Corporate short-termism: An examination of firm and investor national culture

national culture

MSc Thesis for Finance and Accountancy

dr. R.B.H. Hooghiemstra Prepared by: Jos Offerein (s2222647)

Courses: Master’s Thesis Finance (EBM866B20 ) Master’s Thesis Accountancy (EBM869B20 ) Semester: Semester II

June 8, 2017

Abstract

Introduction

National culture has previously been related to firm decision making and thus a substantial number of firm outcomes. Among them are financing structure decisions (Aggarwal, Kear-ney and Lucey, 2012; Zheng et al., 2012), investment decisions (Shao, Kwok and Zhang, 2013), CEO compensation (Tosi and Greckhamer, 2004; Bryan, Nash and Patel, 2015), dividend policy (Shao, Kwok and Guedhami, 2010) and earnings management (Doupnik, 2008; Han et al., 2010). However, one quite general firm outcome has not yet been exam-ined directly: corporate short-termism. This topic has recently (re)gaexam-ined interest among researchers and policymakers. As an example, in the most recent review of the Dutch cor-porate governance code, emphasis was put on improving the long-term orientation of firms. Short-termism refers to an excessive focus on the short-term, which harms the long-term interests of the firm. For example, when managers cut R&D or advertising expenses to gen-erate short-term gains (Sounder et al., 2016; Flammer and Bansal, 2017), or more general, any activity that does not contribute to the long-run value of the firm but merely aims to increase profits in the near future. Firms and specifically their managers, face a vari-ety of incentives and pressures that might lead to such a short-term orientation, although managers themselves prefer to focus on the long-term, according to Bayer, Larcker and Tayan (2014). This study aims to include national culture as a potential determinant of short-termism, alongside previously identified influences like analyst coverage, shareholder time orientation and executive compensation packages, which shape the time orientation of firms. Moreover, this study adopts the view that culture shapes the relationships within a firm, not only differences between firms, in line with Frijns, Dodd and Cimerova, (2016). Specifically, the previously unexamined influence of national culture on the behaviour of shareholders in a firm is taken into account, by verifying whether cultural diversity within the shareholder base contributes to shaping the time orientation of the firm.

This study uses textual analysis to extract a measure of short-termism from quarterly conference calls of the largest 500 firms in Europe. This approach is based on Brochet, Loumioti and Serafeim (2015), whose study is extended by including firm and investor national culture as potential determinants of short-termism. Apart from this, the study includes firm-level determinants and country-level determinants that are either identified in the literature on short-termism, or in the literature on national culture. The results show that previously documented influences persist in Europe as well, although the effects are less robust than those of Brochet et al. (2015). As expected, evidence is found that the three dimensions of Hofstede (2001) that are considered do affect the time orientation that firms express in their conference calls. Especially the long-term orientation dimension is robust across a number of specifications, decreasing the short-term orientation of the firm. The case for the individualism dimension is weaker, but the effect is also as expected; individualism decreases short-termism. Uncertainty avoidance affects short-termism positively, although this effect disappears in alternative econometric models. Furthermore, the cultural diversity index that measures differences between the firm and the shareholder’s national culture does affect short-termism. This measure significantly decreases short-termism, lending support to the view that investors prefer to mingle in firm decision making when they better understand the agency relationships in the cultural environment of the firm.

results to a non-US environment, documenting evidence that the most important drivers of short-termism are also present for European firms. Third, this paper continues the work on identifying economic outcomes that are affected by national culture (Guise, Sapienza and Zingales, 2006; Shao et al., 2013; Hooghiemstra, Hermes and Emanuels, 2015). It adds to this list the implicit time orientation of firms, and offers robust evidence for at least one dimension identified by Hofstede (2001): long-term orientation in a society. Fourth, it extends research on national culture specifically by examining how cultural background shapes the decision making of investors, which has been lacking in the literature thus far. Finally, in examining investor national culture, the literature on cultural diversity is also extended, by building on research initiated by Frijns et al. (2016) on cultural diversity within the firm, instead of between firms or countries (Kogut and Singh, 1988; Beugelsdijk and Frijns, 2010).

The paper proceeds as follows. First, the theoretical background is discussed, along with prior literature on the subject. From this literature review, several hypotheses are postulated. The next section describes how data is collected in order to test these hy-potheses and subsequently the methodology of measuring and extracting measures of the concepts is described. In particular, the approach employed to measure short-termism is discussed extensively. Then, results are presented by means of some descriptive statistics and subsequent regression analysis. Before proceeding to a discussion and conclusion, some robustness checks and additional analyses are presented.

Theory and Hypotheses

Recent research by Dimitratos et al. (2011) established that national culture has significant effects on strategic decision-making procedures in firms. Arguably, the time orientation of firms is the outcome of several related strategic decisions, largely made by the management of the firm, according to the upper-echelon theory (Hambrick and Mason, 1984). In turn, these managers are affected by their cultural background as well (Byrne and Bradley, 2007; Brochet et al., 2016). Thus decision-making within firms is influenced by national culture, and therefore the outcomes of firms are necessarily also influenced. Generally, another part is important in the way modern firms are structured, especially in theories of the firm, such as agency theory (Jensen and Meckling, 1976). This part, naturally, are the shareholders of the firm. Past research has shown that ownership structures or shareholder characteristics have a considerable influence on a variety of firm outcomes (for a recent overview of the concepts studied in the ownership literature, see Boyd and Solarino (2016)). Increasing the specificity to the time orientation studied here, there is also evidence that the time orientation of shareholders’ influences firm outcomes (Derrien, Kecsk´es and Thesmar, 2013; Harford, Kecsk´es and Mansi, 2015). However, there is a lack of evidence on the degree to which culture of shareholders affects firm outcomes and the extent to which national culture is important in explaining corporate short-termism. Starting from the latter, this paper will combine and address these issues.

Firm-level determinants of corporate short-termism

of corporate short-termism: the time orientation of the investor base, the time orientation expressed in compensation packages and the degree of analyst coverage. The literature on these concepts is briefly reviewed below, starting with the time orientation of the investor base. Shleifer and Vishny (1990) noted that arbitrage on financial markets will concentrate on short-term assets because any mispricing in such assets is likely to resolve faster due to their short-term nature. Hence, arbitrageurs exert pressure on management to push for short-term profits in order to resolve the (perceived) mispricing. This corresponds to the expectation that firms with an investor base that is short-term oriented will tend to become overvalued (Stein, 1989; 1996; Polk and Sapienza, 2009; Derrien et al., 2013), since investors exert pressure to increase short-term profits which will not reflect the fundamental value of the firm. The effects also hold at the other end of the time spectrum. Long-term oriented investors can be beneficial for decent corporate investment policies. In two papers, Bushee (1998, 2001) found evidence for the important role of institutional investors in firms’ time horizons. Bushee (1998, 2001) classified institutional investors into three groups and found that transient (short-term) institutional investors tend to increase managerial myopia, whereas dedicated and quasi-index investors do not. Overall, the evidence in favour of long-term oriented institutional investors providing benefits in the form of reduced managerial myopia is overwhelming (Wahal and McConnell, 2000; Aghion, van Reenen and Zingales, 2013; Brochet et al., 2015; Chen, Lin and Yang, 2015) and hence this is an important determinant to account for in research on corporate short-termism.

Related to the discussion on investor base is the common practice of providing stock-based compensation to executives as a method of resolving agency problems between man-agers and shareholders. Recent research has explicitly tied the duration of compensation packages to the time orientation of the firm/investors (Cadman and Sunder, 2014; Gopalan et al., 2014; Edmans, Fang and Lewellen, 2014). Moreover, Dikolli, Kulp and Sedatole (2009) explicitly show that there exists a connection between the investor base, executive compensation and time orientation, illustrating the close relation between these concepts. Furthermore, there is also the straightforward argument that once executives are granted stock-based compensation, they are shareholders with similar interests as regular sharehold-ers. Hence, they have an interest in increasing the share price of the stocks. As noted, the timeframe over which managers with stock-based incentives seek to increase share price will depend on the duration of the compensation package they receive (Gopalan et al., 2014).

Finally, the influence of sell-side analysts following the firm can result in pressure for short-term performance (Fuller and Jensen, 2002; Graham, Harvey and Rajgopal, 2005; He and Tian, 2013, Brochet et al., 2015; DesJardine and Bansal, 2015). These analysts publish their own expectations for the share price of a firm and are thus an important source of information for investors. Moreover, not meeting analyst expectations generally puts a downward pressure on the stock price, hence managers have an incentive to keep up with the (at occasion unrealistic) expectations of analysts (Dechow, Hutton and Sloan, 2000; Hong and Kubik, 2003), which can induce managerial myopia in order to meet these expectations.

Why culture matters

from other countries, regions or groups”. As such, culture determines what people value and deem important in their day-to-day lives. This implies that cultural origins will drive, or at least influence, the decisions people make. To relate the concept of culture to corpo-rate short-termism, one natural view is to examine how cultural values express preferences for risk. However, as Shao et al. (2013) argue, risk is for the largest part captured by the individualism-collectivism dimension of a national culture. For a broader perspective, agency theory provides us with a starting point of why culture might matter for short-termism. Ultimately, short-termism is value decreasing, according to a recent study by Flammer and Bansal (2017), therefore it eventually hurts shareholders. Managers, who are usually assumed to be risk averse and self-interested (Jensen and Meckling, 1976), thus seem to have an incentive to focus on the short-term, which might be exacerbated when provided with short-term incentives (Edmans et al., 2014). Arguably, the relation between managers and principals is partially shaped by the national culture prevalent in the coun-try of origin (Fidrmuc and Jacob, 2010; Hooghiemstra et al. 2015). The programming of the mind that Hofstede (2001) referred to translates to different preferences of what is acceptable to the principal in a principal-agent relationship, or conversely what the agent considers acceptable in pursuing his or her own interests in such a relationship. Hence, the degree of agency problems at least partially hinges on the national culture in the country of origin. Fidrmuc and Jacob (2010) also document empirical evidence on this in the form of higher dividend payouts across culturally diverse countries. Therefore, culture has already provided a contribution to several firm outcomes, and therefore it is expected to (partially) determine the time orientation aspect of agency situations as well.

Through the lens of agency theory, the main channel through which national culture will affect the firm’s time orientation is through the diverging interests of managers and shareholders. It is likely that shareholders and managers have different time orientations1, which might give rise to agency costs. For example, if a shareholder is in for the long-term, whilst the manager is behaving relatively myopic due to his compensation package, cultural heritage, or because he sees scope to extract rents on the short-term due to poor governance, investments are made sub-optimally. The investor would prefer the manager to put more emphasis on, for example, R&D, or in general a more long-term oriented strategy. Thus, within agency theory, there is substantial scope for time orientation to influence agency costs. Moreover, as argued, culture is likely to co-determine the extent of differences in preferences between managers and shareholders. Hence, the need arises to examine how national culture influences the short-term orientation of firms. To do so, a first requirement is to distinguish the dimensions of culture. Several authors have attempted to measure cul-ture, but the Hofstede (2001) dimensions are still the most prevalent2in empirical research. Thus, the Hofstede dimensions will primarily be used to study their effects on corporate short-termism. Consistent with previous research, not all dimensions of national culture will be examined (Doupnik, 2008; Han et al., 2010; Shao et al. 2013; Kanagaretnam, Lim and Lobo, 2014). Only those for which (relatively) sound empirical and/or theoretical results have been obtained will be included. For this study, three dimensions are used: long-term

1

For example due to risk preferences. Risk neutral shareholders might not care that much about the riskiness of projects because they can easily diversify. Managers cannot, and their risk aversion might result in a preferences for the short-term. Alternatively, managers might envisage employment of, say, five years, whilst the shareholder might plan to invest one year, or twenty years.

2

orientation, individualism, and uncertainty avoidance. The brod rationale for this is that individualism and uncertainty avoidance have been most influential in a business environ-ment, whereas the long-term orientation dimension intuitively and theoretically should be important in explaining short-termism.

Long-term Orientation and short-termism

Long-term orientation is defined and explained by Hofstede (2001) as follows: “Every society has to maintain some links with its own past while dealing with the challenges of the present and the future. Societies prioritize these two existential goals differently. Societies who score low on this dimension, for example, prefer to maintain time-honoured traditions and norms while viewing societal change with suspicion. Those with a culture which scores high, on the other hand, take a more pragmatic approach: they encourage thrift and efforts in modern education as a way to prepare for the future.” In a business context, this is arguably not very informative. Moreover, the label of this dimension, long-term orientation, appears to differ from the content that is actually measured by this dimension. Zheng et al. (2012: 486) agree on this, and note the following additional information that Hofstede (2001) provides: “According to Hofstede (2001), virtues and behaviours valued more in long-term orientated societies are directed at the future and include persistence (perseverance), status-based relationship ordering, thrift, and a sense of shame. In contrast, virtues and behaviours valued more in short-term-oriented societies are directed at the past and present and include personal steadiness and stability, protecting face, respecting tradition, as well as reciprocation of greetings, favours, and gifts.” This provides slightly more intuition: long-term oriented societies are directed at the future, whilst short-term oriented societies are directed towards the past and present. These seem to be exactly the characteristics that might influence corporate short-termism.

In the empirical literature, not much work has been devoted to the study of the long-term orientation dimension in the context of time orientation of firms.Nevertheless, the studies that do exist are helpful. Hofstede et al. (2002) document that business leaders in long-term oriented countries are more likely to focus on the long-term, compared to business leaders that reside in short-term oriented countries, who tend to focus on the current year’s profit. This indicates that long-term orientation would decrease short-termism. Similar evidence is obtained by Lewellyn and Bao (2015), who study R&D investments3. These authors find a positive association of long-term orientation with R&D investment. Hence, this provides further empirical evidence that the common-sense interpretation that long-term orientation will reduce short-termism is empirically valid, despite the somewhat unrelated definition. Further empirical evidence is less clear cut, but nevertheless strengthens the case for a negative relation between long-term orientation and corporate short-termism. Antonczyk, Breuer and Salzmann (2014) find evidence that firms in countries with a higher score on the long-term orientation dimension prefer bank-lending based on a relationship with the bank, since this type of funding is available for the long-term. Furthermore, in two similar studies Zheng et al. (2012) and Chang, Wee and Yi (2012) find that long-term oriented so-cieties prefer short maturity debt over debt with a longer maturity. They theorize that this finding implies the following: firms from long-term oriented societies want flexible financing for future chances or threats, and thereby have a longer horizon in mind. This somewhat contrasts the findings of Antonczyk et al. (2014), although Antonzyk at al. effectively ex-amine whether firms prefer the stock market over relationship based debt funding, whereas Zheng et al. and Chang et al. (2012) examine the maturities of debt once firms decide on

3

issuing debt. Hence, the explanations offered do not necessarily contradict each other. In sum, the empirical evidence thus suggests that long-term orientation decreases corporate short-termism. Formally:

Hypothesis 1a: The degree of long-term orientation is negatively related to corporate short-termism.

Individualism and short-termism

Individualism is defined as the degree to which individuals look after themselves. People have a preference for a loosely-knit social framework in individualistic societies, where they only care for their close relatives, not for society as a whole. Collectivist societies represent the opposite: a tightly-knit social framework and a substantially larger circle of people for which one cares are characteristics of such societies. This dimension appears to have no direct relation with the time orientation of firms that have their roots in a more indi-vidualistic or more collective society. However, as past research has shown, the degree of individualism in a country is a dimension that has featured in many empirical studies on firm outcomes (Han et al., 2010; Shao et al., 2013; Kanagaretnam et al., 2014; Hooghiemstra et al., 2015; Lewellyn and Bao, 2015). Many of these outcomes can be related to corporate short-termism.

Shao et al. (2013) examine corporate investment from a risk-taking perspective. Accord-ing to them, risk takAccord-ing is more acceptable in individualistic societies, because individuals care for their own, and risk taking is not perceived as jeopardizing others. Something that more collectivist societies might perceive to be the case when risky investments are made. Since long-term projects are inherently more risky (their payoff is further into the future), individualistic societies would be more likely to support a long-term orientation for firms, according to the reasoning of Shao et al. (2013). Similar results are obtained by Lewellyn and Bao (2015) when they employ Hofstede measures4: individualism increases R&D investment. Further evidence in favour of the risk taking argument is presented in Kanagaretnam et al. (2014), who document that individualistic societies are less conserva-tive in their accounting. Arguably, optimistic accounting represents risk-taking behaviour because it might result in overstating results. Furthermore, Han et al. (2010) document a positive relation between individualism and accrual earnings management. Although this could be interpreted as evidence in favour of a positive relation between individualism and short-termism, earnings management can also be viewed as a tool to make the firm ap-pear more profitable than it currently is, to ensure that shareholders remain satisfied until longer-term projects result in a higher profitability. Thus, accruals management might pre-vail more often in individualistic societies because firms undertake more long-term oriented projects and manage accruals to ensure a constant level of profits. In sum, the evidence thus supports a negative relation between individualism and short-termism. Formally: Hypothesis 1b: The degree of individualism is negatively related to corporate short-termism. Uncertainty Avoidance and short-termism

Uncertainy avoidance deals with the degree to which a society is comfortable with uncer-tainty and unsure outcomes. Therefore, societies that score high on unceruncer-tainty avoidance

4_{Note that these authors have tested their hypotheses with both the House et al. (2004) dimensions and}

tend to stick to what they know, and fear to venture into unknown areas. Hofstede (2001) notes the following on the extreme points of the uncertainty avoidance continuum: “Coun-tries exhibiting strong UAI maintain rigid codes of belief and behaviour and are intolerant of unorthodox behaviour and ideas. Weak UAI societies maintain a more relaxed attitude in which practice counts more than principles.” Uncertainty avoidance therefore seems to be strongly related to the degree of innovation, and research on uncertainty avoidance and innovation is therefore abundant (Venktaraman et al., 1993; Shane, 1995; Mueller and Thomas, 2000; Jones and Davies, 2000; Lewellyn and Bao, 2015). However, also (other) firm outcomes have been related to uncertainty avoidance, such as earnings management (Han et al., 2010), debt maturity (Chang et al., 2012; Zheng et al., 2012) and accounting conservatism (Kanagaretnam et al., 2014).

Most closely related to the definition of uncertainty avoidance presented above, is the work that directly relates the dimension to some measure of innovation. Jones and Davies (2000) argue that low uncertainty avoiding societies foster innovativeness because out-of-the-box thinking by colleagues and employees is more acceptable in such societies. As a firm outcome, Lewellyn and Bao (2015) hypothesize and find that uncertainty avoidance decreases R&D investment, their explanation conforms to that of Jones and Davies (2000): low uncertainty avoiding cultures are more tolerant of new ideas and uncertain outcomes of projects. Note that this closely relates to long-term oriented projects, whose outcome is also inherently uncertain. Hence, this indicates that uncertainty avoidance might increase short-termism. Moreover, recall the discussion of Han et al. (2010) on accruals management in the previous section. These authors also find that uncertainty avoidance decreases discretionary accruals. The reasoning for why this might occur is directly related to Kanagaretnam et al. (2014) who find a positive relation between uncertainty avoidance and accounting conservatism. In high uncertainty avoidance countries, rules for accounting are likely to be more conservative according to Kanagaretnam et al. (2014), yielding less scope for accruals management. Note that also the possibility of accruals management as a positive tool to enable long-term projects reinforces the idea of uncertainty avoidance having a positive relation with short-termism. Uncertainty avoiding cultures prefer the less ambiguous and less uncertain outcomes of short-term projects over long-term projects whose payoffs are inherently more uncertain. Thus, in such cultures, earnings management will occur less because there is no need to cover for reduced profits until long-term projects return their investment. Finally, Chang et al. (2012) and Zheng et al. (2012) both document a negative relation between uncertainty avoidance and debt maturity. The underlying reason seems clear: firms from uncertainty avoiding cultures prefer a higher maturity because this leads to less uncertainty about future financing terms and agreements.

In sum, apart from the simple reasoning that uncertainty avoiding societies prefer short-term payoffs simply because they bring about less ambiguity, the empirical evidence and theoretical underpinnings of this evidence all provide strong support for a positive relation between uncertainty avoidance and short-termism. Formally:

Hypothesis 1c: The degree of uncertainty avoidance is positively related to corporate short-termism.

Investors and National Culture

agents has shown that cultural roots influence the nature of agency relationships (Tosi and Greckhamer, 2004; Crossland and Hambrick, 2007; Li and Harrison, 2008; Liu, 2013; Kish-Gephart and Campbell, 2015; Brochet et al., 2016). The cultural background of investors has received relatively little attention however, despite investors being an integral part in much research on firm outcomes (Boyd and Solarino, 2016). Partially, this might be due to the loose theoretical linkage between investor culture and firm outcomes. The investor base is often broad and diverse, which makes it difficult to assess how the direction of shareholder influence can be traced back to informal institutions that influence these investors’ norms and values in much the same way that this occurs for managers. The belief that investor national culture is important for firm outcomes is reinforced by research on the influence of cultural background on managerial decisions. Since cultural roots have been shown to influence managerial decisions and subsequent outcomes, there seems no theoretical reason why investor culture should not affect shareholders’ perceptions of agency relationships as well (Fidrmuc and Jacob, 2010). Moreover, based on the theoretical foundations discussed next, the concept of firms reacting to shareholder preferences with respect to time orienta-tion seems to be a suitable starting point to examine whether the cultural background of investors matters.

These theoretical foundations for investor preferences affecting firm time orientation trace back to the catering theory of Stein (1989; 1996), which posits that managers can maximize short-term stock prices, or alternatively maximize the long-term value of the firm. If the investor base is entirely long-term oriented, there is no need to maximize short-term stock prices, because investors are in for the long haul and recognize that over a long time period the firm will converge to its fundamental value, hence short-term setbacks do not matter. If the firm investor base consists of more short-term oriented shareholders however, the need to emphasize the short-term arises. If the manager does not, shares are worth less to these short-term shareholders and they hence suffer a loss because they sell their shares before the fundamental value of the firm is realized (Polk and Sapienza, 2009; Derrien et al. 2013). In sum, this thus implies that managers (and hence firms) will cater to their shareholders’ time preferences. An interesting finding that supports this notion to some extent was documented in Dill, Jirjahn and Smith (2014). In Germany, a country which is generally characterized by patient, long-term investors, the authors found that firms owned by foreigners tend to have an emphasis on the short-term. Thus, this provides some initial evidence that the time preferences of investors are likely to be co-determined by the cultural background of the investor base.

Cultural diversity and short-termism

cultural distance between countries as an explanation of the foreign bias in institutional investors’ portfolios. Other examples include Dodd, Frijns and Gilbert (2015) or Karolyi (2016). An example of a study that examines cultural diversity within firms (rather than between countries, or the countries where firms reside) is Frijns et al. (2016) who use an adapted version of the Kogut and Singh (1988) measure to examine cultural diversity within the board of directors. For cultural diversity on boards, Frijns et al. (2016) document a negative relationship with firm performance, although different nationalities on the board can also have positive effects when the firm has more international activities (Masulis et al., 2012; Est´elyi and Nisar, 2016). In sum, cultural diversity has had important implications for a range of topics in finance research. However, the case for investors is, arguably, different from the studies examined thus far. Therefore, we now have the means to examine how the investor base can affect short-termism, but the second point remains: will cultural diversity increase or decrease short-termism.

For this, we will return to the theories discussed previously. Resorting to theory rather than explicit empirical evidence becomes a necessity since previous research on cultural diversity within the investor base of a firm is, to my best knowledge, not available. Note that predictions from other groups that operate within firms, such as boards cannot be immediately translated to investors. Boards are management teams that work together on a regular basis, unlike the investor base of a firm (otherwise firms would find themselves with no need for a board). Thus, the motives for examining cultural diversity among investors are also not comparable to those of Frijns et al. (2016), who name intra-group trust and the degree of task-related and relationship conflict as main consequences of cultural diversity.

Thus, the main foundation for this effect is based on catering theory and agency theory. In catering theory, the general assumption is that management tries to optimize the average investors preferences for the firm’s time orientation (Derrien et al. 2013). Hence, when there are investors from more diverse countries in the firm, the average investor will have a culture that differs stronger from the firm’s national culture. As discussed, this is likely to affect the investor perception of agency problems within the firm. Thus, it is likely that a more cul-turally diverse investor base complicates matters, since the perceptions of agency problems will diverge more and more when the cultural diversity increases. As previous research has shown, investors tend to shy away from more culturally diverse countries, possibly because the aforementioned problems result (Beugelsdijk and Frijns, 2010; Beracha et al., 2014). Therefore, it seems reasonable to posit that investors which are culturally different from the firm in which they invested, are invested there in a passive role, to diversify their hold-ings for example. Hence, managers can roam freely, compared to their counterparts with an all domestic investor base, where investors attempt to meddle in firm policies because they are culturally close to the firm and understand the agency relationship in this culture. From this point, there are two possibilities. First, the managers might be, as agency the-ory predicts, risk averse and prefer short-term projects that boost short-term performance. Second, as Bayer, Larcker and Tayan (2014) noted, they might follow-up on their thrive to focus on prestigious long-term projects without the pressure from their shareholder base, consistent with early catering theory (Stein, 1996). In sum, taking into consideration that most agency problems in Europe are reasonably covered due to governance structures, the second explanation seems likely in this case. Hence, a culturally diverse investor base is expected to have a negative relation with corporate short-termism. This results in the final hypothesis:

Data

The sample on which the specified hypotheses will be tested consists of the 500 largest firms in Europe, according to the list compiled by the Financial Times5. The time period under consideration ranges from 2002-2016. For the earlier years, data availability is limited for some variables, as will be shown in table 1. The first step in the data collection was to gather conference calls through the Factiva database. Within Factiva, calls are extracted from CQ FD Disclosure. Most firms had several consecutive quarters of conference calls available, although some had only limited availability or were completely missing from the database. The sample choice is mainly governed by data availability considerations. When the Factiva database was searched for conference calls of smaller firms, availability of conference calls dropped quickly. For this reason, the largest 500 companies are used. To obtain a multi-country sample with ample variation in cultural background, Europe provides an excellent opportunity to collect data. Although there are substantial cultural differences, we can still

Table 1: Sample selection

Panel A: From quarters to firm-years

Initially collected quarterly conference calls 15,526 Quarterly conference calls that are double 2,096 Quarterly conference calls after accounting for doubles 14,436 Summing word counts over quarters yields firm-years 4,798 Panel B: Unique firm-year aggregated conference calls

2002 47 2003 199 2004 233 2005 243 2006 282 2007 317 2008 323 2009 333 2010 353 2011 401 2012 415 2013 415 2014 427 2015 418 2016 392 Total 4,798

This table summarizes in Panel A how the total number of firm-years was obtained from the initially collected quarterly calls. Moreover, it depicts in panel B how many conference calls/firms are available per year. It does not yet take into account the availability of other data, only conference calls are considered in this table.

5

acknowledge a certain similarity between countries in other respects due to the integrated nature of Europe, hence this geographical area provides a decent opportunity to study the effects of national culture. The choice of time period is again governed by data availability considerations. The Factiva database contains virtually no conference calls of years before 2002. In fact, as mentioned, even in the earlier years of the 2002-2016 time period conference calls are only sparsely available. Hence, it was not possible to collect calls from years before 2002. For obvious reasons, collection of more recent conference calls was also not possible. This sample is comparable in nature to the sample of Brochet et al. (2015) in the sense that they also obtain a cross-industry sample. Alternatively, DesJardine and Bansal (2015) take a different approach when measuring short-termism expressed in conference calls by limiting their data collection to the extraction industry. Although the approach of DesJardine and Bansal (2015) yields a clearer image because industries inherently differ in their time orientation, it is possible to control for differences between industries by including industry

Table 2: Time orientation of firms and industries

Panel A: Most long- and short-term oriented industries

Long-term oriented industries Short-term oriented industries

Aircraft Steel works

Non-Metallic and Industrial Metal Mining Electronic Equipment Beer and Liquor Petroleum and Natural Gas Printing and Publishing Machinery

Utilities Banking

Construction Business Supplies

Construction Materials Electrical Equipment Food products Pharmaceutical Products

Transportation Communication

Trading Consumer Goods

Panel B: Most long- and short-term oriented firms

Long-term oriented firms Short-term oriented firms

Baloise Marine Harvest

Associated British Foods Norsk Hydro

Rolls-Royce Arcelor-Mittal

Bae Systems Transocean

Severn Trent DNB

RELX Swedbank

Pearson Infineon Technologies

Bunzl Gjensidige Forsikring

Hikma Pharmaceuticals Volvo

United Utilities Yara International

fixed effects or by means of correcting observations for industry means. Hence, multiple industries are included to obtain a sample of sufficient size.

The procedure of collecting the conference calls is further elaborated on in the Method-ology section. The final sample eventually consisted of 15,526 quarterly conference calls for 468 unique firms. Table 1, panel A breaks down how this initial number reduced to the eventual sample of 4,798 in terms of firm-years. Note that the reduction in quarterly calls when removing the excess conference calls is 1,090 and not equal to half of 2,0966. This is due to the fact that some quarters even had more than two conference calls. Panel B of table 1 depicts the number of firms (or word counts obtained from the quarterly calls) per year. As becomes clear, most years have a relatively even distribution of available confer-ence calls, although there is a substantial increase compared to the earlier years. Using a textual analysis procedure described in the next section, I arrive at a measure for corporate short-termism at the firm-year level. Employing this data and sorting it by firm or industry, an initial image of which firms and industries are focused on the short-term can be provided. This effectively mimics table 2 from Brochet et al. (2015) for this dataset, which makes it interesting to compare whether similar classifications are obtained. Table 2 provides the aforementioned overview.

Comparing table 2 with the classification in Brochet et al. (2015), Table 2 panel A shows some remarkable differences. Trading is for example included in the list with long-term oriented industries, whilst Brochet et al. (2015) classify it as short-long-term oriented. Similarly, pharmaceutical products is included in the list with short-term oriented industries here, whereas it is classified as a long-term oriented industry in Brochet et al. (2015). The same holds for consumer goods, which also has a reversed classification. Additionally, for Shipbuilding and Railroad equipment it would also hold that classifications are reversed, but this industry had too little observations in this sample to be included in the table (it most likely involved only a single firm). This preliminary classification of industries thus presents some evidence that the European situation is to some extent different from the situation in the United States. A possible explanation for these differences, as this paper explores, might be the cultural differences between countries. Panel B of table 2 lists the firms with the strongest orientation towards the long- or short-term. Consistent with the industry classification, several financial firms are present in the list of short-term oriented firms, such as DNB, Swedbank and Gjensdige Forsikring. In the list of long-term firms are several publishers like Pearson and RELX, corresponding to their industry classification. There are also exceptions however, like Hikma Pharmaceuticals which has a longer-term focus than its industry peers.

Apart from the measure of corporate short-termism, other firm-level data is used in this study. First, some measure of national culture is required for which Hofstede (2001) and House et al. (2004) provide classifications often used in empirical research. Both measures are employed in this study, although the main results are obtained with the measures developed by Hofstede (2001). Second, financial data is gathered, mainly for the purposes of calculating control variables. This financial data is extracted from Compustat Global, where for most firms the most important items were available. Third, I/B/E/S provides the number of analysts that cover each firm in the sample. Fourth, the Thompson Reuters Eikon environment allows the extraction of shareholder history reports, with detailed information on the composition of the investor base over time. Examples of data available through this source are the percentage of shares an investor has, which country the investor is from,

6

what the earliest holding date of this investor is and the share of the investment in a certain firm in the investor’s total portfolio. Unfortunately, this data source is often lacking data in the early years for non-British firms. Finally, some measure of the time duration embedded in the compensation of executives is needed to include this as a driver of potential results. The data for this measure is collected from CapitalIQ, both for all executives as well as the CEO and CFO individually. Unfortunately, this compensation data is not always of high quality and properly available, so including the compensation measure will reduce the number of observations available. Moreover, because several databases are matched it is often the case that for a firm which has analyst data in a certain year the measure for ownership is lacking, which also puts a downward pressure on the number of observations employed in the analysis.

Since this study examines national culture, there are two levels in the data, one firm-level and one on the country-firm-level. In empirical research on national culture it is common to include country level determinants as controls (Han et al., 2010; Li et al., 2011; Shao et al. 2013; Hooghiemstra et al., 2015; Lewellyn and Bao, 2015), such as investor protection and stock-market development. The geographical area examined here is relatively well developed compared to the studies just mentioned, who often employ a range of countries all around the globe. Hence, the need to control for differences in stock-market development and investor protection seems more important in such samples. Nevertheless, although these country level determinants are not the key point of interest in this study, it is not unlikely that they correlate with both short-termism and national culture. If the stock market is well-developed in a certain country, firms listed in such a country are more likely to be influenced by short-term pressures from shareholders, simply because their financing relies strongly on this source. Moreover, when these country-level variables are correlated with national culture, this might present a problem with several of the empirical specifications if these variables are not explicitly accounted for. In sum, it is therefore appropriate to include those variables that have been identified in previous research as important covariates when examining the effects of national culture. As such, this study includes data from the World Bank on GDP, GDP growth and stock-market development. Moreover, the level of investor protection as constructed by Djankov et al. (2008)7 _{is used as a final country-level}

determinant.

Methodology

Measuring Short-termism

Measuring corporate short-termism by analysing conference calls to obtain an indication of the firm’s time orientation is a relatively new approach. The general area of textual analysis has recently surged in the accounting and finance literature, as illustrated in Kearney and Liu (2014)8 who provide an overview of literature on textual analysis in finance. Specifically measuring short-termism has recently been initiated by Brochet et al. (2015) and DesJar-dine and Bansal (2015). Others used textual analysis to measure readability (Merkley, 2014; Lo, Ramos and Rogo, 2017), tone (Huang, Zang and Zheng, 2014), congruency with a norm (Shin and You, 2016) and deceptiveness (Larcker and Zakolyukina, 2012). This selection of

7_{This is an updated version of the investor protection index initially developed by La Porta et al. (1998)}

and is often referred to as the anti-director rights index. Data for this measure is available from the personal website of Rafael La Porta.

8

the literature illustrates that textual analysis can be very informative about firm practices and outcomes. Measuring short-termism is a difficult task. Researchers often use imperfect proxies like shareholder time orientation (Harford et al., 2015) or R&D investment practices (Lewellyn and Bao, 2015) to approximate time orientation. Arguably, the time orientation subconsciously expressed in conference calls provides a better approximation then previ-ously used proxies, because it does not focus on a single aspect that might be part of a long time orientation, such as R&D investment, but rather provides a measure of the aggregate time orientation that results from the decision-making within the firm.

In order to measure short-termism expressed in conference calls, the approaches taken by Brochet et al. (2015) and DesJardine and Bansal (2015) are employed. I will discuss both in turn, although the two are largely similar. Before turning to the actual discussion of the measures, the process of extracting calls is briefly revisited to provide details. Similar to DesJardine and Bansal (2015), I use the Factiva database to extract quarterly confer-ence calls manually. Specifically, within the Factiva database, the CQ FD Disclosure source is selected to extract the conference calls. Further settings are that the language of the transcripts must be in English and the subject selected is “Transcripts” in order to avoid cluttering of the search results. These results were obtained by searching for specific firms. It proved most efficient to select and subsequently download as many quarterly calls in one go as Factiva allows without crashing. After downloading, the accumulated calls were automatically split using a tailor-made script that obtained separate files with each file con-taining only a single call instead of the approximately twenty per file that were downloaded. From here on, textual analysis per call is possible. At this point, the number of word counts as mentioned in table 1, panel A is obtained. Before applying the formulas used by respec-tively Brochet et al. (2015) or DesJardine and Bansal (2015), two more adjustments are necessary to arrive at the final firm-year observations. First, some firms issue more than one call per quarter, for example one for fixed-income investors and one to discuss their quarterly results. Alternatively, some calls, especially the earlier ones, have a morning and afternoon session, or two sessions in different cities. Summing the word counts from double calls and subsequently eliminating the doubles reduces the number of available calls from 15,526 to 14,436. Second, the calls are on a quarterly level, which must be aggregated to an annual level. The same procedure as when dealing with doubles is applied here. All word counts are summed over quarters, and then three out of four quarters are deleted. This results in 4,798 firm-year observations.

To obtain the measure of short-termism, both Brochet et al. (2015) and DesJardine and Bansal (2015) employ a dictionary approach9. This boils down to generating a list of words that is supposedly related to the construct being measured, in this case short-termism. The approach taken specifies certain words as indicators of short-termism, whereas another list consists of words referring to the long-term. Brochet et al. (2015) developed their own word lists, based on their reading of 33,000 lines of conference call transcripts. A full overview of the words used is presented in table 10 in Appendix A. Note that there were also a number of neutral words the authors initially classified as either short or long term. In their subsequent survey distributed among students these neutral words were not confirmed to indicate either the short-term or the long-term. Therefore, these are not included in the measure. These word lists are then employed in a script specifically tailored to count the occurrence of all words in the short-term list, and all the words in the long-term list. This script loops over all quarterly conference calls available. Subsequently, the total count per

9

quarterly conference call for both the short-term and the long-term is depicted as output. In addition, the total number of words in the call is provided as output, together with the date of the conference call. Since no clear link to firm name or identifier is provided in the calls, the word counts are subsequently hand-matched to the other data. The actual measure of short-termism is then calculated as follows:

T imeHorizonBit=

word count short-termit

word count long-termit

. (1)

The B refers to the measure used, Brochet in this case, whilst i denotes firm and t denotes time. Note that defining short-termism like this implies that a higher measure means that a firm is focused more on the short term. Also note that this measure is applied after summing word counts, as described above. Hence, first calls are collected in batch, then they get separated into individual calls, next words related to the short- or long-term are counted, and finally doubles within and between calls are summed and deleted. Only then the measure of short-termism is calculated.

The methodology of DesJardine and Bansal (2015) is largely comparable in nature, although the process followed in order to arrive at their word lists is slightly different. First, the authors manually determined words to be included. In addition, they used a standard word list which included words related to time. Second, three researchers discussed the words and categorized them into short, long or unclear. Only the first two categories are retained to arrive at the words listed in Appendix A, table 11. Finally, the authors checked for the context in which the words occurred. They randomly selected at least 25 occurrences of each word and read the sentences in which these words played a role to assess whether the word is really used the way it was classified. Finally, they take a slightly different approach when defining the actual short-termism variable:

T imeHorizonDit=

word count long-termit

word count short-termit+ word count long-termit

. (2) In this specification, the degree of short-termism is expressed as a proportion of long-term words to the total words related to time horizon. Therefore, a higher value of this measure corresponds to a longer time horizon.

Measuring independent variables

National Culture

Power-distance, masculinity and indulgence appear to have no theoretical relationship with short-termism and are therefore not included. Regarding the GLOBE dimensions, emphasis is on those dimensions that closely correspond to the Hofstede measures, specifically future orientation, institutional collectivism and uncertainty avoidance. The GLOBE project has two alternatives, practices and values. For the table reported in the appendix, the practices variant was used. Although some differences occur when the values alternative is used, these differences are not that large. To assign cultural values to each observation, the country of residence is employed as a proxy for cultural origins. In most cases, this assumption seems appropriate, although it is likely that some bias occurs due to the multinational nature of most firms. When actually matching these values with observations, difficulties occur with RELX and Unilever, since these firms are based in two countries. This problem is solved by taking averages of cultural values for these two countries. Including or excluding these two firms furthermore leaves the results unaffected.

National culture of investors

For the second hypothesis, the cultural diversity of investors has to be determined. Our measure is related to that of Frijns et al. (2016), but slightly different. One difference is that Frijns et al. (2016) employ multilateral differences, so among all board members within the board, every pair is used to calculate the measure. This study adopts a bilateral approach, like in Kogut and Singh (1988), Beugelsdijk and Frijns (2010) or Dodd et al. (2015). For each of the ten largest investors, the difference in each dimension between the firm and every investor is aggregated into a measure of cultural diversity, which is defined according to the following formula:

CDi = J X j∈J v u u t 4 X k=1 (Cki− Ikj)2 Vk . (3)

In this formula, k denotes the cultural dimension, whilst Ckidenotes a firm’s national culture

score on dimension k and Ikj denotes an investors national culture score. Furthermore, J

is the subset among all investors (J ) with shareholdings larger than the cut-off. Finally, Vk denotes the in-sample variance for the corresponding cultural dimension (at the firm

level) and t indicates time. To arrive at the final measure, we divide CDi by the number of

investors J to obtain an average measure of the cultural diversity among the investor base. Note that k is not restricted to only long-term orientation, individualism and uncertainty avoidance. Untabulated tests show that masculinity also significantly affects short-termism, and therefore this cultural dimension is also included in the measure of cultural diversity. The dimensions of power-distance and indulgence do not affect short-termism individually, and hence they are also not included in the cultural diversity measure. To conclude this section, there are also three dummy variables used in the additional analysis that have their closest relation to investor national culture. First, a dummy that equals one if there is a foreign investor among the set of investors J and zero otherwise. Second, a dummy that equals one if there is a UK investor among the set of investors J in those firms that are not located in the UK, and zero otherwise. Third, a dummy that equals one if there is an investor from the US among the set of investors J , and zero otherwise.

Measuring other independent variables

the role of investors’ time orientation on corporate short-termism, sometimes treating them as equivalent (Polk and Sapienza, 2009; Harford et al. 2015). Hence, a measure for the time orientation of investors is included in most specifications. This measure is constructed from the data obtained through Thompson Reuters Eikon. It is tailor made to the data available and consists of the following parts. First, the total percentage of shareholders that are identified in Eikon is calculated. Second, for each of these investors the earliest holding date is subtracted from the current date, yielding the number of months each investor currently possesses shares in the firm. Then, a certain threshold is specified, for example 36 months. If a shareholder has his/her shares for a period ≥ 36 months, his/her shareholdings are included in a sum of all long-term shareholders. Finally, the sum of long-term shareholders is divided by all identified shareholders to obtain a proportion of shareholders oriented towards the long-term. In formula, this boils down to:

Sit = PJ j∈Jsjit PJ sjit with J ⊂ J , (4)

where j identifies each shareholder, i identifies firms and t specifies the year. Moreover, the subset J is comprised of those shareholders whose holding period exceeds the long-term threshold, which is defined as 12, 36, 60 or 120 months. So, four different variants of this variable are specified. The complete set of shareholders in firm i at time t is reflected again by J . Finally, Sit is the variable of interest: the proportion of shareholders who hold their

shares for the long-term. Note that although this specification is expected to pick up the time orientation of shareholders, there is considerable noise. This is to a large extent caused by the fact that Eikon cannot identify all shareholders. Hence, we cannot be sure of the time orientation for these shareholders. However, it seems likely that these shareholders are small and probably do not interfere with the company’s time orientation. Another problem that was already acknowledged is the limited availability of data for non-British firms in early years.

the further calculation of the measures is: LongT ermpit=

Long-term compensationpit

Long-term compensationpit+ Short-term compensationpit

, (5) with p an index to identify which variant of the measure is calculated and i and t defined as usual. The denominator and numerator consist of items available in CapitalIQ, which were classified as either long-term or short-term oriented. Salary, bonus, and immediate stock grants are classified as short-term oriented. Restricted stock grants and amounts paid for a long-term incentive plan (LTIP) are included in long-term pay. Moreover, for both short- and long-term pay there is an “other” category. Finally, the value of options needed to be allocated. Since the vesting period for options is not available, I rely on previous literature, specifically Brochet et al. (2015) who document a positive effect of all stock-based compensation on short-termism. Hence, the options are classified as a short-term reward. This concludes the discussion of how the most important variables are measured. What remains are control variables, both at the firm and the country level.

Measuring control variables

First, control variables at the firm level are defined. Thereafter, a number of country-level determinants often employed in the literature on national culture are discussed. Most control variables at the firm level can be traced back to Brochet et al. (2015), although not all are used in every specification due to the downward pressure on the number of observations of some variables. First, I control for cash flow volatility, defined as the five-year standard deviation of cash flows from operations. Arguably, when more uncertainty exists, firms might have a stronger focus on the short-term to ensure sufficient funds. This measure is expected to be partially covered by industry fixed effects as well. Second, the return on equity is included to account for profitability. This is defined traditionally as net income divided by the book value of equity, or the extent to which firms are able to generate a return on the invested capital. Next, leverage is included, more levered firms need to keep up with interest payments, which might induce a focus on the short-term. This ratio is defined as total debt divided by total assets. Fourth, liquidity is included, measured by current assets divided by current liabilities. This approximates the extent to which a firm is able to meet its short-term obligations. Fifth, the operating cycle is included, as well as the Ohlson (1980) O-score, to measure whether a firm is close to bankruptcy. Finally, size is taken into account by means of the natural logarithm of sales10.

Second, country-level characteristics are included, because previous research has shown that these characteristics are important in cross-country research on national culture (Shao et al., 2013; Hooghiemstra et al., 2015). First, the natural logarithm of GDP per capita is included, to account for differences between wealthier and less wealthy countries. Within the sample of European countries, there is little variation between the countries, nevertheless it is common in culture research to include a measure of country wealth (Shao et al., 2013; Hooghiemstra et al.; 2015; Lewellyn and Bao, 2015). The same holds for the growth rate of GDP, here there could be intuitive reasons why firms might focus on the short-term if the growth in GDP becomes small or negative. This often indicates a recession, which usually implies difficult times for most firms. Furthermore, the anti-director rights index from

10

Djankov et al. (2008) is included. If investors are well protected firms/managers might have larger incentives to cater to their needs, hence there might be a positive relation between investor protection and short-termism, conditional on the investor base being short-term oriented. Next, the level of stock-market development is included. As mentioned before, higher developed stock markets are likely to exert greater pressure on firms in accordance with the time orientation of the investor base. Stock market development is measured as the average of market capitalization to GDP over the entire time period. Averaging is necessary to retain a substantial number of observations, as the data from the World Bank is missing recent years for the United Kingdom, as well as many years for Scandinavian countries. This concludes the country-level variables. All definitions of every variable just discussed, as well as their source and the abbreviated names included in the tables in the next section, are available in table 12 in appendix B.

Empirical specification

The main results of testing the hypotheses specified above are obtained through application of weighted least-squares (WLS) regressions with the weight inversely proportional to the number of observations in a certain country. White robust standard errors are used to account for heteroskedasticity. This approach ensures that the results are not driven by a country that provides the largest part of the observations (in this case the United Kingdom) and is common in cross-country research on culture (Han et al., 2010; Hooghiemstra et al., 2015). However, it does not account for the panel structure of the data. Since we repeatedly observe the same firms over time, it is hard to assume that the error terms are uncorrelated over time. Thus, the standard errors might be biased. Therefore, in the robustness section, two alternatives are used. First, OLS with clustered standard errors is applied. Second, to acknowledge concerns of autocorrelation in the error terms, the panel structure of the data is taken into account more rigorously. This requires to decide upon a specific estimator for the panel data. Since all hypotheses are tested while including time invariant variable (national culture) a random effects specification is necessary to obtain estimates. To approach the generally preferred fixed effects estimator when there might be unobserved heterogeneity that correlates with the explanatory variables, we can include the means over time for each cross-sectional unit for those variables that are time-variant. This serves to specify the individual effects to some extent, and results in the fixed-effects estimator that also allows for time invariant variables to be included. This is known as the Mundlak approach by Mundlak (1978). In general, some form of the following specification is estimated:

T imeHorizonxit= β0+ β1Analystsit+ β2LongT ermInv5it+ β3LongT ermExecit

+ β4CultureDimensionkx+ β5InvestorCulturekit

+ F X f =1 γfF irmControlf it+ C X c=1 λcCountryControlckt

+ Industry dummiesi+ Y ear dummiest+ εit.

In the above formula, x indicates which country, i denotes firm, and t is again the time subscript. Furthermore, k denotes which cultural dimension is examined. Specific to the controls, F denotes the number of firm-level controls, whilst C is its counterpart for the country-level controls. Small f and c indicate which specific firm- or country-level variables is indicated. Note that for hypothesis 3, the investor culture variable is replaced by the index. Hence, the k subscript is dropped in that case. Finally, the T imeHorizonxit variable

Results

Descriptive statistics

To obtain a first impression of differences between countries in the measure of corporate short-termism, table 3 below presents an overview of both the Brochet et al. (2015) and Des-Jardine and Bansal (2015) measures per country. First, from the table it becomes apparent that the number of observations per country is highly skewed, with the UK contributing more than a quarter of all observations. Thus, WLS is applied, as was discussed in the previous section.

Table 3: Short-termism by country for both measures

T imeHorizonB T imeHorizonD

Country N Mean SD Min Max Mean SD Min Max

Austria 35 0.9622 0.3248 0.3879 1,8598 0.1411 0.0402 0.0576 0.2051 Belgium 87 1.0090 0.4193 0.3382 2.6331 0.1520 0.0372 0.0847 0.2759 Cz. Rep. 12 0.5460 0.1404 0.3722 0.8468 0.1357 0.0371 0.0901 0.2083 Denmark 144 0.7969 0.3570 0.1111 1.8696 0.1561 0.0435 0.0167 0.2857 Finland 103 1.000 0.4110 0.3409 2.8148 0.1379 0.0373 0.0704 0.2674 France 723 0.8026 0.3873 0 3.7368 0.1722 0.0559 0.0081 0.3759 Germany 598 0.8576 0.3516 0.0893 3.0724 0.1502 0.0413 0.0498 0.3333 Hungary 12 1.4177 0.3442 0.9443 2.1708 0.1000 0.0167 0.0742 0.1333 Ireland 49 0.5762 0.3500 0.1184 2.2394 0.1986 0.0589 0.1140 0.3744 Italy 260 0.9635 0.4539 0.1327 2.8571 0.1420 0.0555 0.0351 0.3444 Netherlands 232 1.0033 0.7328 0.0556 4.2400 0.1718 0.0806 0.0556 0.3750 Norway 102 1.6841 0.6752 0.4722 4.2621 0.1243 0.03770 0.0551 0.2244 Poland 40 1.0667 0.4293 0.4118 2.1408 0.1062 0.0426 0.0430 0.2162 Portugal 40 0.9654 0.2845 0.5167 1.8529 0.1602 0.0373 0.0597 0.2208 Romania 4 0.7407 0.0682 0.6772 0.8318 0.1371 0.0150 0.1199 0.1515 Russia 99 1.004 0.4766 0.1667 2.6333 0.1401 0.0515 0 0.3043 Spain 218 0.8701 0.3990 0.0682 2.0375 0.1563 0.05998 0.0464 0.3969 Sweden 301 1.2836 0.4424 0.3125 3.0111 0.1275 0.0412 0.0275 0.3565 Switzerland 342 0.7610 0.5079 0.0561 2.9497 0.1837 0.0512 0.0769 0.3782 Turkey 59 1.0933 0.3925 0.4118 2.2784 0.1204 0.0392 0.0496 0.2168 UK 1,338 0.5713 0.3315 0.0405 3.1250 0.2056 0.0594 0.0260 0.4224 Total 4,798 0.8317 0.4734 0 4.2621 0.1696 0.0593 0 0.4224

Kingdom, as well as a similar culture, is the second most long-term oriented country. A surprising result however is the contradiction in the two measures that sometimes occurs. The Netherlands, Ireland, Romania and Belgium are examples of countries for which both measures are above or below the mean in the total sample, which implies that one measure classifies them as long-term oriented whilst the other considers them to be more focused on the short-term. Since most of the countries for which this occurs do not have many observations, it is deemed appropriate to run all analyses with both measures. In sum,

Table 4: Descriptive statistics independent variables

N Mean SD Min Max

Panel A: Culture dimensions

LT O 4,798 59.27 14.28 24 ₈₃ IN D 4,798 73.37 13.03 27 89 U A 4,798 57.92 22.60 23 99 F U T 4,605 4.04 0.471 2.88 4.73 COL 4,605 4.28 0.383 3.60 5.26 GU A 4,605 4.75 0.532 3.09 5.42

Panel B: Investor culture variables

CD Investor 3,890 1.397 1.227 0 5.991

F or Investor 4.062 0.758 0.428 0 1

U S Investor 3,904 0.590 0.492 0 1

U K Investor 3,904 0.144 0.352 0 1

Panel C: Firm-level independent variables

LongT ermInv1 4,048 0.927 0.138 0 1

LongT ermInv3 4,048 0.801 0.236 0 1

LongT ermInv5 4,048 0.685 0.282 0 1

LongT ermInv10 4,048 0.379 0.305 0 1

Analysts 4,586 3.117 0.527 0 4.331

LongT ermT otal 3,697 0.257 0.220 0 1

LongT ermExec 3,668 0.254 0.221 0 1

LongT ermCEO 3,550 0.248 0.226 0 1

LongT ermCF O 2,261 0.243 0.220 0 1

Panel D: Firm-level financial controls

CF O V olatility 3,540 0.0257 0.0217 5.95e-06 0.209 ROE 4,775 0.209 4.029 -18.93 248 Leverage 4,563 0.247 0.153 0 1.514 Liquidity 3,704 1.412 0.841 0.217 19.64 OperatingCycle 3,422 -1,375 578.4 -4,554 76.17 Oscore 3,293 -2.402 2.649 -17.82 75.20 Size 4,758 9.359 1.551 1.411 15.63

Panel E: Country-level controls

GDP 4,406 10.63 0.384 7.998 11.54

%GDP 4,406 0.0133 0.0237 -0.0827 0.263

ST KDV LP 4,406 0.868 0.446 0.111 2.125

ADRI 4,406 0.519 0.269 0.204 0.927

there seems to be considerable variation in country time-orientation, thus warranting the examination of differences in national culture as a possible explanation of this variation.

dimensions are included, as are alternatives for the compensation measure and the investor time orientation measure. Furthermore, three dummies are included, which are used in the additional analysis below. From panel A which contains the national culture variables, we can see that on average the countries in this sample have a relatively high score on most dimensions, especially on individualism. Comparing this to for example Shao et al. (2013) who have an average of 52 on the individualism dimension, this is probably a consequence of including mostly developed Western-European countries in this sample. Comparing the averages to Hooghiemstra et al. (2015) this sample has (again) higher individualism and lower uncertainty avoidance. Most likely due to similar reasons. The descriptive statistics of the cultural diversity index are not that informative, however, the standard deviation shows that there is considerable variation in the degree of cultural diversity among the investor base. From the data, this seems likely. Some firms are wholly owned by domestic investors, as the F or Investor dummy shows. This results in a cultural diversity score of 0, hence it was expected that there would be a large variation in this variable. Finally, we can observe that about 75% of the firms in this sample have at least one foreign investor. Moreover, if there is a foreign investor, about 60% is from the United States, and almost 15% has its roots in the United Kingdom. Although past research has indicated that Anglo-Saxon type capital markets such as the US and the UK tend to focus on the short-term (Jacobs 1991; Porter, 1992), the descriptive statistics of the short-termism measures already revealed that in this sample especially firms from the United Kingdom are focused on the long-term. In the Additional analysis section, these dummies will be used to examine the effect of foreign (Anglo-Saxon) investors.

Next, the set of control variables is presented in panels C-E. Panel C presents firm-level variables that have prominently featured in previous research and present an opportunity to compare whether the same influences as in the United States are present in Europe. More-over, since data limitations required new definitions of both the time orientation of investors and the time orientation embedded in the managerial compensation packages, several vari-ants of these measures are included. Starting with investor time orientation, we can see that more than 90% of investors have held their shares for at least one year. In fact, almost 70% holds their shares for at least 5 years, and a still impressive 37.9% bought shares more than 10 years ago. Hence, the investors known in the Thompson Reuters Eikon database have in general a fairly long holding period. Thus, it might be difficult to identify the effects of long-term shareholders on corporate short-termism. To examine this, appendix D contains table 15, a check to see whether the different specifications of this variable affect short-termism. The Analysts variable is log-transformed and is hence less informative in this form. Finally, the compensation variables are specified, again four different variants are available, as discussed in the Methodology section. As the means and standard devia-tions indicate, there appears to be little difference between the measures. Therefore, it is probably not that important which one is employed, although the LongT ermExec measure seems theoretically the most appropriate one. Furthermore, one important limitation of all these measures is the small number of observations available, which is more severe for the individual executive measures. Ideally, one would like to exclude these measures to save observations, due to the importance of equity incentives in past research on short-termism this is however not feasible (Brochet et al., 2015).

Offerein (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (1) T imeHorizonB 1.00 (2) T imeHorizonD -0.49 1.00 (3) LT O 0.02 -0.08 1.00 (4) IN D -0.25 0.33 -0.18 1.00 (5) U A 0.12 -0.20 0.42 -0.68 1.00 (6) CD Investor 0.06 -0.03 0.26 -0.15 0.11 1.00 (7) LongT ermInv5 -0.03 0.10 -0.09 0.09 -0.09 -0.12 1.00 (8) Analysts 0.18 -0.11 0.12 0.01 0.04 0.16 0.21 1.00 (9) LongT ermExec 0.09 -0.03 0.12 -0.08 -0.02 0.17 0.13 0.11 1.00 (10) ROE -0.11 0.06 -0.08 0.08 -0.18 -0.08 0.02 -0.01 -0.01 1.00 (11) Leverage -0.03 0.03 -0.05 -0.03 0.02 -0.04 -0.00 -0.07 -0.04 -0.03 1.00 (12) Size 0.28 -0.18 0.09 -0.28 0.11 -0.06 0.11 0.33 0.13 -0.07 -0.09 1.00 (13) GDP -0.01 0.14 -0.01 0.42 -0.46 0.14 0.14 0.18 0.17 0.02 -0.02 -0.07 1.00 (14) %GDP -0.03 -0.01 -0.03 -0.02 -0.09 0.00 -0.00 -0.07 0.03 0.15 -0.08 0.01 -0.04 1.00 (15) ST KDV LP -0.21 0.29 -0.02 0.35 -0.36 0.10 0.04 -0.01 0.02 0.11 -0.08 -0.19 0.47 0.06 1.00 (16) ADRI -0.34 0.36 -0.50 0.68 -0.58 -0.29 0.05 -0.21 -0.12 0.13 -0.00 -0.30 -0.02 0.05 0.28 1.00 In this table correlation coefficients between the variables are presented. Note that due to the inclusion of LongT ermExec this

correlation table is based upon a reduced sample. To prevent further reduction, several control variables are omitted. The same holds for the alternative variables available for LongT ermExec. Moreover, the alternative measures for investor time orientation, the GLOBE culture dimensions and some variables used in the additional analysis are omitted to save space and for formatting purposes. As a rule of thumb, all correlations with an absolute value ≥ 0.04 are significant at the 5% level.