Is Japan really a Buy? The corporate governance, cash holdings, and economic performance of Japanese companies

Transcription

1 Is Japan really a Buy? The corporate governance, cash holdings, and economic performance of Japanese companies Kazuo Kato katou@osaka-ue.ac.jp Meng Li mxl120531@utdallas.edu Douglas J. Skinner dskinner@chicagobooth.edu May 2014 Abstract We investigate whether Japan s corporate governance reforms improve economic performance and valuation. Consistent with an overall improvement in governance since 2000, Japanese firms hold less cash and increase payouts to shareholders. However, Japanese firms still hold more cash than US firms, even when we account for the large increase in US firms holdings of cash and marketable securities. In cross section, reductions in (excess) cash and increases in payouts (especially dividends) are associated with improvements in performance. These changes are also related to declines in the influence of the banks (which traditionally sit at the center of Japanese horizontal keiretsu) and with increases in the influence of foreign investors. The market valuation of Japanese firms cash holdings was lower than that of US firms during the 1990s but has now increased to levels closer to those of US firms. Collectively, the evidence suggests that performance improves in those Japanese companies that improve governance but that some Japanese companies continue to lag. Kato is from the Osaka University of Economics; Li is from Jindal School of Management, UT-Dallas; Skinner is from University of Chicago, Booth School of Business. We appreciate comments from Julian Franks (NBER discussant), Yasushi Hamao, Akihiro Noguchi, Akinobu Shuto, Dan Simunic, Tomomi Takada, Arnt Verriest, and workshop participants at Arkansas, Boston College, the Minnesota Empirical Conference, Nagoya, NBER Japan Project, Southern Methodist University, Tilburg, and USC. Skinner acknowledges financial support from the University of Chicago, Booth School of Business. Kato acknowledges financial support from the Japanese Society for the Promotion of Science, Granted-in-Aid for Science Research C

2 1. Introduction The extent to which public firms hold liquidity (cash and marketable securities) is one of the most critical issues facing the CFOs of public companies operating in today s capital markets. Activist investors, especially hedge and focus funds, have become increasingly aggressive in pushing companies to use their cash stockpiles to increase payouts to shareholders. While large US tech firms such as Apple, Microsoft, Cisco, and others have been subject to well-publicized pressure to distribute more cash, the issue is global (Becht et al., 2014). We focus on Japan, which has long been seen as a poster child for poor governance and excessive cash (Jensen, 1989; Rajan and Zingales, 1995; Pinkowitz and Williamson, 2001). 1 At least since French and Poterba (1991), it has been clear that common valuation metrics for Japanese firms differ systematically from those of firms in other countries. At the time French and Poterba wrote their paper (late 1980s), Japanese equity prices seemed too high, with the Nikkei 225 trading at around 80 times earnings. Since the bubble in Japanese real estate and equity prices burst in the early 1990s, Japanese equities have looked cheap by conventional measures. Over the last two decades, Japanese price-to-book multiples have often been well below those of U.S. firms, with many below 1. So interest in buying Japanese equities is longstanding, with many investors, both domestic and foreign, tempted to buy Japanese equities given their valuations. 2 1 The Economist has written numerous articles bemoaning the quality of Japanese corporate governance and the high cash holdings of Japanese firms (for example, see Japan Tobacco: Turning over a new leaf, March 2, 2013; A clash over cash, May 16, 2002; No country for old men, April 29, 1999; Message in a bottle of sauce, November 29, 2007). The recent debacle at Olympus has again focused international attention on the governance practices of Japanese companies (for example, see Pressure on Japan to probe Olympus, Financial Times, October 25, 2011, as well as Michael Woodford s recent book, Exposure, which describes the scandal in detail and relates it to governance problems in Japan (Woodford, 2012)). 2 Milhaupt (2006) cites data showing that in 2000, approximately 13% of 779 non-financial firms on the Tokyo Stock Exchange were trading below their bust up values (measured as cash and cash equivalents plus investment securities minus debt). In spite of the run-up in Japanese equity prices during 2013, at the end of 1

3 A common tactic of activist investors, including those who invest in Japan, is to force undervalued firms to disgorge excess cash to shareholders (Becht et al., 2014; Brav et al., 2009; Hamao et al., 2011). Our goal is to look in an aggregate sense at whether this works in Japan, where many companies have large cash stockpiles and trade at relatively low multiples, attributes that attract activist hedge funds, private equity investors, and other types of value investors. To do this, we study the cash holdings, payout policy, firm performance, and valuation of Japanese companies, and focus on changes around the time Japan reformed its governance practices (circa 2000). Some argue that Japan s governance problems were an important factor in its consistently poor economic performance during the lost decade (Morck and Nakamura, 2001). As Japan s economic problems deepened during the 1990s, there was growing recognition that reforms were needed, including in corporate governance and financial reporting (Hoshi and Kashyap, 2001, 2004; Jackson and Miyajima, 2007). This led to a series of reforms designed to improve the functioning of the Japanese corporate sector and financial system, with many reforms designed to bring Japanese practices into line with those of western countries, often with the US as a model (Nakamura, 2011; Miyajima, 2012). Our paper provides evidence on whether these reforms have been effective, a question on which there is little evidence. We use time-series and cross-sectional variation in Japanese corporate governance to examine the links between excess cash, payout policy, and firm performance. Japan is one of the largest economies in the world, and offers a large and diverse cross-section of public companies, ranging from very large public companies with US-style governance (Sony, 2013 the P/E ratio for the Nikkei 225 was a relatively modest 16.6 (versus 18.2 for the S&P 500) while the P/B ratio was

4 Toyota) to a large number of small- and medium-sized companies with more traditionally Japanese-style governance. We examine the cash holdings of Japanese firms, along with their cash payouts to stockholders, as a concrete way of assessing whether corporate governance improves. Although one can measure corporate governance using various indices, these measures generally have limitations because, first, corporate governance is multi-dimensional and so not subject to direct measurement and, second, because different corporate governance structures are likely to be optimal for different firms (e.g., Larcker et al., 2007). This is especially true in Japan, which allows firms to adopt a western-style model, a more traditional Japanese model, or some form of hybrid model (Jackson and Miyajima, 2007; Milhaupt, 2006), making direct measurement difficult. We provide evidence that Japan s corporate governance reforms have been effective in that there has been a large increase in cash payouts by Japanese companies over the past decade, accompanied by a decline in cash holdings. Nevertheless, we also show that after conditioning on the economic determinants of cash holdings, Japanese firms still hold more cash, on average, than US firms. This result holds in spite of the fact that US firms have increased their holdings of cash and marketable securities over the last two decades (e.g., Bates et al., 2009; Dittmar and Mahrt-Smith, 2007; Pinkowitz et al., 2014). 3 We also show that there is a good deal of cross-sectional variation in the tendency of Japanese firms to increase payouts and reduce cash. There is a strong and economically significant relation between improvements in profitability, declines in excess cash, and increases in cash payouts, particularly dividends. 3 Our findings are robust to the treatment of marketable securities, which are important because US firms holdings of marketable securities increase significantly since the early 1990s while Japanese firms holdings of marketable securities, which are significant in the 1990s, essentially disappear around

5 These changes are more pronounced in firms for which the influence of banks (part of Japan s horizontal keiretsu) declines, the influence of foreign investors increase, and for which managers stockholdings increase. The changes are strongest over , the postreform period. Taken together, these results support the idea that lower cash holdings and higher cash payouts are an integral part of improving corporate governance and performance. We also provide evidence on differences between the market s valuation of cash in US and Japanese firms. Consistent with our main thesis that governance was relatively poor in Japan during the 1990s but improves after 2000, the market valuation of cash for Japanese firms is systematically lower than for US firms during the 1990s, but increases during the 2000s and becomes more comparable to that of US firms. The nature of our evidence makes it hard for us to make strong causal statements. Nevertheless, to date there is virtually no evidence on whether the reforms that have been widely discussed in the wake of Japan s long-standing economic problems have actually had any effect. We present clear evidence of a relationship between changes in cash holdings, payout policy, and firm performance, as well as of a relation between these changes and changes in ownership characteristics that capture elements of Japan s unusual keiretsu ( main bank ) system that is often blamed for its economic problems. As such, we see our evidence as shedding light on the important questions of whether corporate Japan has actually reformed and whether such reforms translate into improved performance. Section 2 provides more details of the Japanese setting. Section 3 details our sample and data. Section 4 provides our evidence on the cash holdings and payout policy of Japanese firms (using US firms as a benchmark), relates changes in cash holdings and payout policy to changes in firm performance, and examines how these changes relate to changes in 4

6 ownership characteristics of these firms. Section 5 then examines how the market s valuation of the cash holdings of Japanese firms compares to that of US firms. Section 6 concludes. 2. Background on Japan Japanese firms are well known for holding high levels of cash. Rajan and Zingales (1995) examine cash holdings of companies across the G7, and find that Japanese firms held substantially more cash than their G7 counterparts in Pinkowitz and Williamson (2001) argue that Japan s main bank system (i.e., the horizontal keiretsu) exacerbates this problem because the banks induce their industrial affiliates to hold excessive cash as a way of expropriating shareholder wealth, a result that reinforces earlier findings on the role of main banks in Japan (Weinstein and Yafeh, 1998). Because large holdings of cash are generally viewed as symptomatic of poor governance, this evidence is consistent with the view that Japanese companies are poorly governed. The Japanese economy has performed persistently poorly over the last two decades. One of the alleged culprits has been Japan s unusual and (some argue) ineffective corporate governance. 4 An important distinguishing feature of the Japanese system is the role of the main bank system and the related keiretsu structure (Aoki et al., 1994; Hoshi and Kashyap, 2001; Morck and Nakamura, 2004). Under this system, creditors, especially banks, play an important role in governance, employees (including management) are seen as an important stakeholders, and shareholders rights are relatively less important than in western capital markets. Since the late 1990s, however, when the Japanese Government introduced an extensive set of reforms designed to remake its financial system, corporate governance has 4 Fukao (2003) and Hoshi and Kashyap (2001) discuss the Japanese financial system and its link to economic performance, especially during the 1990s. Morck and Nakamura (2001) argue that Japanese corporate governance helps explain the poor economic performance of the Japanese corporate sector. 5

7 moved towards a more shareholder-focused model with the goal of improving economic performance. 5 The jury is still out on whether these reforms have led to substantive improvements in Japan s corporate governance practices and the performance of its corporate sector. While in some ways things have improved the importance of the banks and the keiretsu system generally have declined in other ways it seems that the old ways of doing business in Japan remain firmly in place. 6 Some changes that were touted earlier in the 2000s (such as the decline of the keiretsu system and corporate cross-holdings of shares) have partially reversed as incumbent managers seek to protect themselves from an increasingly active market for corporate control, especially from foreign investors. We include the ownership characteristics of Japanese firms, including measures of firms keiretsu affiliations, in a number of our analyses. There is limited evidence on whether Japan s governance reforms have improved corporate performance and increased returns to stockholders. Hamao et al. (2011) provide evidence on the returns to foreign investor activism in Japan between 1998 and They report largely mixed evidence on the ability of activist investors to reform Japanese companies, and widespread adoption of poison pills since Uchida and Xu (2008) provide similar evidence based on an analysis of a few cases of shareholder activism in Japan. 5 Corporate governance reforms include the liberalization of corporate laws that had previously restricted or prevented corporate takeovers, spinoffs, and reorganizations; the use of stock options as management compensation; the availability of stock repurchases (described in more detail below); the introduction of laws allowing shareholders to sue managers (derivative lawsuits); reforms in financial reporting and auditing rules to bring them into line with those in other developed economies; a 2001 law that required banks to reduce their corporate shareholdings; and reforms designed to reduce corporate cross-holding of shares. For more discussion, see Jackson and Miyajima (2007), Milhaupt (2006), Nakamura (2011), Patrick (2004), among others. 6 Milhaupt (2006, p. 3) writes that (o)ver the past decade, the formal institutional environment for Japanese corporate governance has been reformed significantly Yet, despite substantial legal reform and decade after Japan s economic problems emerged, there has been no sea change in Japanese corporate governance practices. See also Jackson and Miyajima (2007). 6

8 3. Sample and data We first compare the cash holdings of Japanese and US firms. The US is often held out as a desirable alternative in discussions of Japanese governance (Nakamura, 2011) and is commonly used as a benchmark in cross-country governance studies generally (e.g., Doidge et al., 2004; Aggarwal et al., 2008). Our data are from WorldScope, collected via Thomson Reuters DataStream (Japanese and US firms) and from Nikkei Financial Quest and Nikkei Quick (Japanese firms). 7 The initial sample includes all Japanese firms listed on the Tokyo Stock Exchange and JASDAQ and all US firms listed on the NYSE, AMEX, and NASDAQ. The sample period is from 1990 to Because our sample period begins in 1990, it encompasses the bursting of the bubble in Japan, which began its economic malaise and so is a suitable starting point for our study. 8 We delete observations with missing total assets and exclude utilities, transportation, and financial firms. Our data on Japanese repurchases comes from Nikkei Quest, supplemented with data drawn directly from financial statements and begins in the latter part of the 1990s when Japanese firms began repurchases. Bates et al. (2009) define cash as the sum of cash and marketable securities in their study of the cash holdings of US firms. Because of possible differences in Japanese firms use of marketable securities, we exclude marketable securities and measure cash holdings simply as cash divided by total assets. 9 To be consistent, we do the same for US firms. However, we replicate all of our analyses using the sum of cash and marketable securities in 7 We also compute certain variables for the US firms using Compustat as a robustness check. For the cash holdings variable, we obtain larger sample sizes using Compustat (Nikkei) for US (Japanese) firms than are available on WorldScope. Nevertheless, our inferences are not affected by the source of the data. 8 Traditional governance practices were still in place in the early 1990s, including almost exclusive reliance by firms on bank financing, the importance of the main bank/keiretsu system, substantial corporate cross-holdings, etc. Reforms began in the mid to late 1990s as the economic problems deepened and there was universal recognition of the need for changes (e.g., Hoshi and Kashyap, 2001). 9 Pinkowitz and Wiliamson (2001) also exclude holdings of marketable securities in measuring cash for Japanese firms. 7

9 place of cash and discuss substantive differences below. In the next section we discuss how Japanese firms holdings of marketable securities change over our sample period. In certain tests we require data on keiretsu membership. Following Dewenter and Warther (1998), we define keiretsu firms as those that belong to one of the largest six large horizontal keiretsu, also known as enterprise groups. We obtain these data from Industrial Groupings in Japan (IGJ, 2001), a standard source of these data, and classify firms as keiretsu firms (if they are classified in IGJ as horizontal keiretsu firms with inclination scores of 2-4) or not (otherwise). Our results are robust to variations of this definition. For other tests we utilize data on the ownership characteristics of Japanese firms, which helps us to measure the effectiveness of governance reforms (for example, declines in corporate cross-holdings are generally seen as reflecting improved governance). In particular, we measure the fractional holdings of financial institutions (both total and within the firm s keiretsu), foreign ownership, and management ownership. These data are from Nikkei, and are only available from 2001 onwards. 4. Evidence 4.1 Differences between Japanese and US industrial firms, We first compare the cash holdings, profitability, and valuation metrics of Japanese and US companies since Table 1, Panel A presents means and medians for cash deflated by total assets, (cash + marketable securities) deflated by total assets, profitability (EBIT/lagged TA), market-to-book ratios, and price-earnings ratios by country and year. 10 Figure 1A plots mean and median ratios of cash/total assets for Japanese firms over the sample period as well as the asset-weighted ratio of cash to total assets (aggregate cash 10 All variables are winsorized to reduce the effect of outliers. Italicized numbers for Japanese firms indicate that the number is significantly different from the corresponding number for US firms at the 1% level or better under two-tailed tests. 8

10 divided by aggregate total assets). Figure 1B plots the mean and median ratios of marketable securities to total assets for Japanese firms. For Japanese firms, mean (median) cash holdings represents 17.2% (15.0%) of assets in 1990, substantially higher than for US firms (at 6.9% and 2.8%, respectively), consistent with previous evidence (e.g., Rajan and Zingales, 1995). The cash holdings of Japanese companies decline through 1997, when the average reaches 11.2% (median 8.8%). This decline presumably reflects the adverse economic conditions in Japan after the bubble burst in After this, average cash holdings increase to 17.2% (median 13.1%) in 2005, numbers still higher than those of US firms but by a much smaller margin (the mean/median for US firms in 2005 is 16.0%/9.6%). The numbers then fluctuate before increasing to 18.5% (14.8%) in 2010 and The increase during the crisis period is consistent with evidence from other countries except the US (Pinkowitz et al., 2014). So Japanese firms still hold more cash than US firms but the differences are smaller than in the early 1990s, due largely to an increase in the holdings of US firms. We revisit this conclusion below, however, using regressions that control for the effect of firm characteristics on cash holdings. Figure 1A plots the asset-weighted ratio of cash holdings ratio for Japanese firms, along with the mean and median of cash/assets. The asset-weighted ratio is close to the median in 1991 and also declines through the 1990s, but then holds roughly constant at 8% through around 2008 after which it increases to 10% in The asset-weighted ratio is consistently below the mean and median, indicating lower cash holdings for the largest 11 Median cash ratios of Japanese are significantly higher than those for US firms (at the 1% level, two tailed) for the full time period. Mean cash ratios are not significantly different from those of US firms in 1996, , 2006, and

11 Japanese firms. 12 The increasing divergence of the three series in Figure 1 implies a substantial increase in the dispersion of cash holdings across Japanese firms over the sample period, especially after This approximately coincides with the introduction of corporate governance reforms, and is consistent with our argument that at least some Japanese firms improve their governance over this period. We exploit this variation in the analyses to follow. When we include marketable securities along with cash (Panel A, Table 1) the conclusions change, largely because of different trends in firms holdings of marketable securities in each country. For US firms, the 1990s saw an increase in holdings of marketable securities to the point that they account for a mean (median) of 15% (5%-6%) of assets from 1998 to 2007 (not reported in tables) after which they fall to 10% (1%) by In contrast (see Figure 1B), Japanese firms hold around 5%-6% (3%) of assets as marketable securities through 1997 after which there is a sharp decline, to around 1% (0%) in 2001, with little subsequent change. 13 The large increase in US firms holdings of marketable securities for most of our sample period means that our conclusions about cash reverse when we compare cash and marketable securities rather than just cash. Beginning in the mid to late 1990s, US firms holdings of cash and marketable securities exceed those of Japanese firms by increasingly large amounts and continue to do so through most of the rest of the sample period. US firms mean (median) holdings of cash and marketable securities are consistently around 12 In contrast, Pinkowitz et al. (2014) report that the largest US firms, primarily US multinationals, hold the largest amounts of cash and marketable securities. 13 There are two non-mutually exclusive explanations for the decline in Japanese firms holdings of marketable securities from 1997 to the early 2000s. First, the decline reflects the general unwinding of corporate crossholdings that begins in the mid-1990s (Miyajima and Kuroki, 2007), assuming that part of the cross-held shares are classified as marketable securities, as seems likely. Second, the sharp decline in 2001 could reflect Japanese firms response to the introduction of mark-to-market accounting for these securities: many Japanese firms reclassified their equity holdings as long-term, apparently to avoid mark-to-market treatment (Urasaki, 2014). 10

12 30% (21%-25%) over compared to 17%-18% (13%) for Japanese firms. One possibility is that while US firms treat marketable securities as being akin to cash and so as part of short-term liquidity, Japanese firms holdings of marketable securities sometimes were more strategic in nature, including cross-holdings of related firms that was part of the keiretsu structures. If this is the case, it seems appropriate to exclude marketable securities for Japanese firms, which are our primary focus (this choice doesn t matter very much after 2001, when most Japanese firms hold essentially no marketable securities). This logic also implies that the appropriate comparison may be between cash holdings for Japanese firms and cash + marketable securities for US firms, in which case we would conclude that US firms hold more short-term liquidity than Japanese firms in every year except 1990 and In any event, these univariate comparisons are not very meaningful without controlling for other differences between these firms, as discussed below. We use accounting profitability (ROA = EBIT on lagged total assets) to compare the economic performance of Japanese and US firms. Consistent with the conventional view that Japanese firms are poorly governed, they are substantially less profitable than US firms for most of this period. 14 Median ROA is significantly lower for Japanese firms in all years except 2001, 2008, and 2011 when there is no reliable difference. Mean ROA is also significantly lower for most of the 1990s, although ROA becomes more volatile for US firms during the 1990s, due to increasing left-skewness in the earnings cross-section, a trend that continues during the 2000s. 15 Japanese firms do not display the tendency of US firms to 14 An alternative perspective, sometimes put forward by Japanese who read the paper, is that the lower profitability of Japanese companies reflects different corporate objectives. That is, shareholder value maximization (and the higher profitability that this implies) is only one of a number of objectives of Japanese managers, who take a more general stakeholder view. As one example, Japanese companies have historically sought to offer lifetime employment to their employees, an objective likely to compromise shareholder value maximization. 15 This increasing left skewness is due to at least two related phenomenon. First, as discussed by Fama and French (2004), there has been a systematic shift in the nature of US publicly-traded firms, with firms tending to 11

13 report large write-downs and losses, and generally show much less cross-sectional dispersion in profitability. 16 The profitability numbers for Japanese firms support the view that the governance reforms that begin around 2000 improve their performance, although there is also a general economic rebound during the early 2000s. Mean (median) ROA improves from 4.0% (3.3%) in 2000 and 2001 to 6.4% (5.7%) in 2006 and 6.0% (5.8%) in The tests below relate these improvements to changes in governance, as measured using excess cash holdings and changes in payout policy. Although the performance of Japanese firms improves over this period, it is still lower than that for the median US firm (median ROA for US firms is 8.5% in 2006 and 8.3% in 2007). As alluded to in the introduction, market-to-book ratios are persistently lower for Japanese firms. During the 1990s, market-to-book ratios for US firms average 3.40 (median 2.13) while those for Japanese firms average 2.36 (median 1.86). [The decade-by-decade numbers are not shown in tables.] During the 2000s the differences widen, with corresponding numbers of 3.50 (2.09) for US firms and 1.84 (1.00) for Japanese firms. The time-series of P/E ratios for Japanese firms also makes it easy to understand why foreign investors have become more interested in these firms. After reaching a peak in 1994, when the mean (median) P/E ratio was 92 (51) the phenomenon that French and Poterba (1991) study P/E ratios for Japanese firms decline consistently. While partly due to a decline in Japanese equity prices, this is also due to a consistent increase in Japanese go public earlier in their life cycles. Second, US firms are reporting losses at an increasing rate (Hayn, 1995; Klein and Marquardt, 2006), and these losses tend to increase in size over time. 16 There is some reason to believe that earnings management, particularly income smoothing, is more prevalent in Japan (e.g., Suda and Shuto, 2005). 12

14 firms EPS, in part due to the fact that Japanese accounting rules changed significantly over this period as part of the overall governance reforms. 17 P/E ratios for US firms show little trend, with the median varying in a tight range around 20 and the mean varying between approximately 30 and 40. Because of the decline in the P/E ratios of Japanese firms, mean and median P/E ratios for Japanese and US companies are similar over 2003 through Given our interest in comparing the cash holdings of Japanese and US firms, we also compare other characteristics of these firms that previous research finds are associated with cash holdings. In Panel B of Table 1, we compare these firms size, net working capital (excluding cash), industry sigma, R&D intensity, leverage, capital expenditures, and operating cash flows. Along with market/book, these are the key determinants of cash holdings (Opler et al., 1999; Bates et al., 2009). To economize on space, we report means and medians for each set of firms by decade (1990s and ). 18 The evidence in Panel B of Table 1 shows that Japanese and US firms differ in a number of respects likely to affect cash holdings. First, Japanese firms have lower investment opportunities than US firms, with lower M/B ratios (above) and R&D intensity throughout most of the sample period. While median R&D intensity is always close to zero, means are 5.4% and 8.8% for US firms versus 1.0% and 1.7% for Japanese firms over the 17 French and Poterba (1991) discuss the fact that the high P/E ratios they observed in the 1980s for Japanese companies was partly driven by differences in accounting pushing down EPS numbers, including the fact that most financial statements in Japan were not consolidated. These accounting differences had largely disappeared by the early 2000s, which possibly explains at least part of the upward trend in Japanese P/E ratios. 18 For each decade we report medians of the annual means and medians. We measure size as the natural log of total assets, leverage as the ratio of long-term debt to total assets, ROE as net income deflated by lagged shareholders equity, net working capital as current assets minus current liabilities minus cash plus short term loans deflated by total assets, R&D as research and development expenditures deflated by sales, capital expenditures as capital expenditures deflated by total assets, and cash flow as funds from operations deflated by total assets. Following Bates et al. (2009), we construct industry sigma as the mean of the standard deviations of EBIT/assets over the past 10 years for firms in a given industry. We define industry using the industry group variable (WC06011) from WorldScope. We use a three-digit code for miscellaneous industry and two-digit codes for all other industries. 13

15 1990s and 2000s, respectively. (All differences across countries are statistically significant at the 5% level or better.) Second, Japanese firms are considerably less volatile than US firms: median industry sigma is 7.1% and 10.8% for US firms compared to only 2.2% and 4.2% for Japanese firms, with similar inferences for means. Other things held constant, these differences imply that Japanese firms would hold less cash than US firms. 19 Third, Japanese firms generally are less levered, with lower operating cash flows, lower capital expenditures, and less working capital than US firms. (Mean leverage for US firms is and versus and for Japanese firms across the two decades, respectively, with similar inferences for medians; median operating cash flows are and for US firms versus and for Japanese firms, with slightly different inferences for means; mean capex for US firms is and versus and for Japanese firms, with similar inferences for medians; mean net working capital is and for US firms versus and for Japanese firms, with similar inferences for medians.) Because these characteristics are generally negatively related to cash holdings, they help explain why Japanese firms tend to hold more cash than US firms. Finally, Japanese firms are larger than US firms (in US dollar terms) in the 1990s but smaller than US firms in the 2000s. Because size tends to be negatively related to cash holdings, this helps explain why Japanese firms have higher cash holdings than US firms in the latter part of the sample period. Overall, Japanese firms differ in a number of respects from their US counterparts, which makes it difficult to interpret simple comparisons of cash holdings. 19 Statements about the relation between firm characteristics and cash holdings are based on previous research that models firms cash holdings. For example, see Pinkowitz et al. (2014) or Opler et al. (1999). Pinkowitz et al. (2014) estimate this model using a large sample of firms drawn from a large cross-section of countries and find that cash holdings are positively related to industry volatility, M/B, and R&D and negatively related to firm size, capital expenditures, cash flows, net working capital, and leverage. 14

16 To compare the cash holdings of Japanese and US firms conditional on firm characteristics, we use variants of the regression model developed by Opler et al. (1999). Based on the transactions costs and precautionary demands for cash, Opler et al. model cash holdings as a function of firm size, a dividend-payer dummy, leverage, market-to-book, industry sigma, net working capitals, R&D intensity, capital expenditures, and cash flow, to which we add profitability and a loss dummy. We use residuals from these regressions later in the paper to measure Japanese firms excess cash holdings, a key variable in our analysis. We first estimate panel regressions for each country, both for the overall period ( ) and by decade ( and ), and report the results in Table 2, Panel A. This allows us to assess whether the economic determinants of cash differ significantly across Japanese and US firms, as well as to gauge how these determinants change over time in each country. If Japanese firms governance improves after the late 1990s reforms, we expect to see improvements in their management of cash in the 2000s. Consequently, when we estimate the regressions for the full time period we include dummies for and to see how cash holdings change after 2000 after conditioning on firm characteristics (and separating the crisis period). Second, we estimate annual cross-sectional regressions that pool US and Japanese firms with available data in a given year, and report the results in Table 2, Panel B. These regressions include a Japan dummy to assess whether the cash holdings of Japanese companies differ from those of US companies conditional on the other variables, and a dummy for Japanese firms affiliated with keiretsu groups. We also estimate these regressions using cash + marketable securities instead of cash as the dependent variable (not reported in tables) and discuss those results when different. 15

17 Table 2, Panel A reports the first set of regressions. 20 For US firms, these regressions have adjusted R-squareds of 32% for , 33% for , and 34% for the overall period. Coefficients on key variables are mostly consistent with Opler et al. (1999) and Bates et al. (2009). Cash holdings are positively related to industry sigma, marketto-book, and R&D intensity, and negatively related to size, dividend payment, leverage, net working capital, capital expenditures, and cash flow. The significance and magnitude of regression coefficients are mostly consistent across sub-periods. The time dummies are insignificant for US firms, indicating that cash holdings for these firms do not increase in the 2000s once we account for changes in firm characteristics (similar to the conclusion of Bates et al., 2009). When we estimate the US firm regressions using cash + marketable securities as the dependent variable (not reported in tables), explanatory power increases markedly, with R- squareds of 48% overall and 49% and 47% for the respective subperiods. Coefficients on key variables increase in absolute value. The fact that these models work significantly better for cash + marketable securities than cash suggests that US firms view marketable securities as part of their short-term liquidity and manage the combination of these amounts. 21 This opposite is true for Japanese firms, as discussed next. For Japanese firms the time dummies are reliably negative for both periods and roughly consistent in magnitude, with coefficients of -3.4% for both and This suggests that, on average, Japanese firms reduce their cash holdings after We estimate these regressions with all available observations. To assess the effect of changes in sample composition over time, we also estimate these regressions after requiring firms to have at least three observations in each decade, with similar results. We cluster standard errors by firm and year. 21 This is also consistent with how the financial press portrays the cash holdings of US firms. For example, the press refers to the approximately $140 billion of cash and marketable securities on Apple s balance sheet (including long-term marketable securities) simply as cash, implying there is no meaningful distinction. As indicated earlier, researchers also include marketable securities as cash for US firms (e.g., Bates et al., 2009; Pinkowitz et al., 2013). 16

18 once we condition on firm characteristics, consistent with governance reforms being modestly effective. The adjusted R-squared for Japanese firms in the full period is 29%, slightly lower than for US firms (34%). The model does not explain cash nearly as well for Japanese firms in the 1990s. For , the adjusted R-squared is 17%, about half of that for US firms in the same period. Two variables that are usually important market-to-book and industry sigma do not load in this period. 22 In the 2000s, the adjusted R-squared for Japanese firms increases to 32%, essentially the same as for US firms, with market-to-book and industry sigma now strongly significant in the expected direction. This shift is generally consistent with Japanese firms managing cash in a more disciplined way (or at least more like US firms) in the post reform period. When we instead estimate these regressions using cash + marketable securities for Japanese firms (not reported in tables), the R-squareds decline, opposite to what we observe for US firms, to 23% (versus 29%) for the full period and to 10% (versus 16%) for the 1990s and 27% (32%) for the 2000s. This is consistent with the idea that cash alone is the relevant liquidity variable for Japanese firms (while cash + marketable securities is the relevant liquidity measure for US firms). Table 2, Panel B reports on the second set of cash regressions. Here we estimate a single regression in each annual cross-section that includes all US and Japanese firms with the requisite data. To measure differences in the cash holdings of Japanese firms these regressions include separate intercept dummies for Japanese firms and Japanese keiretsu firms 22 In addition, the coefficient on the dividend payer dummy is positive for Japanese firms in the 1990s (t = 4.5), opposite to the result for US firms, for which the coefficient is strongly and consistently negative. Because dividends paid by Japanese firms are typically much smaller than those of US firms (see next section), this suggests that while dividends help mitigate free cash flow problems in US firms, they do not do so for Japanese firms, instead serving a more perfunctory role. 17

19 (i.e., both dummies turn on for Japanese keiretsu firms). The idea is to compare the cash holdings of Japanese firms in general, as well as keiretsu firms in particular, to those of US firms conditional on firm characteristics that affect cash holdings. For brevity and ease of presentation, we only tabulate coefficients on the Japan and keiretsu intercept dummies along with number of observations and R-squareds. 23 The coefficient on the Japan intercept is large and reliably positive in all years, indicating that Japanese firms hold more cash than US firms after conditioning on firm characteristics. The differences are large in the early 1990s, ranging from 17% to 20% of total assets over The differences are less pronounced over 1995 to 2001 but remain in the 10% to 14% range. The differences fall in a similar range, 12% to 15%, for the remainder of the sample period, indicating that Japanese firms consistently hold more cash than US firms after conditioning on firm characteristics, even given the fact that US firms increase their cash holdings significantly since The coefficient on the keiretsu dummy is reliably negative in all years except 1998 and 1999 and indicates that these firms hold 2% to 3% less cash than other Japanese firms over most of the time period. 24 We have also estimated these regressions using cash + marketable securities as the dependent variable (not reported in tables). Because US firms generally have larger holdings of marketable securities for most of the sample period (Table 1A), differences are smaller but remain significant. As discussed above, a more appropriate specification may be to use cash for Japanese firms and cash + marketable securities for US firms. The results from this specification (not reported in tables), show that, conditional on firm characteristics, Japanese 23 In most cases, the sign and significance of the coefficients is consistent with that for the regressions reported in Panel A, although for those variables where there was some inconsistency between the results for the US and Japanese firms, the coefficients are less significant. 24 Pinkowitz and Williamson (2001) also find a significantly negative coefficient on a keiretsu dummy in crosssectional cash regressions. 18

20 firms hold significantly more liquidity than US firms in all years except 1996, 1997, and The differences range from 12%-15% in the early 1990s, and increase steadily over the 2000s, from 5% in 2000 and 2001 to 9% in 2007 and 13% in 2010 and Thus, the conclusion that Japanese firms typically hold more cash than US firms is robust to the treatment of marketable securities, and holds up even when we stack the deck in favor of Japanese firms by including marketable securities for US but not Japanese firms. 4.2 Payout policy for Japanese firms It is now well-established that firms distribute cash to shareholders through both share repurchases (buy-backs) and dividends (Grullon and Michaely, 2002; Skinner, 2008). To report on the payout policies of Japanese firms in a meaningful way, we report both dividends and stock repurchases. Restrictions on stock repurchases in Japan were gradually lifted beginning in the mid-1990s (discussed in more detail in the appendix). There is little previous evidence on stock repurchases for Japanese firms. Figure 2 reports aggregate dividends and repurchases by Japanese firms, measured in constant 1991 yen (not reported in tables). While payouts are largely stagnant during the 1990s, the introduction of repurchases in the late 1990s fuels a strong increase in payouts during the 2000s, consistent with improving governance. Repurchases for Japanese firms begin at 33 billion in 1996 and grow quickly to 739 billion in 2001, 1.4 trillion in 2002, 2.6 trillion in 2003 before reaching 3.8 trillion in Dividends are largely flat over 1991 to 2000, varying between 2.2 trillion and 2.5 trillion with no obvious trend. However, over 2001 to 2008 they grow strongly, from 2.7 trillion to 6.4 trillion, before declining with the crisis. Total payouts increase more than threefold over this period, from 3.0 trillion in 2000 to 9.9 trillion in 2008, with dividends accounting for 58% of the increase. The strong growth in total payouts from 2001 to 2007 is consistent with the view 19

21 that corporate governance improves over this period. This trend is similar to that for US firms over this period (Floyd et al., 2013) although in Japan the growth is more heavily tilted towards dividends. 25 To get some sense for the number of Japanese firms that account for these trends, Figure 3 reports the fraction of Japanese firms that pay dividends, repurchase, and that both pay dividends and repurchase by year. The fraction of dividend-payers is above 90% in 1991 but drifts downwards over the 1990s, falling to just less than 80% in After this, the fraction of dividend-payers fluctuates around 80% for the remainder of the sample period. This fraction is much higher than the fraction of dividend-payers in other major economies, including the US (Denis and Osobov, 2008; Floyd et al., 2013). However, dividends paid by Japanese firms tend to be much smaller than those paid in other countries. 26 Figure 3 shows that the fraction of Japanese firms that repurchase increases from close to 0 in 1997 to around 10% from 1999 to 2001 and then fluctuates between 14% and 30% over 2002 to Similar to the US, very few firms repurchase without also paying dividends (i.e., the fraction of firms that pay dividends and repurchase closely tracks the fraction that repurchase). The increased payout by Japanese firms is not shared equally among all payers. Similar to the concentration in payouts observed in the US (DeAngelo et al., 2004), the growth in payouts evident in Figure 2 is concentrated in the set of firms that both pay dividends and repurchase. These firms account for 1/3 of total payouts in 2000 and 2001 and then for between 60% and 74% of payouts over 2003 to 2008 as more firms repurchase 25 Consistent with the idea that Japanese firms are more willing to cut dividends than US firms (Dewenter and Warther, 1998), Japanese industrials cut dividends more aggressively during the crisis than US firms (Floyd et al., 2013). 26 Our data show that annual dividends for the median Japanese firm represent about 0.6% of assets compared to 1.8% of assets for US firms (these amounts do not vary greatly across the sample period). 20

22 (not reported in tables). This means that between 14% and 28% of firms account for the bulk of aggregate payouts (this is the fraction of firms that both repurchase and pay dividends over this period, per Figure 3). 27 We later examine whether across-firm variation in the size of the increase in payouts is related to changes in performance, a key premise of our main argument. 4.3 The relation between excess cash, payout policy, and firm performance Based on our main working assumption that Japanese firms management of cash is a proxy for the quality of governance, we next provide evidence on whether reductions in (excess) cash holdings translate into improvements in performance. We first use the cash holdings regressions to sort sample firms into excess cash deciles (to do this, we use the residuals from the Table 2, Panel A regressions, estimated for the 1990s and 2000s separately, to measure excess cash). We use these decile ranks to investigate whether changes in excess cash are associated with changes in firm performance (changes in ROA). 28 We predict that declines (increases) in excess cash holdings are associated with improved (worsened) performance. We report the results of this analysis in Table 3. We analyze transitions in excess cash over three periods. For each period, we sort firms into excess cash deciles in an initial year and then re-sort in the final year, and report transitions for , , and Although time period choices are inevitably arbitrary, we view 2001 as representing the transition to the reform period of the 2000s because many reforms occur around 2000 (2001 also coincides with the availability of data on the ownership structure of 27 Not surprisingly, firms that both repurchase and pay dividends are larger than those that only pay dividends, which in turn are substantially larger than non-payers. 28 We also use return on equity (ROE) to measure performance, with similar results, as well as replicating all of the analysis in this sub-section using cash + marketable securities in place of cash, once again with similar if not stronger results. 21

23 Japanese firms, which we utilize in the next subsection). Thus, we expect our prediction that changes in excess cash and performance are inversely related is more likely to hold in (when the results of reform should be evident) than in (the heart of the economic malaise). In fact, because the period incorporates the period before reforms, it can be viewed as a placebo period in the sense that, in the absence of meaningful corporate governance reform, there is not likely to be any relation between excess cash holdings and performance. Because is also a post-reform period, we also expect a negative relation although any relation is likely clouded by the effects of the global financial crisis. Table 3 reports the transition matrices (the percentages are based on rows, so that they sum to 100 across rows). Each row comprises observations in a given excess cash decile for the initial year, from 1 to 10, where 1 denotes the lowest excess cash decile and 10 the highest excess cash decile. The columns comprise deciles defined in the same way for the last year. Observations on the diagonal thus represent firms in the same excess cash decile in the first and last years (no change). Observations above the diagonal represent firms for which excess cash holdings increase in relative terms, so that firms move up deciles, and conversely for observations below the diagonal. This analysis requires firms to have data in the first and last years of each period and we lose observations accordingly. If membership in excess cash deciles is independent over time, we would see 10% in all cells. Instead, Table 3 shows evidence of clustering, most notably in cells on the diagonal and just off the diagonal, and particularly in the extremes, indicating that firms relative levels 22

24 of excess cash holdings persist over time, consistent with excess cash being relatively sticky and related to governance. 29 The key test links changes in excess cash to changes in firm performance. For each period, we divide firms into those for which excess cash worsens (improves), as indicated by increases (decreases) in decile membership over the period. If improvements in cash management practices (lower excess cash) result from improved governance, we expect this to translate into improved performance, and vice versa. Table 3 also reports the results of this analysis we compare the change in profitability (ROA) over each transition period for the groups with increases and decreases in excess cash. For , there is little evidence of a relation between changes in excess cash holdings and firm performance, as predicted. The mean (median) change in ROA is -0.6% (0.1%) for firms with increases in excess cash and -0.9% (-0.7%) for firms with decreases in excess cash, numbers that are not different at conventional significance levels. In contrast, there is clear evidence of a negative relation between changes in excess cash holdings and performance over For this period, there is an overall improvement in corporate performance, with a mean (median) increase in ROA of 1.9% (1.9%). As predicted, this improvement is stronger for firms for which excess cash declines. For these firms, the mean (median) change in ROA is 3.0% (2.6%), compared to 0.9% (1.4%) for firms for which excess cash holdings increase (differences significant at better than 1%). There is weaker evidence of a negative relation over For the sample as a whole, mean (median) ROA changes by -1.9% (-1.7%) due to the effects of the global 29 Interestingly, there is an increasing tendency for firms to cluster in the extremes of the diagonal as we move from the earliest to the latest subperiods, perhaps because the governance reforms increasingly force firms into corner solutions where they are extremely good or extremely bad in terms of their governance/cash holdings. 23

25 financial crisis. The mean (median) change in ROA is -2.0% (-1.7%) for firms with increases in excess cash versus -1.2% (-1.2%) for firms with decreases in excess cash, differences that are statistically significant in the predicted direction but only at the 10% level (two-tailed). 30 Another way of assessing the relation between changes in excess cash and changes in performance is to use continuous measures of both variables, and to include changes in payouts in the analysis. To do this, for each period we regress changes in ROA on changes in excess cash and changes in payout, where the latter is measured as the change in total cash payout (deflated by total assets) from the initial year to the final year. Under our main argument, declines in excess cash and increases in cash payouts should both result in improvements in performance. Because dividends represent a stronger commitment than repurchases (e.g., Brav et al., 2005; Grullon and Michaely, 2002; Skinner, 2008), we separate changes in dividends and repurchases. The results are in Table 4. For first sub-period, 1994 to 2001, the coefficient on the change in excess cash is not significant, while that on the change in payout is positive and significant (1.23, t = 2.30), as expected if improved payout is related to improved performance. The overall R-squared is low, at 1.1%. When we separate payout into dividends and repurchases, the coefficient on dividends is large and highly significant (6.82, t = 6.68) while that on repurchases is not significant, and the R-squared increases to 12.2% (the coefficient on excess cash remains insignificant). So increases in dividends are related to improvements in performance, even in the 1990s. The evidence is stronger for the second-subperiod, 2001 to For this period, the coefficient on the change in excess cash is (t = -5.86) while that on the change in 30 When we replicate this analysis using cash + marketable securities rather than cash, the ROA changes are again highly significant for and significant at the 5% level over Results are generally slightly stronger if we use ROE in place of ROA. 24

26 payout is 0.78 (t = 6.75), indicating that improved performance is associated with lower excess cash and increased payouts, consistent with our predictions. The R-squared is 3.8%. When we separate payout into dividends and repurchases, the R-squared increases to 10.9%. The coefficient on dividends is 4.50 (t = 14.6) while that on repurchases is again insignificant, while the coefficient on excess cash remains negative and highly significant, with a coefficient of (t = -5.83). 31 Over , the negative relation between changes in performance and excess cash holdings is no longer significant, as expected due to the crisis. The coefficient on payouts remains positive (0.34 t = 4.37) but the R-squared is only 0.8%. The dividend/repurchases split again improves explanatory power, to 5.8%, with the coefficient on dividends smaller than in prior periods but again highly significant (2.52, t = 12.0). Overall, these results support the view that poor cash management practices are indicative of poor governance in Japanese firms, and that improvements in these practices are associated with improved corporate performance. To more directly tie these results to the governance of these firms, we next examine how these relations relate to changes in the ownership characteristics of Japanese firms using data that is available beginning in Changes in performance, cash holdings, payout policy, and ownership structure of Japanese firms There is a large literature on the unusual features of Japanese governance (Aoki et al., 1994; Dewenter and Warther, 1998; Hoshi et al., 1990; Kang et al., 2000; Kaplan, 1994; Morck and Nakamura, 2001, 2004; Pinkowitz and Williamson, 2001; Weinstein and Yafeh, 1998). At the heart of this literature is the idea that the traditional Japanese system is 31 Causality potentially operates in both directions here: firms can pay higher dividends because they are more profitable, as well as improved governance driving both (dividends are usually dependent on current and past profitability). However, note that in the absence of some minimum level of effective governance, managers of profitable firms would hoard more cash, so that there would be little relation between profitability and dividends. The fact that the strength of the relation (in an R-squared sense) increases after 2000 is consistent with improved governance in this period. Moreover, if causality is running from profitability to payout, the relation with repurchases should also be strong when in fact it is effectively non-existent. 25

27 characterized by relatively high levels of bank ownership, interlocking boards, and corporate cross-holdings, important features of the keiretsu system. While this system was viewed as being effective at improving economic performance during the boom period in Japan (the 1970s and 1980s), questions arose about its viability as the economic downturn of the 1990s continued (Jackson and Miyajima, 2007). To investigate how changes in ownership attributes affect the relation between performance, excess cash, and payout policy, we next examine changes in the ownership structure of Japanese companies. Table 5, Panel A reports mean (median) values for various ownership characteristics of Japanese firms from 2001 to To measure the influence of financial institutions (banks and insurance companies), we report total ownership by financial institutions, ownership by financial institutions within a given keiretsu, and bank loans deflated by assets. The general view from the Japanese governance literature is that the stronger the withinkeiretsu influence of banks, especially the main banks, the weaker is governance (e.g., Weinstein and Yafeh, 1998; Pinkowitz and Williamson, 2001). We also include the levels of management and foreign ownership, which we expect will be associated with better governance and firm performance (although the causality will likely run in both directions). Consistent with the reforms taking effect over 2001 to 2007, the numbers in Table 5, Panel A, show that ownership by financial institutions declines during the 2000s. Average ownership by all financial institutions is 21% (median, 18%) in 2001 and declines to 16% by 2008 and to 15% by 2011 (the medians are 13% and 12%). 32 We report three columns of numbers for the within-keiretsu financial ownership variable to fully capture changes in the 32 Our numbers are broadly consistent with those for the Japanese market as a whole, as reported, for example, in the Tokyo Stock Exchange (TSE) Fact Book (2011). Numbers reported by this source indicate that the decline starts earlier. This publication reports that ownership by all financial institutions in listed Japanese companies, measured on an aggregate market capitalization basis, declines from 42.1% in March 1998 to 30.6% in March 2010, with ownership by city and regional banks declining from 14.8% to 4.3%. Because our ownership data only begins in 2001, our numbers likely understate the effects of reform on ownership. 26

28 influence of the main banks: overall mean, fraction of non-zero observations (which effectively shows the fraction of keiretsu firms), and the median for the non-zero observations. The overall mean declines from 2.1% in 2001 to 0.8% in This is driven by both a decline in financial institutions influence within keiretsu and a decline in the fraction of keiretsu firms (non-zero observations). The fraction of non-zero observations declines from 24% in 2001 to 14% in Within this declining fraction of firms, median ownership of financial institutions declines from 7.6% in 2001 to 4.6% in Also consistent with the declining influence of banks, bank loans as a fraction of assets decline from 26% (median, 24%) of assets in 2001 to 21% (median, 16%) by 2007 and thereafter. Collectively, these results indicate a decline in the influence of the main banks. Most of the decline occurs from 2001 to 2007, consistent with our focus on this period as being most important in understanding the effects of governance reform. Over this same period, the level of foreign ownership increases, consistent with foreign investors being attracted to Japanese equities by their low relative valuations and the prospects for reform. Mean foreign ownership increases from 5% (median, 1%) in 2001 to 9% (5%) in 2006 and 2007, before declining somewhat after this. Finally, we report on management ownership, which is harder to interpret clearly in terms of its association with governance; that is, it can reflect management entrenchment or alternatively be viewed as indicative of better incentive alignment (Morck, Shleifer, and Vishny, 1988). This variable shows little evidence of a trend over 2001 to 2011, with the mean (median) consistently around 9%-10%. Table 5, Panel B, reports correlations between ownership characteristics, firm performance (ROA), excess cash, and payout in Beginning with profitability (ROA), 33 We have also computed correlations between changes in these variables, with similar results. 27

29 the largest correlations are with dividends (0.450), foreign ownership (0.198), management ownership (0.208), and bank loans (-0.285). These correlations are consistent with our working hypothesis that higher payout is an indicator of good governance and so associated with firm performance, that foreign ownership and management ownership are positively associated with good governance, and that bank loans are negatively associated with good governance. Further consistent with these interpretations, the dividend variable is positively associated with foreign ownership (0.215), management ownership (0.183), and negatively associated with bank loans (-0.428). 34 ROA and dividends are both negatively related to within-keiretsu financial ownership ( and , respectively), consistent with this variable being indicative of poor governance. Overall, financial ownership is positively associated with foreign ownership (0.223), strongly positively associated with financial ownership within the keiretsu (0.412, as expected given the overlapping construction of these variables), and negatively associated with managerial ownership (-0.355), making it unclear whether this variable is capturing good or bad governance. Notice that the within-keiretsu financial ownership variable is less negatively associated with managerial ownership (-0.266) and much less positively associated with foreign ownership (0.034) than the overall financial ownership variable, suggesting that it is financial ownership within a keiretsu that adversely affects governance. This makes sense if external financial institutions play a monitoring role while within-keiretsu institutions tend to shield incumbent management from reforms. Table 6 reports regressions of changes in firm performance on changes in excess cash, changes in dividends and repurchases, and changes in the ownership characteristics. 34 The relation between repurchases and these ownership variables is weak. This is expected because payouts are only useful in governance if they represent a commitment to pay out future cash, which is the case for dividends but not repurchases. 28

30 These are the regressions from Table 4 augmented with the ownership change variables. We estimate these regressions for the and subperiods for which we have ownership data. The inclusion of the ownership variables has little effect on the excess cash and cash payout variables. The coefficient on the change in excess cash for continues to be reliably negative, with roughly the same coefficient magnitude and significance as before (-0.13, t = -5.04). This result is not greatly affected by whether the payouts variable is divided into dividends and repurchases. For , this variable again becomes insignificant when the other variables are included. The change in payout variable remains reliably positive for both and when the ownership variables are included, with little change in magnitude of the coefficient or its significance. The same is true for the dividends variable, which again has a larger coefficient and greater significance than the overall payout variable. So the key findings remain largely intact when we include the ownership variables. The inclusion of the ownership variables adds significantly to the explanatory power of these regressions: for , the adjusted R-squared increases from 3.8% to 13.4%, and for from 0.1% to 10.6%. Corresponding increases for regressions that separate the dividend and repurchase variables are from 10.9% to 17.4% over and from 5.8% to 13.3% over Consistent with the idea that foreign ownership is related to good/improving governance, the change in foreign ownership is positively associated with the change in performance over , with a coefficient of 0.12 (t = 4.47). 35 The change in total 35 Once again, we cannot determine if increased foreign ownership is the cause or the result of improved performance; it is likely that both effects are occurring. Hamao et al. (2011) and Uchida and Xu (2008) describe how some foreign (US) investors push for improvements in governance in Japanese firms. A 29

31 financial ownership is also positively related to performance in both periods, consistent with these institutions playing a monitoring role (in contrast, the change in financial ownership within the keiretsu, where the monitoring role may conflict with an entrenchment role of banks, is not related to performance). The change in management ownership is also positively related to performance in (coefficient 0.16, t = 5.47), suggesting that this variable is measuring improved management incentives as opposed to entrenchment. Finally, the change in bank loans is strongly negatively related to performance in both periods, consistent with the argument that main bank influence adversely affects governance and hence performance (most bank loans are made from within the bank group). These results confirm and extend the earlier findings showing that firms that lower their holdings of excess cash and increase dividends enjoy improvements in performance. Importantly, we find that improvements in performance are stronger for those firms that increase levels of foreign ownership, management ownership, ownership by financial institutions generally (but not those within the keiretsu), and that lower their levels of bank loans. These findings support our interpretation that it is improved governance that explains variation in improvements in the performance of Japanese firms. 5. Cash holdings, governance, and valuation Our main research question is whether governance, as manifested in excess cash, improves for Japanese firms. The evidence in Section 4 shows, generally, that lower excess cash holdings translate into better performance, measured using accounting profitability. A different way of addressing this issue is to examine whether improvements in governance are reflected in changes in the market valuation of cash, a common approach in finance (e.g., Dittmar and Mahrt-Smith, 2007; Pinkowitz et al., 2006). These papers find that cash prominent example is Steel Partners, an activist US hedge fund that invests in Japan as well as in the US (e.g., see Message in a bottle of sauce, The Economist, November 29, 2007). 30

32 holdings are valued at lower amounts for countries/firms where governance is poor, the interpretation being that managers can more easily expropriate or mismanage cash when investor protection/governance is weak. We use this approach to investigate two predictions derived from our main thesis: (a) to the extent that cash holdings in Japanese companies are unusually high because of poor governance, cash is valued more highly in U.S. firms than Japanese firms, (b) to the extent that governance in Japan improves over time, we expect any such differences to decline over time (after 2000). To perform this analysis we use two specifications, both of which are based on Fama and French (1998). The first specification follows Pinkowitz et al. (2006): V i,t = α + β 1.Post-Reform t + β 2.E i,t + β 3.dE i,t + β 4.dE i,t+1 + β 5.dNA i,t + β 6.dNA i,t+1 + β 7.RD i,t + β 8.dRD i,t + β 9.dRD i,t+1 + β 10.D i,t + β 11.dD i,t + β 12.dD i,t+1 + β 13.dV i,t + β 14.C i,t + β 15.C i,t.post-reform t + ε i,t (1) where dx t denotes changes in X from t-1 to t, V is the market value of equity plus the book value of debt, E denotes earnings (EBIT), NA denotes net assets (total assets minus cash), RD is research and development expense, D is common dividends, and C is cash. 36 Our focus is on the coefficient on cash, β 14, which we expect to be smaller for Japanese firms than U.S. firms in the 1990s, and to increase for Japanese firms as governance improves after We test the latter prediction by interacting C i,t with a post-reform dummy set to 1 from 2001 to 2007 and 0 otherwise. The second specification is similar and follows Dittmar and Mahrt-Smith (2007): 36 Once again, we have replicated the analyses in this section using cash + marketable securities in place of cash, with similar (somewhat stronger) results. 31

33 V i,t = α + β 1.Post-Reform t + β 2.E i,t + β 3.dE i,t + β 4.dE i,t+2 + β 5.RD i,t + β 6.dRD i,t + β 7.dRD i,t+2 + β 8.D i,t + β 9.dD i,t + β 10.dD i,t+2 + β 11.dNA i,t + β 12.dNA i,t+2 + β 13.dV i,t+2 + β 14.C i,t + β 15.C i,t.post-reform t + ε i,t (2) Here, all variables are deflated by NA t, and dx i,t denotes changes from t-2 to t. Our focus is again on β 14 and β 15, with the expectation that β 14 will be smaller for Japanese firms and that β 15 will be positive for Japanese firms as the cash coefficient increases post reforms. We report the results of these analyses in Table 7. We estimate each regression as an unbalanced panel over by country, with two-way clustering of standard errors. We discuss results for the Pinkowitz et al. (2006) specification first. The regressions have relatively high explanatory power, as expected for these types of levels specifications. For the US, the R-squared is 29.2% while for Japan it is 36.3%, with coefficients on most variables taking the expected signs. Consistent with our predictions, the coefficient on cash is higher for US firms than Japanese firms in the pre-reform years: the coefficient for US firms is 3.48 (t = 7.32) compared to 0.70 (t = 5.15) for Japanese firms, numbers that are statistically different at better than 1% (t = 5.95). The coefficient for Japanese firms increases by 0.70 (t = 2.20) in , consistent with improved governance, and the overall coefficient on cash is no longer significantly different for US and Japanese firms. 37 We obtain similar results for the Dittmar et al. (2007) specification. For the US, the R-squared is 45.6% while for Japan it is 58.1%, with coefficients on most variables again taking the expected signs. Over the coefficient on cash for US firms is 3.97 (t = 5.92) compared to 1.63 (t = 4.15) for Japanese firms, numbers that are statistically different at better than 1% (t = 5.12). The coefficient on cash for Japanese firms again increases in 37 These results are robust to a specification that replaces cash with the change in cash. We have also estimated (1) using annual cross-sections by country (after omitting the period dummy). The coefficients on cash are always larger for US than Japanese firms over 1990 to 2000 (differences are statistically significant in all but one year), and with the differences noticeably smaller (and significant in only three years) over

34 , but in this specification does not reach significance (t = 1.44). And here again we have also estimated the regressions by year, with results again showing larger coefficients on cash for US firms in the 1990s, with differences consistently diminishing in the 2000s. Overall, the evidence from these regressions is largely consistent with our predictions: the valuation of the cash of Japanese firms is low during the 1990s, both in absolute terms and relative to that of U.S. firms, but improves in the 2000s, to levels more comparable to those of U.S. firms Summary and Conclusions We investigate whether the governance practices of Japanese companies, as manifested in their holdings of cash and payout policy, improve over the past two decades, and focus in particular in the period since 2000, which approximately represents a regime shift that improves governance in Japan. While overall levels of cash holdings are now roughly the same for US and Japanese companies (and US firms hold more cash and marketable securities), when we condition on firm characteristics Japanese firms still hold substantially more cash than US firms (even when we include marketable securities). However, Japanese firms hold 3% to 4% less cash (as a fraction of assets) than in the 1990s and collectively have substantially increased their dividends and repurchases since the late 1990s. Consistent with the idea that improvements in governance manifest themselves in lower holdings of cash, there is an inverse relation between changes in the (excess) cash holdings of Japanese firms and changes in their performance since Also consistent with improved governance for some Japanese firms, increases in payouts, especially dividends, are associated with improved performance. 38 We have also examined whether these results differ for keiretsu and non-keiretsu Japanese firms but find little evidence of consistent or reliable differences. 33

35 Ownership attributes associated with good governance, such as foreign ownership and management ownership, increase over the 2000s, while the influence of banks declines. Further, these characteristics are related to changes in firm performance as well as to changes in excess cash holdings and payout policy in ways we might expect if they are picking up the relevant attributes of governance. Finally, we also find that the market valuation of cash differs significantly for Japanese and US firms. In the 1990s, when we posit that governance in Japan is relatively poor, the market valuation of cash is systematically lower for Japanese firms compared to their US counterparts, consistent with market participants discounting their cash holdings due to concerns about Japanese governance. These differences decline after 2000, consistent with generally improved governance in Japan. Overall, our findings support two broad conclusions. First, governance practices in the average Japanese firm have improved since the 1990s, at least as manifested in their management of cash holdings and payouts to shareholders. Further, there is evidence that at least some Japanese firms those that pay both dividends and repurchases now distribute substantial amounts of cash to shareholders on a regular basis. Second, those Japanese firms that improve their management of cash and/or increase payouts enjoy improved performance. These changes are linked to changes in the ownership characteristics of Japanese firms in ways consistent with them being driven by improvements in governance. This evidence offers hope that additional improvements in the governance of Japanese companies can further improve corporate performance and perhaps help reboot private sector economic performance in Japan, an important part of the Abe Government s plan for pulling Japan out of its longstanding economic malaise. 34

36 Appendix: Repurchases in Japan: Institutional details and measurement Beginning in 1994, Japan gradually lifted restrictions on firms ability to repurchase shares. In 1994, Japan modified the law to allow repurchases but only by permanently retiring shares using retained earnings. However, because of uncertainty about the tax treatment of such retirements, managers did not actually use this method until November The special law for stock retirement in June 1997 further liberalized the rules for stock repurchases by removing restrictions over their timing. Under the 1994 regime, firms could only consider plans to make repurchases once a year, at the annual shareholders meeting, and these plans were subject to shareholder approval. Under the special law, once managers obtained approval from shareholders to set up a maximum amount and number of shares for repurchases in the corporate articles, managers could then make repurchases decisions (amount and timing) without shareholder approval. The subsequent amendment of this special law in 1998 and the enactment of the law for evaluation of land in 1999 expanded the components of shareholders equity that were available to fund repurchases: the former included legal capital surplus; the latter added revaluation reserves from land. 40 Related laws liberalized firms ability to make repurchased shares available for management incentive compensation, including stock options. 39 The Japanese Tax Code treats repurchases as a return of net assets to shareholders, that is, the payment is a combination of contributed capital and retained earnings. Similar to the tax treatment of dividends, the Tax Code could recognize that part of the payout corresponding to retained earnings as taxable income. As a result, shareholders receiving cash from such repurchases would pay both capital gains and income taxes, effectively being taxed twice. Because of this treatment, in June 1995, the Japanese tax office announced its intention to exclude repurchases by listed firms from dividend taxation. The corresponding amendment of the Tax Code was enacted in November In general, Japanese law (Commercial Code) has very detailed restrictions regarding firms ability to make dividends and repurchases that are based on various components of shareholders equity. These restrictions are similar in spirit to corporate laws in US states, as well as debt covenants in loan agreements, that typically restrict firms ability to pay dividends to retained earnings or some fraction thereof. The difference is that Japanese legal and accounting practices mean that Japanese firms have numerous categories of capital, retained earnings, and reserves within shareholders equity on the balance sheet. 35

37 The 2001 modified Commercial Code further expanded firms ability to make repurchases. First, for the first time the treasury stock method was permitted. Second, minimum capital restrictions over firms ability to pay out cash as either dividends or repurchases were relaxed. Under the previous law, firms with insufficient capital needed to increase legal retained earnings before making payouts. Under the modified law, firms for which the sum of legal retained earnings and legal capital surplus, instead of legal retained earnings alone, exceeded one quarter of capital stock were exempt from this requirement. Third, firms that met the above capital requirements could reduce the excess portion of the legal capital surplus and add it to retained earnings. Previously, the reduction in legal capital surplus was admitted either to offset loss carry forwards or to increase in capital stock. Further legal changes in 2006 effectively lifted all remaining restrictions on managers ability to return cash to shareholders, including repurchases. Most of the data we use for the study are from Worldscope. However, Worldscope does not separate purchases and sales of common and preferred stock, making it difficult to properly measure repurchases. Consequently, we measure repurchases for Japanese firms using data from Nikkei Quick that covers the relevant corporate news releases, supplemented as necessary by data drawn directly from firms financial statements (managers of Japanese firms report repurchases in disclosures that are required under the securities laws and by the stock exchanges). 36

38 REFERENCES Aggarwal, R., I. Erel, R. Stulz, and R. Williamson Differences in governance practices between US and foreign firms: Measurement, causes, and consequences. Review of Financial Studies 22: Aoki, M Whither Japan s corporate governance? In M. Aoki, G. Jackson, and H. Miyajima (eds). Corporate governance in Japan: Institutional change and organizational diversity. Oxford University Press. Aoki, M., H. Patrick, and P. Sheard The Japanese Main Bank System: An Introductory Overview. In The Japanese Main Bank System. Masahiko Aoki and Hugh Patrick (eds.), Oxford University Press. Bates, T. W., K. M. Kahle, and R. M. Stulz Why do US firms hold so much more cash than they used to? Journal of Finance 64: Becht, M., J. Franks, L. Mayer, and H. Wagner The returns to hedge fund activism: An international study. Working paper, January. Available at Brav, A., J. R. Graham, C. R. Harvey, and R. Michaely Payout policy in the 21st century. Journal of Financial Economics 77: Brav, A., W. Jiang, and H. Kim Hedge fund activism: A review. Foundations and Trends in Finance 4, 3: DeAngelo, H., L. DeAngelo, and D. J. Skinner Are dividends disappearing? Dividend concentration and the consolidation of earnings. Journal of Financial Economics 72: Denis, D. J., and I. Osobov Why do firms pay dividends? International evidence on the determinants of dividend policy. Journal of Financial Economics 89: Dewenter, K. L., and V. A. Warther Dividends, asymmetric information, and agency conflicts: Evidence from a comparison of the dividend policies of Japanese and US firms. Journal of Finance 53: Dittmar, A., and J. Mahrt-Smith Corporate governance and the value of cash holdings. Journal of Financial Economics 83: Doidge, C., G. A. Karolyi, and R. M. Stulz Why are foreign firms listed in the US worth more? Journal of Financial Economics 71: Fama, E., French, K., Taxes, financing decisions, and firm value. Journal of Finance 53,

39 Fama, E. F., and K. R. French New lists: Fundamentals and survival rates. Journal of Financial Economics 73: Floyd, E., N. Li, and D. J. Skinner Payout policy through the financial crisis: The growth of repurchases and the resilience of dividends. Unpublished paper, University of Chicago, February. Available at French, K.R., and J.M. Poterba Were Japanese stock prices too high? Journal of Financial Economics 29: Fukao, M Japan s Lost Decade and its Financial System. The World Economy Vol. 26 No. 3. March 2003: Grullon, G., and R. Michaely Dividends, share repurchases and the substitution hypothesis. Journal of Finance 57: Hamao, Y., K. Kutsuna, and P. Matos U.S.-style activism in Japan: The first ten years. Working paper, USC and Kobe University. February 5. Available at Hayn, C., The information content of losses. Journal of Accounting and Economics 20: Hoshi, T., and A. K. Kashyap Corporate Financing and Governance in Japan: The Road to the Future. Cambridge, Mass.: The MIT Press. Hoshi, T, and A. K. Kashyap Japan s financial crisis and economic stagnation: Journal of Economic Perspectives 18: Hoshi, T., A. Kashyap, and D. Scharfstein The role of banks in reducing the costs of financial distress in Japan. Journal of Financial Economics 27: Jackson, G., and H. Miyajima Introduction: The diversity and change of corporate governance in Japan. In M. Aoki, G. Jackson, and H. Miyajima (eds). Corporate governance in Japan: Institutional change and organizational diversity. Oxford University Press. Jensen, M. C Eclipse of the public corporation. Harvard Business Review (September- October): Kang, J., A. Shivdasani, T. Yamada The effect of bank relations on investment decisions: An investigation of Japanese takeover bids. Journal of Finance 55: Kaplan, S. N Top executive rewards and firm performance: A comparison of Japan and the US Journal of Political Economy 102: Klein, A., and C. Marquardt Fundamentals of Accounting Losses, The Accounting Review 81:

40 Larcker, D., S. Richardson, I. Tuna Corporate governance, accounting outcomes, and organizational performance. The Accounting Review 82: Milhaupt, C.J Lost decade for Japanese corporate governance reform? What s changed, what hasn t, and why. In: M. Blomstrom and S. LaCroix (eds.). Institutional Change in Japan. Routledge, Miyajima, H Institutional change and its economic consequences in Japan: the bright and dark sides of hybridization. Working paper, Wasda University, August. Miyajima, H., and F. Kuroki The unwinding of cross-shareholding in Japan: Causes, effects, and implications. In M. Aoki, G. Jackson, and H. Miyajima (eds). Corporate governance in Japan: Institutional change and organizational diversity. Oxford University Press. Morck, R., and M. Nakamura Japanese corporate governance and macroeconomic problems. In Masao Nakamura, ed. The Japanese Business and Economic System - History and Prospects for the 21st Century. Palgrave, Houndmills, UK, Morck, R., and Nakamura A frog in a well knows nothing of the ocean: A history of corporate ownership in Japan. In: Randall Morck, ed. A Global History of Corporate Governance, Chicago: University of Chicago Press. Morck, R., A. Shleifer, and R. Vishny Management ownership and market valuation: An empirical analysis. Journal of Financial Economics 20: Nakamura, M Adoption and policy implications of Japan s new corporate governance practices after the reform. Asia Pacific Journal of Management 28: Opler, T., L. Pinkowitz, R. Stulz, and R. Williamson The determinants and implications of corporate cash holdings. Journal of Financial Economics 52: Patrick, H Evolving corporate governance in Japan. Unpublished paper. Columbia Business School. February. Pinkowitz, L., R. Stulz, and R. Williamson Does the contribution of corporate cash holdings and dividends to firm value depend on governance? A cross-country analysis. Journal of Finance 61: Pinkowitz, L., R. M. Stulz, and R. Williamson Do US firms hold more cash? Working paper, April. Available at Pinkowitz, L., and R. Williamson Bank power and cash holdings: Evidence from Japan. Review of Financial Studies 14: Rajan, R.G., and L. Zingales What do we know about capital structure? Some evidence from international data. Journal of Finance 50:

41 Skinner, D. J The Evolving Relation between Earnings, Dividends, and Stock Repurchases. Journal of Financial Economics 87: Suda, K., and A. Shuto Earnings management to avoid earnings decreases and losses: Empirical evidence from Japan. Working paper, April. Available at Uchida, K. and P. Xu US Barbarians at the Japan gate: Cross border hedge fund activism. Unpublished paper, University of Kitakyushu and Honsei University. February. Urasaki, N Institutions and accounting standard transformation: Observations from Japan. China Journal of Accounting Research 7: Weinstein, D.E., and Y. Yafeh On the costs of a bank centered financial system: Evidence from the changing main bank relations in Japan. Journal of Finance 53: Woodford, M Exposure. Penguin Books: London. 40

42 Figure'1A:'Cash'holdings'of'Japanese'industrials,' '(mean,' median,'and'aggregate)' 0.200# 0.180# 0.160# 0.140# 0.120# 0.100# 0.080# Asset1weighted# mean# median# 0.060# 0.040# 0.020# 0.000# 1991# 1993# 1995# 1997# 1999# 2001# 2003# 2005# 2007# 2009# 2011# 41

43 Figure'1B:'Holdings'of'marketable'securi8es'by'Japanese'industrials,' 1990?2011'(deflated'by'assets)' 0.070# 0.060# 0.050# 0.040# mean# 0.030# median# 0.020# 0.010# 0.000# 1990# 1991# 1992# 1993# 1994# 1995# 1996# 1997# 1998# 1999# 2000# 2001# 2002# 2003# 2004# 2005# 2006# 2007# 2008# 2009# 2010# 2011# 42

44 Figure 2: Aggregate dividends and repurchases for Japanese industrial firms, (in 1991 constant yen, trillions) Repurchases Dividends