Online Appendix I. 1 1+r ]}, Bψ = {ψ : Y E A S S}, B W = +(1 s)[1 m (1,0) (b, e, a, ψ (0,a ) (e, a, s); q, ψ, W )]}, (29) exp( U(d,a ) (i, x; q)

Σχετικά έγγραφα
Every set of first-order formulas is equivalent to an independent set

2 Composition. Invertible Mappings

Uniform Convergence of Fourier Series Michael Taylor

ST5224: Advanced Statistical Theory II

Ordinal Arithmetic: Addition, Multiplication, Exponentiation and Limit

Other Test Constructions: Likelihood Ratio & Bayes Tests

C.S. 430 Assignment 6, Sample Solutions

Example Sheet 3 Solutions

Statistical Inference I Locally most powerful tests

Lecture 2: Dirac notation and a review of linear algebra Read Sakurai chapter 1, Baym chatper 3

EE512: Error Control Coding

Math 446 Homework 3 Solutions. (1). (i): Reverse triangle inequality for metrics: Let (X, d) be a metric space and let x, y, z X.

Matrices and Determinants

A Two-Sided Laplace Inversion Algorithm with Computable Error Bounds and Its Applications in Financial Engineering

Homework 3 Solutions

ORDINAL ARITHMETIC JULIAN J. SCHLÖDER

A Note on Intuitionistic Fuzzy. Equivalence Relation

derivation of the Laplacian from rectangular to spherical coordinates

Problem Set 3: Solutions

Απόκριση σε Μοναδιαία Ωστική Δύναμη (Unit Impulse) Απόκριση σε Δυνάμεις Αυθαίρετα Μεταβαλλόμενες με το Χρόνο. Απόστολος Σ.

Areas and Lengths in Polar Coordinates

Approximation of distance between locations on earth given by latitude and longitude

2. Let H 1 and H 2 be Hilbert spaces and let T : H 1 H 2 be a bounded linear operator. Prove that [T (H 1 )] = N (T ). (6p)

k A = [k, k]( )[a 1, a 2 ] = [ka 1,ka 2 ] 4For the division of two intervals of confidence in R +

Fractional Colorings and Zykov Products of graphs

Nowhere-zero flows Let be a digraph, Abelian group. A Γ-circulation in is a mapping : such that, where, and : tail in X, head in

Areas and Lengths in Polar Coordinates

Econ 2110: Fall 2008 Suggested Solutions to Problem Set 8 questions or comments to Dan Fetter 1

ANSWERSHEET (TOPIC = DIFFERENTIAL CALCULUS) COLLECTION #2. h 0 h h 0 h h 0 ( ) g k = g 0 + g 1 + g g 2009 =?

Reminders: linear functions

12. Radon-Nikodym Theorem

5. Choice under Uncertainty

3.4 SUM AND DIFFERENCE FORMULAS. NOTE: cos(α+β) cos α + cos β cos(α-β) cos α -cos β

4.6 Autoregressive Moving Average Model ARMA(1,1)

Lecture 13 - Root Space Decomposition II

HOMEWORK 4 = G. In order to plot the stress versus the stretch we define a normalized stretch:

Bayesian statistics. DS GA 1002 Probability and Statistics for Data Science.

6.1. Dirac Equation. Hamiltonian. Dirac Eq.

Lecture 2. Soundness and completeness of propositional logic

w o = R 1 p. (1) R = p =. = 1

SOME PROPERTIES OF FUZZY REAL NUMBERS

Inverse trigonometric functions & General Solution of Trigonometric Equations

Phys460.nb Solution for the t-dependent Schrodinger s equation How did we find the solution? (not required)

Partial Differential Equations in Biology The boundary element method. March 26, 2013

Solutions to Exercise Sheet 5

Homework 8 Model Solution Section

SCITECH Volume 13, Issue 2 RESEARCH ORGANISATION Published online: March 29, 2018

CHAPTER 101 FOURIER SERIES FOR PERIODIC FUNCTIONS OF PERIOD

Lecture 21: Properties and robustness of LSE

Jesse Maassen and Mark Lundstrom Purdue University November 25, 2013

Commutative Monoids in Intuitionistic Fuzzy Sets

Math221: HW# 1 solutions

ω ω ω ω ω ω+2 ω ω+2 + ω ω ω ω+2 + ω ω+1 ω ω+2 2 ω ω ω ω ω ω ω ω+1 ω ω2 ω ω2 + ω ω ω2 + ω ω ω ω2 + ω ω+1 ω ω2 + ω ω+1 + ω ω ω ω2 + ω

Section 8.3 Trigonometric Equations

Pg The perimeter is P = 3x The area of a triangle is. where b is the base, h is the height. In our case b = x, then the area is

Finite Field Problems: Solutions

DIRECT PRODUCT AND WREATH PRODUCT OF TRANSFORMATION SEMIGROUPS

Coefficient Inequalities for a New Subclass of K-uniformly Convex Functions

Lecture 15 - Root System Axiomatics

New bounds for spherical two-distance sets and equiangular lines

Concrete Mathematics Exercises from 30 September 2016

Parametrized Surfaces

Continuous Distribution Arising from the Three Gap Theorem

( y) Partial Differential Equations

MINIMAL CLOSED SETS AND MAXIMAL CLOSED SETS

CRASH COURSE IN PRECALCULUS

Chapter 6: Systems of Linear Differential. be continuous functions on the interval

SOLUTIONS TO MATH38181 EXTREME VALUES AND FINANCIAL RISK EXAM

D Alembert s Solution to the Wave Equation

Generating Set of the Complete Semigroups of Binary Relations

ΚΥΠΡΙΑΚΗ ΕΤΑΙΡΕΙΑ ΠΛΗΡΟΦΟΡΙΚΗΣ CYPRUS COMPUTER SOCIETY ΠΑΓΚΥΠΡΙΟΣ ΜΑΘΗΤΙΚΟΣ ΔΙΑΓΩΝΙΣΜΟΣ ΠΛΗΡΟΦΟΡΙΚΗΣ 6/5/2006

Lecture 34 Bootstrap confidence intervals

Math 6 SL Probability Distributions Practice Test Mark Scheme

Appendix to Technology Transfer and Spillovers in International Joint Ventures

6.3 Forecasting ARMA processes

Section 9.2 Polar Equations and Graphs

1. Introduction and Preliminaries.

Problem Set 9 Solutions. θ + 1. θ 2 + cotθ ( ) sinθ e iφ is an eigenfunction of the ˆ L 2 operator. / θ 2. φ 2. sin 2 θ φ 2. ( ) = e iφ. = e iφ cosθ.

Section 7.6 Double and Half Angle Formulas

Fourier Series. MATH 211, Calculus II. J. Robert Buchanan. Spring Department of Mathematics

SOLUTIONS TO MATH38181 EXTREME VALUES AND FINANCIAL RISK EXAM

CHAPTER 25 SOLVING EQUATIONS BY ITERATIVE METHODS

SCHOOL OF MATHEMATICAL SCIENCES G11LMA Linear Mathematics Examination Solutions

ECE598: Information-theoretic methods in high-dimensional statistics Spring 2016

ENGR 691/692 Section 66 (Fall 06): Machine Learning Assigned: August 30 Homework 1: Bayesian Decision Theory (solutions) Due: September 13

HOMEWORK#1. t E(x) = 1 λ = (b) Find the median lifetime of a randomly selected light bulb. Answer:

Srednicki Chapter 55

The semiclassical Garding inequality

Practice Exam 2. Conceptual Questions. 1. State a Basic identity and then verify it. (a) Identity: Solution: One identity is csc(θ) = 1

Second Order RLC Filters

Local Approximation with Kernels

Second Order Partial Differential Equations

Space-Time Symmetries

DESIGN OF MACHINERY SOLUTION MANUAL h in h 4 0.

PROPERTIES OF CERTAIN INTEGRAL OPERATORS. a n z n (1.1)

Exercises to Statistics of Material Fatigue No. 5

Homomorphism in Intuitionistic Fuzzy Automata

Risk! " #$%&'() *!'+,'''## -. / # $

Solution Series 9. i=1 x i and i=1 x i.

Arithmetical applications of lagrangian interpolation. Tanguy Rivoal. Institut Fourier CNRS and Université de Grenoble 1

Transcript:

Online Appendix I Appendix D Additional Existence Proofs Denote B q = {q : A E A S [0, +r ]}, Bψ = {ψ : Y E A S S}, B W = {W : I E A S R}. I slightly abuse the notation by defining B q (L q ) the subset of B q in which q is uniformly Lipschitz continuous in s with condition L q, i.e., B q (L q ) = {q B q : q(a, e, a, s) q(a, e, a, s ) < L q s s, (a, e, a), s, s }. Similarly, B ψ (L ψ ) = {ψ B ψ : ψ (d,a ) (a, e, a, s) ψ (d,a ) (a, e, a, s ) < L ψ s s, (d, a ), (a, e, a), s, s }. B W (L ψ ) = {W B W : W (i, e, a, s) W (i, e, a, s ) < L W s s, (i, e, a), s, s }. Define mapping T = (T, T, T ) : B q B ψ B W B q B ψ B W as following: T q(a, e, a, s; q, ψ, W ) = + r {s[ m (,0) (g, e, a, ψ (0,a ) (e, a, s); q, ψ, W )] e E +( s)[ m (,0) (b, e, a, ψ (0,a ) (e, a, s); q, ψ, W )]}, (9) T ψ (d,a ) (e, a, s; q, ψ, W ) = ξ (d,a ) (x; q, ψ, W )Ω(g g) + ( ξ (d,a ) (x; q, ψ, W ))Ω(g b), (0) T W (i, e, a, s; q, ψ, W ) = γ C + log[ where, m (d,a ) (i, e, a, s; q, ψ, W ) = (d,a ) M(x;q) U(d,a ) (i,e,a,s;q) U( d, ( d, ã ) M(e,a,s;q) U(d,a ) (i, x; q) + β i EW (i, e, a, ψ (d,a ) (e, a, s)) a ) (i,e,a,s;q) + β i EW (i, e, a, ψ (d,a ) (e, a, s))], (), for (d, a ) M(e, a, s; q), + β i EW (i, e, ã, ψ ( d,ã ) (e, a, s)))) 0, for (d, a ) / M(e, a, s; q); () odd (d,a ) (e, a, s; q, ψ, W ) = U(,0) (b,e,a,s) β b EW (b, e, a, ψ (d,a ) (e, a, s))) β gew (g, e, a, ψ (d,a ) (e, a, s))) + β b EW (b, e, a, ψ (d,a ) (e, a, s))) U(,0) (g,e,a,s) + β gew (g, e, a, ψ (d,a ) (e, a, s))) ( d, a ) M(e,a,s;q) U( d, a ) (i,e,a,s;q) ( d, a ) M(e,a,s;q) U( d, a ) (i,e,a,s;q) ( d, a ) M(e,a,s;q) U( d, a ) (e,a,s,x;q) ( d, a ) M(e,a,s;q) U( d, a ) (e,a,s,x;q) + β gew (g, e, ã, ψ ( d, a ) (e, a, s)) + β b EW (b, e, ã, ψ ( d, a ) (e, a, s))) + β gew (g, e, ã, ψ ( d, a ) (e, a, s)) + β b EW (b, e, ã, ψ ( d, a ) (e, a, s))) (), for (d, a ) = (0, a ), for (d, a ) = (, 0) 78

ξ (d,a ) (e, a, s; q, ψ, W ) = + s s odd (d,a ) (e, a, s; q, ψ, W ) (4) The goal is to find large enough, s.t. (L q, L ψ, L W ), the image of T is contained in B q (L q ) B ψ (L ψ ) B W (L W ). Lemma 6. >, > 0, the function h(c) = c ) over c (0, ˉc] satisfies following properties: (i) lim c 0+ h(c) = 0; (ii) lim c 0+ h (c) = 0; (iii) h (c), c (0, ˉc]; (iv) h(c) ˉc. Proof. (i) is obvious. For (ii), consider Consider the transformation t = c which by L Hospital s rule is equal to 0. For (iii), consider h (c) = c c )., then lim c 0+ h (c) = lim t t) (( ), t + h (c) = c c ) + c c ) = c c )(c + c ). Therefore, the maximum of h (c) is attained at c = ( ), and h (c ) = ( ) ) = By Mean Value Theorem, we get (iv).. Now I show that B W can be chosen a set of functions with bounded value. Lemma 7. W ( ) and W ( ) s.t. if W ( ) W (i, e, a, s) W ( ), i, e, a, s, then W ( ) T W (i, e, a, s; q, ψ, W ) W ( ), i, e, a, s; q, ψ. Moreover, W ( ) can be chosen that is increasing in ; W ( ) can be chosen to be W that does not depend on. 79

Proof. Since default is always an option, Therefore, U(d,a ) (e, a, s, x; q) + β i EW (i, e, a, ψ (d,a ) (e, a, s))) (d,a ) M(e,a,s) [φ i (c (,0) (e, a, s) v(e))] ) β i W ( )). T W (i, e, a, s; q, ψ, W ) γ C + [φ i (c (,0) (e, a, s) v(e))] + β i W ( ). Note I have used the property the consumption upon default c (,0) (e, a, s) does not depend q. Thus fix i it suffices to set the lower bound s.t. W ( ) = γ C + [φ i (c (,0) (e, a, s) v(e))] + β i W ( ). Taking the lower across i we thus have W ( ) = min i {b,g} Since >, we have W ( ) is increasing in. (γ C + [φ i (c (,0) (e, a, s) v(e))] ). β i Next, since >, U (d,a ) (i, e, a, s; q) < 0, (i, e, a, s). Therefore, T W (i, e, a, s; q, ψ, W ) γ C + log(n A + ) + β i W ( ), where N A is the cardinality of set A. Therefore, it suffices to set which does not depend on. W ( ) = max i b,g β i (γ C + log(n A + )), Next I establish Lipschitz condition for T W. I first show the following intermediate function is Lipschitz continuous. From now on denote L q, L ψ, L W as Lipschitz conditions of q, ψ, W, respectively. The underlying assumptions in each following lemma is that q, ψ, W are Lipschitz continuous (note it does not mean I have imposed the image of T to be Lipschitz). 80

Consider function μ (d,a ) (i, e, a, s; q) defined as μ (d,a ) (i, e, a, s; q) = c (d,a ) (e, a, s; q) () ), for (d, a ) M(e, a, s; q), 0, for (d, a ) / M(e, a, s; q). Lemma 8. μ (d,a ) (i, e, a, s; q) μ (d,a ) (i, e, a, s ; q) L q C μ s s, (d, a ), i, e, a, s, where C μ is a constant that does not depend on. Further, μ (d,a ) (i, e, a, s; q) for some constant C μ 0 that does not depend on. Proof. Case, If (d, a ) M(e, a, s; q) and (d, a ) M(e, a, s ; q), using Lemma 6 and Mean Value Theorem, μ (d,a ) (i, e, a, s; q) μ (d,a ) (i, e, a, s ; q) C μ 0 c (d,a ) (e, a, s; q) c (d,a ) (e, a, s ; q) max a A a L q s s. Case, suppose (d, a ) M(e, a, s; q) and (d, a ) / M(e, a, s ; q). a = 0. Now I claim Notice in this case, c (d,a ) (e, a, s; q) s s a L q Suppose not, then we have s s < c(d,a ) (e, a, s; q) a L q, which implies c (d,a ) (e, a, s ; q) c (d,a ) (e, a, s; q) L q a s s < c (d,a ) (e, a, s; q), which implies c (d,a ) (e, a, s ; q) > 0 that contradicts (d, a ) / M(e, a, s ; q). Therefore, μ (d,a ) (i, e, a, s; q) μ (d,a ) (i, e, a, s ; q) = μ (d,a ) (i, e, a, s; q) c (d,a ) (e, a, s; q) 0 max a A a L q s s 8

Setting C μ = maxa A a we have proved the first part. For the second part, apply (iv) of Lemma 6 we have μ (d,a ) (i, e, a, s; q) where ˉc can be any constant greater than the maximum total consumption attainable, which can be set as ˉc = we max n(e max )( τ) a min + r + a max + Tr v(n(e max )). Setting C μ 0 = ˉc, we have proved the second part. We review the following properties about Lipschitz continuity. Lemma 9. Suppose functions f (x), g(x) are defined on closed interval [x, x] and are bounded in sup norm f and g, and f (x), g(x) are Lipschitz continuous in x with condition L f and L g, respectively. Then the following are true: () L α f α L f, α R; () L f +g L f + L g ; () L f g L f g + L g f ; (4) For function h that the compound evaluation h( f ) is defined, suppose h( ) is Lipschitz continuous with condition L h, then L h( f ) L h L f. If h is differentiable with bounded derivative h, then L h( f ) h L f. Lemma 0. T W (i, e, a, s ; q, ψ, W ) T W (i, e, a, s; q, ψ, W ) < L T W s s, where L T W = [C T W ( )L q + C T W ( )L W L ψ ], where C T W ( ) > 0 is decreasing in, and C T W ( ) > 0 is decreasing in. Proof. First, default is always an option so the sum of exponential value is bounded ˉc 8

below, denoted by LB( ): U(d,a ) (i, e, a, s; q) + β i EW (i, e, a, ψ (d,a ) (e, a, s))) (d,a ) M(e,a,s;q) min [Tr + ( τ)we min n(e min ) n(e min )] + β i {b,g} i W ( )) = LB( ). Notice that LB( ) is increasing in since > and W ( ) is increasing in (Lemma 7). Denote L ( f ) as the Lipschitz condition for the function f w.r.t. s. Notice can be written as U(d,a ) (i, e, a, s; q) + β i EW (i, e, a, ψ (d,a ) (e, a, s)) (d,a ) M(x;q) μ (d,a ) (i, e, a, s; q) β i EW (i, e, a, ψ (d,a ) (e, a, s))), (d,a ) Y because μ (d,a ) (i, e, a, s; q) assigns 0 to (d, a ) / M(e, a, s; q). Using Lemma 9, L (β i EW (i, e, a, ψ (d,a ) (e, a, s)))) β i β i W )L W L ψ. And using Lemma 8, L (μ (d,a ) (i, e, a, s; q)) C μ L q, μ (d,a ) (i, e, a, s; q) C μ 0, 8

Therefore, using Lemma 9, C μ 0 Therefore, Therefore, L [ L (μ (d,a ) (i, e, a, s; q) β i EW (i, e, a, ψ (d,a ) (i, e, a, s)))) L (β i EW (i, e, a, ψ (d,a ) (e, a, s)))) + β i W )L (μ (d,a ) (i, e, a, s; q)) (d,a ) M(x;q) U(d,a ) (i, e, a, s; q) β i C μ 0 β i W )L W L ψ + C μ L q. β g C μ 0 β g W )L W L ψ + C μ L q. + β i EW (i, e, a, ψ (d,a ) (i, e, a, s)))] [β g C μ 0 β g W )L W L ψ + C μ L q ](N A + ). Setting L [T W (i, e, a, s; q, ψ, W ) = L [log( U(d,a ) (i, e, a, s; q) + β i EW (i, e, a, ψ (d,a ) (e, a, x))))] (d,a ) M(x;q) LB( ) [β gc μ 0 β g W )L W L ψ + C μ L q ](N A + ) C T W ( ) = LB( ) β gc μ 0 β gw )(N A + ) C T W ( ) = LB( ) Cμ (N A + ). Since LB ( ) > 0 and is increasing in, we have proved the results. Lemma. m defined in Equation is Lipschitz continuous in s with condition L m = [C m ( )L q + C m ( )L W L ψ ], where C m ( ) > 0 is decreasing in, and C m ( ) > 0 is decreasing in. 84

Proof. Notice m (d,a ) (i, e, a, s; q, ψ, W ) = μ (d,a ) (e, a, s; q) β i EW (i, e, a, ψ (d,a ) (e, a, s)) ( d, ã ) Y μ( d,ã ) (e, a, s; q) β i EW (i, e, ã, ψ ( d,ã ) (e, a, s))). As shown in Lemma 0, L [μ (d,a ) (i, e, a, s; q) β i EW (i, e, a, ψ (d,a ) (e, a, s)))] β g C μ 0 β g W )L W L ψ + C μ L q. L [ ( d,ã ) Y μ ( d,ã ) (e, a, s; q) β i EW (i, e, ã, ψ ( d,ã ) (e, a, s)))] [β g C μ 0 β g W )L W L ψ + C μ L q ](N A + ) Apply Lemma 9, μ ( d,ã ) (e, a, s; q) β i EW (i, e, ã, ψ ( d,ã ) (e, a, s))) LB( ) ( d,ã ) Y L (m (d,a ) (i, e, a, s; q, ψ, W )) LB ( ) [β gc μ 0 β g W )L W L ψ + C μ L q ](N A + ) β g W ) +[β g C μ 0 β g W )L W L ψ + C μ L q ] LB( ) = [ LB ( ) (N A + ) β g W )C μ 0 + LB( ) Cμ 0 ]β g β g W L W L ψ Setting and C m( ) = [ LB ( ) (N A + ) β g W ) + LB( ) ]Cμ +[ LB ( ) (N A + ) β g W ) + LB( ) ]Cμ C m( ) = [ LB ( ) (N A + ) β g W ) + LB( ) ]Cμ 0 β g β g W ). Since LB( ) is increasing in, we have proved the result. Similarly we can show L q 85

Lemma. odd defined in Equation is Lipschitz continuous in s with condition L odd = [C odd ( )L q + C odd ( )L W L ψ ], where C odd ( ) > 0 is decreasing in, and C odd ( ) > 0 is decreasing in. Lemma. ξ defined in Equation 4 is Lipschitz continuous in s with condition L ξ = [C ξ ( )L q + C ξ ( )L W L ψ ] + C ξ, where C ξ ( ) > 0 is decreasing in, C ξ ( ) > 0 is decreasing in, and C ξ depend on. > 0 does not Proof. Notice 0 < s s s <, s s s s and d s s ds = s [ s, s ] Using Lemma 9, L [ s s odd (d,a ) (e, a, s; q, ψ, W )] s + L odd( s ) L [ + s s odd (d,a ) (e, a, s; q, ψ, W ) ] s + L odd( s ) s + ( s )( [C odd ( )L q + C odd ( )L W L ψ ]). Then we set C ξ ( ) = ( s )Codd ( ) C ξ ( ) = ( s )Codd ( ) 86

C ξ = s Lemma 4. T ψ defined in Equation 0 is Lipschitz continuous in s with condition L T ψ = [C T ψ ( )L q + C T ψ ( )L W L ψ ] + C T ψ, where C T ψ ( ) > 0 is decreasing in, C T ψ ( ) > 0 is decreasing in, and C T ψ > 0 does not depend on. Proof. Apply Lemma 9, L T ψ (Ω(g g) + Ω(g b))l ξ (Ω(g g) + Ω(g b))[ (C ξ ( )L q + C ξ ( )L W L ψ ) + C ξ ]. Lemma 5. Tq defined in Equation 9 is Lipschitz continuous in s with condition L T q = [C T q ( )L q L ψ + C T q ( )L W L ψ] + C T q, where C T q ( ) > 0 is decreasing in, C T q ( ) > 0 is decreasing in, and C T q > 0 does not depend on. Proof. Apply Lemma 9, L T q + r (sl ml ψ + + ( s)l m L ψ + ) + r + ( + s s)l ψ[ [C m( )L q + C m( )L W L ψ ]]. Lemma 6. L q, L ψ, L W, (L q, L ψ, L W ) s.t. > (L q, L ψ, L W ), L T W < L W. Proof. From Lemma 0, L T W = [C T W ( )L q + C T W ( )L W L ψ ]. 87

Choose (L q, L ψ, L W ) s.t. C T W ( )L ψ < / and This can be done since lim. C T W ( )L q < L W Then we have > (L q, L ψ, L W ), L T W < L W. = 0 and C T W ( ) and C T W ( ) are decreasing in Lemma 7. L q, L ψ C T ψ, L W, (L q, L ψ, L W ) s.t. > (L q, L ψ, L W ), L T ψ < C T ψ. Proof. From Lemma 4, L T q = [C T q ( )L q L ψ + C T q ( )L W L ψ] + C T q. L q, L ψ < C T ψ, L W, choose (L q, L ψ, L W ) s.t. C T q. This can be done since lim in. Similarly, we can establish [C T q ( )L q L ψ + C T q ( )L W Lψ ] < = 0 and C T ψ ( ) and C T ψ ( ) are decreasing Lemma 8. L q C T q, L ψ, L W, (L q, L ψ, L W ) s.t. > (L q, L ψ, L W ), L T q < C T q. Through the above three lemmas, we have the following Lemma 9. For L q = C T q, L ψ = C T ψ, L W =, s.t. >, L T q C T q, L T ψ C T ψ, L T W. Proof. Choose (L q, L ψ, L W ), (L q, L ψ, L W ), (L q, L ψ, L W ) from Lemma 8, Lemma 7, and Lemma 6, respectively. Choose = max{ (L q, L ψ, L W ), (L q, L ψ, L W ), (L q, L ψ, L W )}. Then the result follows Lemma 8, 7, and 6. Up to now we have established that by choosing large enough, T maps B q (L q ) B ψ (L ψ ) B W (L W ) to itself. Now we prove the mapping T is continuous. Denote as sup norm. 88

Lemma 0. The intermediate function μ (d,a ) (i, e, a, s; q) defined in Lemma 8 is continuous in q with sup norm for q B q (L q ). Proof. Consider μ (d,a ) (i, e, a, s; q) μ (d,a ) (i, e, a, s; q ) under different cases. Case : (d, a ) M(e, a, s; q), and (d, a ) M(e, a, s; q ), by Mean Value Theorem μ (d,a ) (i, e, a, s; q) μ (d,a ) (i, e, a, s; q ) c (d,a ) (e, a, s; q) c (d,a ) (e, a, s; q ) max a A a q q Case : (d, a ) M(e, a, s; q), and (d, a ) / M(e, a, s; q ), Notice in this case, a = 0. Now I claim c (d,a ) (e, a, s; q) q(a, e, a, s) q (a, e, a, s) a Suppose not, then we have which implies q(a, e, a, s) q (a, e, a, s) < c(d,a ) (e, a, s; q) a, c (d,a ) (e, a, s; q ) c (d,a ) (e, a, s; q) q(a, e, a, s) q (a, e, a, s) a < c (d,a ) (e, a, s; q), which implies c (d,a ) (e, a, s; q ) > 0 that contradicts (d, a ) / M(e, a, s; q ). Therefore, μ (d,a ) (i, e, a, s; q) μ (d,a ) (i, e, a, s; q ) = μ (d,a ) (i, e, a, s; q) 0 c (d,a ) (e, a, s; q) 0 q(a, e, a, s) q (a, e, a, s) a max a A a q q. Similarly for case where (d, a ) / M(e, a, s; q), and (d, a ) M(e, a, s; q ). And for 89

case 4 (d, a ) / M(e, a, s; q) and (d, a ) / M(e, a, s; q ) we have μ (d,a ) (i, e, a, s; q) μ (d,a ) (i, e, a, s; q ) = 0. Summarize over all four cases we have μ (d,a ) (i, e, a, s; q) μ (d,a ) (i, e, a, s; q ) Therefore μ (d,a ) (i, e, a, s; q) is continuous in q with sup norm. max a A a q q. Lemma. T W (i, e, a, s; q, ψ, W ) is continuous in (q, ψ, W ) with sup norm on B q (L q ) B ψ (L ψ ) B W (L W ). Proof. Write T W (i, e, a, s; q, ψ, W ) as T W (i, e, a, s; q, ψ, W ) = γ C + log( (d,a ) Y μ (d,a ) (i, e, a, s; q) β i EW (i, e, a, ψ (d,a ) (e, a, s)))) Since (d,a ) Y μ (d,a ) (i, e, a, s; q) β i EW (i, e, a, ψ (d,a ) (e, a, s))) is uniformly bounded positive below (by LB( ) established in Lemma 0), μ (d,a ) (i, e, a, s; q) is continuous in q, W (i, e, a, s) is Lipschitz continuous in s, therefore, T W (i, e, a, s; q, ψ, W ) is continuous in (q, ψ, W ). Lemma. m (d,a ) (i, e, a, s; q, ψ, W ) defined in Equation is continuous in (q, ψ, W ) with sup norm on B q (L q ) B ψ (L ψ ) B W (L W ). Proof. Write m (d,a ) (i, e, a, s; q, ψ, W ) as following m (d,a ) (i, e, a, s; q, ψ, W ) = μ (d,a ) (e, a, s; q) β i EW (i, e, a, ψ (d,a ) (e, a, s)) ( d, ã ) Y μ( d,ã ) (e, a, s; q) β i EW (i, e, ã, ψ ( d,ã ) (e, a, s))). Note the following properties. () W is uniformly bounded above and below. () W is Lipschitz in s. () μ is continuous in q. (4) ( d, ã ) Y μ( d,ã ) (e, a, s; q) β i EW (i, e, ã, ψ ( d,ã ) (e, a, s))) is uniformly bounded above and uniformly positively bounded below (established in the proof of Lemma 7). Therefore, we have m is continuous in (q, ψ, W ). We next establish Lemma. T q(a, e, a, s; q, ψ, W ) is continuous in (q, ψ, W ) with sup norm on B q (L q ) B ψ (L ψ ) B W (L W ). 90

Proof. Consider (q n, ψ n, W n ) (q, ψ, W ) with sup norm, we have max m(d,a ) (i, e, a, ψ (0,a ) (e, a, s); q, ψ, W ) m (d,a ) (i, e, a, ψ (0,a ) i,e,a n (e, a, s); q n, ψ n, W n ),e,a,s max m(d,a ) (i, e, a, ψ (0,a ) (e, a, s); q, ψ, W ) m (d,a ) (i, e, a, ψ (0,a ) (e, a, s); q n, ψ n, W n ) i,e,a,e,a,s + max i,e,a,e,a,s m(d,a ) (i, e, a, ψ (0,a ) (e, a, s); q n, ψ n, W n ) m (d,a ) (i, e, a, ψ (0,a ) n (e, a, s); q n, ψ n, W n ) First, lim n max m(d,a ) (i, e, a, ψ (0,a ) (e, a, s); q, ψ, W ) m (d,a ) (i, e, a, ψ (0,a ) (e, a, s); q n, ψ n, W n ) = 0 i,e,a,e,a,s since m( ; q, ψ, W ) is continuous in (q, ψ, W ). Second, m (d,a ) (i, e, a, ψ (0,a ) (e, a, s); q n, ψ n, W n ) m (d,a ) (i, e, a, ψ (0,a ) n (e, a, s); q n, ψ n, W n ) L m ψ ψ n since m (d,a ) (i, e, a, s; q, ψ, W ) is uniformly Lipschitz continuous in s with condition L m, as established in Lemma. lim n Therefore, max m(d,a ) (i, e, a, ψ (0,a ) (e, a, s); q, ψ, W ) m (d,a ) (i, e, a, ψ (0,a ) i,e,a n (e, a, s); q n, ψ n, W n ) = 0,e,a,s Therefore, consider T q( ; q, ψ, W ) T q(a, e, a, s; q, ψ, W ) = + r {s[ m (,0) (g, e, a, ψ (0,a ) (e, a, s); q, ψ, W )] e E +( s)[ m (,0) (b, e, a, ψ (0,a ) (e, a, s); q, ψ, W )]}, lim max T q(a, e, a, s; q, ψ, W ) T q(a, e, a, s; q n, ψ n, W n ) = 0 n a,e,a,s Similarly we can prove Lemma 4. T ψ(a, e, a, s; q, ψ, W ) is continuous in (q, ψ, W ) with sup norm on B q (L q ) B ψ (L ψ ) B W (L W ). 9

We next apply Schauder fixed point theorem on the mapping T. Lemma 5. s.t. >, T defined in Equation 9, 0, and has fixed points. Proof. First, from Lemma 7, for large enough, B W can be chosen the set of functions uniformly bounded by W () and W. (Notice W ( ) is increasing in ). Fix L q = C T q, L ψ = C T ψ, L W =. From Lemma 9, s.t. >, L T q C T q, L T ψ C T ψ, L T W. Now denote T : B q (C T q ) B ψ (C T ψ ) B W () B q B ψ B W as mapping that agrees with T on the subset B q (C T q ) B ψ (C T ψ ) B W (). Then T(B q (C T q ) B ψ (C T ψ ) B W ()) B q (C T q ) B ψ (C T ψ ) B W (). And from Lemma,, and 4, T is continuous. By Arzelà Ascoli theorem, the set of bounded Lipschitz functions B q (C T q ) B ψ (C T ψ ) B W () is compact. All conditions applying Schauder fixed point theorem are satisfied. Therefore, T has fixed points, which are also fixed points of T. Now given q, ψ, W and the induced policy function m( ; q, ψ, W ), I establish the existence of stationary probability measure. Denote M Φ as the set of probability measure Φ over (I E A S, P(I) P(E) P(A) B(S)), where P( ) is the power set of underlying discrete sets, and B(S) is the Borel algebra of [s, s]. Denote T 4 : M Φ M Φ defined as following: T 4 Φ(i, e, a, S ; Φ) = +(a = 0) i,e,a Ω(i i)γ(e e) m (0,a ) (i, e, a, s; q, ψ, W )(ψ (0,a ) (e, a, s) S )Φ(i, e, a, ds) a Ω(i i)γ(e e)m (,0) (i, e, a, s; q, ψ, W )(ψ (,0) (e, a, s) S )Φ(i, e, a, ds), i,e,a We prove T 4 is continuous in Φ under the weak topology. Similar proof strategies have been used in Cao (06). We first have the following: Lemma 6. f Lipschitz continuous (with condition L f ) and sequence of Φ n that is weakly converging to Φ lim f (s )T 4 Φ(i, e, a, ds ; Φ n ) = f (s )T 4 Φ(i, e, a, ds ; Φ) n s S s S depends on (r, τ, Tr) and other model parameters. i, e, a, S B(S) (5) 9

i, e, a, S B(S). Proof. Consider = lim f (s )[ n s S = lim [ n +(a = 0) +(a = 0) s i,e,a = lim [ n s i,e,a lim f (s )T 4 Φ(i, e, a, ds ; Φ n ) n s S Ω(i i)γ(e e) m (0,a ) (i, e, a, s; q, ψ, W )(ψ (0,a ) (e, a, s) ds )Φ n (i, e, a, ds) a s i,e,a Ω(i i)γ(e e)m (,0) (i, e, a, s; q, ψ, W )(ψ (,0) (e, a, s) ds )Φ n (i, e, a, ds)] Ω(i i)γ(e e) a m (0,a ) (i, e, a, s; q, ψ, W ) s i,e,a +(a = 0) i,e,a f (s )(ψ (0,a ) (e, a, s) ds )Φ n (i, e, a, ds) s S Ω(i i)γ(e e)m (,0) (i, e, a, s; q, ψ, W ) f (s )(ψ (,0) (e, a, s) ds )Φ n (i, e, a, ds)] s S {s:ψ (0,a ) (e,a,s) S } i,e,a Ω(i i)γ(e e) m (0,a ) (i, e, a, s; q, ψ, W ) f (ψ (0,a ) (e, a, s))φ n (i, e, a, ds) a {s:ψ (0,a ) (e,a,s)} S Ω(i i)γ(e e)m (,0) (i, e, a, s; q, ψ, W ) f (ψ (0,a ) (e, a, s))φ n (i, e, a, ds)] The second equal changes order of integration since the inner integral is finite (m is positive bounded below ). Since f, m, ψ are all bounded and Lipschitz in s ( f is bounded because f is Lipschitz and defined on a compact set), by definition of weak convergence the above = i,e,a +(a = 0) i,e,a {s:ψ (0,a ) (e,a,s) S } {s:ψ (0,a ) (e,a,s) S } Ω(i i)γ(e e) m (0,a ) (i, e, a, s; q, ψ, W ) f (ψ (0,a ) (e, a, s))φ(i, e, a, ds) a Ω(i i)γ(e e)m (,0) (i, e, a, s; q, ψ, W ) f (ψ (0,a ) (e, a, s))φ(i, e, a, ds) = f (s )T 4 Φ(i, e, a, ds ; Φ) s S Therefore, T 4 ( ; Φ) is continuous in Φ with weak topology, we have Lemma 7. T 4 has fixed points. Proof. Since S is convex and compact, M Φ is a convex and compact set. And since T 4 is continuous. The conditions of Schauder fixed point theorem are satisfied. T 4 has fixed points. Now I construct the remaining parts of equilibrium out of q, ψ, W, Φ. 9

Proof for Proposition, restated in Lemma 8 Lemma 8. For any r > 0, τ [0, ), and Tr 0, (r, τ, Tr) s.t. > (r, τ, Tr), a stationary equilibrium SCE(r, τ, Tr) exists. Proof. Given (r, τ, Tr), w = z η η. Using Lemma 5, we have, s.t. >, (q, ψ, W ) that is a fixed point of operator T. Then pick and fix (q, ψ, W ), using Lemma 7, we have Φ that is a fixed point of operator T 4. Now pick and fix a Φ. Construct the following (sequentially): W(i, e, a, s) = W (i, e, a, s) i = ˉr B = [ π = ˉπ = a m (0,a ) (i, e, a, s; q, ψ, W )q(a, e, a, s)a Φ(di, de, da, ds)]( + r) n(e) = v (( τ t )w t e) N = Γ(e)n(e)e e E Pro f it = zn wn G = B + Pro f it + τwn Tr B + r Y = zn Then one verifies all the equilibrium conditions are satisfied. 94