The Dominant Presence of Conglomerates in Emerging Markets
Enis Fesci
10425721
Bachelor’s Thesis – Bsc Economics and Business (Economics and Finance)
Supervisor: Dr J. J. G. Lemmen
Abstract
Large, diversified conglomerates dominate the corporate world in emerging markets.
Conglomerates formed as a result of the conglomerate merger wave in the late 1960s, but most in developed markets dissolved shortly after due to their inefficiencies. In emerging markets, however, conglomerates continued to thrive to present day. Emerging markets are characterized by elements such as political instability and higher stock market and exchange rate volatility that make them a different macroeconomic environment than developed markets. In this study a predictive model was created to test the effect of being a conglomerate, being a conglomerate in an emerging market or a developed market and total assets on two financial performance measures: the Jensen’s alpha and the Sharpe ratio. I found evidence that conglomerates have a lower Jensen’s alpha but no evidence that they have a lower Sharpe ratio than
non-conglomerates. There was no evidence that conglomerates in emerging markets have better financial performance than conglomerates in developed markets. There was strong evidence that firms with greater total assets have lower financial performance.
Key Words: Conglomerates, Emerging Markets, Firm Performance JEL Classification: G14
2
Statement of Originality
This document is written by Student Enis Fesci who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3
Table of Contents
1 Introduction --- 4
2 Literature Review --- 5
2.1 The Conglomerate Merger Wave --- 5
2.2 The Break Up of Conglomerates in the U.S. --- 7
2.3 The Success of Conglomerates in Emerging Markets --- 8
2.4 Should Emerging Markets Dismantle Conglomerates as Developed Markets Did? --- 9
2.5 Distinct Characteristics of Emerging Markets --- 11
2.6 Family Control and Pyramidal Ownership --- 13
2.7 Government Ownership --- 14
2.8 Table of Past Empirical Studies --- 15
2.9 Summary of Literature Review --- 19
3 Methodology --- 20
3.1 Regression Model --- 20
3.2 Regression 1 – Jensen’s Alpha as the Dependent Variable --- 21
3.3 Regression 2 – Sharpe Ratio as the Dependent Variable --- 24
3.4 Hypotheses --- 25
4 Data --- 26
5 Regression Outputs and Analyses --- 28
5.1 Regression 1 Output and Analysis --- 29
5.2 Regression 2 Output and Analysis --- 31
5.3 Hypothesis Testing and Decisions on Hypotheses --- 32
6 Conclusion and Limitations --- 34
6.1 Summary of Study and Conclusion --- 34
6.2 Limitations and Opportunities for Future Research --- 36
References --- 37
Appendices --- 41
Appendix 1 – Calculation of Alpha for Koç Holding and Wal-Mart Stores --- 41
Appendix 2 – Market Indices Used --- 42
Appendix 3 – Sample of Public Firms used in Regression --- 42
Appendix 4 – Merger Waves and the Conglomerate Merger Wave --- 45
4
1 Introduction
The two largest corporations in the United States and Turkey based on revenues are Wal-Mart Stores and Koç Holding, respectively. In the 2014 fiscal year Wal-Mart Stores had a total revenue of $476 billion and Koç Holding’s revenue for the same period was $31 billion. These two corporations are also leaders in their own countries when it comes to total number of employees, with over 2.2 million people employed at Wal-Mart Stores and over 85,000 people employed at Koç Holding (Wal-Mart Stores, 2015) (Koç Holding, 2015).
When asked “in what sector do these firms operate in?”, in the case of Wal-Mart Stores the answer is simple: they operate hypermarkets, discount stores and groceries. The industry that Wal-Mart Stores is in can be summarized in one word: retail. Answering the same question about Koç Holding is not as simple.
Koç Holding is a key player in Turkey’s automotive manufacturing industry. They also dominate Turkey’s energy market, refining, storing and distributing oil and natural gas, also producing and distributing electricity. Koç Holding owns and operates one of Turkey’s largest and most successful commercial banks. They build ships for leisure, cargo and defense.
Koç Holding is one of Europe’s largest durable goods producer. Also, they own and operate one of the most prestigious universities in the country, as well as a prestigious elementary and high school. Other industries that Koç Holding has operations in are food, retail, construction and IT.
Starting out in the electricity, construction and automotive industry when first
incorporated in 1938 (Colpan & Jones, 2016, p.75) Koç Holding grew to over eight decades to the large diversified conglomerate it is today, operating in numerous industries. Koç Holding, along with a handful of other family owned and operated diversified conglomerates dominate Turkey’s corporate world.
This phenomenon does not only occur in Turkey, but is a recurring pattern among the emerging economies, in which conglomerates such as Reliance Industries in India, Astra
International in Indonesia and DRB-HICOM in Malaysia have a dominant presence. Meanwhile in the developed world, core competencies and focus is the key strategy of the largest and most successful corporations (Khanna & Palepu, 1997).
In this study I will attempt to find a reason for the dominance of conglomerates in emerging markets and their much less significance in developed markets by answering my
5
research question: are diversified conglomerates more financially successful in emerging
markets than they are in developed markets?
In the next section I will conduct a literature review, studying the history of the
emergence of conglomerates, and the key differences between developed markets and emerging markets that led to conglomerates failing and dissolving in developed markets and thriving and dominating in emerging markets. Following the literature review, I will formulate my
methodology to quantify and measure the financial success of conglomerates in emerging markets and developed markets using a multiple regression ordinary least squares analysis, and I will conclude this study by analyzing the regression results based on the theoretical and
empirical findings of this study.
2 Literature Review
2.1 The Conglomerate Merger Wave
One of the mysteries of corporate finance is why there are periods when there are much higher mergers and acquisitions and periods when merger activity is much lower. In the period 1963-1964 there were 3,311 total acquisition announcements, while in 1968-1969 there were 10,569 acquisition announcements. In both the period from 1979 to 1980 and from 1990 to 1991 there were approximately 4,000 acquisition announcements while the late 1980s and late 1990s there was much more merger activity, with 9,278 announcements in 1999 alone. (Rhodes‐Kropf & Viswanathan, 2004, p. 2685).
There is clear evidence of the existence of merger waves: periods of high merger activity followed by periods of lower merger activity (Harford, 2005, p.532). Researchers have different theories as to why merger waves occur and what drives them. Harford (2005) supports the neoclassical explanation of merger waves, which is that merger waves occur in response to specific industry shocks that lead to large scale reallocation of assets (Harford, 2005, p.532). The clustering of shock-driven (be they economic, regulatory or technological shocks) industry merger waves leads to an aggregate “merger wave”. While there are also mergers that occur due to managers’ “market timing” attempts, they will not lead to collective merger waves (Harford, 2005, p.559). Contrarily, Rhodes‐Kropf & Viswanathan (2004) believe that market timing
6
attempts and valuation have a bigger role in the merger wave phenomenon. They show that merger waves and waves of cash and stock purchases can be driven by periods of overvaluation and undervaluation of the stock market (Rhodes‐Kropf & Viswanathan, 2004, p. 2710).
In this study, there will be focus on a particular merger wave that occurred during the late 1960s. This is the conglomerate merger and it led to the diversification of firms and the rise of large conglomerates (Matsusaka, 1993, p. 357). The definition for conglomerate mergers that Gribbin (1976, p. 19) uses is “conglomerate mergers are mergers between companies that do not produce similar products and where neither is an actual or potential supplier of the other”. This means that a merger is a conglomerate merger when a firm acquires another firm that operates in a completely unrelated industry to the acquiring firm.
There are several possible reasons why such a trend existed in the late 1960s. Hubbard & Palia (1999, p. 1149) found evidence that firms merged to form their own internal capital
markets due to the lack of well-developed external capital markets. Internal capital markets can be efficient by allowing conglomerates to allocate funds to its different divisions with the best opportunities, and this is a valuable real option in the presence of significant external financing costs (Hubbard & Palia, 1999, p. 1132). Rhodes‐Kropf & Viswanathan’s theory of valuation (2004) is another possible reason behind the conglomerate merger wave – the abnormal returns by bidding firms in the late 1960s on the diversified acquisition announcements suggest that the market awarded diversification (Hubbard & Palia, 1999). The positive returns on announcements eliminates managerial incentives such as “empire building” in which case the returns would be negative (Matsusaska, 1993, p. 376). Gribbin suggests that “diversification arises as a
consequence of excess capacity which firms try to extinguish by growth” (1976, p.35). A final explanation is that rather than anticipated efficiencies, an accounting gimmick was the driver behind the conglomerate merger wave (Matsusaska, 1993, p. 376). Conglomerates acquired companies with lower price-earnings (P/E) ratios than their own in order to increase their own earnings per share (EPS) and boost stock prices (although there is no evidence that firms which inflated EPS with this method earned positive returns). Appendix 4 shows a chart of the merger waves that occured from 1926 to 2010 including the conglomerate merger wave of the late 1960s.
7
2.2 The Break Up of Conglomerates in the U.S.
Whatever the cause may be of the conglomerate merger wave, diversification has not been beneficial for firms in the U.S. (Berger & Ofek, 1995, p.59). While diversification led to small benefits such as interest tax shields resulting from higher debt capacity and the ability of conglomerates to immediately realize tax savings by offsetting losses in some divisions against profits in others, these savings were only equal to 0.1% of the sales and were too small to offset the value loss caused by diversification (Berger & Ofek, 1995, p.59). Whilst Healy, Palepu & Ruback (1992) found that in general large mergers significantly improved operating cash flow returns, Melicher & Rush (1973) found as a result of their empirical study that conglomerate firms performed comparably to non-conglomerate firms but did not achieve outstanding results. Whilst the returns were similar, conglomerate firms’ securities posed greater financial risk than non-conglomerate firms, and this was evident in the pricing of their securities (Melicher & Rush, 1973, p.387).
The firms that once saw immediate positive returns on the announcement of a conglomerate merger were now trading at a “diversification discount” due to the market’s knowledge that diversified firms were systematically less profitable than focused firms (Servaes, 1996, p.1216). While diversification of activities was seen as a risk-reducing activity, recent studies have shown that conglomerate mergers had increased the vulnerability of the financial system (Laeven & Levine, 2007). Few firms, such as General Electric, succeeded in
diversification, however most went on the path of de-diversification to focus on their core competencies (Servaes, 1996, p. 1223).
Matsusaka (1993, p. 377) presents three possible explanations for the sudden rise and fall of the conglomerate firms. Firstly, because a reason of the formation of conglomerates was to internalize capital markets due to inefficient external capital markets, once the external capital markets improved the incentives to internalize them disappeared and firms no longer had a need to diversify. Secondly, because conglomeration was a new idea at the time, the first firms to diversify captured the advantages of diversification immediately, but once many firms followed, the advantages disappeared. It may be that the founders of the first conglomeration had a special talent for diversification, but the imitators did not manage to see the big gains the early movers did. The third explanation is that because conglomerates were an innovation at the time,
8
managers has no past data to examine. The future projections and benefits of diversification were based on theories and estimates. For any or all of these reasons, the trend of diversification was over and there was a return to specialization.
2.3 The Success of Conglomerates in Emerging Markets
While core competencies and focus was the corporate strategy for Western firms such as Wal-Mart Stores after dismantling the conglomerates formed in the late 1960s – large, diversified conglomerates are the dominant form of enterprise in emerging markets (Khanna & Palepu, 1997, p.41). Diversified conglomerates like Koç Holding in Turkey are present and strong in emerging countries such as Brazil, Chile, China, India, Indonesia, South Korea, Mexico, Pakistan, Thailand and many more emerging economies. These diversified firms operate in multiple industries as Koç Holding does, and generally consist of legally independent firms that are bound together by formal ties (such as equity) and informal ties (such as family) (Khanna & Yafeh, 2007, p. 331).
The characteristics of emerging markets created an environment in which large diversified conglomerates could succeed. Large conglomerates can imitate the functions of several institutions that are only present in advanced economies (Khanna & Palepu, 1997, p. 41). Western companies have access to advanced technology, cheap financing and sophisticated managerial expertise (Khanna & Palepu, 1997, p.51). In the absence of institutions providing the above and other services, conglomeration may be the best way to compete with the Western companies. Lack of skilled labor, limited enforcement of contracts, poor rule of law and other institutional deficiencies can all give rise to a conglomerates that are forced to generate all of these services internally for the benefit of its individual business segments (Khanna & Rivkin, 2001, p. 339) Koç Holding opened its own high school, prestigious university and owns one of Turkey’s largest banks. When Koç Holding needs skilled workers in their ship-building business segment, they can hire their best engineering graduates. When banks do not want to lend to one of Koç Holding’s business segments because they perceive the project to be too risky, or they offer loans with very high borrowing costs, they can just turn to the bank owned by their parent company, in which case most if not all of the adverse selection and moral hazard problem will be eliminated.
9
Conglomerates in emerging markets are more likely to receive bailouts from banks (Kim, 2014, p.316). Kim (2004) developed a model that showed that especially in countries with weak institutions (most emerging economies), for standalone firms in financial distress, the optimal bailout policy can be no bailout. In the case of conglomerates in countries with weak institutions, the lack of information on each business segment may cause the bank to find that the firm is not in a bad enough shape to be liquidated, so bailout can be an optimal policy.
Due to the markets being de-regulated only recently in emerging economies, large powerful conglomerates may be the only local firms capable of finding the capital and
management skills to take advantage of the new opportunities created by the deregulation of the markets (Khanna & Palepu, 1999, p.303).
Conglomerates in emerging markets also enjoy government support because in most cases they were formed with this support (Khanna & Rivkin, 2001, p.352) which is an advantage for any firm.
For the above reasons, after the conglomerate merger wave of the late 1960s, Western conglomerates broke up and focused on their core competencies, whilst the conglomerates in emerging markets kept growing larger and stronger and are the dominant form of enterprise in emerging markets in the present day.
2.4 Should Emerging Markets Dismantle Conglomerates as Developed Markets Did?
As emerging markets develop and open up to global competition, consultants and foreign investors increasingly pressure conglomerates in emerging markets to follow in the footsteps of the West and dismantle their conglomerates, decreasing the scope of their business activities and focusing on their core competencies (Khanna & Palepu, 1997, p.41).
There were several concerns regarding the successful management of a large, diversified conglomerate. When a conglomerate is diversified and made up of disparate business segments, there is little benefit obtained by sharing activities among the segments. When businesses are so disparate and large, the central management starts making mistakes in the management of these businesses and in the evaluation of the strategies and actions of the managers of the individual business segments (Bhattacharyya & Rahman, 2003, p.110).
10
Another reason to de-diversify is that when emerging markets opened up to global competition, global conglomerates could not effectively compete with the specialized Western competition, and the consistent advice of management consulting firms is that in order to
compete with such firms, conglomerates must divest their unrelated businesses and focus on their core competencies (Bhattacharya & Rahman, 2003, p.105).
There are also fears that under the current structure, investment analysts will not be able to analyze each business segment properly – conglomerate executives will shuffle funds from one company to another according to financial needs (Khanna & Palepu, 1997, p. 51).
Breaking up conglomerates could also help reduce the large debt burdens that some of them carry (Khana & Palepu, 1999, p. 126). Politicians and commentators also argue that although they contributed to economic growth in earlier decades, conglomerates are now the cause of economic difficulties such as slow growth and financial crises (Almeida & Wolfenzon, 2006, p.100). All of the above are reasons to consider if it is time that emerging economies follow in the footsteps of the developed world and dismantle the conglomerates that dominate the private sector.
Khanna & Palepu (1999) argue that rushing to dismantle the conglomerates that fill the current institutional voids in emerging markets can do more harm than good. The inefficiencies of the private sector may intensify and social distress may rise, which the emerging economies cannot handle as well as developed economies. If conglomerates decide to de-diversify and there is a “fire sale” of assets, fairly and reliably estimating the value of each business segment of the many conglomerates will be difficult. If the government attempts to force the conglomerates to dismantle, in defense they may use their large economic and political power to block the attempts and lead to an overall hostile tension between the countries’ economic and political leaders. Such tensions arose when Korea’s government blamed the Korean conglomerates for the Asian crisis due to profiteering and advocated their break up whilst the conglomerates responded by blaming the government (Khanna & Palepu, 1999, p.133).
Instead of rushing the process, Khanna & Palepu (1999) argue that in the short term governments and advisors should encourage conglomerate management to pursue internal reforms that improve performance, efficiency and their ability to substitute for market institutions. Transparency of the management should be encouraged in order to be able to analyze reliably the performance of the individual business segments (Khanna & Palepu, 1997,
11
p.51). In the long term, emerging market governments should focus on building and improving their market institutions (Khanna & Palepu, 1999, p.126) so the transition from large
conglomerates to specialization will be smooth and be a result of market equilibrium as it was in the West. Almeida & Wolfenzon (2006, p. 130) also believe that the break-up of conglomerates will not be voluntary, therefore the government must pursue policies that discourage their formation.
2.5 Distinct Characteristics of Emerging Markets
In order to fully understand the success and contemplate the future of conglomerates in emerging markets, it is important to fully consider all the characteristics of emerging markets that separate them from developed markets.
The first question that must be answered is: “what is an emerging market?” While most analysts define emerging markets according to characteristics such as size and growth, Khanna & Palepu (1997, p. 42) believe the most important criterion is how well an economy helps its buyers and sellers come together. They state that ideally, an economy has a range of institutions to facilitate a well-functioning market, and that developing countries severely lack the
institutions that help mitigate the three sources of market failure, which they define as information problems, misguided regulations and inefficient judicial systems.
Emerging markets are in between advanced economies who have all of the institutions needed and developing markets who are severely lacking in them. They have developed some of the institutions that encourage commerce but there are still institutional voids which must be filled by companies (most commonly conglomerates) so they can perform the basic function of the institutions themselves (Khanna & Palepu, 1997, p. 42).
Links between dramatic political events and large market movements suggest that political risk can have effects on stock returns in emerging markets (Diamonte, Liew & Stevens, 1996, p. 71). Developed markets are less vulnerable to political risk. However, during the past 10 years, there was a convergence in political risk between developed markets and emerging
markets (Diamonte, Liew & Stevens, 1996, p. 75) as emerging economies became politically safer and developed markets became politically riskier.
12
The behavior of stock returns in developed markets and emerging markets are also different. Emerging market stock returns historically have large volatility and high average returns. However, emerging market returns have a low correlation with developed country returns, meaning including emerging market assets can significantly reduce portfolio volatility and increase expected returns (Harvey, 1995, p. 811). Unlike in developed markets, the global unconditional asset pricing models are unable to be used to predict emerging market returns (Harvey, 1995, p. 812). Stock prices move together more in poor economies than in rich economies (Morck, Yeung & Yu, 2000, p. 215). Less respect for private property by the government in emerging markets and stronger legal protection in developed economies lead to high synchronicity in emerging markets and lower synchronicity in developed markets (Morck, Yeung & Yu, 2000, p. 258). The function of an efficient stock market is to process information and help guide capital to its best economic use (Morck, Yeung & Yu, 2000, p. 259). Morck, Yeung & Yu (2000, p. 259) argue that if stock price movements in emerging markets are due to politically driven shifts in property rights, inaccurate information from the stock market could cause poor allocation of capital, adversely affecting economic growth. This will have an effect on the performance of conglomerates (and non-conglomerates) in emerging markets.
Aguiar & Gopinath (2004, p.24) use a standard business cycle model to explain important differences between emerging markets and developed markets. Frequent regime changes and dramatic reversals in monetary, fiscal and trade policy lead to volatile fluctuations in
productivity. Developed markets, on the other hand, are characterized by a more stable trend (Aguiar & Gopinath, 2004, p. 2). Fluctuations in policy and productivity will also have an effect on the performance of conglomerates (and non-conglomerates) in emerging markets.
Maćkowiak (2007, p. 2519) finds that external shocks play an important role in emerging markets. There is strong evidence that the central issue for monetary policy in emerging markets is how to stabilize the economy in response to external shocks, such as a change in U.S.
monetary policy. Even if an emerging market chooses a fixed exchange rate regime and
abandons independent monetary policy, the economy must be stabilized in response to external shocks with adjustments to fiscal policy (Maćkowiak, 2007, p. 2519). Adjustments to either monetary or fiscal policy in emerging markets will directly affect the performance of conglomerates and non-conglomerates in emerging markets. Maćkowiak (2007) shows that
13
emerging markets are more sensitive to changes in U.S. monetary policy than the U.S. market itself.
All of the above factors are key differences between developed and emerging markets, and are all important to consider when analyzing the performance of conglomerates and non-conglomerates in developed and emerging markets.
2.6 Family Control and Pyramidal Ownership
Because most conglomerates in emerging markets are family owned or family controlled (Khanna & Yafeh, 2007, p. 331), the dynamics of family ownership and control is an important factor to consider when assessing the success of conglomerates in emerging markets. Almeida & Wolfenzon (2006) show that by using a pyramidal structure, families can own very little shares in different business segments of the conglomerate but still have full control. It is common for conglomerate firms to adopt a pyramidal ownership structure (Almeida & Wolfenzon, 2006, p. 2637).
In the pyramidal ownership structure, a family that owns 50% of firm 1 which in turn owns 50% of firm 2, in effect, owns 25% of firm 2. However, because firm 1 has complete control over firm 2 and the family has complete control over firm 1, the family has complete control over firm 2 even though it owns only 25% of its equity. (Almeida & Wolfenzon, 2006, p. 2638).
This is an important issue in the case of conglomerates in emerging markets, because even though the equity of the business segments may not be concentrated to a family or certain individuals, control can be and often is. Centralized family control is a key feature of
Koç Holding (Colpan & Jones, 2016), Turkey’s largest conglomerate, and Reliance Industries, India’s largest conglomerate, and this most likely affects (positively or negatively) their success compared to other firms. Villalonga & Amit (2006, p. 414) found interesting results on this issue. They found that additional value is created when the founder of the firm is the CEO, but value is destroyed when descendants of the founder become the CEO. This leads to a “family control discount” as power is passed down from generation to generation.
14
2.7 Government Ownership
Yiu, Bruton & Lu (2005) find that government ownership and affiliation is also an important determining factor in the performance of conglomerates in emerging markets. In their study of Chinese conglomerates, they found that there is a negative relationship between government ownership and the return on assets (ROA) of the conglomerate firm. They also find that there is a negative relationship between the number of previous government officials as senior managers in a conglomerate and ROA. This shows that, based in evidence from China, conglomerates
15
2.8 Table of Past Empirical Studies
Below is a table of past empirical studies carried out that are related to this study.
Author Region Period Method Independent & Control
Variable/s
Dependent Variable/s Results
Yiu, Bruton & Lu (2005) China 1998 Multiple regression Age of conglomerate Government ownership Number of government officials in management Founding dummy Acquisiton intensity Internal cabability development International diversification Control: Number of employees Current ratio Group diversification Industry competitiveness Conglomerate performance (ROA) Age of conglomerate negatively related to performance Negative relatonship between government ownership and performance Performance lower when previous government officials as management
16
(moderately supported)
Firms that pursue more acquisitions perform better Melicher & Rush (1972) USA 1966 - 1971 Difference test Conglomerates vs Non-conglomerates Conglomerate Performance: Jensen’s alpha Sharpe ratio Treynor ratio Conglomerates have slightly higher Jensen’s alpha than non-conglomerates, however there is no statistical evidence that the Jensen’s alpha of conglomerate firms is non-zero. There is also no statistical evidence of a difference of Sharpe ratio or Treynor ratio between
17 conglomerates and non-conglomerates. Weston & Mansinghka (1971) USA 1960 - 1968 Difference test Conglomerates vs random sample of non-conglomerate firms
Forbes firms vs multi-industry firms
Conglomerats vs Fortune 500 firms.
P/E ratio
Debt/net worth
5 year annual increase in sales, 5 year annual increase in EPS,
Price gain for 5 year period,
5 year return on debt and equity, 5 year average retirn on equity (ROE).
10 year compund annual growth rates of EPS, net profit to shareholder’s equity.
Conglomerate firms
outperformed other firms in all growth measures. Earnings performance of conglomerates measured by net income to shareholder’s equity is higher for conglomerates, but difference is not statistically significant. Conglomerate firms are more leveraged than
18 non conglomerates. Johnson & Soenen (2003) USA 1982 - 1998 Multiple regression Book-to-market ratio
Sustainale growth rate
Profitability (ROA)
Size
Capital structure
Liquidity
Cash conversion cycle
Earnings volatility
R&D expenditures
Advertising expenditures
Sharpe ratio,
Jensen’s alpha
Economic Value Added (EVA) Positive relationship between certain accounting performance indicators (independent variables) and the risk adjusted performance measures Sharpe ratio and Jensen’s alpha, and the EVA.
ROA is a significant indicator for all three dependent variables.
19
2.9 Summary of Literature Review
Conglomerates formed as a result of the conglomerate merger wave in the late 1960s. However, due to inefficiencies, diversified conglomerates could not compete in developed markets and by the 1980s most had divested their unrelated businesses to focus on core competencies. In emerging markets, however, conglomerates continued to grow and dominate the private sector.
As emerging economies opened up their markets, consultants advised emerging market conglomerates to divest their unrelated firms and focus on their core competencies in order to be able to compete with the developed market firms, however there are issues such as
underdeveloped external capital markets in emerging economies that may make this not such a good idea.
Emerging markets are characterized by many differences to developed markets such as increased political instability and higher stock and currency market volatility that may make emerging markets a more suitable environment for conglomerates than developed markets.
Family and government ownership/control is also an important factor to consider. By using a pyramidal ownership structure, families can have complete control of a conglomerate and its subsidiaries, even if it owns a small fraction of their equity. Value is created when the founder is the CEO, but destroyed when descendants become CEO. Conglomerates have poorer
performance when owned by the government and also when previous government officials are in senior management.
Previous empirical studies have been made investigating the performance of conglomerates vs conglomerates in developed markets and conglomerates vs non-conglomerates in emerging markets, but there are no empirical studies that investigate how conglomerates perform in emerging markets compared to their performance in developed markets. Studying this could give an insight into the dominant presence of conglomerates in emerging markets.
I have developed a strong theoretical background that will help us answer the main research question of this study: are diversified conglomerates more financially successful in
emerging markets than they are in developed markets? I have established that conglomerates
dominate the private sector in emerging markets and that emerging markets have certain characteristics that separate them from developed markets, but do these characteristics make
20
conglomerates more financially successful than they are in developed markets, and is this why they are still present and growing? In the next section, I will build a regression model in order to be able to answer the research question empirically.
3 Methodology
3.1 Regression Model
In this study, the hypotheses developed in section 3.4 regarding conglomerate firm performance in emerging and developed markets will be tested using OLS multiple regression analysis. The population is all the large firms in both developed and emerging economies listed on the stock exchange. The sample will be 60 firms, a sufficiently large to assume an approximate normal distribution. Because a randomized controlled experiment is not possible in this case (as with most economic experiments), the samples will not be completely random, to ensure that a sufficient number of firms that are in developed/emerging countries and are conglomerates/no-conglomerates are in the sample in order to have enough data to test the relationship between the dependent and independent variables. The main reason for this is that, if selected completely at random, the firm selected (especially if based in a developed economy) will most likely not be a conglomerate, as there are very few left in existence. When it comes to conglomerate firms, data will be on the level of the entire holding company.
Johnson & Soenen conducted a study using a sample of 478 firms that tested the relationship between ten company specific indicators of superior performance (such as ROE, ROA, advertising expenses as a percentage of sales and other success indicators) and three stock price success measures as the dependent variable: the Sharpe ratio, the Jensen’s alpha and the ‘Economic Value Added’ measures (2003). They found that “especially large, profitable
companies, with efficient working capital management (i.e. relative short cash conversion cycles) and certain degree of uniqueness (measured by advertising spending relative to sales) outperform the sample average on the three performance measures” (Johnson & Soenen, 2003, p. 368). The ROA was also a strong indicator of financial performance. This is an important finding, because for this study I will use the Sharpe ratio and Jensen’s alpha as dependent variables as Melicher & Rush did (1973). They are both risk adjusted measures that can be used to compare companies in
21
different countries, and Johnson and Soenen (2003) showed that these measures have a
significant positive relationship with accounting based performance measures, further validating their use as performance measures. Measures such as the ROA cannot be used in this study as Yiu, Bruton and Lu did in their study of Chinese conglomerates (2005) because such measures do not account for the difference in interest rates and market returns between countries. While a ROA of 7% may be impressive in a country like USA with a 2% risk free rate and 3% market return, it is unimpressive in a country like Turkey with a 10% risk free rate and a 12% market return. The Jensen’s alpha and Sharpe ratio adjust for these differences, and are more appropriate to this study because the sample contains firms from 15 different countries.
Using a similar approach as Melicher & Rush (1973), two regressions will be conducted, one with Jensen’s alpha as the dependent variable and one with the Sharpe ratio as the dependent variable. These will be the ‘financial success’ measures of the firms in my sample. The
independent variables will be whether the firm is a conglomerate, whether it is a conglomerate in an emerging market and the natural logarithm of the firm’s total assets.
3.2 Regression 1 – Jensen’s Alpha as the Dependent Variable
The relationship to be tested in the first regression is:
𝐴𝐿𝑃𝐻𝐴𝑖 = 𝛽0+ 𝛽1𝐶𝑂𝑁𝐺𝑖+ 𝛽2(𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) + 𝛽3𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 + 𝑢𝑖
𝐴𝐿𝑃𝐻𝐴𝑖 is the dependent variable of the above regression model, measured by the Jensen’s alpha of the firm, the average excess return of the firm’s stock in excess of the risk adjusted predicted excess return over a period of time. This is the average excess return the firm realized minus the risk adjusted predicted excess return. It is a performance measure of the firm that is equal to the alpha (Jensen’s alpha, α) in the capital asset pricing model (CAPM) model.
The simple CAPM model states that:
𝑅𝑒𝑡𝑢𝑟𝑛𝑂𝑛𝐸𝑞𝑢𝑖𝑡𝑦𝑖 = 𝑅𝑖𝑠𝑘𝐹𝑟𝑒𝑒𝑖 + 𝛽(𝑅𝑒𝑡𝑢𝑟𝑛𝑀𝑎𝑟𝑘𝑒𝑡𝑖− 𝑅𝑖𝑠𝑘𝐹𝑟𝑒𝑒𝑖) + 𝑢𝑖
22
𝐸(𝑅𝑒𝑡𝑢𝑟𝑛𝑂𝑛𝐸𝑞𝑢𝑖𝑡𝑦𝑖) = 𝑅𝑖𝑠𝑘𝐹𝑟𝑒𝑒𝑖 + 𝛽[𝐸(𝑅𝑒𝑡𝑢𝑟𝑛𝑀𝑎𝑟𝑘𝑒𝑡𝑖− 𝑅𝑖𝑠𝑘𝐹𝑟𝑒𝑒𝑖)]
By subtracting 𝑅𝑖𝑠𝑘𝐹𝑟𝑒𝑒𝑖 from both sides, the CAPM model can be rewritten as:
𝐸(𝑅𝑒𝑡𝑢𝑟𝑛𝑂𝑛𝐸𝑞𝑢𝑖𝑡𝑦𝑖) − 𝑅𝑖𝑠𝑘𝐹𝑟𝑒𝑒𝑖 = 𝛽[𝐸(𝑅𝑒𝑡𝑢𝑟𝑛𝑀𝑎𝑟𝑘𝑒𝑡𝑖− 𝑅𝑖𝑠𝑘𝐹𝑟𝑒𝑒𝑖)]
where 𝛽 is the expected percentage increase of firm return for a 1% increase in market return, measuring the firm’s volatility relative to the market. Jensen (1968, p.393) showed that by adding a constant, 𝛼, to the above CAPM model, the regression of the model will not be constrained to one without an intercept, and that the new intercept will measure the amount in which a security (or portfolio) beat its risk adjusted expected return on average. The CAPM model with the new constant, called the Jensen’s alpha (𝛼) can be written as:
𝐸(𝑅𝑒𝑡𝑢𝑟𝑛𝑂𝑛𝐸𝑞𝑢𝑖𝑡𝑦𝑖) − 𝑅𝑖𝑠𝑘𝐹𝑟𝑒𝑒𝑖 = 𝛼 + 𝛽[𝐸(𝑅𝑒𝑡𝑢𝑟𝑛𝑀𝑎𝑟𝑘𝑒𝑡𝑖 − 𝑅𝑖𝑠𝑘𝐹𝑟𝑒𝑒𝑖)]
The values of 𝛼 and 𝛽 can be estimated for each firm in my sample using the OLS estimator formulas and each firm’s monthly return, the monthly market return (measured by the monthly return of the main market index located in the same country and measured in the same currency as the stock exchange on which the firm is listed) and historical monthly risk free rates (measured by the annual yield of 10 year government bonds of the country of the stock exchange on which the firm is listed). The regression model for each firm in my sample will be in the following form:
(𝑅𝐸𝑇𝐸𝑗 − 𝑅𝐸𝑇𝑅𝐹𝑗) = 𝛼 + 𝛽(𝑅𝐸𝑇𝑀𝑗 − 𝑅𝐸𝑇𝑅𝐹𝑗) + 𝑢𝑗
where 𝑅𝐸𝑇𝐸𝑗 is the return on equity on the firm’s stock for month 𝑗, 𝑅𝐸𝑇𝑅𝐹𝑗 is the risk free return for month 𝑗, 𝑅𝐸𝑇𝑀𝑗 is market return for month 𝑗 and 𝑢𝑗 is the error term.
(𝑅𝐸𝑇𝐸𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗), the excess return of the firm’s equity for month 𝑗, is the dependent variable. (𝑅𝐸𝑇𝑀𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗), the excess return of the market for month 𝑗, is the independent variable. 𝛼
23
is the vertical axes intercept, and 𝛽 is the slope. 𝛽 is estimated by 𝛽̂ using the formula for the OLS estimator of the slope coefficient:
𝛽̂ =𝐶𝑜𝑣[(𝑅𝐸𝑇𝐸𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗), (𝑅𝐸𝑇𝑀𝑗 − 𝑅𝐸𝑇𝑅𝐹𝑗)] 𝑉𝑎𝑟(𝑅𝐸𝑇𝑀𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗)
and 𝛼 is estimated by 𝛼̂ using the formula for the OLS estimator of the intercept: 𝛼̂ = (𝑅𝐸𝑇𝐸̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ − 𝛽̂(𝑅𝐸𝑇𝑀𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗) ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗)
= (𝑅𝐸𝑇𝐸̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ − 𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗)
[𝐶𝑜𝑣[(𝑅𝐸𝑇𝐸𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗), (𝑅𝐸𝑇𝑀𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗)]
𝑉𝑎𝑟(𝑅𝐸𝑇𝑀𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗) ] × (𝑅𝐸𝑇𝑀̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ = 𝐴𝐿𝑃𝐻𝐴𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗) 𝑖
The predicted value of 𝛼 for the firms in my sample is the dependent variable in main the regression model, 𝐴𝐿𝑃𝐻𝐴𝑖, and will be calculated using 𝑚 = 80 observations of monthly historical returns between (and including) April 2008 and November 2014. Graphical
representations of the calculation of the beta and the alpha for Wal-Mart Stores and Koç Holding are shown in Appendix 1.The beta is the slope coefficient of the trend line and the Jensen’s alpha, the dependent variable of the main regression, is the vertical axis intercept of the trend line.
𝐶𝑂𝑁𝐺𝑖 is a dummy variable that is equal to 1 if the firm is a diversified conglomerate and 0 if the firm is a non-conglomerate.
(𝐸𝑀𝐸𝑅𝒊× 𝐶𝑂𝑁𝐺𝑖) is a dummy interaction term that is equal to 1 if the firm is both a diversified conglomerate and based in an emerging country and 0 otherwise. The coefficient of this variable measures the extra effect that being a conglomerate will have if the firm is based in an emerging country rather than a developed economy. The effect on the Jensen’s alpha on the firm being a conglomerate is:
∂𝐴𝐿𝑃𝐻𝐴𝑖
24
If 𝐸𝑀𝐸𝑅𝒊 = 0, (the firm is based in a developed country) the conglomerate effect will be 𝐵1. If 𝐸𝑀𝐸𝑅𝒊= 1 (the firm is based in an emerging country), the conglomerate effect will be 𝐵1+ 𝐵2.
𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 is a control variable which is equal to the natural logarithm of the total
assets of each firm for the year ended in 2014 measured in US dollars. 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 is included as a control variable because the more assets a firm has, the more difficult it is to effectively manage those assets and create positive returns. This is especially true for large, diversified
conglomerates. We want to measure the isolated effect of being a conglomerate and being a conglomerate in an emerging market therefore our model will be controlled for the size of the firm. The size is measured as the natural logarithm of total assets and not the value of the total assets because we want to control for the percentage increase in total assets not the unit increase of total assets. Apart from controlling for the effect of firm size on alpha, adding this variable will give empirical evidence on weather firm size has an effect on performance or not.
3.3 Regression 2 – Sharpe Ratio as the Dependent Variable
The relationship to be tested in the second regression is:
𝑆𝐻𝐴𝑅𝑃𝐸𝑖 = 𝛽0+ 𝛽1𝐶𝑂𝑁𝐺𝑖+ 𝛽2(𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) + 𝛽3𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 + 𝑢𝑖
𝑆𝐻𝐴𝑅𝑃𝐸𝑖 is the dependent variable of the second regression model to be tested. 𝑆𝐻𝐴𝑅𝑃𝐸𝑖 is the Sharpe ratio of the stock, and is the expected excess return of a stock or a portfolio per unit of volatility. The Sharpe ratio is named after economist William F. Sharpe, who introduced the concept five decades ago (Sharpe, 1966). The formula for the Sharpe ratio is as follows:
𝑆ℎ𝑎𝑟𝑝𝑒 𝑅𝑎𝑡𝑖𝑜 = 𝐸(𝑆𝑡𝑜𝑐𝑘 𝑅𝑒𝑡𝑢𝑟𝑛 − 𝑅𝑖𝑠𝑘 𝐹𝑟𝑒𝑒 𝑅𝑎𝑡𝑒) 𝑉𝑜𝑙𝑎𝑡𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑆𝑡𝑜𝑐𝑘 𝑅𝑒𝑡𝑢𝑟𝑛
Using the data in our sample, the Sharpe ratio will be calculated as
𝑆𝐻𝐴𝑅𝑃𝐸𝑖 =
𝑅𝐸𝑇𝐸𝑗− 𝑅𝐸𝑇𝑅𝐹𝑗 ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ √𝑉𝑎𝑟(𝑅𝐸𝑇𝐸̅̅̅̅̅̅̅̅)𝑗
25
The definitions of the variables 𝑅𝐸𝑇𝐸𝑗, 𝑅𝐸𝑇𝑅𝐹𝑗 and the regressors 𝐶𝑂𝑁𝐺𝑖, (𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖)
and 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 carry forward from the previous model with Jensen’s alpha as the dependent variable. The effect on the Sharpe ratio on the firm being a conglomerate is:
∂𝐴𝐿𝑃𝐻𝐴𝑖
∂𝐶𝑂𝑁𝐺𝑖 =𝐵1+ 𝐵2𝐸𝑀𝐸𝑅𝒊
If 𝐸𝑀𝐸𝑅𝒊 = 0, (the firm is based in a developed country) the conglomerate effect will be 𝐵1. If 𝐸𝑀𝐸𝑅𝒊= 1 (the firm is based in an emerging country), the conglomerate effect will be 𝐵1+ 𝐵2.
3.4 Hypotheses
Because of the efficiency problems of running a conglomerate firm experienced especially in developed markets, the “conglomerate effect” on the success measures of the firm (the Jensen’s alpha and Sharpe ratio) will be negative, mirroring Berger & Ofek’s findings (1995) that
diversification had value-reducing effects. However, the macroeconomic environment and capital markets make of emerging markets make them a more suitable environment for conglomerate firms to succeed, and this is the reason that they are present and still strong in such countries such as India, Turkey, Malaysia and Indonesia. Therefore the “conglomerate effect” when the firm is based in an emerging market will be greater than it is when the firm is based in a developed country. Also, larger firms (especially conglomerates) will be more difficult to manage, therefore firms with more total assets will yield lower returns.
Accordingly, for the first regression model:
𝐴𝐿𝑃𝐻𝐴𝑖 = 𝛽0+ 𝛽1𝐶𝑂𝑁𝐺𝑖+ 𝛽2(𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) + 𝛽3𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 + 𝑢𝑖
my null and alternate hypotheses are:
𝐻𝑜: 𝛽1 = 0 vs alternate hypothesis 𝐻1: 𝛽1 < 0 𝐻𝑜: 𝛽2 = 0 vs alternate hypothesis 𝐻1: 𝛽2 > 0
26
𝐻𝑜: 𝛽3 = 0 vs alternate hypothesis 𝐻1: 𝛽3 < 0
In words, I hypothesize that in the first regression model with 𝐴𝐿𝑃𝐻𝐴𝑖 as the dependent variable, the coefficient of 𝐶𝑂𝑁𝐺𝑖 will be negative, the coefficient of (𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) will be positive
and the coefficient of 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 will be negative.
Similarly, for the second regression model:
𝑆𝐻𝐴𝑅𝑃𝐸𝑖 = 𝛽0+ 𝛽1𝐶𝑂𝑁𝐺𝑖+ 𝛽2(𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) + 𝛽3𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 + 𝑢𝑖
my null and alternate hypotheses are:
𝐻𝑜: 𝛽1 = 0 vs alternate hypothesis 𝐻1: 𝛽1 < 0 𝐻𝑜: 𝛽2 = 0 vs alternate hypothesis 𝐻1: 𝛽2 > 0
𝐻𝑜: 𝛽3 = 0 vs alternate hypothesis 𝐻1: 𝛽3 < 0
In words, I hypothesize that in the second regression model with 𝑆𝐻𝐴𝑅𝑃𝐸𝑖 as the dependent variable, the coefficient of 𝐶𝑂𝑁𝐺𝑖 will be negative, the coefficient of (𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) will be
positive and the coefficient of 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 will be negative.
4 Data
The data was selected by doing online research. For several emerging and developed countries, conglomerates and non-conglomerates were selected at random, with the only criteria being that they are listed on the stock exchange, so the values of their Jensen’s alpha and Sharpe ratio could be calculated. For each company, share price data was downloaded from Yahoo! Finance for a total of 81 months (80 returns) (April 2008 to December 2014). Using data from the risk free rate (monthly yields of 10 year government bonds downloaded from investing.com) and market returns (returns of market indices obtained from Yahoo! Finance– see Appendix 2) the Jensen’s
27
Alpha and Sharpe ratio for each observation were calculated using the formulas in section 3.2 and 3.3.
Assigning a value to the CONG variable was based on personal judgement. Rather than how many subsidiaries are owned by the company, the number of unrelated industries the company had operations in were more important. Royal Dutch Shell, for example owns several subsidiaries, but they are almost all related to the exploring drilling, manufacturing and
distribution of oil. Since they are in one industry, the value assigned to the CONG variable is 0. A total of 60 random firms were selected. The only non-random component of the selection process was ensuring that there was enough data on emerging market conglomerates, emerging market non-conglomerates, developed market conglomerates and developed market non-conglomerates. A sample of 60 observations is sufficient to assume approximately normally distributed errors.
A table showing the descriptive statistics of the variables is shown below. A total list of the firms used in this study can be found in Appendix 3. There were no large outliers and therefore the third OLS assumption is not violated. There were also no issues with missing or incomplete data.
Descriptive Statistics CONG EMER*CONG LNTOTASS ALPHA SHARPE
Mean 0.4833 0.2333 3.7385 0.0049 0.0783 Standard Error 0.0651 0.0551 0.2157 0.0012 0.0118 Median 0 0 3.7844 0.0046 0.0794 Standard Deviation 0.5039 0.4265 1.6705 0.0090 0.0916 Sample Variance 0.2540 0.1819 2.7906 0.0001 0.0084 Kurtosis -2.0653 -0.3392 -0.0755 1.0492 -0.1712 Skewness 0.0684 1.2935 0.0957 0.2763 -0.1311 Range 1 1 7.9997 0.0491 0.4015 Minimum 0 0 0.0488 -0.0195 -0.1439 Maximum 1 1 8.0485 0.0297 0.2577 Sum 29 14 224.3072 0.2913 4.6975 Count 60 60 60 60 60
28
Below is a cross correlation table showing the correlation coefficients between the independent variables.
Correlation CONG EMER*CONG LNTOTASS
CONG 1
EMER*CONG 0.5704 1
LNTOTASS -0.2331 -0.3014 1
The highest absolute value of the correlation between the independent variables is 0.5704, the correlation between CONG and EMER*CONG. It can be concluded that multicollinearity will not be a problem in the regression analyses.
5 Regression Outputs and Analyses
For the firms listed in Appendix 3, using the calculated values for the Jensen’s alpha and the Sharpe Ratio as dependent variables, and the status of the firm as a
conglomerate/non-conglomerate based in an emerging/developed market and the natural logarithm of the firm’s total assets as the independent variables, two regression analyses were run. The output of the first regression model with the Jensen’s alpha as the dependent variable is shown in section 5.1. The output of the second regression model with the Sharpe ratio as the dependent variable is shown in section 5.2. In section 5.3 we will formally test the hypotheses formulated in section 3.3. The program used for the regression analyses was EViews. White heteroskedastic robust standard errors were used to correct for the possibility that the errors are not homoscedastic.
29
5.1 Regression 1 Output and Analysis
Below is the output of the first regression analysis that was run in EViews with the Jensen’s alpha (𝐴𝐿𝑃𝐻𝐴𝑖) as the dependent variable. The regression tests the model:
𝐴𝐿𝑃𝐻𝐴𝑖 = 𝛽0+ 𝛽1𝐶𝑂𝑁𝐺𝑖+ 𝛽2(𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) + 𝛽3𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 + 𝑢𝑖
Dependent Variable: ALPHA Method: Least Squares Date: 01/24/16 Time: 19:54 Sample: 1 60
Included observations: 60
White heteroskedasticity-consistent standard errors & covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.0118 0.0036 3.2734 0.0018
CONG -0.0036 0.0023 -1.5411 0.1289
EMER*CONG 0.0004 0.0036 0.1073 0.9150
LNTOTASS -0.0014** 0.0007 -2.0992 0.0403
R-squared 0.0843 Mean dependent var 0.0049
Adjusted R-squared 0.0353 S.D. dependent var 0.0090
S.E. of regression 0.0088 Akaike info criterion -6.5564
Sum squared resid 0.0044 Schwarz criterion -6.4168
Log likelihood 200.6918 Hannan-Quinn criter. -6.5018
F-statistic 1.7194 Durbin-Watson stat 1.9087
Prob(F-statistic) 0.1734
∗= significant at 10%,∗∗= significant at 5%,∗∗∗= significant at 1%
The R-squared tells us the fraction of the variation of the dependent variable explained by the variation of the independent variables. Because the R-Squared increases each time an
30
independent variable is added to the regression model, the Adjusted R-Squared corrects for this. The Adjusted R-Squared in the above output shows that only 3.5% of the variation of the dependent variable 𝐴𝐿𝑃𝐻𝐴𝑖 is explained by the variation of the independent variables 𝐶𝑂𝑁𝐺𝑖,
(𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) and 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖.
The F statistic tests the entire regression model. The null hypothesis is that all the
coefficients are equal to zero, and the alternative hypothesis is that at least one of the coefficients is not equal to zero. The significance level of the computed F statistic is 17.3%, therefore the null hypothesis should not be rejected. This shows that the regression model is not a good fit.
The signs of the coefficient are all as predicted. The “conglomerate effect” on the Jensen’s alpha is negative as seen by the negative sign of the coefficient of 𝐶𝑂𝑁𝐺𝑖, the effect becomes less negative in emerging markets as seen by the positive sign of the coefficient of (𝐸𝑀𝐸𝑅𝑖×
𝐶𝑂𝑁𝐺𝑖), and firms with greater total assets have a lower Jensen’s alpha as seen by the negative sign of the coefficient of 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖.
However, of the three coefficients calculated by the multiple regression analysis, only the coefficient of 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 is significant at 5%.
31
5.2 Regression 2 Output and Analysis
Below is the output of the second regression analysis that was run in EViews with the Sharpe ratio (𝑆𝐻𝐴𝑅𝑃𝐸𝑖) as the dependent variable. The regression tests the model:
𝑆𝐻𝐴𝑅𝑃𝐸𝑖 = 𝛽0+ 𝛽1𝐶𝑂𝑁𝐺𝑖+ 𝛽2(𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) + 𝛽3𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 + 𝑢𝑖 Dependent Variable: SHARPE
Method: Least Squares Date: 01/24/16 Time: 19:57 Sample: 1 60
Included observations: 60
White heteroskedasticity-consistent standard errors & covariance
Variable Coefficient Std. Error t-Statistic Prob.
C 0.1397 0.0339 4.126185 0.0001
CONG -0.0232 0.024547 -0.947115 0.3476
EMER*CONG 0.0056 0.035010 0.160692 0.8729
LNTOTASS -0.0138* 0.007291 -1.890615 0.0639
R-squared 0.065428 Mean dependent var 0.078292
Adjusted R-squared 0.015361 S.D. dependent var 0.091578
S.E. of regression 0.090872 Akaike info criterion -1.894381
Sum squared resid 0.462436 Schwarz criterion -1.754758
Log likelihood 60.83144 Hannan-Quinn criter. -1.839767
F-statistic 1.306816 Durbin-Watson stat 1.635538
Prob(F-statistic) 0.281257
32
The Adjusted R-Squared in the above output shows that only 1.5% of the variation of the dependent variable 𝑆𝐻𝐴𝑅𝑃𝐸𝑖 is explained by the variation of the independent variables 𝐶𝑂𝑁𝐺𝑖, (𝐸𝑀𝐸𝑅𝑖 × 𝐶𝑂𝑁𝐺𝑖) and 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖.
The significance level of the computed F statistic is 28.1%, therefore the null hypothesis should not be rejected. This shows that the regression model is not a good fit.
As was the case with the first regression, the signs of the coefficients are all as predicted. The “conglomerate effect” on the Sharpe ratio is negative as seen by the negative sign of the coefficient of 𝐶𝑂𝑁𝐺𝑖, the effect becomes less negative in emerging markets as seen by the positive sign of the coefficient of (𝐸𝑀𝐸𝑅𝑖× 𝐶𝑂𝑁𝐺𝑖), and firms with greater total assets have a
lower Sharpe ratio as seen by the negative sign of the coefficient of 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖.
However, of the three coefficients calculated by the multiple regression analysis, only the coefficient of 𝐿𝑁𝑇𝑂𝑇𝐴𝑆𝑆𝑖 is significant at 10%.
5.3 Hypothesis Testing and Decisions on Hypotheses
Using the preconditions that the regression output conducts a two sided test on the coefficient and the coefficients in the outputs all have the sign predicted in the hypotheses, we can conduct the one sided hypothesis tests simply by halving the p-values in the regression outputs.
Regression 1
𝐻𝑜: 𝛽1 = 0 vs alternate hypothesis 𝐻1: 𝛽1 < 0
p-value = 0.1289
2 = 0.06445 – reject 𝐻𝑜at 10%
There is enough statistical evidence to conclude that conglomerates have a lower Jensen’s alpha than non-conglomerates at 10% significance, but not at 5% significance (moderately supported).
𝐻𝑜: 𝛽2 = 0 vs alternate hypothesis 𝐻1: 𝛽2 > 0
p-value = 0.9150
33
There is not enough statistical evidence to conclude that conglomerates have a higher Jensen’s alpha in emerging markets than they do in developed markets.
𝐻𝑜: 𝛽3 = 0 vs alternate hypothesis 𝐻1: 𝛽3 < 0
p-value = 0.0403
2 = 0.02015 – reject 𝐻𝑜 at 2.5% significance level.
There is enough evidence at a 2.5% significance level to conclude that firms with more assets have a lower Jensen’s alpha.
Regression 2
𝐻𝑜: 𝛽1 = 0 vs alternate hypothesis 𝐻1: 𝛽1 < 0
p-value = 0.3476
2 = 0.1738 – do not reject 𝐻𝑜
There is not enough statistical evidence to conclude that conglomerates have a lower Sharpe ratio than non-conglomerates.
𝐻𝑜: 𝛽2 = 0 vs alternate hypothesis 𝐻1: 𝛽2 > 0
p-value = 0.8729
2 = 0.43645 – do not reject 𝐻𝑜
There is not enough statistical evidence to conclude that in conglomerates have a higher Sharpe ratio in emerging markets than they do in developed markets.
𝐻𝑜: 𝛽3 = 0 vs alternate hypothesis 𝐻1: 𝛽3 < 0
p-value = 0.0639
2 = 0.03195 – reject 𝐻𝑜 at 5% significance level.
There is enough evidence at a 5% significance level to conclude that firms with more assets have a lower Sharpe ratio.
34
6 Conclusion and Limitations
6.1 Summary of Study and Conclusion
Large, diversified conglomerates dominate the corporate world in emerging markets.
Conglomerates formed as a result of the conglomerate merger wave of the late 60s, however shortly after dissolved in developed markets due to their inefficiencies. In emerging markets, however, conglomerates continued to thrive to the present day. Emerging markets are
characterized by several differences to developed markets such as political instability and stock market and exchange rate volatility that make them a different macroeconomic environment.
I developed a multiple regression model to test the effect on being a conglomerate, being a conglomerate in an emerging market and total assets on two financial success measures, the Jensen’s alpha and the Sharpe ratio. The two models were tested using OLS multiple regression analyses.
In the first model with the Jensen’s alpha as the dependent variable, we found moderate statistical evidence that conglomerates have a lower Jensen’s alpha than non-conglomerates (p-value 6.4%). We found no statistical evidence that the previous effect is “less negative” or positive whether the conglomerate is in an emerging market or developed market (p-value 43.6%). There is strong statistical evidence that firms with greater total assets have a lower Jensen’s alpha (p-value 3.2%). The overall model was not a good fit for the data (p-value 17.3%).
In the second model with the Sharpe ratio as the dependent variable, we found no statistical evidence that conglomerates have a lower Sharpe ratio than non-conglomerates (p-value 17.4%). We found no statistical evidence that the previous effect is “less negative” or positive whether the conglomerate is in an emerging market or developed market (p-value 45.8%). There is strong statistical evidence that firms with greater total assets have a lower Sharpe ratio (p-value 2.0%). The overall model was not a good fit for the data (p-value 28.1%).
This result contradicts the findings of Melicher & Rush (1973), who found that
conglomerates performed slightly better than non-conglomerates in the Jensen’s alpha measure. My study shows the opposite, in addition to showing the result does not change whether the conglomerates are based in emerging markets or developed markets. This could be due to the following reason: while Melicher & Rush used data from 1966 to 1971, I used more recent data
35
between 2008 and 2014. It may be the case that five decades ago during the height of the
conglomerate merger wave, investors had high hopes for conglomerates based on theories and no past data, hence inflating returns and financial performance of conglomerate assets. However, in the present day, investors are wiser and have past information that makes conglomerates less attractive than focused firms.
Similarly, the result concerning the effect of total assets on the Jensen’s alpha and the Sharpe ratio contradicts Johnson & Soenen’s findings (2003). While they found that firms with greater total assets had on average a higher Jensen’s alpha and Sharpe ratio, our regression analyses shows that firms with greater total assets had on average a lower Jensen’s alpha and Sharpe ratio. This interesting contradiction could be due to a similar reason as above: for the firms in their sample, Johnson & Soenen (2003) used data from 1982 to 1998. The data for the firms in our sample was from 2008 to 2014. It may be that between the years 1982 and 1998 investors saw increased total assets as a sign of power and financial success, heavily invested in these firms and inflated their returns, therefore increasing their Jensen’s alpha and Sharpe ratio. However between the years 2008 to 2014, investors were probably wiser and saw greater total assets as a burden and more difficult to efficiently manage and achieve high returns, therefore leading to a lower Jensen’s alpha and Sharpe ratio.
Because the fit of both models was not good and the 2 sided p-value for the variable EMER*CONG was so high, I ran a restricted regression analysis for each model without including this variable. The outputs of the restricted regression can be found in Appendix 5. In these restricted regression analyses, based on statistical evidence, the conclusions regarding the effect of the variable CONG and LNTOTASS are the same for both models. However, the fit of the model significantly improves for both the first model (p-value of F-test 8.2% from 17.3%, Adjusted R-Squared 5.2% from 3.5%) and the second model (p-value of F-test 14.7% from 28.1%, Adjusted R-Squared 3.2% from 1.5%). This further validates our conclusions regarding the conglomerate effect and the effect of total assets on the Jensen’s alpha and the Sharpe ratio of a firm.
In conclusion, based on the empirical findings of this study, the answer to the research question “are diversified conglomerates more financially successful in emerging markets than
they are in developed markets?” is no. If conglomerates are more dominant in emerging markets
36
socio-economic and cultural characteristics that separate emerging markets and developed markets are the reason for this trend. It may be the time to listen to the management consultants who advise conglomerates to divest their unrelated businesses, but it is important to remember the points made by Khanna & Palepu (1999). Divestment and focus must be encouraged, not forced. Otherwise tensions could arise as they did in South Korea between conglomerates and the government during the Asian Crisis, which is harmful to the already volatile capital markets of emerging countries.
6.2 Limitations and Opportunities for Future Research
There were several limitations in this study that could be improved to yield more accurate results. These limitations and improvements are important to consider because the signs of the
coefficients were all as I had predicted in the hypotheses, but they were not all statistically
significant. Improvements to the study could result in them being statistically significant and give more information on the performance of conglomerates in emerging markets.
A larger sample size would have led to more accurate results due to better approximation by the normal distribution and smaller standard errors.
There was only one measure of diversification, a binary variable CONG which was equal to a 1 if the firm is a diversified conglomerate and 0 if the firm is a focused firm. However, firms have different degrees of diversification, and this was not measured. While Koç Holding may be more diversified and in more industries than Astra International, for both observations 1 was recorded as the value for the variable CONG. The effect of family/government ownership and control was also shown in past empirical evidence (Yiu, Bruton & Lu, 2005) (Villalonga & Amit, 2006) to have an effect on conglomerate performance but not considered in our regression model.
This missing information could lead to omitted variable bias, which in turn leads to the violation of the first OLS assumption (that the independent variable and the error term must not be correlated). When the first OLS assumption is violated, the estimators for the coefficient of the slope of the independent variables of the population regression function are biased and
inconsistent, even for large sample sizes. Either measuring the degree of diversification and family/government control or using other econometric tools such as instrumental variable regression analysis to counter omitted variable bias will improve the accuracy of the results.
