while holding bankruptcy costs constant (μ = 0. 12) are shown in Figure 3. As σ
increases and riskier investment projects are being financed, the ω-leverage schedule
shifts up (top left panel). Although increases in the default threshold are very small
for most reasonable values of leverage, they translate, for a given leverage, into a
considerably higher likelihood of default (top right panel). The resulting jump in ex-
pected bankruptcies accordingly causes a significant upward shift in both the external
finance premium and the credit spread schedules over a much wider range of leverage
(bottom panels).
The size of the bankruptcy parameter μ, as well as that of σ ,is an open empir-
ical question. Moreover, the potential variability of these two parameters over the
course of the business cycle has important implications for the behavior of the ex-
ternal finance premium, a key unobservable variable determining the magnitude of
the financial accelerator. To shed light on these questions, we constructed a new
dataset linking firm-level balance sheet variables to credit spreads and market-based
measures of expected default risk, which allows us to estimate directly the volatility
of idiosyncratic risk and the magnitude of bankruptcy costs implied by the BGG
debt-contracting problem for the 1997-2003 period. Estimates of these structural
parameters, in turn, permit us to examine the behavior of the model-implied external
finance premium during the most recent economic downturn.
10