Event Studies in M&A Research

The initial version of this article was written by Prof. Christopher Kummer. It has since been expanded and updated by the website owners. Credit for the original work remains with Prof. Christopher Kummer.

The merger and acquisition (M&A) announcement event study is the single most developed application of the event study methodology in corporate finance. Over a 40-year empirical record, researchers and practitioners have used it to answer one deceptively simple question: when a deal is announced, does the stock market expect it to create or destroy value, and for whom? Because share prices in reasonably efficient markets impound new public information almost immediately, the abnormal return earned by the acquirer, the target, or the combined entity in a short window around the announcement serves as a forward-looking, market-based estimate of the deal's expected wealth effect. That makes the event study the right tool for M&A: it isolates the value attributable to the transaction itself from the ordinary, market-wide and industry-wide movements that would have happened anyway.

If a combination is expected to create value, the merging companies' stocks should appreciate; if the market judges a deal value destroying, prices should fall. The simplest intuition is this: read the announcement abnormal return as the market grading the deal in real time. The target's jump is the takeover premium becoming public, and the acquirer's flat-to-negative reaction is the market's bet on whether the buyer is overpaying. The literature that operationalizes this logic converges on a robust and remarkably stable set of stylized facts, summarized next. One clarification belongs up front, because recent work has sharpened it: the announcement abnormal return measures the market's expectation of the deal's wealth effect at the moment of announcement, not the value the deal ultimately realizes. The two are related but, as the section below on what CARs do and do not measure explains, they are far from identical.

What the research shows

The empirical canon rests on a clear foundational lineage: the Jensen and Ruback (1983) survey first assembled the stylized facts, Bradley, Desai and Kim (1988) established the synergy and its division between target and bidder, and Asquith, Bruner and Mullins (1983) established the relative-size result; Andrade, Mitchell and Stafford (2001) and Betton, Eckbo and Thorburn (2008) are the modern syntheses, and Renneboog and Vansteenkiste (2019) is the most authoritative recent survey. Across thousands of deals and dozens of surveys, three headline results recur:

• Targets win big. Target shareholders earn large, positive announcement abnormal returns, on the order of +15% to +30% over a short window, with tender offers and hostile or contested deals running higher (often above 30%) than friendly mergers. Jensen and Ruback (1983) put the canonical figures at roughly +30% in tender offers and +20% in mergers; the modern handbook survey of Betton, Eckbo and Thorburn (2008) reports a public-target announcement cumulative abnormal return (CAR) of about +22% over the (-1,+1) window. On recent data, Renneboog and Vansteenkiste (2019) report a target mean of about +27.36%, positive in 91.1% of deals.

• Acquirers roughly break even. Bidder announcement CARs are, on average, statistically indistinguishable from zero and are often slightly negative around the announcement. Andrade, Mitchell and Stafford (2001) report an acquirer 3-day CAR of about -0.7% (insignificant); Betton, Eckbo and Thorburn document a bidder announcement CAR near zero with roughly half of all bidders experiencing negative reactions; Renneboog and Vansteenkiste report an acquirer mean of about -0.83%, positive in only 40.0% of deals. The dispersion around this average is enormous: the mean hides large winners and large losers. Asquith, Bruner and Mullins (1983) first showed that bidder announcement returns are significantly positive when the target is large relative to the acquirer, foreshadowing the cross-sectional structure documented below.

• The combined entity gains modestly. The value-weighted combined (bidder plus target) announcement CAR is positive and significant, in the region of +1.5% to +3.6%. Bradley, Desai and Kim (1988), the seminal synergy-division study, put the upper bracket: a successful tender offer raises the combined value of target plus acquirer by an average of about +7.4%, with the target earning roughly +32%, and they showed that across legal sub-periods (pre-1968, the Williams Act era, and the early 1980s) the synergistic gain stayed positive but its division shifted decisively toward targets as bidder competition and regulation rose. M&A creates value on average, but that value accrues overwhelmingly to target shareholders. Andrade, Mitchell and Stafford report a combined 3-day CAR of about +1.8% (with a full-sample target CAR of about +16%).

These directional facts have been confirmed repeatedly. Bruner (2002), synthesizing more than 100 scientific studies from 1971 to 2001, frames the decision-maker takeaway crisply: targets earn sizeable positive returns, bidders roughly earn their required return (value conserving on average, not value destroying), and the combined effect is positive. The practical implication is not that "M&A pays" or "M&A does not pay" in the aggregate, but that deal and firm characteristics reliably separate the winners from the losers. The best-documented of those characteristics follow.

State of the evidence in the 2020s. The directional stylized facts have held up under replication on recent data, but two canonical mechanisms have been materially qualified since 2015. First, the all-stock penalty (below) is now understood to be largely an equity-issuance financing signal rather than deal-level destruction (Golubov, Petmezas and Travlos, 2016). Second, the flat "acquirers lose" average is composition-dependent: Dessaint, Eckbo and Golubov (2025) show that the near-constant 40-year average acquirer CAR masks a rise of as much as about 5 percentage points in the common synergy component, offset by a fall in the bidder-specific component. The canon is intact in direction; its magnitudes and interpretations are not static.

Method of payment: cash beats stock

Method of payment is the dominant acquirer-side driver. Cash-financed bids earn higher acquirer and combined returns than stock-financed bids, and all-stock acquirers earn significantly negative announcement CARs. The seminal result is Travlos (1987): stock-financed bidders earn significantly negative announcement returns while cash bidders earn approximately zero. The mechanism is the equity-overvaluation signal of Myers and Majluf (1984): managers are more willing to pay with stock they believe is overvalued, so an all-stock offer is read by the market as bad news about the bidder's standalone value. Andrade, Mitchell and Stafford make the split concrete: for the no-stock subsample, the combined CAR is about +3.6% (target +20.1%, acquirer +0.4%), whereas for the stock subsample the combined CAR is only about +0.6% (target +13.0%, acquirer -1.5%). These payment subsample target figures (+20.1% and +13.0%) are higher and lower, respectively, than the +16% full-sample target CAR reported in Table 1, which pools both payment types. Mixed cash-and-stock offers typically fall between the two and, in some samples, earn relatively favourable acquirer returns, because they let the bidder signal confidence with cash while sharing valuation risk with the target. The stock penalty persists into the long run (see below).

Two important recent qualifications keep this from being a settled "cash always beats stock" story. First, the stock penalty is largely a financing signal, not deal-level destruction. Golubov, Petmezas and Travlos (2016) use seasoned-equity-offering (SEO) announcement returns plus propensity-score matching to isolate the share-price drop attributable to the equity-issuance decision itself; net of that financing-signal effect, stock-financed acquisitions are not value destructive and method of payment loses essentially all of its cross-sectional explanatory power. Second, the long-run "stock acquirers underperform" narrative is reversed under clean identification. Savor and Lu (2009) compare stock acquirers whose deals succeeded against those whose deals failed for exogenous reasons (antitrust or regulatory blocks, competing bids, target shocks): successful stock acquirers outperform the exogenously-failed group by about +13.6%, +22.2% and +31.2% (buy-and-hold) over one, two and three years, and by about +20.9%, +19.5% and +25.2% in calendar-time portfolios, with no such effect for cash bids. The interpretation is that using overvalued equity as acquisition currency creates value for long-run holders when the deal closes; the apparent stock-acquirer underperformance reflects the standalone overvaluation, not the acquisition. Savor and Lu is also the identification strategy that disciplines the endogeneity of payment method, discussed in the methodology section.

The listing effect: public versus private targets

The target's listing status is the second best-documented cross-sectional result. Acquirers tend to lose when buying public targets but gain when buying private firms or parent-controlled subsidiaries. Fuller, Netter and Stegemoller (2002) establish this cleanly with a within-bidder design (firms making five or more deals), holding bidder quality fixed: the same serial acquirers earn about +1.90% (significant) on private-target deals versus about +0.57% (insignificant) on public-target deals over the (-2,+2) window. The causal channel is a liquidity discount that the bidder captures: Officer (2007) shows unlisted targets sell at a discount, largest when the parent is liquidity constrained, so acquirers of listed targets earn about -0.38% versus about +1.48% for unlisted targets. Crucially, for private targets the usual stock penalty reverses: paying with stock can be value creating, because it aligns incentives (the private owners become monitoring blockholders) and may reflect a deferred-tax discount. The acquirer gain concentrates where private targets are intangible-heavy or financially constrained, a modern refinement of the Officer and Fuller-Netter-Stegemoller channel. The buyer's identity matters too: private-equity and other financial buyers draw a different announcement reaction from strategic (operating) acquirers, and negotiated mergers differ from tender offers in both premium and completion odds (the strategic-versus-financial distinction goes back to Healy, Palepu and Ruback, 1997).

Acquirer size: small buyers gain, large buyers lose

Moeller, Schlingemann and Stulz (2004) document a robust size effect: small acquirers earn positive announcement CARs regardless of payment method or target status, while large acquirers earn significantly less, with a large-minus-small differential of roughly -2 percentage points. The equally-weighted average therefore masks systematic value destruction by large firms, which tend to pay higher premiums.

Aggregate dollar losses and the glamour effect

The companion study, Moeller, Schlingemann and Stulz (2005), shows why percentage and dollar-weighted views can diverge sharply. Acquirers lost about $0.12 per dollar spent in 1998 to 2001 (roughly $240 billion in total, with bidder losses exceeding target gains by $134 billion), versus only about $0.016 per dollar ($7 billion) in the entire 1980s. A handful of very large deals by extremely high-valuation ("glamour") firms drives the aggregate loss. This connects to the valuation effect documented by Rau and Vermaelen (1998): low book-to-market "glamour" acquirers underperform in the long run while high book-to-market "value" acquirers do better, consistent with a hubris and extrapolation story in which the market over-credits glamour managers' acquisition ability. The same q-based pattern appears in Servaes (1991): total, target and bidder gains are larger when the target has a low Tobin's q and the bidder a high q, a clean foundational complement to the glamour-versus-value result.

Relative deal size and serial acquirers

Relative deal size amplifies the sign and magnitude of bidder returns: a large target relative to the acquirer makes a private-target deal more positive and a large public stock deal more negative. The foundational source is Asquith, Bruner and Mullins (1983), who showed that bidder announcement abnormal returns are significantly positive and rise with the relative size of the target to the acquirer; the within-bidder relative-size results of Fuller, Netter and Stegemoller echo it four decades later. Deal-design facts matter too, including toeholds, hostility, and cross-border status. Serial or program acquirers tend to see returns decline across successive deals; Moeller, Schlingemann and Stulz (2005) note that serial acquirers with extreme valuations eventually suffer losses as inorganic growth becomes unsustainable.

Runup versus markup

A substantial part of the target's total gain accrues before the official announcement, as pre-bid runup driven by leakage, anticipation, and toehold accumulation. Schwert (1996) decomposes the target's total premium into a pre-bid runup and a post-announcement markup, finds the post-announcement markup is the larger component (on the order of 60% of the total premium) while the pre-bid runup accounts for roughly 35% to 40%, and shows the two components are largely uncorrelated. The economic implication is important for both interpretation and design: runup is not a substitute for the markup but an added cost to the bidder, and a too-narrow event window will understate the target's true reaction.

Note also that the takeover premium (the offer price over a pre-bid reference) is a different number from the target's window CAR. The premium is larger: classic estimates average 40% to 50% over the price four weeks prior, and Eckbo, Malenko and Thorburn (2020) report a mean unconditional premium of about 43% (median about 37%) for listed targets over the pre-bid price. The target's window CAR is smaller than the headline premium for two reasons: part of the gain is already in the price as runup, and the market discounts the offer for the probability that the deal does not complete. Eckbo (2009) reviews the contest dynamics behind this wedge: initial bidders win only about two-thirds of control contests, rivals that enter win roughly twice as often as they appear, and toeholds and the winner's-curse adjustment shape both the premium and the bidder's runup-forced markup.

Long-run drift (and why it is contested)

A separate, more contested literature documents negative long-run post-acquisition abnormal returns, concentrated in stock-financed and glamour acquirers. Loughran and Vijh (1997) report five-year buy-and-hold abnormal returns (BHAR) of about -25.0% for stock-merger acquirers versus +61.7% for cash tender-offer acquirers, so method of payment predicts long-run as well as announcement performance. Agrawal and Jaffe (2000), surveying the long-window literature, conclude there is strong evidence of post-merger underperformance.

These long-run results should be treated cautiously, because the long-horizon design is itself the source of much of the apparent drift. The modern resolution is characteristic matching: Bessembinder and Zhang (2013) show that acquirer long-run abnormal returns (1980 to 2005) are statistically indistinguishable from zero once event firms are matched to control firms not only on size and book-to-market but also on idiosyncratic volatility, liquidity, return momentum, and capital investment, so the documented drift is largely a control-firm-matching (bad-model) artifact rather than an acquisition effect. The identification-driven counterweight to the Loughran-Vijh narrative is Savor and Lu (2009) (above): using exogenously-failed deals, successful stock acquirers actually outperform failed ones by about +22% over two years and +31% over three (BHAR), with no effect for cash, so the apparent stock-acquirer underperformance reflects pre-deal overvaluation, not the deal. Method also matters for inference: BHAR significance requires skewness-adjusted or bootstrapped test statistics (Lyon, Barber and Tsai, 1999), not ordinary t-tests, and Fama (1998) favours calendar-time (Jensen-alpha) monthly portfolios because monthly returns are less skewed and cross-sectional dependence is absorbed automatically; BHAR and calendar-time designs can give conflicting answers. Short-window announcement CARs are far more robust, which is the question the tools on this site are built to answer.

A 2010s caveat: the canon is period-dependent

The classic "acquirers lose" narrative is conditional on its sample period. Alexandridis, Antypas and Travlos (2017) show that after 2009 acquirer announcement returns turned positive and significant, stock-for-stock deals stopped destroying value, and public acquisitions turned significantly positive. Even mega-deals (at least $500 million) delivered gains to acquirers of about +$62 million around announcement, a roughly $325 million improvement on the 1990 to 2009 norm, with combined synergistic gains exceeding +$542 million; the authors link the upturn to post-crisis improvements in acquirer corporate governance. Dessaint, Eckbo and Golubov (2025) add that the near-constant 40-year average is itself a composition artifact: controlling for bidder composition, the common (synergy) component of acquirer returns has risen by as much as about 5 percentage points relative to the 1980s, offset by a fall in the average bidder-specific component, so synergies have grown but become less bidder-specific. The lesson for any analyst is methodological as much as substantive: stylized facts are sample-period-conditional, and a study run on recent data should not assume the pre-crisis pattern.

Deal anticipation: why bidders look like they break even

The most consequential recent refinement to the "acquirers break even" fact is that the near-zero bidder CAR is, in large part, a measurement artifact of anticipation. Because the market partially prices in likely acquirers and likely deals before any formal announcement, the announcement-date abnormal return captures only the surprise component, not the full wealth effect. Cai, Song and Walkling (2011) show that less-anticipated bids earn significantly higher announcement returns, and that anticipation also explains the positive reaction of an acquirer's industry rivals and the well-known decline in serial-acquirer returns: once the market expects a firm to keep buying, each subsequent deal carries less surprise. Tunyi (2021) quantifies the gap directly: acquirers in genuinely unanticipated deals earn 7-day CARs of about +5.4% to +7.5% (against roughly zero in the pooled sample), rising to about +11.6% over a wider (-20,+20) window. In other words, bidders that the market did not see coming do create measurable announcement value; the canonical zero is the average of those surprises with the deals that were already in the price. This both adds a major recent finding and supplies the mechanism behind the serial-acquirer decline noted above. It also carries a direct methodological warning for any bidder event study, discussed in the methodology section.

Behavioral drivers: hubris and CEO overconfidence

The cross-sectional patterns above (large buyers overpaying, glamour acquirers destroying value, the size effect) have a long-standing behavioral interpretation. The theoretical anchor is Roll's (1986) hubris hypothesis: even in an efficient market, bidder managers overpay because they are subject to a winner's curse and to overconfidence in their own valuations, which predicts exactly the observed combination of large target gains, bidder losses, and roughly zero combined gains. Malmendier and Tate (2008) make this measurable: CEOs classified as overconfident (from their personal option-exercise behavior) are about 65% more likely to make an acquisition, tilt toward diversifying and cash-financed deals, and draw a significantly more negative market reaction, on the order of -90 basis points versus -12 basis points for non-overconfident CEOs. Overconfidence is thus a concrete, codable cross-sectional variable for the second-stage regression, alongside the managerial-stake and excess-cash drivers discussed below.

Table 1: M&A Research Results Based on Event Studies

Author	Sample Size & Description	Sample Period	Returns to Acquiring Firm Shareholders
Megginson, Morgan and Nail (2004)	92 focus-decreasing mergers	1977-1996	Announcement period abnormal returns	Buy-and-hold abnormal returns
				Year 1	Year 2	Year 3
			-4.39%	-8.61%	-19.56%	-18.47%
Chari, Ouimet and Tesar (2010)	Sub-Sample	1988-2002	Symmetric 3-week event window around the week of the announcement
			Raw returns		Market-adjusted returns
	346 transactions involving a developed market acquirer and an emerging-market target		3.05%		2.43%
	92 transactions involving a developed market acquirer gaining majority control of an emerging-market target		5.66%		3.99%
Bouwman, Fuller and Nain (2009)		1979-2002	3-day CARs		2-year BHARs
	1090 High-market acquisitions		-0.04%	High-market minus Low-market acquisition = 1.28%	-11.32%	High-market minus Low-market acquisition = -8.04%
	1004 Low-market acquisitions		-1.31%		-3.28%
Rau and Vermaelen (1998)	212 mergers involving public targets and value acquirers	1980-1991	3-year bias-adjusted CAR = 9.87%
Loughran and Vijh (1997)		1970-1989	5-year post-acquisition BHAR
	385 stock mergers		-25.0%
	111 cash tender offers		61.7%
Conn, Cosh, Guest and Hughes (2005)		1984-1998	CARs over a symmetric 3-day event window around the announcement day
	705 acquisitions of public targets by U.K. public firms		-0.82%
	3615 acquisitions of private targets by U.K. public firms		0.86%
Moeller, Schlingemann and Stulz (2004)	Small U.S. acquirers, 3-day announcement CAR	1980-2001	Small acquirers = +2.32%
	Size effect (large minus small acquirer)		Large minus Small = -2.24%
Andrade, Mitchell and Stafford (2001)	3,688 mergers, 3-day combined / target / acquirer CAR	1973-1998	Combined +1.8% / Target +16% / Acquirer -0.7%
Betton, Eckbo and Thorburn (2008)	15,000+ control bids; public-target CAR (-1,+1) and bidder CAR	1980-2005	Target ~+22% / Bidder ~0 (worst case: large + all-stock + public = -2.2%)
Renneboog and Vansteenkiste (2019)	Modern survey; pooled acquirer and target announcement CARs	Survey (short-window announcement CARs)	Acquirer -0.83% (positive in 40.0%) / Target +27.36% (positive in 91.1%)

What announcement CARs do and do not measure

A definitive treatment has to be honest about the limits of its central measure. The announcement CAR is a market expectation: it is forward looking, fast, and replicable, which is precisely why it dominates the empirical literature (the great majority of top-journal M&A studies use the acquirer CAR as their proxy for deal quality). It is not, however, the same thing as the value the deal ends up creating.

The strongest recent statement of this caveat is Ben-David, Bhattacharya, Huang and Jacobsen (2026), who study about 47,000 acquisitions from 1980 to 2018 and find that acquirer announcement CARs are essentially uninformative about realized deal outcomes such as subsequent goodwill impairment, abnormal operating performance, and even deal completion. Observable deal and firm characteristics (relative size, method of payment, target listing status, leverage) predict five-year outcomes with a return spread of roughly 8% to 11%, whereas the CAR-implied signal delivers under 3%. A strategy of completing positive-CAR deals and withdrawing from negative-CAR ones would have produced about -5% relative five-year returns versus the opposite policy, and the dollar CAR moves six to thirteen times more with acquirer-level news than with deal-specific news.

The practical takeaways are constructive, not nihilistic. First, the announcement CAR remains the right instrument for the question event studies actually answer: how did the market revise its expectations when the deal became public? Second, it should be read as an expectation conditioned on what was already anticipated (see the deal-anticipation discussion above), not as a verdict on realized value. Third, for a question about whether a deal worked, pair the announcement CAR with characteristic-based controls and, where data allow, with longer-horizon or operating-performance evidence rather than relying on the short-window reaction alone. This is the single most important credibility upgrade for a finance-research audience, and it sharpens rather than undermines the case for running the announcement study carefully.

Other firm, deal and managerial value drivers

Beyond the headline cross-sectional results, the literature documents several additional drivers of M&A success, useful both as hypotheses to test and as controls in a cross-sectional regression.

Product-line diversification is value destroying. The relatedness of the merging firms' businesses is positively correlated with stock returns (Megginson, Morgan and Nail, 2004). Conglomerate deals earn the lowest returns, while divestitures or spin-offs that sharpen focus tend to enhance shareholder value.

Cross-border diversification is value destroying, unless the target is in an emerging market. Acquirers of domestic targets generally enjoy larger short- and long-term gains than acquirers of overseas targets, and deals that also diversify activity fare worse. Returns are higher, though, when there is little co-movement between the target's and the acquirer's home markets, and the market rewards developed-market acquirers in emerging markets (Chari, Ouimet and Tesar, 2010), especially when the deal transfers majority control in R&D-intensive or brand-intensive industries.

Managerial hubris during bull markets is value destroying. Although the market reacts more favourably to deals announced in bull markets, acquisitions initiated in bullish markets underperform on two- and three-year buy-and-hold returns (Bouwman, Fuller and Nain, 2009). Managers are more prone to hubris in upswings; bearish-market deals tend to be more cautiously motivated by realistic synergy expectations.

Acquisitions financed with excess cash do not pay. Announcement returns are negatively correlated with the acquirer's cash reserves (Harford, 1999): a cash stockpile removes the discipline of external financing, making value-destroying investment more likely.

High managerial stake is a value driver. Returns to buyers' shareholders rise with managers' equity stakes (You, Caves, Henry and Smith, 1986; Healy, Palepu and Ruback, 1997). Consistent with this, leveraged buyouts that align managerial and shareholder interests create value, and CEO ownership has a strong positive impact on long-term returns (Cosh, Guest and Hughes, 2006).

Bondholders and the wealth-transfer question

Not all of the equity announcement gain is new synergy: part of it can be a transfer between claimholders. A combination can lower the merged firm's default risk through coinsurance (two imperfectly correlated cash-flow streams), which raises existing bond values, while a leverage-increasing or risk-shifting deal can lower them. The classic evidence is mixed and modest in magnitude (Asquith and Kim, 1982; Billett, King and Mauer, 2004, who find target bondholders gain while acquirer bondholders can lose, especially in below-investment-grade deals). The implication for event-study design is direct: an equity-only study can over- or under-state true synergy when debtholder effects are material, so a complete who-gets-what accounting tracks the debt claim as well as the two equity claims.

Worked examples

The stylized facts are easier to internalize against concrete deals and a concrete calculation.

From daily abnormal returns to a CAR: a step-by-step table

The cumulative abnormal return is nothing more than the sum of the daily abnormal returns over the event window, where each daily abnormal return is the actual return minus the expected (normal) return predicted by the chosen model. Suppose a target's stock, around a takeover announcement on day 0, behaves as follows, with the expected return estimated from a market model fitted over a clean pre-event estimation window:

Table 2: Building a (-1,+1) target CAR from daily abnormal returns

Day	Actual return (R)	Expected return (market model)	Abnormal return (AR = R minus expected)
-1	+2.4%	+0.3%	+2.1%
0	+18.8%	+0.4%	+18.4%
+1	+3.7%	+0.2%	+3.5%
(-1,+1) CAR	sum of the three daily ARs		+24.0%

The arithmetic is deliberately simple: CAR(-1,+1) = +2.1% + 18.4% + 3.5% = +24.0%. The day-0 abnormal return dominates because that is when the takeover information hit the market, exactly the discrete target jump the stylized facts describe. This is precisely the construct the Abnormal Return Calculator (ARC) produces for every event in your sample: it fits the expected-return model on the estimation window, computes the daily abnormal returns, cumulates them over the event window you specify, and then runs the significance tests on the resulting CARs.

The all-stock acquirer penalty: AOL and Time Warner (announced January 10, 2000)

This is the textbook illustration of the target-wins, stock-acquirer-loses, glamour-overvaluation pattern. On announcement, target Time Warner jumped sharply (on the order of +39%) while the acquirer, AOL (trading near its dot-com-peak valuation, the archetypal glamour bidder paying with overvalued stock), reacted negatively and shed a large share of its value over the following month. The market read an all-stock offer from a richly valued buyer exactly as the Travlos and Myers-Majluf logic predicts: as a signal about the bidder's own price. It remains the canonical historical all-stock archetype.

A modern cash deal: Microsoft and Activision Blizzard (announced January 18, 2022)

For a clean contemporary counterpart that shows the cross-section in one named episode, take Microsoft's all-cash bid for Activision Blizzard at $95.00 per share (about $68.7 billion), a premium of roughly 45% over the pre-bid price of about $65. Three things happened at once, all of them textbook:

• The target jumped, but not all the way to the offer. Activision closed announcement day up about +25% (to roughly $82), still about 14% below the $95 offer. That residual gap is the merger-arbitrage spread, and it encodes the market's assessment of completion (here, regulatory) risk, the live counterpart to the completion-probability adjustment discussed in the methodology section.

• The acquirer dipped. Microsoft fell about -2% on the day, the small negative bidder reaction the canon predicts even for a strategically coherent cash deal.

• The antitrust dimension was explicit. The U.S. Federal Trade Commission moved to block the deal in December 2022; the arb spread widened and narrowed with the litigation, and the deal ultimately closed in October 2023. The episode therefore ties together the target jump, the acquirer dip, the arb-spread-as-completion-probability, and the antitrust thread covered in the practitioner section below.

Constructing the combined return: a numeric mini-example

The combined-entity CAR is a market-value-weighted average of the two firms' CARs, with weights equal to their pre-announcement market capitalizations. Suppose the bidder has a pre-announcement market cap of $10 billion and the target $2 billion, and that over the (-1,+1) window the target earns +25% and the acquirer -1%. The combined CAR is:

CAR_combined = (10 / 12) x (-1%) + (2 / 12) x (+25%) = -0.83% + 4.17% = +3.3%.

The example shows the headline arithmetic of the field at a glance: a positive combined number (value created on average) sitting on top of a negative acquirer number (value transferred to the target), with the small bidder weight on a large target gain doing the work. This is exactly the construct the Abnormal Return Calculator produces when you supply both legs of the deal.

Common misconceptions and pitfalls

Beyond academia: how practitioners use M&A event studies

The same machinery serves several applied audiences outside finance scholarship.

Securities litigation and damages

This is plausibly the single largest commercial application of disclosure and M&A event studies. In Rule 10b-5 fraud-on-the-market cases, the event study has become the court-preferred and, in the words of several courts, "almost obligatory" method for proving loss causation, materiality, price impact and damages (the dedicated Litigation use case treats this application in full). The legal architecture rests on three decisions: Basic Inc. v. Levinson (1988) established the fraud-on-the-market presumption of reliance; Dura Pharmaceuticals v. Broudo (2005) required loss causation to be proven separately (a corrective-disclosure price drop, not merely an inflated purchase price); and Halliburton Co. v. Erica P. John Fund (2014, "Halliburton II") permitted defendants to use event-study evidence of the absence of price impact to rebut the Basic presumption at class certification. Courts commonly apply a 95% statistical-significance convention to the corrective-disclosure stock-price move, and admissibility under Daubert turns on the rigor of the study. An event study that fails to show a significant move can defeat class certification by rebutting the presumption of price impact. The robustness battery this site implements (the standardized cross-sectional BMP test, the Kolari-Pynnonen cross-correlation adjustment, and non-parametric rank and sign tests) is exactly the tooling a litigation expert needs to survive cross-examination, especially given the well-documented low power of single-firm event studies.

Antitrust and competition analysis (rival-firm returns)

A horizontal merger is plausibly anticompetitive if it lets the merged firm and its rivals exercise market power, and plausibly efficiency enhancing if it threatens rivals with a tougher competitor. These two stories have opposite predictions for rivals' stock prices on announcement: market power makes rivals rise, expected efficiency gains leave them flat or falling. Eckbo (1983) is the foundational application: merging firms earned about +25% combined and rivals also rose, but rivals' returns did not fall when a merger was antitrust-challenged, which rejects the collusion hypothesis in favour of efficiency. Eckbo and Wier (1985) applied the rival-return test directly to Hart-Scott-Rodino-challenged horizontal mergers and again found little evidence the challenged deals were collusive. Regulators run these studies themselves: the FTC Bureau of Economics retailing-mergers study read positive rival reactions around May Co. / Associated Dry Goods (1986) and American Stores / Lucky Stores (1988) as possible evidence of anticompetitive effect. A definitive treatment must also state the indispensable caveat: McAfee and Williams (1988) applied an event study to a merger known ex post to be anticompetitive and found no significant rival reaction, demonstrating the low statistical power of single-deal event studies for antitrust screening (rivals are often large, high-variance firms). The method links directly to our Competitive Dynamics application.

Merger arbitrage and the deal-completion signal

The completion-probability (truncation) adjustment discussed abstractly in the methodology section has a live, tradable counterpart: the merger-arbitrage spread, which event-driven funds read off the tape every day. For a cash deal, the market-implied probability of completion is approximately the current target price divided by the offer price, so an arb spread is the inverse of the completion odds. This shortcut is valid only for cash deals: for a fixed-exchange-ratio stock deal the target price tracks the (moving) acquirer price, so the spread must be measured against the current value of the share-exchange package rather than a fixed offer, and collars are a contractual mechanism that bounds how far that package value can drift. On announcement the median target jumps about +27% and then trades at roughly a 3.5% spread to the offer; the Microsoft-Activision example above showed about a 14% spread reflecting regulatory risk. Base rates practitioners price on: about 90% of announced deals complete overall, but only about 38% of hostile bids versus about 82% of friendly ones. This is the same probability the analyst must divide out to recover the conditional (completion) deal effect from the unconditional announcement CAR.

Fairness opinions, valuation and board post-mortems

Boards, advisors and valuation experts use announcement CARs and premiums as market benchmarks: most targets and a substantial minority of acquirers obtain a fairness opinion, and the announcement reaction is a natural external check on whether the negotiated terms looked value creating to the market. Acquirers increasingly review the market's announcement reaction as one input (read, per the caveat above, as an expectation rather than a verdict) when evaluating their own programs; strategy and investor-relations teams at serial acquirers track their own running announcement CARs as a scorecard, and a declining reaction can signal that the market has stopped crediting the strategy, the anticipation effect made concrete. The same event-study machinery underlies expert-witness and consulting practice at the major economic-consulting firms, where ARC plus the significance-test battery is the standard robustness tooling for damages and competition testimony.

How to run this kind of event study

The general workflow (estimation window, expected-return model, event window, abnormal returns, significance testing) is covered in our Introduction to Event Study Methodology and the step-by-step Event Study Application Blueprint. M&A studies add several event-type-specific design choices that matter a great deal in practice.

Event-date identification is the central hazard

Use the first public announcement (or rumor) date, not the completion date. M&A announcements are notoriously leak-prone, and anchoring on the wrong date materially biases the measured reaction. Mulherin and Simsir (2015) quantify the hazard precisely: the SDC "Date Announced" is biased in about 24% of deals because an earlier merger-related event (a rumor, a search-for-buyer announcement, or an "original date announced") already moved the price; targets earn about +12.6% around the original date announced and a further about +11% around the SDC date, so a study that anchors only on the database date can miss roughly half of the target reaction. The fix is to use the earliest merger-related disclosure date as the event date; Fuller, Netter and Stegemoller (2002) hand-verified 500 SDC announcement dates and found them correct in more than 90% of cases, so verification is feasible. Our Event Date Identifier (EDI) is built to recover the original date, and News Analytics (CATA) can help screen the surrounding news flow for leakage and confounders.

Sample construction and data screens

Sample construction is itself a result-determining methodology choice, not a neutral preliminary. Netter, Stegemoller and Wintoki (2011) show that the literature's standard screens (public-target-only, deal value disclosed, a minimum deal-size threshold) systematically oversample large public deals and shift the headline acquirer CAR: in the broad unscreened SDC universe, acquirers gain in most takeovers and the merger-wave pattern attenuates. Data quality compounds the problem where inference is most fragile: Barnes, Harp and Oler (2014) find that hand-collected data is more accurate than Thomson/SDC and that SDC errors cluster in smaller, high book-to-market acquirers with weak announcement reactions, exactly the subsample that drives the size and listing effects. The practical rule: report your screens explicitly, justify the minimum-size and listing filters, and verify dates and deal flags for the small and value-tilted acquirers where measurement error most distorts cross-sectional CAR inference.

Run acquirer and target as separate samples

Acquirer and target are different event types and require different workflows. Targets show a discrete jump at announcement and are the right sample for premium and efficiency questions; acquirers show small, noisy reactions that demand tighter controls and larger samples. The combined-entity return is a separate construct: build it as a market-value-weighted two-firm portfolio of bidder and target, with weights equal to their pre-announcement market capitalizations, CAR_combined = w_A x CAR_A + w_T x CAR_T, where w_A and w_T are the bidder and target market-cap shares (see the worked example above).

Adjust for anticipation and completion probability

Two M&A-specific biases pull the measured bidder CAR toward zero, and a careful design addresses both. The first is partial anticipation: because the market prices in likely acquirers and likely deals in advance, a single announcement-date study captures only the surprise, so short-window bidder CARs systematically understate the full wealth effect for anticipated and serial acquirers (Cai, Song and Walkling, 2011; Tunyi, 2021). Where possible, include an ex-ante acquisition-probability proxy as a control, or report an unanticipated-deal subsample, so the surprise component is isolated. The second is the completion-probability (truncation) adjustment: the announcement CAR is a probability-weighted expectation, approximately P(complete) x (value conditional on completion). To recover the conditional deal effect, divide by an estimated completion probability or restrict attention to the surprise component; the observable arb spread (see the merger-arbitrage subsection) is a ready estimate of P(complete) for cash deals. Ignoring this biases the CAR toward zero, most severely for contested or heavily regulated deals, and it is part of why the announcement reaction diverges from realized outcomes.

Window choice: leakage versus confounders

Window selection trades off two opposing forces. Leakage and anticipation argue for windows that begin well before the announcement (for example a runup window such as (-42,-1), or a wider (-20,+1)) to capture the pre-bid runup documented by Schwert. Confounding events argue for the shortest possible window (the modern default is the tight 3-day (-1,+1), or (0,+1)). A common and defensible compromise is to report both a tight announcement window and a wider runup-inclusive window, and, for targets, to report the runup and markup windows separately because they measure economically distinct quantities. The older 21-day (-10,+10) window is still seen, but for clean announcement inference the short window is now standard. The estimation window should end before the runup begins (for example (-300,-60) or (-250,-46)) so leakage does not contaminate the normal-return benchmark; Dessaint, Eckbo and Golubov (2025) emphasize this point and note that the dollar CAR partly reflects acquirer-level rather than deal-specific news.

Confounders, clustering and test statistics

Confounding events are pervasive in M&A (simultaneous earnings releases, guidance, competing bids, deal revisions or withdrawals); screen each event and consider excluding clustered observations or shrinking the window. Two M&A-specific statistical pitfalls deserve attention. First, event-induced variance: announcement-day return variance rises sharply, biasing ordinary t-tests toward over-rejection. Second, cross-sectional dependence and event clustering: deals cluster by merger wave and by industry, inflating test statistics. Both call for variance-robust and dependence-robust tests: the standardized cross-sectional (BMP) test of Boehmer, Musumeci and Poulsen, the cross-correlation-robust adjustment of Kolari and Pynnonen, and non-parametric rank and sign tests. For any long-horizon design, note that the short-window battery is not enough: BHAR requires skewness-adjusted or bootstrapped tests (Lyon, Barber and Tsai, 1999) and Fama (1998) prefers calendar-time monthly portfolios. See our overview of Significance Tests for the full battery.

Expected-return model and the second stage

For short windows the market model is the default and the choice of normal-return model has little effect. For cross-sectional comparisons across acquirers tilted by size, book-to-market or momentum, and for any long-run design, prefer Fama-French three- or five-factor or the Carhart four-factor model; see Expected Return Models. The research payoff is the second stage: a cross-sectional regression of per-event CARs on deal characteristics (a payment dummy, a public/private dummy, relative deal size, acquirer size, book-to-market, toehold, hostility, cross-border), with method of payment included as the Travlos-standard baseline control. Where data permit, add behavioral and governance controls (a CEO-overconfidence measure, excess cash, the managerial equity stake) and an anticipation proxy (an ex-ante acquisition-probability estimate), since the surprise content of the deal is itself a determinant of the CAR. One caution on interpretation: method of payment is not exogenous. Cash versus stock is a bidder choice correlated with the bidder's own (over)valuation, so the payment dummy is best read as an endogenous indicator rather than a clean treatment; the clean way to discipline this endogeneity is the failed-deal natural experiment of Savor and Lu (2009), comparing completed against exogenously-failed stock bids rather than raw stock-versus-cash comparisons. Finally, distinguish the equally-weighted percentage view from the dollar-weighted aggregate (Moeller, Schlingemann and Stulz, 2005): a positive average CAR does not imply aggregate wealth creation when a few mega-deals dominate the dollar totals, so report both.

Recommended default design at a glance

For a standard acquirer or target announcement study, the following design maps one-to-one onto the input fields of the Abnormal Return Calculator and is a defensible starting point that you can then justify departing from:

• Event date: the first public announcement or rumor date, verified against news archives, not the completion date.

• Samples: run acquirers and targets separately; build the combined return as a market-cap-weighted two-firm portfolio.

• Event windows: a tight (-1,+1) (or (0,+1)) for clean announcement inference, plus a wider runup-inclusive window such as (-42,-1) or (-20,+1); for targets, report runup and markup windows separately.

• Estimation window: end it before the runup begins, for example (-300,-60) or (-250,-46), with the standard 120 to 250 trading days.

• Expected-return model: the market model for short windows (model choice barely matters there); Fama-French three- or five-factor or Carhart four-factor for size, value or momentum-tilted cross-sections and any long-run design.

• Significance tests: the standardized cross-sectional BMP test for event-induced variance, the Kolari-Pynnonen adjustment for cross-sectional dependence and clustering, plus non-parametric rank and sign tests; skewness-adjusted or bootstrapped tests and calendar-time portfolios for any long-run claim.

• Second stage: regress per-event CARs on deal and firm characteristics (payment, listing status, relative size, acquirer size, book-to-market, toehold, hostility, cross-border), with anticipation and behavioral controls where available, and report both percentage and dollar-weighted views.

Run it with our tools

The applications on this site implement the workflow above end to end:

• Abnormal Return Calculator (ARC) is the core tool for the M&A question. Supply separate acquirer and target event files (announcement date as the event date), choose your estimation and event windows, pick an expected-return model (Market Model, CAPM, Fama-French 3-factor, Fama-French 5-factor, Carhart 4-factor, or comparison-period mean), and run the full battery of more than a dozen parametric and non-parametric significance tests built to handle clustered M&A samples. ARC outputs per-event CARs that you can export directly into the second-stage cross-sectional regression on deal characteristics.

• Event Date Identifier (EDI) helps pin down the true first announcement date and detect leakage, the single most important design decision in an M&A study.

• News Analytics (CATA) screens the news flow around the deal for confounding announcements and for the tone of coverage.

• Abnormal Volume Calculator (AVC) and Abnormal Volatility Calculator (AVyC) complement the return analysis: trading-volume and volatility reactions corroborate that information actually arrived and help distinguish anticipated from surprise announcements.

Frequently asked questions

Why do acquirer stock prices often fall on a merger announcement?

On average, acquirer announcement CARs are statistically close to zero and often slightly negative (about -0.7% in Andrade, Mitchell and Stafford; about -0.83% in Renneboog and Vansteenkiste). Three forces pull the bidder reaction down: the market discounts the risk of overpaying (Roll's hubris hypothesis), an all-stock offer signals that the bidder believes its own equity is overvalued (Travlos; Myers and Majluf), and much of the wealth effect is already anticipated and in the price, so only the surprise shows up on the announcement date (Cai, Song and Walkling; Tunyi). The reaction is more positive for cash, small, private-target, and genuinely unanticipated deals.

How much do target shareholders typically gain?

Target shareholders earn large positive announcement returns, on the order of +15% to +30% over a short window, higher (often above 30%) in tender offers and hostile or contested deals. Jensen and Ruback put the canonical figures at about +30% in tender offers and +20% in mergers; Renneboog and Vansteenkiste report a modern mean of about +27.36%, positive in 91.1% of deals.

Is the takeover premium the same as the target's abnormal return?

No. The premium (offer price over a pre-bid reference, typically 40% to 50%, about 43% mean in Eckbo, Malenko and Thorburn) is larger than the target's window CAR. Two reasons: part of the gain is already in the price as pre-bid runup (roughly 35% to 40% of the premium in Schwert, the post-announcement markup being the larger component), and the market discounts the offer for the probability the deal does not complete.

What event window should I use?

For clean announcement inference, use a tight 3-day (-1,+1) (or (0,+1)). Add a wider runup-inclusive window such as (-42,-1) or (-20,+1) to capture leakage, and for targets report the runup and markup windows separately. The estimation window should end before the runup begins (for example (-250,-46)) so leakage does not contaminate the normal-return benchmark.

Do mergers create value?

On average yes, but modestly and unevenly. The combined (bidder plus target) announcement CAR is positive and significant, about +1.5% to +3.6% (up to about +7.4% for tender offers in Bradley, Desai and Kim), so M&A is value conserving to value creating in aggregate (Bruner). The value accrues overwhelmingly to target shareholders, and deal and firm characteristics, not a blanket verdict, separate the winners from the losers.

Does it matter whether a deal is paid in cash or stock?

Historically yes: all-stock bidders earned significantly negative announcement returns while cash bidders earned about zero (Travlos). Recent work qualifies this in two ways: net of the equity-issuance financing signal, stock-financed deals are not value destructive (Golubov, Petmezas and Travlos), and under clean identification successful stock acquirers actually create long-run value relative to exogenously-failed ones (Savor and Lu). The headline cash-beats-stock pattern is real but is largely a financing signal, not deal-level destruction.

Why do acquirers look like they break even on average?

Largely because of anticipation. The market partially prices in likely acquirers and likely deals before any formal announcement, so the announcement-date abnormal return captures only the surprise. Genuinely unanticipated bidders earn meaningfully positive CARs (about +5.4% to +7.5% over 7 days in Tunyi); the canonical zero is the average of those surprises with deals already in the price.

Does a positive announcement CAR mean the deal actually worked?

Not reliably. The announcement CAR is the market's expectation, not realized value. Ben-David, Bhattacharya, Huang and Jacobsen show that acquirer announcement CARs are essentially uninformative about realized five-year outcomes (goodwill impairment, operating performance, even completion); observable characteristics predict outcomes far better. Pair the CAR with characteristic controls and longer-horizon evidence to ask whether a deal worked.

Related use cases

Mergers and acquisitions are one of several corporate events analyzed with this methodology. See the closely related pages on Divestitures, Alliances and Joint Ventures, and Competitive Dynamics; the methodologically adjacent Earnings Announcements and Litigation applications; and the broader Comparative Event-Type Analyses. For the full catalogue, return to the Practical Applications overview.