Alliances and Joint-Ventures

A strategic alliance or joint venture is a voluntary, dated, and largely unanticipated decision by two or more firms to cooperate: to exchange, share, or co-develop products, technologies, or services. Because the announcement of such a deal is a discrete piece of public information, and because (unlike a merger or acquisition) it transfers no control and displaces no management, the event study method (ESM) is the natural and unusually clean tool for measuring how capital markets value the decision. It isolates the abnormal stock return attributable to the announcement from normal market-wide movements, and so quantifies whether investors judge the cooperation to create or destroy value for each partner. The stakes are not marginal: Anand and Khanna (2000) estimate that alliances account for roughly 6% to 16% of the total market capitalization of US firms, so the value created or destroyed at these announcements is a first-order economic question.

Following Gulati (1998: 293), strategic alliances include all "voluntary agreements between firms involving exchange, sharing, or co-development of products, technologies, or services. They can occur as a result of a wide range of motives and goals, take a variety of forms, and occur across vertical and horizontal boundaries." It is useful to distinguish two broad governance forms, because they differ in commitment, information content, and typical market reaction:

• Equity joint ventures (JVs): the partners create a separate, jointly owned entity and contribute equity to it. The ownership stake is a credible, observable commitment.
• Non-equity (contractual) alliances: the partners cooperate through a contract (co-development, co-marketing, licensing, supply, or distribution) without forming a shared entity.

Across both forms, the recurring empirical finding is a small but typically positive average announcement return to the participating firms, with the interesting story being the cross-sectional heterogeneity: which deals, partners, and industries create the most value. The aggregate evidence is best described as mixed in magnitude (estimates vary with sample, industry, period, and whether returns are equally or value weighted), but the cross-sectional drivers are remarkably stable across studies. One intuition organizes the whole literature: the announcement return is the market's estimate of the deal's net present value to each partner, divided by that partner's market capitalization, which is exactly why a deal that barely registers for a giant can be transformational for a small partner. The magnitudes, drivers, and methodological subtleties are summarized below.

This use case has not stood still since the 1980s manufacturing and information-technology joint ventures that anchor the seminal studies. The dominant present-day form is the biopharma co-development and out-licensing deal: a large originator partners with a smaller, pipeline-dependent firm, often through milestone payments, an equity stake in the partner, or an option to acquire. These structures sit between a pure contractual alliance and a full equity JV, and they are exactly the small-partner, high-technology-content deals the literature predicts will move the smaller partner sharply. A reader who follows current deal flow (the 2025 to 2026 wave of large biopharma collaborations is a vivid example) is reading the same event the method was built to price.

The basic logic is a two-step design. In a first step, abnormal returns around the alliance announcement date are estimated for each publicly traded partner. In a second step, these abnormal returns are regressed on industry-, firm-, and deal-level explanatory variables to learn why some firms profit more than others.

What the research shows

Headline findings at a glance

The table below collects the canonical short-window results so the central pattern is visible in one view: a tight cluster of small positive abnormal returns, with the action in the cross-section. Every magnitude is labelled with its exact event window, because mixing a one-day figure with a two-day figure unlabelled is the single most common way these numbers get misquoted.

Study	Sample / period	Window	Avg abnormal return	Key cross-sectional driver
McConnell & Nantell (1985)	136 firms, 210 JVs; 1972-1979	[-1,0] (two-day)	+0.73% (z=4.10)	Smaller partner +1.10% vs larger +0.63%
Chan et al. (1997)	345 alliances; 1983-1992	Day 0 (announcement day)	+0.64%	No wealth transfer; technical > non-technical
Koh & Venkatraman (1991)	239 IT firms, 175 JVs; 1972-1986	Short window	Equity JV significant; licensing / marketing / supply not significant	Governance form drives value
Das, Sen & Sengupta (1998)	119 alliances; 1987-1991	Short window	Weak overall; technology > marketing (null)	Less-profitable / smaller firms gain more
Anand & Khanna (2000)	>2,000 JVs & licensing; SDC 1990-1993	Announcement window	Rises with experience (JVs only)	Learning curve present for JVs, absent for licensing
Kale, Dyer & Singh (2002)	US firms; dedicated alliance function	Announcement window	+1.35% (with function) vs +0.18% (without)	Dedicated alliance capability
Koval et al. (2024)	27,512 firm-alliance obs; 1990-2020	[-1,0] / [-1,+1]	+0.62% / +0.69% (t>13)	Technology effect conditional on firm size / R&D

A small, reliably positive announcement reaction

The alliance and JV event study literature is one of the oldest and most robust applications of the method in strategy and finance, and its central stylized fact is consistent: the representative alliance is value creating, not zero-sum. The seminal joint venture study, McConnell and Nantell (1985), examines 136 firms across 210 joint ventures and documents a significantly positive two-day [-1,0] announcement abnormal return of about +0.73% (z=4.10, 67% positive). To read that figure plainly: the average parent firm's stock outperformed its market-model benchmark by about three-quarters of one percent over the two days around the announcement. Crucially, because the parents' managements stay intact, JVs isolate the synergy hypothesis from the management-displacement hypothesis as the source of combination gains: the wealth effect must come from expected synergy, not from disciplining or replacing managers. Chan, Kensinger, Keown and Martin (1997), studying 345 strategic alliances over 1983 to 1992, find a significant positive abnormal return of about +0.64% on the announcement day (day 0; wider windows are sometimes quoted higher, so it is worth labelling the figure by window), and, importantly, find no evidence of wealth transfer between partners: gains are shared synergy rather than one partner expropriating the other.

These magnitudes are representative. Short-window cumulative average abnormal returns (CAARs) to the average partner typically cluster around +0.5% to +1.0%, and sector samples range higher (technology and biotechnology deals can reach several percent for the smaller, pipeline-dependent partner). The reaction is concentrated in a tight window (for example [-1,0] or [-1,+1]), is incorporated quickly, and shows little robust long-run drift, consistent with roughly efficient pricing of the announcement.

Recent and meta evidence (2010 onward): the result has held up

Almost all of the canonical anchors above are pre-2010, which invites the obvious question of whether the effect is an artifact of older, smaller samples. It is not. The headline modern replication is Koval, Zaefarian and Iurkov (2024), the largest alliance event study run to date: 27,512 firm-alliance observations across 3,936 unique US firms over 1990 to 2020. They report pooled CAARs of +0.62% over [-1,0], +0.69% over [-1,+1], and +0.67% over [-1,+2], all with t-statistics above 13, and the market model and a four-factor model give near-identical results. These magnitudes are almost indistinguishable from the seminal figures, which makes the cross-era line striking: +0.73% in 1985, +0.64% in 1997, +0.62% to +0.69% in 2024. The small positive announcement effect has survived four decades and tens of thousands of deals. This is a documented replication, not a repeated assertion.

Modern large-sample evidence also sharpens the small-partner story with hard numbers. In biopharma, alliance and partnership announcements earn CARs on the order of +1.8% over [-1,0] and [-1,+1], materially above the cross-industry average, concentrated in emerging and mid-stage biotech rather than in mature big pharma. Cho, Singh and Lo (2024), analyzing 503,107 news releases for 1,012 biopharma firms over 2000 to 2022, confirm that biotech and small-cap firms show substantially larger abnormal-return magnitudes than pharma and large-cap firms, and that collaboration, partnership, and JV news produces positive but modest reactions that rank below acquisitions and FDA-approval news. That ranking is exactly what the page's framing predicts: an alliance transfers no control and is a weaker, more reversible signal than a completed acquisition. On the learning side, the single-study evidence is now backed by meta-analysis: Wang, Jiang and Dong (2022) pool 143 empirical studies and find the alliance-experience-to-performance link is positive but contingent on the type of experience and the outcome measure, which is precisely why individual studies disagreed on magnitude.

Technology beats marketing, but the ranking is context-dependent

The type of cooperation conditions the market reaction. Das, Sen and Sengupta (1998), studying 119 alliances over 1987 to 1991, find only weak support for a positive average reaction across all alliances, but a robust cross-sectional result: technological alliances earn significantly higher abnormal returns than marketing alliances, because greater knowledge content and contractual ambiguity carry larger expected value. The same pattern appears in Chan et al. (1997): for horizontal alliances (partners in the same broad industry), deals that pool or transfer technical knowledge create significantly more value than non-technical (marketing or distribution) deals. Koh and Venkatraman (1991), examining 175 joint ventures among 239 information-technology firms over 1972 to 1986, were among the first to document that equity JV announcements raise market value, and they supply the cleanest governance-form result in the literature: equity JVs and technology-exchange agreements produced significant positive reactions, while licensing, marketing, and supply agreements showed no significant market reaction. The form of cooperation, and not only its existence, drives value.

That said, the technology-over-marketing ranking is real but not universal, and a definitive treatment should say so. Koval et al. (2024) show the technology premium is interactive: technological alliances create value for small and R&D-intensive firms but are value-neutral, or even value-destroying, for large firms, and smaller firms gain more specifically from JV-structured deals while high leverage can turn broad-scope alliances value-destroying. The ordering can also reverse by market context: in Korean evidence (Journal of Business Research, 2013), non-technological marketing alliances raised firm value more than technological alliances, even as the inverse size effect held. The defensible claim is therefore conditional: technology and R&D content typically carries the larger reaction, especially for small, research-intensive partners, but firm size, R&D intensity, leverage, and national context all move the result.

Partner characteristics: the small-partner effect

The single most durable cross-sectional fact is the relative-size effect first documented in McConnell and Nantell (1985), and its precise shape is more interesting than the usual shorthand. The smaller partner earns a much larger percentage abnormal return (+1.10%, z=2.77) than the larger partner (+0.63%, z=2.30). In dollars, however, the gains are comparable and in fact slightly larger for the bigger partner: an average abnormal dollar gain of about $4.54M for the small partner versus $6.65M for the large partner. The two facts are the mirror image of each other, and both follow from the same arithmetic: percentage return equals dollar gain divided by market capitalization, so a similar dollar synergy is a big percentage for the small partner and a rounding error for the giant. McConnell and Nantell also normalize the gain by resources committed to the venture, yielding a scaled premium (Q) of roughly 0.23 to 0.54, squarely in the range documented for merger and tender-offer premiums. This single ratio is the cleanest way to resolve the dollar-versus-percent ambiguity: it asks how much value each partner captured per dollar it put at risk.

The practical reporting rule follows directly: scale abnormal returns by firm value, report the scaled premium where resources committed are observable, and analyze partners separately rather than pooling them. Das, Sen and Sengupta (1998) reinforce the pattern from a different angle: less-profitable (and smaller) firms gain more than more-profitable firms, and for marketing alliances the profitability coefficient is negative, meaning the market reads a strong firm's marketing alliance as comparatively weak news.

Equity stakes as credible commitment

Governance form is not just a classification variable; it is a signal. The page's organizing intuition is a commitment ladder: more binding, less reversible structures carry more information and tend to move prices more. Cho, Singh and Lo (2024) document exactly this ordering in biopharma, where the average reaction rises from joint venture to partnership to merger to acquisition. The cleanest direct evidence for the equity mechanism is Allen and Phillips (2000): corporate block-equity purchases that are accompanied by an alliance, joint venture, or other product-market agreement generate significantly larger excess returns to target shareholders than equity blocks without such an agreement, and the targets subsequently raise investment and operating cash flow relative to peers. An equity stake bundled with cooperation is read as a credible, costly commitment, which is why equity JVs typically show larger and cleaner reactions than loose contractual alliances (Koh and Venkatraman, 1991). In modern deal flow the same logic appears as milestone-plus-equity collaborations and option-to-acquire structures, where an accompanying corporate-venture investment is itself read as a credibility signal.

Learning and alliance capability

Firms appear to get better at creating value through alliances. Anand and Khanna (2000), analyzing more than 2,000 joint ventures and licensing deals, document a clear learning curve: announcement value rises with a firm's accumulated alliance experience, the effect is present for joint ventures but not for licensing, and it is strongest for research JVs and weakest for marketing JVs, exactly where contractual ambiguity (and thus the scope for learning) is greatest. Kale, Dyer and Singh (2002) operationalize this organizationally: firms with a dedicated alliance-management function earn substantially higher abnormal stock-market returns at alliance announcement (on the order of +1.35%, versus about +0.18% for firms without such a function) and report greater long-term alliance success (roughly 63% versus 50%). As noted above, the meta-analytic evidence (Wang, Jiang and Dong, 2022) confirms the experience-performance link is positive but contingent, so the learning effect should be reported as real but moderated rather than as a uniform constant.

The marketing-science branch extends this from firm-level capability to the firm's whole network position. Swaminathan and Moorman (2009), studying 230 marketing-alliance announcements in the software industry, find significant positive announcement-window returns whose size depends on the announcing firm's alliance network: network efficiency and density help most at moderate levels, and a firm's marketing-alliance capability has a positive effect, while network reputation and centrality have no significant effect. The lesson for an alliance event study is that the right cross-sectional moderators may include not just this deal and these two partners, but the structure of each partner's accumulated alliance portfolio.

Relatedness, information asymmetry, and industry spillovers

Reactions are larger where information problems are more severe. Reuer and Koza (2000) show that JV announcement abnormal returns are larger where adverse-selection and information-asymmetry problems are greater (for example international and cross-industry JVs), supporting an information-economics rationale for when JVs create value. Announcements also reprice firms beyond the partners themselves: Oxley, Sampson and Silverman (2009) show that an alliance moves the valuation of non-participating rivals, who can gain or lose depending on whether the deal signals industry-wide cooperation ("detente") or an intensifying competitive threat ("arms race"). This industry-level spillover is both a substantive finding and a methodological warning: the "control" set in an alliance study is not automatically clean.

Extension to financial services and bondholders

The findings generalize beyond manufacturing and technology. Amici, Fiordelisi, Masala, Ricci and Sist (2013), studying US and European banking over 1999 to 2009, find that value creation is concentrated in JVs (rather than looser alliances), is larger when the partner is a non-bank financial firm, and is larger for deals enabling geographic (cross-border) expansion, confirming the cross-sector robustness of the JV-structure and diversification effects. The wealth gain is also broad across claimants rather than a transfer: joint ventures and alliances create modest positive value for bondholders as well as shareholders, indicating genuine firm-value creation rather than risk shifting. The method travels even further afield: applied event studies of airline alliances (for example codeshare and membership announcements in the Canadian airline industry) show the same machinery works in any vertical a practitioner happens to work in.

Short-run reward, long-run caution

Announcement returns capture expectations, and expectations can be revised. Harmancioglu, Griffith and Yilmaz (2019), tracking 270 international co-development alliances with latent-growth modeling, find that investors reward firms in the short term but tend to punish them over the long term: sharing innovation resources reads as an advantage at announcement and as a transaction hazard later. This sharpens the standard caveat that short-window CARs measure anticipated, not realized, value, and it is a particular concern for the open-ended, knowledge-pooling deals (research JVs, co-development) where contractual ambiguity is greatest. Koval et al. (2024) make the same methodological point with modern tools, stressing that buy-and-hold and calendar-time methods can give materially different long-horizon results because expected-return mis-specification compounds. The lesson is consistent across four decades: short-window CARs are the well-identified workhorse, and long-run claims are fragile.

A worked numeric example

To make the arithmetic concrete, consider a modern co-development alliance of the kind that now dominates the use case: a large pharmaceutical originator agrees to co-develop and co-commercialize a clinical-stage asset held by a much smaller biotechnology partner, with a sizeable upfront payment plus milestone payments. All numbers below are illustrative, chosen to show the mechanics and the small-partner effect, not drawn from a specific deal.

An abnormal return is the actual return minus the market-model expected return: AR_t = R_t - (alpha + beta × R_mkt,t). Take the small biotech partner with an estimated beta of 1.4 and alpha of approximately 0, and run the three event days:

• Day -1: stock +1.20%, market +0.30%. AR = 1.20 - (1.4 × 0.30) = 1.20 - 0.42 = +0.78%.
• Day 0: stock +7.50%, market +0.40%. AR = 7.50 - (1.4 × 0.40) = 7.50 - 0.56 = +6.94%.
• Day +1: stock +0.90%, market +0.20%. AR = 0.90 - (1.4 × 0.20) = 0.90 - 0.28 = +0.62%.

Summing across the window: CAR[-1,+1] = 0.78% + 6.94% + 0.62% = +8.34%. Now run the same three days for the large pharma originator, with beta 0.9, and the cumulated abnormal return comes to about +0.35% over [-1,+1]. The contrast is the whole point. Suppose the small partner has a market capitalization of $1.2B and the large partner $180B. The dollar value changes are then roughly comparable in magnitude even though the percentage reactions differ by more than twenty-fold:

Partner	Market cap	CAR[-1,+1]	Implied $ value change
Small biotech	$1.2B	+8.34%	~ +$100M
Large pharma	$180B	+0.35%	~ +$630M

This is McConnell and Nantell's relative-size effect made tangible: the dollar gains are of comparable order (here slightly larger for the giant, mirroring the +$6.65M-versus-+$4.54M pattern in the original sample), while the percentage reaction is overwhelming for the small partner and a rounding error for the large one. Reporting the two partners separately, and scaling by firm value, exposes the small-partner effect that pooling would hide. For the second stage, each row is coded with its moderators: the deal is flagged as R&D or co-development (not marketing), as cross-border or domestic, by the relative size of the partners, and by structure (here a milestone-plus-equity-stake collaboration). A single illustrative row of the request file, with one row per listed partner per deal, might carry columns such as firm_id, event_date, role, alliance_type, structure, partner, relative_size. The same moderator columns are then carried through into the regression described below.

How to run this kind of event study

For the conceptual foundations see our introduction to event study methodology; for a step-by-step workflow see the event study application blueprint. The points below are the specifics that matter for alliances and joint ventures, several of which are harder here than for mergers and acquisitions.

• Identify the event date carefully, and understand why the window is two days. Alliances are often disclosed gradually (letter of intent, memorandum of understanding, definitive agreement, launch) and many smaller deals receive only a single press release. The relevant event is the first credible public announcement of the alliance, not the legal signing or JV incorporation. McConnell and Nantell (1985) give the textbook rationale for the canonical [-1,0] window: the news-index or newswire date (their day 0 was the Wall Street Journal publication date) typically lags the first public release by a day, and one cannot tell whether the news reached the market before or after the close, so a two-day window absorbs the event-date ambiguity. Anchor on the first newswire or business-press date, verify it in a news database, and test robustness across alternative anchor dates because leakage and ambiguous dating are common.
• Treat the deal as a dyadic, two-observations event. Each alliance generates one observation per publicly traded partner. Decide whether the unit of analysis is the deal or the firm, and report partner-level abnormal returns separately (smaller versus larger partner), because pooling masks the size effect (McConnell and Nantell, 1985). Note that many partners, and most JV "children," are private and unlisted, which biases samples toward larger, listed firms.
• Report both percentage and dollar abnormal returns, and consider the scaled premium. The percentage effect is large for small partners while the dollar effect can be modest; reporting market-value-weighted dollar abnormal returns alongside raw CARs is central to interpreting who captures the synergy. Where the resources each partner commits are observable, the scaled premium Q (abnormal dollar gain divided by resources committed) is the cleanest single statistic for comparing partners and benchmarking against merger premiums (McConnell and Nantell, 1985). Disclose the weighting scheme explicitly: Park, Borde and Choi (2008) show that value-weighted and equally-weighted portfolio returns can differ materially, because the large percentage returns sit in the small partner, which partly explains why reported magnitudes vary across studies.
• Code the alliance type ex ante. The primary cross-sectional moderators (equity JV versus non-equity contractual alliance; technology or R&D versus marketing or distribution; horizontal versus vertical by industry overlap; domestic versus international) should be coded before estimation, since they are the heart of the second-stage regression.
• Use short, symmetric windows. Because the effect is small, narrow windows ([-1,0], [0,+1], [-1,+1]) maximize statistical power; wider windows ([-2,+2], [-5,+5]) add noise and pick up confounds and should be reported as robustness, not as the primary result. This short-window discipline is the central methodological recommendation of McWilliams and Siegel (1997), who show that conclusions in management event studies can reverse once window choice and confounds are handled properly.
• Screen aggressively for confounding events. Alliance press releases frequently bundle earnings, dividends, product launches, or simultaneous deals, and confounding is more acute here than for many event types (McWilliams and Siegel, 1997). McConnell and Nantell (1985) provide a ready-made template: they searched the news index for earnings, dividend, financing, capital-expenditure, and merger announcements in the window, dropped 30 contaminated firms, and the non-contaminated [-1,0] result held at +0.85% (z=3.45). Screen the event window and either exclude contaminated observations or report results with and without them.
• Handle event clustering and rival spillovers. Alliances cluster by industry and time (technology, biotechnology, banking), inducing cross-sectional correlation in abnormal returns, and rivals themselves react (Oxley, Sampson and Silverman, 2009). Use portfolio or calendar-time approaches, or correct test statistics for event-date clustering, rather than treating observations as independent, and be cautious using industry indices or rival firms as a clean benchmark.
• Correct for thin trading in small-cap partners. Small biotechnology partners often trade thinly, which biases beta and understates variance, so the naive market model can mislead exactly where the effect is largest. Use a Scholes-Williams or Dimson beta correction, and pair the return study with the abnormal-volume check below to confirm the announcement actually traded.
• Choose the benchmark and statistics to match a small, noisy effect. Use a market model or a Fama-French or Carhart factor model over a clean estimation window (commonly 200 to 250 trading days ending well before the event); for small-cap-heavy samples (biotechnology) control for size and value factors. McConnell and Nantell triangulated significance three ways: a standardized-prediction-error parametric test (Dodd-Warner), a non-parametric Wilcoxon signed-rank median test, and a comparison-period (Masulis) cross-check, exactly the multi-test discipline a small, clustered effect demands. Because the effect is small and clustering inflates rejection rates, prefer standardized cross-sectional tests (Patell, BMP) alongside a non-parametric sign or rank test; note that the BMP test specifically corrects for the event-induced variance increase that an alliance announcement causes, which is the reason a naive t-test over-rejects here. See our significance tests documentation and the expected return models overview; full sources are in the references.
• Treat the sample as self-selected. Firms choose to form alliances, and only successfully negotiated deals are ever announced, so the abnormal returns you measure describe a selected population. Interpret cross-sectional results as conditional associations rather than as the causal effect of "doing an alliance," and note the selection caveat alongside the confounding-events screen.
• Remember that announcement returns measure expectations, not realized value. Unlike a completed acquisition, an alliance announcement does not signal a finished transaction: deals can be non-binding or never executed, and the short-run reaction may be partly reversed later (Harmancioglu, Griffith and Yilmaz, 2019). Short-window CARs are the well-identified workhorse; long-run buy-and-hold or calendar-time claims are fragile and benchmark-sensitive, so foreground the short-window evidence.

The second stage is a cross-sectional regression of each partner's abnormal return on the coded moderators. A representative specification is CAR_i = a + b₁(equity JV) + b₂(technology or R&D) + b₃(relative size) + b₄(international) + b₅(prior alliance experience) + b₆(alliance-function dummy) + e_i, where relative size is the partner's market value divided by the firm's own. The literature predicts b₂ > 0 (technology beats marketing, at least for small and research-intensive firms), b₃ < 0 (the smaller own firm earns the larger percentage return), and b₅, b₆ > 0 (experience and a dedicated alliance function add value). Reporting the regression alongside the headline CAARs is what turns a descriptive result into the explanatory account that this use case is really about.

Common misconceptions and pitfalls

The cautions above recur often enough to be worth collecting in one place. The most common ways an alliance event study goes wrong are interpretive, not computational:

• A positive CAR is an expectation, not a verdict. The announcement return measures the market's estimate of the deal's value at announcement, not whether the alliance ultimately succeeded. Short-run rewards can reverse (Harmancioglu, Griffith and Yilmaz, 2019), so do not read a positive day-0 reaction as proof the cooperation worked.
• Do not pool the partners. Collapsing both firms into one CAAR hides the small-partner effect, which is the most informative feature of the data. Always report the smaller and larger partner separately and scale by market capitalization.
• Screen bundled news. Alliance press releases routinely arrive alongside earnings, guidance, product, or other deal news, so confounding screening matters more here than for most event types.
• The sample is self-selected. Only negotiated, announced deals exist in the data, so cross-sectional results are conditional associations, not the causal effect of forming an alliance.
• Long-run BHAR is fragile. Buy-and-hold and calendar-time abnormal-return estimates are highly model-sensitive; foreground the short-window CARs and treat long-horizon claims as tentative.
• Mind thin trading and event-induced variance. Small-cap partners need beta corrections and variance-robust tests (BMP); a naive t-test will over-reject precisely where the effect is most interesting.

Who uses this, and why

Alliance and JV event studies are run by four distinct communities, each with a different question and a natural entry point among our tools:

• Event-driven and biotech investors and sell-side analysts read the announcement-day reaction in real time to size the small-partner effect and judge whether the market is pricing a milestone-laden deal as transformational or marginal. Start with the Abnormal Return Calculator.
• Corporate development and alliance-management offices benchmark their own announcement returns against peers and use the evidence (Kale, Dyer and Singh's +1.35% versus +0.18% for firms with a dedicated alliance function) to justify building that capability. ARC plus the cross-sectional regression is the workflow.
• Antitrust authorities and competition economists read the abnormal returns of non-participating rivals to judge whether a JV or alliance is pro- or anti-competitive under the DOJ/FTC guidelines for collaborations among competitors, following the Oxley-Sampson-Silverman "detente versus arms race" logic. Run ARC on a rival portfolio.
• Securities-litigation and damages experts price an alliance disclosure, or its later collapse or quiet unwinding, for materiality, price impact, and loss causation. The same event-study machinery used in court applies directly; see our guide on event studies in securities litigation (Basic v. Levinson, Halliburton II, Daubert), because a misrepresented or reversed alliance is a litigated, value-relevant disclosure.

To make the small-partner effect tangible, three named episodes are illustrative. Summit Therapeutics and Akeso's ivonescimab collaboration (December 2022; about $500M upfront to Akeso plus up to roughly $4.5B in milestones) is the modern co-development archetype, and Summit's stock subsequently ran several hundred percent. The 2025 Pfizer-Metsera-Novo Nordisk episode (a $4.9B offer countered at $8.5B) shows a partnership-then-acquisition announcement repricing a small partner dramatically and triggering a bidding war. The Sony Ericsson JV unwind is the two-partner equity-JV-structure case: on the announcement that Sony would buy out Ericsson's 50% stake, both parents rose on the order of 4% to 5%. In each, the convention is one row per listed partner, with day 0 set to the first credible press release.

Run it with our tools

Our calculators implement this two-step workflow end to end and are free to use:

• Abnormal Return Calculator (ARC): the core tool. Build a request file of announcement dates with one row per publicly traded partner, pick the estimation and event windows ([-1,0] or [-1,+1] as primary), choose the market model or a factor model, and select the test statistics (including standardized and non-parametric tests). Run separate batches for equity JVs and contractual alliances, and for technology versus marketing deals, then add alliance-type, relative-size, relatedness, and prior-experience variables for the cross-sectional regression.
• Abnormal Volume Calculator (AVC) and Volatility Calculator (AVyC): corroborate the small return reaction with abnormal trading volume and volatility around the announcement, which is especially useful when confounding events make the return signal ambiguous, and as the thin-trading check for small-cap partners flagged above.
• Event Date Identifier (EDI): given the multi-stage alliance timeline (letter of intent, definitive agreement, launch), EDI parses large volumes of press releases and returns the dates mentioned in the text, helping you pin down the correct day 0.
• News Analytics (CATA): classify and code alliance press releases at scale, for example to separate equity JVs from contractual alliances, flag technology versus marketing language, or measure tone, before running the return study. In practice the coding step in the worked example above is itself a CATA job rather than hand-coding.

Frequently asked questions

Do strategic alliances create shareholder value?

On average, yes, but modestly. Short-window cumulative abnormal returns to the participating firms cluster around +0.5% to +1.0%, a result that has replicated from McConnell and Nantell's +0.73% in 1985 to Koval et al.'s +0.62% to +0.69% in 2024. The gains are shared synergy, not a zero-sum transfer between partners (Chan et al., 1997).

Which creates more value, a joint venture or a contractual alliance?

Equity joint ventures typically show larger and cleaner reactions. Koh and Venkatraman (1991) found equity JVs and technology exchanges significant while licensing, marketing, and supply agreements showed no significant reaction, and Allen and Phillips (2000) show that bundling an equity stake with an alliance raises partner returns. The intuition is a commitment ladder: more binding structures carry more information.

Why does the smaller partner's stock jump more?

Because the percentage return equals the dollar gain divided by market capitalization. The same dollar synergy is a large percentage for a small firm and a rounding error for a giant. McConnell and Nantell (1985) found the smaller partner earned +1.10% versus +0.63% for the larger, even though the dollar gains were comparable ($4.54M versus $6.65M).

Technology or marketing alliance: which gets the bigger reaction?

Technology and R&D alliances usually earn higher abnormal returns than marketing alliances (Das, Sen and Sengupta, 1998; Chan et al., 1997), because of their knowledge content. The ranking is context-dependent, though: it is strongest for small, research-intensive firms, can vanish or reverse for large firms (Koval et al., 2024), and reverses entirely in some markets such as Korea.

What event date and window should I use?

Use the first credible public announcement (newswire or business press), verified in a news database, as day 0, and a short symmetric window, with [-1,+1] as the field-standard primary window and [-1,0] also common. The two-day window absorbs the ambiguity about whether news hit before or after the close.

Does a positive announcement return mean the alliance will succeed?

No. The announcement CAR is the market's expectation at announcement, not a measure of realized success. Harmancioglu, Griffith and Yilmaz (2019) show short-run rewards can reverse over the long run, so a positive day-0 reaction is a forecast, not a verdict.

Related use cases

Alliances and joint ventures sit alongside other corporate-event applications of the method. See mergers and acquisitions (the change-of-control cousin, where the same relatedness and payment logic applies) and divestitures (the mirror-image restructuring decision), and competitive dynamics (for the rival and spillover effects that alliance announcements trigger). Because alliance disclosures are litigated, value-relevant events, see also our guide to event studies in securities litigation. For the full catalogue of applications, return to the overview of practical applications.