Individualism, National Cultural Change and Momentum
Profits
Tim de Heij
University of Groningen
Faculty of Economics and Business
Supervisor dr. M.M. Kramer
Semester 1, 2013-2014
Abstract
This study examines the relationship between cultural differences and momentum profits using local currency denominated international stock market data. Cross-country cultural differences are quantified with the individualism index developed by Taras, Steel and Kirkman (2012), which is related to overconfidence and self-attribution bias. Significant cross-country differences in momentum profits exist and the momentum anomaly is highly persistent. The individualism measure is positively related to trading volume and momentum profits, however not to volatility. Furthermore, negative average momentum profits for Asian countries in the sample change to positive over the sample periods 1990-1999 to 2000-2012.
JEL classification: G020, G12, G15, Z13
1
1. Introduction
An important belief underlying the efficient market hypothesis (EMH) is that when predictable patterns in returns exist, they are quickly canceled out by investor seeking to exploit them. However, this assumption does not seem to hold for price based momentum strategies. Although documentation of momentum strategies dates back to the early 1990s, when Jagadeesh and Titman (1993) find that common stocks in the United States which produce the best (worst) returns over the past 3 to 12 months will maintain their good (bad) performance over the consecutive 3 to 12 months. Strategies based on the momentum anomaly continue to yield excess returns in the following years (Jagadeesh and Titman, 2001). The momentum effect is not limited to the U.S. market but found in most stock markets around the world, with the exception of a several Asian markets1. Despite the widespread and persistent evidence proving the existence of the momentum effect, no consensus has been reached among financial economists on the drivers of momentum profits nor the dispersion of momentum profits among international equity markets.
Previous research on the drivers of momentum profits has concentrated on the extent to which momentum profits are correlated with stock characteristics and attempts to provide behavioral explanations for the momentum anomaly. In a nutshell, the evidence indicates that momentum profits are caused by delayed reaction to firm-specific information. One explanation of this evidence is that investors tend to underreact to new firm-specific information. Another explanation is that the delayed reaction is an overreaction by investors, who either react to new information with a delay or who like to chase past winners.
Chui, Titman and Wei (2010) find that cross-country differences in momentum profits are substantial and persist over time. They propose that national cultural differences may be related to behavioral biases, and consequently, cross-country cultural differences may explain cross-country dispersion in the profitability of momentum
1
2 strategies. To measure cross-country differences in culture they use the individualism index developed by Hofstede (2001), hypothesizing that it is related to overconfidence and self-attribution biases. Their results show that individualism is positively correlated with momentum profits. Although Hofstede’s cultural dimensions have been used in numerous studies to successfully explain and predict a wide range of business and economics phenomena or processes, major critiques exist.2 In addition, many have assumed that cultures are stable over time. Hofstede viewed theories of cultural change as “naïve” (Hofstede, 2001, p. 34) and pointed out that cultural change would be slow. However, there is growing empirical evidence that cultural dimensions slowly change over time. This evidence suggests that change is certainly not rapid, but it appears to be happening faster than expected by Hofstede.3 Taras, Steel and Kirkman (2012) meta-analyze the relationship between Hofstede’s four cultural dimensions and several organizational outcomes to mitigate the critiques on the Hofstede data. In addition they conclude that, on a global level, national cultures change towards values typical to Western free market societies. For example, there seems to be a steady trend towards higher Individualism in general 4.
This study uses international stock market data and the individualism index developed by Taras, Steel and Kirkman (2012) to examine if the cross-country differences in momentum profits can indeed be explained by cultural differences in individualism. Subsequently, cross-country differences in individualism are hypothesized to explain why investors in one country are more likely to exhibit overconfidence and self-attribution bias. Considering that individualism does not directly measure the two behavioral biases, this study includes a correlation assessment for individualism with stock market trading volume and volatility. Research has shown that investors who are prone to exhibit overconfidence and self-attribution bias generate excess trading volume and volatility (Daniel et al. (1998), Glaser and Weber (2009)).
2 Examples can be found in human resource management (Ramamoorthy and Carroll, 1998), international trade and cooperation (Kogut and Singh, 1988), marketing (Yeniyurt and Townsend, 2003), accounting and audit (Yamamura, Frakes, Sanders, & Ahn, 1996), entrepreneurship (Mueller & Thomas, 2001) and more recently finance (Chui, Titman, and Wei, 2010).
3 See for an example Inglehart and Baker (2000), who provide evidence for substantial cultural change using a sample that comprises 75% of the world’s population.
3 While most research on cross-country differences in momentum profits use U.S. dollar denominated data, this study takes the perspective of a local investor and presents the results in local currencies. Hereby circumventing the limitation that the obtained results are biased by exchange rates. For example, converting U.S. dollar denominated momentum returns for an investor holding a euro denominated bank account possibly leads to different returns. An additional motivation is provided by Nitschka (2013), who argues that from the perspective of the U.S. investor, historical, relatively low foreign stock market returns are associated with currently low foreign currency excess returns and vice versa. This implies that past, high cumulative stock market returns, signal foreign currency appreciation. In other words, the study suggests that momentum in stock market returns contain information about exchange rate changes that are not captured by interest rate differentials5.
The analysis in part one and two of the results reveal that although trading volume is positively related to individualism, volatility is not. Indicating mixed proof that individualism is related to overconfidence and self-attribution bias. Part three shows that there are significant cross-country differences in momentum profits which are highly persistent. In addition, the momentum profits increase in magnitude for many countries between the sub-periods 1990-1999 and 2000-2012. Part four shows that cross-country differences in momentum profits can be explained by cultural differences in individualism. In addition, as cultures change and come to accept values traditionally ascribed to Western free market cultures so do the patterns in stock market returns. The global trend towards higher individualism appears to translate itself to an increase in overconfidence and self-attribution bias, exhibited by investors while making investment decisions, subsequently generating higher momentum profits.
The remainder of this paper is structured as follows. Section 2 reviews the literature which forms the motivation and basis for this study. In Section 3 the data employed in this study are described. Section 4 reports the methodology used and the descriptive
5
4 statistics. Section 5, part one and two present the results pertaining to individualism, trading volume and volatility. Part tree shows the results for international dispersion of momentum profits and part four the relationship of momentum profits with individualism. Section 6 provides the conclusion and discussion. Finally, Section 7 discusses limitations and suggestions for future research.
2. Literature review
This section presents the literature which forms the basis of the research. Discussed in turn are; market efficiency, the momentum effect, explanations for the momentum effect, how individualism and momentum effects are related through overconfidence as well as self-attribution bias, the individualism measures, the relationship of overconfidence and self-attribution bias with trading volume and volatility, to conclude with the hypotheses.
2.1 Market efficiency
5 notable of these trends is the momentum effect (Asness, Moskowitz and Pedersen, 2013). A popular term for the observation that stock returns exhibit substantial serial correlation. With the use of a momentum investment strategy systematical excess returns can be achieved employing only historical return data and thus challenges even the weak form of the EMH.
2.2 Momentum
6 After the initial documentation of the momentum effect much empirical research has been done on the topic. Rouwenhorst (1998), who uses the same methodology as Jegadeesh and Titman (1993), researches momentum strategies for twelve European countries in the period 1980 to 1995. He also finds momentum profits of about 1% per month. Griffin, Ji and Martin (2005) study momentum around the world for the period 1975 to 2000. They find that both price and earnings momentum strategies yield substantial abnormal profits. In addition, they conclude that foreign momentum strategies are much less correlated with U.S. momentum strategies when compared to market index correlation, meaning that momentum strategies benefit more from diversification than market index strategies. Most interestingly is their observation that momentum is present when markets rise and fall as well as when the economy is expanding and contracting. Moreover, although momentum profits are less volatile than the market or regional indexes, very negative returns are also documented. These negative returns are dubbed reversal in finance literature. Bhojraj and Swaminathan (2006) find proof for both momentum and reversal using a dataset comprising of 38 international stock market indexes. Winners perform better than loser for the first 3 to 12 months, In the two following years, loser outperform winners.
There are also important exceptions found in equity markets. Hameed and Kusnadi (2002) investigate Hong Kong, Malaysia, Singapore, South Korea, Taiwan and Thailand and document meager proof of momentum in these Asian markets. They conclude that the factors that spark the momentum phenomenon in the U.S. and Europe do not exist in Asian markets. Moreover, Chui, Titman and Wei (2003) find that Indonesia and South Korea exhibit negative momentum profits and find only slightly positive profits for Japan. On the contrary, Kang, Liu and Ni (2002), document substantial excess profits for momentum strategies in the Chinese market.
2.3 Explanations for the momentum effect.
7 the weak form of “market inefficiency," others have argued that the profits from these strategies are either compensation for risk, or alternatively, the product of data mining. First, although ruling out data mining as a possible explanation for the momentum effect is difficult, the robust and persistent empirical proof for the phenomenon points to its unimportance. Second, risk based explanations proposed by Conrad and Kaul (1998) argue that the excess returns of momentum strategies come from cross-sectional variation in expected returns and not due to any predictable time-series variations in stock returns. They find that stocks which have high (low) unconditional expected return rates in adjoining time periods are expected to have high (low) realized rates of returns in both periods which are also found by Lo and MacKinlay (1990); Jegadeesh and Titman (1993). To put it differently, momentum strategies have, on average, positive returns. Hence, profits of a momentum portfolio are on average positive in any holding period. To test this hypothesis Jagadeesh and Titman (2001) research the post-holding period of momentum portfolios and find that winner stocks underperform losers in the subsequent 13 to 60 months in the U.S. market. Bhojraj and Swaminathan (2006) find the same evidence for indexes around the world, where after a 3 to 12 month holding period, winners underperform losers for two years. In addition, rational momentum models6 will need to assume absurd levels of risk-aversion to explain the magnitude of momentum profits (about 12%) found in most equity markets. Moreover, Daniel (2011) concludes that the market risk of momentum portfolios varies dramatically, but does not appear to explain the variation in the premiums earned by momentum strategies.
With data mining ruled out and the failure of rational models to fit the momentum anomaly academic literature turned to behavioral models for a more suiting explanation. As mentioned before most of the models assume that the momentum effect is caused by the serial correlation of individual stock returns. Serial correlation however, can be caused by under-reaction or delayed overreaction. If the serial correlation is caused by under-reaction, expectations are positive abnormal returns during the holding period followed by normal returns in the following period. However, if the abnormal returns
8 are caused by delayed overreaction, then expectation are that the abnormal momentum returns in the holding period will be followed by negative returns given that the delayed overreaction must be subsequently reversed. Empirical research, discussed in the previous section, finds mixed proof for reversal in the long-term, pointing to evidence for both under-reaction and delayed overreaction.
The first explanation for under-reaction was proposed by Barberis, Shleifer and Vishny (1998). The conservatism bias, identified by Edwards (1968), suggests that investors tend to underweight new information when they update their prior investments. Resulting in slow price adjustments to new information, however once the information is completely incorporated in prices; predictability about future stock performance diminishes. Second, Grinblatt & Han (2005) provide evidence in line with the disposition effect, which suggests that loss-averse investors tend to hold on to their past losers and sell their past winners. Third, George & Hwang (2004) show that stocks perform well after hitting their 52-week highs, that provides evidence of anchoring.
9 traders. Informed investors acquire signals pertaining to future cash flows but neglect information captured by the price history. Technical traders base their investment decisions on a limited history of prices, but do not incorporate and firm specific information. Information used by the informed group of investors is reflected in prices with a delay, resulting in under-reaction which in turn generates momentum profits. The technical traders reasoning are based on past prices which pushes prices of past winners above their fundamental values. The short-term profitability eventually reverses as prices eventually return back to fundamental values in the longer term. Expectations are updated rationally by investors, conditional on their information sets; however return predictability is established because of the fact that each group uses only part of all information available to update their expectations.
10 Now that an overview of the most important behavioral explanations of the momentum effect is established, the remaining part of this paper will focus on the international differences in overconfidence and the self-attribution bias as potential drivers for the empirical divergence of momentum profits in equity markets. The previously discussed model of Daniel et al. (1998) will eventually be used to connect individualism with momentum profits. The following sections will discuss in turn; how individualism and momentum effects are related through overconfidence as well as self-attribution bias, the individualism measures plus their shortcomings and individualism related to trading volume and volatility.
2.4 Individualism and Momentum, linked by self-attribution bias and overconfidence
While it may be natural to be unsure of your knowledge in the case of general knowledge, studies have shown that people can also be quite overconfident in their fields of expertise. This has been shown for such occupations as market forecasters,
Behavioral biases Stock Price reactions Abnormal returns Momentum Under-reaction Conservatism bias Disposition effect Anchoring Informed investors Delayed-overreaction Technical investors Conservatism bias + represetativeness heuristic Self-attribution bias and overconfidence
11 investment bankers, business managers, lawyers, and medical professionals (Barber and Odean, 1999). Thus, overconfidence afflicts experts as well as amateurs. There is also evidence that the extent of overconfidence may be a function of demographics. Most reliable is the difference in the degree of overconfidence between men and women, with men tending to be more overconfident than women (Lundeberg, Fox and Punćcohaŕ, 1992). Moreover, Chui, Titman, and Wei (2010) find that cultural differences influence the returns of momentum strategies across international equity markets. The individualism index by Hofstede (2001) is employed to measure cross-country cultural differences. They make a case for the observation that individualism is related to overconfidence and self-attribution bias. It must be noted that individualism is not a perfect or direct measure for the two behavioral biases. Cultural research distinguishes societies based on an individualism versus collectivism spectrum. Individualism is related to the degree to which people in a country tend to have an independent rather than an interdependent self-construal, and the reverse is the case for collectivism (Hofstede (2001). According to Markus and Kitayama (1991, p.226) the independent self-construal is: “A conception of the self as an autonomous, independent person. In order to satisfy one self, others do not have to play an important role.” The interdependent self-construal is that people view themselves: “not as separate from the social context but as more connected and less differentiated from others.” (Markus and Kitayama, 1991, p.227).
12
Figure 2. The conceptual framework of the link between individualism and momentum
By reviewing results obtained from cross-cultural psychological experiments and surveys, Markus and Kitayama (1991) and Heine et al. (1999) document that people in individualistic cultures, such as the United Kingdom, have the tendency to view their abilities as above average. On the other hand, people in more collectivistic cultures, for example most Asian nations, do not share this self-image. Van den Steen (2004) and others propose that when people are overconfident with respect to their own abilities, they have the tendency to overrate the accuracy of their prospects, which is the cocept of overconfidence in the momentum model discussed by Daniel et al. (1998). In contrast, people from collectivist countries tend to have high self-monitoring (Church et al., 2006). High self-monitoring stems from cultural values centered on behaving appropriately and adapting to different social situations, meaning that people adjust their behavior to the expectations of the social situation by observing social cues (Biais et al. (2005). They find in the same research that self-monitoring mitigates the cognitive bias caused by overconfidence, which is studied by observing trading behavior in an artificial financial market characterized by asymmetric information. Overconfidence in cross-cultural psychology literature can refer to either overconfidence about general knowledge or overconfidence in comparison with an individual’s peers (Yates, Lee, and Shinotsuka, 1996). They conclude that individualism is undoubtedly associated with peer-comparison overconfidence. Consequently, overconfidence with respect to success
Momentum profits
High Low
Overconfidence and self-attribution bias
High Low
Self-construal
Independent self Interdependent self
Individualism spectrum
13 in relation to an individual’s peers induces investors to overemphasize the validity of their own information, called miscalibration (Van den Steen, 2004). Hence, the type of overconfidence in momentum literature is similar to research on individualism.
The link between individualism and self-attribution biases has been researched by Markus and Kitayama (1991) and Kagitcibasi (1997). Both emphasize, that in order to promote and maintain self-esteem, individuals in individualistic cultures exhibit pervasive overconfidence and self-attribution bias. Nurmi (1992, p. 70) documents: “This cross-cultural difference in self-attribution bias is typically explained by Western individualism and the collectivist orientation of Eastern cultures.” Moreover, when trying to find cultural universals among societies (Moghaddam, 1998, p.167) concludes: “Evidence suggests that the pattern of self-serving biases found in societies more supportive of independent selves, such as the United States, is not always found in societies in which interdependent selves receive stronger encouragement, such as Japan”.
Now that a clear empirical and theoretical link for individualism with both overconfidence and self-attribution bias is established the behavioral model by Daniel et al. (1998), that links overconfidence and self-attribution bias to momentum profits, will be shortly revisited. Find the conceptualization of the model in figure 3. For a more
detailed description of the model by Daniel et al. (1998) see page 9. Considering the
importance of adequately operationalizing the individualism concept in the subsequent empirical research the next two sections are devoted to a review of the individualism measures.
14
2.4 Hofstede’s Individualism measure
The individualism index is the results of a cross-country psychological survey in 1967-1973. The survey was conducted with the help of 88,000 IBM employees in 72 countries. Out of responses on 14 questions about the subjects attitudes toward work and private lives the individualism index was calculated. Hofstede’s cultural dimensions have been used in thousands of studies to successfully explain and predict a wide range of business and economics phenomena and processes. Examples can be found in human resource management (Ramamoorthy and Carroll, 1998), international trade and cooperation (Kogut and Singh, 1988), marketing (Yeniyurt and Townsend, 2003), accounting and audit (Yamamura, Frakes, Sanders, and Ahn, 1996), entrepreneurship (Mueller and Thomas, 2001) and more recently finance (Chui, Titman, and Wei, 2010). Despite the wide acceptance and use, major critiques on the relevance of Hofstede’s scores are the single organizational design, the small sample and the age of the data, which is from 1973.
2.5 Individualism, a meta-analysis
15 sample comprised working age people (mean, 33.4 years), in the middle or upper middle-class with both genders almost equally represented (male 47.4%) and fairly well educated (overall mean, 15.5 years of schooling). The demographics did not vary noticeably across countries. The outcomes of their meta-analysis point in the direction of cultural change. It appears cultures on average change to values typically ascribed to Western free market societies. Important for this study is the movement towards higher individualism. As can be seen in table 1 of the appendix, the values and relative rank positions have clearly changed for most of the countries. For instance, many Asian cultures show a substantial increase in individualism. On the contrary, typical individualistic societies as the United Kingdom and the United States move more towards the collectivistic pole. As a result, the index values of Hofstede (2001) have become more outdated with every passing year. In addition, for countries that have experienced a substantial change towards either individualism or collectivism the indices can be very inaccurate. The same fact hold for single survey studies on cross-country cultural differences. For the reasons stated above, this research will use the individualism scores of Taran, Steel and Kirkman (2012).
2.6 Individualism, Trading volume and Volatility
The behavioral biases discussed above are not solely related to the momentum effect in empirical literature. It is found that behavioral biases also generate excess trading volume and volatility in stock markets7. The relationship between trading volume and overconfidence is the following. Overconfident investors trade more, because they overestimate the precision of their information. Moreover, Odean (1998) argues that trading by overconfident investors leads to excess volatility. Previous theoretical and empirical studies also indicate that overconfidence together with the self-attribution bias generate excess trading volume and volatility (Daniel et al. (1998), Glaser and Weber (2009)).
The literature discussed above can be summarized in the three hypotheses shown below, which will be answered in the subsequent parts of the thesis.
16
3. Data description
3.1 Individualism data
Taras, Steel and Kirkman (2012) included studies that reported cultural values of their participants measured using models and methodology comparable with those devised by Hofstede in their meta-analysis. A common challenge in meta-analysis is the tradeoff between sample size and comparability. The comparability concerns allows for fewer studies to be included. Relaxing inclusion standards generates a larger sample size. A larger sample size increases the generalizability, reliability and allows the assessment of more target populations over a wider range of conditions at more precise degrees of time. To assure the optimal balance between sample size and method variance, Taras, Steel and Kirkman (2012) used a content validation approach in which multiple of the data coders drive instrument compatibility through comparative item analysis. This type of expert assessment is common for meta-analyses in general and for those solely focused on Hofstede’s framework (Oyserman, Coon and Kemmelmeier, 2002; Taras, Kirkman, and Steel, 2010). The main information taken from the studies comprised of the sample means along the four cultural dimensions, sample size, year, country and sample demographics. As in the IBM study, underrepresented countries are grouped into cultural regions following Hofstede’s original clustering scheme with some adjustments to account for changes in geopolitics of the 1990s (Taras, Steel and Kirkman, 2012). The following regions were formed: Africa (South Africa was analyzed separately), Arab countries, Caribbean, Central America, South America, Asian Republics of former USSR, Baltic Republics of former USSR, and Slavic Republics of former USSR. To provide a foundation for the analysis, Taras, Steel and
H1: Trading volume is higher in more individualistic countries. H2: Volatility is higher in more individualistic countries.
17 Kirkman (2012) converted all culture data into a common metric after which they used the procedures described by Hunter and Schmidt (2004) for d-scores to derive the meta-analytic indices for each country/region along the four cultural dimensions. To arrive at a country score estimate the mean values of the samples were used, which is mathematically similar to taking the average of individual responses in typical cross-cultural research. The country scores can be found in table 1 of the appendix. Although the conclusions are based on the meta-analysis individualism measure regressions with Hofstede (2001) scores are included as an extra comparison.
3.2 Country trading volume data
As in other research by Griffin, Nardari and Stulz (2007) and the Worldbank a country’s trading volume (TN) is measured as its trading volume by value divided by the market capitalization in the respective month, both denominated in U.S. dollars. This is the only data in the thesis measured in U.S. dollars. The data are extracted from DataStream. The data for these measures are available in U.S. dollars. In addition, an alternative measure for turnover is extracted from the Global Financial Development Database8 (GFDD), which is computed the same way as the ratio obtained from DataStream and also denominated in U.S. dollars. The GFDD database contains an extensive dataset of financial system characteristics for 203 economies.
3.3 Stock return data
Local currency monthly stock returns, including 37 countries for January 1990 to December 2012, are extracted from DataStream International. As mentioned in the introduction, local return data is used to circumvent the limitation that the obtained results are biased by exchange rates. For example, converting U.S. dollar denominated momentum returns for an investor holding a euro denominated bank account possibly leads to different returns. An additional motivation is provided by Nitschka (2013), who argues that from the perspective of the U.S. investor, past, relatively low foreign stock market returns are associated with currently low foreign currency excess returns and
8
18 vice versa. Therefore, returns extracted in U.S. dollar possibly include changes in the exchange rate which will subsequently bias results. Common stocks, both domestic and foreign, which are listed on the major stock exchange(s) in each country are included. The active stock and dead stocks for each country are extracted, the dead stocks to mitigate the possible survivorship bias in the sample. Due to differences in data availability, the starting date for every country varies. However for most countries there is data in 1990. In order to determine the past 6-month cumulative returns on individual stocks as well as to measure the returns on the momentum portfolios, stocks are also required to each have a return history of at least 8 months. Since a reasonable number of stocks to form momentum portfolios is needed, each country is required to have a minimum of 30 stocks that meet the stock selection criteria in any month during the sample period. In addition, every country momentum portfolio is required to have a return history of at least 5 years. The total return index is extracted from DataStream. Looking at equity s total return is usually considered a more accurate measure of performance. Specific for this research; since the formation and holding periods of the momentum portfolios number 6 months and not, for example, weeks the total return index is more appropriate measure of performance. From the total return index monthly stock returns are calculated using the following formula.
(2)
Similar to Bekaert, Harvey, and Lundblad (2007), this study uses the squared monthly returns to obtain the monthly volatility per stock. The average stock volatility in country j in month t ( ) is the average of the squared monthly returns on the stocks in month t in country j,
(3)
19 deviation of the return on the national stock market indexes by Bloomberg is collected from the GFDD for each country.
3.4 Control variables
Although there is little empirical evidence when it comes to determinants of cross-country trading volume previous research has included the political risk index (PR) from the International Country Risk Guide (ICRG) as a proxy for liquidity cost (Chui, Titman, and Wei, 2010). In addition, differences in information asymmetries between countries are controlled for by using an indicator for financial market development; the ratio of total private credit to GDP (Credit) which is extracted from the GFDD (Stulz and Williamson, 2003). For volatility, in addition to the Credit variable, this study controls for the standard deviation of GDP growth rates which positively relates to volatility (Du and Wei, 2004). Real GDP growth rates are collected from the United Nations Conference on Trade and Development (UNCTAD) statistics database. Also, Bekaert and Harvey (1997) find that the ratio of market capitalization to GDP (Mcap) is negatively related to volatility. This measure is extracted from the GFDD. Table 1 summarizes the various variables discussed above.
4. Methodology
4.1 Trading volume and individualism
In order to determine the relationship between trading volume and individualism the following equation is estimated,
(4)
in which all variables have the cross-sectional component j referring to the countries in our sample. The time series component t, is for individualism by decade, for Volatility and Turnover monthly, and for Political Risk and Credit annually.
4.2 Volatility and individualism
20 monthly stock volatility is regressed against individualism and the previously discussed control variables,
in which the cross-sectional component j refers to countries. GDP volatility only has a cross-sectional component. The subscript t refers for Volatility to months, for Credit and Mcap to years.
Table 1. Definitions of individualism, trading volume, volatility, momentum and control variables
Continues on the next page
Variables Abbreviation Type Period Description Source
Individualism
Hofstede IndHof Cross-section Single cross-section A higher score indicates a more individualistic society
Hofstede (2001)
Meta IndMeta Cross-section and
time series by decade
1990s and 2000s A higher score indicates a more individualistic society
Taras and Kirkman (2012)
Stock Market
Stock returns RI Cross-section and
monthly time series
1-1990 to 12-2012 Monthly returns are calculated from local currency using equation (2)
Datastream
Stock Market Volatility
Volatility DS Cross-section and monthly time series
1-1990 to 12-2012 Market Volatility in month t is calculated using equation 3 Bekaert, Harvey, and Lundblad (2007), Datastream Market turnover
Turnover DS Cross-section and monthly time series
1-1990 to 12-2012 A country’s trading volume is measured as its monthly turnover by value, divided by the market capitalization in the respective month in US dollar Griffin, Nardari and Stulz (2007), Datastream Momentum Momentum Profits
Momentum Cross-section and monthly time series
1-1990 to 12-2012 A zero cost investment strategy using overlapping portfolio’s with a formation and holding period of 6 months. Stocks are assigned to the winner or loser portfolio based on past cumulative returns.
Jagadeesh and Titman (1993)
Control
Political Risk Political Risk Cross-section and Annual time-series
1-1990 to 12-2008 The PR rating is calculated by assigning risk points to a pre-set group of factors, termed political risk components.
21
4.3 Momentum portfolio formation
The methodology for portfolio construction is based on the earlier mentioned procedure by Jegadeesh and Titman (1993). For each country the profitability of momentum strategies is calculated with a formation and holding period of 6 months. At the end of every month, the whole set of stocks per country are ranked in ascending order based on the formation period cumulative returns. The lowest one third of the stocks is grouped in the “Loser” (L) portfolio, the top one third in the “Winner” (W) Portfolio. Hereafter, these similarly weighted portfolios are held for 6 months. Due to the smaller sample sizes in some countries, the cut-off point of one third is used instead of the 10% by Jegadeesh and Titman (1993). Moreover, to mitigate the effect of the bid-ask bounce9 and the lead-lag effect10 there is a waiting period of one month between the formation
9 The bid-ask bounce is the bouncing of trade prices between the bid and ask sides of the market. It introduces a systematic bias to the data which can cause serious problems in analysis.
10
A lead–lag effect in economics describes the situation where one (leading) variable is correlated with the values of another (lagging) variable at later times. Lo and MacKinlay (1990) found that in some circumstances there is a lead-lag effect between large-capitalization and small-capitalization stock-portfolio prices.
Financial market development
Credit Cross-section and Annual time-series
1-1990 to 12-2011 Total amount of domestic private debt securities (amount outstanding) issued in domestic markets as a share of GDP. It covers data on long-term bonds and notes, commercial paper and other short-term notes.
Bank for International Settlements
Financial market Depth
Mcap Cross-section and
Annual time-series
1-1990 to 12-2011 Stock market capitalization to GDP (%) Standard & Poor's, Global Stock Markets Factbook and supplemental S&P data Financial Market efficiency
Turnover S&P Cross-section and Annual time-series
1-1990 to 12-2011 Stock market turnover ratio (value traded/capitalization) (%) Standard & Poor's, Global Stock Markets Factbook and supplemental S&P data Financial Market Stability
Volatility BB Cross-section and Annual time-series
1-1990 to 12-2011 Volatility of stock price index is the 360-day standard deviation of the return on the national stock market index.
Bloomberg
GDP growth
volatility GDP volatility Cross-section 1990-2012
22 and holding period. Overlapping portfolios are created to both avoid the waiting period and increase the power of the tests. Figure 1 below depicts the procedure employed.
Figure 1. Momentum portfolio formation strategy (6,6)
At the end of every month a new portfolio is created. Both the portfolios and the
underlying stocks are equally weighted instead of value weighted to increase the weight of smaller stocks. The psychological biases discussed previously are more important for individual domestic investors who predominantly trade smaller stocks (Chui, Titman and Wei, 2010). The weight of a portfolio is 1/H due to overlapping holding periods. If a stock has a return of zero in the formation period it is excluded from the sample to ensure the results are not biased by small or illiquid stocks. The return of a zero cost momentum strategy is then measured at the end of the month by subtracting the
weighted loser portfolio returns from the weighted winner portfolio returns (W-L). The momentum data presented in the analysis section starts in September 1990. Although the first 5 months of the sample have less than six overlapping portfolios, they are included in the sample.
4.4 Descriptive Statistics
Table 2 shows the descriptive statistics for all variables used in this study. To reduce the non-normality of the dependent variables, they are transformed by natural logarithm. In addition, to ensure that positive numbers are not changed to negative ones, all observations have one added before taking the natural logarithm. The IndMeta measure is transformed by adding 2 to each observation, which leads to a better fit with the other
23 regression data, in particular the other individualism measure IndHof. Volatility DS is created by multiplying the volatility measure calculated from the DataStream data by 100, making the values more comparable with the remainder of the data.
Table 2. Descriptive statistics for Turnover, Volatility, Individualism, control variables and momentum data for the period 2-1990 to 12-2012
Descriptives Mean Median Maximum Minimum Std. Dev. Observations Variables Turnover DS 51.429 41.283 1136.708 0.001 42.868 9725 Turnover S&P 76.707 61.496 377.242 1.559 54.482 9096 Volatility DS 4.611 2.476 777.367 0.108 15.973 10143 Volatility S&P 26.731 23.398 151.130 8.977 13.506 8124 IndHof 0.893 0.920 1.700 0.020 0.420 10212 IndMeta 2.154 2.100 3.790 0.660 0.672 9816 Credit 28.384 23.307 193.407 0.007 27.028 8436 Political Risk 77.926 80.440 99.164 36.400 11.722 8196 Mcap 73.651 52.374 569.462 0.179 70.289 9108 GDP volatility 2.860 2.490 5.810 1.324 1.105 10212 Return loser -0.480 -0.394 69.571 -92.691 8.371 9897 Return winner 0.345 0.624 73.268 -63.040 6.956 9897 Momentum 0.826 0.916 117.718 -57.585 4.752 9897 Transformed Turnover DS (ln) 3.628 3.744 7.037 0.001 0.918 9725 Turnover S&P (ln) 4.098 4.119 5.933 0.444 0.738 9096 Volatility DS (ln) 1.342 1.246 6.657 0.103 0.680 10143 Volatility S&P (ln) 3.188 3.153 5.018 2.195 0.427 8124
The number of observations per variable differs due to the time period covered. In addition, some countries in the sample have missing values for one or more variables because data in the specific time period was not gathered, usually in the first years of the sample.
24 one would expect. One reason could be the difference in time interval, the DataStream variable is measured monthly whereas the S&P variable on a yearly basis. The same holds for the two volatility measures. In addition, the volatility measure based on the DataStream data is calculated after the screening process described in the previous section, whereas the Bloomberg measure represents index volatility. Both volatility measures are positively correlated with both turnover measures, however, only statistically significant for their origin turnover counterparts. The IndMeta variable is positively correlated with both turnover variables as one would expect but negatively correlated to both volatility measures. IndHof is positively correlated to the turnover and volatility variables based on the DataStream data yet negatively correlated to the turnover S&P and Bloomberg volatility variables. Quite remarkably the correlation between IndHof and IndMeta is very low (0.060). Regardless of the difference in time interval a higher correlation would be more plausible. The Hofstede measure, as discussed before, seems outdated, exhibiting almost no correlation with the recently published meta-individualism measure of Taras, Steel and Kirkman (2012).
Table 3. Pair wise correlation matrix
Pair wise correlation statistics with p-values shown in the parentheses for the period 2-1990 to 12-2008.
25
5. Results
In the first and second part present the results of the OLS estimations for the relationship between trading volume, volatility and individualism. The third part shows the average monthly returns of the country momentum portfolio strategies. Part four is devoted to the analysis between momentum profits and individualism. The regression method in part one and two is pooled OLS with corrected standard errors for cross-sectional (country) correlation and heteroskedasticity using White’s standard errors.11 This model does not distinguish between the various countries in the sample, i.e. it does not allow for any heterogeneity or individuality in the cross-section. In the subsequent tables, column 1 and 2 present results based on the trading volume and volatility data obtained from DataStream, column 3 and 4 present the results using the S&P trading volume and Bloomberg volatility data. Column 1 will be leading in the analysis since it uses the meta-individualism measure of Taras, Steel and Kirkman (2012) combined with the DataStream data which are more detailed than the S&P and Bloomberg data. Column 2 serves as an extra illustration using the individualism measure by Hofstede used in previous research on the topic by Chui, Titman and Wei (2010).
5.1 Trading Volume and individualism
Table 4 reports the regression results for trading volume regressed on individualism and the control variables. Because there is data for political risk until the end of 2008, the regressions are estimated from 2-1990 to 12-2008. The number of cross-sectional entities and number of observations differs per regression due to missing values for some country variables in different months. For column 1 and 2 both individualism measures are positively related to trading volume and statistically significant at the 1% level (β=0.381, β=0.163), supporting hypothesis 1 and consistent with previous research by Chui, Titman and Wei (2010). Surprisingly, in column 4 the Hofstede measure changes to a negative sign while the meta-analysis measure of individualism in column 3 remains positive and significant. The meta-individualism measure is more detailed and is constructed because the Hofstede measure has been losing its precision with
11
26 every year and, in some cases, can be presently misleading. Therefore, the conclusion based on the trading volume analysis is that the meta-analysis individualism measure serves as a proxy for overconfidence and self-attribution bias.
The estimated coefficient for Credit, which captures the differences in asymmetric information i.e. financial market development, is not statistically significant. The coefficients for political risk and volatility are positive and statistically significant. The positive sign for political risk indicates that countries with lower liquidity cost tend to have higher trading volume, which is in line with expectations. The positive and significant sign for volatility indicates that trading volume is affected by the uncertainty
Table 4. Results for individualism and stock market trading volume
Presented below are the OLS estimates of the coefficients related to market trading volume. Panel 1 and 2 presents equation 4 with the market trading volume of country j in month t measured as the market dollar trading volume divided by the index’s market capitalization in month t obtained from the DataStream Global Index. Panel 3 and 4 report regression 4 using trading volume data from the Standard & Poor's, Global Stock Markets Factbook for country j and year t which is computed similarly. The natural logarithm of market trading volume is regressed on the meta individualism index (IndMeta) or Hofstede’s individualism index (IndHof), total private credit as a percentage of GDP (Credit), the political risk index of the ICRG (Political Risk) and the natural logarithm of market volatility (Volatility DS) from Datastream or the 360-day standard deviation of the return on the national stock market index for each country collected from Bloomberg (Volatility BB). All variables are updated monthly or annually, for a more detailed description of the variables see table 1. The sample period is 2-1990 to 12-2008. I use White cross-section standard errors to compute robust standard errors. P-values are shown in parenthesis under the coefficients.
Turnover DS (ln) Turnover S&P (ln)
27 of information flows. More uncertainty of information flows thus leads to higher trading volume, which is consistent with theory.
5.2 Volatility and Individualism
Table 5 summarizes the results for the OLS estimations of stock market volatility regressed on individualism and control variables. Column 1 and 3 show that, at the 1% level, the meta-individualism measure is neither statistically significant related to stock market volatility calculated from individual stocks nor index volatility. For column 2, consistent with the findings of Chui, Titman and Wei (2010), the coefficient of the Hofstede’s measure for individualism is positive and significantly related to stock market volatility. However, for index volatility the sign changes to negative. All regression signs are in line with those reported in correlation table 3. Moreover, regressing the non-logarithmic (non-transformed) volatility measure on the independent variables yields similar coefficient signs with little variation in statistical significance. Based on the preceding analysis, this study finds no proof for hypothesis 2 using the individualism measure by Taras, Steel and Kirkman (2012). The meta-individualism measure appears not to be a proxy for overconfidence and self-attribution bias based on the relationship with volatility. However, when employing the outdated measure of Hofstede (2001), volatility is positively related to individualism, which is consistent with the findings of Chui, Titman and Wei (2010).
28
5.2.1 Robustness
Omitted variables can bias OLS estimation results. These can be controlled for using a fixed-effects model with country-fixed effects and time-fixed effects. This model controls for all time-invariant differences between the countries, so the estimated coefficients of the fixed-effects models are not due to omitted time-invariant characteristics. Because the most important independent variable, individualism has near similar values for each time-series observation per cross-section, equation 4 and 5
Table 5. Results for individualism and stock market volatility
This table presents the OLS estimates of the coefficients related to stock market volatility. Panel 1 and 2 present equation 5 with the average monthly squared stock returns of country j in month t computed from the return index of Datastream (Volatility DS). Panel 3 and 4 show equation 5 using trading volume data for the 360-day standard deviation of the return on the national stock market index for each country j and annual t collected from Bloomberg (Volatility BB). The natural logarithm of volatility is regressed on the meta individualism index (IndMeta) or Hofstede’s individualism index (IndHof), total private credit as a percentage of GDP (Credit), the volatility of per capita GDP growth rates (GDP volatility and the market value expressed as a ratio of GDP. The sample period is 2-1990 to 12-2011. I use White cross-section standard errors to compute robust standard errors. P-values are shown in parenthesis under the coefficients.
29 are estimated only with period-fixed effects. The preceding conclusions do not change based on this model specification. To assess the need for random-effects specification Hausman tests are applied to and rejected at a 1% significance level for all regressions.
5.3 Returns of Country Momentum Portfolios
This section presents the profitability of momentum strategies per country. Table 6 presents the momentum profits for 37 countries, in the period 1990-1999, 2000 to 2012 and for the whole sample period. Profits are depicted in monthly average returns (%) of the momentum portfolio (W-L)12. Also included in the table are the individualism indexes, Hofstede as a single observation and the meta-analysis per period similar to that of the momentum profits. The results show that both the average individualism score and average momentum profits of the countries in our sample have increased. The average monthly momentum profits for the whole sample amount to 0.826%, which corresponds to about 10% per year. Previous research for Europe (Rouwenhorst, 1998) and the U.S. (Jagadeesh and Titman, 1993, 2001) document momentum profits of about 12% per year. For Asia much lower or negative momentum profits are found (Hameed and Kusnadi (2002); Chui, Titman and Wei, (2003)). Since our sample comprises of countries from all continents 10% appears very plausible. Furthermore, average monthly momentum profits increase from 0.622% to 0.971% between the sub-periods 1990-1999 and 2000-2012. This corresponds to an increase in annual returns from about 8% to approximately 12%. In all but two countries, pursuing a momentum strategy yields positive returns. Brazil (-0.340%) and Turkey (-0.493) show negative momentum profits over the whole sample period. The largest momentum profits are found in Sweden (1.873%), United Kingdom (1.750%), The Netherlands (1.743%), South Africa (1.615%) and Denmark (1.597%). All these countries have a relatively high score on the individualism index indicating preliminary proof for hypothesis three. In the sub-period 1990-1999 there are six countries with negative momentum profits; Brazil (-0.984%), Israel (-0.641), Indonesia (-0.170%), Malaysia (-0.029), South Korea (-0.070%) and Turkey (-0.381%). Moreover, Japan (0.039%) barely shows positive momentum profits. These finding are
30
Table 6. Country Average Momentum Profits and Individualism
For each country the profitability of momentum strategies is calculated with a formation and holding period of 6 months. At the end of every month, the whole set of stocks per country are ranked in ascending order based on the formation period cumulative returns. The lowest one third of the stocks is grouped in the “Loser” (L) portfolio, the top one third in the “Winner” (W) Portfolio. Hereafter, these similarly weighted portfolios are held for 6 months. To increase the power of the tests, overlapping portfolios are constructed. The winner (loser) portfolio is an overlapping portfolio that consists of “W” (“L”) portfolios in the previous 6 ranking months. Returns on these portfolios are measured 1 month after ranking. Returns on these portfolios in month t are computed as (average cumulative returns of the stocks in these portfolios in month t divided by average cumulative returns of these stocks in month (t − 1) − 1). Returns on the winner and loser portfolios are the simple average of the returns on the six “W” and the six “L” portfolios, respectively. The momentum portfolio (W – L) is a zero-cost, winner-minus-loser portfolio. Presented below are the average monthly returns (%) of the momentum (W – L) portfolios per country in local currency for the period 1990-1999, 2000-2012 and total sample. The country-average portfolio is an equally weighted portfolio. Corresponding t-statistics are in parentheses. In addition the individualism scores of Hofstede (2001) and meta-individualism scores of Taras, Steel and Kirkman (2012) are presented.
Individualism Momentum (W-L)*
Country N= 37 Hofstede Meta-analysis
1990-1999 2000-2012 1990-1999 2000-2012 Total Argentina 0.75 -0.71 -0.12 0.905 (0.98) 0.314 (0.93) 0.561 (1.29) Australia 1.70 0.93 0.98 0.583 (2.48) 1.286 (3.90) 0.992 (4.59) Austria 0.27 0.902 (3.31) 1.107 (3.92) 1.021 (5.12) Belgium 1.09 0.69 0.67 0.788 (3.86) 0.775 (2.41) 0.781 (3.79) Brazil 0.75 -0.71 -0.12 -0.984 (-1.71) 0.123 (0.52) -0.340 (-1.50) Canada 1.29 0.37 0.78 1.173 (4.25) 1.316 (4.07) 1.256 (5.70) Chile 0.75 -0.71 -0.12 0.616 (2.02) 0.811 (3.81) 0.729 (4.11) China 1.17 -0.19 0.02 0.532 (0.96) 0.142 (0.56) 0.298 (1.11) Denmark 1.05 0.47 0.77 0.981 (2.52) 2.039 (4.45) 1.597 (5.09) Finland 0.60 -0.04 0.933 (2.55) 1.007 (3.29) 0.976 (4.17) France 0.92 0.18 0.60 0.640 (1.26) 1.604 (3.44) 1.201 (3.48) Germany 0.76 0.31 0.40 0.769 (3.99) 1.604 (4.55) 1.255 (5.66) Greece 0.55 -0.73 -0.08 0.863 (2.22) 0.449 (0.94) 0.622 (1.93) Hong Kong 0.96 -0.12 0.09 0.418 (1.06) 0.678 (2.17) 0.569 (2.32) India 0.02 -0.42 -0.28 0.296 (0.88) 0.083 (0.22) 0.172 (0.65) Indonesia 1.41 0.01 -0.170 (-0.31) 0.398 (1.20) 0.160 (0.53) Ireland 0.88 0.35 0.42 1.267 (2.91) 1.577 (3.92) 1.447 (4.89) Israel 0.23 0.98 0.89 -0.641 (-2.03) 0.838 (3.35) 0.220 (1.09) Italy 1.13 -0.13 0.32 0.534 (2.00) 1.289 (4.15) 0.974 (4.57) Japan 0.10 -0.14 0.11 0.039 (0.11) 0.387 (1.68) 0.241 (1.19) Malaysia 0.92 -0.95 -0.93 -0.029 (-0.06) 0.630 (2.24) 0.355 (1.34) Mexico 1.51 -1.26 -1.03 1.437 (3.03) 0.795 (2.72) 1.063 (4.07) Netherlands 1.29 0.94 1.07 1.523 (5.41) 1.901 (5.23) 1.743 (7.20) New Zealand 1.25 1.37 1.03 0.565 (1.55) 1.910 (5.28) 1.348 (5.13) Norway 0.84 0.65 0.95 1.178 (3.50) 1.665 (6.58) 1.462 (7.17) Poland 0.47 -0.37 -0.24 0.948 (0.82) 1.494 (4.49) 1.280 (2.57) Portugal 0.88 -0.38 -0.22 0.130 (0.35) 0.381 (0.93) 0.276 (0.97) Singapore 1.17 -0.97 0.19 1.582 (4.82) 1.639 (8.37) 1.615 (9.07) South Africa 0.68 0.37 0.83 0.525 (1.08) 0.855 (2.46) 0.717 (2.51) South Korea 1.25 -0.37 0.45 -0.070 (-0.10) 0.821 (2.23) 0.448 (1.27) Spain 0.1 0.51 0.83 0.667 (1.82) 0.668 (2.13) 0.668 (2.81) Sweden 0.92 1.54 1.79 1.679 (3.78) 2.013 (4.89) 1.873 (6.19) Switzerland 0.8 0.93 0.737 (3.48) 1.100 (3.41) 0.948 (4.57) Taiwan 1.29 -0.46 -0.47 0.235 (0.57) 0.365 (1.14) 0.311 (1.23) Thailand 1.17 -1.34 -0.73 0.445 (0.83) 0.457 (1.48) 0.452 (1.57) Turkey 0.47 -0.30 -0.45 -0.381 (-0.79) -0.573 (-2.32) -0.493 (-2.00) United Kingdom 1.66 0.82 0.33 1.442 (7.50) 1.971 (7.06) 1.750 (9.63) Average 0.89 0.01 0.28 0.622 (7.42) 0.971 (17.68) 0.826 (17.58)
31 in line with what previous research has documented for similar time periods (e.g., Chui, Titman and Wei (2003), Hameed and Kusnadi (2002)). In subsample 2000-2012, momentum profits have substantially increased in most countries. Only Turkey (-0.573%) still exhibits negative momentum profits. The highest momentum profits in for 2000-2012 are found in Sweden (2.013%), Denmark (2.039%), the United Kingdom (1.971%), Norway (1.910%) and The Netherlands (1.901%). The four Asian countries that exhibited negative momentum profits in the 1990-1999 period do not appear to do so in the 2000-2012 period. To check if the cross-country differences in momentum profits are persistent the Spearman rank correlation is computed between the country average momentum profits in the two sub-samples. The Spearman rank correlation is
0.690 (p-value = 0.00). Figure 4 presents a more graphical picture of table 6. The country average momentum profits and the meta-individualism scores per country are presented in a scatterplot for both sub-periods. Both lines corresponding to a sub-period. These fitted lines both indicate a positive relationship between a country’s average momentum profits and individualism. In addition, the accompanying equations show that the intercept, slope and the goodness of fit of the linear relationship increases from 1990-1999 to 2000-2012. The intercept, indicating higher momentum profits on
y = 0.2037x + 0.6108 R² = 0.0551 y = 0.6606x + 0.7948 R² = 0.4002 -2 -1 0 1 2 3 -2 -1 0 1 2 Individualism Momentum
32 average, the slope a stronger relationship between momentum profits and individualism, and the goodness of fit a significant better fit of the line with the data points. In the following section the relationship between momentum profits and individualism is further analyzed.
5.4 Portfolio Analysis of Momentum Profits and Individualism
This section explores the relationship between individualism and the profitability of momentum strategies. Countries are grouped based on their individualism score. The monthly average momentum portfolios are divided into four quantiles depending on the meta-individualism score by Taras, Steel and Kirkman (2012). The quantiles differ per period due to different individualism scores of the underlying countries. Table 7 reports the monthly (%) returns of these portfolios and their differences for three periods, the whole sample period 1990-2012, and two sub-periods; 1990-1999 and 2000-2012. The results reveal that in all periods except one, momentum profits continually increase with the score of the individualism index. This exception is for the period 1990-1999 where momentum profits for the most collectivistic group of countries are higher than the second quantile, however lower than the more individualistic quantiles; three and four. In addition, it is the only value in all the momentum quantiles that is not highly significant.
33 To summarize, the preceding results point to a highly significant relationship between individualism and momentum profits, thus supporting hypothesis three.
Although this study finds comparable momentum profits to what previous research documents (Chui, Titman and Wei, 2010). The returns on both the Winner and Loser portfolios are substantially lower. A reason for this can be that this research does not
Table 7. Momentum Profits and Individualism
This table presents monthly momentum profits (%) for country average portfolios computed from local currency returns. The country average portfolio equally weighs every country specific momentum (W-L) portfolio in this portfolio. See table 4 for a more detailed explanation about the formation of the momentum (W-L) portfolios. The countries monthly average portfolios are divided into four quantiles depending on the meta-individualism score by Taras, Steel and Kirkman (2012). The table reports three periods, the whole sample period 1990-2012, and two split samples; 1990-1999 and 2000-2012. The quantiles differ per period due to different individualism scores of the countries in the sample periods. Corresponding t-statistics are in the parentheses
Individualism Quantile Observations Winner (W) Loser (L) W minus L 1990-2012 Collectivistic 1 < - 0.3 2237 0.635 0.258 0.377 (3.23) (1.13) (2.75) 2 [-0.30, 0.10) 2492 0.367 -0.085 0.452 (2.28) (-0.46) (5.22) 3 [0.10, 0.69) 2276 0.089 -1.028 1.116 (0.85) (-7.34) (12.73) Individualistic 4 ≥ 0.69 2512 0.301 -1.017 1.318 (3.03) (-7.75) (17.03) All 9517 0.346 -0.476 0.822 (4.79) (-5.47) (16.69)
Individualistic minus collectivistic -0.334 -1.275 0.941
(-1.52) (-4.85) (5.98) 1990-1999 Collectivistic 1 < -0.46 896 1.232 0.760 0.472 (3.73) (1.93) (1.75) 2 [-0.46, -0.12) 989 0.517 0.275 0.241 (1.48) (0.75) (1.31) 3 [-0.12, 0.51) 1008 0.382 -0.461 0.844 (2.12) (-2.07) (6.36) Individualistic 4 ≥ 0.51 1008 0.583 -0.317 0.900 (3.42) (-1.60) (8.38) All 3901 0.663 0.043 0.620 (4.98) (0.29) (6.94)
Individualistic minus collectivistic -0.649 -1.077 0.428
(-1.75) (-2.44) (1.48) 2000-2012 Collectivistic 1 < -0.12 1260 0.278 -0.170 0.448 (1.36) (-0.69) (3.85) 2 [-0.12, 0.19) 1416 0.210 -0.284 0.495 (1.18) (-1.32) (4.75) 3 [0.19, 0.81) 1536 -0.118 -1.508 1.390 (-0.91) (-7.98) (11.97) Individualistic 4 ≥ 0.81 1404 0.170 -1.256 1.427 (1.32) (-7.11) (13.62) All 5616 0.126 -0.836 0.962 (1.57) (-8.09) (17.27)
Individualistic minus collectivistic -0.107 -1.086 0.979
34 employ a stock screening process based on market capitalization like Chui, Titman and Wei (2010). Another explanation could be that it is due to the fact that returns in this research are extracted in local currencies and not U.S. dollars. Also, the Winner portfolios in the most collectivistic quantile appear to generate higher average monthly returns (%) than all other quantiles in the three sample periods. The trend observed is that Winner portfolio returns decrease until the most individualistic quantile, in which the returns increase. The same trend is observed for the loser portfolios, where also the most collectivistic quantile of countries has the highest monthly average return. In addition, the dispersion of momentum profits between individualistic and collectivistic countries appears to be more driven by differences in the returns of the Loser portfolios than Winner Portfolios. Because these findings are not part of the main structure of the thesis and are possibly caused by limitation of study they will be discussed in the appropriate place, Section 7.
6. Discussion and conclusions
This study examines the relationship between cultural differences and the dispersion of momentum profits across countries. With the use of a new measure for individualism by Taras, Steel and Kirkman (2012) and a sample using local currency return data over more than 21 years including 37 countries.
Similar to the findings of Chui, Titman and Wei (2010), the first and second part of the research show that the individualism index of Hofstede (2001) is positively related to both trading volume and volatility. However, the individualism measure used in this study is only positively related to trading volume and not to volatility. These results show mixed proof for the notion that individualism is indeed related to overconfidence and self-attribution bias. A possible explanation is that the individualism index is related to other country specific attributes which correlate differently with volatility.
35 2001) document 12% per year. For Asia much lower or negative momentum profits are found (Hameed and Kusnadi (2002); Chui, Titman and Wei, (2003)). Furthermore, average monthly momentum profits increase from about 8% per year to approximately 12% between the sub-periods 1990-1999 and 2000-2012. Moreover, over the whole sample period all but two countries exhibit positive momentum profits. Only in Turkey the use of a momentum investment strategy yields negative profits in both sub-periods. The negative momentum profits found in several countries, most notably those in the Asian continent, for the first sub-period, disappear in the next. It appears Asian cultures have become more prone to exhibit overconfidence and self-attribution bias. Socio-economic research by Acker and Duck (2007) supports this observation. They conclude, using a stock-market game and predictions of examination marks that Asians are consistently more overconfident than British while all are equally prone to self-attribution bias. The older studies on cross-cultural variation in the self-self-attribution bias summarized by Nurmi (1992, p. 70) that explain differences in the self-attribution bias along the “Western” versus “Eastern” dimension do not appear valid anymore.
Part four presents the portfolio analysis to assess the relationship between cross-country momentum profits and individualism. The average monthly returns on a zero-cost (long minus short) momentum portfolio are 0.822% higher for countries ranked in the most individualistic quantile compared with the most collectivistic quantile. The difference in returns is robust for both sub-periods and very significant. The evidence supports the idea that culture can have a substantial effect on stock return patterns. Investors in more individualistic countries appear to interpret information differently. One interpretation of the results comes from the earlier discussed model of Daniel et al. (1998). It seems that in collectivistic cultures investors put minor weight on their own information and major weight on the general agreement of their peers. More specifically, investors in collectivistic cultures are less prone to exhibit overconfidence and self-attribution bias while making investment decisions. This manifests itself in lower momentum profits.
36 countries. As cultures change and come to accept values traditionally ascribed to Western free market cultures so do the patterns in stock market returns. The global trend towards higher individualism appears to translate itself to an increase in overconfidence and self-attribution bias exhibited by investors while making investment decisions, subsequently generating higher momentum profits.
