Seasoned equity offering announcements and the returns on European bank stocks and bonds

ABSTRACT

We analyse, by means of appropriate event studies, the returns following seasoned equity offering announcements made by western European banks between 2008 and 2014. Consistently with the pertinent literature on non-financial companies, we find that shareholders experience negative returns. We highlight that the same occurs for bondholders, although not surprisingly to a smaller extent. Overall, our results show that seasoned equity offering announcements play an important signalling role also in the banking industry, despite the tight regulation and supervision by banking authorities, which should in principle reduce the impact on pricing of the information asymmetries about banks’ financial conditions.

KEYWORDS:

• Seasoned equity offerings
• SEO
• excess bond returns
• excess stock returns
• event study
• security issuance

JEL CLASSIFICATION:

• G14
• G30
• G32

Acknowlegment

The authors would like to thank for their helpful comments the participants at the 16th Annual Meeting of the European Economic and Finance Society - 2017 Conference at the University of Ljubljana.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Kalay and Shimrat (1987) find a negative effect on bond returns, while Elliott, Prevost, and Rao (2009) find a positive return. To the best of our knowledge, no evidence concerning bond returns exists for European markets, let alone focusing specifically on the banking industry.

² In particular, managers may decide to issue equity when they perceive their firm to be overvalued. If this is the case, a SEO announcement would induce rational investors to purchase additional equity only at a discount: both stock values and bond values would then be negatively affected.

³ Clearly, this holds under the assumptions that the demand curve of stocks is downward sloping with respect to price (i.e. stocks are normal goods, with a negative price elasticity), and that market prices reflect the equilibrium between supply and demand.

⁴ Leland and Pyle (1977) discuss how an entrepreneur may reveal his confidence in the company by retaining a large fraction of ownership. By issuing new equity, he may reduce his share, sending a negative signal about the future cash flows. Myers and Majluf (1984) contend that managers acting in the interest of current shareholders would issue new equity only when perceive it to be overvalued, again sending a negative signal to the market. Miller and Rock (1985) suggest that unexpected reductions in dividends or relying on external equity financing can be interpreted as a sign of current cash flows being below expectations, again providing negative information to the market.

⁵ An estalished view in the literature argues that certain corporate events may trigger conflicts of interest between shareholders and bondholders. See, e.g. Scott (1992) for the potential conflicts associated to leveraged buyouts, Klock, Mansi, and Maxwell (2005) for the wealth effects of anti-takeover provisions, or Maxwell and Stephens (2003) for the wealth effects of stock repurchases.

⁶ They estimate a cumulative standardized excess premium bond return of −2% in the [−15, +15] event window.

⁷ Trading frequency in the bond market is often lower than in the stock market, so that bond prices may not be available with the required frequency. To overcome this limitation of the analysis, multiple-day returns (whenever daily returns are not available) are often used to estimate the mean and the standard deviation of daily returns. See, among others, Handjinicolaou and Kalay (1984) for a description and application of this procedure. We adopt a more restrictive approach, by only focusing on bonds for which there are no missing observations at the daily frequency.

⁸ We exclude bonds with lower maturity because their daily returns may be more sensible to monetary policy (or to the overall conditions of the money market) than to the quality of the bank itself.

⁹ See Appendix 6 for a description of the procedure we adopt to convert the credit ratings issued by the main rating agencies into a unique numerical credit score.

¹⁰ As we are focusing on Europe, we select an index that tracks the performance of Euro-denominated bonds issued by European banks. For non-Euro denominated bonds, we use Euro-converted prices in order to eliminate price fluctuations determined by exchange rate movements.

¹¹ This index measures the fluctuation range expected by the market, i.e. the implied volatility of the ‘Euro STOXX 50 Index’, extrapolated from the prices of a basket of options on the same index quoted at or out of the money.

¹² Note that the dummy is set to zero even if the equity issue takes place while having a national government as one of the shareholders of the bank, as long as the bank has exited the financial support program and paid back emergency bailout funds (i.e. if the equity issue takes place at market conditions).

¹³ See MacKinlay (1997) for a thorough review of the event study methodology.

¹⁴ As already noted in Footnote 7, and contrary to previous studies, we only consider bonds for which we have a complete time series of daily prices during the event window. This allows us to work with actual market returns rather than with estimates of daily returns.

¹⁵ To account for potential event-induced variance in stock returns, we use the test proposed by Boehmer, Musumeci, and Poulsen (1991) – as modified by Kolari and Pynno¨Nen (2010) – in evaluating the statistical significance of CAARs.

¹⁶ The generalized sign test proposed by Cowan (1992) is a non-parametric test that allows us to properly account for our somewhat limited sample size.

¹⁷ The dummy Greece is never statistically significant. As shown in Table 1, only two equity issues in our sample have been completed by Greek banks, both in 2013 (i.e. in the post-crisis sub-period) and with government intervention. In both cases, however, the banks have been able to raise enough funds from private investors in order to be still considered as privately run under the terms of the Greek bailout program.

¹⁸ Quite surprisingly, when focusing on the sub-samples accounting for the pre-crisis and post-crisis periods, we find that the positive effect of Government is confined to the pre-crisis period only, while after the crisis the corresponding coefficients are not statistically significant. This notwithstanding, it is worth emphasizing that the regressions in Table 8 do not control for individual bank characteristics. Indeed, the picture changes quite substantially when taking bank characteristics into account, with stock returns being positively correlated with government intervention both in the pre-crisis and post-crisis sub-samples, as we show when discussing the results of the multivariate analysis in Table 9.

¹⁹ We also test additional variables to account for the leverage effect on abnormal returns: the total debt to total assets ratio, the total debt to book equity ratio, and the total amount of the equity issue divided by total debt outstanding. In the article, we only report the ratio between the size of the equity issue and total long-term debt outstanding because it is the one with the highest statistical significance in all the event windows. Nonetheless, the sign of the relevant coefficient – when significant – remains positive, irrespective of the indicator used as independent variable.

²⁰ It is important to stress that our analysis is limited to banks with traded stocks and bonds, which are likely to be the largest banks in the industry. Given that we find a positive relationship between bank size and stock CARs, our estimate of the average effect may underrate the true average negative effect of the announcement. Moreover, banking is a peculiar industry, with strict and highly specific regulations, so that our results – here and in the following sub-sections – may not directly apply to firms operating in other sectors.

²¹ See, for example, the requirements imposed both by the United States and the European Union on the so-called global list of systemically important banks (G-SIBs), and on the domestic systemically important banks (D-SIBs) as defined by the Financial Stability Board.

²² Note that there are no SEOs by Greek banks in our sample for the pre-crisis period. The finding of higher returns for banks in IIPS countries with respect to those in more financially sound countries, although somewhat striking, may reflect the fact that SEO announcements are more likely to be expected in financially weaker countries, so that observing one from a bank in a financially stronger country has a deeper negative content.

²³ We experiment with other covariates, such as return on assets, return on equity, volatility of the bank’s stock prices in the fifty-two weeks before the issue, and the average rating of the bank’s outstanding bonds. None of them is statistically significant, and their exclusion does not affect the coefficients and statistical significance of any other independent variables.

²⁴ Observations are clustered as each bank in the sample may have multiple bonds outstanding. The statistical significance of CARs is evaluated using standard errors corrected for clustered observations.

²⁵ As for equity, we investigate various sub-periods within the [−30, +30] time interval, finding again significant CARs only within a [−5, +5] event window. Within this time interval, we focus on the same event windows – i.e. [−5, +5], [−2, +2], [0, +2] – used in the analysis of stock returns.

²⁶ We also test for differences in CARs between investment and non-investment grade bonds, by simply regressing bond CARs on the dummy Investment. The resulting coefficient is not statistically significant in all event windows.

²⁷ The same multivariate regressions are run also including as covariates the stock CARs in the corresponding event window, and the credit score of the individual bond. The results of these regressions are not reported, being these two variables never statistically significant.

²⁸ As in the case of the bond-level analysis, there is no evidence of statistically significant differences in the average CARs between investment and non-investment grade banks.

²⁹ The rating of each synthetic bond corresponds to the weighted average rating of every bank outstanding (rated) bonds. Hence, with a slight abuse of notation, we take a bank to coincide with its synthetic bond.

³⁰ Note that there are no banks with investment grade synthetic bonds in Greece over the period we consider.

³¹ We also run the same bank-level analysis by splitting our sample into the two sub-samples corresponding to the period before the beginning of the European debt crisis and that following it. The corresponding results are reported in Appendix 7. The relevance of such analysis is constrained by the observation that for the pre-crisis period we only have one SEO announcement for a non-investment grade bank, so that a model allowing for the interactions between the dummy Investment and the other regressors cannot be estimated. Furthermore, when adding interaction terms in the post-crisis period, IIPS cannot be interacted with Investment due to the presence of perfect collinearity. Similarly, the dummy Government cannot be interacted with Investment, even in the entire sample, because of collinearity. Hence, results are not fully comparable with those at the bond level.

³² Indeed, our multivariate regression analysis indicates that stock and bond CARs are not significantly correlated, and that on average the value of bonds diminishes, even if leverage and hence default risk decrease following a SEO.

³³ Note that this technique is commonly adopted. For example, the data and information provider Markit uses a similar procedure to determine the ‘Markit iBoxx Rating’ that is used to discriminate between investment and non-investment grade bonds in the construction of bond market indexes.