly significant after EU accession. These results once again bring attention to the CAPM and its possible feature to better explain stock returns in this study after EU accession.

As some additional information and following the approach we took in the OLS and OLS dummy regressions this study will once again adjust the periods for Croatia, run the panel data regression and check the results, which can be found in Appendix D (Table 15). To include all three countries, the first period becomes 01/2000 – 02/2005 and the second period becomes 01/2007 – 01/2019. The results show that BR is significant at 5% for the first period with coefficient of -0.0105 and that STOXX becomes significant in the second period with a coefficient of 1.078.

The economic implication behind the results from Table 7 is similar to what was discussed for the previous two regressions. The output for Bulgaria and Romania indicates that economic activity has higher explanatory power before joining and interest rates have higher explanatory power, after joining the EU. When Croatia is added, it can be observed that interest rates are related with stock returns before joining the EU. The market index has explanatory power after EU accession for both cases. Again, these results can have practical implication for investors that consider such factors in their investment and diversification decisions, as well as for governments of countries with Emerging markets that will become EU members.

4.4 Further discussion and summary of the results

If we consider again the OLS and OLS Dummy regressions, we can observe that the trade balance (TB) and the Consumer Price Index (CPI) exhibited their highest coefficients and significance levels before joining EU (particularly for Romania since no macroeconomic variables were found to be significant before EU accession for Croatia and Bulgaria). For the OLS model, TB and CPI showed coefficients of 0.0139 and -3.1992, respectively (TB significant at 1% and CPI at 5%). For the OLS Dummy model, 0.0142 and -3.2521, respectively (both significant at 1%).

If we consider the same variables after joining EU we can see that the only the OLS Dummy model finds some significance. TB has a coefficient of 0.0041 and is significant at 10% and CPI has a coefficient of 0.3512 and is significant at 5%. Comparing these specific findings shows that the two variables had stronger effect before joining EU.

27

potential direction might be that each country that joins the EU, becomes under supervision of The Single Supervisory Mechanism. This mechanism is exercised by the European Central Bank (ECB) and applies rules and checks on the national authorities (Eurostat). These actions can possibly make changes in such manner that stock returns became related with interest rates. Looking at the OLS and OLS Dummy results for STOXX, we can see that for Bulgaria and Romania (OLS) it goes from insignificant before EU accession, to significant at 1% with a coefficient of 0.9588 for Bulgaria and significant at 1% with a coefficient of 1.2424 for Romania, after EU accession (OLS Dummy shows 5% significance for Bulgaria, 0.9587 and 1% significance for Romania, 1.2424). These results show that STOXX can be considered to have higher explanatory power once EU accession occurs, which as already discussed can be seen as a support to the CAPM for that period.

For Croatia we observe different results. The OLS model shows STOXX to be significant at 1% (0.9059) before joining the EU, while no significance was found for the period after joining. The OLS Dummy model showed that STOXX is significant at 1% (0.9075) before joining and significant at 1% (0.1099) after joining the EU. These findings and potential reasons for them were already discussed in the previous section and additional regressions were performed for Croatia by examining the periods before and after it became an official member for the EU (Appendix D). In this scenario the OLS shows STOXX to be significant at 1% (0.3993) before joining EU and to be significant at 1% (1.0112). We can see that in this manner STOXX had stronger effect on stock returns in the second period. The results are similar for the OLS Dummy.

Summarizing the Panel Least Squared regressions for Bulgaria, Croatia and Romania we can see that BR was significant at 10% before EU accession with a coefficient of -0.0065 and that STOXX was significant after EU accession at 1%, with a coefficient of 0.2579. The same procedure, performed only for Bulgaria and Croatia (due to the fact that they join at the same time and no observations will be cut) showed TB to be significant at 10% with a coefficient of 0.0089 before joining the EU. After joining the EU, BR and STOXX became significant at 10% (-0.0138) and 1% (1.0959), respectively. Again we can see that STOXX has explanatory power after EU accession.

5. Conclusion

28

mean that joining the EU changes the degree to which factors can influence stock returns. Specifically trade balance (TB), and Consumer Price Index (CPI) exhibited consistently stronger and more significant results throughout the study, for the period before joining the EU. On the other hand, Central bank interest rate (BR) was found to have higher explanatory power after EU accession. Finally, the results for the market index (STOXX), showed that it has high explanatory power after EU accession for Bulgaria and Romania. Being the only global risk factor in this study, the significance of the market factor can be viewed as a consequence of the occurring integration when a country joins the EU. Although the results for Croatia presented STOXX to be significant before EU accession, after adjusting the periods to be divided by the moment when Croatia became an official candidate for the EU, rather than the moment when it joined, the results were stronger for the second period. The overall results for Croatia might be explained by its prolonged accession process, as discussed throughout this study.

A potential limitation of this study is the unavailability and frequency of relevant proxies that represent the economic activity, inflation and interest rates for this Emerging market countries. This paper picked Croatia, Bulgaria and Romania since they joined the EU most recently and may have undergone similar integration processes that may apply for future members. Nevertheless, next researches may consider adding some of the 10 countries that joined in 2004, although so many members joining all at once may have some effect, on both integration and the stock market, which needs to be addressed for.

6. Bibliography

Antoniou, A., Ergul, N., Holmes, P., 1997. Market efficiency, thin trading and non-linear behaviour: evidence from an Emerging market. European Financial Management 3, 175– 190.

Aslanidis, N., Savva, C.S., 2008. Stock market integration between new EU member states and the Euro-zone. Working Papers 2072/13263. Universitat Rovira i Virgili, Department of Economics.

Asprem, M., 1989. Stock prices, asset portfolios and macroeconomic variables in ten European countries. Journal of Banking & Finance 13(4-5), 589-612.

Baker, M., Wurgler, J., 2006. Investor Sentiment and the Cross-Section of Stock Returns. The Journal of Finance 61(4), 1645-1680.

Bekaert, G., Harvey, C., Lundblad, C., Siegel, S., 2013. The European Union, the Euro, and equity market integration, Journal of Financial Economics 109(3), 583-603.

29

Bodie, Z., 1976. Common stocks as a hedge against inflation. The Journal of Finance 31, 459-470.

Campbell, J.Y., 1987. Stock returns and the term structure. Journal of Financial Economics 18(2), 373-399.

Campbell, J.Y., 1993. Understanding Risk and Returns. NBER Working paper, No. 4554. Cambridge: National Bureau of Economic Research.

Cappiello, L., Engle, R.F., Sheppard, K., 2006. Asymmetric Dynamics in the Correlations of Global Equity and Bond Returns. Journal of Financial Econometrics 4(4), 537–572. Chen, NF., Roll, R., Ross, S., 1986. Economic Forces and the Stock Market. The Journal of

Business 59, 383-403.

Chen, SS., 2009. Predicting the bear stock market: Macroeconomic variables as leading indicators. Journal of Banking & Finance 33(2), 211-223.

Cheng, A., 1996. Economic Factors and Stock Markets: Empirical Evidence from the UK and the US. International Journal of Finance & Economics 1, 287-302.

Chow, G.C., 1960. Tests of Equality between Sets of Coefficients in Two Linear Regressions. Econometrica 28, 591-605.

Christiansen, C., Ranaldo, A., 2009. Extreme coexceedances in new EU member states' stock markets. Journal of Banking & Finance 33(6), 1048-1057.

Dhrymes, P.J., Friend, I., Gültekin, N.B., 1984. A critical reexamination of the empirical evidence on the arbitrage pricing theory. The Journal of Finance 39(2), 323-346.

Edison, H., Levine, R., Ricci, L., Sløk, T., 2002. International financial integration and economic growth. Journal of International Money and Finance 21(6), 749-776.

Eurostat. European Neighborhood Polity And Enlargement Negotiations. ec.europa.eu Eurostat. Tables, Graphs and Maps Interface (TGM) table. ec.europa.eu

Eurostat. Guide to Banking Supervision, September 2014. consilium.europa.eu

Fama, E.F., 1965. The Behavior of Stock-Market Prices. The Journal of Business 38(1), 34-105.

Fama, E.F., 1981. Stock Returns, Real Activity, Inflation, and Money. The American Economic Review 71(4), 545-565.

Fama, E.F., 1990. Stock Returns, Expected Returns, and Real Activity. The Journal of Finance 45(4), 1089-1108.