The impact of the Bank of Japan’s exchange traded fund and corporate bond purchases on firms’ capital structure

Abstract

This study examines how the Bank of Japan’s (BOJ’s) exchange-traded fund (ETF) and corporate bond (CB) purchases affect the capital structure of Japanese listed firms. The results first suggest that the BOJ’s ETF purchases reduced the equity risk premium for firms included in the purchases, thereby lowering their cost of equity. These firms issued more stocks and became less dependent on bond debt and bank loans than control firms, resulting in a lower level of leverage. On the other hand, the BOJ’s CB purchases lowered the risk premium of bonds eligible for such purchases, which implies a decline in the cost of bond debt. Firms eligible for CB purchases issued more bonds, while reducing bank debt but to a smaller extent, thus having higher leverage than firms not eligible for CB purchases. Categorizing firms into four groups based on whether their stocks were included in ETF purchases and/or their bonds were eligible for CB purchases further shows that the policy impacts on firm financing have differed depending on which group a firm fell in. These results indicate that the BOJ’s ETF and CB purchases have had a considerable impact on the cost of capital, securities issuance activities, and corporate capital structure.

1 Introduction

In the aftermath of the 2008 global financial crisis, with the nominal interest rates reaching their zero-lower bound, central banks in many advanced countries started to adopt unconventional monetary policies to provide additional monetary stimulus and revitalize the economy. One of the key unconventional policy tools introduced after the crisis was the large-scale asset purchases (LSAPs), under which different types of risk assets were purchased by major central banks including the Federal Reserve (Fed), the Bank of Japan (BOJ), the European Central Bank (ECB), and the Bank of England. These purchasing programs have been subsequently expanded and maintained to this day, and are expected to ease financial market conditions, thereby stimulating corporate investment and job creation.¹ In this spirit, the effects of LSAPs on the corporate sector in general and corporate financing activities in particular are of first-order importance; however, the previous literature provides limited evidence on this aspect.

Against this background, this study aims to examine the impacts of the BOJ’s purchases of exchange-traded funds (ETFs) and corporate bonds (CBs) on firms’ capital structure and financing choices. The ETF and CB purchases were parts of the BOJ’s LSAPs that was unprecedented at the time. Specifically, since the introduction of Comprehensive Monetary Easing (CME) policy in 2010, the BOJ has purchased four types of risk assets, namely, commercial paper, CBs, ETFs, and Japan real estate investment trusts (J-REITs) on a large scale.² Through the purchase of ETFs and CBs, the BOJ indirectly owns firms’ floating stocks and directly owns corporate bonds, so that the intervention likely affects the financing conditions of firms whose stocks or bonds are included in the purchases.

The research questions this study attempts to answer are the following: (1) Did the BOJ’s ETF and CB purchases affect the firms’ access to external financing sources? (2) Did the ETF and CB purchasing program have a significant impact on firms’ capital structure and their financing choices? And (3) Are there any substitution effects among financing sources, such as between bank loans and bond debt, or equity and debt capital?

Regarding the impacts of the BOJ’s risk asset purchases, the program’s stated objective is to enhance monetary easing by encouraging a decline in risk premiums. This purpose can be achieved through the following mechanism. In times of financial market uncertainty, investors may require a significant risk premium to compensate them for holding risky assets, and asset prices may fall substantially. The BOJ’s asset purchases may increase asset prices and valuations, and reduce price volatility, thereby lowering the risk premium on those purchased assets. Previous studies have provided evidence supporting this mechanism, that the ETF purchases by the BOJ raise stock prices (e.g., Charoenwong et al., 2021; Harada & Okimoto, 2021) and reduce equity risk premium (e.g., Adachi et al., 2021; Katagiri et al., 2022), while the CB purchases lead to a decrease in credit spreads (Suganuma & Ueno, 2018). Notably, firms’ cost of capital (i.e., cost of equity, cost of debt) can be reduced if the risk premiums are lowered. This has important implications for countries like Japan, where the risk-free rate is close to zero and cannot further decline.

Therefore, I hypothesize that the BOJ ETF purchases may have reduced the cost of equity capital for treatment firms, providing them with more favorable equity market conditions than control firms. To raise capital, ETF-treatment firms may have timed the market and relied more on stock issuance, e.g., as suggested by the equity market timing (see Baker & Wurgler, 2002; Huang & Ritter, 2009). In addition, these firms may have relied less on debt financing, lowering their leverage ratios overall. On the other hand, the BOJ CB purchases may have reduced the yield on eligible bonds, making it easier for issuing firms of such bonds to access the public bond market. Eligible firms will likely have increased their bond issuance and may have also substituted bank loans with bond debt, but the magnitude of such a substitution effect may have been small. As a result, it is likely that the leverage ratios of eligible firms increased following the intervention. Moreover, the policy impact may have differed depending on whether these firms are subject to both purchase programs, only ETF purchases, only CB purchases, or neither of them. To the best of the author’s knowledge, this study is the first to examine the impact of the BOJ purchases of risky assets on capital structure.

In the baseline analysis, several datasets are utilized to test the above hypotheses. The primary data is for non-financial firms listed on the first and second sections of the Tokyo Stock Exchange (TSE1 and TSE2) covering the period from 2009 to 2018 obtained from Nikkei NEEDS Financial Quest (Nikkei FQ). To compute equity and bond risk premium, I further collected the beta for individual stocks and market risk premium from Bloomberg, bond transactions data published by the Japanese Securities Dealers Association (JSDA), and the risk-free rate from the Ministry of Finance website. Macroeconomic data were from the databases of the World Bank and the International Monetary Fund. Using this comprehensive dataset, a difference-in-differences (DID) framework was applied to estimate the impact of the purchasing programs on risk premiums and firms’ capital structure, exploiting the fact that it is possible to identify stocks being indirectly purchased via ETFs and bonds being eligible for CB purchases by the BOJ. Meanwhile, the policy impact on firms’ stock and bond issuance (i.e., the probability of issuance and issuance amounts) was examined using a logit and a tobit model. In addition, the combined impact of ETF and CB purchases was also considered by categorizing firms into four groups based on whether their stocks were included in ETF purchases and/or their bonds were eligible for CB purchases by the BOJ. In the supplementary analysis, to quantify the magnitude of the treatment effect more precisely, changes in the BOJ’s total purchasing amount and the amount allocated to each type of ETF were taken into account to measure the treatment firms’ different exposure levels to BOJ’s ETF purchases.

The results show that as hypothesized, the ETF and CB purchases by the BOJ have had a considerable impact on corporate capital structure with the following major findings. First, the BOJ’s ETF purchases led to a reduction in beta value and equity risk premium of firms included in such purchases. These firms issued stocks more frequently with larger amounts than before, and became less dependent on bond issuance and bank debt, resulting in a lower leverage ratio than firms in the control group. Second, following the introduction of the BOJ’s CB purchases, the credit spreads of eligible bonds decreased. Eligible firms (firms whose bonds were eligible for CB purchases) have been issuing bonds more frequently with larger amounts, while reducing bank debt by a smaller extent, thus having a higher leverage ratio than ineligible firms. Third, categorizing firms into four groups—firms that were subject to both purchases, only ETF or CB purchases, or neither ETF nor CB purchases—shows that the policy impacts on firms’ leverage ratio have differed depending on which group a firm fell in. After the policy intervention, firms included in BOJ ETF purchases only saw a decrease in their leverage. In contrast, the leverage ratio of firms in the group eligible for CB purchases only increased. Firms in the group subject to both ETF and CB purchases strengthened their bond and stock issuance activities; however, the impact on their leverage ratio is small, because the positive and negative impacts cancelled each other out. Fourth, the evidence further suggests that the positive effect of ETF (CB) purchases on stock (bond) issuance is 1.3–2.2 times greater for firms subjected to solely ETF (CB) purchases than for those subjected to both purchases. Presumably, firms included in both purchases have lower costs of equity and bond debt, and can flexibly choose to issue bonds or stocks to raise capital. In contrast, firms included in solely ETF purchases have a lower cost of equity, but their cost of bond debt is relatively high. As a result, these firms relied more on stock issuance and may replace bonds with stocks to take advantage of the market conditions. A similar argument could be made for firms included in solely CB purchases. Supplemental analyses and robustness checks confirm these findings.

The remainder of the study is organized as follows. Section 2 provides a review of the literature. Section 3 describes the institutional framework of the BOJ’s ETF and CB purchases, while Sect. 4 presents the hypotheses and describes the empirical approach and data used. Next, Sect. 5 provides summary statistics, the regression results and robustness checks. Finally, Sect. 6 offers conclusions and suggestions for future research.

2 Literature review

2.1 Previous studies on capital structure

Capital structure theory originated with the seminal paper by Modigliani and Miller (1958). The relaxation of some of Modigliani and Miller’s (1958) restrictive assumptions has led to the formation of some well-known theories of capital structure: the trade-off theory, pecking order theory, signaling theory, and market timing theory.

There are two strands of trade-off theory, both of which are based on the concept that firms set a target debt-equity ratio by weighing up the various benefits and costs of borrowing. The first strand considers the trade-off between the costs of financial distress and the benefits of interest tax shield, namely, that interest payments are tax-deductible (Miller, 1977; Modigliani & Miller, 1963). On the other hand, the second strand focuses on the agency costs and the benefits of leverage (Jensen & Meckling, 1976; Myers, 1977). These studies found that while leverage has agency benefits because it increases managers’ ownership concentration and commitment, agency costs arise when there are conflicts of interest between stakeholders, especially when firms are in financial distress.

Next, the pecking order theory and signaling theory are based on the concept of asymmetric information. Signaling theory was first proposed by Ross (1977), who argued that managers may use leverage as a credible signal to investors of the firm’s value and ability to generate future cash flow, so that superior firms will choose a sufficiently high leverage that other firms in the industry cannot imitate. Furthermore, when managers have private information about the value of the firm, investors will discount the price they are willing to pay for a new equity issue due to adverse selection. The introduction of adverse selection into theoretical models led to the emergence of the pecking order hypothesis, which states that to minimize adverse selection costs related to the issuance of securities, firms will rely on internal finance first, followed by the issuance of debt, and finally the issuance of equity (Myers & Majluf, 1984).

Empirical studies on capital structure have focused on testing the implications of the various theories. They were able to show causal relationships between firms’ choice of capital structure on the one hand and various firm characteristics, including the tangibility of their assets, corporate tax rate, as well as their size, profitability, firm age, and growth opportunities, on the other. However, findings on the direction in which some of these firm-specific factors affect capital structure are rather mixed. For example, some researchers found a positive link between firms’ profitability and their leverage ratio (e.g., Fama & French, 2002; Frank & Goyal, 2009; Rajan & Zingales, 1995; Titman & Wessels, 1988), while others found a negative link (e.g., Myers, 1984; Myers & Majluf, 1984). Given these conflicting results, Graham and Leary (2011) argue that empirical research has failed to explain the heterogeneity in firms’ debt composition, changes in leverage, and decision to issue securities. They further claim that it is not only corporate demand for outside financing that determines firms’ capital structure; rather, the supply of capital may potentially also affect their capital structure.³

With this background, several recent studies have verified the role of supply factors in capital structure decisions by employing two major approaches. The first approach focuses on segmentation in debt markets between financially constrained bank-dependent firms and unconstrained firms with access to arms-length lenders (Faulkender & Petersen, 2006), or examines different types of shocks to intermediary capital (e.g., as in Khwaja & Mian, 2008; Leary, 2009; Rice & Strahan, 2010; Sufi, 2009). Meanwhile, the second approach investigates the impact of market conditions, revealing that the changes in capital supply conditions (e.g., equity and credit market conditions) are relevant for capital structure choices and firms’ financing behavior, as in Baker and Wurgler (2002) and Huang and Ritter (2009).⁴ More specifically, Baker and Wurgler (2002) firstly proposed the market timing theory, which suggests that firms tend to issue equity when their market valuations are high and repurchase equity when their market valuations are low. Huang and Ritter (2009) used the equity risk premium as a proxy for the cost of equity and presented evidence that firms are more likely to issue equity when the relative cost of equity is low. Consistent with the market timing theory, their result implies that when the equity risk premium is low, external equity is not necessarily more expensive than external debt. The present study is closely related to this second approach, considering that the BOJ’s purchases of ETFs and CBs can be regarded as exogenous changes in the supply side of capital, and could potentially affect firms’ access to external financing sources.

2.2 Literature on the impact of central banks’ LSAPs

The preceding subsection provided a review of capital structure theories and indicated the importance of market conditions as determinants of firms’ capital structure. The LSAPs implemented by central banks are potential sources that may lead to changes in market conditions, and the BOJ was one of the first to introduce such measures.

Regarding the BOJ’s purchases of CBs and/or ETFs, a number of studies have found significant impacts of the policy on different aspects, including credit spreads, the stock prices and stock valuations, or firm performance and investment. In particular, for the study on the purchases of corporate bonds, Suganuma and Ueno (2018) show that the BOJ’s purchases of corporate and government bonds lead to a decrease in Japanese firms’ credit spreads through several channels: the default risk channel, the local and global supply channels, and the risk-taking channel. For the studies on ETF purchases, several authors focus on the policy impact on stock prices and equity risk premium. Barbon and Gianinazzi (2019) propose a theoretical asset pricing model which provides them with testable predictions about the effect on equity prices. Their results highlight that the ETF purchases by the BOJ generate a positive and persistent effect on stock prices, which in turn can reduce the cost of equity capital for Japanese firms. Harada and Okimoto (2021) share the same view that the BOJ’s interventions have a considerable impact on daily stock prices by showing that the afternoon returns of Nikkei 225 stocks have been significantly higher than those of non-Nikkei 225 stocks on days when the BOJ purchased ETFs. Meanwhile, Adachi et al. (2021) find evidence that the BOJ’s ETF purchases have lowering effects on equity risk premium. A similar conclusion is reached by the study of Katagiri et al. (2022), which indicates that such a lowering effect was achieved through a decrease in market beta for Japanese stocks.

The literature also focuses on other aspects of ETF purchases beyond the impact on stock prices and risk premium. Gunji et al. (2021) estimate the influence of the BOJ’s ETF purchases on corporate performance. Their DID analysis suggests that the policy has lowered firms’ profits. Charoenwong et al. (2021) find evidence consistent with Harada and Okimoto (2021) that the BOJ’s ETF purchases boost share prices and share valuations. The authors further argue that the ETF purchases encourage firms to make use of equity financing, increase firms’ cash holdings and short-term investment, but the policy does not prove to be an effective way to stimulate real tangible capital investment.

In short, there are a number of studies examining the impact of risky asset purchases by the BOJ from a variety of angles. Although none of these studies focused on firms’ capital structure, the results they obtained provide important evidence that helps explain the underlying mechanism that will be discussed in Sect. 4.1 of how this policy may affect firms’ capital structure.

The BOJ’s unconventional monetary policy of risky asset purchases was followed by the ECB. Several studies concentrate on LSAPs implemented by the ECB and investigate the effects on firms’ capital structure. For instance, Grosse-Rueschkamp et al. (2019) examine the ECB’s Corporate Sector Purchase Programme (CSPP) implemented since 2016, under which the ECB purchases corporate sector bonds. The authors propose a “capital structure channel” of monetary policy: direct corporate bond purchases by central banks reduce the bond yields of firms whose bonds are eligible for these purchases. These firms then substitute bank loans with bond debt. This relaxes banks’ lending constraints and enables banks to increase lending to firms whose bonds are not eligible for such purchases. The existence of this channel is corroborated by Arce et al. (2021) and Betz and De Santis (2022). While both studies conclude that the CSPP has led to an increase in bank lending to SMEs, Betz and De Santis (2022) argue that the spillover effect is independent of the quality of bank balance sheets. On the other hand, Adelino et al. (2022) show that the CSPP can have spillover effects on firms’ capital structure through the “trade credit channel”: eligible firms under the CSPP can pass on the additional funding liquidity to their customers through trade credit, thereby enhancing their competitive position in the market.

However, the current study differs from that of Grosse-Rueschkamp et al. (2019) and other existing studies (e.g., Adelino et al., 2022; Arce et al., 2021; Betz & De Santis, 2022) in a number of ways. First, whereas these studies focus on CB purchases only, the present study focuses on purchases of both CBs and ETFs. This focus of interest not only reflects the difference between the ECB’s and BOJ’s LSAPs, but also allows me to gauge the combined effects of the ETF and CB purchases, namely, the effects on firms that were subject to both purchases, only ETF or CB purchases, or neither of them. Second, this study finds that LSAPs (BOJ purchases of ETFs and CBs) have had considerable direct effects on firms’ capital structure. In contrast, previous studies show that the impact of LSAPs (ECB purchases under the CSPP) on firms’ leverage ratio is insignificant. Presumably, the difference is due to a smaller extent of substitution between bond and bank debt for listed firms in Japan than that in ECB’s member countries. Third, while Grosse-Rueschkamp et al. (2019) use a treatment group dummy variable, this study takes one step further in its supplementary analysis by measuring the different levels of exposure of treatment firms to ETF purchases, thereby quantify the magnitude of the treatment effect more precisely.

To sum up, previous studies have identified some potential channels through which LSAPs may affect corporate capital structure. However, to date, there are no studies investigating the impact on this aspect of LSAPs other than the CB purchase program by the ECB. To the best of the author’s knowledge, this study is the first to examine the impact of multiple asset purchase programs (BOJ purchases of ETFs and CBs) on corporate financing activities, and to find evidence of considerable direct impacts of LSAPs on firms’ capital structure.

3 The BOJ’s purchases of risky assets

This section provides an overview of the BOJ’s policy of purchasing risky assets. It starts with a brief review of unconventional monetary policy in Japan and then describes the BOJ’s policy of purchasing ETFs and CBs, concentrating on the period from 2009 to 2018 (i.e., the analysis period of this study).

3.1 Unconventional monetary policy in Japan

Japan was the first country to introduce unconventional monetary policies, namely the zero interest rate policy during February 1999 to August 2000, and quantitative easing (QE) during March 2001 to March 2006. Following the implementation of QE, in 2010, the BOJ introduced Comprehensive Monetary Easing (CME) and expanded it in the following years to mitigate the negative impact of the global financial crisis in 2008. The most important element of CME was the credit easing policy, under which the BOJ purchased risky assets, including commercial paper, CBs, ETFs and Japan real estate investment trusts (J-REITs) through the establishment of the Asset Purchase Program to compress the risk premium.

Although CME led to an improvement in the economy and rising prices, that improvement did not last. Therefore, in April 2013, the BOJ introduced Quantitative and Qualitative Monetary Easing (QQE). Under this policy, the BOJ conducted operations to increase the monetary base at an annual pace of about 60–70 trillion yen and set an inflation target of 2 percent. In addition, the BOJ increased its purchases of Japanese government bonds (JGBs) and risky assets, especially ETFs, on a massive scale.

QQE quickly had a considerable impact and contributed to the economic recovery, but the price stability target of 2 percent was not attained. In January 2016, it introduced QQE with a Negative Interest Rate. Further, in September 2016, the BOJ announced the adoption of QQE with Yield Curve Control, which consists of two main components. The first is “yield curve control,” under which the BOJ controls short-term and long-term interest rates through market operations. The second component is an “inflation-overshooting commitment,” under which the BOJ commits itself to expanding the monetary base until the yearly increase in the CPI exceeds the 2 percent price stability target.

3.2 The BOJ’s ETF and CB purchasing policy

A key component of the BOJ’s unconventional monetary policy, as mentioned, is the Asset Purchase Program. To give an impression of the scale of the program and the relative importance of the different types of asset purchases, Fig. 1 shows the total outstanding amount of the BOJ’s commercial paper, CB, ETF and J-REIT holdings from 2010 to 2019.

Fig. 1

Total outstanding amount of BOJ’s risk asset holdings. Source: Based on data from the Bank of Japan (https://www.boj.or.jp/en/statistics/boj/other/acmai/release/index.htm). This figure utilizes data from the BOJ’s balance sheet to depict the total outstanding amount of risk assets the BOJ holds, including commercial papers, corporate bonds, ETFs, and J-REITs, from 2010 to 2019

At the beginning of 2011, when the BOJ started to buy risk assets, the outstanding amount of commercial paper was the largest, followed by CBs, ETFs, and J-REITs, respectively. However, since the introduction of QQE in 2013, the outstanding amount of ETFs has increased dramatically and has surpassed all other risk assets purchased by the BOJ: the BOJ’s ETF holdings rose from 0.185 trillion in 2010 to 2.5 trillion in 2013 and 23.5 trillion at the end of 2018, whereas the outstanding amount of CBs, J-REITs, and commercial paper holdings since 2013 has remained more or less unchanged. Because of the large scale of such purchases and the continuing nature of this intervention, the BOJ’s ETF and CB purchases can be regarded as a positive supply shock in the availability of funds for firms that were subjected to the purchases.

3.2.1 Outline of ETF purchases

To date, the BOJ is the only central bank in the world that indirectly holds company stocks through the purchase of ETFs.⁵ According to the BOJ’s principal terms and conditions,⁶ the BOJ purchases ETFs via trust banks, and from the beginning of ETF purchases in 2010 until March 2021, such purchases focused on two main types of ETFs: ETFs that track the Tokyo Stock Price Index (TOPIX) and ETFs that track the Nikkei 225 Stock Average (Nikkei 225).⁷

Furthermore, through the purchase of ETFs, the BOJ indirectly holds firms’ stocks and has become a top shareholder of many listed companies. Some researchers (e.g., Gunji et al., 2021; Samikawa & Takano, 2018) have tried to estimate the percentage of shares the BOJ holds in various companies. Stocks that have higher Nikkei 225 and/or TOPIX weights are likely to have a higher percentage of shares held by the BOJ.

Table 1 presents the 26 companies with the highest BOJ shareholding rates as of July 2018. The table indicates that Advantest, Fast Retailing, Taiyo Yuden, TDK, and Family Mart UNY Holdings are the five companies with the highest BOJ indirect shareholding ratios. Remarkably, on a floating stock basis, the ratio of shares indirectly held by the BOJ for Fast Retailing—a company that owns various fashion brands such as UNIQLO and GU—was 88.3 percent as of July 2018, substantially higher than for all other companies.

Table 1 Companies with highest indirect BOJ shareholding ratios as of July 2018

3.2.2 Outline of purchases of CBs

Unlike in the case of ETFs, where the companies’ whose stocks were purchased by the BOJ are clearly known, data on companies whose bonds were purchased by the BOJ are highly confidential. However, the BOJ does publish the principal terms and conditions for outright purchases of CBs,⁸ including which CBs are eligible for purchase, the purchasing methodology, and the maximum outstanding amount to be bought. Specifically, eligible CBs must fulfill the general criteria, have a remaining maturity of 1–3 years and have a rating of BBB or higher by an eligible rating agency or, if they do not have a rating of BBB or higher, must be fully guaranteed by a company rated BBB or higher by an eligible rating agency. Thus, the CB purchase program focuses on short-term bonds with a high credit rating.

Regarding the purchasing methodology, the BOJ relies on multiple-price competitive auctions, where counterparties bid their desired yield at which they wish to sell CBs to the BOJ. The CBs purchasing price will then be determined using the yield from the auction, and the maximum outstanding amount of CBs by a single issuer that the BOJ shall purchase is 100 billion yen.

3.2.3 Purchasing rules for ETFs and CBs

Table 2 provides an overview of the BOJ’s major revisions of the ETF and CB purchases since the introduction of these programs until the end of 2018. With the purpose of ensuring stability in financial markets and facilitating corporate financing, the BOJ started the outright purchases of CBs in February 2009.⁹ The program ended on December 31, 2009, before restarting again in October 2010 under the CME policy.

Table 2 Changes in the BOJ’s ETF and CB purchase programs

When the BOJ introduced CME, the purchasing of risk assets was meant to be temporary. However, in practice, the BOJ did not abolish the policy at the end of 2011 as scheduled but in fact extended the program. From 2010 to 2012, the maximum purchasing amount for CBs always exceeded that for ETFs, thus it seems reasonable to conclude that before the introduction of QQE the BOJ put more emphasis on the purchase of CBs than ETFs.

Since October 2012, the eligibility criteria and the amount of CB purchases remained stable. In contrast, following the introduction of QQE, the purchasing rules for ETFs have been adjusted several times and the ETF purchase program has been substantially expanded in scale. As a result, the BOJ’s ETF holdings exceeded the amount of CBs holdings in September 2014, as shown in Fig. 1. Further, on July 29, 2016, the BOJ decided to increase the annual purchase amount from 3.3 trillion to 6 trillion yen and has maintained this annual pace up until now. On September 21, 2016, and July 31, 2018, the BOJ announced changes to the amount of money allocated for the purchase of each type of ETF.

Because the amount of each ETF purchased was to be roughly proportional to the total market value of that ETF issued, the BOJ for a long time—from 2010 until September 2016—focused largely on purchases of Nikkei 225 ETFs and spent more than half of its “budget” on ETFs that track this index. However, following the change in September 2016, the BOJ has gradually put more weight on ETFs tracking the TOPIX instead of the Nikkei 225. In March 2021, the BOJ announced that from April 2021 only ETFs whose prices track the TOPIX shall be purchased.¹⁰ The large purchases of Nikkei 225 ETFs have led to concerns about the possible distortion of stock prices (e.g., see Barbon & Gianinazzi, 2019) and negative effects on corporate performance and governance, so that the BOJ may have adjusted the allocation of purchases to mitigate this type of impact.

4 Methodology and data

This section provides an overview of the methodology and data used in this study. Specifically, Sect. 4.1 presents the hypotheses regarding the impact of the BOJ’s ETF and CB purchases that will be examined. Next, Sect. 4.2 outlines the methodology used for the baseline analysis, while Sect. 4.3 describes the data and variables employed.

4.1 Hypotheses

This subsection presents various hypotheses that will be empirically examined in Sect. 5. Regarding the BOJ’s direct purchases of risk assets including CBs and ETFs, the programs are expected to boost asset prices and valuations, and reduce price volatility, especially in times of market uncertainty or turmoil. This will lead to a decrease in the risk premium on those assets purchased by the BOJ. A number of studies confirm this mechanism by showing that the ETF purchases raise stock prices (Barbon & Gianinazzi, 2019; Charoenwong et al., 2021; Harada & Okimoto, 2021) and reduce equity risk premium (Adachi et al., 2021; Katagiri et al., 2022), while CB purchases reduce credit spreads (Suganuma & Ueno, 2018). Given this, the present study hypothesizes that the BOJ’s purchases of ETFs and CBs may have had a considerable impact on the cost of capital, securities issuance activities, and corporate capital structure.

Specifically, the BOJ’s ETF purchases may have led to a reduction in equity risk premium and in the cost of equity capital for firms whose stocks were included in the purchases, providing them with more favorable equity market conditions than control firms. To raise capital, these treatment firms may have timed the market and issued more stocks as implied by the equity market timing (e.g., see Baker & Wurgler, 2002; Huang & Ritter, 2009). Also, considering that these firms are large public firms that may not be significantly financially constrained, they may become less dependent on bond issuance and bank debt, thus lowering their leverage ratios overall. The first hypothesis could be presented as follows:

Hypothesis 1

(impact of the BOJ’s ETF purchases): Following the adoption of ETF purchases, firms whose stocks were included in the purchases experienced a decline in equity risk premium. These firms increased their stock issuance activities and substituted debt financing with equity financing. Consequently, these ETF-treatment firms decreased their leverage ratios relative to control firms.

Next, BOJ purchases of CBs may have reduced the yield on eligible bonds, making it cheaper for issuing firms of such bonds to raise capital through the public bond market. Following the introduction of CB purchases, eligible firms will likely have increased their bond issuance when they need to raise external capital. They may also have substituted bank debt with bond debt, although the magnitude of such a substitution effect may have been small because of the benefits provided by long-term relationships with banks such as the mitigation of information asymmetry problems and the loosening of loan terms (e.g., fewer requirements for collateral and covenants). In sum, the leverage ratios of eligible firms likely increased following the intervention. Therefore, the second hypothesis is constructed:

Hypothesis 2

(impact of the BOJ’s CB purchases): Following the introduction of CB purchases, the risk premium on eligible bonds was reduced, making it easier for issuing firms of such bonds to access the public bond market. Firms whose CBs were eligible for BOJ purchases strengthened bond issuance activities. There may exist a substitution effect between bond debt and bank debt; however, the magnitude of such effect may have been small. As a result, eligible firms increased their leverage ratio relative to firms whose CBs were not eligible for BOJ purchases.

Finally, firms can be divided into four groups depending on whether both their stocks and bonds (Both), only their stocks (Stocks only), only their bonds (Bonds only), or neither (Neither) were eligible for the BOJ’s ETF and CB purchases. Given Hypothesis 1 and 2, it is possible that after the policy intervention, the leverage ratio of firms in the Bonds only group may have increased, while firms in the Stocks only group may have seen a decrease in their leverage. Firms in the Both group (i.e., their stocks were included in ETF purchases and their bonds were eligible for CB purchases) may have strengthened their bond and stock issuance activities; however, their leverage ratio may have remained more or less unchanged because the positive and negative impacts may have cancelled each other out. Therefore, the following hypothesis is also examined:

Hypothesis 3

(combined impact of ETF and CB purchases): Some firms may have been eligible for both ETF and CB purchases, while others may have been eligible for one of the two or none, and the impact of the program may have differed depending on which group a firm fell in.

4.2 Methodology

In the main analysis, to investigate whether the above hypotheses are correct, I start by examining the impact of the BOJ’s ETF and CB purchases on risk premium using the DID approach. Next, the impact of the policy interventions on firms’ issuance of stocks and bonds is examined using a logit model. After that, the DID model is applied again to estimate the impacts of the purchasing programs on firms’ capital structure. In addition to studying the effect of each policy on the variables of interest separately, I also examine the two policies together to gauge their simultaneous impact. The baseline model specifications for the analyses on risk premium, firms’ securities issuance and capital structure are explained in turn as follows.

4.2.1 BOJ purchases and risk premiums

To examine the first part of Hypothesis 1, the effect of the BOJ’s ETF purchases on equity risk premium as well as the systematic risk are estimated by regressing the following model:

$${Y}_{it}=\alpha +{\updelta }_{ETF}\left({N225}_{it}\times {ETFPost}_{t}\right)+{X}_{it}^{{^{\prime}}}\beta +{u}_{i}+{v}_{t}+{\varepsilon }_{it}$$

(1)

where the dependent variable (${Y}_{it}$) is either adjusted beta value (the proxy for systematic risk¹¹) or equity risk premium of firm $i$ in year $t$; ${N225}_{it}$ is a dummy variable that equals one if firm $i$ is a Nikkei 225 component firm in year $t$ and zero otherwise; $ETF{Post}_{t}$ is a dummy variable that equals one for the ETF purchases treatment period, i.e., the period after the introduction of QQE on April 5, 2013. ${X}_{it}$ is a vector of firm-specific control variables (tangibility of assets, firm size, profitability, market to book (MB) ratio, firm age);${u}_{i}$ and ${v}_{t}$ are firm and time fixed effects, and ${\varepsilon }_{it}$ is the error term. In model (1), ${\updelta }_{ETF}$ is the estimator of the average treatment effect on the treated (ATT).

Furthermore, as discussed in Sect. 3.2, from October 2010 until before the announcement of the QQE policy, the BOJ put greater emphasis on purchases of CBs than ETFs. By contrast, since April 2013, the ETF purchases program has been significantly expanded and the BOJ’s holdings of ETFs have come to significantly outstrip holdings of CBs and other risk assets. Therefore, in the baseline analysis, I choose the period after the introduction of QQE in 2013 as the treatment period for ETF purchases, and the period after the introduction of CME in 2010 as the treatment period for CB purchases. In addition, because the BOJ focused largely on purchases of Nikkei 225 ETFs and spent more than half of its “budget” on ETFs that track this index until 2016, in the main analysis, Nikkei 225 component firms are the ETF-treatment group and non-Nikkei 225 firms listed on TSE1 and TSE2 are the ETF-control group.¹²

Next, given Hypothesis 2, the impact of the BOJ’s CB purchases on bond risk premium is examined. As explained in Sect. 3.2.2, the BOJ only publishes general information on what CBs are eligible for purchases, namely, CBs with a remaining maturity of 1–3 years and a rating of BBB or higher. These criteria are used to identify bonds eligible for CB purchases. Using the treatment group thus defined, the following DID model is employed:

$${Y}_{ikt}=\alpha +{\updelta }_{CB}\left({BondEligible}_{ikt}\times {CBPost}_{t}\right)+{X}_{ikt}^{{^{\prime}}}\beta +{u}_{i}+{v}_{t}+{\varepsilon }_{it}$$

(2)

where ${BondEligible}_{ikt}$ is a dummy variable that equals one if bond $k$ of firm $i$ satisfy the eligibility criteria for BOJ purchases in year $t$ (i.e., having a remaining maturity of 1 year to 3 years and a rating of BBB or higher) and zero otherwise; ${CBPost}_{t}$ is a dummy variable that equals one for the intervention period (i.e., the period after the introduction of CME on October 28, 2010) and zero otherwise; ${X}_{ikt}$ is a vector of control variables consisting of bond characteristics (remaining maturity, coupon rate) and firm-specific factors (tangibility of assets, firm size, profitability, firm age); ${u}_{i}$ and ${v}_{t}$ are firm and time fixed effects, and ${\varepsilon }_{it}$ is the error term. For the dependent variable (${Y}_{ikt}$), a proxy for bond risk premium—the credit spread—is used. Therefore, in model (2), coefficient ${\updelta }_{CB}$ measures the intention-to-treat (ITT) effect.

4.2.2 BOJ purchases and securities issuance

The next issue of interest is the impact of the BOJ’s ETF and CB purchasing programs on firms’ stock and bond issuance activities. To examine this impact, I first collect the historical data of stock and bond issuance to construct stock and bond issuance dummies, which are used as the dependent variables (${Y}_{it}$), then estimate the following logit model (model (3)):

$$P\left({Y}_{it}=1|{{\varvec{X}}}_{it}\right)=\frac{\text{exp}({{\varvec{X}}}_{it}{\varvec{\beta}}+{u}_{i}+{v}_{t})}{1+\text{exp}({{\varvec{X}}}_{it}{\varvec{\beta}}+{u}_{i}+{v}_{t})}$$

(3)

where $P\left({Y}_{it}=1|{{\varvec{X}}}_{it}\right)$ is the response probability (the probability of success), which indicates the probability that a firm issued stocks or bonds during a given year. ${{\varvec{X}}}_{it}$(${X}_{1it}$,…,${X}_{Kit}$) is a vector of independent and control variables and ${u}_{i}$ and ${v}_{t}$ are firm and time fixed effects.

Using this model specification, to examine the impact of ETF purchases on the probability that a firm issued stocks or bonds, the stock and bond issuance dummies (${Stock\_issuance}_{it}$, ${Bond\_issuance}_{it}$) are regressed on the interaction term between the Nikkei 225 dummy and the ETF purchases treatment period dummy (${N225}_{it}\times {ETFPost}_{t}$) (as in model (1)) and firm-specific factors (i.e., firms’ interest tax shield, cash ratio, tangibility, size, profitability, current ratio, and age).

Next, to investigate whether there is a causal relationship between the probability that a firm issued stocks or bonds and its eligibility for CB purchases, stock and bond issuance dummies are regressed on firm-specific control variables and the interaction term ${CBEligible}_{it}\times {CBPost}_{t}$, where ${CBEligible}_{it}$¹³ is a dummy variable that equals one for firms whose CBs satisfy the eligibility criteria for BOJ purchases in year $t$ and zero otherwise, and ${CBPost}_{t}$ is the CB purchases treatment period dummy.

Moreover, to gauge the interaction between the BOJ’s CB and ETF purchases, firms are categorized into four groups based on whether their stocks and bonds were subjected to the ETF and/or CB purchases as follows:

$${Both}_{it}=1\Leftrightarrow{N225}_{it}\times {ETFPost}_{t}=1\; \& \;{CBEligible}_{it}\times {CBPost}_{t}=1;$$

$${Bonds\_only}_{it}=1\Leftrightarrow {N225}_{it}\times {ETFPost}_{t}=0\; \& \;{CBEligible}_{it}\times {CBPost}_{t}=1;$$

$${Stocks\_only}_{it}=1\Leftrightarrow {N225}_{it}\times {ETFPost}_{t}=1\; \& \;{CBEligible}_{it}\times {CBPost}_{t}=0;$$

$${Neither}_{it}=1\Leftrightarrow {N225}_{it}\times {ETFPost}_{t}=0\; \& \;{CBEligible}_{it}\times {CBPost}_{t}=0;$$

After that, stock and bond issuance dummies can be regressed on three group dummies ($Both, Bonds\_only, Stocks\_only)$. In this regression, firm-specific factors are also controlled for.

4.2.3 BOJ purchases and firms’ capital structure

After exploring the impact on securities issuance, the DID framework is again used to address the impacts on firms’ capital structure. To estimate the effect of the BOJ’s ETF purchases, the following model is regressed:

$${Z}_{it}=\alpha +{\updelta }_{ETF}\left({N225}_{it}\times {ETFPost}_{t}\right)+{X}_{it}^{{^{\prime}}}\beta +{u}_{i}+{v}_{t}+{\varepsilon }_{it}$$

(4)

where ${N225}_{it}\times {ETFPost}_{t}$ is the interaction term between the Nikkei 225 dummy and the ETF purchases treatment period dummy as in model (1); ${X}_{it}$ is a vector of control variables¹⁴ consisting of firm-specific factors (size of interest tax shield, cash ratio, tangibility of assets, firm size, profitability, current ratio, market to book (MB) ratio, firm age) and macroeconomic control variables (gross domestic product (GDP) growth, consumer price index (CPI)); ${u}_{i}$ and ${v}_{t}$ represent firm and time fixed effects, and ${\varepsilon }_{it}$ is the error term. For the dependent variable (${Z}_{it}$), three alternative proxies for firms’ leverage (total leverage, short-term leverage, long-term leverage) and three variables related to firms’ debt structure (their bond ratio, short-term bank loan ratio, and long-term bank loan ratios) are used to examine the effect of the BOJ’s ETF purchases on firms’ capital structure and debt structure. In this model, ${\updelta }_{ETF}$ is the estimator of the ATT.

Similarly, the following DID model is employed to examine the impact of the BOJ’s CB purchases on firms’ capital structure:

$${Z}_{it}=\alpha +{\updelta }_{CB}\left({CBEligible}_{it}\times {CBPost}_{t}\right)+{X}_{it}^{{^{\prime}}}\beta +{u}_{i}+{v}_{t}+{\varepsilon }_{it}$$

(5)

where the dependent variable, ${Z}_{it}$, is a proxy for leverage or debt structure as in model (4); ${CBEligible}_{it}\times {CBPost}_{t}$ is the interaction term between the CB-eligibility dummy and the CB purchases treatment period dummy; ${X}_{it}$ is a vector of the same control variables as in model (4); ${u}_{i}$ and ${v}_{t}$ represent firm and time fixed effects, and ${\varepsilon }_{it}$ is the error term. In this model, coefficient ${\updelta }_{CB}$ measures the ITT effect.

Last but not least, the combined effect of ETF and CB purchases on firms’ capital structure is examined. The model is specified as follows:

$${Z}_{it}={\gamma }_{0}+{\gamma }_{1}{Both}_{it}{+{\gamma }_{2}Bonds\_only}_{it}{+{\gamma }_{3}Stocks\_only}_{it}+{X}_{it}^{{^{\prime}}}\beta +{u}_{i}+{v}_{t}+{\varepsilon }_{it}$$

(6)

where ${Z}_{it}$ is a proxy for leverage or debt structure as in model (4) and (5); $Both, Bonds\_only,$ and $Stocks\_only$ are firm group dummies based on their eligibility for ETF and CB as defined in the preceding subsection; ${X}_{it}$ is a vector of control variables, ${u}_{i}$ and ${v}_{t}$ are firm and time fixed effects, and ${\varepsilon }_{it}$ is the error term. Model (6) makes it possible to compare the leverage of firms in the above three groups (Both, Bonds only, Stocks only) and firms whose bonds and stocks are both ineligible for purchases by the BOJ (in the Neither group).

4.3 Data and variables

To examine the impact of the BOJ’s ETF and CB purchases, this study focuses on non-financial firms listed on the TSE1 and TSE2 during the period from 2009 to 2018. All dependent variables used throughout the study can be divided into three main categories: (i) proxies for systematic risk and risk premiums, (ii) proxies for stock and bond issuance, and (iii) proxies for leverage and debt structure. In the baseline analysis, the key independent variables are related to the BOJ’s purchasing program and consist of the interaction terms N225*ETFPost, BondEligible*CBPost and CBEligible*CBPost, as well as three firm group dummies based on their eligibility for ETF and CB purchases by the BOJ (Both, Bonds only, Stocks only). I also employ a set of control variables, including bond and firm-specific characteristics, and macroeconomic conditions. Table 3 provides definitions of all the variables used in this study,¹⁵ while Appendix Table 16 explains the predicted signs for the impact of the control variables on firms’ leverage.

Table 3 Definitions of all variables

Turning to the data sources, several datasets are utilized. First, to calculate the equity risk premium for estimation, the annual beta for individual stocks and the market risk premium are collected from Bloomberg. Using a two-year estimation period with weekly returns and the TOPIX index as the relative market index, Bloomberg estimates beta of a stock listed on the TSE1 or TSE2 by regressing the historical returns of the stock on the constant term and the historical returns of the TOPIX index. Meanwhile, the proxy for bond risk premium (i.e., monthly credit spreads of individual bonds) is computed based on the compounded yield of each bond from the OTC Bond Transactions database published by the JSDA, and the JGBs interest rate (with the same maturity) from the Ministry of Finance website. The eligibility dummy of each bond and bond characteristics control variables are also obtained based on the JSDA’s bond transactions data. Macroeconomic data are from the databases of the World Bank and the International Monetary Fund.

Next, firm data—the primary data of this study—is obtained from various databases of Nikkei FQ. Specifically, annual financial statement data used for calculating proxies for leverage and debt structure, as well as firm characteristics are collected from the Corporate Financial Statements database. Corporate Attribute Data provides necessary corporate information such as company name, establishment date (used for calculating firm age), industry group, and listing information. Stock prices data are from Stock Price and Indicators database. To generate variables related to securities issuance activities, Corporate Action Data of Nikkei FQ, which covers historical data of firms’ new bond and stock issuances (including the date of issuance and the amount raised) are also used.

To restrict the sample to non-financial firms, firms belonging to the following four industry categories: (i) banks, (ii) securities and commodity futures, (iii) insurance, and (iv) other financing business are excluded from the sample using the TSE industry group code. In the sample, listing information of firms is compiled based on data for 2018. To construct the treatment and control groups precisely, the analyses do consider the entry and exit of firms from the Nikkei 225 index by identifying all component changes in the Nikkei 225 index during the estimation period. Data of constituent changes in the Nikkei 225 index are from the website of Nikkei Indexes.¹⁶ Further, to remove outliers, firms that have a leverage ratio of less than zero or higher than one are excluded from the sample. Finally, the baseline sample includes 2517 firms, of which 2008 are listed on TSE1 and 509 are listed on TSE2. Of the 2008 TSE1 firms, 196 firms are components of the Nikkei 225 index. In the CB purchases intervention period, on average, 280 firms were eligible for the CB purchasing program, accounting for 11 percent of the total number of firms.

5 Empirical results

This section presents the results of the empirical analysis based on the approach described in the preceding section. It starts with the presentation of summary statistics, then discusses the results of regression analyses, and finally provides additional analyses to check the robustness of the baseline regression results.

5.1 Summary statistics

Table 4 presents summary statistics of the key variables used in the study. The figures are the averages for the observation period overall from 2009–2018.

Table 4 Summary statistics

Starting with the firm variables, the average adjusted beta is 0.883. Using the Capital Asset Pricing Model (CAPM), the equity risk premium is calculated based on the adjusted beta and the market risk premium, and its average value is 10.6 percent. The mean of the stock issuance dummy is slightly higher than that of the bond issuance dummy, suggesting that firms are more likely to issue bonds than stocks in any given year. Looking at the proxies for leverage and debt structure, the mean of firms’ leverage ratio is 48.5 percent. On average, bond debt accounts for only 1.8 percent of total assets, which is much smaller than the average percentage of short-term and long-term bank loans with 7.5 percent and 10.7 percent. Further, the summary statistics of firm-specific variables indicate that the firms included in the sample are quite heterogeneous. For example, looking at the minimum and maximum values, firms differ substantially in terms of their size, cash ratio, profitability, etc.

Turning to the bond variables, including the credit spreads, the BondEligible dummy, the bond maturity, and coupon rate, these summary statistics are obtained using the bond sample. The average credit spread is 0.44 percent, and bonds that satisfy the eligibility criteria for CB purchases account for 25.5 percent of the total number of bonds. An average bond has a remaining maturity of 4.8 years and a coupon rate of 1.34 percent.

5.2 Regression analysis

5.2.1 Impact of BOJ purchases on risk premiums

First, the treatment effects of the BOJ’s ETF purchases on CAPM beta and equity risk premium are estimated. The adjusted beta is employed, considering that it is calculated based on the assumption that the beta tends to revert toward the market average of 1.0 over time, and thus better reflects the future beta than the historical beta (raw beta). Using the CAPM approach, the equity risk premium for each stock is then calculated by multiplying the adjusted beta by the market risk premium. Employing model specification (1), the analysis yields the results reported in Table 5.¹⁷

Table 5 Impact of BOJ ETF purchases on beta value and equity risk premium (ERP)

In Table 5, the estimation results using the full sample (data for all non-financial firms on the TSE1 and TSE2) are reported in the first two columns. Following the expansion of the ETF purchases in 2013, Nikkei 225 firms have a 5.19 percentage points lower adjusted beta and a 1.31 percentage point lower equity risk premium than non-Nikkei 225 firms. In addition, because the relative market index used to estimate the beta is the TOPIX index, to address the potential bias, the analysis in the columns (3) and (4) employ data only for firms that are constituents of the TOPIX index (i.e., firms listed on the TSE1). The results are similar to those obtained using the full sample and consistent with Hypothesis 1, that the BOJ’s ETF purchases have reduced the market beta and equity risk premium, which will lead to a decrease in the cost of equity.

Next, the effect of the BOJ’s CB purchases on bond risk premium is examined using model (2) and bond transactions data collected from JSDA. The results are presented in Table 6. In column (1), the credit spread is regressed on the interaction term BondEligible*CBPost (the key independent variable) and bond characteristics (the natural log of remaining maturity and the coupon rate). In column (2), firm-specific control variables are included in the model by combining the bond data set from JSDA with the firm data set obtained from Nikkei FQ. Time and firm fixed effects are included in these estimations. As can be seen, following the introduction of CB purchases by the BOJ, the credit spreads of bonds eligible for CB purchases are significantly lower than that of ineligible bonds. This evidence is in line with Hypothesis 2 that the CB purchases have reduced the bond risk premium and thus making it cheaper for issuing firms of such bonds to raise capital through the bond market.

Table 6 Impact of BOJ CB purchases on credit spreads

5.2.2 Impact of BOJ purchases on firms’ securities issuance

In the next step, the impact of the BOJ’s asset purchases on the probability that firms issue new stocks and bonds is examined. The first two columns in Panels A–C of Table 7 present the average marginal effects obtained using the logit model with two-way fixed effects (model (3)). To address the potential bias in fixed effect estimations of the logit model due to the incidental parameter problem (Neyman & Scott, 1948), this analysis implements the analytical bias correction method proposed by Cruz-Gonzalez et al. (2017).

Table 7 Impact of BOJ asset purchases on securities issuance

First, the impacts of the ETF purchases on firms’ securities issuance are examined by regressing the stock or bond issuance dummies on the N225*ETFPost, controlling for firm characteristics, and firm and time fixed effects. The results, as shown in column (1) of Table 7A, suggest that the probability of stock issuance is 1.83 percentage points higher for Nikkei 225 firms. This implies that because ETF purchases have made it cheaper for Nikkei 225 firms to issue equity, they may have taken advantage of this opportunity to increase their stock issuance, as predicted by the equity market timing and Hypothesis 1. On the other hand, the probability that Nikkei 225 firms issued bonds decreased by 1.22 percentage points after the introduction of the ETF purchase program (column (2)), indicating that Nikkei 225 firms substitute bond financing with equity financing, which is also consistent with Hypothesis 1.

Next, a similar analysis was conducted to examine the impact of the BOJ’s CB purchasing program on the probability that firms issued bonds or stocks. Columns (1) and (2) of Table 7B show the estimation results of the logit regression model. The marginal effect of the CBEligible*CBPost indicates that the probability of bond issuance is 6.12 percentage points higher for firms whose bonds are eligible for CB purchases by the BOJ, which supports Hypothesis 2. Meanwhile, the probability of stock issuance is not significantly different between eligible and ineligible firms.

Subsequently, the stock or bond issuance dummies are regressed on three firm group dummies and control variables to gauge the simultaneous impact of the ETF and CB purchases. Columns (1) and (2) in Table 7C show that belonging to Both group increased the probability that firms issued bonds and stocks by 3.70 and 1.68 percentage points compared to firms in Neither group, but the size of the effect is much smaller than that on firms included in solely ETF or CB purchases. This evidence suggests that the magnitude of the effects of ETF (or CB) purchases on firms’ securities issuance depends on whether the firms were subjected to the CB (or ETF) purchases.

To supplement the findings obtained from the logit model, columns (3) and (4) in Panels A–C of Table 7 report the results from a linear probability model. Stock and bond issuance dummies are again utilized as the dependent variables, while controlling for firm characteristics, and firm and time fixed effects. Across all three analyses, the signs of the coefficients in the linear probability model align with the results obtained from the logit estimations. Overall, the evidence suggests that the ETF and CB purchases by the BOJ have changed firms’ stock and bond issuance behavior in the primary market,¹⁸ and thus may have a substantial impact on corporate capital structure.

5.2.3 Impact of the BOJ’s ETF purchases on firms’ capital structure

First of all, it is necessary to check whether the parallel trend assumption—a key prerequisite for the DID method—holds. This assumption requires that if there were no treatment or policy changes, the changes in outcomes over time would be the same in both the treatment and the control group. To this end, Fig. 2 depicts the trend in the average leverage of Nikkei 225 and non-Nikkei 225 firms.

Fig. 2

Testing for parallel trend—BOJ ETF purchases. This figure plots the trend in the average leverage ratio of Nikkei 225 firms (solid blue line) and non-Nikkei 225 firms (solid red line) from 2009 to 2018. The leverage ratio is calculated as the total liabilities scaled by total assets. The vertical line denotes the separation between the pre-intervention period and the intervention period for ETF purchases

As can be seen, the average leverage ratios of the treatment group and the control group follow a common trend before the expansion of the BOJ’s ETF purchase policy in 2013, suggesting that the parallel trend assumption is not violated.¹⁹ In the ETF purchases treatment period, although the average leverage of the treatment and control groups both decreased, the mean of Nikkei 225 firms’ leverage declined more sharply by 4.5 percentage points compared to that of non-Nikkei 225 firms by only 2.7 percentage points. This indicates that the BOJ’s ETF purchases may have reduced the leverage of firms whose stocks were included in ETF purchases, and the following regression results yield a similar finding.

Table 8 reports the estimation results when the fixed effects regression is conducted using model (4) specified in Sect. 4.2. The treatment effect of the BOJ’s ETF purchases on firms’ capital structure is the coefficient on N225*ETFPost. The results indicate that the BOJ’s ETF purchasing program reduced the leverage, long-term leverage, bond, and long-term bank loan ratios of the treatment group (Nikkei 225 firms) relative to the control group (non-Nikkei 225 firms), which support Hypothesis 1. Notably, as shown in columns (1) and (3), Nikkei 225 firms have a 0.85 percentage point lower leverage ratio and a 1.12 percentage point lower long-term leverage ratio than non-Nikkei 225 firms. This decline in leverage ratios is in line with the evidence on the effects of ETF purchases on equity risk premium and securities issuance activities: experiencing a decline in equity risk premium (which implies an increase in stock valuations), Nikkei 225 firms relied more on equity financing to raise outside capital, presumably to take advantage of more favorable equity market conditions, while decreasing the amount of bank loans and bond debt. In addition, most of the coefficients on the control variables are significant and the signs are as predicted, including the interest tax shield, cash ratio, tangibility, size, profitability, and current ratio.

Table 8 Impact of BOJ ETF purchases on firms’ capital structure

5.2.4 Impact of the BOJ’s CB purchases on firms’ capital structure

Next, the impact of the BOJ’s CB purchases on firms’ capital structure is examined. To check whether the parallel trend assumption holds, Fig. 3 plots the trend in the average bond ratio of eligible firms and non-eligible firms (i.e., firms whose bonds are not eligible for the CB purchases).

Fig. 3

Testing for parallel trend—BOJ CB purchases. This figure plots the trend in the average bond ratio of CB-eligible firms (solid blue line) and non-eligible firms (solid red line) from 2009 to 2018. The bond ratio is calculated as the bonds and convertibles scaled by total assets. The vertical line denotes the separation between the pre-intervention period and the intervention period for CB purchases.

Figure 3 shows that the average bond ratios of the treatment and control groups follow a common trend during the pre-intervention period, implying that the parallel trend assumption is not violated. Looking at the change over the intervention period, it can be seen that the mean bond ratio of the control group decreased slightly, but is generally stable over time. In contrast, the mean bond ratio of the treatment group was much higher than before the policy intervention.

Turning to the estimation results, by employing model (5), estimates of the ITT effect are obtained and reported in Table 9.

Table 9 Impact of BOJ CB purchases on firms’ capital structure

As shown in Table 9, following the introduction of CB purchases, firms whose CBs are eligible for BOJ purchases have a 0.95 percentage point higher overall leverage, a 1.12 percentage point higher long-term leverage, and a 1.43 percentage point higher bond ratio than other firms, however, there is no significant impact on short-term leverage. Furthermore, the short-term and long-term bank loan ratios of eligible firms are 0.35 percentage points and 0.14 percentage points lower than those of ineligible firms after controlling for firm characteristics and macroeconomic factors, as well as including firm and time fixed effects in all estimations. This evidence is in line with Hypothesis 2 that after the introduction of CB purchases under the CME policy in 2010, relative to ineligible firms, eligible firms used more bond debt and replaced part of their bank loans with bond debt. The decrease in bank debt is smaller than the increase in bond debt, resulting in a higher level of leverage overall.

Regarding the impact of firm-specific and macroeconomic control variables on firms’ leverage, as shown in Tables 8 and 9, the results of the baseline analysis are generally consistent with the predictions shown in Appendix Table 16. Specifically, the regression results suggest that firms’ leverage ratio is positively correlated with the interest tax shield and size, and negatively correlated with firms’ cash ratio, tangibility, profitability, current ratio, and age as well as GDP growth. However, contrary to expectation, the coefficients on the CPI are negative, while the impacts of the current MB ratio (a measure for investment opportunities) on proxies for leverage are generally weak.

5.2.5 Combined effect of ETF and CB purchases on firms’ capital structure

Last but not least, the combined effect on firms’ capital structure is examined using model (6). The results are presented in Table 10 and confirm Hypothesis 3 that the effects of the BOJ’s ETF and CB purchases on corporate capital structure are not uniform across firms that differ in terms of the eligibility of their stocks and bonds for BOJ purchases.

Table 10 Combined effect on firms’ capital structure

Specifically, Table 10 implies that firms belonging to the Bonds only group (i.e., firms that have solely better access to the public bond market following the introduction of the program) have a higher overall leverage ratio, long-term leverage ratio and bond ratio, while having a lower short and long-term bank loan ratios than firms in the Neither group. Meanwhile, firms belonging to the Stocks only group (i.e., firms that have solely better access to the public stock market following the introduction of the program) have a 1.13 percentage point lower long-term leverage than firms in the Neither group. Inclusion only in the BOJ’s ETF purchases also has an impact on firms’ leverage overall (column (1)) and their long-term bank loan ratio, but the coefficients are significant only at the 10 percent level. Turning to firms belonging to the Both group, the leverage and debt ratios of these firms are not significantly different from those in the Neither group. This result is not very surprising, since firms in the Both group have easier access to both the public bond market and the public stock market, so that the impacts of CB purchases and ETF purchases on firms’ leverage ratios, which work in opposite directions, seem to have cancelled each other out.

5.3 Additional analyses and robustness checks

In this subsection, I perform various supplemental analyses to check the robustness of the baseline findings. I start by applying multiple treatment groups or a continuous treatment variable in the DID model to estimate the effect of the ETF purchases more precisely. Next, I explore the impacts of the BOJ’s ETF and CB purchasing programs on securities issuance amounts and stock repurchases. Furthermore, I discuss the potential existence of the spillover effects and finally utilize an adjusted dataset to mitigate the potential noises associated with stock exchange transfers.

5.3.1 Multiple treatment groups for ETF purchases

In the baseline analysis, Nikkei 225 firms were used as the ETF-treatment group, while non-Nikkei 225 TSE1 firms and TSE2 firms were used as the control group. However, it could be argued that the inclusion of non-Nikkei 225 TSE1 firms in the control group for ETF purchases may bias the estimation of treatment effects, since part of the BOJ’s ETF purchases is of those tracking the TOPIX (as mentioned in Sect. 3.2.1). To address this concern, the DID analysis in Table 11 employs both TOPIX and N225 component firms as the treatment groups and firms listed on the TSE2 as the control group. The set of control variables consists of firm-specific factors and macroeconomic variables as applied in the baseline models.

Table 11 Impact of BOJ ETF purchases: Multiple treatment groups

The table shows the treatment effects of the BOJ’s ETF purchases on the leverage and debt indicators for both N225 and TOPIX component firm. TOPIX is a dummy variable that equals one for TOPIX component firms and zero otherwise. The coefficients on N225*ETFPost and TOPIX*ETFPost in columns (1) and (3) are both negative and statistically significant, indicating that the BOJ’s ETF purchases have led to a reduction in the leverage and long-term leverage ratios of firms whose stocks were included in ETF purchases. Specifically, the estimated treatment effect on leverage is -0.7 percentage point for TOPIX firms, and the total effect is -1.4 percentage points for Nikkei 225 firms (the total effect is the sum of two coefficients since N225 firms are also constituents of the TOPIX index). This evidence is in line with the baseline results, and further supports the findings on the impacts of the BOJ’s ETF purchases.

5.3.2 Continuous treatment variables and the ETF purchase program

One important feature of the main analysis is that all the treatment variables used were dummy variables. Moreover, as explained in the preceding subsection, using non-Nikkei 225 TSE1 and TSE2 firms as the control group for ETF purchases may lead to potential bias. To address this issue and further examine the differential impacts on firms included in the ETF purchases, I take changes in the BOJ’s total purchasing amount and the amount allocated to each type of ETF into account by employing a continuous treatment variable: the BOJ’s indirect shareholding ratio, which was previously introduced in Sect. 3.2.

To calculate the BOJ’s indirect shareholding ratio, I follow the method proposed by Gunji et al. (2021).²⁰ Additional data are obtained from the websites of BOJ and Nikkei Indexes. In particular, the authors calculate the BOJ’s indirect shareholding ratio of TOPIX component firms as follows:

$${\rho }_{t}=\frac{{X}_{t}^{TOPIX}}{{\sum }_{i=1}^{N}{p}_{it}{q}_{it}}$$

where ${\rho }_{t}$ is the BOJ’s shareholding ratio of TOPIX component firms in year $t$; ${X}_{t}^{TOPIX}$ is the total outstanding amount of TOPIX ETFs purchased by the BOJ in year $t$; ${p}_{it}$ and ${q}_{it}$ are the stock price and the number of outstanding shares of firm $i$ in year $t$. The authors assume for simplicity that the free float ratio is constant for all stocks and all periods, resulting in the same indirect shareholding ratio (${\rho }_{t}$) for all TOPIX component firms in each year $t$. Note that ${\rho }_{t}$ is equal to zero for TSE2 firms.

Moreover, the BOJ’s indirect shareholding ratio of Nikkei 225 firms is:

$${\theta }_{it}=\frac{50{X}_{t}^{N225}}{{\overline{p} }_{it}{q}_{it}{d}_{t}{N225}_{t}}$$

where ${\theta }_{it}$ is the BOJ’s shareholding ratio of Nikkei 225 component firm $i$ in year $t$; ${X}_{t}^{N225}$ is the total outstanding amount of Nikkei 225 ETFs purchased by the BOJ in year $t$; ${\overline{p} }_{it}$ is the presumed par value of firm $i$ in year $t$; ${q}_{it}$ is the number of outstanding shares of firm $i$ in year $t$; ${d}_{t}$ is the Nikkei 225 divisor in year $t$; and ${N225}_{t}$ is the Nikkei 225 index in year $t$.${\theta }_{it}$ equals zero for non-Nikkei 225 firms.

In the analysis, the BOJ’s indirect shareholding ratio will be referred to as the “exposure variable” for short. The total exposure variable,²¹ which measures the extent to which each firm is “exposed” to ETF purchases by the BOJ, is the sum of ${\rho }_{t}$ and ${\theta }_{it}$, since the stocks of Nikkei 225 firms are subject to both TOPIX and Nikkei 225 ETF purchases by the BOJ. Using this variable as a continuous treatment variable, I repeat the analysis based on model (1) to verify the impact of the ETF purchases on CAPM beta and equity risk premium. As presented in Table 12, following the introduction of ETF purchases in 2010,²² firms that are highly exposed to the ETF purchases have lower market beta and equity risk premium than less exposed firms. The results are robust to using different firm samples (i.e., the sample of all firms listed on the TSE1 and TSE2, or the sample of TSE1 firms only). The results obtained from Table 12 are in line with that from Table 5, reinforcing that the BOJ’s ETF purchases have reduced the cost of equity for firms whose stocks are included in the purchases.

Table 12 Impact of ETF purchases on beta value and ERP: robustness checks

Next, using the continuous treatment variable, the impact on capital structure is considered. Similar to model (4), leverage and debt indicators are regressed on the total exposure variable and control variables. The results are shown in Table 13A. In addition, I regress the leverage and debt ratios on the Nikkei 225 exposure variable (${\theta }_{it}$) and the TOPIX exposure variable (${\rho }_{t}$) to examine whether Nikkei 225 ETF purchases or TOPIX ETF purchases played a larger role in capital structure changes. The results are shown in parts B and C of Table 13, respectively.

Table 13 Continuous treatment variables and BOJ ETF purchases

The three parts of Table 13 show that the estimated treatment effect on leverage is -1.1 percentage points in the case of total exposure, -0.4 percentage points in the case of Nikkei 225 exposure, and -2.1 percentage points in the case of TOPIX exposure. These estimates show that following the introduction of the BOJ’s ETF purchases in 2010, firms with greater exposure to the purchases, as measured by three proxies for the BOJ’s indirect shareholding ratio, have lower leverage and debt ratios than firms with less exposure. This implies that the impact of ETF purchases on firms in the treatment group is heterogeneous and likely depends on the weight of each firm in the Nikkei 225 or TOPIX indexes as well as changes in the BOJ’s purchasing rules, i.e., the purchasing amount and allocation of purchases.²³ Moreover, the results in parts B and C of Table 13 indicate that the TOPIX ETF purchases have a larger effect on highly exposed firms than the Nikkei 225 ETF purchases, suggesting that the BOJ’s adjustment of its purchases from Nikkei 225 ETFs to TOPIX ETFs is an appropriate action.

5.3.3 Policy impact on securities issuance amounts

In the baseline estimations presented in Sect. 5.2.2, I used a logit model to examine the effect of the BOJ’s asset purchases on firms’ decision to raise or not to raise external financing in the market. However, it is possible that some firms issue securities frequently but each time issue only a very limited amount of bonds or stocks. Therefore, in the following analysis, I further explore the effect of the BOJ’s ETF and CB purchasing program on securities issuance amounts by applying a tobit model. A tobit model is a limited dependent variable model, and such models are widely used when the dependent variable is censored from below at zero, as is the case here.

The average marginal effects from the tobit model estimation are presented in Table 14. Like in the analysis of the probability of securities issuance in Sect. 5.2.2, I separately regress proxies for the amounts of stock and bond issuance on the interaction term N225*ETFPost (Panel A), or the interaction term CBEligible*CBPost (Panel B), or firm group dummies (Panel C), controlling for firm-specific factors as well as firm and year fixed effects.

Table 14 Impact of BOJ asset purchases on securities issuance amounts: Tobit model

The marginal effects on stock issuance amounts of N225*ETFPost in Panel A as well as Both and Stocks only in Panel C of Table 14 suggest that, on average, the actual values of Log(1 + Stock issuance amounts) are 17.90 to 34.57 percentage points higher for firms whose stocks were included in ETF purchases by the BOJ. Meanwhile, the marginal effects on bond issuance amounts of CBEligible*CBPost in Panel B as well as Both and Bonds only in Panel C indicate that the Log(1 + Bond issuance amounts) are 26.74 to 57.70 percentage points higher for firms whose bonds were eligible for CB purchases than for firms whose bonds were not eligible. This result suggests that treatment firms, as a consequence of the decline in bond and/or equity issuance costs due to BOJ asset purchases, have increased their bond and/or stock issuance: not only did they issue securities more frequently (as implied by Table 7), but they also issued securities with considerably larger amounts than before.

Last but not least, evidence from Panel C of Tables 7 and 14 indicates that the combined effect of the BOJ’s ETF and CB purchases on securities issuance does exist. The positive effect of ETF (or CB) purchases on targeted firms’ probability and amount of stock (or bond) issuance is 1.3 to 2.2 times greater if these firms were not subjected to the other policy—CB (or ETF) purchases. Presumably, when there is a need to raise external capital, firms included in both ETF and CB purchases can flexibly choose to issue bonds or stocks. In contrast, firms included in solely ETF purchases have a lower cost of equity, but their cost of bond debt is relatively high compared to firms included in CB purchases. As a result, firms included in solely ETF purchases relied more on stock issuance and may have tended to replace bonds with stocks to take advantage of the market conditions. A similar argument could be made for firms included in solely CB purchases.

5.3.4 Policy impact on stock repurchases

In addition to stock issuances, another important type of corporate action that will lead to changes in corporate capital structure is stock repurchases. As implied by the market timing theory, firms tend to buy back equity when their shares are undervalued; however, because the BOJ’s purchases of risk assets provide support to asset prices and valuations, the stocks of Nikkei 225 component firms are less likely to be undervalued. Therefore, it is possible that following the expansion of BOJ ETF purchases, Nikkei 225 firms may have reduced their stock repurchase activities relative to non-Nikkei 225 firms.

On the other hand, firms may choose to buy back their stock for a variety of reasons, such as to optimize their capital structure, send signals to investors, distribute profits or excess cash to shareholders, improve their financial ratios, reduce stock dilution, or for other corporate governance purposes. Gunji et al. (2021) argued that because the BOJ purchases ETFs via trust banks, the exercise of voting rights is delegated to these banks. However, trust banks, as passive investors, are more likely to focus on the short term and may not exercise sufficient monitoring. Therefore, if Nikkei 225 firms are concerned about such a potential impact, or are uncertain about future policy changes, their stock repurchase activities may have increased relative to those of non-Nikkei 225 firms.

To investigate this issue, I first collect stock repurchases data from Corporate Action Data of Nikkei FQ. Next, I apply logit and tobit models to examine the effect of the BOJ’s ETF purchases on the probability that firms conduct stock repurchases, and the effect on the amount of stock repurchases if they are conducted. However, the results obtained (not shown here) are not statistically significant, probably because the impact of the BOJ’s ETF purchases on firms’ stock repurchases could go in either direction, and any effects may have cancelled each other out.

5.3.5 Accounting for general equilibrium effects or spillovers

As mentioned in Sect. 5.2.3, although the parallel trend assumption for the DID analysis has been met, a factor that could bias our results is the potential violation of the stable unit treatment value assumption (SUTVA). SUTVA requires that the treatment of one unit does not affect the outcomes of other units, meaning there should be no general equilibrium effects or spillovers generated by the treatment (as indicated in Rubin, 1990; Heckman et al., 1998; and Wooldridge, 2010). However, in the context of this study, SUTVA may be violated due to the spillovers of the BOJ’s purchasing program on firms not eligible for such purchases through (i) bank lending channel, (ii) changes in government bond yield, and (iii) the stock and corporate bond markets. In this subsection, I discuss each type of spillover and how it may affect the obtained estimates.

I start by examining the spillover effect through the bank lending channel. Specifically, the obtained results indicate that, following the policy intervention, eligible firms raised external capital from public markets more frequently and in larger amounts, while significantly reducing their bank debt compared to control firms. This suggests that the demand for bank loans of eligible firms has decreased. Consequently, banks having more lending capacity may have redirected lending to ineligible firms in the control group, which leads to an increase in these firms’ leverage ratio. If such spillover effects exist, the treatment effect will be biased. I apply the following strategy to investigate the spillover effects of the BOJ’s ETF and CB purchases on ineligible firms.²⁴

First, I identify banks that have experienced a substantial increase in lending capacity due to a significant decline in bank loan demands (referred to as “highly exposed banks”) and those that have experienced a small increase in lending capacity resulting from a small decline in bank loan demands (referred to as “less exposed banks”). I then compare the changes in bank loans of listed firms that have transaction relationships with each of these types of banks. If the spillover effect exists, there should be a greater increase in bank loans for firms transacting with highly exposed banks compared to those transacting with less exposed banks. In this scenario, the treatment effect of a lower leverage ratio among firms eligible for ETF purchases might be overestimated.

The above-mentioned strategy is implemented as follows. I begin by collecting additional data on loans from banks to publicly listed firms at the individual loan level from the Nikkei FQ to compute ${Bank ETF \left(CB\right) exposure}_{jt}$ variables. These variables measure the degree to which bank j has been exposed to the BOJ’s ETF (CB) purchases in year t once the BOJ started to make such purchases:

$${Bank\; ETF\; \left(CB\right) exposure}_{jt}=\frac{\sum {\text{Loan amount provided to ETF }\left(\text{CB}\right)\text{ treatment firms}}_{jt}}{\sum {\text{Loan amount provided to all firms on TSE}1\&2}_{jt}}\times {Post}_{t}$$

where ETF treatment firms are Nikkei 225 component firms; CB treatment firms are firms whose bonds are eligible for CB purchases; ${Post}_{t}$ is a dummy variable that equals one for the intervention period (i.e., the period after the introduction of CME on October 28, 2010) and zero otherwise.

Next, the spillovers on ineligible firms are estimated using the propensity score matching difference-in-difference (PSM-DID) approach.²⁵ I consider firms not eligible for the CB purchases to be firms whose bonds have been ineligible for CB purchases throughout all years, while firms not eligible for the ETF purchases are those listed on the TSE2.

To apply the PSM-DID method, firm observations are divided into two periods: the pre-intervention period and the intervention period. Covariates $X$ used for matching firms in the treatment and control groups are the average values of firm characteristics (firm tangibility, size, profitability, and age) in the pre-treatment period. The outcome of interest, $\Delta {Y}_{i}\equiv {Y}_{i,post}-{Y}_{i,pre}$, is the difference between the average values of $Y$ in the pre-intervention and the intervention period, where $Y$ represents the log of loan amount or loan ratio.²⁶ The binary treatment variable $D$ is the $Firm ETF\left(CB\right) exposure$, which equals one for firm $i$ if the average of $Bank ETF\left(CB\right) exposure$ of firm $i$’s top three main banks in the intervention period is above the mean.²⁷

Table 15 reports the estimates of the average treatment effects (ATE) of ETF and CB purchases on ineligible firms and the corresponding p-values.

Table 15 Spillover of BOJ purchases on ineligible public firms: PSM-DID

As can be seen from the table, the p-values are larger than 0.1, meaning that the ATE estimates for the spillover effects of the BOJ’s ETF and CB purchases on ineligible firms’ bank borrowing are statistically insignificant.²⁸ Plotting the propensity score density and the balance box (not shown for brevity) shows that the overlap assumption and the balancing condition are not violated. Because the spillover effect is not significant for ineligible publicly listed firms, we can infer that there is no violation of SUTVA due to the spillover effect through bank lending channel, at least in this analysis on listed firms.

In addition to the bank lending channel, the BOJ's asset purchases might have a spillover effect through changes in government bond yield. Fukui and Yagasaki (2023) pointed out that before the introduction of QQE with Yield Curve Control in 2016, the BOJ’s ETF purchases caused an increase in long-term interest rate, measured by the 10-year JGB yield. Although the JGB yield, being an indicator of the risk-free interest rate, has the potential to affect the cost of capital and corporate financing activities, any impact if it exists will affect all firms, irrespective of their eligibility for the BOJ’s ETF and CB purchases. Therefore, it is unlikely that this spillover would introduce bias to the estimates on the effect of the BOJ’s purchases.

Another possibility is that the BOJ’s ETF and CB purchases might generate spillover effects in the stock and bond markets, presumably by motivating investors searching for yield (such as banks and insurance companies) to buy ineligible stocks and/or bonds. As a result, ineligible public firms may have experienced an increase in stock and/or bond prices and a decline in their cost of capital, thereby strengthening their issuance of equity and/or bond debt. In this case, the BOJ’s purchasing program would affect the leverage ratio of treatment and control firms in the same direction, and the treatment effect would be underestimated. At the moment, the study is unable to verify the spillover effect occurring through this channel due to the lack of validation tools.²⁹ However, if this spillover effect exists, the treatment effect obtained will be a conservative estimate (to the extent that the true effect is underestimated).

5.3.6 Adjusting the sample

Finally, there is a concern that during the analysis period, on July 16, 2013, the TSE and the Osaka Securities Exchange (OSE) merged, and 1100 firms listed on the OSE were transferred to the TSE. Although only a small percentage of these firms were transferred to the TSE1 and TSE2 (37 firms to the TSE1, 162 firms to the TSE2, and the remaining were transferred to the Mothers and JASDAQ), this change in the components of the TSE may affect the construction of treatment and control groups as well as the estimates of treatment effects. Therefore, to mitigate the potential impact of noises associated with stock exchange transfers, an adjusted data set is constructed by excluding all firms involved in the merger of TSE and OSE from the original sample. Data on companies transferred from OSE to TSE following the merger are collected from the Corporate Attribute—Listed Market Information database of Nikkei FQ. Rerunning the baseline analysis on capital structure (model specifications (3)–(6)) using this new sample, the results remain essentially unchanged as shown in the Appendix (Tables 17, 18, 19, 20), confirming the robustness of the findings of the baseline analysis.

6 Conclusion

This study focused on the effects of the BOJ’s ETF and CB purchasing program on the capital structure of Japanese listed firms spanning the period from 2009 to 2018. To measure the treatment effect of the policy intervention on firms’ capital structure, the DID framework was employed. In addition, to understand the mechanism by which the ETF and CB purchases affect corporate capital structure, the policy impacts on equity and bond risk premiums, as well as on firms’ stock and bond issuance activities have also been investigated.

The results first suggest that the BOJ’s ETF purchases reduced the equity risk premium for firms included in ETF purchases, thereby lowering the cost of equity. These firms actively issued more stocks and became less dependent on bond debt and bank loans than control firms, resulting in a lower level of leverage. On the other hand, the BOJ’s CB purchases lowered the risk premium of bonds eligible for such purchases, which suggests a decline in the cost of bond debt. Firms whose bonds were eligible for CB purchases issued more bonds, while reducing bank debt but to a smaller extent, thus having a higher leverage ratio than ineligible firms. Moreover, categorizing firms into four groups based on whether their stocks were included in ETF purchases and/or their bonds were eligible for CB purchases shows that the policy impacts on firms’ leverage ratio and securities issuance activities have differed depending on which group a firm fell in.

Importantly, in the present study, the obtained results suggest that changes in market conditions brought about by the BOJ purchases may have affected firms’ capital structure and securities issuance, supporting the market timing theory proposed by Baker and Wurgler (2002). That is, firms try to time the market when issuing debt or equity: firms included in ETF and/or CB purchases appear to have taken advantage of the increase in the prices and valuations of their securities as well as a relative reduction in the cost of equity and/or cost of bond issuance compared to the cost of other forms of capital. While Baker and Wurgler (2002) provide evidence that the current capital structure of firms is strongly related to the historical market valuations of the stock, this study shows that the firms’ capital structure is affected by exogenous changes, namely, the BOJ’s asset purchases that boost the valuations of their securities.

Turning to the policy implications, through a joint investigation into two large-scale asset purchases implemented by the BOJ, this study has discovered evidence that the ETF and CB purchase programs can improve financing conditions of public firms whose stocks and/or bonds are included in the purchases. This is achieved because the BOJ’s asset purchases have reduced the cost of capital and prompted eligible firms to raise external funds from the public market. Experiencing a negative shift in demand for bank loans, although highly exposed banks do not use the idle funds to extend loans to ineligible publicly listed firms (as shown in Table 15), there is a possibility that these banks may have increased their credit allocation to financially constrained and inherently riskier SMEs. The finding that the BOJ’s asset purchases can facilitate corporate financing activities is meaningful, especially when the economy experiences an abrupt shock like COVID-19.

However, the BOJ also needs to consider the potential problem when implementing asset purchases on a large scale, as the interventions can cause securities prices to substantially deviate from their fundamental values, potentially exacerbating the asymmetric information problem between firms and investors. Moreover, firms included in the ETF and CB purchases are large public firms that already have an advantage in raising external funds vis-à-vis public firms that are not included. Therefore, the BOJ’s asset purchases may intensify inequality across firms if financially unconstrained can raise capital more easily, while relatively more constrained ineligible public firms cannot.

While this study provides evidence on the effects of BOJ asset purchases on firms’ capital structure, a range of issues warrant further investigation. First, the results indicate that the demand for bank loans of treatment firms has reduced. However, supplemental analysis in Sect. 5.3.5 reveals that banks that have experienced a substantial increase in lending capacity due to a decline in loan demands do not extend more loans to ineligible public firms. Several research questions remain unanswered: Do banks highly exposed to the BOJ’s purchases extend more loans and/or offer more favorable loan terms to private SMEs? Moreover, how does the change in eligible firms’ loan demand affect various aspects of banking, such as banks’ total lending, investment in securities, risk-taking, and profitability?

Second, it would be beneficial to extend the period of analysis, and compare the policy impacts in the earlier phase (as in the present study) and the later phase (such as during the outbreak of COVID-19). Following the spread of COVID-19, the BOJ announced in April 2020 that it would temporarily expand the purchases of risk assets by (i) doubling its annual ETF purchases from 6 trillion yen to a maximum of 12 trillion yen, (ii) loosening the maturity criteria for CB purchases, whereby corporate bonds with a remaining maturity of more than 3 years and up to 5 years were also eligible for the purchases, as well as (iii) conducting additional CB purchases.³⁰ Given the large scale of this intervention, the policy might have significantly prevented declines in stock and bond prices and mitigated the impact of adverse shocks due to COVID-19 on corporate financing activities. Another notable change in recent years is that from April 2021, the BOJ only purchases TOPIX ETFs. The decision to stop allocating the budget to Nikkei 225 ETFs might be a strategic move of the BOJ to address concerns about price distortion in the stock market. Additionally, while the BOJ has reduced its purchases of ETFs from 2022 onwards, it remains highly possible that this monetary policy tool will be increasingly utilized during economic downturns.

Third, although the BOJ’s ETF and CB purchases have changed firms’ security issuance behavior, further analysis is needed to understand whether the BOJ’s interventions affect firm performance and investment activities, or ownership structure and corporate governance. The literature has suggested that passive management can improve corporate governance through the increases in independent directors and equal voting rights, or the removal of takeover defenses such as poison pills (Appel et al., 2016), but can also adversely affect the governance by strengthening the power of managers and decreasing the fraction of new independent directors (Schmidt & Fahlenbrach, 2017). Therefore, exploring how the BOJ’s ETF purchases program, which leads to an increase in passive ownership, affects corporate governance would be an important direction.

Appendix

See Tables 16, 17, 18, 19 and 20.

Table 16 Control variables

Appendix Tables 17, 18, 19 and 20 are related to the discussion in Sect. 5.3.6. It provides the replication results of the main regressions (Tables 7, 8, 9, 10) using an adjusted data set. Specifically, to mitigate the potential impact of noises associated with stock exchange transfers during the estimation period, namely, due to the merger between TSE and OSE in 2013, all firms involved in this merger are excluded from the sample. After revisiting the baseline analyses using the adjusted dataset, the same pattern of results was obtained. That is, the sign and magnitude of coefficients of Tables 17, 18, 19 and 20 resemble those in Tables 7, 8, 9 and 10 confirming that the findings are robust and consistent with the hypotheses presented in Sect. 4.1.

Table 17 Replication of Table 7—policy impact on securities issuance

Table 18 Replication of Table 8—impact of ETF purchases on firms’ capital structure

Table 19 Replication of Table **9—impact of CB purchases on firms’ capital structure

Table 20 Replication of Table **10—combined effect on firms’ capital structure