Financial Fragility and Economic Performance

The model’s economy has two periods and agents endowed with a nonconsumable input good w_i (interpretable as net worth), who choose to store it (earning a safe return r), to lend it, or to become entrepreneurs who pay a personal evaluation cost e(θ) to learn a project’s privately-observed success probability p before deciding whether to undertake it, with a successful project paying gross return R > r and a failed one paying zero. Because a project requires exactly one unit of input and most individuals have w_i < 1, external finance is necessary, and the authors stress that “net worth is not measured by the total equity the firm issues,” but by the personal stake of the entrepreneur – the ratio w/(1-w) is the borrower’s balance-sheet position. The paper is explicit that the agency problem it isolates does not require debt as the financing instrument: because the model permits general state-contingent contracts and observable project outcomes, “our analysis assumes contingent contracting and observable returns, [so] our model economy may be interpreted as a pure equity economy; thus, all our results go through even in the complete absence of debt instruments.”

The agency problem rests on two assumptions: a project’s success probability p, once evaluated, is private information to the entrepreneur, and it is not publicly verifiable whether a claimed evaluation actually took place (so non-entrepreneurs could falsely claim to have evaluated and rejected a project). Absent these asymmetries, a social planner’s first-best solution sets the reservation success probability at p_fb = r/R (any project with success probability at or above this threshold should be undertaken, since its expected return just covers the input’s opportunity cost in storage) and chooses the fraction of entrepreneurs m_fb so that the marginal evaluator’s net expected surplus equals the marginal evaluation cost, e(m_fb).

Proposition 4.2 characterizes the optimal contract (when p < r/R) as effectively a debt-like contract: the entrepreneur receives C_0 = rw if the project fails to proceed or is not undertaken, zero consumption (C_u = 0) if an undertaken project fails, and keeps the full upside if it succeeds – a structure forced by two binding constraints, a moral-hazard constraint that payments to non-proceeding entrepreneurs cannot be so generous that non-entrepreneurs are tempted to fake an evaluation, and a limited-liability constraint (C_u ≥ 0) that caps how severely a failed entrepreneur can be punished.* Proposition 4.1 shows this optimal contract always induces an incentive-compatible reservation probability p* that is less than or equal to the social optimum r/R – that is, “the borrower may have an incentive to be insufficiently selective when deciding whether to undertake his project” – because the intermediary’s ability both to reward selective entrepreneurs and to punish unselective ones is constrained by these two conditions.

Proposition 4.3 shows that whenever net worth w is below one, both the equilibrium reservation probability p and the fraction of entrepreneurs m are strictly below their first-best levels, and both are monotonically increasing in w; at w = 1 – full collateralization – the economy exactly attains the unconstrained first-best, p = r/R and m = m_fb.** Intuitively, raising w shifts the curve relating p* to the marginal evaluation cost e(m) up and to the right in the paper’s phase-diagram (Figure 2), so greater collateralization directly reduces how far the agency problem distorts both the quality and the quantity of investment undertaken.

Proposition 4.4 shows that if the unconditional expected gross project return (evaluated at the lowest possible reservation probability, p=0) is less than even the most efficient entrepreneur’s evaluation cost – formally, ∫pR dH(p) - r < e(0) – then there exists a strictly positive threshold level of net worth, w̲, below which the investment market “collapses”: no positive level of investment (m=0) can be sustained in equilibrium at all.* This is the paper’s formal definition of “financial fragility”: as net worth falls toward this threshold, the economy first suffers worsening underinvestment and misallocation (fewer, lower-average-quality projects), and then, below w̲, a complete cessation of investment, which the authors note “declines in borrower net worth induce” directly through the general-equilibrium mechanism rather than through any exogenous credit-rationing assumption.

The paper’s key finding is that “the critical determinant of the potential effectiveness of policy…is the degree of observability of individual entrepreneurial skills”: if entrepreneurial ability (evaluation cost) is completely unobservable to the planner, “policy can do little,” whereas “perfect observability of skill levels implies, remarkably, that policy can drive the economy to the unconstrained social optimum,” with intermediate observability yielding intermediate effectiveness. When skill is wholly unobservable, the best the planner can do is a second-order tax-on-success/subsidy-to-storage scheme that raises p* at the cost of lowering m, and the authors say they “don’t take [this] tax policy…very seriously,” since its net welfare effect “would likely be insignificant” and it rests on the artificial assumption that the planner has literally no information about who the efficient entrepreneurs are.

If the planner can perfectly observe each individual’s evaluation cost, a simple policy – giving a net lump-sum subsidy of (1 - w_i) to every individual whose evaluation cost is at or below the first-best marginal cost e(m_fb) – fully collateralizes exactly the entrepreneurs who should be active, driving the economy to the unconstrained first-best; the authors stress “this result cannot be achieved by the private economy alone, even by using lotteries, coalitions, or other devices,” because “it works because it directly attacks the cause of financial fragility, low borrower net worth.” The paper shows this transfer policy retains real efficacy under weaker observability conditions too – for instance, when the planner can only tell whether an entrepreneur’s cost is above or below the first-best marginal cost, or when entrepreneurs can only understate (not overstate) their own skill, or when entrepreneurs can emit a costly signal of skill (e.g., by taking on debt to build a “track record”) – though with costly signaling it is “never desirable to fully collateralize entrepreneurs,” since the marginal signaling cost of chasing a larger subsidy eventually outweighs the shrinking agency-cost benefit as the economy nears the first-best.

The authors interpret their partial-observability transfer result as “supporting the view that, in the face of large and imperfectly provided for shocks to borrower net worth, debtor ‘bailouts’ by the government are sometimes a useful alternative to letting the financial system collapse or operate at an inefficient level,” while explicitly cautioning that transfers to borrowers are not Pareto-improving (they help borrowers at the rest of the population’s expense) and that they “do not want to be interpreted as favoring a generalized and ongoing bailout policy.” They single out the U.S. in the 1930s as close to the ideal case for a bailout – those needing help were “easily identified,” the shock was “clearly systemic rather than idiosyncratic,” and the New Deal debt-readjustment programs could credibly be presented as “one-time-only” – and argue that the government must weigh the short-run benefits of a bailout against the longer-run cost of “excessive risk-taking and other inefficiencies that the prospect of future bailouts creates,” concluding that “optimal bailouts will be rare and will be in response only to large, systematic shocks.”

The authors connect their mechanism to the severity of the Great Depression (citing Bernanke 1983 on debt-deflation), to postwar episodes such as the agricultural credit crisis, and – more speculatively – to the cross-country link between financial development and economic development, suggesting that “an important cause of the primitive and fragmented state of LDC financial markets is the generally low wealth level of potential borrowers,” and that the historical emergence of a wealthy mercantile class has been “an important pre-condition for the development process.” They also report that a multi-period extension they were pursuing (allowing entrepreneurs to invest repeatedly and maintain ongoing lender relationships) leaves the paper’s basic messages “largely unchanged,” but adds that “financial factors provide an independent source of output persistence,” since current economic shocks affect net worth, which then affects investment efficiency, and hence output, in subsequent periods – a dynamic amplification-and-propagation channel distinct from the static misallocation result emphasized in this paper.