CAN OVERSEAS INVESTMENT IMPROVE EARNINGS QUALITY?

Transcription

1 CAN OVERSEAS INVESTMENT IMPROVE EARNINGS QUALITY? ABSTRACT Junjian Gu* Nagoya University Dan Hu* Nagoya University It is widely assumed that foreign investment is good for companies in several ways. However, there are few studies addressing whether overseas investment can benefit companies earnings qualities. Using a sample of Japanese firms overseas investments in global markets, this paper tries to find the relationship between overseas investment and companies earnings qualities. In this paper, we use the propensity score matching model to control for differences in firms characteristics between overseas investment Japanese firms and non-overseas investment Japanese firms and regression models to test 6 hypotheses on the associations among earnings qualities, legal systems, and degrees of development. We find that although overseas investment can reduce earnings management risk, it can also impair reporting conservatism. Further analyses show that firms in common law or developed countries have higher level of earnings qualities. JEC Classifications: M41, D92, Q56, K42 Keywords: overseas investment, earnings quality, propensity score matching Corresponding Author s Address: gu.junjian@f.mbox.nagoya-u.ac.jp INTRODUCTION Nowadays, globalization of local firms is an important topic for every country, as it is one of the main drivers of economic growth. As of 2015, 32 Japanese firms are listed on overseas stock markets (17 in the U.S., 11 in European markets, and 7 in Asian markets). In addition, around 1,100 listed Japanese firms have overseas investment in 110 countries. The total number of Japanese overseas subsidiaries in 2014 reached 26,100; in which the compound annual growth rate (CAGR) of the subsidiaries is 4.7% from Based on this trend, it can be expected that more local Japanese firms will expand into global markets. What reasons lead so many local firms to expand into global markets? We can answer this question from two perspectives. From the firms perspective, overseas investment can be an effective way to enter into a foreign market and acquire important natural resources (e.g. precious metals, fossil fuels, etc.). Overseas investment is also a mean for companies to reduce their production cost if labor market is cheaper and regulations are less restrictive in the target foreign market. From the foreign country s perspective, overseas investment can provide a source of external capital and higher revenue, which can lead to economic development. Overseas investment can also expose national and local governments, local businesses, and citizens to new business practices, management techniques, economic concepts, and technology that will help them develop local businesses and industries. However, from investors perspective, is this a good phenomenon? Those international firms seem to have a high reputation, but does it work for investors? Historically, international firms that had a high reputation among investors have been involved accounting scandals such as Enron (2001) 2, WorldCom (2002) 3, and Saytam (2009) 4 before the exposure of terrible news. These cases raise investors suspicion that international firms do not always have high reputations or high quality accounting information. Investors should focus on some important indexes of earnings quality, helping themselves to decide the quality of companies accounting statements which lead to a better investment decision. This paper aims to identify the relationship between overseas investment and earnings qualities from the perspective of countries legal systems and degrees of development. Firm managers must use specific accounting standards when issuing financial statements. But even with those rules, it is still hard to determine when and how much accounting manipulation has taken place. There is enough noise in the accounting data to influence earnings quality. There are several methods to measure earnings quality. Dechow et al. (2010) in their review paper discuss specific proxies for earnings quality. They discussed three categories of earnings quality proxies: earnings properties (i.e., persistence, residuals from accrual models, smoothness, and target beating), investor responsiveness to earnings, and external indicators of earnings misstatements. This paper will use several variables mentioned above as proxies of earnings quality to find whether overseas investment improves earnings qualities. La Porta et al. (1998) find that common-law countries (e.g. United States, United Kingdom, Australia, etc.) generally have the strongest legal protection for investors, while French-civil-law countries (e.g. France, Brazil, etc.) and German and Scandinavian-civil-law countries (Japan, Germany, Finland, etc.) have weaker 737

2 legal regimes. They document that common law countries have a higher level of protection for minority shareholders compared to civil law countries. In conclusion, common law countries show evidence of superior investor protection and thus contribute to more mature and stable capital markets. Based on these findings, legal systems are linked with the level of legal protection for investors, which may lead to higher or lower earnings qualities. In this paper, we will try to test the relationship between overseas investment and earnings qualities with these two legal systems (i.e. common law system and civil law system). Solow (1956) suggested that the lack of capital in developing countries was the principal cause for their poor economic growth compared to developed ones. Low growth is the outcome of weak fundamentals and a shaky institutional structure. Rajagopalan et al. (2014) indicate that investors protection and legal origin play a significant role in the equity market volatility for all countries. High transparency coupled with low corruption creates more confidence amongst investors in developed countries as compared to emerging markets and this reduces market volatility. Based on these findings, the different degree of development is linked with weak or strong fundamentals and institutional structure, which may lead to higher or lower earnings qualities. In this paper, we will further test the relationship between overseas investment and earnings qualities from the perspective of target countries degree of development. Recent accounting research suggests the potential threat of selection bias in research when researchers compare large and small accounting firms that have different characteristics. To relieve the selection bias problem, we will use propensity score matching (PSM) model, a rigorous matching method, to control for differences in firm characteristics of paired samples between Japanese firms with overseas investment and Japanese firms without overseas investment. Empirical research on the association between overseas investment and earnings qualities is still limited. In this paper, we consider the effect of earnings qualities, legal systems and degrees of development, and carry out multivariate regression tests for 6 hypotheses on these three factors. We attempt to answer the following three questions: what kind of companies make overseas investment? Can overseas investment improve all types of earnings qualities? Do the legal systems and degree of development of the invested countries effect earnings qualities? Based on our results, we find that Japanese firms with lower advertisement expenditure, lower quick ratio, higher capital ratio, higher research and development expenditure, higher interest expense, higher fixed assets rate, higher return on asset, and larger size, have a higher tendency to make overseas investment. In addition, we find that overseas investment can reduce earnings management risk, but it can also impair reporting conservatism. Lastly, we find that firms under the influence of common law or developed countries have higher level of earnings qualities. This study contributes to the accounting literature in the following ways. First, we provide empirical evidence on the relationship between earnings qualities and overseas investment from the perspective of different legal systems and degrees of development. We also use two different dimensions (quantity and percentage) to represent the extent of the influence of common law countries and developed countries on firms. Second, we use a sample of Japanese firms with and without overseas investment, which are then matched by PSM model. Third, compared to prior research that focused on country differences in earnings quality using one aspect of earnings quality proxy, this study is one of the first to explore country differences in earnings qualities represented by three accounting-based earnings proxies. The findings of this study have implications not only for investors but also for security analysts, auditors, standard setters, and other accounting information users in enhancing their understanding of legal systems and degree of development differences, and their impact on earnings qualities. HYPOTHESES DEVELOPMENT This paper discusses the relationship between overseas investment and companies earnings qualities. International markets have different legal system, investor protection, development degree and many other institutional factors that are assumed to affect earnings qualities differently. Perspective of Legal Systems Earnings Management Haw et al. (2004) use the sample of 25,210 observations from 9 East Asian and 13 West European countries to examine the pervasiveness of income management induced by control cash-flow divergence and whether legal and extra-legal institutions mitigate such behavior. They find that multivariate regression result shows that higher income management is significantly limited in strong investor protection regime. Several other studies also suggest that there is a negative correlation between investor protection and earnings management (Shen et al., 2007; Cahan et al., 2008); that is, better investor protection can mitigate earnings management. Additionally, 738

3 La Porta et al. (1998) find that common-law countries generally have the strongest legal protection for investors, while civil-law countries have weaker legal regime. Using the above rationale, we formulate the first hypothesis: risk. H1 Firms under greater influence of common law countries have lower level of earnings management Value Relevance of Accounting Information Hung (2000) use a sample of 17,743 firm-year observations for the period 1991 to 1997 to examine whether the accrual accounting system negatively affects the value-relevance of accounting measure in countries with weak shareholder protection. They find that higher use of accrual accounting (i.e. high value of accrual index) is negatively associated with the value relevance of earnings and return-on-equity. This relationship however is weakened in countries with strong shareholder protection. Defond et al. (2007) use 29,867 firm-year observation from 26 countries for the period 1995 to 1999 to investigate cross-country difference in the market reaction to earnings announcements. They find that annual earnings announcements have significantly greater information content in countries with strong investor protection environment. Additionally, La Porta et al. (1998) document that common law countries practice a higher level of protection on minority shareholders compared to civil law countries. Using the above rationale, we formulate the second hypothesis: H2 Firms under greater influence of common law countries have higher level of value relevance. Reporting Conservatism In common law capitalist economies such as Australia, Canada, New Zealand, the United Kingdom and the United States, the economic interests of shareholders, creditors, auditors, regulators and other stakeholders drive accounting practices in the direction of conservative reporting and away from neutral or aggressive reporting (Bushman et al., 2006; Roychowdhury et al., 2007). Manganaris et al. (2011) use a sample of 321 public firms that operate in the financial sector of four European countries 5 for the period 1999 to 2008 to investigate whether conservatism exists during the period and whether conservatism level has changed over this period. They find that conservatism exists in all countries before the IFRS adoption and that its level has decreased after 2005 only in France and Germany. They also find that UK exhibits the highest level of conservatism when examining the entire time period. Using the above rationale, we formulate the third hypothesis: H3 Firms under greater influence of common law countries have higher level of reporting conservatism. Perspective of Degrees of Development Earnings Management Leuz et al. (2003) use the sample of 70,995 firm-year observations for the period 1990 to 1999, to provide comparative evidence on corporate earnings management across a broad range of countries. They find that outsider economies characterized by large stock markets, dispersed ownership, and strong legal protection display the lowest level of earnings management. They also find that insider economies characterized by less-developed stock markets, concentrated ownership, and weak legal protection display the highest level of earnings management. Additionally, as mentioned above, common-law countries generally have the strongest legal protection for investors (La Porta et al., 1998). These findings indicate that less-developed countries decrease legal protection for investors, and the weakened legal protection would result in higher earnings management. Japanese firms investing in developed countries are subject to more stringent laws and regulations, and are therefore exposed to a stronger investor protection environment. It follows that the risk of earnings management of these firms will be diminished. Using the above rational, we formulate the fourth hypothesis: risk. H4 Firms under greater influence of developed countries have lower level of earnings management Value Relevance of Accounting Information Kang et al. (2005) use a sample of 1,898 firm-year observations from 1993 to 1999 representing 41 countries to examine whether disclosure transparency affects the value-relevance of accounting information for cross-listed 739

4 firms. They find that firms cross-listed from developed countries provide more value-relevant information because these countries are characterized to have more transparent disclosures. Similarly, developed countries investments such as ones from Japanese companies are exposed to the threat of litigation for less transparent disclosures, a stricter regulatory and enforcement system than their non-overseas investment counterparts. Using the above rationale, we formulate the fifth hypothesis: H5 Firms under greater influence of developed countries have higher level of value relevance. Reporting Conservatism Kung et al. (2008) use the samples of China, Hong Kong and Taiwan by including firm-year observations for the year 1994 to 2003, to examine cross-country differences in financial reporting quality by contrasting the influence of country-specific institutional factors on financial reporting incentives for conservative accounting practices. They find that Hong Kong is the most conservative economy, followed by Taiwan and China, which suggesting that legal system, regulatory environment and political economy of different countries (i.e. developing countries and developed countries) are the pivotal forces which drive accounting conservatism. In addition, Rajan et al. (2003) mention that in a high corruption environment, the expectation is that legal enforcement is weak and that corporate transparency and the quality of accounting information are questionable. Therefore, it is expected that developed countries which generally have less corrupt business practices will disclose conservative accounting information to the market. Using the above rationale, we formulate the sixth hypothesis: METHODS Matching Process H6 Firms under greater influence of developed countries have higher level of reporting conservatism. Many previous researches use matching methods to evaluate the effect of a treatment by comparing the treated and the non-treated units (e.g. cross-country and non-cross-country, overseas investment and non-overseas investment, etc.) in an observational study. However, previous research using sample matching method did not analyze the sample using a more rigorous matching model (Seetharaman et al., 2002; Choi et al., 2009). This paper resolves the selection bias problem by using the propensity score matching (PSM) method as our research method. In this paper, we use the PSM model to match the sample on a broad range of firm characteristics to examine what kind of Japanese companies have tendency to make overseas investment. PSM model matches observations based on the probability of undergoing a treatment, which in our case is the probability of Japanese firms overseas investing. This model generates a sample in which both the overseas investment Japanese firms and without overseas investment Japanese firms are matched to have similar characteristics. We use a logistic model to estimate the probability of Japanese firms make overseas investment. Following Horaguchi (1992) we estimate the propensity score using the following logistic regression model: 8 OVERSEAS COUNTROL Id Yd i, t 0 k Kit, i, t k 1 (1) Where COUNTROL denotes overseas investment control variables. All of the variables are defined in Table 1. We include 8 overseas investment controls (CAP, ADV, R&D, INT, INTEN, QR, ROA, and LNTA). As shown in Table 1, the 8 overseas investment control variables contain the firm information about capital ratio, advertisement expenditure, research and development expenditure, interest expense, fixed asset ratio, quick ratio, return on asset, and firm size. We use propensity score analysis with nonparametric regression (i.e. kernel-based matching). This method uses propensity scores derived from multiple matches to calculate a weighted mean that is used as a counterfactual. As such, kernel-based matching is a robust estimator (Guo et al., 2010). Regression Model Our tests on six hypotheses are based on cross-sectional regressions of the each dependent variable of earnings quality (i.e. DA, RELEVANCE, and CONSERVATISM) on a number of test and control variables. We use the following cross-sectional regression models: 740

5 DAit, 12 RELEVANCE OVERSEAS NUMCOMMON PERCOMMON COUNTROL Id Yd CONSERVATIS i, t 0 1 i, t 2 i, t 3 k Kit, i, t k 1 M it, (2) DAit, 12 RELEVANCE OVERSEAS NUMDEVELOPED PERDEVELOPED COUNTROL Id Yd CONSERVATISM i, t 0 1 i, t 2 i, t 3 k Kit, i, t k 1 it, Each cross-sectional regression model contains three test variables. Test variable OVERSEAS in both model (2) and mode (3) tests for the correlation between overseas investment and earnings qualities. In model (2) we use two measures to represent the extent of influence of common law countries on firm. One is the number of common law countries where firm is investing, represented by test variable NUMCOMMON. The other is the percentage of common law countries where firm is investing, represented by test variable PERCOMMON. Similarly, in model (3) we use two measures to represent the extent of influence of developed countries on firm. One is the number of developed countries where firm is investing, represented by test variable NUMDEVELOPED. The other is the percentage of developed countries where firm is investing, represented by test variable PERDEVELOPED. Quantity (i.e. NUMCOMMON and NUMDEVELOPED) and percentage (i.e. PERCOMMON and PERDEVELOPED) are two different dimensions and it is necessary to discuss the issue with these two dimensions. Where CONTROL denotes earnings qualities control variables. All of the variables are defined in Table 1. We include 12 earnings qualities controls (FEE, BIG4, REV, LOSS, LEV, ISSUE, CFO, FAGE, OC, SALES, ROA, and LNTA). As shown in Table 1, the 12 earnings qualities control variables contain the firm information about audit fees, auditor, change in receivable, loss, debt ratio, financial funding, cash flow, firm age, operating cycle, sales, return on asset, and firm size. (3) Variable definitions about model (1), (2), and (3) are summarized in Table

6 TABLE 1. DEFINITION OF VARIABLES ABOUT MODEL (1), (2), AND (3) Dependent Variables Test Variables Control Variables Variable Name OVERSEAS CAP Propensity Score Marching Regression Test Model (1) Model (2) Model (3) Description Variable Name Description Variable Name Description Dummy variable equals to 1 if a firm was investing in overseas and 0 otherwise. Capital ratio, measured by ownership equity t / Total assets t-1. DA RELEVANCE CONSERVATIS M OVERSEAS NUMCOMMON PERCOMMON ADV Advertisement expenditure t / net sales t. BIG4 Increasing discretionary accruals t measured by performance-matched modified Jones model. The value would be replaced by 0 if discretionary accruals was minus. Value relevance of accounting information based on regression of annual earnings on contemporaneous stock returns over a period of at least 10 years beginning in 2005 and ending in Report conservatism based on Basu (1997) s model over a period of at least 10 years beginning in 2005 and ending in Dummy variable equals to 1 if a firm was investing in overseas and 0 otherwise. The number of total common law countries where a firm was investing. The percentage of total common law countries where a firm was investing. FEE The natural log of audit fees t. OVERSEAS NUMDEVELOPED PERDEVELOPED Dummy variable equals to 1 if a firm was investing in overseas and 0 otherwise. The number of total developed countries where a firm was investing. The percentage of total developed countries where a firm was investing. Dummy variable equals to 1 if one Big4 auditor (i.e. KPMG, EY, PwC, DTT) was engaged by the firm and 0 otherwise. R&D Research and development expenditure t / net sales t. REV (Receivable t Receivable t-1) / total assets t. INT Interest expense t / debt t. LOSS Dummy variable equals to 1 if a firm was net loss at the end of year t. INTEN Fixed assets t / total assets t-1. LEV (Long term debt t plus debt in current liabilities t) / total assets t-1. QR Quick ratio, measured by liquid assets t / current liabilities t. ISSUE Dummy variable equals to 1 if a firm was fundraising more than 5% of total assets and 0 otherwise. ROA Return on asset, measured by net income t / total assets t-1. CFO (Operating cash flow t Operating cash flow t-1) / total assets t. LNTA Natural logarithm of total assets at the end of year t. FAGE Firm s age. OC The natural log of operating cycle rule, measured by log (365 / purchases average inventories / credit sales average accounts receivable). SALES Net sales growth rate t. ROA Return on asset, measured by net income t / total assets t-1. LNTA Natural logarithm of total assets at the end of year t. 742

7 743

8 Proxies of Earnings Qualities Proceedings of the Asia Pacific Conference on Business and Social Sciences 2015, Kuala Lumpur This paper investigates the relationship between 3 earnings qualities (earnings management, value relevance of accounting information, and reporting conservatism) and overseas investment. Earnings Management We use increasing discretionary accrual to represent the risk of earnings management. Discretionary accruals are estimated with the performance-matched modified Jones model (Kothari et al., 2005). The model is estimated by year and by industry, scaled by lagged assets: TA (1/ ASSETS ) ( SALES REC ) PPE (4) i, t 0 1 i, t 2 i, t i, t 3 i, t i, t All of the variables are defined in Table 2. From model (4), we can obtain each firm s estimated modified discretionary accrual and we subtract the estimated modified discretionary accrual of the closest ROA firm in the same industry and year. The resulting error term is the performance-matched modified discretionary accrual measure. The higher the value of increasing discretionary the greater the risk of earnings management is. Value Relevance of Accounting Information We measure the value relevance of accounting information based on regression of annual earnings on contemporaneous stock returns over a period of at least 10 years beginning in 2005 and ending in 2014 as follow (Francis et al. 1999; Bushman et al., 2004): RET E ( ) E ( ) (5) i, t i, t i, t i, t MVi, t 1 MVi, t 1 All of the variables are defined in Table 2. This model suggests that both the current earnings level and the earnings changes level have explanatory power on returns. Thus, the value relevance of earnings is measured by the t-statistics of the slope coefficients of the earnings and change in earnings variables (Francis et al., 1999). The higher coefficients of (α1+α2) are interpreted as more timely accounting information. Reporting Conservatism According to prior researches (Basu, 1997; Pope et al., 1999); we use the following model to measure conservatism over a period of at least 10 years beginning in 2005 and ending in 2014: E MV it, it, 1 NGE RET NGE RET 0 1 i, t 2 i, t 3 i, t i, t i, t (6) All of the variables are defined in Table 2. The asymmetry in the speed of bad news recognition versus good news in earnings supports the description of conservatism according to Basu (1997) who claims that negative returns are recognized faster in earnings than the positive ones. Thus, the higher the value of (α2+α3) / α2 the greater the level of conservatism is. Variable definitions about model (4), (5), and (6) are summarized in Table

9 TABLE 2. DEFINITION OF VARIABLES ABOUT MODEL (4), (5), AND (6) Earnings Quality: Discretional Accrual Earnings Quality: Relevance Earnings Quality: Conservatism Model (4) Model (5) Model (6) Variable Variable Variable Description Description Name Name Name Description TA Total accruals equals (net income before extraordinary items minus operating cash flows from continuing operations t) / total assets t-1. RET ASSETS Total assets t-1 E SALSE REC PPE Sample Selection Change in sales from year t-1 to year t / total assets t-1. Change in accounts receivable from year t-1 to year t / total assets t-1. (Net property, plant and equipment in year t) / total assets t-1. E MV The 15-month stock return ending 3 months after the end of fiscal year t. Earnings before extraordinary items, discontinued operations, and special items. Change in earnings before extraordinary items, discontinued operations, and special items. Market capitalization, measured by stock price t times total share number t. E/MV NEG RET NEG RET Earnings before extraordinary items, discontinued operations, and special items / Market capitalization, measured by stock price t times total share number t. Dummy variable equal 1 if RET t is negative and 0 otherwise. The 15-month stock return ending 3 months after the end of fiscal year t. Variable NEG times variable RET. Japanese firms making overseas investment provide an ideal opportunity to research firms operating under different legal systems and degrees of development. We use a sample of Japanese firms making overseas investment in global market and Japanese firms not making overseas investment in global market to account for different characteristics between these two groups. Using this data allows our paper to provide an insight for discussing the association between overseas investment and earnings quality under different legal systems and degrees of development. TABLE 3. SAMPLE SELECTION AND DESCRIPTION Listed companies for fiscal years 2005 to 2014 (ending in March) 19,490 * (less) Financial companies -1,220 (less) Financial data unavailable -6,632 (less) Stock data unavailable -541 (less) Auditing data unavailable -1,646 (less) Employee data unavailable -175 (less) Relevance and conservatism data unavailable -1,946 ** (less) Discretionary accruals data unavailable -20 *** (less) Propensity score matching data unavailable -80 Full available sample 7,230 **** (less) Data unmatched -257 ***** Propensity score matching sample 6,973 ***** Downloaded data from NEEDS Database using the criteria: accounting year-end at the end of March. ***** To estimate the value of relevance and conservatism require at least 10 years' period. ***** To estimate performance-matched modified Jones model requires more than 20 points of per year or per industry data. ***** The value of relevance and conservatism and discretionary accruals are estimated with full available sample. ***** Propensity score matching sample is matched by the Epanechnikov kernel model. (Bandwidth is 0.06, 5% of the treatment observations) Table 3 presents information on the sample s selection process. For our analyses, we use firm-year data from 2005 to Information about audit fees, auditors and average working age of employees is obtained from the EOL database. Information about stock and financial statements is 745

10 obtained from the NEEDS Financial Quest database. Information about Japanese firms overseas investment is obtained from the Toyo Keizai s Overseas Japanese Companies database. Information of legal system is obtained from the WORLD FACTBOOK of Central Intelligence Agency. Information about countries degrees of development is obtained from data center of UNCTAD. After restricting our sample to firms with fiscal year ended as of March 31, excluding financial companies and missing data, our full available sample consists of 7,230 firm-years. Then after using matching method, we get the final propensity score matching sample which consists of 6,

11 Over seas Proceedings of the Asia Pacific Conference on Business and Social Sciences 2015, Kuala Lumpur TABLE 4. SAMPLE DESCRIPTION BY EACH INDUSTRY AND YEAR Over seas Over seas Over seas Over seas Over seas Non Non Non Non Non Non Non Non Non Non Glass and Ceramics Products % -2.45% Rubber Products % 0.00% Services % -0.96% Pulp and Paper % 0.00% Pharmaceutical % 0.00% Wholesale Trade % 0.00% Chemicals % -0.63% Machinery % -0.96% Metal Products % 0.00% Construction % 0.00% Retail Trade % 0.00% Information & Communication % -0.96% Foods % -0.49% Oil and Coal Products % 0.00% Textiles and Apparels % 0.00% Iron and Steel % -2.75% Electric Appliances , % -0.76% Nonferrous Metals % -2.01% Transportation Equipment % 0.00% Total , % -0.63% Table 4 presents sample description by each industry and year. According to industrial classification of Tokyo Stock Exchange, there are 19 types of Japanese industries making overseas investment. Electric appliances, machinery, chemicals, and construction are top four industries that make positive overseas investment. The percentage of Japanese firm making overseas investment is around 70 percent of total Japanese listed firms. Over seas Over seas Over seas Over seas Firm-y ear Total Oversea s CAGR Non 747

12 TABLE 5. SAMPLE DESCRIPTION BY OVERSEAS INVESTMENT COUNTRIES AND YEAR Civil Law 2% Mixed 98% Mixed 31% Developed 4% Common Law 69% Common Law 2% Mixed 57% Developing 23% Civil Law 41% Non-overseas 30% Overseas 70% Mixed 8% Developing 14% Developed 2% Developed Developing 78% 15% Mixed 83% Common Law 3% Civil Law 11% Mixed 14% Developing 86% Mixed 73% Mixed 86% Firm-year CAGR Developed countries only % Common law countries only Developing countries only % 1.87% Mixed % Developed countries only Overseas Civil law countries only Developing countries only % -0.87% -0.02% Mixed % Developed countries only % Mixed Developing countries only % 0.08% Mixed , % Non-overseas , % Total ,

13 Table 5 presents sample description by overseas investment country and year. From the perspective of legal system, most of the Japanese listed firms (86%) overseas invest in more than two countries where legal systems are different. Only 169 (3%) and 529 (11%) firm-years represent Japanese firms investing only in common-law countries and only civil-law countries, respectively. Among the group of the only common-law countries investment, over three quarters of firms choose to invest in developed countries only. From the perspective of the degree of development, most of the Japanese listed firms (73%) overseas invest in more than two countries and where degrees of development are different. Only 190 (4%) and 1,120 (23%) firm-years represent developed countries only investment and developing countries only investment, respectively. Among the group of developed countries only investment, around 69% of firms choose to invest only in common law countries. As a side notice, among the final propensity score matching sample, 491 listed Japanese firms have overseas investment in 93 countries. Information of each country s legal system, degree of development, and number of Japanese investment firm is listed in Appendix 1. RESULTS Logistic Model TABLE 6. CORRELATIONS AMONG THE VARIABLES FOR PROPENSITY SCORE MATCHING OVERSEAS CAP ADV R&D INT INTEN QR ROA LNTA OVERSEAS 1 CAP ADV R&D INT INTEN QR ROA LNTA Correlations are based on 7,230 firm-year observations. Pearson correlations in the lower diagonal. Table 6 presents the correlations among the variables for propensity score matching. There is a low level of correlation among the all variables, which means there is no multicollinearity problem. TABLE 7. LOGIST REGRESSION FOR EXTIMATING PROPENSITY SCORE U.S. listed Coefficient Std. Err. z-value _cons *** CAP ** ADV ** R&D *** INT *** INTEN *** QR *** ROA *** LNTA *** Id & Yd included likelihood Pseudo R-squared No. Obs. 7,230 *,**,*** Indicates significance at the 0.10, 0.05 and 0.01 levels, respectively. 749

14 Table 7 shows the results of the logistic regression from estimating the propensity score. The logistic model results indicate that 8 variables are significant: CAP, ADV, R&D, INT, INTEN, QR, ROA, and LNTA. It shows that Japanese firms with higher capital ratio, lower advertisement expenditure, higher research and development expenditure, higher interest expense, higher fixed assets rate, lower quick ratio, higher return on asset, and larger size, have a higher tendency to make overseas investment. TABLE 8. RESULTS OF PROPENSITY SCORE MATCHING Treatment assignment Support Off support On support Total Untreated 0 2,088 2,088 Treated 257 4,885 5,142 Total 257 6,973 7,230 Propensity score matching sample is matched by the Epanechnikov kernel model. (Bandwidth is 0.06, 5% of the treatment observations) Table 8 shows the results of propensity score matching. The full available sample consists of 7,230 firm-years, out of which 2,088 and 5,142 firm-years represent non-overseas investment Japanese firms and overseas investment Japanese firms, respectively. After using the Epanechnikov kernel model, the propensity score matching sample consist of 6,973; out of which 2,088 and 4,885 firm-years represent non-overseas investment Japanese firms and overseas investment Japanese firms, respectively. Descriptive Statistics 750

15 TABLE 9. CORRELATIONS AMONG THE VARIABLES FOR MODEL (2) DA OVERS EAS NUMC OMMO N DA 1 OVERSEAS NUMCOMMON PERCOMMON FEE PERCO MMON FEE BIG4 REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA BIG REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA NUMC OMMO N RELEV ANCE OVERS EAS PERCO MMON FEE BIG4 REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA RELEVANCE 1 OVERSEAS NUMCOMMON PERCOMMON FEE BIG REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA CONSE RVATIS M OVERS EAS NUMC OMMO N CONSERVATISM 1 OVERSEAS NUMCOMMON PERCO MMON FEE BIG4 REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA 751

16 PERCOMMON FEE BIG REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA Correlations are based on 6,973 firm-year observations. Pearson correlations in the lower diagonal. Table 9 presents the correlations among the variables for model (2). There is a low level of correlation among the all variables, which means there is no multicollinearity problem. TABLE 10. CORRELATIONS AMONG THE VARIABLES FOR MODEL (3) OVERS EAS NUMDE VELOPE D PERDE VELOPE D FEE BIG4 REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA DA DA 1 OVERSEAS NUMDEVELOPED PERDEVELOPED FEE BIG REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA NUMDE VELOPE D RELEV ANCE OVERS EAS RELEVANCE 1 OVERSEAS NUMDEVELOPED PERDEVELOPED PERDE VELOPE D FEE BIG4 REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA 752

17 FEE BIG REV LOSS Proceedings of the Asia Pacific Conference on Business and Social Sciences 2015, Kuala Lumpur LEV ISSUE CFO FAGE OC SALES ROA LNTA CONSE RVATIS M OVERS EAS NUMDE VELOPE D PERDE VELOPE D FEE BIG4 REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA CONSERVATISM 1 OVERSEAS NUMDEVELOPED PERDEVELOPED FEE BIG REV LOSS LEV ISSUE CFO FAGE OC SALES ROA LNTA Correlations are based on 6,973 firm-year observations. Pearson correlations in the lower diagonal. Table 10 presents the correlations among the variables for model (3). There is a low level of correlation among the all variables, which means there is no multicollinearity problem. 753

18 TABLE 11. DESCRIPTIVE STATISTICS Propensity score matching sample Overseas sample by different legal system Overseas sample by different degree of development Non-overseas Overseas Difference Non-common-law Common-law Difference Non-developed Developed Difference in Means countries countries in Means countries countries in Means Variable Name Mean Std. Dev. Mean Std. Dev. (t-statistic) Mean Std. Dev. Mean Std. Dev. (t-statistic) Mean Std. Dev. Mean Std. Dev. (t-statistic) DA *** *** *** RELEVANCE ** CONSERVATISM *** *** *** OVERSEAS NUMCOMMON *** *** *** PERCOMMON *** *** *** NUMDEVELOPED *** *** *** PERDEVELOPED *** *** *** FEE *** *** *** BIG *** * *** REV LOSS *** ** ** LEV * ** ** ISSUE CFO FAGE *** * OC *** *** *** SALES * * ROA *** *** *** LNTA *** *** *** No. Obs 2,088 4, ,931 1,120 3,765 % of Total 29.9% 70.1% 19.5% 80.5% 22.9% 77.1% *,**,*** Indicate significance at the 0.10, 0.05 and 0.01 level. Propensity Score Matching sample is matched by the Epanechnikov kernel model. (Bandwidth is 0.06, 5% of the treatment observations) 754

19 Table 11 presents the descriptive statistics for the propensity score matching sample, overseas sample by different legal system, and overseas sample by different degree of development. The propensity score matching sample consists of 6,973 firm-years, out of which 2,088 (29.9%) and 4,885 (70.1%) firm-years represent non-overseas investment Japanese firms and overseas investment Japanese firms, respectively. The overseas sample categorized by different legal system consists of 4,885 firm-years, out of which 954 (19.5%) and 3,931 (80.5%) firm-years represent non-common-law countries investment Japanese firms and common-law countries investment Japanese firms, respectively. The overseas sample categorized by different degree of development consists of 4,885 firm-years, out of which 1,120 (22.9%) and 3,765 (77.1%) firm-years represent non-developed countries investment Japanese firms and developed countries investment Japanese firms, respectively. The descriptive statistics for propensity score matching sample indicate that non-overseas investment Japanese firms and overseas investment Japanese firms have significantly different earnings qualities in earnings management and reporting conservatism. The descriptive statistics for overseas sample categorized by different legal system indicate that non-common-law countries investment Japanese firms and common-law countries investment Japanese firms have significantly different earnings qualities in earnings management and reporting conservatism. The descriptive statistics for overseas sample categorized by different degree of development indicate that non-developed countries investment Japanese firms and developed countries investment Japanese firms have significantly different earnings qualities in earnings management, value relevance of accounting information, and reporting conservatism. Regression Results TABLE 12. OVERSEAS INVESTMENT AND EARNINGS QUALITIES ANALYSIS (PERSPECTIVE OF LEGAL SYSTEMS) Dependent Variable: DA Dependent Variable: RELEVANCE Dependent Variable: CONSERVATISM Expec ted Multivar iate Expec ted Multivar iate Expec ted Multivar iate Variable Name Sign Estimate t-statistic Sign Estimate t-statistic Sign Estimate t-statistic cons OVERSEAS *** *** NUMCOMMON ** *** ** PERCOMMON *** FEE *** ** BIG ** ** REV *** LOSS *** LEV *** ISSUE *** CFO *** FAGE *** OC *** *** SALES * ROA *** LNTA *** *** *** Id & Yd Included Included Included Adj R-squared No. Obs. 6,973 6,937 6,937 *,**,*** Indicate significance at the 0.10, 0.05 and 0.01 level. In Table 12, we find a negative and significant OVERSEAS coefficient of in the left column, suggesting that the treatment effects of overseas investment Japanese firms are significantly different from those of non-overseas investment Japanese firms with respect to earnings quality in earnings management. Then we find a negative and significant NUMCOMMON coefficient of in the left column, and a positive and not significant PERCOMMON coefficient of in the left column, thus the H1 is supported by quantity not percentage dimensions that firms under greater influence of common law countries (investing in more common law countries) have lower level of earnings management risk. We also find a negative and non-significant OVERSEAS coefficient of in the middle column of Table 12, suggesting that the treatment effects of overseas investment Japanese firms are not significantly different from those of non-overseas investment Japanese firms with respect to earnings quality in value relevance of accounting information. Then we find a positive and significant NUMCOMMON coefficient of in the middle column, and a negative and not significant 755