Erratum: Corrigendum to “Multidimensional endogenous gridpoint method: Solving triangular dynamic stochastic optimization problems without root-finding operations” (Economics Letters (2015) 135 (72–76)(S0165176515003067)(10.1016/j.econlet.2015.07.033))

Assumption A1 in  Iskhakov (2015) is incomplete. The correct version should read: Assumption A1 Transition rules for the continuous state variables Xt1,…,XtM, j∈{1,…,M} can be expressed in the form (1)Xt+1j=χj(f1(Xt1,xt1,st),f2(Xt1,xt1,Xt2,xt2,st),…,fM(Xt,xt,st),st,st+1,ξt+1)such that the post-decision states f(t)=(f1(t),…,fM(t))∈RM admit the following structure (2)f1(t)=f1(Xt1,xt1,st),f2(t)=f2(Xt1,xt1,Xt2,xt2,st),…fM(t)=fM(Xt1,xt1,…,XtM,xtM,st)=fM(Xt,xt,st), where χj(⋅) and fj(⋅) are deterministic differentiable functions, the partial derivatives of fj(t) take the form (3)∂fj(t)∂xti=∂fj(Xt1,xt1,…,Xtj,xtj,st)∂xti=gij(f1(t),…,fj(t),st),i≤j for each j=1…M and arbitrary functions gij, and it holds ∂fj(t)∂xtj≠0 and ∂fj(t)∂Xtj≠0; ξt+1∈RK are idiosyncratic shocks. The additional condition (3) ensures that ∂fk(t)∂xtj does not depend on the continuous state and decision variables in period t other than through the post decision states. These partial derivatives had been implicitly treated as constants in Section 3. I apologize for the error and express my gratitude to Jeppe Druedahl and Thomas Jørgensen at the University of Copenhagen for pointing it out. With the above correction, the example in Section 4 is invalid. Yet, the main message of the paper is unchanged: triangular dynamic stochastic optimization problems with certain structure can be solved with the multidimensional version of the endogenous grid point method without any root-finding operations.