Merger arbitrage backtest: methodology for announced takeovers
The merger arbitrage backtest calculates what betting on announced takeovers of listed US companies since 2001 would have earned: buy the target after the news, sell the acquirer short in stock deals, and hold until the deal closes or fails. This methodology describes what the code behind the Merger arbitrage page does.
List of deals from SEC filings
A takeover target has to file documents with the SEC: for a merger the definitive proxy statement for the shareholder vote (Form DEFM14A), for a tender offer the recommendation statement on the offer (Form SC 14D9). The nightly run searches both forms for each quarter since 2001 in the SEC full-text search, without amendments (“/A”), taking each company's first filing of the quarter. From the text it reads:
- Consideration per share: cash (“right to receive $… per share in cash”), acquirer stock (“… shares of … common stock”), or both; if there are several hits, the most frequent value counts. With stock it records the acquirer's name.
- Acquirer's proxy: acquirers also call a vote when they issue new shares. If the text names the filing company as the one whose shares are given as consideration, it is not a target and drops out. Blank-check companies (industry code 6770) drop out as well: their vote is about buying a company, not a takeover.
- Agreement date: from sentences such as “On March 28, 2019, … entered into an Agreement and Plan of Merger”.
- Ticker: from the text (“listed on the NYSE under the symbol …”), each occurrence assigned to the nearest name before it (target or acquirer). If missing, from the press release on the day of the news (“(NYSE: …)”). All tickers found remain candidates; the prices decide which one is right.
Announcement and outcome
The company's list of all SEC filings provides the date and time of the news and the outcome:
- Announcement: the company's first filing on or up to 7 days after the agreement date (8-K, proxy communication, tender offer communication, Rule 425 communication). If it was received after 16:00 New York time, the first day after the news is the next trading day. If the agreement date is missing from the text, the last filing on a material agreement (8-K Item 1.01) in the 200 days before the proxy counts.
- Completion: the target is delisted (Form 25 or 25-NSE) or terminates its registration (15-12B, 15-12G). The completion date is the filing on the closing (8-K with Item 2.01, 3.01, or 5.01) in the 10 days before, otherwise the delisting date.
- Failure: termination of the agreement (8-K with Item 1.02 without closing) before a delisting. Without such a filing, a deal counts as failed if the company is still filing annual or quarterly reports more than 15 months later; then the date is missing, and the deal drops out of the calculation (the page counts it).
- Pending: deals without an outcome only count once it is known. The nightly run checks them again every 30 days.
Prices and position
For each deal with a known outcome, the nightly run loads the prices of the target and, in stock deals, the acquirer, and stores only derived values. A ticker matches the target if its price series begins before the news, the price at entry is between 0.5 and 1.25 times the offer, and, for a completed deal, the series ends at most 30 days after it. The acquirer must have prices until after the outcome. If none matches, the deal drops out with a reason.
The position is one share of the target; in stock deals the acquirer is also sold short at the exchange ratio. Its value relative to the target's price at entry:
- growth factor of target and acquirer since entry, including distributions
- price of acquirer and target at entry
- short sale as a share of the amount invested
At completion, the offer (cash plus ratio times the acquirer's price on that day) replaces the target's last price. On failure, the bet ends at the first close after the filing. The short seller pays the acquirer's distributions; this is included in the growth factor g_A.
- Entry: first close after the news (the jump on the day of the news is not part of the arbitrage), or 5 or 20 trading days later.
- Exit: at the outcome or after at most one year (250 trading days) at the market price.
- Costs: a percentage per purchase and sale, also on the short sale (default 0.1%); at completion there is no sale, because the offer is paid out or exchanged. Plus an annual borrow fee on the part sold short (default 0.5%).
- Annualized: (1 + return) to the power of 250 divided by the trading days held, at least five days.
Implied probability and filters
The spread between price and offer shows how confident the market is that the deal will close. In a simplified view, the price falls back to its level before the news if the deal fails; then:
- target's price at entry
- close before the news
- value of the offer at entry
The page caps p at 0 to 100% and compares it in bands (below 50%, 50 to 70, 70 to 80, 80 to 90, 90 to 95, 95% and above) with the share that actually closed. If the actual share is above the implied one, the market was too cautious, and the bet paid off on average.
Filters only use information at announcement: announcement period, consideration (cash, stock, mixed), type (merger or tender offer), deal value (offer times shares outstanding according to the last figure before the news), premium over the price before the news, spread at entry, target sector, and P/E and value to revenue at the offer from the figures of the last fiscal year at the SEC. The result shows the same deals by group, with the market's expectation versus actual completion; groups with fewer than five deals are left out.
The portfolio of all bets weights all open bets equally each month, without costs; months without a bet are held in cash without interest. The beta to the market is the slope of the monthly returns against the S&P 500.
What research shows about merger arbitrage
Mitchell and Pulvino (2001) found over several decades of takeovers a return profile like selling a put option on the market: in calm and rising markets no relationship with the market, in sharply falling markets a close one, because more deals fail then. Baker and Savaşoglu (2002) treat merger arbitrage as a case of limited arbitrage: the risk of individual deals is hard to diversify, and arbitrageurs have limited capital, so spreads can stay wider than the risk of failure alone justifies. Both can be checked in the backtest: in the portfolio's beta and maximum drawdown and in the comparison of implied and actual probability.
Limitations
- Only targets listed in the US and filing with the SEC; takeovers outside the US are missing.
- Text instead of a database: consideration and dates come from the text of the filings. Elections (cash or stock), ratio adjustments, contingent value rights (CVR), and raised offers are not captured by the backtest; matching against prices catches gross errors.
- Competing bids: in a bidding war, the proxy can name an offer that is later topped. The outcome then counts, while the consideration stays the one from the filing.
- Outcome without a date: failed deals without a termination filing are missing; these tend to be less prominent cases.
- Simplified hedge: short sale without recall and with a fixed borrow fee, no costs for capital and collateral.
Frequently asked questions
What is the spread in a takeover?
The gap between the offer and the target's current price, usually as a percentage of the price. It is the profit if the deal closes as announced and the price for the risk that it fails.
Why is the acquirer sold short in stock deals?
Because the value of the offer then moves with the acquirer's price. If you buy the target and sell the acquirer short at the exchange ratio, you hedge this fluctuation and keep only the spread and the risk of failure.
How does the backtest measure whether a deal failed?
By the termination of the agreement in an SEC filing (8-K Item 1.02 without closing). If the target is delisted afterward, the deal counts as completed.
What does implied probability mean?
The probability of completion suggested by the target's price, assuming it falls back to its level before the news if the deal fails. It is an approximation: often the price does not fall back all the way, or another bidder comes along.
Why is the annualized return so much higher than the return per deal?
Because many deals close after a few months. A spread of 3% in three months corresponds to about 12% annualized. Across all deals, the portfolio of all bets evens this out again, because the money is not always invested.
Literature
- Baker, Malcolm; Savaşoglu, Serkan (2002): Limited arbitrage in mergers and acquisitions. Journal of Financial Economics 64 (1), pp. 91–115. doi:10.1016/s0304-405x(02)00072-7
- Mitchell, Mark; Pulvino, Todd (2001): Characteristics of Risk and Return in Risk Arbitrage. The Journal of Finance 56 (6), pp. 2135–2175. doi:10.1111/0022-1082.00401
This page describes how Portfolio Lab calculates; past results are not a forecast. All terms in the glossary.