Features & Departments Where's the risk in risk arbitrage?
IN PRINT ARCHIVE CIR Spring 1999
|Where's The Risk in Risk Arbitrage?|
|By Andrew Karolyi and John Shannon|
This study examines the profit potential of risk arbitrage in Canadian mergers and acquisitions (M&A). The sample comprises 37 deals valued at over $50 million that took place during 1997. The returns and risks associated with this type of investment strategy were measured relative to intrinsic value of the target companies for such deals, but also relative to the specific features of different deals. The risk and return of risk arbitrage as a strategy is particularly interesting in the Canadian setting because of the unique industrial structure of the market, the extent of cross-border dealing and the smaller size of the deals. It is well-known that resource stocks represent a disproportionately large component of the capitalization of the Canadian market. But, it is also worth noting that, according to Crosbie and Company, the two hottest sectors for Canadian M&A is utilities and oil and gas, the latter comprising 14 of 37 deals in our study. Second, of the 279 deals in the first quarter of 1998, 118 or 40 percent are cross-border deals (77 represent Canadian takeovers of foreign firms, another 41 foreign acquisitions of Canadian targets).3 Finally, while a $1 billion deal will often top the list of largest deals in any quarter in Canada, similar lists in the U.S. will top $5 to $10 billion.4 This size difference could limit the liquidity of the risk arbitrage market in Canada.
It was found that the average return to the risk arbitrage strategy yielded 4.78 percent in excess of the TSE 300 stock index. With the average duration of these risk arbitrage strategies lasting 57 days, this figure translates into an annualized excess return of 33.9 percent. The resulting abnormal returns are insensitive to various attributes of the deal, such as the average number of days to close the deal (or fail), the market capitalization or industry sector of the target company, or the means of payment for the transaction. Though readers are cautioned of the limited scope of the findings (i.e. Canada, 1997 data, deals over $50 million), they are offered as a puzzle and question why the profit potential of risk arbitrage has been underestimated to date in Canada.
Several researchers, including Dodd and Ruback and Bradley, Desai and Kim provide empirical evidence that corporate acquisitions by tender offers provide significant and positive abnormal returns to the shareholders of both the target and bidding firm.5 Bradley et al. look specifically at unsuccessful offers and show that target firms in those cases experience no wealth change unless target shareholders reject one bid and accept another made by a rival firm. In those cases, the unsuccessful bidder realized a significant wealth loss following rejection. Most of the share price reactions occur on the day of the announcement by the rival firm. For the target, if the initial bid is rejected because it is too low and if no subsequent bid materializes, a greater positive abnormal return is achieved. This initial gain, however, tends to dissipate over the next two years. Several studies have extended the findings by Dodd and Ruback and Bradley et al. to predict tender offer success from information available to the potential bidding firm before the conveyance of the tender offer.6
The most closely related article to this study is Dukes, Frohlich and Ma which examined 761 tender offers in the U.S. between 1971 and 1985.7 They determined that 82 percent of the transactions were profitable yielding average abnormal returns of 24.6 percent over 52.4 days on average (annualized yield of 171 percent). They also found that the abnormal return increases inversely with the level of success probability, which is measured by the largest price spread relative to some pre-announcement benchmark relative to the tender premium based on the offer price.
Using Crosbie and Company's database, all announced takeover bids for Canadian publicly-traded targets in 1997 were identified where the market capitalization of the target company was over $50 million.8 The average transaction represented about $572 million in capitalization with a typical duration of 57.3 days (Table 1). The largest contingent of risk arbitrage candidates originated in the oil and gas sector with 14 deals. The oil and gas transactions were typically shorter in duration (44 days) and on average smaller in size ($410 million). Deals in relatively more closely-regulated industries such as financial services, paper and forest products and communications and media take relatively longer to complete than average. Though not reported in Table 1, all but three deals (91 percent) were successful by year end. This is very close to the 89 percent success rate for tender offers in the larger Dukes et al. U.S. sample.
|RISK ARBITRAGE SPREADS|
Returns for the risk arbitrage deals are summarized in Table 1 and Figure 1, which present the mean raw and adjusted spreads and betas. The spreads are annualized using the formulas noted in the Endnotes. The betas are computed from daily returns during the pre-announcement year and up to the ten days before the announcement date itself. Note that the duration of each offerand consequently the holding period for the risk arbitrage strategyvaries, so that comparisons across deals are not straightforward. Nevertheless, the annualized equivalent returns provide some guidelines. The average risk arbitrage spread was 7.81 percent. On an annualized basis, with an average duration of 57.3 days, the spread translates into 52.25 percent. To put these results into context, the largest return was examined (20.2 percent or 115.4 percent annualized) for the Polymer takeover of Dominion Textile in December 1997 where the tender price of $12.10 per share ultimately closed at $14.55 after 64 days. This consumer products deal was not unusual for companies in that sector, as the other three deals generated the highest average spread of 13.04 percent (64.11 percent annualized). Financial services and oil and gas risk arbitrage deals followed closely averaging about 9 and 7 percent (more than 43 percent annualized), respectively. 9
Over the year, the TSE 300 index return 13.47 percent. As a result, the 52.25 percent overall annualized risk arbitrage spread would have easily cleared the market benchmark, particularly if we consider that the average pre-announcement period betas for the target stocks was 0.517. To address this benchmark issue more formally, the adjusted spreads were also computed in Table 1. These measure the spreads on the target firms from announcement date to close/failure date in excess of the TSE 300 return (without dividends reinvested) over the same horizon. On average, the adjusted spread was 4.79 percent which translates into an annualized 33.9 percent.
Figure 1 shows that the distribution of raw and adjusted spreads for M&A transactions is significantly positively skewed. In fact, only five of the 37 deals yielded zero or negative raw spreads and none declined by more than 6 percent. By contrast, eleven deals yielded more than 10 percent. For the adjusted spreads, the distribution shifts closer to zero, but the number of large positive outcomes (eight deals exceeding 10 percent) exceeds the large negative outcomes (-8.55 percent is lowest adjusted spread for the Ranger Oil takeover of Elan Energy).
|WHERE'S THE RISK?|
Several types of comparisons were used to assess the risk involved in the business of risk arbitrage. First, the annualized return from a conservative buy-and-hold strategy was evaluated by benchmarking the risk arbitrage spreads with the return on the TSE 300 index. These results were presented in Table 1 and Figure 1. This naïve benchmark captures relative performance, but, of course, ignores the risk exposures from investing in these target companies.
A second method used to compare risk and return of different investments is to compute the risk-adjusted excess return. This is usually done by first estimating the systematic risk, or beta, of the investment and then subtracting the corresponding portfolio of the market (TSE 300 index) return. A positive excess return would suggest that there is profit beyond the compensation justified by the underlying risk. As Dukes et al. point out, the concept of adjusting for systematic risk is less relevant in the case of M&A deals because the target firms usually experience significant reorganization or restructuring in response to tender offers. As a result, the risk structure of the firm may change dramatically around the time of the tender offer. While the estimate of risk using conventional statistical procedures may be biased, the magnitude of the changes may still provide clues on relative risk. For each deal, the pre-announcement period beta for the target firm was computed using daily returns in the year before announcement and an "in play" beta using the daily returns from 10 days following the announcement to the close of the deal. The 10 days before and after the announcement date were excluded in order to minimize the influence of the unusually large returns during the announcement period. Table 1 shows that the pre-announcement betas of the targets averaged 0.517, much lower than expected for the risky nature of the deals in which the companies are involved.10 In unreported results, the "in play" betas were even lower with an average of 0.393. Some of the most dramatic declines occur for those sectors that experience the highest risk arbitrage spreads, including consumer products and paper and forest products. Only for one sector, financial services with its four deals, did the in-play beta increase relative to its pre-announcement beta, and this is driven primarily by an outlier in the Investors Group takeover of Quorum Growth in May 1997. This comparison is problematic not only because of measurement problems, but also because during the "in-play" period especially for all-cash offers the movement of the market is unlikely to have any impact on the price of the target firm. The betas are more likely to reflect non-market specific influences, such as new bidders, changes in the bid price or an increase in the probability that the new bid will be successful.11
Though problems inherent in these risk measures must be acknowledged, risk arbitrageurs surprisingly appear to benefit not only from spreads that outperform a buy-and-hold, but also from changing risks associated with these strategies. To capture both effects together, Figure 2 shows the cumulative abnormal returns to the risk arbitrage strategy in event time from 50 days prior to the announcement date until the deal closed of failed.12 The cumulative pre-announcement returns--which the arbitrageur would not accrue--represent about 5 percent up to the announcement date. The most significant single-day jump of 4 percent occurs on the announcement date and by Day 50 following the announcement the abnormal return cumulates to a total of 17 percent. The risk arbitrageur takes advantage of the cumulative returns from the close of the announcement date only and would have achieved on a risk-adjusted basis 5.95% over 50 days, or 43.4 percent annualized.
The third method used to incorporate the risks of the risk arbitrage strategy focuses on deal-specific elements. The deal element consists of factors such as the expected time to close, necessary regulatory approvals, strategic fit of the two entities, conditions of the agreement, payment mechanisms (cash versus stock or mixed offers), and the financial health of the bidder and target. These factors are distinct from value-specific elements which relate to the intrinsic valuation of the target, such as other recent comparable transactions, potential competing bidders, the company's ultimate worth relative to the current bid price.13 To evaluate these deal-specific factors, multivariate regression analysis of the raw and adjusted spreads was performed against a series of variables which relate to deal-specific factors. These factors include the number of days to close, the pre-announcement betas, the pre-announcement (2-week and 12-month prior) stock price run-up and an oil and gas sector dummy variable. Days to close is a potential proxy for the likelihood of success as it may signal the impact of subsequent offers from competing bidders, which are not double counted in the sample.14 Also included are value-relevant risk attributes of the target as control variables, including the capitalization of the target, and price-to-sales and price-to-book ratios based on the share price two weeks prior to the announcement.
Table 2 presents the summary statistics from the regression analysis presented separately for the raw and adjusted non-annualized spreads. In the first regression using raw spreads, it was found that the market capitalization variable is marginally significant with a positive coefficient of 0.051. That spreads are positively related to size runs counter to our priors from the size effect literature which suggests that the smaller, less liquid deals would be riskier and should command a higher risk premium. Another variable that is significantly negatively related to raw spreads is the 2-week pre-announcement run up. That is, the profits that risk arbitrage yield are smaller if there is a larger run-up immediately before the announcement. Schwert 15 suggests that the relationship between the run-up and the "mark-up" (what we call risk arbitrage spreads) may be a reflection of the total control premium paid by the acquirer. To the extent that a large run-up reflects either trading by insiders or toehold acquisitions by potential bidders, he predicts that the mark-up does not have to be as large to achieve control, so that the mark-up and run-up would be substitutable. While he finds little evidence of correlation between markups and run-ups, our evidence is consistent with his prediction. Interestingly, while we would expect the duration of the deal to best represent the indirect proxy for the risk of deal failure (based on previous studies), the coefficient is positive, but not significant statistically. Overall, the intercept coefficient is significantly positive indicating the robustness of the risk arbitrage spreads to these deal-specific attributes. The coefficient of determination (or R2) for the regression is 30.9 percent, or 7.96 percent adjusted for the number of variables in the regression.
|The second regression repeats the analysis for the adjusted spreads (non-annualized). In this case, none of the deal-specific variables are statistically significant, including the market capitalization and the pre-announcement run-up, though the signs of the coefficients are similar. The days to close variable has a larger coefficient than with the raw spreads, but it is still insignificant. The intercept coefficient for the adjusted spreads regression is lower, as expected, but it is also still significant. The R2 drops dramatically to 13.5 percent or -1.5 percent adjusting for the number of variables in the regression, which indicates that the explanatory power of these deal-specific factors is poor.|
The evidence from this study has several interesting implications. It is possible for average investors in Canada to participate in risk arbitrage, long considered the domain of more informed, skilled traders. Such investors would have earned significantly higher than normal profits in the process of risk arbitrage. Most merger and acquisition studies show that shareholders do not necessarily lose the incremental wealth accumulated between announcement of the acquisition and the resolution date, even if the tender offer fails. This is the key factor in the significant spreads in risk arbitrage and the findings of this study support this argument. Not only do risk arbitrage investors earn higher returns than for a conservative buy-and-hold strategy, but also the magnitude of their excess returns are insensitive to a number of deal-specific attributes, such as the number of days to close, payment method, size of the deal and the pre-announcement share price run-up.
It remains to be seen if these risk-return patterns in the M&A market in Canada in 1997 are an anomaly or an exception that proves the rule. One indisputable fact is that the opportunities to evaluate such a strategy will continue as the M&A market of the 1990s continues to blossom in Canada.
|Andrew Karolyi is a professor, Fisher College of Business, Ohio State University in Columbus and John Shannon works on the equity derivatives desk, Nesbitt Burns Inc., in Toronto|
This research was prepared as an independent study (Business 699) by John Shannon, MBA 1998 at the Rischard Ivey School of Business, University of Western Ontario, under the supervision of Professor Andrew Karolyi. The views expressed in this study are solely those of the authors and do not necessarily represent opinions of Nesbitt Burns, Inc. We are grateful to comments of Craig Dunbar, Steve Foerster, Bruce Langstaff (Bunting Warburg) and especially Paul Halpern. All remaining errors are our own.
1. Crosbie and Company (Toronto Globe & Mail, March 16, 1998) report that in the first quarter of 1998, 279 deals worth $42 billion were announced, a 75 percent increase from the first quarter of 1997.
2. Seminal studies of this phenomenon include M. Bradley, A. Desai and E. H. Kim, 1983 "The rationale behind interfirm tender offers: information or synergy," Journal of Financial Economics, pp. 183-206, and P. Dodd and R. Ruback, 1977, "Tender offers and stockholder returns: an empirical analysis," Journal of Financial Economics, pp. 351-374. A useful summary is found in G. Wyser-pratte, 1982, Risk Arbitrage II, (new York University's Salomon Brothers Center, New York), and F. Weston, K. Chung and J. Siu, 1998, Takeovers, Restructuring and Corporate Governance (Prentice Hall, New York). A recent contribution by F. Cornelli and D. Li, "Risk Arbitrage in Takeovers" (University of Michigan working paper, 1998) models the informational asymmetry between management and arbitrageurs in the takeover process.
3. See Globe & Mail, March 16, 1998, ibid.
4. Fidelity's Merger Fund manages about $300 million in assets for risk arbitrage. See Barron's "Mutual Funds World's Safest Stock Fund," September 18, 1995).
5. See Bradley et al. (1983, ibid.) and Dodd and Ruback (1977, ibid.).
6. See R. Walkling, "Predicting tender offer success: A logistic analysis," Journal of Financial and Quantitative Analysis, pp. 461-478, W. Samuelson and L. Rosenthal, 1986, "Price movements as indicators of tender offer success" Journal of Finance, pp. 481-499, and K. Brown and M. Raymond, 1986, "Risk arbitrage and the predication of successful corporate takeovers," Financial Management, pp. 54-66.
7. W. Dukes, C. Frohlich and C. Ma, 1992, "Risk arbitrage in tender offers," Journal of Portfolio Management, Summer, pp. 47-55.
8. We limited our sample only to those deals that were both announced in and completed/failed in 1997. Further, following Dukes et al. (1992), we eliminated from our sample any transactions with a higher degree of complexity in locking in a spread on the announcement date, such as the existence of a "boot" in a tender offer. A "boot" involves some security exchanges at an undetermined stock ratio and which would require the arbitrageur to make a valuation estimate on the bidder's stock.We assumed that the best price that the arbitraguer could acquire the stock at was the closing price on the day of the announcement or the closing price on the day after announcement, if the announcement was made after the market close. The end price received by the arbitrageur is the offer price or the price at the end of the trading day that the companny announces publicly that the takeover bid has failed. If the tender offer is successful, the spread is computed as,
SPREADi = (Pioffer - Piannounce)/ Piannounce x 365/daysi,
where Pioffer is the tender offer price for the ith offer, Piannounce is the announcement date price for the ith offer, and daysi is the duration of the deal in days. If the tender offer fails, the return is computed as,
SPREADi = (Piclose - Piannounce)/ Piannounce x 365/daysi,
where Pi>close is the actual market closing price on the expiration date of the failed offer. The latter return is computed assuming that the arbitrageur sells the shares back to the market when the offer is not accepted. We also compute adjusted spreads by subtracting the TSE return (without dividends reinvested) matched over the same deal duration period. The adjusted spreads are annualized in the same way.
9. Cambridge's takeover of Markborough in June 1997 was the sole real estate deal and it generated a 23.37 percent spread over a 42 day period (203% annualized).
10. The average betas of these companies are likely low because of the "intervaling" effect in betas due to non-trading factors.
11. I thank Paul Halpern and Bruce Langstaff for clarifying this point.
12. We compute the raw daily returns for each target stock and subtract the TSE 300 index return adjusted by the company's pre-announcement beta. These abnormal returns are averaged across all 37 target companies in event time (with the announcement date equal to Day 0) and cumulated from Day -50 to Day +50 or the close/fail date, whichever comes first.
13. Wyser-Pratte (1982, ibid., Chapter 1) defines this dichotomy of deal and value elements.
14. For example, London Insurance Group was acquired by Great-West Life Company at $34 after 140 days, but only after Royal Bank initiated with the first bid at $28. We thank Paul Halpern and Bruce Langstaff for clarifying this important point.
15. W. Schwert (1996) "Markup pricing in mergers and acquisitions," Journal of Financial Economics, pp. 153-192, shows that the pre-bid runup and the post-announcement increase in the target's stock price his reference to 'markup' are generally uncorrelated. n