Corporate finance has one question, and it was installed in 1958. Before then the field asked what firms should do: how much to borrow, what to pay out, which projects to fund, questions answered by taxonomy and seasoned judgment because there was no benchmark against which a practice could be wrong. Modigliani and Miller replaced that question with a sharper one. If, in a frictionless market, the value of the firm is untouched by how it is financed, then every observed financing regularity is evidence of a friction, and a theory of corporate finance is a theory of which assumption fails. This essay traces the sequence of answers. It is a history of a question, not a catalog of results, and it treats individual papers as evidence for claims about how the question changed rather than as entries in a bibliography.

The eras that organize what follows are defined by the class of friction each proposed, not by their dates, and they overlap in calendar time because a field can ask two questions at once. The field tried the tax code and the bankruptcy courts first, and found the frictions too small. It tried the people inside the firm, their private information and their private interests, and produced theories that explained everything and therefore discriminated among nothing. It tried the allocation of power when contracts fall silent, and derived the same security from three incompatible primitives. It then turned on its own instruments and found that the regressions meant to detect the frictions could not identify them, just as the object those regressions studied, the dispersed-ownership public industrial firm, began to leave the sample. The friction left standing at the end is the one the frictionless benchmark deleted at birth: the financier in the middle, the intermediary whose own balance sheet decides which firms feel which frictions. The essay ends there, where the history of asset pricing, approached from the other side of the same balance sheet, ends too.

Prologue: The Field Before the Question (1900-1957)

Corporate finance before 1958 was a field without a question. It had subject matter in abundance: securities, promotions, reorganizations, capital budgets, dividend customs. What it lacked was a proposition that could be wrong. A field accumulates when its claims can collide; the pre-1958 literature was built so that collision was impossible, each book a self-contained codification of practice, each rule of thumb answerable only to experience, never to another rule. The best work of the period is impressive precisely in the way a legal treatise is impressive, and the comparison is not an insult, because the treatise was the genre its authors chose. Three texts show the tradition at its ceiling, and show that the ceiling was not stupidity but the absence of a benchmark against which practice could be wrong.

Arthur Stone Dewing’s Financial Policy of Corporations, the field’s dominant text for four decades, organizes corporate finance by instrument and life-cycle stage: the securities a corporation issues, how it is promoted, how it administers income, how it expands, how it fails (Dewing, 1953). Its central intellectual device is a taxonomy, not a theorem. Capital stock comes in two types; par value is a fiction; no-par stock has the “advantage of truthfulness.” Dewing himself was more careful than his successors’ caricature of him: he explicitly separated statements of fact from statements of policy, warned that apparent inductive principles are summaries of limited experience, and distinguished “capital” (the real wealth of the business) from “capitalization” (the artificial, ultimately fictitious representative values issued against it) (Dewing, 1953). That last distinction is a near-miss of historic proportions. A reader who takes it seriously is one step from asking whether rearranging the fictitious claims can change the real value underneath. Dewing never asks. He treats capitalization as a matter of institutional form and expedient policy, and the question sits on his page for forty years, unasked.

Joel Dean’s Capital Budgeting (1951) shows the tradition at its analytical peak, and the peak is a decision system, not a theory. Dean built the first systematic framework for the investment decision: rank proposals by prospective rate of return, fund from the top until capital runs out, let the marginal project define the cut-off rate, and reject the payback rule (Dean, 1951). His worked example sets $20 million of capital requests against $14 million of supply and rations the difference (Dean, 1951, Ch. I). This is genuine progress, and it is progress of a specific, bounded kind: the cost of capital enters as an engineering input, and the aversion to external finance is documented as managerial fact rather than derived from any friction. Dean called the space between the pure theory of investment and management’s actual decision a “no man’s land” (Dean, 1951, p. 2). He mapped it; he did not annex it.

Between Dewing’s taxonomy and Dean’s decision system sits the tradition’s whole intellectual range: description of forms at one end, codified managerial judgment at the other, and at neither end a benchmark. Ask what the tradition would say to a firm choosing between debt and equity and the answer is a list of considerations; ask what it would say to a firm that chose wrongly and there is no answer, because without a valuation theorem there is no such thing as wrongly. The prologue’s most instructive figure is David Durand, because Durand asked the question and then declined to answer it. His 1952 NBER paper lays out two ways to value a levered firm: capitalize net operating income for the whole firm, in which case total value is invariant to the debt-equity split ($20 million in his worked example, whatever the bond issue); or capitalize net income to shareholders, in which case each $2.5 million of debt adds $1.5 million of value (Durand, 1952, p. 227). The first is Modigliani-Miller Proposition I in everything but proof. Durand even gestured at the arbitrage argument, acknowledging that arbitragers would normally erase excess yield differentials (Durand, 1952, pp. 230-231). But he believed the frictions were the reality and the invariance the abstraction: restricted institutional investors let firms exploit a “super premium for safety,” so he built a compromise schedule running between his two extremes and declared the conflict unresolved (Durand, 1952, p. 233, p. 237). He was explicit that cost-of-capital research was blocked until someone resolved it. Six years later someone did, by insisting on exactly the frictionless benchmark Durand’s instincts compromised away. The prologue thus ends with the field holding all the pieces: the invariance conjecture, the arbitrage mechanism, the tax advantage of debt, even (on Durand’s last pages) the observation that managers’ interests diverge from stockholders’ (Durand, 1952, pp. 260-261). What it lacked was the willingness to let a theorem be true in a world that visibly contradicted it.

Era I: The Theorem (1958-1963)

Did Modigliani and Miller prove a result about firms, or did they abolish the field’s question and install a new one? The claim this section defends is the second, stronger reading: irrelevance is the founding act of modern corporate finance because it converted “what should firms do” into “which assumption fails.” A reader can resist this. One can hold that MM proved a substantive economic theorem, that the value of a firm is the capitalized value of its assets, and that the reframing reading is a later imposition by a field that needed a founding myth. The evidence for the stronger reading is not in the 1958 paper’s stated intent, which was modest, but in what every subsequent era did with it, which was not.

Start with what the paper actually says, because almost nobody does. The 1958 paper is not addressed to capital-structure specialists; it opens by observing that three audiences, the corporate-finance specialist, the capital-budgeter, and the macroeconomist, each need a theory of the effect of financial structure on valuation and none has one (Modigliani and Miller, 1958, p. 261). The machinery begins with the assumption everyone forgets to count as an assumption: firms sort into equivalent-return classes, within which every firm’s earnings stream is a scale copy of every other’s, so that shares within a class are perfect substitutes up to a scale factor (Modigliani and Miller, 1958, pp. 266-267). The risk class is what makes arbitrage possible before anyone has a model of risk pricing; it is 1958’s substitute for the asset-pricing theory that did not yet exist. On that scaffolding sit three propositions. Proposition I: the market value of any firm is independent of its capital structure, V equals expected earnings capitalized at the rate appropriate to its risk class (Modigliani and Miller, 1958, p. 268). Proposition II: the expected yield on equity rises linearly in leverage, i = rho + (rho minus r)(D/S), so the cheapness of debt is exactly consumed by the rising required return on the levered equity (Modigliani and Miller, 1958, p. 271). Proposition III, the one Dean should have been waiting for: the investment cut-off rate is rho, the class capitalization rate, regardless of how the investment is financed (Modigliani and Miller, 1958, p. 288). The prologue’s two traditions are answered in one stroke: Durand’s NOI-versus-NI conflict is resolved in favor of NOI, with MM citing Durand as the source of the taxonomy (Durand, 1952), and Dean’s cut-off rate, which he treated as an engineering input, is derived and shown independent of the financing mix that so exercised his clients.

The proof mechanism is arbitrage, not equilibrium in preferences: if Proposition I failed, an investor could sell one income stream and buy an identical, cheaper one, borrowing on personal account to undo or redo the firm’s leverage, and the switch is profitable regardless of the investor’s attitude toward risk (Modigliani and Miller, 1958, pp. 269-271). MM’s own gloss on the mechanism is the milk parable: the firm that issues debt is selling the cream and the skim milk separately, and arbitrage guarantees the whole milk fetches the same price either way (Modigliani and Miller, 1958, pp. 279-280). The device matters more than the destination. Proofs by arbitrage require no assumptions about who investors are or what they want, only that two identical things cannot sell at two prices, and the home-made-leverage step tells you exactly what the theorem needs: investors who can borrow on the same terms as firms. Every later era can be read as an account of who cannot. The field learned to ask what breaks the arbitrage, and each answer became an era: the tax collector stands inside the cash flows (Era II), the informed manager cannot be arbitraged against (Era III), control is not a cash flow at all and cannot be replicated in a personal account (Era IV), and the intermediary through whom the borrowing runs has a balance sheet of its own (Era VI). That is why the theorem generalized while the specific propositions did not.

Two features of the founding paper are systematically forgotten, and both are load-bearing for this history. The first is that MM ran regressions. Section I.E tests Proposition I on samples of 43 electric utilities and roughly 42 oil companies: the cost of capital regressed on leverage is flat and insignificant (for the utilities, x = 5.3 + .006d with r = .12; for oil, x = 8.5 + .006d with r = .04), against the traditional view’s predicted negative slope (Modigliani and Miller, 1958, p. 282). The equity-yield regressions come out strongly positive (r = .53 in both samples), consistent with Proposition II (Modigliani and Miller, 1958, p. 285). By modern standards the test is weak: crude proxies, near-zero correlations on the first regression, no identification to speak of. The field kept the theorem and quietly discarded its first empirical test, which is itself a fact about the field: the paper survived as mathematics after failing to persuade as measurement. The second forgotten feature is that the tax analysis of 1963 is already present in 1958, filed under “small.” The 1958 paper works through the corporate tax and reaches a tax-adjusted cost of capital that falls with leverage (Modigliani and Miller, 1958, p. 294), but MM judge the gain minor, equity money “only 25 per cent more expensive than debt money” (Modigliani and Miller, 1958, pp. 295-296). The friction that would define the next era enters the literature inside the founding paper, priced and dismissed by its own authors.

The 1963 correction repriced it. MM concede that the 1958 proof mishandled taxes because the after-tax returns of two same-class firms are not proportional when their leverage differs (Modigliani and Miller, 1963, p. 434), and derive the formula every student now learns: the value of the levered firm is the unlevered value plus the tax rate times the debt (Modigliani and Miller, 1963, p. 436). The mechanism is unchanged from 1958; what changes is the capitalization rate on the tax shield, the sure stream tau-R now discounted at the certainty rate rather than the risky rate. I stress this because the trade-off theory’s first blade is therefore not a new idea entering the field but an old idea repriced, and the repricing was performed by the theorem’s own authors in a paper framed as an apology for an algebra error. And here the founding papers execute their most consequential move: having enlarged the tax advantage, MM refuse the corner solution it implies. The 1963 conclusion states that the existence of the tax gain need not mean firms “should at all times seek to use the maximum possible amount of debt,” and lists the reasons: personal taxation of interest income, limits imposed by lenders, and the need to preserve flexibility, a reserve of untapped borrowing power (Modigliani and Miller, 1963, p. 442). Read that list again. Personal taxes are Miller (1977). Lender-imposed limits are the agency and control literatures. Flexibility is the top answer CFOs will give Graham and Harvey in 2001. The founding papers contain, in one closing paragraph, the research program of the next four decades, stated as reasons to distrust their own headline formula.

The 1961 dividend paper belongs here as the companion extension: irrelevance carried from the financing margin to the payout margin. The field credits it with the informational-content-of-dividends concession that seeds the signaling literature (Miller and Modigliani, 1961). For this history, the point is its reception and benchmark role rather than a reconstruction of every step in the paper’s internal proof. Black’s dividend puzzle is a puzzle only against the 1961 benchmark, and the signaling program of Era III grows from the concession the paper is credited with. Both debts are therefore stated as reception claims rather than overloaded as new textual claims.

MM thought they were writing a partial-equilibrium cost-of-capital paper, reconciling the finance specialist, the capital-budgeter, and the macroeconomist, with novelty claimed for Proposition II and the arbitrage proof, and with the general-equilibrium extension explicitly deferred (Modigliani and Miller, 1958, p. 261, p. 271). The 1958 conclusion describes the model’s assumptions, atomistic competition, statics, easy market access, as drastic simplifications adopted in order to come to grips with the problem at all, to be relaxed toward greater realism (Modigliani and Miller, 1958, p. 296). The field decided the durable contribution was the reframing: after 1958, a theory of capital structure is a theory of which assumption fails. I side with the field over the authors here, and it is worth saying why: intent is not the measure of a founding act. What makes 1958 the founding paper is not that MM meant to reorganize the field but that the reorganization is what the paper does when read, the way a constitution founds a state whatever its drafters thought they were drafting. The gap between the two readings is the era’s reversal, and it sets the pattern every later era repeats. But the reframing reading leaves a residue the theorem cannot digest. If financing choices are irrelevant, the observed behavior of financial officers is a standing anomaly: they deliberate, they preserve debt capacity, they treat the choice as consequential, and MM’s own 1963 paragraph concedes them reasons. If the theorem is right, why do CFOs visibly agonize over these choices? The next era’s answer was that the agony is rational because the tax code and the bankruptcy courts make it so. That answer, mechanical frictions, is where the field went first, and it is the era that fails most instructively.

Era II: Mechanical Frictions (1963-1976)

The first answer to “which assumption fails” was the most literal one: the tax code subsidizes debt, bankruptcy penalizes it, and the two together pin down an optimal capital structure for every firm. The claim is contestable, and this era is the rare case where the claim’s own authors supplied the refutation within the era’s lifetime. It failed on its own terms, and the failure is the era’s contribution.

Kraus and Litzenberger (1973) gave the trade-off its rigorous form, and the form deserves respect before the attack arrives: this is not intuition dressed in algebra but a theorem inside a complete-markets state-preference model, the most disciplined pricing framework then available. Value additivity across states forces the decomposition: the value of the levered firm equals the unlevered value plus the corporate tax rate times the market value of debt, minus the after-tax present value of bankruptcy costs, with the model specifying exactly which states trigger the costs, those where earnings fall short of the debt obligation (Kraus and Litzenberger, 1973). The formalization contained two warnings its canonizers ignored. Optimal debt is found by enumeration over states, the optimum sits at one of the finitely many earnings levels, not at a smooth first-order condition; and firm value need not be concave in leverage at all: value at an interior debt level can sit strictly below the value at both adjacent levels, so the textbook inverted U is a possibility, not a theorem (Kraus and Litzenberger, 1973). The paper the field remembers as the clean inverted-U benchmark proved, in the same pages, that the inverted U is not guaranteed. This is the first instance of a pattern that recurs through this history: a paper canonized for its tidy result while its own caveat, which undercuts the tidy result, is forgotten.

Miller’s 1977 Presidential Address is the era’s defining dispute, and it is worth being precise about what kind of dispute it is, because it is narrower and deadlier than the textbook version. Kraus-Litzenberger and Miller do not disagree about whether debt carries a corporate tax advantage; the tax term is the leading positive entry in the trade-off equation and the numerator of Miller’s gain-from-leverage formula alike. The entire dispute is over the magnitude of one coefficient in equilibrium. Miller’s attack has two prongs. The magnitude prong: for conspicuously low-levered firms like IBM and Kodak, balancing the tax gain against bankruptcy costs is “the recipe for the fabled horse-and-rabbit stew, one horse and one rabbit” (Miller, 1977, p. 264), the tax benefit being the horse; the footnote adds that the leading bankruptcy-cost calibration would require distress to consume 40 percent of asset value. Behind the joke is a stylized fact the trade-off theory has never answered: the debt ratio of the typical nonfinancial corporation in the 1950s was little changed from the 1920s, though the corporate tax rate had risen from 10-11 percent to 52 percent (Miller, 1977, p. 264). A friction that quintuples without moving the choice it supposedly governs is not the binding friction. The equilibrium prong: once personal taxes on interest and equity income enter, the gain from leverage is zero when the personal-tax disadvantage of interest income exactly offsets the corporate deduction (Miller, 1977, p. 267). Firms sort into leverage clienteles, “one clientele is as good as the other” (Miller, 1977, p. 269), and there is an aggregate equilibrium quantity of debt but no optimum for any firm. Miller even anticipated the behavioral objection, that firms do not reason this way, and answered it with an evolutionary argument: surviving heuristics are compatible with equilibrium (Miller, 1977).

Steelmanned, the two sides fight to an instructive draw, and it is worth insisting on how narrow the battlefield is, because the textbook version stages it as a clash of theories when it is an argument about one coefficient. The trade-off advocate concedes Miller’s personal-tax arithmetic and holds ground anyway: the framework survives a smaller tax term, bankruptcy costs are real and state-contingent, and a determinate optimum needs only a positive net benefit; the model even has honesty built in, having flagged its own non-concavity. The Miller advocate answers that the defense has given up the theory’s content: a trade-off whose tax term is nearly zero and whose cost term is a rabbit explains neither the level nor the stability of observed leverage, and a theory reduced to “some positive net benefit exists” no longer predicts which firms borrow. The field’s eventual verdict split the difference in a way neither side proposed: it rejected Miller’s firm-level irrelevance once non-debt tax shields were shown to restore interior optima, and kept his method, the insistence that capital-structure theory reason about the supply and demand for corporate debt and the identity of the marginal bondholder (Miller, 1977). A paper arguing for irrelevance was absorbed as a lesson in equilibrium reasoning. The era’s reversals all run this way: Kraus-Litzenberger, remembered for the benchmark their caveat undercut; Miller, remembered for the method his conclusion rode in on.

Black’s “The Dividend Puzzle” (1976) is the era’s terminal document, and in idea-order it comes last even though it is contemporaneous with Miller’s address, a four-page essay in a practitioner journal that has outlived shelves of theorems, because it presumes the trade-off machinery is built and attacked and then shows the machinery cannot answer the field’s most elementary question: why do firms pay dividends at all? Black runs the MM elimination procedure on dividends, and the procedure is the essay’s real content: start from irrelevance, add frictions one at a time, and watch each fail to carry the observed behavior. Taxes cannot explain dividends; they say firms should pay little or none, since every dollar of dividend is a dollar the tax code punishes relative to the repurchase that replicates it (Black, 1976, p. 6). Transaction costs cannot explain them; a repurchase plan arranges away the small investor’s need for homemade income (Black, 1976, p. 6). Signaling gets partial credit and then fails at the decisive margin: managerial information explains why dividend changes move prices, and cannot explain why the level of dividends exists, since a no-dividend convention would carry the same information in its deviations (Black, 1976, pp. 6-7). The creditor-expropriation channel, dividends as a transfer from bondholders to stockholders, is real and he judges it too small or too negotiable to carry the phenomenon (Black, 1976, p. 7). The essay ends in a confessed dead end: we do not know why firms pay dividends, and we do not know what firms or investors should do about it (Black, 1976, p. 8). The field received the confession as a signpost, “the dividend puzzle,” and the reception was correct in a way Black did not intend. Look at where his elimination argument kept snagging: managerial information, creditor conflict, the inside of the firm. Miller’s address snags in the same place; he flags the cross-section of leverage as understudied Miller (1977), p. 272 and defends rational modeling with an evolutionary argument rather than a theory of the manager. Both papers name frictions located in the boardroom, information asymmetry and the agency of debt, and both decline to build them. Mechanical frictions had been tried and found too small. The frictions that matter might not be in the tax code but in the boardroom, and the next era walked through the door this one left open.

Era III: Human Frictions (1976-1990)

The second answer to “which assumption fails”: the people inside firms know things outsiders do not, and they act on their own account. In its strong form, the era’s claim is that capital structure is a signaling and governance technology, and that every observed financing regularity, leverage, dividends, maturity, the takeover wave, is downstream of one of two human frictions, asymmetric information or the manager-owner conflict. The claim can be resisted from inside the era’s own record. Its foundational information paper describes itself as one of many possible stories (Myers and Majluf, 1984), and its most public empirical claim, that takeovers are efficient discipline, rests on a bidder-return estimate of roughly zero that its own authors flagged as the finding most likely to age badly (Jensen and Ruback, 1983).

The era’s anchor is Jensen and Meckling (1976), and the anchor deserves the extended treatment usually reserved for MM, because the gap between what the paper is and what the field made of it is the widest in this history. What the paper is: a theory of ownership structure. Jensen and Meckling say so, deliberately, distinguishing their question from “capital structure” because the object they want to determine is the split between inside equity, outside equity, and debt, that is, who owns the firm, not merely how it is levered (Jensen and Meckling, 1976). The machinery is built in three moves. First, dissolve the firm: it is not an individual with a utility function but a “nexus for a set of contracting relationships,” a legal fiction through which conflicting individual objectives are brought into equilibrium (Jensen and Meckling, 1976). Second, define agency costs as a sum: monitoring expenditures by the principal, bonding expenditures by the agent, and the residual loss, the value destroyed by the divergence that survives monitoring and bonding (Jensen and Meckling, 1976). Third, and this is the move that does the work, price the conflict into the security issue itself: a rational outside buyer anticipates the manager’s behavior at any given ownership split and pays accordingly, so the owner-manager who sells equity bears the full expected value of his own future misbehavior in the price he receives (Jensen and Meckling, 1976). Agency costs are not imposed by wicked managers on trusting shareholders; they are borne by the seller, which is why the seller has the incentive to invent the institutions that reduce them. The same pricing logic runs on the debt side: lenders anticipate that a levered owner-manager will shift into wilder projects, since equity keeps the upside and debt eats the downside, and they price the anticipated shifting into the interest rate, so the owner bears the agency costs of debt too (Jensen and Meckling, 1976). Optimal ownership structure minimizes the total: the agency costs of outside equity (perquisites, shirking, the monitoring they provoke) against the agency costs of debt (risk-shifting and the bonding and bankruptcy apparatus debt requires). Because each cost rises in its own instrument, an interior mix exists, and capital structure acquires, for the first time since 1958, a determinate optimum that owes nothing to the tax code. That is the paper’s actual theorem, and it is a theorem about an owner-managed firm selling claims, not about General Motors.

Now the reversal. Jensen and Meckling stated, in the paper, that its “most serious limitation” was that it had not been applied to the large modern corporation whose managers own almost no equity, and they deferred the control machinery, the right to hire and fire, tender offers, proxy fights, to “a future paper” (Jensen and Meckling, 1976). The field did not wait for the future paper. It took the nexus-of-contracts aside as a general theory of organizations, spun the debt-agency section into the debt-overhang and free-cash-flow programs, and built the entire takeover literature inside the gap the authors had marked as unexplored. The theory of everything that “agency theory” became was constructed out of a paper about the ownership split of an owner-managed firm, and the construction happened in the deferred territory, not the delivered one. Meanwhile the question Jensen and Meckling actually posed, what determines the insider-outsider equity split, proved empirically stubborn and faded from the headline. I know of no cleaner case of a field promoting a paper beyond its authors’ stated scope while abandoning the paper’s stated subject.

The information branch runs parallel, and its papers were presented in the same rooms in the same year. The branch’s shared move is to ask what a financing choice can say that an announcement cannot. Talk is free, so talk is worthless; a signal persuades only if mimicking it would hurt the mimic. Ross (1977) makes the debt level itself the message: with managerial payoffs tied to a known incentive schedule that penalizes bankruptcy, a separating equilibrium exists in which only high-quality firms can afford high leverage, because the low type’s weighted gain from mimicry falls short of its weighted bankruptcy penalty (Ross, 1977). The result restores capital-structure relevance with no taxes anywhere in the model, which is its historical function: it proves the MM assumptions can fail through information alone. Ross treated the signal as nearly costless, since in equilibrium no firm actually incurs the penalty; the field kept the conceptual win, that a financing choice can move the market’s perceived risk class, and treated the model itself as fragile, resting on an exogenous incentive schedule and admitting multiple equilibria (Ross, 1977). Leland and Pyle (1977), from the same 1976 AFA meetings, make the entrepreneur’s retained stake the message, at a real cost: signaling forces owners to hold more of their own project than diversification would counsel, so the information problem is solved by paying a risk-bearing toll (Leland and Pyle, 1977). Their closing pages contain one of this history’s longest-fused ideas: a few paragraphs offered as preliminary thoughts, suggesting that the same logic implies an institution that certifies borrowers by holding their claims, a theory of financial intermediation (Leland and Pyle, 1977). The aside outlived the paper; its lineage runs to Diamond (1984) and, through it, to the era in which this history ends. Myers (1977) supplies the branch’s most durable machine: with risky debt of face value P outstanding, equityholders exercise the growth option only when its value exceeds investment plus P, so states exist in which positive-NPV projects die, and the loss falls on the firm that issued the debt (Myers, 1977). His decomposition of firm value into assets-in-place and growth options became the market-to-book control in every capital-structure regression since, which is its own small reversal: a theory of borrowing remembered as a variable definition. Myers and Majluf (1984) close the branch by converting adverse selection into a financing hierarchy. In their setup, a manager acting for existing shareholders issues equity only when the shares are dear, so an equity issue is bad news in equilibrium, the announcement moves the price against the issuer, and the firm prefers internal funds to debt and debt to equity; financial slack has value because without it good projects are passed up (Myers and Majluf, 1984). The pecking order is a theorem about one specific information structure, and its authors said as much, calling it one story among many and flagging the passive-shareholder assumption as load-bearing (Myers and Majluf, 1984). The field canonized the hierarchy and dropped the modesty.

The control branch is older than its anchor and more combative. Manne (1965) supplies the premise: control of a corporation is a valuable asset in its own right, the share price is an index of managerial performance, and the takeover is the mechanism by which bad management is priced and displaced (Manne, 1965). Note what Manne is remembered for and what he wrote: an antitrust intervention, arguing against per-se hostility to mergers, whose governance kernel outlived its policy occasion (Manne, 1965). Everything in the control branch runs on Manne’s premise, and the premise goes essentially unexamined for two decades: if price does not track managerial efficiency, the disciplinary reading of takeover returns collapses into a repricing story. Grossman and Hart (1980) then do to Manne what Miller did to the trade-off: accept the mechanism’s premise and break its arithmetic. Proper management is a public good among dispersed shareholders. Each small holder, asked to tender at a price below the post-raid value, prefers to hold and free-ride on the raider’s improvement; so the raider must pay full post-raid value, captures no surplus, and, facing any positive cost, stays home (Grossman and Hart, 1980). Value-improving raids do not occur unless the corporate charter deliberately permits dilution of minority holders, self-dealing salaries, asset sales at favorable prices, squeeze-outs, so that the raider can be paid in coin taken from the shareholders who do not tender (Grossman and Hart, 1980). Sit with that result: the disciplining device works only if shareholders pre-authorize their own expropriation. The takeover mechanism’s fuel is a private benefit orthogonal to efficiency.

The takeover wars of the 1980s were the era’s public dispute, discipline against looting, and no academic argument in this history spilled into the newspapers, the Delaware courts, and the Congress the way this one did. The discipline side had the numbers. Jensen and Ruback’s 1983 survey aggregates the event studies: successful targets earn abnormal returns around 30 percent in tender offers, a weighted average of 29.1 percent across seven studies whose individual estimates run from 16.9 to 34.1 percent, and about 20 percent in mergers; bidders earn roughly 4 percent and zero respectively; only unsuccessful bidders lose (Jensen and Ruback, 1983). The authors read the gains as real wealth creation from the transfer of control rights, argue affirmatively against the monopoly-power alternative, and report that managerial actions harmful to shareholders are hard to find except those that eliminate a bidder, targeted repurchases and standstill agreements (Jensen and Ruback, 1983). Jensen (1986) supplied the motive and made it operational: free cash flow, cash beyond what positive-NPV projects can absorb, which managers of low-growth empires retain and waste rather than disgorge. Debt is the fix, the “control hypothesis”: a dividend promise can be cut with impunity, but missed debt service hands creditors the right to force the question, so exchanging debt for equity bonds the manager’s promise in a way no announcement can, and the market prices the bonding, with leverage-increasing exchange offers earning two-day gains of up to 21.9 percent (Jensen, 1986). The oil industry was the exhibit: the ten largest oil firms held $48.5 billion of 1984 cash flow, 28 percent of the total for the top 200 firms in Dun’s Business Month survey, in an industry that needed to shrink; Jacobs estimated some $200 billion of recoverable inefficiency across the sector; the Gulf, Getty, and Conoco takeovers created shareholder gains exceeding $17 billion (Jensen, 1986). On this reading the LBO is the theory made flesh: a transaction whose entire design, concentrated equity, crushing leverage, managers forced to own what they run, is a Jensen-Meckling ownership structure imposed by force on exactly the firms Jensen’s diagnosis selected, cash-rich, low-growth, and public.

The looting side had the theorem. Grossman-Hart’s free-rider result says the discipline story cannot pay for itself; what pays is dilution, the raider’s charter-sanctioned ability to extract value from the shareholders who do not tender, and a mechanism fueled by dilution selects for extraction as readily as for improvement (Grossman and Hart, 1980). The looting reading of the 1980s did not need to deny the target premia; it needed only to ask where the premium comes from, improvement or transfer, and the era’s event studies cannot distinguish the two, because both raise the target’s price. And the discipline side’s own scoreboard contains the anomaly the theorem predicts: bidder returns of roughly zero, the number Jensen and Ruback themselves flagged as the finding that aged least well, later pushed negative by the hubris literature (Jensen and Ruback, 1983). A disciplinary market in which the disciplinarians earn nothing is exactly what the free-rider theorem says a disciplinary market must look like, when it exists at all. The theory predicted the empirical soft spot before the empiricists conceded it. I score the takeover wars a draw with asymmetric honors: the survey is right about realized target gains, the theorem is right about why the mechanism is not self-financing, and the two are answers to different questions, what happened to prices and who could have paid for it, that the era mistook for one.

The era also never reconciled its own two branches, and the two share more than a premise; they share a hinge paper. Myers (1977) belongs to both: debt overhang is simultaneously an information-era underinvestment result and, as Myers himself framed it, an agency cost induced by risky debt, in explicit alignment with Jensen-Meckling (Myers, 1977). Yet the branches assign debt opposite signs. In the information branch, debt is a credential: Ross’s high types lever up, Myers-Majluf’s firms issue debt because it signals less bad news than equity. In the control branch, debt is a leash: Jensen’s managers are bonded by it against their will. Same instrument, opposite valence, no model in the era carries both frictions at once. A field that cannot say whether debt is a boast or a muzzle has not yet found its primitive. And the deepest gap is the one the era’s prescription papers walk past. Align incentives, says agency theory; write the contract. But Jensen-Meckling deferred the question of who holds authority when the contract runs out; Grossman-Hart showed the takeover answer requires a charter provision allocating a control-contingent private benefit; Myers-Majluf’s whole result turns on an unmodeled assumption about which shareholders act (Jensen and Meckling, 1976; Grossman and Hart, 1980; Myers and Majluf, 1984). Agency theory says align incentives, but who decides when contracts are silent? The authors of the free-rider theorem spent the next decade answering exactly that question, and their answer opens the next era.

Era IV: Power Frictions (1986-2003)

Eras III and IV share a calendar decade; they differ by question, not by date: Era III asks how to align incentives, Era IV asks who holds the decision right when contracts are silent. The era’s claim, stated so it can be fought: the value of a financial claim is determined by who controls the decisions the contract left unspecified, not by the cash flows the contract promised. A security is a bundle of rights before it is a bundle of cash flows. The claim can be attacked at its foundation, and was: the incompleteness it rests on sits on a knife-edge assumption (investments non-contractible ex ante, contractible ex post) that Grossman and Hart conceded and Maskin and Tirole later pressed (Grossman and Hart, 1986). It can also be attacked from below, by the era’s own second wing: rights are worth only what a court will enforce, so the primitive is not the contract but the legal system standing behind it (La Porta et al., 1998).

The era’s founding paper is Grossman and Hart (1986), and its founding move is a definition. Ownership is the purchase of residual control rights: the right to decide every use of the asset not explicitly assigned by contract (Grossman and Hart, 1986). When contracts are complete, ownership is empty, MM logic extended to control. When contracts are incomplete, ownership is everything, and it is symmetric in a way the era’s predecessors missed: giving party A the residual right necessarily takes it from party B, so integration does not remove the distortion, it relocates it. Under firm-i control, manager i overinvests and manager j underinvests in relationship-specific assets, because the party without the residual right anticipates being held up in the ex post bargain (Grossman and Hart, 1986). There is no allocation that gets both margins right; the theory of the firm becomes the theory of the least-bad assignment of a scarce, non-neutral right. Hart and Moore (1990) generalize from two wealthy owners to many agents, including employees who own nothing, and in doing so isolate what ownership actually is: the right to exclude others from the asset, which determines each agent’s outside option in the ex post bargain and therefore each agent’s ex ante willingness to invest in the relationship (Hart and Moore, 1990). The prescriptions follow from indispensability and complementarity, and their worked example prices the logic exactly: a chef facing a replaceable skipper-owner gets one-third of a 240 surplus (below his 100 cost, so he does not invest), while facing an indispensable tycoon-owner he gets half (so he does) (Hart and Moore, 1990). Who should own the yacht is decided by whose investment the ownership pattern protects, and the answer changes with nothing but the bargaining geometry.

Aghion and Bolton (1992) turn this apparatus into corporate finance by dropping one assumption, and the assumption they drop is wealth. Grossman-Hart-Moore agents are rich enough to buy any asset; the interesting financing problem begins when the entrepreneur is penniless (Aghion and Bolton, 1992). Then control cannot simply be sold to whoever values it most; it must be pledged, and the optimal pledge is contingent: the entrepreneur holds control in good states, the investor seizes it in bad ones. That contingent allocation is what debt is (Aghion and Bolton, 1992). This is the era’s central theoretical object, and it is worth pausing on how different it is from every earlier derivation of debt. Debt is no longer a tax shield, a signal, or a bond against free cash flow; it is a state-contingent constitution.

The era claimed ancestors, and the claim is a reversal performed on the dead. Townsend (1979) had derived debt from costly state verification: audit only when the borrower reports low, pay a flat amount otherwise (Townsend, 1979). Gale and Hellwig (1985) proved the standard debt contract optimal on the same primitive (Gale and Hellwig, 1985). Both papers understood themselves as information economics in the principal-agent tradition; Townsend called his paper a first step and conceded it could not explain the coexistence of shares and debt (Townsend, 1979). The control era reread “audit in default” as “control shifts to the lender in default” and installed both papers as its antecedents, a reading their machinery never contained (Townsend, 1979; Gale and Hellwig, 1985). DeMarzo and Duffie (1999) later derived debt a third time, from adverse selection, as the least information-sensitive monotone claim a liquidity-motivated issuer sells (DeMarzo and Duffie, 1999). Hold the three derivations side by side, because their coexistence is the era’s quiet catastrophe: costly verification, incomplete contracts, and adverse selection, three mutually independent frictions, each delivering the same security as its optimum. Debt is over-determined. Observing the world full of debt tells you that a friction binds; it cannot tell you which. The era gained explanatory completeness and lost falsifiability in the same decade.

The era’s second wing relocated the primitive from the contract to the courthouse, and the relocation happened through a survey, which is itself worth noticing: the field’s direction changed not through a new theorem but through a reframing of what the accumulated theorems were about. Shleifer and Vishny’s 1997 survey compresses governance into a single question: how do suppliers of finance get anything back from managers who control sunk, specific capital (Shleifer and Vishny, 1997)? The question is pure incomplete-contracts thinking, the survey stands in the Grossman-Hart-Moore tradition rather than the older agency one, and its answer names two mechanisms, legal protection of investors and concentrated ownership, with the observation that large investors substitute for weak law (Shleifer and Vishny, 1997). What made the survey a hinge rather than a summary was its two confessed puzzles: why legal protection varies across countries, and who holds equity where protection is weak. Those two open questions, listed as loose ends, became the next research program. Their refusal to rank the American, German, and Japanese systems, because the evidence does not tell us, is the honest statement of where the field stood (Shleifer and Vishny, 1997). La Porta, Lopez-de-Silanes, Shleifer, and Vishny then made investor protection a variable. External stock-market capitalization held by outsiders is 60 percent of GNP in common-law countries and 21 percent under French civil law; IPOs per million people, 2.2 against 0.2; moving a country’s anti-director-rights index from the French-origin average (1.76) to the common-law average (3.39) is associated with roughly 19 points of capitalization to GNP (La Porta et al., 1997). Law and Finance codified the index and the legal-origin classification across 49 countries, argued that legal origin, transplanted by conquest and colonization, can be treated as exogenous, and stated the wing’s theoretical challenge plainly: the rights that make a security valuable are conferred and enforced by law, so an identical contract means different things in Paris and Delaware (La Porta et al., 1998). This is aimed at the control wing’s heart. Aghion-Bolton’s contingent constitution presupposes the enforcement LLSV measure. The wing’s later history belongs to a different volume, the index was recoded, the exogeneity contested, the authors’ own hedge (that their measures capture some though not all of the gap) proved prophetic (La Porta et al., 1997; La Porta et al., 1998), but the relocation of the question survived every recoding: after LLSV, no theory of financial contracting could take the courthouse for granted.

The era closed with an empirical victory, and the victory is genuine, jurisdictionally captive, and routinely oversold, all three at once. Kaplan and Strömberg (2003) did something the field had spent four eras not doing: they read the contracts. Two hundred thirteen venture-capital financings into 119 portfolio companies, clause by clause, and the incomplete-contracts theory turns out to be written into the documents. Cash-flow rights, board rights, voting rights, and liquidation rights are allocated separately, not bundled the way a textbook security bundles them; the allocations are contingent on observable performance, with the entrepreneur gaining rights as milestones are met and the investor gaining them as they are missed; 18 percent of financings carry state-contingent board provisions that hand the investor board control in bad states, transfers “not triggered simply by default on a debt payment,” which is Aghion-Bolton’s mechanism rather than Townsend’s; and convertible preferred, the instrument in 204 of 213 rounds, is best read as the device that unbundles all four rights inside one security (Kaplan and Strömberg, 2003). Aghion-Bolton’s blackboard object exists, coded by lawyers who never read the paper. That is as clean as empirical vindication gets in this field. But the sample is one country, drafted under the world’s strongest enforcement, and it holds the legal regime fixed by construction; it validates the control wing and is silent on the legal wing (Kaplan and Strömberg, 2003). The era’s dispute, contract or courthouse, ends adjudicated on exactly one side. Kaplan and Strömberg thought they were generating stylized facts for future theory and scored several theory classes as performing well; the field kept the incomplete-contracts vindication and discarded the rest of the report card (Kaplan and Strömberg, 2003), the era’s own reversal pattern applied to its final paper.

The era’s reversals run in one direction, and the direction is diagnostic. Townsend and Gale-Hellwig, information economists, were reread as control theorists. Grossman-Hart thought it was repairing a gap in transaction-cost economics and conceded its insurance-industry evidence lacked quantitative completeness; it became the founding document of property-rights theory (Grossman and Hart, 1986). Hart-Moore thought it was answering Coase on the boundaries of the firm; it became the microfoundation for security design and law-and-finance, applications outside its stated frame (Hart and Moore, 1990). Aghion-Bolton offered a “pecking-order theory of governance structures”; the field kept the contingent-control result and let the framing fade (Aghion and Bolton, 1992). DeMarzo-Duffie wrote a liquidity theory of security design; it is remembered as an adverse-selection pillar and, in a flag its authors raised only in passing, the charter for securitization and tranching (DeMarzo and Duffie, 1999). Every paper in the era was promoted toward the control-rights center of gravity, beyond its authors’ stated intent. When a field systematically enlarges its papers in one direction, the direction is telling you what the field believes its own question is.

Four eras of theory now stood complete, and their joint product was an embarrassment of sufficiency: taxes could explain leverage, signaling could explain leverage, agency could explain leverage, control allocation could explain leverage, and debt itself had been derived from three incompatible primitives. Every observed capital structure was consistent with something. Four eras of theory, and the field still could not tell which friction binds in the data. The next era was not another friction. It was the discovery that the field’s measuring instruments could not answer the question either.

Era V: The Measurement Crisis (1988-2010)

The era’s claim is not that capital markets are imperfect; every era since 1963 claimed that. The claim is harsher: the regressions the field had been running to detect its frictions were uninterpretable, because the coefficients everyone read as measures of a friction did not measure that friction monotonically or at all. MM fails, this era says, but for two decades the profession could not prove it failed for the reasons it claimed. A reader can fight this: one can hold that the original designs were sound and the critiques idiosyncratic. The record of the era’s three disputes is the answer, and this section is organized around the disputes, not the papers.

First dispute: the sensitivity war. Fazzari, Hubbard, and Petersen (1988) made the asymmetric-information theory operational with one proxy. Sort firms by dividend payout; low-payout firms are presumptively constrained; read the cash-flow coefficient in an investment regression as the tightness of the constraint. The gradient came out large and monotone, 0.461, 0.363, 0.230 across the three payout classes, with a t-statistic of 12.1 on the difference between the extremes, and it survived an unusual battery: instrumenting with lagged Q, first and second differences, the sales accelerator, the neoclassical cost of capital, industry splits, balance-sheet liquidity (Fazzari et al., 1988). FHP’s best defense is genuinely good, and it should be stated the way they stated it: they never claimed the coefficient’s level measured anything, only that its ordering across pre-sorted classes tracks a constraint story and embarrasses a frictionless one, since under perfect markets retention behavior should predict nothing about investment and no measurement-error story explains why the bias should line up with payout policy (Fazzari et al., 1988). Kaplan and Zingales (1997) attacked the premise underneath the ordering, and they attacked it inside FHP’s own sample, which is what made the attack undeniable. Reading the annual reports and 10-Ks of FHP’s 49 low-dividend firms year by year, they classified each firm-year by whether the firm could in fact have invested more, and found the ordering inverted: the least constrained firms show the highest sensitivities, 0.702 for the never-constrained against 0.340 for the likely-constrained and 0.180 for the possibly-constrained, with the fifteen firms never constrained in any subperiod reaching 0.779 and Hewlett-Packard alone running between 0.97 and 1.15 (Kaplan and Zingales, 1997). In roughly 85 percent of firm-years the firms could have invested more had they chosen to, and nearly 40 percent of the firms, HP included, could have done so in every single year; a purely quantitative re-sort on interest coverage reproduces the inversion, so the result is not an artifact of KZ’s judgment calls (Kaplan and Zingales, 1997). These are not distressed firms mismeasured; they are firms with slack, sorted into the “constrained” bin by a proxy that mistakes a payout choice for a constraint. The theoretical companion is worse for the proxy than the reclassification: the sensitivity need not be monotonic in constraint tightness for entirely standard production functions, so even a perfect sort would not license the inference (Kaplan and Zingales, 1997). The dispute did not resolve on substance; FHP replied, KZ confined their inversion to the FHP sample and left its generality open, and the deep question KZ named, why the least constrained firms behave as if internal cash matters, stayed unanswered (Kaplan and Zingales, 1997). It resolved on method, which is the era’s signature outcome: after KZ, a high investment-cash flow sensitivity, by itself, stopped counting as evidence of a financing constraint. The field’s most-used friction detector was decommissioned without anyone establishing what the needle had been measuring.

Second dispute: models versus CFOs. Graham and Harvey (2001) put the question to the people the theories are about, 392 CFOs, and the answers embarrassed both leading theories at once. Read the ranking as a referendum on this history’s first five eras. Financial flexibility ranks first among capital-structure concerns (59.4 percent rating it important) with credit ratings second (57.1 percent); the interest tax shield, Era II’s horse, rates a middling 2.07 on their scale; distress costs, Era II’s rabbit, 1.24; personal taxes, Miller’s whole equilibrium, 0.68; the free-cash-flow discipline of debt, Era III’s proudest export, 0.33, a rejection of Jensen from the corner office (Graham and Harvey, 2001). Every friction the theory eras proposed sits at the bottom of the practitioners’ list, and the top of the list is the reason MM gave in 1963 for not borrowing the maximum. NPV and the CAPM are meanwhile ubiquitous in project selection, so the managers are not innocent of the theory; they use the tools and disbelieve the frictions (Graham and Harvey, 2001). Both trade-off and pecking order lose to an answer, flexibility, that neither theory dignified with microfoundations, though MM had listed it in 1963 and Graham-Harvey’s CFOs merely repeat it back across four decades. The authors hedged that surveys measure beliefs, not actions (Graham and Harvey, 2001); the field discarded the hedge and made the paper the canonical citation for the theory-practice gap. I read the dispute less as models-versus-managers than as a second decommissioning: the first dispute retired the field’s favorite regression, this one retired the presumption that the canonical frictions were the ones doing the work.

Third dispute: rational versus behavioral managers. With the firm-level proxy discredited, the residual variation had to live somewhere, and three papers moved it into the manager. The move was overdetermined: Era V had shown that firm characteristics explain a modest share of policy variation and that the explained share resists structural interpretation, so the field went looking for the missing variance in the one place its regressions had never controlled for, the person making the decision. The three papers disagree about what a manager is, a market-timer, a psychology, or a fixed effect, and the disagreement is the dispute. Baker and Wurgler (2002) put it in history: leverage is the cumulative residue of past attempts to time the equity market, their external-finance-weighted market-to-book is the strongest single determinant of leverage, and the effect persists at horizons past ten years (Baker and Wurgler, 2002). The strong version, that there is no target and capital structure is just sediment, did not survive; later work found partial rebalancing, and the field kept only the finding that history matters (Baker and Wurgler, 2002). Malmendier and Tate (2005) put it in psychology, and their instrument deserves its fame: overconfidence measured from revealed portfolio behavior, CEOs who hold their own exercisable options to expiration, a bet on the firm no diversified rational agent makes. Overconfident CEOs show investment-cash flow sensitivity of 0.2339 above their peers, and the effect concentrates in equity-dependent firms, 0.4990 in the top constraint quintile (Malmendier and Tate, 2005). Note the recursion: the behavioral program’s evidence is the very sensitivity KZ had discredited as a constraint measure, now reinterpreted as a trait detector. The proxy outlived its meaning. Bertrand and Schoar (2003) put it in identity: tracking managers across firms, manager fixed effects raise the explained variance of policy regressions by over four points of R-squared beyond firm effects, and the authors explicitly disclaim causality since executives are not randomly assigned (Bertrand and Schoar, 2003). The era’s honesty pattern repeats: KZ confined their claim, RZ called the theory-proxy link weak, Bertrand-Schoar refused the causal reading, and in each case the field’s citation practice stripped the disclaimer. The bar was raised by authors policing their own overreach and lowered by readers ignoring the police report.

The era’s reversal pattern is a single reversal repeated seven times, and stating it once saves stating it seven times: every constructive contribution was absorbed as a negative or methodological result. FHP offered a proxy for a friction and became the founding specimen of the crisis, the paper whose identifying assumption everyone attacks (Fazzari et al., 1988). KZ answered one narrow question about one sample and became a general teaching lesson, its scope caveats ignored (Kaplan and Zingales, 1997). Graham-Harvey offered a Lintner-style descriptive survey and became a license to weigh CFO beliefs against models (Graham and Harvey, 2001). Baker-Wurgler proposed a theory and kept only a fact (Baker and Wurgler, 2002). Malmendier-Tate proposed a friction and bequeathed a measurement paradigm, the managerial trait read from revealed portfolio behavior (Malmendier and Tate, 2005). Bertrand-Schoar, the exception that proves the rule, was received exactly as framed, an existence result, and became a design template (Bertrand and Schoar, 2003). The field, in short, spent the era converting answers into instruments.

Under the three disputes ran the methodological hinge. Rajan and Zingales (1995) took the U.S. leverage regularities abroad and found the G-7 looks alike once accounting is harmonized (Germany’s adjusted debt-to-capital falls from 38 to 18 percent, Japan’s from 53 to 37) and that four factors are robust everywhere: tangibility positive, market-to-book negative, size positive, profitability negative (Rajan and Zingales, 1995). That last sign is the era’s thorn: the trade-off theory’s most direct prediction is that profitable firms, with more taxable income to shield and lower distress risk, should borrow more, and in every country they borrow less. The four-factor specification became the standard controls of every leverage paper since, while the authors’ own verdict, that the theoretical underpinnings of the correlations remained unresolved, became the era’s epitaph (Rajan and Zingales, 1995). Rajan and Zingales (1998) then supplied the identification template: industries that depend on external finance grow disproportionately faster in countries with developed financial systems, an industry-country interaction design with both fixed effects absorbed, the credibility revolution imported into finance in one specification (Rajan and Zingales, 1998). The era’s legacy is that design, not any answer: a raised identification bar, with the substantive question, which friction binds, still open. And the bar rose at a strange moment. The era learned to measure the public industrial firm exactly as the public industrial firm began to leave the sample. Identification got clean just as the object of study started disappearing.

Era VI: The Firm Escapes (2008-present)

Canonical corporate finance solved the wrong firm. That is this era’s claim, stated at full strength so it can be refused: the theory built across five eras assumed a dispersed-ownership public industrial firm choosing its capital structure against tax, information, and contracting frictions, and by the time the field’s instruments could credibly measure which friction binds, the marginal firm was private, intangible-heavy, and funded through intermediaries whose own balance sheets are the binding constraint. The refusal is available and respectable: the public firm shrank in count but not obviously in economic weight, and the frictions found where they bind hardest, small bank-dependent borrowers, may be the same frictions the canon named, measured at last where they are largest. The era’s evidence sits on three fronts, and the fronts share a destination.

Financing becomes macro. The front’s methodological signature is one estimator, and the estimator is the era’s answer to the measurement crisis. Khwaja and Mian (2008) exploit Pakistan’s 1998 nuclear-test shock to dollar deposits: the same firm borrowing from two banks, one hit harder than the other, with firm fixed effects absorbing everything about the firm’s demand for credit, so the difference across its lenders identifies supply. A one percent fall in a bank’s liquidity cuts its lending to the same firm by 0.6 percent; OLS without the fixed effects understates the channel at 0.46, and the gap between the two numbers is a measure of exactly the demand-supply confounding that Era V taught the field to fear (Khwaja and Mian, 2008). The distributional finding matters more than the average: large, politically connected, conglomerate-attached firms substitute across lenders almost fully (aggregate borrowing coefficient 0.04), while small firms cannot substitute at all (0.87) and eat the shock (Khwaja and Mian, 2008). Note what the estimator does philosophically: it identifies the friction by throwing away all variation in the firm’s own choices. Five eras of theory about the firm’s decision, and the cleanest identification in the field’s history comes from a design in which the firm’s decision is a nuisance parameter. Chodorow-Reich (2014) carried the design to U.S. employment through the crisis. The friction that makes it work is relationship stickiness: firms stay with their pre-crisis syndicates at rates seven times what lender market shares would predict, so a firm’s exposure to Lehman-damaged lenders is as good as predetermined (Chodorow-Reich, 2014). The dispersion in lender health was enormous, U.S. Bancorp cut syndicated lending about 14 percent after Lehman against Credit Suisse’s 79 percent, and firms attached to healthy syndicates were nearly four times likelier to get a crisis loan; a move from the 10th to the 90th percentile of lender health is worth 4 to 5.5 points of employment growth, with credit disruption accounting for between one-third and one-half of the employment decline at small and medium firms in the sample, a partial-equilibrium figure the author labels as such (Chodorow-Reich, 2014). The effect concentrates where theory says it must: significant for small, opaque, bond-market-excluded firms, absent for large ones (Chodorow-Reich, 2014). The intermediation theory this work reaches back to was written decades earlier and ran as a parallel literature, and one paragraph of acknowledgment is owed rather than a separate track. Diamond and Dybvig (1983) showed the deposit contract that supports optimal risk sharing carries a second, self-fulfilling run equilibrium, with deposit insurance as the remedy; Diamond (1984) turned “banks are special” from assertion into theorem, the intermediary as delegated monitor whose diversification across many projects drives the delegation cost toward zero and makes the arrangement incentive-compatible; Holmström and Tirole (1997) made intermediary capital an aggregate state variable and drew the map the crisis literature would need, separating shocks to firm collateral from shocks to intermediary capital and showing each hits a different class of borrower (Diamond and Dybvig, 1983; Diamond, 1984; Holmström and Tirole, 1997). Corporate finance did not absorb these papers when they were written; it absorbed them after 2008, when intermediary health became the visible driver of firm outcomes, and the absorption is the era’s clearest reversal: a literature the field had filed under “banking” for twenty-five years was retrofitted as the microfoundation of corporate finance’s newest front. The reversal here is temporal rather than substantive, the papers meant what they said, and the field simply took a generation to decide they were corporate finance.

Theory goes dynamic and structural. The second front turned the field’s accumulated “anomalies” into artifacts, and its instrument was a discipline the field had never imposed on itself: make the model generate data, then treat the generated data exactly as an empiricist would treat real data. Strebulaev (2007) calibrated a dynamic trade-off model in which firms refinance infrequently because refinancing is costly, simulated economies from it, and ran the profession’s own regressions on the output. The model reproduces the negative profitability-leverage slope (around minus 0.32, overlapping the empirical range) that a generation had read as the trade-off theory’s refutation, more than ten standard errors below zero, and it reproduces Welch’s leverage-passivity coefficient almost exactly (Strebulaev, 2007). The mechanism is mundane and devastating: between refinancings, profits accumulate in equity, so profitable firms mechanically drift to low leverage even though every firm in the model is a textbook trade-off optimizer; the cross-section is dominated by firms optimally away from their targets, and comparative statics computed at refinancing points were never the right predictions for it. The tests did not mean what they said, which is this era’s echo of Kaplan-Zingales, aimed at the other camp. Hennessy and Whited (2005) reached the same verdict by an independent route, indirect inference on a dynamic model with no target leverage ratio, reproducing the anomalous signs (Hennessy and Whited, 2005); two methods sharing no machinery converging on the same acquittal is why the acquittal stuck. Their 2007 structural estimation then put magnitudes on the frictions themselves: external equity costs equivalent to an 8.3 percent underwriting fee on average, 10.7 percent for small firms against 5.0 for large, and bankruptcy costs of 10.4 percent of capital (Hennessy and Whited, 2007). Their subtlest result rearms Era V’s dispute: raising true financing costs in the model makes the standard constraint indexes, including the Kaplan-Zingales index, decline, because a high-cost firm substitutes toward equity and looks unconstrained to the proxy (Hennessy and Whited, 2007). The constraint measures are equilibrium outcomes, not gauges. DeMarzo and Sannikov (2006) announced the front’s theoretical reach: an optimal dynamic contract under hidden cash flows implemented by a credit line, long-term debt, and equity, with leverage falling in past profitability (DeMarzo and Sannikov, 2006). The claim here is limited to the paper’s announced contribution rather than treated as a full reconstruction of the model.

The era’s defining dispute sets this front against the first, and the two do not meet on common ground, which is the finding. Strebulaev’s position, at full strength: cross-firm reduced-form regressions are uninformative, since a one-friction model mechanically reproduces them. The credit-supply position, at full strength: the structural parameters are estimated on Compustat panels of public firms, exactly the population that shrank 39 percent from its peak, while the binding friction lives in cross-lender variation the structural models do not represent at all (Hennessy and Whited, 2005; Doidge et al., 2017; Khwaja and Mian, 2008). One program denies the informativeness of cross-firm variation; the other builds causality on cross-lender variation. They agree, from opposite directions, on a single empirical regularity: the friction concentrates in small firms (the structural estimates say small firms pay roughly double for external funds (Hennessy and Whited, 2007); the credit-supply estimates say small firms cannot substitute across lenders), and they disagree about what the friction is, a flotation-and-bankruptcy wedge or an inability to replace a damaged lender. The era ends with its two best programs answering different questions and sharing only a size gradient.

The modal firm changes. The third front is three separate breaks in the canon’s object, unified by locating the break outside the capital-structure choice. Doidge, Karolyi, and Stulz (2017) count the disappearance: U.S. listings peaked at 8,025 in 1996 and stood at 4,101 in 2012, a drop of roughly half per capita, and a listing gap opens relative to what U.S. characteristics predict from cross-country regularities. The decomposition matters: the gap is driven by delists, mostly acquisitions, and by a collapse in new lists, and the convenient explanation fails on their evidence, since the private-equity share of public-firm acquisitions is no higher after 1996 than before (Doidge et al., 2017). The authors’ own conclusion is the era’s thesis said flatly: the existing paradigm does not explain the gap, another is needed (Doidge et al., 2017). Kaplan and Schoar (2005) measured the form that absorbed the escapees, and the measurement deflates the escape narrative: private-equity funds return roughly the public market net of fees (PME 0.96 equal-weighted, 1.05 value-weighted), so the limited partners who fund the private firm earn approximately what an index fund pays, while performance persists across a GP’s successive funds (coefficient 0.54, strong in venture at 0.69, weak in buyouts at 0.17), a persistence public-market vehicles do not show and a sign the surplus is real and captured at the GP level (Kaplan and Schoar, 2005). The firm did not escape to a form that pays its investors more; it escaped to a form whose intermediaries capture the difference, which is the era’s thesis in miniature. Crouzet and Eberly (2023) broke the valuation-to-investment mapping, which is a quieter way of saying they broke Q theory for the modern firm: with intangible capital and rents treated jointly in a Q-plus framework, measured valuations reflect rents and unmeasured capital as well as investment incentives, the two are confounded when estimated separately, and intangibles account for something between 40 percent and two-thirds of the post-1990 gap between valuations and investment (Crouzet and Eberly, 2023). The regression that Era V fought over, investment on Q and cash flow, presumed a firm whose capital is on its balance sheet; the modal firm’s capital increasingly is not. Azar, Schmalz, and Tecu (2018) broke the objective function itself: when the same diversified institutions own competing airlines, route-level concentration measured through common ownership (the MHHI delta) predicts fares 3 to 7 percent higher, 10 to 12 in their strongest design built around the BlackRock-BGI merger, than separate ownership would deliver (Azar et al., 2018). The airline point estimate is contested and now widely regarded as fragile, and should be read as contested (Azar et al., 2018); the conceptual point survives the fragility, because a firm whose marginal shareholder holds the industry does not maximize own value, and nothing in five eras of theory tells us what it maximizes instead. Era III assumed the shareholder wants value maximized and the manager resists; the common-ownership firm inverts the cast, with a manager who would happily maximize and a shareholder base whose portfolio prefers he did not.

Each front subtracts one assumption the canon never listed as an assumption: that the firm’s lender is interchangeable, that the firm’s capital is tangible and priced, that the firm is public, that the firm maximizes its own value. None of these was an assumption in the sense that the risk class or the tax rate was an assumption; they were the background against which assumptions were stated, which is why no era thought to relax them. I date the era’s opening to 2008 rather than to any of its papers because the crisis is what converted the background into foreground: a year in which firm outcomes visibly depended on lender identity made Khwaja-Mian’s estimator, published months before Lehman, retroactively the era’s founding method (Khwaja and Mian, 2008). The question the era leaves is whether anything remains of the canonical firm for the canon to be about. What remains, on the evidence of every front at once, is the party the frictionless benchmark deleted at birth: the financier in the middle. The era’s own tools point past the firm to the intermediary as the unit of analysis, and that is where the history of asset pricing, told the same way, already stands.

Epilogue: Two Fields, One Balance Sheet

Asset pricing and corporate finance are usually told as separate histories, one about the demand for securities, one about their supply. Told side by side, they are the same history run on the two sides of one balance sheet, and they end in the same place.

Both fields were founded by an irrelevance result that deleted the financier. In asset pricing, the frictionless consumer prices every claim through a stochastic discount factor; intermediaries are a veil. In corporate finance, MM made the liability side a veil: the cream and the skim milk fetch the whole milk’s price, so it cannot matter who holds which claim, or through whom. Both foundations were arbitrage arguments, both were right about their assumptions, and both fields then spent decades cataloguing which assumption fails. Asset pricing’s catalogue ran through preferences, beliefs, and market structure; corporate finance’s ran through taxes, information, agency, control, and measurement, the eras of this history. The catalogues were compiled independently, by different people, in different journals, and they converged on the same missing character.

Corporate finance arrived first from the theory side without noticing. Leland and Pyle’s afterthought section suggested that the logic of retained-equity signaling implies an intermediary that certifies by holding (Leland and Pyle, 1977); Diamond built the theorem (Diamond, 1984); Holmström and Tirole made the intermediary’s own capital the state variable (Holmström and Tirole, 1997). Corporate finance filed all of it under banking and went on refining theories of the firm’s own choice. It took 2008 to force the filing error into the open: Khwaja-Mian and Chodorow-Reich showed that which bank a firm borrowed from, and how healthy that bank’s balance sheet was, moved lending and employment at the firm level, with the smallest firms unable to substitute away (Khwaja and Mian, 2008; Chodorow-Reich, 2014). The firm’s outcomes were priced off its lender’s balance sheet. Asset pricing, over the same years, reached the mirror conclusion: the marginal investor in most asset classes is a levered institution, and the price of risk moves with intermediary capital. A history of asset pricing ends at “whose m, with how much capital.” A history of corporate finance ends at “whose credit, through which intermediary.” These are the same sentence with the balance sheet read from opposite sides: an asset pricer asks whose discount factor prices the claim, a corporate financier asks whose balance sheet originates it, and both answers name the intermediary.

The convergence reframes each field’s founding theorem. MM irrelevance and frictionless pricing were never two benchmarks; they are one benchmark stated twice, the assumption that the middle of the financial system is passive plumbing. Every era of this history proposed a friction inside or around the firm: the tax collector, the informed manager, the self-serving manager, the silent contract, the uninterpretable regression. Each explained something; none survived contact with the data as the binding constraint. The friction that survived the measurement crisis best was not inside the firm at all, it was the identity and health of the firm’s lender, a variable no era before the sixth thought to put on the right-hand side. There is a lesson about theory choice in that sequence. The field spent its middle decades multiplying frictions faster than it could test them, until debt itself was over-determined and every capital structure was consistent with something; the friction that finally identified cleanly was the one attached to an actor with its own balance sheet, its own shocks, and its own paper trail.

Meanwhile the firm itself slipped the field’s grasp, going private, going intangible, funded by funds: Doidge-Karolyi-Stulz counted the exits, Kaplan-Schoar showed the private form returns the public market net of fees while its general partners keep the difference (Doidge et al., 2017; Kaplan and Schoar, 2005). The modal financing arrangement of the coming decades, a private, intangible-heavy firm funded by a levered intermediary that is itself funded by institutions, contains three balance sheets where the canonical theory modeled one. Diamond and Dybvig’s closing observation, that any institution holding illiquid assets against very short-term claims inherits run risk (Diamond and Dybvig, 1983), applies to at least two of the three, and the argument stops at the chartered bank’s edge; following the run into the shadow system would require a different evidentiary base.

So the two histories end as one instruction. Seventy years were spent discovering, from both directions, that the frictionless benchmark’s missing ingredient was the financier in the middle. Neither field chose this destination; each was dragged there by its own anomalies, asset pricing by prices its consumers could not rationalize, corporate finance by firms its frictions could not explain, and the independence of the two forced marches is the best evidence the destination is real. The next history of thought in this subject, written a generation from now, will likely not be a history of asset pricing or of corporate finance. It will be a history of intermediation, and both fields’ histories will read as its prologue: two fields, one balance sheet, and the slow recognition that the interesting party was the one standing between the household and the firm, holding both their claims, deciding which of them gets to be frictionless.