The Options Market Reaction to Bank Loan Announcements

January 1996 is the first month for which options data are available on Ivy DB OptionMetrics.

Data download from Dealscan as of February 2012.

We follow the aforementioned studies and consider each facility to be a separate observation for the sample, given loan characteristics as well as loan spreads may very well vary across facilities (Kim et al. 2011, pp. 1,166 and footnote 15). At this point, Billett et al. (2011) mention that each of their events represents a deal, rather than a facility (footnote 3) and Cook et al. (2003) make use of only the facility listed first in each of the deals, but find no material difference in their results if they not control for multiple facilities with the same borrower–lender combinations (footnote 21). However, different facilities corresponding to the same deal may possess different loan characteristics (e.g. different maturities for different tranches of the deal), so we follow from previous research focusing the analysis on the facility level, while we address this issue by making use of standard errors clustered at the firm level in all regression analysis that follows (Graham et al. 2008; Petersen 2009; Gow et al. 2010; Kim et al. 2011). On several instances, the expressions ‘loan facility’ and ‘loan’ are used in text without distinction, by making reference to loan facilities each time.

In cases where a number of different borrowers were mentioned for a particular facility in Dealscan, matching with Compustat firms was performed for the borrowing firm which was mentioned first.

More specifically, a so-called ‘power search’ according to company name was performed in Lexis/Nexis, by indicating that the desired level of relevance is ‘strong’.

In case we find an announcement mentioning that ‘a borrower is either seeking the loan or expecting to receive the loan or that a lead bank is inviting syndicate members to participate in a loan’ as stated by Maskara and Mullineaux (2011a, p. 686), we then consider that an announcement is indeed made, even if the loan is not considered finalised or a ‘done deal’. Nevertheless, from the moment a certain amount of publicity pre-exists about a loan that is subsequently concluded (and thus appears in Dealscan), we do classify it as announced.

For a detailed description of loan syndication, see also Maskara and Mullineaux (2011b).

Maskara and Mullineaux (2011b), using data from Dealscan, mention that non-syndicated loans are significantly smaller than those that are syndicated, with Dealscan reporting roughly equal number of bank loans and syndicated loans, at least during their 1985–1999 sample period. (footnote 3). We consider that the greater representation of syndicated loans in our sample may also stem from the fact that the 12,928 tranches representing the sample of firms with data on Compustat and Optionmetrics matched with loan data from Dealscan may very well represent larger than average firms from Dealscan.

We further use the average of call and put implied volatility (Truong et al. 2012) among our robustness checks. Since both the call and the put implied volatilities from Ivy are close-to-the-money, the average and delta-interpolated implied volatilities are almost indistinguishable (average correlation coefficient in excess of 0.90). Results are unaffected by the use of average, call-only or put-only implied volatility.

In accordance with past research, we employ Dealscan’s ‘All-in Spread Drawn’ item as our facility spread proxy, which is equivalent to the amount paid by the borrower in basis points (b.p.) over LIBOR or LIBOR equivalent for each dollar drawn (Graham et al. 2008). Please also refer to Appendix A for detailed variable definitions.

The information asymmetry proxies we employ and the exact calculation of our IA composite index follows the work of Maskara and Mullineaux (2011a); Gomes and Phillips (2010); Leary and Roberts (2010), and Chung and Zhang (2014), as well as the work of the researchers they cite (Dierkins 1991; Krishnaswami et al. 1999; Krishnaswami and Subramaniam 1999; Bharath et al. 2009). Maskara and Mullineaux (2011a) actually use a total of six information asymmetry proxies in constructing their composite index; however, we focus on the four proxies mentioned in the text (and explained in Appendix A) which represent the common ground in the aforementioned studies.

Table 4 reports information on a more limited number of time windows compared to Table 3 for brevity. Results in Table 4 are qualitatively similar (actually stronger) when using median rather than mean values. Moreover, tests in Table 4 have been re-calculated (where applicable) for all loans that have been announced by any party (by the firm or any other party), but remain qualitatively similar if repeated for loans announced by the firm only.

In the first-stage equation, we make use of a bottom-line negative earnings dummy variable while Maskara and Mullineaux (2011a) use a relevant variable for negative EBITDA, to account for the effect of bottom-line earnings on the probability that a loan is secured or not. Findings remain qualitatively similar when a size regressor (natural logarithm of market value of equity) is added in the probit model regression, in line with one alternative model specification used by Maskara and Mullineaux (2011a), and the same applies when estimating an ordered probit model, in which the dependent variable takes different values depending on whether the bank loan was announced by the firm itself, by parties other than the firm, or not announced at all (untabulated data). Results from the probit model estimation (and also overall results from Tables 5 and 6) remain qualitatively similar in case the dependent variable takes the value of one if the loan is announced by any party (rather than the firm itself) and zero otherwise (untabulated results). We chose, however, to report results when the dependent variable is equal to one when the loan is announced by the firm, in line with Maskara and Mullineaux (2011a), as such loans are expected to be more strongly related to information asymmetries. Finally, results from Tables 5 and 6 are robust to the inclusion of industry dummy variables (according to 2-digit SIC codes (untabulated results) in addition to year indicator variables for both stages of the Heckman (1979) estimation procedure.

Fields et al. (2006) find that bank loan renewal announcement returns are more likely to be positive for firms with poorer stock price performance up to the announcements, while Hadlock and James (2002) relate the level and volatility of prior year returns to the probability of a bank agreement over a public debt issue taking place. In this respect, we have repeated the analysis by including regressors for the level and volatility of market performance (well and above the volatility of operating cash flows) and (untabulated) results remain qualitatively similar.

Cash flow volatility is observed to negatively relate to changes in the TSIV for time windows around or after the loan announcement event, indicating stronger decreases in the slope of the TSIV curve for firms with lower cash flow volatility, or firms posing fewer concerns with respect to the historical variability of their operating cash flows.

A straddle is a combination of a call option and a put option on the same underlying asset, with the same maturity and strike price. If an announcement is made during the weekend or a holiday (52 announcements), the strategy is executed from the next trading day onwards.

Again, for strategy B, for all announcements made during the weekend or a holiday, the strategy is executed from the next trading day onwards.

If one were to include the ‘multiplier’, the long positions initiated would cost 100 times more, but also 100 times more would be the proceeds from the short positions. Moreover, note that by ignoring margin calls for uncovered short positions, our results ignore occasions where the investor would have to finance the maintenance of the margin account level over the 10-day horizon out of her own pocket.

By total fee, the sum of any base fee plus the per-contract fee is meant. The fee structure imposed is essentially the average from ten different brokers through which an individual investor could trade.

Phillips and Smith (1980) identify the bid-offer spread as the largest cost facing option investors. Although options exchanges only set upper limits for the bid-ask spreads, which may range from $0.25 to $1.00 per transaction (see Hull 2003), the actual quoted or effective bid-offer spreads at which investors trade equity options are below that, but still substantial. For example, in his study on the liquidity of the CBOE equity options, Vijh (1990) reports an average bid-ask spread of 21.3 cents, with less than 1.5 % of the bid-ask spreads in his sample ever exceeding 50 cents. The study by Amin and Lee (1997), that reports an average effective bid-ask spread of 16.6 cents on equity option trades surrounding earnings announcement, offers an example of an equity option trading strategy with statistically significant profits, which turn to losses once the bid-ask spreads are accounted for (see Amin and Lee 1997, Table 5 and the discussion that follows it).