Computed with taste shocks (discrete choice methods) for B′ and d. Fortran parallelized with OpenMP.
Value and default option:
V\left(y, B\right) = \rho_D \log \left\\{ \exp\left[ \dfrac{V^d(y)}{\rho_D} \right] + \exp \left[ \dfrac{V^r(y, B)}{\rho_D} \right] \right\\}
Pr(d=1∣y,B)=exp[ρDVd(y)]+exp[ρDVr(y,B)]exp[ρDVd(y)]=1+exp[ρDVr(y,B)−Vd(y)]1
Default value:
V^d\left(y\right) = u\left[h\left(y\right)\right] + \beta \mathbb{E}_{y'|y} \left\\{ \gamma V\left( y', 0 \right) + (1-\gamma) V^d\left(y'\right) \right\\}
Repayment values:
W(y,B,B′)=u[y−κB+q(y,B′)(B′−(1−δ)B)]+βEy′∣yV(y′,B′)
Vr(y,B)=ρBlog∑B′exp[ρBW(y,B,B′)]
Choice probabilities for B′:
Pr(B′=x∣y,B)=∑iexp[ρBW(y,B,i)]exp[ρBW(y,B,x)]
Bond price schedule:
q(y,B′)=1+r1Ey′∣yPr(d=0∣y′,B′)[κ+(1−δ)Q(y′,B′)]
with
Q(y′,B′)=∑B′′Pr(B′′∣y′,B′)q(y′,B′′)
Functional forms and shocks:
u(c)=1−σc1−σ−1
h(y)=y−max0,λ0y+λ1y2
logy′=−(1−ρ)2(1−ρ2)σy2+ρlogy+σyε,ε∼N(0,1)