CEO Tenure, Corporate Social Performance, and Corporate Governance: A Korean Study

: The purpose of this study was to examine the association between Chief Executive O ﬃ cer (CEO) tenure and corporate social performance with the moderating e ﬀ ect of governance. We investigated whether new CEOs and CEOs in their last year of service were more focused on short-termism than CEOs of other periods. Speciﬁcally, we tested whether these CEOs reduced social performance that demands immediate expenditure and expect payo ﬀ s in the long run. We also tested whether good governance can mitigate such behaviors, because not all CEOs of the same tenure will act the same, depending on the monitoring environments surrounding them. We employed ordinary least squares (OLS) method and the moderator models using data from the Korean listed companies from 2012 to 2016. Test results showed that only the CEOs of their last year reduced social performance. However, when we considered corporate governance, we found that both groups of CEOs reduced social performance, and that good governance mitigated the adverse e ﬀ ects of the two periods on Corporate Social Responsibility (CSR). Speciﬁcally, we tested board independence, board frequency, CEO duality, and board diversity, and found that, for all but board independence, the negative e ﬀ ects of the two periods on social performance were decreased.


Introduction
Prior research has studied the influence of CEO tenure on earnings manipulation, performance, or audit fees [1][2][3][4][5][6][7][8][9][10][11]. A group of studies that investigated the earnings manipulation of the first year of CEO tenure generally have founded "big bath" [12][13][14][15], whereas studies on CEOs' short-termism during the last year of his/her tenure showed inconsistent results [1][2][3][4][5][6]11,[15][16][17]. Unlike the big bath studies that generally focused on the first year of a new CEO, Ali and Zhang [8] tested the early three years and the last one year and concluded that in both periods CEOs manipulate earnings upwards. Limited studies on CEO tenure have investigated effects other than earnings management, such as audit fees, industry specialist auditors, and board monitoring [7,9,10,18]. Also, only a limited number of studies have considered CEO tenure in studying its impact on corporate social responsibility (hereafter, CSR); however, except for one recent paper [19], researchers have instead focused on different aspects (CEO power [20,21] or CSR disclosure [22]).
The purpose of this study is to discover the impact of managerial behaviors on current discretionary investments that are expected to benefit in the long run when a CEO is mostly motivated to pursue short-term goals. Specifically, this study tested the effect of CEO tenure on CSR in terms of its early years and last year, following Ali and Zhang's [8] research design. Unlike Chen et al. [19] who tested

The Early Stage of CEO Tenure
The early or later stage of CEO tenure has been the interest of many researchers. As for the first year, because compensation contracts are not expected to consider the first year of a new CEO, and a new CEO is assumed not to be responsible for the past performance, an incoming CEO may have an incentive to minimize earnings by deferring new income and recognizing write-down as much as possible [16,23]. This aggressive recognition of losses as the predecessor's responsibility positively affects the successors because it lowers the benchmark point against which the new managers will be evaluated [13,23]. Studies have discovered this "big bath" behavior for the first year of CEO tenure [2,[12][13][14][15][16]23,24]. For example, in DeAngelo's [14] paper, the new management tends to decrease earnings because the former management is to be blamed and because such a bath allows a big rebound the next year. Elliott and Shaw [13] reported that one-third of write-offs reported as a special item occurred during CEO changes. Goodwill impairment also occurs in the early years of CEO tenure [24]. As this research generally focused on the new CEO's downward earnings management or first-year write-offs, relatively fewer studies have investigated the CEO's behavior after the initial year. Although the former CEO can explain the first-year performance, a new CEO has a motivation to increase earnings to dispel skepticism about his/her ability to successfully manage the firm. On the basis of this idea, Ali and Zhang [8] assumed that a CEO's early years would be related to aggressive income recognition. A different view comes from Chen et al. [19], who suggested that at the early stage CEOs may not only try to signal their ability by showing short-term performance, but also will have incentives to invest in long-term projects because enduring initial investment can pay off later, which makes it easier for the CEO to achieve excellent performance later in his/her tenure. Test results supported both views. In Ali and Zhang's [8] research, CEOs increased discretionary accruals in their second and third years. In Chen et al.'s [19] test, during their second and third years, CEOs achieved higher CSR scores than the remainder of their tenure. Chen et al.'s [19] approach was in line with the idea that CSR is a form of investment that provides positive future cash flow [25].

The Final Year of CEO Tenure
Another group of researchers has focused on CEOs' last year [1][2][3]5,8,15,16,26,27]. As earning-based performance measures induce a CEO's short-termism, in his/her final year the CEO with earning-based compensation tends not to consider long-term investments such as R&D [1]. Another explanation is the external carrier concern. A positive relationship exists between firm performance during the pre-retirement period and the possibility of getting post-retirement jobs [26], which implies that the carrier concern may cause CEOs of the last year to boost firm performance as they initially did. Some studies have focused on the mandatory retirement at the age of 65 [1][2][3]26,27], not only because the retiring CEO does not need to consider his/her carrier anymore but also because such departures are neutral from dismissal due to poor performance.
Although Murphy and Zimmerman [2] denied this "horizon problem" of a CEO's last year by showing that such a decrease in discretionary expense only appears when performance was poor, many pieces of research have joined to explore the possibility of losing a long-term perspective just before CEO departure. However, as early researchers such as Dechow and Sloan [1] and Murphy and Zimmerman [2] did not agree with each other, later studies also failed to find consistent results. From a non-routine CEO change sample, Pourciau [15] discovered an earning-decreasing accrual and write-offs not only in the first year but also in the final year of a CEO's departure. Wells [16] also failed to find the final year's earnings management behavior. Kalyta [3] discovered a horizon problem when the CEO had a retirement benefit based on stock performance around the last years.
Hazarika et al.'s [5] study approached the subject from a different perspective. They showed that earnings management causes forced CEO turnover; therefore, the last year of the replaced CEO has a higher level of earnings management than the periods of the successor CEO. According to McClelland et al. [6] and Antia et al. [4], the horizon problem is related to the tendency to maintain the status quo. Their studies empirically proved that CEO's long tenure and older age create a horizon problem in terms of higher agency cost, lower accrual quality, and deteriorated future performance.

Hypothesis
Stakeholder theory assumes various levels of stakeholder involvement, from managers' mere consideration of some stakeholders to active stakeholder participation [28], implying that each firm's level of CSR varies depending on the environment. Each CEO's ethical attitude is also affected by environmental factors such as financial performance and the pressure from shareholders [29].
Although a CEO's tenure is a continuum with a beginning and an end, except in the case of Ali and Zhang [8], not many studies have empirically considered the initial multiple years of CEO tenure and terminal stage together. Instead, studies generally have focused on what changes between the former and successor CEOs. Ali and Zhang [8] showed that earnings management was higher both in the first three years and the last year. As a manager who intends to increase earnings will try to control Sustainability 2020, 12, 99 4 of 17 not only accrual recognition but also a firm's real activities [30], we may expect that CSR spending will be reversely related to earnings manipulation in this period of shorter management horizon. Moreover, Ali and Zhang [8] also showed that CEOs of those periods spend less discretionary expenses. Given that the involvement in CSR activities as part of the long-term investments eventually requires certain costs, expenditure, or any form of cash outflow, we may expect that a CEO who urgently needs to maximize his/her short-term performance will not easily devote his/her firm's resources to CSR activities.
Up until now, only one study has tested the relationship between tenure and CSR performance [19]. Chen et al. [19] argued that tenure and CSR have a negative linear relationship, implying that CEOs initially invest more in CSR than later years; however, their results were inconsistent with implications from Ali and Zhang [8] and other studies. The literature has shown that CEOs tend to manipulate their performance in the first few years. For example, Pourciau [15] discovered a clear pattern of big bath for the first year of CEO tenure as lower earnings and more write-offs were observed. Their second year was a year of great reversal-a significant increase in earnings and near-normal level of write-offs. In Ali and Zhang's [8] study, discretionary accruals were higher in the second and third years of CEO tenure compared with other periods, and abnormal discretionary expenses were smaller in the first three years. Because the market and board heavily rely on current performance to evaluate a new CEO, and because even a competent CEO may lose his/her opportunity as his/her first performance was insufficient, a new CEO has an incentive to boost his/her performance in the early stage of tenure [8]. According to Kuang et al. [31], because CEOs just hired outside the company have more incentive to prove their abilities while having less expectation to survive, they increase earnings in the early years of their tenure until they attain enough of a reputation to quit their short-term perspective. Because earnings manipulation is negatively related to corporate social performance [32][33][34], we may expect that the CSR level will be low at the early stage of CEO tenure. If CSR is a type of investment [25], managers who want to maximize their short-term profits may be reluctant to spend more CSR costs at an early stage. Even a new CEO wants to act morally, his/her limited power and lack of firm-specific knowledge may provide a dispersal of responsibility and information asymmetry to the new CEO, which makes him/her hesitate to initiate new decisions [35].
However, Chen et al. [19] argued that CSR, as a form of long-term investment, can work as a means to dispel skepticism about the new CEO's ability. Their implicit assumption is that the corporate governance structure generally influences a new manager to have a long-term perspective. As a result, in order to achieve future stable performance, a new CEO may initiate long-term investments at his/her initial stage. However, Chen et al.'s [19] argument does not explain other reported behaviors of new managers, such as the first-year "big bath" and subsequent short-term income-increasing earnings management reported in previous studies [8,[12][13][14][15]31]. Such behaviors of the new CEO do not align with shareholder interests, implying that a CEO who is newly employed and whose position is not yet secured may lack a long-term perspective that supports shareholder value. For such CEOs who need immediate performance at the beginning of their tenure, discretionary expense minimization might be another primary concern. We believe that such CEOs cannot afford to be interested in long-term investments with uncertain returns, such as CSR. Chen et al.'s [19] interpretation also cannot explain the differential impact of governance on CEO behavior. Studies have found that not all governance has the same influence on earnings management [36][37][38] or CSR [39,40]. It is reasonable to assume that only good governance effectively monitors its new agent to act for shareholders from the beginning.
Beyond only their early years, a CEO may also show short-termism in his/her last year of tenure due to his/her loss of career concern. As a CEO approaches retirement, his/her concern about his/her next employment decreases [1,[41][42][43]. Another explanation is related to corporate performance. As an opportunistic CEO tries to avoid the possibility of being fired, he/she may try to maximize short-term profits through opportunistic accounting decisions [15] or adjusting the firm's real activities [1]. Even honest CEOs with poor performance may use earnings management to convey insider information that the firm's real prospect is not as reported [15]. No studies have tested the effect of the horizon Sustainability 2020, 12, 99 5 of 17 problem on CSR; however, we can speculate that a CEO who wants abnormal earnings will try to minimize discretionary costs such as CSR.
Although prior studies have failed to provide consistent results on the horizon problem [1][2][3][4][5][6]11,[15][16][17], Ali and Zhang [8] showed that, only after they considered the early years and the last year together in one model, the final year's earnings management is positive and statistically significant along with the initial periods' discretionary profit maximization. Therefore, we employed Ali and Zhang's [8] model and applied it to a Korean sample.
Even though Korea is one of the developed countries, its unique position in Asia may offer the possibility that the executives behave differently from the West. Prior studies support this idea by showing that the legal system and culture influence accounting practices [44,45] and CSR [46][47][48]. However, results may be a matter to be confirmed empirically. Specifically, according to Ringov and Zollo [46], power distance and masculinity are negatively related to CSR. However, Ho et al.'s [48] study showed that power distance and masculinity increase CSR. Moreover, Ho et al. [48] and Thanetsunthorn [49] showed that collectivism and uncertainty avoidance facilitates CSR, whereas Ioannou and Serafeim [50] concluded that individualism is related to a high level of CSR. Because we cannot draw consistent conclusions about the effects of culture, although Korea is classified as low power distance, low masculinity, and high uncertainty avoidance, and Korea's collectivism was once very high but has recently declined [51], it is uncertain how cultural features affect Korea's CSR differentially from the West. We assume that as one of the highly opened and developed economies, there is no fundamental reason that the current Korean companies act differently from the United States. Therefore, we expect that the relationship between tenure and CSR of Korean companies would not be different from what Ali and Zhang's [8] earnings management study on the U.S. data implies. As CEOs have a motivation to maximize their short-term performance in their first years and at the last year of tenure, we assume that not only the discretionary accrual and discretionary expense show differential characteristics, as reported in Ali and Zhang [8], but also that a firm's CSR will be affected. As CSR is a type of long-term investment or at least requires some level of discretionary expenditure, CEOs who need to maximize short-term performance, in other words, CEOs of their early years and the terminal year, will decrease their corporate commitment to CSR activities. Hypothesis 1. CSR during the first few years and the last year of a CEO's tenure is lower than the CSR of the CEO's other periods.
Prior studies have shown that good governance mitigates agency problems. Beasley [36] showed that more independent boards decrease the probability of fraud. According to Xie et al. [38], boards and audit committees with more active and financially knowledgeable members are associated with lower earnings management. Other studies have demonstrated that good governance decreases the cost of capital [52,53], increases the firm value [54,55], and provided stability during the global financial crisis [56]. Studies also have proven that good governance structure improves firms' CSR activities [57][58][59]. As the monitoring mechanism that solves the agency problems, it is expected that good governance enforces a CEO not to keep short-termism; as a result, CSR investment will no longer be a reluctant choice for a CEO immediately after his/her new job or just before his/her resignation.
Ali and Zhang [8] showed that strong monitoring reduces the positive effects of early years on earnings management. They used institutional ownership, analyst following, board independence, and audit committee independence and found that the coefficients of the intersection variables between the first few years dummy and the governance variables were negative and statistically significant. As more earnings manipulation means less CSR investments [32][33][34], we expected a similar moderating effect of governance on the relationship between tenure and CSR. Unlike Ali and Zhang [8], who did not test the governance effect for the last year, we predicted the moderating effect of governance also on the horizon problem. Chen et al. [19] also confirmed the governance effect by testing the intersection between tenure and board independence. However, Chen et al.'s [19] Table 5 actually showed that only after they considered governance, their hypothesis that generally tenure and CSR has a negative linear relationship can be confirmed, because after introducing the intersection variable between governance and tenure, the original negative effect of the CEO tenure on CSR is then reversed to the statistically and economically significant positive effect. Therefore, we will predict our second hypothesis by referring to Ali and Zhang [8] rather than Chen et al. [19].

Hypothesis 2.
Good monitoring mitigates the negative effect of the first few years and the last year of CEO tenure on CSR.

Model and Data
The empirical model is a modification of Ali and Zhang's [8] research design. CSR = α + β 1 ceofirst + β 2 ceolast + β 3 size + β 4 lev + β 5 roa + β 6 cur + β 7 growth + β 8 age + β 9 maj +β 10 for + β 11 mjda + β 12 ten + β 13 big + β 13 audithour + ε This model tests whether the CSR levels of the initial years (ceofirst) and the last year of CEO tenure (ceolast) are different from the other years, after controlling for corporate and audit-related factors that may affect the level of CSR investment. On the basis of Ali and Zhang's [8] earnings management research that discovered positive signs in both periods, we believe that it is reasonable to expect that the coefficients of both dummy variables (ceofirst and ceolast) are negative. The ceofirst variable represents the first two years or three years and the ceolast variable means the last year of CEO tenure. For ceofirst we tested the first two years and three years together. For the dependent variable (CSR) we used KEJI (Korean Economic Justice Index), the most commonly used CSR index in Korea. The KEJI is provided by the Institute of Economic Justice and in 2017 they disclosed their 26 th announcement of their index and related awards for outstanding companies. Their 2017 awards were for CSR activities in 2016, which contained the KEJI index until 2016.
The control variables were company-or auditor-related characteristics that may influence corporate's CSR decision, such as basic characteristics of firm/auditor (size, age, big), debt burden (lev), profitability, liquidity and growth (roa, cur, growth), and outside monitoring devices (maj, for, big, ten, audithour). For audithour we used the partner hours because partners are the most influential position in terms of deciding the level of audit quality [60]. We included the CEO's earnings management measured by modified Jones model [61] to control the effect of management choice between aggressive financial reporting and CSR investment (mjda).
For corporate governance, we employed the characteristics of inside monitoring such as board independence, board meeting frequency, CEO duality, and board diversity. Board independence (independence) is the percentage of outside directors on the board. Board meeting frequency (frequency) is the natural log of the number of board meetings. For independence and frequency, we tested the dummy versions also, giving 1 when the variables were above the median.
CEO duality (duality) is a dummy variable giving 1 to the CEO who is also the chairman of the board. Board diversity (diversity) is a dummy variable giving 1 to the board that contains woman directors. Because Korean companies often do not include female directors on the board and the companies that employ female directors on the board commonly do not hire multiple female directors, the original data were already close to dummy variables. Therefore, we included only the dummy version of the variable in our test. All variables are defined in Appendix A.
The data consists of 332 firm-years that include companies listed in the two Korean stock markets (Korea Composite Stock Price Index, or KOSPI, and, KOrea Securities Dealers Automated Quotation, or KOSDAQ) during 2012-2016. Although the two stock markets contain around 1400 companies each year, data availability was decreased due to the nature of the KEJI, which measures the CSR activities of 300~400 companies each year. We have a complete list from the institution; however, the list was Sustainability 2020, 12, 99 7 of 17 before their 2018 year-end announcement, which contains the CSR index until 2017. Therefore, we could not include the CSR score for the year 2017 and beyond. Another significant limitation was some of our governance and audit-related measures. The hand-collecting characteristic of these variables limited our time horizon to within four years and decreased our data size. We used ordinary least squares (OLS) method for all models and utilized the intersection variables for the moderator models. Each model is explained below each corresponding table. Table 1 show descriptive statistics for the variables our test model employs. CSR is the natural log of KEJI and showed a mean of 4.120. For ceofirst, we used two versions: ceofirst1 for the first two years of CEO tenure, and ceofirst2 for the first three years. The table shows that the first two (three) years of CEO tenure occupied 16.4% (29.1%) of the total CEO tenure. The variable ceolast represents the last year of CEO tenure and corresponded to almost 10% of the total tenure. Considering that the mean value of CEO tenure (unreported) was around five years, we can assume that the missing values were removed relatively more frequently in the early and late years of CEO tenure. The mean value of independence shows that independent directors made up one-third of the board on average. The dummy version of independence (independence_dum) had a mean value of 0.620, implying that around the median value of independence (0.333) many companies shared the same ratio. The variable frequency is a natural log of the number of board meetings, and its dummy version (frequency_dum) showed that 53.9% of total samples were above the median (2.398). The table also shows that 69% of CEOs also served as chairman of the board (ceo_duality), and only 13.3% of companies had female directors on the board (diversity). As some variables in Table 1 are not the raw data but the natural logarithm version to enhance data distribution and coefficient interpretation, care should be taken as they do not show the original levels. The correlation matrix in Table 2 shows that there was no significant relationship among CSR, ceofirst, and ceolast, implying that CSR was not different in the beginning and ending periods of tenure, contrary to our expectation. Table 2 also shows that only some governance variables, such as independence and ceo_duality, were positively related to CSR. However, these correlations could be superficial because other factors that influence each variable were not controlled. The relationships between CSR and control variables that affect CSR were positive for size, roa, cur, and for, indicating that more prominent and profitable firms, as well as firms with foreign ownership, tend to have better CSR performance, and were negative for lev, age, and maj, meaning that debt financing, matured firm life cycle, and ownership concentration reduced CSR investment. Table 3 shows the test results for hypothesis 1. In columns 1 and 4 of Table 3, we failed to find statistically significant results for the early years of CEO tenure as the two versions of ceofirst (ceofirst1: the first two years and ceofirst2: the first three years) did not show different results. In contrast, we found that in the last year of CEO tenure (ceolast) CSR was decreased by 1.3% for one unit change of the independent variable, regardless of whether or not the two versions of ceofirst were included in the regression. These results were not consistent with the implication from Ali and Zhang [8] and might be in line with Chen et al. [19]. Although Ali and Zhang's [8] early years variable was defined to mean the first three years of CEO tenure, their yearly test results showed that the second year had the strongest effect followed by the third year, with no statistically significant results for the first and the other years of CEO tenure. Because our test results of insignificantly positive or negative beginning and significantly negative ending are more consistent with Chen et al.'s [19] negative linear relationship, we tested the linear model and some other dummy models (Table 4). Due to space constraints, we did not report test results of control variables from Table 4.  1. Note: t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1; variable description is avaialble in Appendix A. Test model is as follows: CSR = α + β 1 ceofirst + β 2 ceolast + β 3 size + β 4 lev + β 5 roa + β 6 cur + β 7 growth + β 8 age + β 9 maj + β 10 for + β 11 mjda + β 12 ten + β 13 big + β 14 audithour + ε. Column headings represent dependent variable and choice of independent variables. Industy and year effects were controlled for but not reported in the model or as numbers. 1. Note: t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1; Variable description is available in Appendix A. Test model for the first column is as follows: CSR = α + β 1 ceotenure + β 2 size + β 3 lev + β 4 roa + β 5 cur + β 6 growth + β 7 age + β 8 maj + β 9 for + β 10 mjda + β 11 ten + β 12 big + β 13 audithour + ε. Test model from the second columns are as follows: CSR = α + β 1 secondyeart (thirdyear) + β 2 ceolast + β 3 size + β 4 lev + β 5 roa + β 6 cur + β 7 growth + β 8 age + β 9 maj + β 10 for + β 11 mjda + β 12 ten + β 13 big + β 14 audithour + ε. Columns represent dependent variable and each option of independent variables. Industy and year effects were controlled for but not reported in the model or as numbers. Table 4 column 1 shows that Korean list companies' tenure and CSR did not have a linear relationship. In the second year of CEO tenure, CSR declined but was statistically insignificant (columns 2 and 3). Test results for year three were also statistically insignificant, but this time coefficients were positive (columns 4 and 5), implying that CEOs' short-termism begins to stop in the third year. For all the test results in Table 4, celoast were statistically significant, but all variations of ceofirst were not. Finally, in an unreported analysis due to space unavailability, the first year dummy was insignificant with β = −0.001, t = −0.073. Therefore, we decided to keep our current selection of variables of interest, the first two or three years. Table 5 focuses on hypothesis 2, which assumes that governance mitigates the adverse effects of ceofirst and ceolast. As seen in panel A, for the two versions of board independence we failed to find any moderating effect of governance. Moreover, in panel A both ceofirst and ceolast lost their statistical significance. For all the tables hereafter, as the test results of ceofirst1 and ceofirst2 were very similar, we show test results of ceofirst1 only.  1. Note: t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1; Variable description is available in Appendix A. Test model is as follows: CSR = α + β 1 ceofirst + β 2 ceolast + β 3 governance + β 4 ceofirst × governance + β 5 ceolast × governance + β 6 size + β 7 lev + β 8 roa + β 9 cur + β 10 growth + β 11 age + β 12 maj + β 13 for + β 14 mjda + β 15 ten + β 16 big + β 17 audithour + ε. For ceofirst Table 5 reported test results of ceofirst1 only; for governance Table 5 provied test results of independent, indepependent_dum, frequency, frequency_dum, ceo_duality, and diversity. Columns represent dependent variable and each options of independent variables; industy and year effects were controlled for but not reported in model or as numbers.

Multivariable Regression Analysis
Panel B of Table 5 shows that board meeting frequency mitigated the negative effects of ceofirst and ceolast, consistent with hypothesis 2. In panel B, coefficients of ceofirst and ceolast are shown to be negative, the direction of decreasing CSR. However, the joint effect of meeting frequency and tenure influenced CSR in a positive direction, offsetting the negative effect of ceofirst and ceolast to some extent.
Panel C of Table 5 also provides test results for the remaining two governance variables, ceo_duality and diversity. Unlike frequency on Panel B in which both ceofirst and ceolast had statistically significant test results, the ceo_duality model from columns 1 to 3 of panel C shows the statistically significant results for ceofirst only. Nevertheless, the interpretations of the empirical results of frequency and ceo_duality were identical. First, a tenure variable interfered with CSR, and next, a governance variable emerged to stop it. In the case of ceo_duality, the positive values of coefficients were large enough to cancel all the negative effects of ceofirst.
The diversity model from column 4 to 6 of Table 5 panel C shows the statistically significant results for ceofirst only. This time we might have to slightly change our interpretation. Governance in terms of diversity facilitated CSR, and the early years' effect mitigated it. In general, not all the variables of interest in the test models were statistically significant; however, it was common for most of the models to share the general characteristics of a decreasing pattern for the early and/or last years of tenure and an offsetting pattern when governance was considered.

Conclusions
In this paper, we tested the impact of CEOs' early and late years on CSR performance by examining the OLS model in which the influences of the early years and the last year dummy on the KEJI index score were tested. The early years of CEO tenure are the time of a CEO's short-termism because of his/her vulnerable position [8,13,14,16,23,31]. In the final year, a CEO may also lose his/her long-term perspective due to either his/her career concerns or loss of concern [1,15,[41][42][43]. Ali and Zhang [8] tested the two periods and discovered that in both periods CEOs increase discretionary accruals and reduce discretionary expenses. As an increase in earnings management decreases CSR [32][33][34], and CSR investment can be one type of discretionary expenditure CEOs want to minimize at the time of short-termism, we expected that the effect of these periods on CSR would be similar to that of Ali and Zhang [8]. Because a manager who wants to maximize his/her short-term profit may not have room for allowing expenditures in a long-term project such as CSR, CSR investment will be decreased.
As expected, we discovered that CSR was decreased around the two periods, the beginning and ending of CSR tenure, and good governance can mitigate this occurrence, which is consistent with what can be inferred from Ali and Zhang [8].

Discussions
There have been two recent studies that provide different implications on the opportunism by CEO tenure [8,19]. In this study, we followed Ali and Zhang [8] and did not accept Chen et al. [19]. One reason was because of the test result consistency. Our results are different from Chen et al. [19] because we failed to find the linear relationship between tenure and CSR described in Chen et al. [19] while obtaining test results consistent with Ali and Zhang [8]. Another reason is the theory consistency. Chen et al.'s [19] model, which claims that new CEOs do more CSR than older CEOs, cannot explain the opportunistic behaviors of the newly arrived managers described in prior studies [8,[12][13][14]31]. In contrast, Ali and Zhang's [8] work is in line with the literature. Our last reason as to why we accept Ali and Zhang's [8] model is that their model also considered the horizon problem, the short-termism peculiarity that occurs around the last-year [1][2][3][4][5][6][7].
We believe that Chen et al. [19] did not consider the influence of governance enough. Their Table 5 shows that only after considering the governance and tenure effects together, their main finding appeared that claims a linearly negative impact of tenure on CSR. After separating the influence of governance, their coefficient of tenure instead turned into a positive value, opposite to their main claim. A more realistic interpretation is that only firms with strong governance will be able to minimize the short-termism of CEOs in their first and final years. We believe that to test such consideration, Ali and Zhang's [8] model that assumes a non-linear relationship fits the purpose of the study better.
Our findings, which are more consistent with the theories previous studies are based upon, will contribute more to the literature. For example, as our findings imply that CEO turnover is related to a lower level of CSR activity, and that good governance mitigates the underinvestment in CSR, stakeholders such as civil activists and regulators who expect a higher level of social involvement from the company may be able to know which companies require higher levels of governance. The limitation of this paper is that we did not thoroughly investigate the moderating effects of governance. Only after knowing which part of governance affects the extent of the relationship can current findings give practical implications completely. Moreover, we did not try to answer the reason as to why previous studies have reported different results. We expect that further studies may determine the cause of inconsistency among the literature.
Author Contributions: J.H.C. provided the research idea, performed the main literature review and the main data analysis, and wrote this paper. S.K. reviewed the entire process, provided the literature review, and wrote this paper. A.L. prepared the data, was involved in the initial idea development and initial literature review, and completed the final review of this paper. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding. governance a continuous or dummy variable that shows a level of corporate monitoring independent the ratio of outside directors to the total number of directors in the board independent_dum a dummy variable 1 if a firm's independent is above the median in the same industry frequency natural log of the number of board meeting during a year frequency_dum a dummy variable 1 if a firm's frequency is above the median in the same industry ceo_duality a dummy variable 1 if a CEO is also the chairman of the board of directors diversity a dummy variable 1 if a company's board contains woman director(s)

Control variables:
size natural log of total asset lev total liability divided by total assets roa net income dividend by total assets cur current assets dividend by current liabilities growth the difference between this year's sales and last year's sales, divided by last year's sales age natural log of the difference between current year and the founded year plus 1 maj natural log of the largest shareholder's ownership percentage multiplied by 100 for natural log of the foreign ownership percentage multiplied by 100 mjda discretionary accrual measured following the method of Dechow et al. [58] ten natural log of auditor tenure, the number of annual contracts renewed between the company and the auditor big a dummy variable 1 if the auditor who signed the audit contract with the company is one of the big four large accounting firms audithour partner hours spent and reported in the audit report (required by Korean regulation)