Opportunistic or rational behaviour: Evidence from debt retiring IPOs

Transcription

1 Opportunistic or rational behaviour: Evidence from debt retiring IPOs Pengda Fan Graduate School of Economics, Kyushu University , Hakozaki, Higashiku Fukuoka JAPAN An early version of this paper was presented at IFABS Asia 2017 Ningbo China Conference, the annual meeting of the Japan Finance Association. I am grateful to Konari Uchida, Yusuke Kinari, Minoru Otsubo, Kazuo Yamada, Katsuhiko Okada, Henok Neguse for their helpful comments. This research have received "Mizuho Securities Endowment, Kyoto University Excellence in Research Award, 2017" Corresponding author. Graduate School of Economics, Kyushu University , Hakozaki, Higashiku, Fukuoka JAPAN. Tel.: kopdada@hotmail.co.jp 1

2 Abstract This paper attempts to examine whether debt retirement is motivated by opportunistic or rational incentives. We find that interest spread before IPO is positively associated with the level of bank debt repayment at the year of IPO. This positive relation is robust to control for the endogeneity problem and selection bias. As for the effect of debt retirement, we show that firms retiring more bank debt experience significantly greater reductions in interest spread. Furthermore, debt retiring IPOs simultaneously increase their bank debt financing and invest more during the post-ipo period, which contributes to relatively better long-term operating performance. Those results clearly indicate that debt retiring IPOs are not necessarily motived by opportunistic incentive, but by the improvement of debt financing conditions. Finally, we find that expropriation by banks is more likely to be the source of high interest burdens for bank-dependent firms. JEL classification code: G21; G30; G31; G32 Key words: Rent expropriation; Interest spreads; IPOs; Debt retirement 2

3 1. Introduction Bank debt is an important financing source for private companies (Berger and Udell, 1995, 2002), and it is well documented that bank monitoring help mitigate problem arising from information asymmetry (Diamond, 1984). However, firms reliance on bank financing also incur significant costs. On the one hand, it increases the risk of bankruptcy. On the other hand, it may provide banks with monopoly power, which enables them to expropriate wealth from borrowing companies (Rajan, 1992; Sharpe, 1990). Both stories predict that banks are likely to impose high interest spread on bank-dependent firms. Meanwhile, IPOs (initial public offerings) provide firms a vital opportunity to reduce leverage though equity issuance and interest burdens (Pagano et al., 1998; Schenone, 2010). In addition, it also enables bank-dependent firms to repay existing bank debt by using proceeds from IPO, on which high interest spread is likely charged. Then, as a result of the improved financing conditions, bank-dependent firms can increase capital expenditures during the post-ipo period. This idea indicates that debt retirement is particularly advantageous for bank-dependent firms with high interest burdens prior to IPO. Indeed, a non-trivial number of firms indicate that their primary use of IPO proceeds is bank debt retirement (e.g., 42% of sample firms in Busaba et al. (2001) and 31% of those in Dunbar and Foerster (2008). 1 Many previous studies, however, suggest that firms that state bank debt repayment as the primary use of proceeds are driven by opportunistic incentives (Amor and Kooli, 2017; Andriansyah and Messinis, 2016; Bray and Peterson, 2009; Wyatt, 2014). They conjecture that debt retiring IPOs are those which lack investment opportunities and employ a timing strategy of conducting IPO to pay down debt when operating 1 When a firm files to go public, it must disclose the intended use of IPO proceeds. 3

4 performance is unsustainably high, which leads to long-term underperformance. One potential reason for the dark side of debt retirement behaviour is that they collectively identify debt retiring IPOs by relying on the statement disclosed in IPO prospectuses. Given that actual use might be different from the stated one, it may be more informative to focus on what firms really do after the IPO (Amor and Kooli, 2017; Kim and Weisbach, 2008). This paper aims to examine whether debt retirement is motivated by opportunistic or rational incentives, by focusing on the actual use of IPO proceeds. We address this issue by using Japanese IPO data during 2001 to Japanese data is advantageous for addressing our hypothesis because the traditional Japanese system is a bank-oriented one, suggesting that bank debt is likely a predominant financing source for Japanese private companies. 2 Besides, it is well documented that firms reliance on bank debt also provides Japanese banks an opportunity to expropriate wealth from borrowing companies (Pinkowitz and Williamson, 2001; Weinstein and Yafeh, 1998; Wu et al., 2009; Wu and Yao, 2012), indicating that bank-dependent firms in Japan are more likely to bear high interest burden prior to IPO. We identify debt retiring IPOs based on actual bank debt retired at the IPO year. We find that interest spread before IPO is positively associated with the level of bank debt repayment at the year of IPO. This positive relation is robust to control for the endogeneity problem and selection bias. As for the effect of debt retirement, we show that firms retiring more bank debt experience significantly greater reductions in interest spread. Furthermore, debt retiring IPOs actually increase bank financing as well as capital expenditures during the post-ipo period, which contributes to relatively 2 In addition, bank-affiliated venture capitals (BVCs) occupy significant presence in the Japanese venture capital market. Hellmann et al. (2008) and Sun and Uchida (2016) show evidence that firms owned by BVCs tend to borrow from their parent banks. 4

5 better long-term operating performance. These clearly results indicate that debt retiring IPOs are not necessarily motived by opportunistic incentives, but by the improvement of debt financing conditions. Additionally, we find that outstanding bank debt (scaled by total liabilities) and main bank equity ownership are positively associated with firm s interest spread preceding IPO, after controlling for the effect of leverage. Given that information asymmetry has been documented to be less severe for subsidiaries of public listed firms (Ghosh et al., 2012; Prezas et al., 2000), we premise that extraction by banks may be less evident for them. Consistent with our premise, we find that outstanding bank debt is not associated with interest spread for the subsidiaries, suggesting that expropriation by banks is more likely to be the source of high interest burdens, which motivates debt retirement at the IPO. This paper makes a significant contribution to the literature. To the best of our knowledge, this is the first paper to show it is interest burdens that motivate debt retirement at the IPO. Although many previous studies relying on the information stated in prospectuses contend that firms citing debt retirement as the intended use of proceeds are the worst underperformers, our findings, instead, highlight that debt retiring IPOs substantially improve their debt financing conditions, actually increase more capital expenditures and exhibit relatively better long-term operating performance by focusing on the actual use of IPO proceeds. Secondly, our research also contributes to the literature of banks expropriation problems by showing that while bank financing is advantageous for private firm subject to high information asymmetry, it also incurs significant costs. The remainder of the paper is organized as follows. Section 2 presents a literature review and hypothesis. Section 3 descries sample selection and data collection. Section 5

6 4 presents empirical results. Section 5 deals with additional analyses. Section 6 is a brief summary and conclusion of this research. 2. Literature review and hypothesis The literature closely related to our paper focuses on the use of equity proceeds. Andriansyah and Messinis (2016) and Amor and Kooli (2017) examine whether the intended use of proceeds can explain the well-known long-term underperformance of IPO firms. They maintain that firms specifying debt retirement as the intended use of proceeds in the prospectuses are those which lack investment opportunities and take advantage of market condition by issuing overvalued stock. Consistent with their argument, they document that debt retiring IPOs invest much less and show the worst long-term operating performance. The dark side of debt retirement is also observed at the time of SEOs (seasoned equity offerings). Bray and Peterson (2009) investigate the relation between seasoned equity issuers' stated use of proceeds and their subsequent long-term performance. They find that while issuers with specific investment plan do not display underperformance, issuers citing debt retirement as the intended use of proceeds experience poor performance in the subsequent three years. They also attribute those results to market-timing behaviour of debt retiring IPOs. In addition, Wyatt (2014) argue that the designation of the proceeds for debt retirement is not informative about future investment opportunities and cash flows, thus it is expected to be associated with high ex ante uncertainty. Indeed, they find that debt retirement is positively related to initial return. However, the above-mentioned evidence is particularly puzzling given that a non-trivial number of firms indicate that their primary use of IPO proceeds is bank debt 6

7 retirement. If debt retirement is really motivated by opportunistic incentives, a more plausible strategy is to conceal the real intention since they are not obligated to commit to the statement in IPO prospectuses. Thus, it is important to uncover the potential benefits and the real motivation of debt retirement at the IPO. We hypothesize that debt retirement is motivated by rational incentive. Specifically, IPOs enable firms to repay existing bank debt by using proceeds, on which high interest spread is likely charged and to improve debt financing conditions. Then the improved financing condition should be expected to contribute to better long-term operating performance through increasing capital expenditures during the post-ipo period. Those facts give rise to the following hypothesis: Hypothesis: Bank-dependent firms bearing high interest burdens go public to retire existing bank debt and improve debt financing conditions. 3. Sample selection and data We collect information of firms that went public in the Japanese stock market during the period of 2001 to 2014 from Japanese IPO White Papers. Financial institutions and utilities are removed from the sample. This pre-filtering leads us to discard very few records and leave us a sample made up of 1474 IPOs. For those companies, we manually collect stated objectives of the IPO from prospectuses, which have been available since Firms short- and long-term average interest spreads are available from the FinancialQuest since We use 10-year Japan Government Bond yield as a proxy for risk-free rate, which is available from the Bank of Japan web-site. Financial data (such as bank debt and bank debt repayment) is also taken from the Nikkei NEEDS 7

8 FinancialQuest database. We limit our attention to firms for which the weighted average interest spread for the year before IPO (Year -1) are available. It is also worth noting that, after the second filtering, the sample is restricted to firms with outstanding bank debt prior to IPO. Our final sample consists of 714 firms. Table 1 presents sample distribution by year and stated objectives of IPO. 3 The frequency of IPO significantly declines in the second half of our sample period due to the global financial crisis in We define debt-retiring IPOs as firms that state bank debt retirement as their full or partial use of IPO proceeds. Table 1 suggests that about 5% of sample companies state bank debt retirement as their full usage of IPO proceeds in prospectuses. Although approximately 27% of sample firms cite that one of the primary usages is to retire bank debt, they also mention that capital (or R&D) expenditures as another use of IPO proceeds, which makes it difficult to assess the effect of debt retirement. As argued by Amor and Kooli (2017), it is important to control what IPO firms really do after the offering, rather than relying on the information released in the prospectuses. Motivated by the potential contamination, this research identifies debt retiring IPOs upon the actual level of debt retirement in the IPO year. 4 Specifically, we equally divide sample companies (714 companies) into four groups upon the short- and long-term bank debts retired in the IPO year scaled by total liabilities at the year before IPO (REPAY). Those in the top (bottom) quantile of REPAY are defined as High-Retiring IPOs (Low-Retiring IPOs). 5 3 Sample size in Table 1 is slightly smaller than those in subsequent analyses due to the lack of information on prospectuses (81 companies). 4 Kim and Weisbach (2008) also examine the ex-post use of IPO proceeds by relying on accounting measures. 5 We find that the proportion of firms stating bank debt retirement as their full use of IPO 8

9 As a measure of interest burden, we employ three interest spreads. SSPREAD is short-term interest spread (average short-term interest rate minus 10-year Japanese government bond yield). LSPREAD is long-term interest spread (average long-term interest rate minus 10-year Japanese government bond yield). ASPREAD is the weighted average of SSPREAD and LSPREAD. As with previous studies, we further take the natural logarithm transformation of interest spread (SSPREAD, LSPREAD and ASPREAD) since the distribution of interest spread is heavily positive skewed (Campello et al., 2011; Goss and Roberts, 2011; Graham et al., 2008). All continuous variables in this research are winsorized at the top and bottom 1% values. Table 2 presents those variables separately for High-Retiring IPOs and Low-Retiring IPOs. It indicates that High-Retiring IPOs retire 47% (28.5%) of short-term (long-term) bank debt at the IPO year, which is substantially larger than that of Low-Retiring IPOs. Importantly, High-Retiring IPOs bear significantly higher interest burden than Low-Retiring IPOs, which is consistent with our hypothesis that high interest burdens motivate debt retiring IPOs. We also find that firms with large bank debt are more likely to retire bank debt, compared to those with small bank debt, indicating that bank-dependent firms bear higher interest burdens than non-bank-dependent firms. High-Retiring IPOs also have significantly higher leverage than Low-Retiring IPOs, which is consistent with the argument of Pagano et al. (1998) that firms rebalance capital structures through IPOs. With regard to other control variables, we find that High-Retiring IPOs are smaller, younger, high sales growth firms with fewer tangible assets, which is consistent with proceeds for High-Retiring IPOs is only 5%, which is even lower than that of Low-Retiring IPOs (10.2%). In addition, unreported table shows that there is not significant difference in most of the firm characteristics between debt-retiring and non-debt-retiring IPOs, suggesting that actual use probably differs from the stated one. 9

10 the common wisdom that firms with high information asymmetry are more likely to rely on bank financing. 4. Empirical results 4.1. The determinants of bank debt retirement in the IPO year This section examines our hypothesis after controlling for other firm-specific characteristics, as well as the endogeneity concern and selection bias. Panel A of Table 3 implements regression of bank debt repayment at the IPO. The dependent variable for Model (1) to (3) is REPAY (short- and long-term bank debts retired in the IPO year scaled by total liabilities at the year before IPO). Model (1), (2) and (3) engender positive and significant signs on SSPREAD, LSPREAD and ASPREAD respectively, indicating that interest spread is an important driver of debt retirement. We also implement a Logit regression. The dependent variable for Model (4) to (6) is the binary dummy variable (HIGH-REPAY) that takes on a value of one (zero) for those in the top (bottom) quantile of REPAY. Again, we find that firms with high interest burden are more likely to retire large amount of existing bank debt at IPO year. 6 With respect to other control variables, ROA is negatively linked to the debt-retiring decision, suggesting that firms with relatively low operating performance are more likely to conduct debt retirement, which is inconsistent with the opportunistic story. Furthermore, we find not clear evidence that debt retiring IPOs tend to issue overvalued stock when the industry-wide market valuation is high, as Industry MTB Ratio (the median of industry's market-to-book ratio at the year before IPO) is not associated with 6 Unreported results show that BANKDEBT also carries a positive and significant coefficient, suggesting that firms relying on bank debt are more likely to retire existing debt. 10

11 the debt retirement in all models. In contrast, small and high growth firms are more likely to conduct debt retirement, which is more consistent with our hypothesis that high information asymmetry firms tend to depend on bank financing, and high interest burdens in turn motivate debt retiring IPO. One might criticize that High-Retiring IPOs differ from Low-Retiring IPOs with regard to various firm-specific characteristics associated with the level of debt retirement. To address this endogeneity concern, we compare SSPREAD, LSPREAD and ASPREAD between High-Retiring IPOs and their matching Low-Retiring IPOs. Specifically, for each of High-Retiring IPOs, a matching Low-Retiring IPOs is selected by using propensity score matching from those in the lowest two quantiles of REPAY. Results are presented in Panel B of Table 3. Model (1) in Panel A is used to estimate the propensity score. Again, we find significant difference in interest spread between High-Retiring IPOs and their matching IPOs. For instance, the SSPREAD is for High-Retiring IPOs, which is significantly higher than that of matching firms, (p=0.001). Another concern is that our IPO firms are non-random selection, which may create a self-selection bias. To further address this concern, we employ the Heckman 2-stage procedure. In the first stage, we estimate a logit model where we set the dependent variable IPOD, equal to one if the firm goes public and zero otherwise. Results are presented in Panel C of Table3. In the first stage regression, we find that highly leveraged firms are more likely to go public. Both the variables that measure a firm s financing need (capital expenditures and sale growth ratio) increase the probability of going public. Finally, Industry MTB Ratio also engenders a positive and highly significant sign, suggesting that favourable market condition is also an important determinant of the going-public decision. Importantly, in the second stage regression, all the interest spread variables carry positive and significant signs even after 11

12 controlling the potential selection bias. Taken together, results in Table 3 support our hypothesis that firms suffering from high interest burdens go public to retire existing bank debt Interest spread around IPO Results so far suggest that firms incurring high interest spread tend to conduct debt repayment at IPO year. To further examine our hypothesis, we investigate whether High-retiring IPOs decrease more interest burden. Table 4 reports the fixed effects regression of interest spreads around IPO. The dependent variables are SSPREAD, LSPREAD and ASPREAD respectively. POSTIPO takes on a value of one for post-ipo period (Year 0 to Year 3), zero for pre-ipo period. POST-IPO * HIGH-REPAY is the interaction term between POSTIPO and HIGH-REPAY (the binary dummy variable that takes on a value of one (zero) for those in the top (bottom) quantile of REPAY). If High-Retiring IPOs retire a large amount of existing bank debt (including early redemption of long-term bank debt) on which high interest spread is likely charged, and successfully roll over it at a lower cost, a negative coefficient is expected for POST-IPO * HIGH-REPAY. Consistent with our premise, POST-IPO * HIGH-REPAY engenders negative and significant signs for SSPREAD, LSPREAD and ASPREAD, with the coefficient on LSPREAD being the most significant one, suggesting that the effect of retiring debt on interest spread is more evident when more long-term bank debt is retired at IPO. The estimated coefficients in Model (3) suggest that, for High-Retiring IPOs, they substantially improve ASPREAD by 0.033, which amount to 6.2 million Japanese yen for the average sample firms. 7 In contrast, POSTIPO carries positive and significant signs on LSPREAD and ASPREAD, suggesting that for Low-Retiring IPOs, interest burden actually increase during the 7 [exp(0.033)-1]*1.567%* (11725 million) = 6.2 million Japanese Yen. 12

13 post-ipo period. Overall, these results clearly suggest that High-retiring IPOs significantly improve their debt financing conditions even after controlling for firm characteristics associated with creditworthiness (PUBLICDEBT, LEVERAGE, Tangible Assets, ROA, and so on) Post-IPO financing behaviours and capital expenditures We have shown evidence that firms retire bank debt successfully decrease interest burden, which is beneficial for bank-dependent firms. We next trace post-ipo financing patterns and capital expenditures to see what debt retiring IPOs actually do during the post-ipo period. If debt retiring IPOs, as argued by Andriansyah and Messinis (2016) and others, are motivated by market-timing and deleveraging considerations, it predicts that they do not increase bank debt and invest less during the post-ipo period. If debt retiring IPOs are motivated by the improvement of debt financing conditions, it predicts that, they increase bank debt since IPO significantly increase the negotiation power against banks, which allows them to roll over it in better terms. With respect to investment intensity, this story predicts that debt retiring IPOs do not invest less since the reduced financial constraints enable them to increase capital expenditures. To examine this issue, we denominated bank debt (public debt) and capital expenditures by the total assets of the firms in the year preceding IPO and the results are presented in Table 5. Panel A clearly indicates that, while High-Retiring IPOs retire large amount of bank debt at IPO year, they substantially increase bank debt as well as public debt than Low-Retiring IPOs do during the subsequent three years. Pagano et al. (1998) contend that firms go public to decrease their leverage through equity issuance. We extend their analyses by showing evidence that bank-dependent firms, one the one hand, use IPO proceeds to repay high-interest existing bank debt, on the other hand, increase bank financing simultaneously. 13

14 Panel B presents the change of capital expenditures for subsamples. Inconsistent with the opportunistic story, we find that High-retiring IPOs significantly invest more both prior to and after IPO (except for Year 3) than Low-Retiring IPO do, suggesting that debt retiring IPOs seems to be growth firms with more investment opportunities Post-IPO operating performance Since debt retiring IPOs substantially improve their debt financing conditions by retiring high-interest bank debt and rolling it at a lower cost, and consequently increase more capital expenditures, we predict that High-Retiring IPOs do not underperform Low-Retiring IPOs. As with Amor and Kooli (2017), we examine the change of the industry-adjusted operating performance, both before and after interest expenses. OCF_TA is operating cash flow to total assets. OCF_SALES is operating cash flow to sales. NCF_TA is operating cash flow after interest expenses to total assets. NCF_SALES is operating cash flow after interest expenses to sales. We compute industry-adjusted change in operating performance of an IPO firm as the difference between its change in operating performance and the median change in operating performance of all non-ipo firms in its industry. Panel A of Table 6 compares the change of operating performance between High-Retiring IPOs and Low-Retiring IPOs. 9 In terms of operating performance before interest payment (OCF_TA), we observe a pervasive decline for Low-Retiring IPOs over a period of three years after IPO. In contrast, High-Retiring IPOs show improvement in IPO year and a less-obvious decline in the subsequent three years, indicating that the improved financing conditions also positively affect the investment 8 Similar but weaker results are obtained when matched sample analyses are conducted. 9 When OCF_SALES and NCF_SALES are compared, we also find that High-Retiring IPOs significantly outperform Low-Retiring IPOs at the IPO year, but do not underperform during the subsequent three years. 14

15 efficiency. In terms of operating performance after interest payment (NCF_TA), again, we do not find clear evidence of underperformance for High-Retiring IPOs. Panel B shows the results of fixed effects model, where the dependent variables are OCF_TA, OCF_SALES, NCF_TA and NCF_SALES respectively. POSTIPO takes a value one for post-ipo period (Year 0 to Year 3), zero for pre-ipo period. POST-IPO * HIGH-REPAY is the interaction term between POSTIPO and HIGH-REPAY. In all models, the coefficients of POSTIPO are negative, indicating that the well-documented long-term underperformance is at least partly due to the underperformance of Low-Retiring IPOs. As regard to POST-IPO * HIGH-REPAY, we find positive and significant coefficients. On average, High-Retiring IPOs outperform Low-Retiring IPOs by approximately 4.3% in terms of OCF_TA. Besides, the estimated coefficients clearly suggest that High-Retiring IPOs do no show long-term underperformance, rejecting the idea that debt retiring IPOs tend to conduct IPO when operating performance is overstated. Taken together, these results highlight that debt retirement at the IPO is not motivated by opportunistic incentives, but motivated by the improvement of financing conditions. 5. Further Analyses In this section, we examine the source of high interest burden for bank-dependent firms. As argued before, high interest burdens for bank-dependent firms can be attributable to bankruptcy risk as well as to expropriation by banks. To disentangle these two stories, we start our empirical analyses by running regression of interest spreads in Year -1 for entire IPO firms. Then, we conduct regression analyse separately for two subsamples upon the hypothesized extent of expropriation by banks. Finally, as with Takahashi and Yamada (2015), we take advantage of financial information of 15

16 private firm, which enable us to examine whether rent extraction by banks is prevalent among private firms. In Japan, private firms with more than 500 shareholders and more than 10 million yen of public equity issued, and private firms with more than 100 million yen of private equity issued, are required to file with the Ministry of Finance. This implies that those Japanese private firms, collected by Nikkei NEEDS FinancialQuest database, are qualified candidates for public listing. 10 While not reported, we find that BANKDEBT of 0.48 for IPO firms is significantly higher compared to 0.44 for private firms. With respect to interest burden, we find that short-term (long-term and weighted-average) interest spread for IPO firms is significantly higher than that of private firms, indicating that bank-dependent private firms have an incentive to improve their debt financing conditions through IPO Bank reliance and interest spreads for IPO firms We start our empirical analyses by running OLS regression of interest spreads in Year -1 to examine the source of high interest spread. 11 The key independent variable is BANKDEBT, which is measured as the sum of short- and long-term bank debts scaled by total liabilities. LEVERAGE and PUBLICDEBT are also included. We stress that our bank reliance measures are scaled by total liabilities to disentangle effects of bank reliance from leverage effects. Since LEVERAGE has a relatively high correlation with BANKDEBT (0.55), which might lead to potential multicollinearity, we then use the residual term from the regression of LEVERAGE on BANKDEBT as a control variable 10 After limiting private sample to those with available interest spread data (with outstanding bank debt), our private sample consists of firm-year observations from 2000 to 2014 (1481 private firms). 11 We cannot employ fixed effects model here since interest spread data is only available from the fiscal year preceding IPO for most of our sample firms (574 out of 714 IPO firms). 16

17 (LEVERAGE Residual) instead of LEVERAGE. 12 Various control variables are also adopted. Since large and old firms are likely to incur low interest burden due to their credit worthiness, natural logarithm of total assets (Ln(Assets)) and firm age (Firm Age) from foundation are adopted (Hale and Santos, 2009). 13 High interest spread may be charged on firms with high uncertainty and information asymmetry. We address the issue by using sales growth rate. Finally, tangible assets over total assets (Tangible Assets) is included since tangible assets can be used as collateral, which mitigates problems arising from information asymmetry (Santos and Winton, 2008). Panel A of Table 7 presents OLS regression of interest spreads preceding IPO. The coefficients of BANKDEBT are positive and significant at the 1% level in the regressions of short-term interest burden (SSPREAD), long-term interest burden (LSPREAD) and average interest spread (ASPREAD). The marginal effect of bank debt is also economically significant. One standard deviation (0.209) increase in BANKDEBT is associated with a increase in SSPREAD, which amounts to an increase of 9.5 million Japanese Yen based on the mean value of short-term bank loan (5324 million Japanese Yen) and the mean value of short-term interest spread before log transformation (1.615%). 14 Consistent with a common view that creditors charge high interest rates on less credit worthy companies, LEVERAGE Residual has a positive and significant coefficient. We stress that the positive association between BANKDEBT and interest spread is obtained after controlling for leverage effects. With respect to other control variables, large companies tend to incur relatively low interest spread. Sales Growth Ratio is positively associated with interest spread 12 Similar results are obtained when LEVERAGE and BANKDEBT are included simultaneously. 13 Firms foundation years are obtained from IPO White Papers. 14 [exp(0.105)-1]*1.615%* (5324 million) = 9.5 million Japanese Yen. 17

18 probably because banks view growing companies as risky borrowers. Longer maturity is also associated with higher interest burden. In addition to capital structure variable, we also employ an alternative proxy variable for bank-firm relationship. Given that monitored bank debt in Japan is usually delegated to the main bank (Wu et al., 2009), and that Weinstein and Yafeh (1998) find that firms with main banks pay higher average interest spread on their liabilities than those without main banks, we include MBOWN (ownership by main bank) in the interest spread regressions. Results are presented in Model (4) to (6). Consistent with the story that bank-dependent firms are likely to suffer from expropriation by banks, we find that MBOWN engenders a positive and significant sign, suggesting that main bank extract rents through higher interest rates. Importantly, the coefficient for BANKDEBT is almost unchanged. As further robustness checks, we also include the number of lending banks, since Houston and James (1996) argue that lending from multiple banks is an effective way to curb expropriation by banks. However unreported results suggest that multiple banking relationships have little effect in curbing holdup problems by banks while our main results remains unchanged. The stylized fact we document is consistent with the rent-extraction story, but they may be also consistent with other alternative story, such as the effect of interest-bearing debt. Since public debt is only sparsely used, it is extremely difficult to distinguish whether the positive relationship is due to bank debt or due to interest-bearing debt. To further confirm banks expropriation, we test a key cross-sectional prediction that is unique to rent-extraction story. Specifically, we separately implement OLS regression analyses for two subsamples upon the hypothesized extent of expropriation by banks. Since Ghosh et al. (2012) and Prezas et al. (2000) document that IPOs of public listed firms (equity carve-out) are less underpriced than independent IPOs because the 18

19 information asymmetry is less severe for carve-out, we premise that subsidiaries of public listed firms are less vulnerable to extraction by banks. 15 To test this premise, we add an interaction term between PARENTD (takes on a value of one for subsidiary of public listed firm and zero otherwise) and BANKDEBT and the results are presented in Panel B. It shows that the coefficient of BANKDEBT is unaffected by the inclusion of the interaction term, suggesting that expropriation by banks is prevalent among independent IPOs. Importantly, PARENTD * BANKDEBT engenders negative and significant signs in all models and the combined effect of BANKDEBT for the subsidiaries on interest spread is negligible, supporting that high interests spread is probably due to expropriation by banks Bank reliance and interest spreads for private firms One may further raise the concern that OLS regression lacks control for the time-invariant firm-specific characteristics. To address this concern, we employ fixed effects model for private firms, where the dependent variables are SSPREAD, LSPREAD and ASPREAD. Results in Panel A of Table 8 suggest that BANKDEBT is positively related to SSPREAD. A one standard deviation increase in BANKDEBT increases the interest burden by 7.8 million Japanese Yen. As with Schenone (2010), we further include the a quadratic BANKDEBT term, BANKDEBT 2. Similar to theirs, we find a U-shaped relation between BANKDEBT and interest spread. This suggests that banks can successfully lock in their clients after providing about 35% of the total liabilities. As with Table 7, we further include the interaction terms PARENTD * BANKDEBT 15 As public debt in Japan is only sparsely used by private firms, it is difficult to examine whether firms with public debt issuance prior to IPO are less likely to suffer from extraction by banks. 19

20 and PARENTD BANKDEBT 2. Again, we find that extraction by banks is less severe for IPOs of public listed firms. The estimated coefficients in Model (6) suggests that, banks can lock in independent firms after providing 34% of total liabilities, which is significantly lower than that of subsidiaries (40%). Taken together, these results suggest that IPO may provide bank-dependent firms an opportunity to mitigate expropriation by banks. 6. Conclusion A non-trivial number of IPOs indicates that the primary use of IPO proceeds is debt retirement. Given that retiring existing bank debt, on which high interest spread is likely charged, is a rational decision for bank-dependent firms, it is surprising that many previous studies suggest that debt retirement is motivated by opportunistic incentives. By relying on the actual level of debt retirement at the IPO year rather than on the statement from IPO prospectuses, we investigate whether debt retirement is motivated by opportunistic or rational incentives. We find that interest spread before the IPO are positively associated with the debt repayment level at the year of IPO. Debt retiring IPOs show a significantly greater decline in interest spreads. Furthermore, debt retiring IPOs significantly increase their bank debt financing and invest more during the post-ipo period, which contributes to relatively better long-term operating performance. These results suggest that, debt retirement is actually beneficial for bank-dependent firms. In addition, consistent with the rent-extraction story, we find that bank debt (main bank equity ownership) is positively associated with interest spread prior to IPO. 20

21 Reference Amor, SB, Kooli, M. Intended use of proceeds and post-ipo performance. Quarterly Review of Economics and Finance 2017;65; Andriansyah, A, Messinis, G. Intended use of IPO proceeds and firm performance: A quantile regression approach. Pacific-Basin Finance Journal 2016;36; Berger, AN, Udell, GF. Relationship lending and lines of credit in small firm finance. Journal of Business 1995; Berger, AN, Udell, GF. Small business credit availability and relationship lending: The importance of bank organisational structure. Economic Journal 2002;112;F32-F53. Bray, DE, Peterson, DR. Intended use of proceeds and the long-run performance of seasoned equity issuers. Journal of Corporate Finance 2009;15; Busaba, WY, Benveniste, LM, Guo, R-J. The option to withdraw IPOs during the premarket: empirical analysis. Journal of Financial Economics 2001;60; Campello, M, Lin, C, Ma, Y, Zou, H. The real and financial implications of corporate hedging.journal of Finance 2011;66; Diamond, DW. Financial intermediation and delegated monitoring. Review of Economic Studies 1984;51; Dunbar, CG, Foerster, SR. Second time lucky? Withdrawn IPOs that return to the market. Journal of Financial Economics 2008;87; Ghosh, C, Petrova, M, Feng, Z, Pattanapanchai, M. Does IPO Pricing Reflect Public Information? New Insights from Equity Carve Outs. Financial Management 2012;41;1-33. Goss, A, Roberts, GS. The impact of corporate social responsibility on the cost of bank loans. Journal of Banking & Finance 2011;35; Graham, JR, Li, S, Qiu, J. Corporate misreporting and bank loan contracting. Journal of Financial Economics 2008;89; Hale, G, Santos, JA. Do banks price their informational monopoly? Journal of Financial Economics 2009;93; Houston, J, James, C. Bank information monopolies and the mix of private and public debt claims. Journal of Finance 1996;51; Kim, W, Weisbach, MS. Motivations for public equity offers: An international perspective. Journal of Financial Economics 2008;87;

22 Pagano, M, Panetta, F, Zingales, L. Why do companies go public? An empirical analysis. Journal of Finance 1998;53; Pinkowitz, L, Williamson, R. Bank power and cash holdings: Evidence from Japan. Review of Financial Studies 2001;14; Prezas, AP, Tarimcilar, M, Vasudevan, GK. The Pricing of Equity Carve Outs. Financial Review 2000;35; Rajan, RG. Insiders and outsiders: The choice between informed and arm's length debt. Journal of Finance 1992;47; Santos, JA, Winton, A. Bank loans, bonds, and information monopolies across the business cycle. Journal of Finance 2008;63; Schenone, C. Lending relationships and information rents: Do banks exploit their information advantages? Review of Financial Studies 2010;23; Sharpe, SA. Asymmetric information, bank lending, and implicit contracts: A stylized model of customer relationships. Journal of Finance 1990;45; Takahashi, H, Yamada, K. IPOs, growth, and the impact of relaxing listing requirements. Journal of Banking & Finance 2015;59; Weinstein, DE, Yafeh, Y. On the costs of a bank centered financial system: Evidence from the changing main bank relations in Japan. Journal of Finance 1998;53; Wu, X, Sercu, P, Yao, J. Does competition from new equity mitigate bank rent extraction? Insights from Japanese data. Journal of Banking & Finance 2009;33; Wu, X, Yao, J. Understanding the rise and decline of the Japanese main bank system: The changing effects of bank rent extraction. Journal of Banking & Finance 2012;36; Wyatt, A. Is there useful information in the use of proceeds disclosures in IPO prospectuses? Accounting & Finance 2014;54;

23 Table 1 Sample distribution This table indicates yearly distribution of our sample IPOs. Debt-retiring IPOs are those that state bank debt retirement in the prospectus as a primary (full or partial) use of IPO proceeds. All other IPOs are defined as non-debt-retiring IPOs. Year Debt-retiring IPOs (Full use of proceeds) Debt-retiring IPOs (Partial use of proceeds) Non-debt-retiring IPOs Total IPOs [24%] 11 [11%] 63 [65%] [9%] 23 [34%] 38 [57%] [6%] 14 [20%] 51 [74%] [5%] 23 [26%] 61 [69%] [10%] 16 [22%] 49 [68%] [6%] 42 [50%] 37 [44%] [7%] 12 [26%] 31 [67%] [0%] 3 [27%] 8 [73%] [0%] 3 [50%] 3 [50%] [0%] 1 [17%] 5 [83%] [0%] 4 [24%] 13 [76%] [6%] 2 [11%] 15 [83%] [3%] 6 [19%] 24 [77%] [0%] 7 [33%] 14 [67%] 21 Total 54 [5%] 167 [27%] 412 [68%]

24 Table 2 Summary statistics This table reports summary statistics for High-Retiring IPOs and Low-Retiring IPOs. The entire sample consists of 714 IPOs in Japan during 2001 to We equally divide sample companies into four groups upon the short- and long-term bank debts retired in the IPO year scaled by total liabilities at the year before IPO (REPAY). Those in the top (bottom) quantile of REPAY are defined as High-Retiring IPOs (Low-Retiring IPOs). SREPAY (LREPAY) is short-term (long-term) bank debts retired in the IPO year scaled by total liabilities at the year before IPO. SSPREAD is short-term interest spread (average short-term interest rate minus 10-year Japanese government bond yield). LSPREAD is long-term interest spread (average long-term interest rate minus 10-year Japanese government bond yield). ASPREAD is the weighted average of SSPREAD and LSPREAD. We further take natural logarithm transformation of interest spread as with previous studies. BANKDEBT is short- and long-term bank debts over total liabilities. LEVERAGE is total liabilities over total assets. Ln (Assets) is natural logarithm of total assets. CASH is cash and its equivalents over total assets. Sale Growth Ratio is percentage sales growth ratio from previous year. ROA is operating income divided by total assets. Data preceding the IPO are presented (Year -1). All continuous variables are winsorized at the top and bottom one percent values. P-values are for mean (median) difference between. Full sample High- Retiring IPOs Low-Retiring IPOs P- value REPAY [0.200] N= [0.593] N= [0.062] N=179 SREPAY [0.041] [0.368] [0.001] N=714 N=178 N=179 LREPAY [0.115] [0.251] [0.046] N=714 N=178 N=179 SSPREAD [4.940] [5.139] [4.884] N=714 N=178 N=179 LSPREAD [5.245] [5.331] [5.188] N=714 N=178 N=179 ASPREAD [5.088] [5.224] [4.993] N=714 N=178 N=179 BANKDEBT [0.492] [0.587] [0.312] N=714 N=178 N=179 LEVERAGE [0.718] [0.730] [0.708] N=714 N=178 N=179 Ln (Assets) [8.811] [8.370] [9.077] N= 714 N=178 N=179 CASH [0.170] [0.205] [0.163] N=714 N=178 N=179 Sale Growth Ratio [0.161] [0.282] [0.127] N=710 N=177 N=176 ROA [0.086] [0.093] [0.078] N=714 N=178 N=179 Industry MTB Ratio [0.917] [0.938] [0.869] N=714 N=178 N=179 ***: Significant at the 1% level; Significant at the 5% level; Significant at the 10% level 0.000*** [0.000***] 0.000*** [0.000***] 0.000*** [0.000***] 0.000*** [0.000***] 0.000*** [0.000***] 0.000*** [0.000***] 0.000*** [0.000***] 0.030** [0.028**] 0.000*** [0.000***] 0.006*** [0.008***] 0.000*** [0.000***] [0.010***] 0.004*** [0.004***] 24

25 Table 3 Determinants of bank debt retirement in IPO year This table explores the determinants of bank debt retirement in IPO year. The dependent variable for Model (1) to (3) in Panel A is REPAY (short- and long-term bank debts retired in the IPO year scaled by total liabilities at the year before IPO). The dependent variable for Model (4) to (6) in Panel A is the binary dummy variable (HIGH-REPAY) that takes on a value of one (zero) for those in the top (bottom) quantile of REPAY. SSPREAD is short-term interest spread (average short-term interest rate minus 10-year Japanese government bond yield). LSPREAD is long-term interest spread (average long-term interest rate minus 10-year Japanese government bond yield). ASPREAD is the weighted average of SSPREAD and LSPREAD. We further take natural logarithm transformation of interest spread (SSPREAD, LSPREAD and ASPREAD) as with previous studies. LEVERAGE is total liabilities over total assets. Ln (Assets) is natural logarithm of total assets. CASH is cash and its equivalents over total assets. Sale Growth Ratio is percentage sales growth ratio from previous year. ROA is operating income divided by total assets. Industry MTB Ratio is the median of industry's market-to-book ratio at the year before IPO. CAPEXP is capital expenditures scaled by lagged PPE. Panel B reports the comparison of SSPREAD, LSPREAD and ASPREAD between High-Retiring IPOs and their matching Low-Retiring IPOs. For each of High-Retiring IPOs, a matching Low-Retiring IPOs is selected by using propensity score matching from those in the lowest two quantiles of REPAY. Model (1) in Panel A is used to estimate the propensity score for sample companies. t-statistics are for difference test between High-Retiring IPOs and matching Low-Retiring IPOs. Panel C reports the Heckman 2-stage procedure. In the first stage, we estimate a logit model where we set the dependent variable IPOD, equal to one if the firm goes public and zero otherwise. All continuous variables are winsorized at the top and bottom one percent values. All estimations include industry and year dummies (not reported). 25

26 Panel A Model (1) Model (2) Model (3) Model (4) Model (5) Model (6) Dependent variable REPAY REPAY REPAY HIGH-REPAY HIGH-REPAY HIGH-REPAY SSPREAD 0.079*** (3.05) 0.583** (2.39) LSPREAD 0.087** (2.43) 0.896** (2.32) ASPREAD 0.081** (2.14) 0.570* (1.66) Control variable: LEVERAGE (-0.11) (0.07) (0.04) (1.28) (1.24) (1.38) Ln (Assets) *** (-4.07) *** (-4.44) *** (-4.34) *** (-3.42) *** (-3.38) *** (-3.49) CASH (0.01) (0.28) (0.17) (0.48) (0.79) (0.61) Sale Growth Ratio 0.084** (2.12) 0.089** (2.25) 0.086** (2.16) 0.588** (2.20) 0.661** (2.42) 0.616** (2.28) ROA (-0.35) (-0.49) (-0.39) * (-1.70) * (-1.82) * (-1.77) Industry MTB Ratio (-0.17) (-0.03) (-0.14) (1.14) (1.15) (1.10) Constant (1.47) (0.85) (1.04) (-0.31) (-0.93) (-0.29) Year dummy YES YES YES YES YES YES Industry dummy YES YES YES YES YES YES N R Panel B 1 vs 1 Matching SSPREAD LSPREAD ASPREAD High-Retiring IPOs Matching Low-Retiring IPO t-statistics 4.12*** 3.46*** 3.61*** N

27 Panel C Model (1) Model (2) Model (3) Model (4) Model (5) Model (6) First-Stage Second-Stage First-Stage Second-Stage First-Stage Second-Stage Dependent variable IPOD REPAY IPOD REPAY IPOD REPAY SSPREAD 0.080*** (3.06) LSPREAD 0.090** (2.34) ASPREAD 0.084** (2.41) Control variable: LEVERAGE (-0.38) (-0.20) (-0.25) Ln (Assets) *** (-2.97) *** (-3.26) *** (-3.11) CASH (-0.24) (0.08) (-0.08) Sale Growth Ratio 0.077** (2.42) 0.084*** (2.63) 0.079** (2.46) ROA (-0.29) (-0.43) (-0.33) Industry MTB Ratio (-0.48) (-0.33) (-0.51) Lambda (-0.96) (-0.89) (-1.06) Firm Age *** (-17.62) *** (-17.62) *** (-17.62) Tangible Assets *** (-5.67) *** (-5.67) *** (-5.67) CAPEXP (1.18) (1.18) (1.18) Industry MTB Ratio 1.547*** (15.12) 1.547*** (15.12) 1.547*** (15.12) Year dummy YES YES YES YES YES YES Industry dummy NO YES NO YES NO YES Constant *** (-8.25) 0.590* (1.70) *** (-8.25) (1.29) *** (-8.25) (1.50) N ***: Significant at the 1% level; Significant at the 5% level; Significant at the 10% level 27

28 Table 4 Interest spread around IPO This table reports the interest spreads around IPO. The dependent variables are SSPREAD, LSPREAD and ASPREAD respectively (fixed effects model are employed). POSTIPO takes a value one for post-ipo period (Year 0 to Year 3), zero for pre-ipo period. POST-IPO * HIGH-REPAY is the interaction term between POSTIPO and HIGH-REPAY (the binary dummy variable that takes on a value of one (zero) for those in the top (bottom) quantile of REPAY). PUBLICDEBT is public debts scaled by total liabilities. LEVERAGE is total liabilities over total assets. Ln (Assets) is natural logarithm of total assets. Firm Age is firm age from foundation. Sale Growth Ratio is percentage sales growth ratio from previous year. ROA is operating income divided by total assets. Tangible Assets is tangible assets over total assets. MATURITY is measured as the ratio of long term bank debts to total bank debts. All continuous variables are winsorized at the top and bottom one percent values. All regressions include year dummies (not reported). z-statistics based on heteroskedasticity-consistent method are shown in parentheses. Model (1) Model (2) Model (3) Dependent variable SSPREAD LSPREAD ASPREAD POSTIPO (1.18) 0.064*** (2.69) 0.051** (2.12) POST-IPO * HIGH-REPAY * (-1.76) *** (-3.00) ** (-2.48) Control variable: PUBLICDEBT (-1.61) (1.41) (0.35) LEVERAGE 0.458*** (2.61) (1.60) 0.386*** (2.87) Ln (Assets) ** (-2.49) (-0.42) ** (-2.11) Firm Age (-1.53) (-1.40) *** (-5.00) Sale Growth ratio (0.95) (-0.34) (0.79) ROA (-0.31) (-0.51) (-0.42) Tangible Assets (-0.35) (-0.60) (-1.19) MATURITY 0.238*** (3.52) Constant 8.431*** (4.58) 7.664*** (4.87) 7.317*** (16.30) Year dummy YES YES YES N R ***: Significant at the 1% level; Significant at the 5% level; Significant at the 10% level 28