Advanced Corporate Finance — Revision Hub

The capital-structure literature is one long argument about what — if anything — pins down leverage. Modigliani–Miller's irrelevance benchmark gives way to Myers–Majluf's asymmetric-information pecking order (internal funds → debt → equity), which Shyam-Sunder–Myers claim beats the trade-off model — though their own power tests warn the horse-race can't really discriminate. Graham–Harvey's CFO survey finds managers care more about financial flexibility, credit ratings and EPS dilution than tax shields, while Baker–Wurgler argue leverage is just the cumulative residue of past equity-market timing. Then Lemmon–Roberts–Zender and Fukui–Mitton–Schonlau deliver the deflating punchline: a stable firm fixed effect — present even before the IPO — explains far more of leverage than any theory's pet covariates. On the issuance side DeAngelo et al. and Dittmar et al. show SEOs follow cash need and behavioural anchoring, not rational timing, and Mota–Siani extend the lens to debt and investor-base design. The thread agrees on one thing: no single theory wins, and persistent firm heterogeneity dwarfs the textbook determinants.

Security design asks why firms issue hybrids rather than straight debt or equity. Dutordoir et al.'s survey concludes the four classic rationales for convertibles — risk-shifting, risk-uncertainty/adverse-selection, backdoor equity, and sequential financing — each win under some method or geography, with none dominating. Brown et al. supply the modern mechanism: most convertibles are bought by convertible-arbitrage hedge funds who delta-hedge by shorting the stock, so the issue is really a cheap, backdoor way to distribute equity exposure — and the announcement-date short selling is information-free hedging, not a bad signal. Together they reframe convertibles from a pure adverse-selection story (Myers–Majluf's 'debt plus an option') toward an investor-demand / market-microstructure story.

M&A research splits into 'who gains?' and 'what explains the cross-section?'. Andrade–Mitchell–Stafford set the stylised facts — combined announcement gains are positive (~+2%) but targets capture essentially all of it while acquirers earn ~0, and stock deals fare worse. Harford explains the timing: waves need an industry shock and abundant capital liquidity, and the clustering of cash/divisional deals rejects the pure overvaluation story. The cross-section turns out to be largely firm-specific — Golubov–Yawson–Zhang find acquirer skill is a persistent fixed effect (echoing Lemmon–Roberts–Zender in capital structure). The rest isolate information channels: banks (Ivashina et al.) and advisors (Golubov et al. on reputation; Eaton et al. on biased valuation comps) as intermediaries, and the payment method as signal — Malmendier et al. (cash reveals target value), Dutordoir et al. (cash dodges arbitrage shorting), Filipovic–Wagner (cash + intangibles talk signals overoptimism). 'Cash' means something different to each, which is why payment choice stays unsettled.

Which mechanisms actually discipline managers? La Porta et al. make legal origin foundational: common-law countries protect investors better, and where law is weak, concentrated ownership emerges as a substitute — a point Denis–McConnell's survey generalises into the controlling-shareholder-vs-minority agency problem that dominates outside the US/UK. Inside the firm, Gompers–Ishii–Metrick price governance directly (their 24-provision G-index 'democracy-minus-dictatorship' portfolio earned ~8.5%/yr in the 1990s — though later work questions whether that was causal or a one-decade artifact). On boards, Yermack shows small boards command higher valuations, and Adams–Akyol–Verwijmeren refine 'good board' from independence toward skill composition. Lie's backdating study is the clean-identification highlight — post-grant abnormal returns contain a market-wide component executives cannot forecast, so the grants must have been backdated. The common lesson: external law and internal board/ownership structure are substitutes, and the binding constraint differs by country.

What do large investors do — to prices, and to managers? Gompers–Metrick show institutional growth tilted demand toward big, liquid stocks and explains ~half their relative price rise (a demand, not information, effect). Yan–Zhang qualify this sharply: only short-term, high-churn institutions predict returns and earnings, so 'smart money' is a horizon-specific subset, not institutions in general. Bharath–Jayaraman–Nagar move from prices to governance — blockholders discipline managers via a credible threat of exit, identified with the same liquidity that enables informed trading in Yan–Zhang. The agreement: the composition and horizon of ownership matter far more than its level.

These three price the consequences of the public/private margin — specifically information and ownership. Badertscher et al. show private opacity is priced by bond markets: private issuers pay 26–60bp more (partly justified by higher realised defaults). Gogineni et al. open the private firm's black box: both vertical (owner–manager) and horizontal (majority–minority) agency problems independently — and more than additively — destroy operating performance. Ortiz et al. show transparency is causal and valuable on the M&A margin: mandatory disclosure roughly doubles a private firm's acquisition probability by cutting acquirers' search costs. They agree the public/private choice is a genuine trade-off — opacity and concentrated ownership carry real costs — undercutting any simple 'private firms escape agency and short-termism' narrative.

Both papers ask whether AI is a first-order driver of firm value, and both answer yes — through different lenses, and largely agreeing. Babina et al. use slow-moving firm-level AI hiring and find AI raises growth via product innovation, concentrating gains in large 'superstar' firms. Eisfeldt–Schubert zoom in on the generative-AI structural break (ChatGPT) and find an ~8–10% market revaluation of exposed firms almost overnight, plus that LLMs are useful research tools. They differ in method (a multi-year labour-stock instrument vs a high-frequency event study) and object (realised real growth vs anticipated price revaluation) — leaving open whether prices have correctly anticipated AI rents or run ahead of them.