yuanmeng001

Inﬁnitesimal Methods in Mathematical Economics

Inﬁnitesimal Methods in Mathematical Economics
Robert M. Anderson1 Department of Economics and Department of Mathematics University of California at Berkeley Berkeley, CA 94720, U.S.A. and Department of Economics Johns Hopkins University Baltimore, MD 21218, U.S.A.
January 20, 2008
1The author is grateful to Marc Bettz¨uge, Don Brown, HungWen Chang, G´erard Debreu, Eddie Dekel-Tabak, Greg Engl, Dmitri Ivanov, Jerry Keisler, Peter Loeb, Paul MacMillan, Mike Magill, Andreu Mas-Colell, Max Stinchcombe, Cathy Weinberger and Bill Zame for their helpful comments. Financial support from Deutsche Forschungsgemeinschaft, Gottfried-WilhelmLeibniz-F¨orderpreis is gratefully acknowledged.
Contents
0 Preface v
1 Nonstandard Analysis Lite 1 1.1 When is Nonstandard Analysis Useful? . . . . 1 1.1.1 Large Economies . . . . . . . 2
1.1.2 Continuum of Random Variables . . . 4
1.1.3 Searching For Elementary Proofs . . . 4 1.2 Ideal Elements 5 1.3 Ultraproducts . 6
1.4 Internal and External Sets . .9
1.5 Notational Conventions . . . . . . . . . . . . . 11
1.6 Standard Models . . . . . . . . . . . 12
1.7 Superstructure Embeddings. . . 14
1.8 A Formal Language . . . . . . . . . . . . . . . 16
1.9 Transfer Principle . . . . . . . . . . . . . . . . 16
1.10 Saturation . . . . . . . . . . . . . .

. 18
1.11 Internal Deﬁnition Principle . . . . . . . . .

. 19
1.12 Nonstandard Extensions, or Enough Already with the Ultraproducts . . . . . . . . . . . . . 20
1.13 Hyperﬁnite Sets . . . . . . . . . . . . . .

. . . 21
1.14 Nonstandard Theorems Have Standard Proofs 22
2 Nonstandard Analysis Regular 23 2.1 Warning: Do Not Read this Chapter . . . . . 23
i
ii CONTENTS
2.2 A Formal Language . . . . . . . . . . .

. . . . 23
2.3 Extensions of L . . . . . . . . . 25
2.4 Assigning Truth Value to Formulas . . . . . .

28
2.5 Interpreting Formulas in Superstructure Embeddings . . . . . . . . . . . . . . . . . . .

. . 32
2.6 Transfer Principle . . . . . . . . . . .

. . 33
2.7 Internal Deﬁnition Principle . . . . . . . . . . 34
2.8 Nonstandard Extensions . . . . . . . . . . . . 35
2.9 The External Language . . . . . . . . . . . . . 36
2.10 The Conservation Principle . . . . . . . .

. . 37
3 Real Analysis 39 3.1 Monads . . . . . . . . . . . . . . . . . . . .

. 39
3.2 Open and Closed Sets . . . . . . . . . . . . .

44
3.3 Compactness . . . . . . . . . . . . . 45
3.4 Products . . . . 48
3.5 Continuity . . . . . . . . . . . . . . 49
3.6 Diﬀerentiation . . . . . . . . . . . . . . . . . . 52 3.7 Riemann Integration . . . . . . . . . . . . . . 53
3.8 Diﬀerential Equations . . . . . . . . . . . . . . 54
4 Loeb Measure 57 4.1 Existence of Loeb Measure . . . . . . . . . . . 57 4.2 Lebesgue Measure . . . . . . . . . . . . . . . . 61 4.3 Representation of Radon Measures . . . . . . 62 4.4 Lifting Theorems . . . . . . . . . . . . . . . . 63 4.5 Weak Convergence . . . . . . . . . . . . . . . 65
5 Large Economies 69 5.1 Preferences . . . . . . . . . . . . . . . . . . . 70 5.2 Hyperﬁnite Economies . . . . . . . . . . . . . 74 5.3 Loeb Measure Economies . . . . . . . . . . . . 74 5.4 Budget, Support and Demand Gaps . . . . . . 75 5.5 Core . . . . . . . . . . . . . . . . . . . . . . . 76
CONTENTS iii
5.6 Approximate Equilibria . . . . . . . . . . . . . 98 5.7 Pareto Optima . . . . . . . . . . . . . . . . . 98 5.8 Bargaining Set . . . . . . . . . . . . . . . . . 98 5.9 Value . . . . . . . . . . . . . . . . . . . . . . . 99 5.10 “Strong” Core Theorems . . . . . . . . . . . . 99
6 Continuum of Random Variables 101 6.1 The Problem . . . . . . . . . . . . . . . . . . 101 6.2 Loeb Space Construction . . . . . . . . . . . . 103 6.3 Price Adjustment Model . . . . . . . . . . . . 105
7 Noncooperative Game Theory 111
8 Stochastic Processes 113
9 Translation 115
10 Further Reading 119
A Existence Proof 121
iv CONTENTS
Chapter 0
Preface
Nonstandard analysis is a mathematical technique widely used in diverse areas in pure and applied mathematics, including probability theory, mathematical physics, functional analysis. Our primary goal is to provide a careful development of nonstandard methodology in suﬃcient detail to allow the reader to use it in diverse areas in mathematical economics. This requires some work in mathematical logic. More importantly, it requires a careful study of the nonstandard treatment of real analysis, measure theory, topological spaces, and so on. Chapter 1 provides an informal description of nonstandard methods which should permit the reader to understand the mathematical Chapters 3, 4 and 9, as well as Chapters 5, 6, 7 and 8 which provide applications in Economics. Chapter 2 provides a precise statement of the fundamental results on nonstandard models, using tools from mathematical logic. Most readers will ﬁnd it better to browse through the proofs in Chapter 3 before tackling Chapter 2; many readers will choose to skip Chapter 2 entirely! The proof of the existence of nonstandard models is trivial but tedious, and so is deferred to Appendix A.
v
vi CHAPTER 0. PREFACE
This is an expanded version of Anderson (1992), a Chapter in Volume IV of the Handbook of Mathematical Economics. This ﬁrst handout contains Chapters 1–3 and the Bibliography. Chapters 4 and 5 and part of Chapter 6 are done, apart from minor revisions to be made based on the experience in the ﬁrst few weeks of teaching; they will be handed out in a few weeks. The remainder of the the chapters listed in the contents are phantoms at this point; they will be writtenand distributedover the course of the semester.
Chapter 1
Nonstandard Analysis Lite
1.1 When is Nonstandard Analysis Useful?
Nonstandard analysis can be used to formalize most areas of modern mathematics, including real and complex analysis, measure theory, probability theory, functional analysis, and point set topology; algebra is less amenable to nonstandard treatments, but even there signiﬁcant applications have been found. Complicated —δ arguments can usually be replaced by simpler, more intuitive nonstandard arguments involving inﬁnitesimals. Given the dependence of work in mathematical economics on arguments from real analysis at the level of Rudin(1976) or Royden (1968), a very large number of papers could be signiﬁcantly simpliﬁed using nonstandard arguments. Unfortunately, there is a signiﬁcant barrier to the widespread adoption of nonstandard arguments for these kinds of problems, a barrier very much akin to the problems
1
2 CHAPTER 1. NONSTANDARD ANALYSIS LITE
associated with the adoption of a new technological standard. Few economists are trained in nonstandard analysis, so papers using the methodology are necessarily restricted to a small audience. Consequently, relatively few authors use the methodology if more familiar methods will suﬃce. Therefore, the incentives to learn the methodology are limited. Accordingly, the use of nonstandard methods in economics has largely been limited to certain problems in which the advantages of the methodology are greatest.
1.1.1 Large Economies
Most of the work in Economics using nonstandard methods has occurred in the literature on large economies. Suppose χn : An → P × Rk + is a sequence of exchange economies with |An|→∞. In other words, An is the set of agents and |An| the number of agents in the nth economy, P a set of preferences, and χn assigns to each agent a preference and an endowment vector. A natural approach to analyzing the sequence χn is to formulate a notion of a limit economy χ : A → P×Rk +. This limit economy can be formulated with A being eithera nonatomic measure space or a hyperﬁnite set–a nonstandard construction. The properties of measure spaces and hyperﬁnite sets are closely analogous. Indeed, using the Loeb measure construction which we describe in Chapter 4, a hyperﬁnite set can be converted into a measure space. The theory of economies with a hyperﬁnite set of agents is analogous in many respects to the theory of economies with a measure space of agents. However, there are certain phenomena that can occur in hyperﬁnite economies which are ruled out by the measuretheoretic formulation. For the most part, these relate to situations in which a small proportion of the agents are en
1.1. WHEN IS NONSTANDARD ANALYSIS USEFUL? 3
dowed with, or consume, a substantial fraction of the goods present in the economy. In the hyperﬁnite context, certain conditions inherent in the measure-theoretic formulation can be seen to be strong endogenous assumptions. Using hyperﬁnite exchange economies, we can state exogenous assumptions which imply the endogenous assumptions inherent in the measure-theoretic formulation, as well as explore the behavior of economies in which the endogenous assumptions fail.
The power of the nonstandard methodology is seen most clearly at the next stage, in which one deduces theorems about the sequence χn from the theorems about the limit χ. A central result known as the Transfer Principle asserts that any property which can be formalized in a particular formal language and which holds for χ must also hold for χn for suﬃciently large n. Viewed in this context, the Transfer Principle functions as a sweeping generalization of the convergence theorems that can be formulated using topology and measure theory. The Transfer Principle converts results about the limit economy χ into limiting results about the sequence χn in a few linesof argument. Consequently, for those properties which hold both in measure-theoretic and hyperﬁnite economies, nonstandard analysis provides a very eﬃcient tool to derive limit theorems for large ﬁnite economies. On the other hand, in the situations in which the behavior of the measure-theoretic and hyperﬁnite economies diﬀer, it is the hyperﬁnite economy rather than the measure-theoretic economy which captures the behavior of large ﬁnite economies. The literature on large economies is discussed in Chapter 5.
4 CHAPTER 1. NONSTANDARD ANALYSIS LITE
1.1.2 Continuum of Random Variables
Probability theory is currently the most active ﬁeld for applications of nonstandard analysis. Since probabilistic constructions are widely used in general equilibriumtheory, game theory and ﬁnance, these seem fruitful areas for further applications of nonstandard methodology. Nonstandard analysis provides an easy resolution of the so-called continuum of random variables problem. Given a standard measure space (A,A,μ) and an uncountable family of independent identically distributed random variables Xλ : A→ R, one would like to assert that there is no aggregate uncertainty; in other words, the empirical distribution of the Xλ’s equals the theoretical distributuion with probability one. Typically, however, the set of a ∈ A for which the empirical distribution of the Xλ’s equals the theoretical distribution is not measurable; by extending μ, it can be assigned any measure between 0 and 1. Note that, with a large ﬁnite number of random variables, the measurability issue does not even arise. Thus, the problem is a pathology arising from the formulation of measure theory, and not a problem of large ﬁnite systems. If A is a hyperﬁnite set, and μ is the Loeb measure, then the empirical distribution does equal the theoretical distribution with probability one. Moreover, as in the large economies literature, the Transfer Principle can be used to deduce asymptotic properties of large ﬁnite systems. The Continuum of Random Variables Problem is discussed in Chapter 6.
1.1.3 Searching For Elementary Proofs
Nonstandard analysis allows one to replace many measuretheoretic arguments by discretecombinatoric arguments. For example, the Shapley-Folkman Theorem plays the same role
1.2. IDEAL ELEMENTS 5
in hyperﬁnite sets as Lyapunov’s Theorem playsin nonatomic measure spaces. Thus, nonstandard analysis is an eﬀective tool for determining exactly which parts of a given proof really depend on arguments in analysis, and which follow from more elementary considerations. Occasionally, it is possible to replace every step in a proof by an elementary argument; as a consequence, one obtains a proof using neither nonstandard analysis nor measure theory. Examples are discussed in Chapter 9.
1.2 Ideal Elements
Leibniz’ formulation of calculuswas based on the notion of an inﬁnitesimal. Mathematics has frequently advanced through the introduction of ideal elements to provide solutions to equations. The Greeks were horriﬁed to discover that the equation x2 = 2 has no rational solution; this problem was resolved by the introduction of the ideal element √2; ultimately, the real numbers were deﬁned as the completion of the rationals. Similarly, the complex numbers were created by the introduction of the ideal element i = √−1. Leibniz introduced inﬁnitesimals as ideal elements which, while not zero, were smaller than any positive real number. Thus, an inﬁnitesimal is an ideal element providing a solution to the family of equations
x>0; x<1,x<
1 2
,x<
1 3
,.... (1.1)
Inﬁnitesimals played a key role in Leibniz’ formulation of calculus. For example, the derivative of a function was deﬁned as the slope of the function over an interval of inﬁnitesimal length. Leibniz asserted that the real numbers, augmented by the addition of inﬁnitesimals, obeyed all the same rules
6 CHAPTER 1. NONSTANDARD ANALYSIS LITE
as the ordinary real numbers. Unfortunately, neither Leibniz nor his successors were able to develop a formulation of inﬁnitesimals which was free from contradictions. Consequently, in the middle of the nineteenth century, the –δ formulation replaced inﬁnitesimals as the generally accepted foundation of calculus and real analysis. In 1961, Abraham Robinson discovered that model theory, a branch of mathematical logic, provided a satisfactory foundation for the use of inﬁnitesimals in analysis. In the remainder of Section 1, we will provide an informal description of a model of the nonstandard real numbers. In Section 2, we will provide a formal description of nonstandard models, along with a precise statement of the rules of inference which are allowed in reasoning about nonstandard models. The proof of the underlying theorems which justify the rules of inference will be presented in the Appendix.
1.3 Ultraproducts
A very simple construction which produces elements with inﬁnitesimal properties is RN, the space of real sequences. We can embed R into RN by mapping each r ∈ R to the constant sequence ¯ r =(r,r,r,...). Now consider the sequence deﬁned by xn = 1 n. LetR++ denote the set of strictly positive real numbers. Given any r ∈ R++, observe thatxn < ¯ rn for all but a ﬁnite number of values of n. In other words, if we were to deﬁne a relation x< F y ⇐⇒ xn 1.3. ULTRAPRODUCTS 7
mals. For example, consider y deﬁned by yn = 2 forn odd and yn = 0 forn even. Neither y< F ¯1 nor¯ 1 ∈U ; 4. if A ⊂N, eitherA∈Uor N\A ∈U . Remark 1.3.2 Note that by item 4 of Deﬁnition 1.3.1, we must have either {2,4,6,...}∈Uor {1,3,5,...}∈U, but not both by items 1 and 3.
Proposition 1.3.3 Suppose N = A1∪...∪An with n ∈ N, and Ai ∩Aj = ∅ for i
= j. ThenAi ∈Ufor exactly one i. Proof: Let Bi = N\Ai. If there is no i such that Ai ∈U , then Bi ∈Ufor each i. Then∅ = B1 ∩(B2 ∩...∩(Bn−1 ∩ Bn)...)∈Uby n−1 applications of Property 1 of Deﬁnition 1.3.1, which contradicts Property 3 (since ∅ is ﬁnite). Thus, Ai ∈Ufor some i. IfAi ∈Uand Aj ∈Uwith i
= j, then ∅ = Ai ∩ Aj ∈U, again contradicting Property 3. Thus, Ai ∈Ufor exactly one i. Deﬁnition 1.3.4 The equivalence relation =U on RN is deﬁned by x =U y ⇐⇒ {n : xn = yn}∈U . (1.3)
8 CHAPTER 1. NONSTANDARD ANALYSIS LITE
Given x ∈ RN, let [ x] denote the equivalence class of x with respect to the equivalence relation =U. The set ofnonstandard real numbers, denoted *R and read “star R”, is {[x]:x ∈ RN}. Any relation on R can be extended to *R. In particular, given [x],[y]∈ *R, we can deﬁne [x] U [y] is true. Proof: Let A = {n ∈ N : xn y n}. By Proposition 1.3.3,exactly one of A,B or C is in U. Example 1.3.6 Let x∈ RN be deﬁned by xn = 1 n, and ¯ r = (r,r,...) forr ∈ R. Ifr>0, then {n : xn < ¯ rn}∈U, sinceits complement is ﬁnite. Thus, [ x] U ¯ m for all m∈ N. Given any function f : R → R, we can deﬁne a function *f :*R→ R by *f([x]) = [(f(x1),f(x2),...)]. (1.5)
In other words, *f is deﬁned by evaluating f pointwise on the components of x.
1.4. INTERNAL AND EXTERNAL SETS 9
1.4 Internal and External Sets
In order to work with the nonstandard real numbers, we need to be able to talk about subsets of *R. We extend the ultraproduct construction to sets by considering sequences in (P(R))N, whereP(R) is the collection of all subsets of R, and extending the equivalence relation =U from Deﬁnition 1.3.4. Deﬁnition 1.4.1 Suppose A,B ∈ (P(R))N,[ x] ∈ *R. Wedeﬁne an equivalence relation = U by A =U B ⇐⇒ {n : An = Bn}∈U . (1.6) Let [A] denote the equivalence class of A. We deﬁne
[x]∈U [A] ⇐⇒ {n : xn ∈ An}∈U .1 (1.7) Note that [A] isnot a subset of *R; it is an equivalence class of sequences of sets of real numbers, not a set of equivalence classes of sequences of real numbers. However, we can associate it with a subset of *R in a natural way, as follows.
Deﬁnition 1.4.2 (Mostowski Collapsing Function) Given A∈ (P(R))N, deﬁne a set M([A])⊂*R by M([A]) ={[x]∈ *R :[x]∈U [A]}. (1.8) A setB ⊂*R is said to be internal if B = M([A]) for some A ∈ (P(R))N; otherwise, it is said to beexternal. A function is internal if its graph is internal.
1As above, the reader will have no trouble verifying that the deﬁnition does not depend on the choice of representatives from the equivalence classes.
10 CHAPTER 1. NONSTANDARD ANALYSIS LITE
Deﬁnition 1.4.3 Suppose B ⊂ R. Deﬁne *B = M([A]), where A ∈ (P(R))N is the constant sequence An = B for all n ∈ N. Example 1.4.4 The set of nonstandard natural numbers is *N = {[x]∈ *R :{n : xn ∈ N}∈U} . (1.9) Let ¯ N = {[¯ n]:n ∈ N}. Then¯ N ⊂ *N. Indeed, ¯ N is a proper subset of *N, as can be seen by considering [x], where xn = n for all n ∈ N. Ifm ∈ N, {n : xn =¯ mn} = {m}
∈U , so [x]
=[¯ m]. Proposition 1.4.5 *N\ ¯ N is external. Proof: Suppose *N\ ¯ N = M([A]). We shall derive a contradiction by constructing [y]∈ *N\ ¯ N with [y]
∈ M([A]).Let J = {n : An ⊂ N}. We may choosex ∈ RN suchthat xn ∈ An \ N for n ∈ N \ J, andxn = 0 forn ∈ J.Therefore, {n ∈ N : xn ∈ N} = ∅, so [ x]
∈ *N. SinceM ([A]) ⊂ *N,[ x]
∈ M([A]), so N \ J
∈U, soJ ∈U . Without loss of generality, we may assume that An ⊂N for all n ∈ N. For m ∈ N, letTm = {n ∈ N : m
∈ An}; since [¯ m]
∈M ([A]), Tm ∈U. Form∈ N∪{0}, let Sm = {n ∈ N : An ⊂{m,m +1,m+2,...}}. (1.10) Then Sm = ∩m−1 k=1 Tk, soSm ∈U . S0 = N. Let S∞ = ∩m∈NSm = {n ∈ N : An = ∅}. IfS∞ ∈U, thenM([A]) =∅, a contradiction since *N\ ¯ N
= ∅ by Example 1.4.4. Hence,S ∞
∈U . Deﬁne a sequence y ∈ RN by yn = m if n ∈ Sm+1\Sm+2,y n = 0 forn ∈ S∞. Then{n : yn ∈ N} = N\S∞ ∈U, so[ y] ∈ *N. Givenm ∈ N, {n : yn = m} = Sm+1 \Sm+2 ⊂
1.5. NOTATIONAL CONVENTIONS 11
N\Sm+2
∈U, so [ y]
∈ ¯ N, and hence [y]∈ *N\ ¯ N. However, {n : yn ∈ An}⊂S∞
∈U, so [ y]
∈ M([A]), so M([A])
= *N\N. Corollary 1.4.6 ¯ N is external.
Proof: Suppose ¯ N = M([A]). Let Bn = N\ An for eachn ∈ N. M([B]) ⊂ *N. Suppose [y] ∈ *N; we may assume without loss of generality that yn ∈ N for all n ∈ N. Then [y]∈ M([B])⇔{n ∈ N : yn ∈ Bn}∈U ⇔{n ∈ N : yn ∈ An}
∈U⇔[y]
∈ M([A]). (1.11) Thus, M([B]) = *N\ ¯ N, so *N\ ¯ N is internal, contradicting Proposition 1.4.5.
1.5 Notational Conventions
It is customary to omit *’s in many cases. Note ﬁrst that we can embed R in *R by the map r → [¯ r]. Thus, it is customary to view R as a subset of *R, and to refer to [¯ r] as r. Thus, we can also write N instead of the more awkward ¯ N. Basic relations such as <,>,≤,≥are written without the addition of a *. Functions such as sin, cos, log, ex, |•|(for absolute value or cardinality) are similarly written without *’s. Consider the function g(n)=Rn. Ifn is an inﬁnite natural number, then *Rn is deﬁned to be (*g)(n); equivalently, it is the set of all internal functions from {1,...,n} to *R. The summation symbolrepresents a function from Rn to R. Thus, ifn is an inﬁnite natural number and y ∈ *Rn, (*)n i=1yi is deﬁned. It is customary to omit the * from summations, products, or Cartesian products.
12 CHAPTER 1. NONSTANDARD ANALYSIS LITE
Thus, the following expressions are acceptable: ∀x∈ *R ex > 0; (1.12) ∃n∈ *N n i=1 xi =0 . (1.13) 1.6 Standard Models
We need to be able to consider objects such as topological spaces or probability measures in addition to real numbers. This is accomplished by considering a superstructure. We take a base set X consisting of the union of the point sets of all objects we wish to consider. For example, if we wish to consider real-valued functions on a particular topological space (T,T), we take X = R∪T. The superstructure is the class of all objects which can be obtained from the base set by iterating the operation of forming subsets; we will refer to it as the standard model generated by X.
Deﬁnition 1.6.1 Suppose X is a set, all of whose members are atomic, i.e. ∅
∈ X and no x∈ X contains any elements. Let X0 = X; (1.14) Xn+1 = P( n k=0 Xk)∪X0 (n =0 ,1,2,...) (1.15) where P is the power set operator, which associates to any set S the collection of all the subsets of S. Let
X =
∞ n=0Xn; (1.16) X is called the superstructure determined by X. For any set B ∈X, letFP(B) denote the set of all ﬁnite subsets of B.
1.6. STANDARD MODELS 13
The superstructure determined by X contains representations of subsets of X, functions deﬁned on X, Cartesian products of subsets of X, and indeed essentially all the classical mathematical constructions that can be deﬁned using X as the initial point set.2 The exact form of the representation can become quite complicated; fortunately, we need never work in detail with the superstructure representations, but only need to know they exist. The following examples illustrate how various mathematical constructions are represented in the superstructure. Example 1.6.2 An ordered pair (x,y) ∈ X2 is deﬁned in set theory as {{x},{x,y}}. x,y ∈X 0, so{x}∈X 1 and {x,y}∈X 1, so{{x},{x,y}}∈X2. Example 1.6.3 A function f:A → B, whereA,B ⊂ X, can be represented by its graph G = {(x,f(x)):x ∈ A}. From the previous example, we know that each ordered pair (x,f(x)) in the graph G is an element of X2, soG∈X 3. Example 1.6.4 The set of all functions from A to B, with A,B ⊂ X, is thus represented by an element of X4. Example 1.6.5 If N ⊂ X, ann-tuple (x1,...,xn) ∈ Xn can be represented as a function from{1,...,n}to X. Thus, if A ⊂X, thenAn is an element of X4. Example 1.6.6 Xn is an element of Xn+1. Example 1.6.7 Let (T,T) be a topological space, so that T is the set of points and T the collection of open sets. Take X = T. ThenT∈X 2. 2Indeed, formally speaking, the deﬁnition of each of these constructions is expressed in terms of set theory; see for example Bourbaki(1970).
14 CHAPTER 1. NONSTANDARD ANALYSIS LITE
Example 1.6.8 Consider an exchange economy with a set A of agents and commodity space Rk +. LetX = A ∪ R. An element of Rk + is a k-tuple of elements of X, and hence is an element of X3. A pair (x,y) withx,y ∈ Rk + can be viewed as an element of R2k, and so is also an element of X3. A preference relation is a subset of Rk +×Rk +, so it is an element of X4. A preference-endowment pair (a,e(a)) with e(a) ∈ Rk + is an element of X6. The exchange economy is a function from A to the set of preference-endowment pairs, so it is an element of X9. Remark 1.6.9 If Z ∈X, thenZ ∈X n for some n; thus, there is an upper bound on the number of nested set brackets, uniform over all elements z ∈ Z. In particular, the set {x,{x},{{x}},{{{x}}},...} is not an element of the superstructure X. Moreover,X is not an element of X.
1.7 Superstructure Embeddings Given a standard model X, we want to construct a nonstandard extension, i.e. a superstructure Y and a function *:X→Ysatisfying certain properties. Deﬁnition 1.7.1 Consider a function * from a standard model X to a superstructure Y. A∈Yis said to be internal if A ∈ *B for some B ∈X, and external otherwise. The function * :X→Yis called a superstructure embedding3 if 1. * is an injection; 2. X0 ⊂Y 0; moreoverx∈X 0 ⇒*x = x. 3. *X0 = Y0; 3Some of the properties listed can be derived from others.
1.7. SUPERSTRUCTURE EMBEDDINGS 15
4. *Xn ⊂Y n; 5. *(Xn+1 \Xn) ⊂Y n+1 \Y n (n =0 ,1,2,...); 6. x1,...,xn ∈X⇒*{x1,...,xn}= {*x1,...,*xn}; 7. A,B ∈X=⇒{A ∈ B ⇔ *A∈ *B}; 8. A,B ∈X=⇒ (a) *(A∩B)=*A∩*B; (b) *(A∪B)=*A∪*B; (c) *(A\B)=*A\*B; (d) *(A×B)=*A×*B; 9. If Γ is the graph of a function from A to B, withA,B ∈ X, then *Γ is the graph of a function from *A to *B; 10. A ∈ *Xn,B∈ A⇒ B ∈ *Xn−1; 11. A internal, A⊂ B, B ∈ *(P(C))⇒ A∈ *(P(C)). A ∈Yis said to be hyperﬁnite if A ∈ *(FP(B)) for some set B ∈X(recall FP(B) is the set of all ﬁnite subsets of B). Let *X denote {y ∈Y: y is internal}. A function whose domain and range belong to Y is said to be internal if its graph is internal.
Example 1.7.2 Suppose X = R. TakeY =*R, deﬁned via the ultraproduct construction. Let Y be the superstructure constructed with Y as the base set. Then * as deﬁned by the ultraproduct construction is a superstructure embedding. Note that Y1 contains both internal and external sets; thus, the embedding * is not onto.
16 CHAPTER 1. NONSTANDARD ANALYSIS LITE
1.8 A Formal Language
“I shall not today attempt further to deﬁne the kinds of material I understand to be embraced within that shorthand description; and perhaps I could never succeed in intelligiblydoing so. But I know it when I see it.” Justice Potter Stewart, concurring in Jacobellis v. Ohio, 378 U.S. 184 at 197.
In order to give a precise deﬁnition of a nonstandard extension, one must deﬁne a formal language L; see Chapter 2 for details. In practice, one quickly learns to recognize which formulas belong to L. The formal language L is rich enough to allow us to express any formula of conventional mathematics concerning the standard model X, with one caveat: all quantiﬁers must be bounded, i.e. they are of the form ∀x ∈ B or ∃x ∈ B where B refers to an object at a speciﬁc level Xn in the superstructure X. Thus, the quantiﬁer ∀f ∈F (R,R), where F(R,R) denotes the set of functionsfrom R to R, is allowed; the quantiﬁers ∀x∈Xand ∀x are not allowed.
1.9 Transfer Principle
Leibniz asserted, roughly speaking, that the nonstandard real numbers obey all the same properties as the ordinary real numbers. The Transfer Principle gives a precise statement of Leibniz’ assertion. The key fact which was not understood until Robinson’s work is that the Transfer Principle cannot be applied to external sets. Thus, the distinction between internal and external sets is crucial in nonstandard analysis. Given a sentence F ∈Lwhich describes the stan
1.9. TRANSFER PRINCIPLE 17
dard superstructureX, we can form a sentence *F by making the following substitutions: 1. For any set A∈X, substitute *A; 2. For any function f : A →B with A,B ∈X, substitute *f. 3. For any quantiﬁer over sets such as ∀A ∈P (B) or ∃A ∈P (B), where B ∈X, substitute the quantiﬁer ∀A ∈ *(P(B)) or ∃A ∈ *(P(B)) which ranges over all internal subsets of *B. 4. For any quantiﬁer over functions such as∀f ∈F(A,B) or ∃f ∈F (A,B), where F(A,B) denotes the set of functions from A to B for A,B ∈X, substitute the quantiﬁer ∀f ∈ *(F(A,B)) or ∃f ∈ *(F(A,B)) which ranges over all internal functions from *A to *B.
We emphasize that quantiﬁers in *F range only over internal entities. The Transfer Principle asserts that F is a true statement about the real numbers if and only if *F is a true statement about the nonstandard real numbers.
Example 1.9.1 Consider the following sentence F: ∀S ∈P(N)[ S = ∅∨∃n ∈ S ∀m∈ Sm≥n]. (1.17) F asserts that every nonempty subset of the natural numbers has a ﬁrst element. *F is the sentence ∀S ∈ *(P(N)) [S = ∅∨∃n ∈ S ∀m ∈ Sm≥n]. (1.18) *F asserts that every nonempty internal subset of *N has a ﬁrst element. External subsets of *N need not have a ﬁrst element. Indeed, *N\N has no ﬁrst element; if it did have a
18 CHAPTER 1. NONSTANDARD ANALYSIS LITE
ﬁrst element n, thenn−14 would of necessity be an element of N, but then n would be an element of N.
1.10 Saturation
Saturation was introduced to nonstandard analysis by Luxemburg (1969). Deﬁnition 1.10.1 A superstructure embedding * from X to Y is saturated5 if, for every collection {Aλ:λ ∈ Λ} with Aλ internal and |Λ|< |X|, λ ∈Λ Aλ = ∅=⇒∃λ1,...,λn n i=1 Aλi = ∅. (1.19) One can construct saturated superstructure embeddings using an elaboration of the ultraproduct construction described above. To make the saturation property plausible, we present the following proposition.
Proposition 1.10.2 Suppose *R is constructed via the ultraproduct construction of Section 1.3. If {An : n ∈ N} is a collection of internal subsets of *R, and∩n∈NAn = ∅, then ∩n0 n=1An = ∅ for some n0 ∈ N. Proof: Since An is internal, there is a sequence Bnm (m ∈ N) such that An = M([Bn•]). If A1∩•••∩An
= ∅ for all n ∈ 4We can deﬁne a function f : N→N∪{0}by f(m)=m−1. Thenn −1 is deﬁned to be *f(n). It is easy to see that, if n =[x] forx∈RN,n −1 = [(x1 −1,x2 −1,...)]. 5Our use of the term “saturated” is at variance with standard usage in nonstandard analysis or model theory. Commonly used terms are “polysaturated” (Stroyanand Luxemburg (1976))or|X|-saturated. Model theorists use the term “saturated” to mean that equation 1.19 holds provided |Λ|< |X|.
1.11. INTERNAL DEFINITION PRINCIPLE 19
N, we can ﬁnd [xn]∈ *R with [xn]∈ A1∩•••∩An for each n. Note that xn ∈ RN, so letxnm denote the m’th component of xn. Then{m : xnm ∈ B1m ∩•••∩Bnm}∈U. We may assume without loss of generality that xnm ∈ B1m∩•••∩Bnm for all n and m. Deﬁne [z] ∈ *R by zm = xmm. Then {m : zm ∈ Bnm}⊃{n,n +1 ,•••}∈U. Thus, [ z] ∈ An forall n, so∩n∈NAn
= ∅. Theorem 1.10.3 Suppose * : X→Yis a saturated superstructure embedding. If B is internal and x1,x2,... is a sequence with xn ∈ B for each n ∈ N, there is an internal sequence yn with yn ∈ B for all n ∈ *N such that yn = xn for n ∈ N. Proof: Let An = { internal sequences y : yi = xi (1 ≤ i ≤ n),y i ∈ B(i ∈ *N)}. Fix b ∈ B. If we considery deﬁned by yi = xi(1 ≤ i ≤ n), and yi = b for i>n, we see that An
= ∅. By saturation, we may ﬁnd y ∈∩ n∈NAn. Theny is an internal sequence, yn ∈ B for all n ∈ *N, andyn = xn for all n ∈ N.
1.11 Internal Deﬁnition Principle
One consequence of the Transfer Principle, the Internal Definition Principle, is used suﬃciently often that it is useful to present it separately. The informal statement of the Internal Deﬁnition Principle is as follows: any object in the nonstandard model which is describable using a formula which does not contain any external expressions is internal. For a formal statement, see Deﬁnition 2.7.1.
Example 1.11.1 The followingexamples will helpto clarify the use of the Internal Deﬁnition Principle. 1. If n ∈ *N, {m ∈ *N : m>n } is internal.
20 CHAPTER 1. NONSTANDARD ANALYSIS LITE
2. If f is an internal function and B is internal, then f−1(B) is internal. 3. If A,B are internal sets with A⊂ B, then{C ∈P(B): C ⊃ A}, the class of all internal subsets of B which contain A, is internal. 4. {x ∈ *R : x 0} is not internal; the presence of the external expression x 0 renders the Internal Deﬁnition Principle inapplicable.
1.12 Nonstandard Extensions, or Enough Already with the Ultraproducts
Deﬁnition 1.12.1 A nonstandard extension of a standard modelX is a saturated superstructure embedding * : X→Y satisfying the Transfer Principle and the Internal Deﬁnition Principle.
As we noted above, the real numbers R are deﬁned as the completion of the rational numbers Q. The completion is constructed in one of two ways: Dedekind cuts or Cauchy sequences. In practice, mathematical arguments concerning R never refer to the details of the construction. Rather, the construction is used once to establish the existence of a set R satisfying certain axioms. All further arguments are given in terms of the axioms. In the same way, the ultraproduct construction is used to demonstrate the existence of nonstandard extensions. Nonstandard proofs are then stated wholly in terms of those properties, without reference to the details of the ultraproduct construction.
1.13. HYPERFINITE SETS 21
1.13 Hyperﬁnite Sets Deﬁnition 1.13.1 Suppose that A ∈Xand * : X→Y is a nonstandard extension. Let FP(A) denote the set of ﬁnite subsets of A. A setB ⊂ *A is said to be hyperﬁnite if B ∈ *(FP(A)).
Example 1.13.2 Suppose m is an inﬁnite natural number. Consider B = {k ∈ *N : k ≤ m}. The sentence ∀m∈ N {k ∈ N : k ≤m}∈FP (N) (1.20) is true in the standard model X. By the Transfer Principle, the sentence
∀m∈ *N{k ∈ *N : k ≤m}∈*(FP(N)) (1.21) is true, so B is hyperﬁnite.
Remark 1.13.3 The Transfer Principle implies that hyperﬁnite sets possess all the formal properties of ﬁnite sets.
Theorem 1.13.4 Suppose * : X→Yis a nonstandard extension. If B ∈Xand n ∈ *N \ N, then there exists a hyperﬁnite set D with |D| Proof: See Theorem 2.8.3.
Proposition 1.13.5 Suppose that * :X→Yis a nonstandard extension. Suppose that B is hyperﬁnite and A⊂ B, A internal. Then A is hyperﬁnite.
Proof: See Proposition 2.6.5.
22 CHAPTER 1. NONSTANDARD ANALYSIS LITE
1.14 Nonstandard Theorems Have Standard Proofs
Although nonstandard proofs never make use of the details of the ultraproduct construction, the construction shows that the existence of nonstandard models with the assumed properties follows from the usual axioms of mathematics. Any nonstandard proof can be rephrased as a proof from the usual axioms by reinterpretingeach line in the proof as a statement about ultraproducts. Consequently, any theorem about the standard world which has a nonstandard proof is guaranteed to have a standard proof, although the proof could be exceedingly complex and unintuitive. The important point is that, if we present a nonstandard proof of a standard statement, we know that the statement follows from the usual axioms of mathematics.
Chapter 2
Nonstandard Analysis Regular
2.1 Warning: Do Not Read this Chapter
There is nothing innately diﬃcult about the mathematical logic presented in this Chapter. However, learning it does require a perspective that is a little foreign to many mathematicians and economists, and this may produce a certain degree of intimidation. Do not worry! At the ﬁrst sign of queasiness, rip this Chapter and Appendix A out of the book, and go on to Chapter 3!
2.2 A Formal Language In this Section, we specify a formal language L in which we can make statements about the superstructure X. One must ﬁrst specify the atomic symbols, the vocabulary in which the language is written. Next, one speciﬁes grammar rules which
23
24CHAPTER 2. NONSTANDARDANALYSISREGULAR
determine whether a given string of atomic symbols is a valid formula in the language. The grammar rules are described inductively, showing how complicated formulas can be built up from simpler formulas. Implicit in the grammar rules is a unique way to parse each formula, so that there is only one sequence of applications of the rules that yields the given formula. Together, the atomic symbols and the grammar rules determine the syntax of the language. All formulas will be given the conventional mathematical interpretation; the formal speciﬁcation of the interpretation process is given in Section 2.4. Deﬁnition 2.2.1 The atomic symbols of L are the following: 1. logical connectives ∨∧⇒⇔¬ 2. variables v1 v2 ... 3. quantiﬁers ∀∃ 4. brackets [ ] 5. basic predicates ∈ = 6. constant symbols Cx (one for each x ∈X) 7. function symbols Cf (one for each function f with domain x, range y and x,y ∈X) 8. set description symbols { : } 9. n-tuple symbols (,)
Deﬁnition 2.2.2 A ﬁnite string of atomic symbols is a formula of L if it can be derived by iterative application of the following rules:
2.3. EXTENSIONS OF L 25
1. If F is a variable or constant symbol, then F is a term;
2. If F is a term and Cf is a function symbol, then Cf(F) is a term;
3. If F is a formula and G is a term, F does not contain the quantiﬁer ∀vi, the quantiﬁer ∃vi, or a term of the form{vi ∈ H : J}, andG does not contain the variable vi, then{vi ∈ G : F} is a term; 4. If F1,...,Fn are terms, then {F1,...,Fn} and (F1,...,Fn) are terms; 5. If F and G are terms, then [F ∈ G] and [ F = G] are formulas; 6. If F is a formula, [¬F] is a formula; 7. If F and G are formulas, [F ∨G], [F ∧G], [F ⇒ G], and [F ⇔ G] are formulas; 8. If F is a formula and G is a variable, a constant symbol, or a term, and neither F nor G contains ∃vi, ∀vi, or a term of the form {vi ∈ H : J}, then [ ∃vi ∈ G[F]] and [∀vi ∈ G[F]] are formulas. 2.3 Extensions of L In practice, mathematics is always written with a certain degree of informality, since strict adherence to the formulas of a formal language like L would make even simple arguments impenetrable. One of the key ways a formal language is extended is through the use of abbreviations. It is important to be able to recognize whether or not a given formula, expressed with the degree of informality commonly used in
26CHAPTER 2. NONSTANDARDANALYSISREGULAR
mathematics, is in fact expressible as a formula in the formal language. In the following examples, we show how to construct formulas in L using the grammar rules and abbreviations; we also give some examples of strings of atomic symbols which are not formulas.
Example 2.3.1 The symbol string ∀v1[v1 = v1] is not a formula, because the quantiﬁer is not of the correct form. However, ∀v1 ∈ Cx[v1 = v1] is a formula ofL provided x is an element of the superstructure X. It is important that, every time a quantiﬁer is used, we specify an element of the superstructure X over which it ranges.
Example 2.3.2 The symbol strings
∧[v1∀CX (2.1) v1 ∈ v2 ⇔¬ (2.2) ∀[v2 ∈ v3] (2.3) are not formulas because they cannot be constructed by iterative application of the grammar rules.
Example 2.3.3 The induction axiom for the natural numbers N is the following:
∀v1 ∈ CP(N) [[v1 = ∅]∨[∃v2 ∈ v1 [∀v3 ∈ v1[(v2,v3)∈ Cx]]]] (2.4) where x = {(n,m) ∈ N2 : n ≤ m}. Note that, although we could describe the set x in our language L, there is no need for us to do so; sincex ∈X, we know there is a constant symbol Cx corresponding to it already present in our language.
2.3. EXTENSIONS OF L 27 Given x ∈X, we can substitute x for the more cumbersome Cx, the constant symbol which corresponds to x. We can omit all brackets[,] except those which are needed for clarity. We can substitute function and relation symbols such as f(x) or m≤ n for the more awkward Cf(x) or (m,n)∈ Cx where x is the graph of the relation “≤”. Finally, we can substitute more descriptive variable names (by using conventions such as small letters to denote individual elements, capital letters sets, etc.).
Example 2.3.4 Using the above abbreviations, the induction axiom can be written as ∀S ∈P(N) ∃n∈ S ∀m ∈ Sn≤m; (2.5) since we know this is an abbreviation for a formula in L, we know that we can treat it as if it were in L. Example 2.3.5 The grammar of the language L does not permit us to quantify directly over functions. However, we can easily get around this restriction by representing functions by their graphs. Given A,B ∈X, letF(A,B) denote the set of all functions from A to B,G(A,B) the set of graphs of functions in F(A,B). G(A,B) ∈X. Consider the function FAB :G(A,B)×A→ B deﬁned by FAB(Γ,a)=f(a) wheref is the function whose graph is Γ. (2.6) Note that FAB ∈F (G(A,B) × A,B), so there is a func-tion symbol in L corresponding to FAB. The formula ∀f ∈ F(A,B)[G(f)] is an abbreviation for ∀Γ∈G (A,B)[G(FAB(Γ,•)]. (2.7) For example, the sentence ∀f ∈F([0,1],R)∃x∈ [0,1]∀y ∈ [0,1]f(x)≥ f(y) (2.8)
28CHAPTER 2. NONSTANDARDANALYSISREGULAR
which asserts (falsely) that every function from [0,1] to R assumes its maximum is an abbreviation for the following sentence in L: ∀Γ∈G([0,1],R)∃x∈ [0,1]∀y ∈ [0,1]F[0,1]R(Γ,x) ≥ F[0,1]R(Γ,y) (2.9) Example 2.3.6 As noted in Section 1.5, summation is a function deﬁned on n-tuples or sequences. Thus, expressions liken∈N xn are abbreviations for formulas in L. Example 2.3.7 Suppose x,y ∈X n. We can deﬁne a relation ⊂ by saying that x ⊂ y is an abbreviation for ∀z ∈ Xn[z ∈ x ⇒ z ∈ y]. In the induction axiom discussed in the previous example, we did not write ∀S ⊂ N ∃n ∈ S ∀m∈ Sn≤m (2.10) because the quantiﬁer is required to be of the form ∀S ∈ C for some set C ∈X. A key issue in nonstandard models is the interpretation of quantiﬁers over subsets; for this reason, we will not use the abbreviation ⊂ within quantiﬁers.
2.4 Assigning Truth Value to Formulas
We shall be interested in interpreting formulas in the standard universe, as well as in nonstandard models. For this reason, it is important that we specify precisely the procedure by which formulas are interpreted. In the standard world, every formula F in L is interpreted according to the conventional rules for interpreting mathematical formulas. Thus, the logical symbol ∨ will have the conventional interpretation “or,” while the quantiﬁer ∀ has the conventional
2.4. ASSIGNING TRUTH VALUE TO FORMULAS 29
interpretation “for all.” If all the variables appearing in F appear within the scope of a quantiﬁer, then F can be assigned an unambiguous truth value; if not, then the truth value of F depends on how we choose to interpret the variables which are not quantiﬁed. This leads to the following set of deﬁnitions.
Deﬁnition 2.4.1 Anoccurrence of a variable vi ina formula F is bound if there is a formula E such that F = ...E... , E =[∀vi ∈ G[H]] or E =[∃vi ∈ G[H]] or F ={vi ∈ G : H} (2.11) and the occurrence of vi is inside E. If the occurrence of vi is not bound, it is said to be free. If every occurrence of each variable in F is bound, F is said to be a sentence. Deﬁnition 2.4.2 An interpretation of L in X is a function I mapping {v1,v2,...} to X. In the next deﬁnition, we show how to extend an interpretation so that it assigns an element of X to each term and a truth value t (“true”) or f (“false”) to each formula.
Deﬁnition 2.4.3 The value of a constant symbol or set description J under the interpretation I, denoted I(J), and the truth value of a formula F under the interpretation I, denoted I(F), are deﬁned inductively as follows:
1. For any constant symbol Cx, I(Cx)=x (i.e. every constant symbol is interpreted as the corresponding element of X). 2. If F is Cf(G), where Cf is a function symbol and G is a term, then I(F)=f(I(G)) if I(G) is deﬁned and I(G) is an element of the domain of f; otherwise,I(F) is undeﬁned.
30CHAPTER 2. NONSTANDARDANALYSISREGULAR
3. If J is a term {vi ∈ H : K}, then I(J)={x ∈ I(H):I (K)=t where I (vi)=x and I (vj)=I(vj) forj
= i} (2.12) if I(H) is deﬁned, and I(J) is undeﬁned otherwise.
4. (a) If F is (F1,...,Fn), then I(F)=⎧ ⎪ ⎨ ⎪ ⎩ (I(F1),...,I(Fn)) if I(F1),...,I(Fn) are all deﬁned; undeﬁned otherwise. (2.13) (b) If F is {F1,...,Fn}, then I(F)=⎧ ⎪ ⎨ ⎪ ⎩ {I(F1),...,I(Fn)} if I(F1),...,I(Fn) are all deﬁned; undeﬁned otherwise. (2.14) 5. (a) If F is [G ∈ H], where G and H are terms, then I(F)=t if I(G) andI(H) are deﬁned and I(G)∈ I(H), while I(F)=f otherwise. (b) If F is [G = H], where G and H are terms, then I(F)=t if I(G) andI(H) are deﬁned and I(G)= I(H), while I(F)=f otherwise. 6. If F is [¬G], then I(F)=t if I(G)=f and I(F)=f if I(G)=t. 7. (a) If F is [G ∨ H] thenI(F)=t if I(G)=t or I(H)=t and I(F)=f otherwise; (b) If F is [G∧ H] thenI(F)=t if I(G)=t and I(H)=t and I(F)=f otherwise;
2.4. ASSIGNING TRUTH VALUE TO FORMULAS 31
(c) If F is [G ⇒ H] thenI(F)=t if I(G)=f or I(H)=t and I(F)=f otherwise; (d) If F is [G ⇔ H] thenI(F)=t if I(G)=t and I(H)=t or I(G)=f and I(H)=f and I(F)=f otherwise. 8. (a) If F is [∀vi ∈ G[H]], then I(F)=t if I(G) is deﬁned and, for every x ∈X , I (H)=t, where I (vi)=x and I (vj)=I(vj) for every j
= i; I(F)=f otherwise. (b) If F is [∃vi ∈ G[H]], then I(F)=t if I(G) is deﬁned and there exists some x∈ I(G) such that I (H)=t, whereI (vi)=x and I (vj)=I(vj) for every j
= i; I(F)=f otherwise. Remark 2.4.4 (The King of France Has a Beard) It would be possible, but exceedingly tedious, to restrict our language so that it is impossible to write formulas that contain expressions of the form f(x) wherex is outside the domain of f. We have instead taken the route of allowing them in the language, and have speciﬁed a more or less arbitrary assignment of truth value. In practice, such nonsensical formulas are easily spotted, and so no problems will arise. Note that, assuming for concreteness that f : N →N, our assignment of truth value has the following consequences: • I(f(π) = 0) =f. • The formula f(π)
= 0 is ambiguous. It is not an element of L. – If it is an abbreviation for ¬[f(π) = 0], then I(f(π)
= 0) =t. – If it is an abbreviation for (f(x),0) ∈ C where C = {(y,z)∈ N2 : y
= z}, thenI(f(π)
= 0) =f.
32CHAPTER 2. NONSTANDARDANALYSISREGULAR
Proposition 2.4.5 If F is a sentence, then I(F) is independent of the interpretation I.
Proof: Let vi be a variable which occurs in F. Since every occurrence of vi is bound, I(F) is independent of I(vi) by items 3 and 8 of Deﬁnition 2.4.3. Thus, I(F) is independent of I.
Deﬁnition 2.4.6 Suppose F is a sentence. We say F holds in X if I(F)=t for some (and thus every) interpretation I of L in X, andF fails in X if I(F)=f for some (and thus every) interpretation I of L in X.
2.5 Interpreting Formulas in Superstructure Embeddings Suppose that * is a superstructure embedding from X to Y. We shall deﬁne the ﬁrst of two languages which we will use to describe Y. Deﬁnition 2.5.1 The internal language *L is deﬁned in exactly the same way as the language L, except that the set of constant symbols is {Cx : x ∈ *X} = {Cy : y ∈ Y,y internal} and the set of function symbols is {Cf : f is an internal function from x to y for some x,y ∈ *X}. Aninterpretation of *L in *X is a function I :{v1,v2,...}→*X. Given an interpretation I of *L in *X, we can extend it to assign truth values to formulas in *L in exactly the same way as in Deﬁnition 2.4.3.
Remark 2.5.2 Deﬁnition 2.5.1 forces the interpretation of a variable to be internal. The astute reader may wonder
2.6. TRANSFER PRINCIPLE 33
whether, given an interpretation I, I(J) is internal for complex terms J such as {vi ∈ G : F} or Cf(F). If the superstructure embedding satisﬁes the Internal Deﬁnition Principle, and I is an interpretation of *L in *X, thenI(J) is internal for every term in *L. However, since the assignment of truth values in the previous deﬁnition makessense whether or not I(J) is internal, we will defer the formal discussion of the Internal Deﬁnition Principle to Section 2.7. Proposition 2.5.3 If F is a sentence in *L, thenI(F) is independent of the interpretation I. Deﬁnition 2.5.4 Suppose F is a sentence in *L. We sayF holds in *X if I(F)=t for some interpretation I of *L in *X, andF fails in Y if *I(F)=f for some interpretation I of *L in *X.
2.6 Transfer Principle Deﬁnition 2.6.1 Given an interpretation I of L in X, we can associate an interpretation *I of *L in *X by specifying (*I)(vi) = *( I(vi)) for each variable vi. Given a formula F ∈L, we associate a formula *F ∈ *L by replacing each constant symbol Cx (x ∈X) with the constant symbol C*x and each function symbol Cf with the function symbol C*f. Deﬁnition 2.6.2 A superstructure embedding * from X to Y satisﬁes the Transfer Principle if, for every formula F ∈L and every interpretation I of L in X, I(F)=t if and only if*I(*F)=t. (2.15) Proposition 2.6.3 If a superstructure embedding * from X to Y satisﬁes the Transfer Principle, and F is a sentence in L, thenF holds in X if and only if *F holds in *X.
34CHAPTER 2. NONSTANDARDANALYSISREGULAR
Remark 2.6.4 The astute reader may have noticed a very important point arising in the case that F involves quantiﬁers over sets or other objects not in X0. Assume for concreteness that we are considering a formula F =[ ∀v1 ∈C P(X)[G]]; I(F)=t if the property speciﬁed by G holds for every subset of X.* I(F)=t if the property speciﬁed by G holds for every element of *(P(X)) ⊂P(*X)=P(Y0); thus, the property speciﬁed by G need hold only for internal subsets of Y0; it need not hold for external subsets of Y0. Proposition 2.6.5 Suppose that the superstructure embedding * from X to Y satisﬁes the Transfer Principle, B is hyperﬁnite, and A⊂ B, A internal. Then A is hyperﬁnite. Proof: Since B is hyperﬁnite, B ∈ *(FP(C)) for some C ∈ X. SinceA is internal, and A ⊂ *C, A∈ *P(C) by item 11 of Deﬁnition 1.7.1. The sentence ∀B ∈FP(C)∀A∈P (C) [[ ∀x∈ C[x∈ A ⇒x ∈ B]] =⇒ [A∈FP (C)]], (2.16) which asserts that a subset of a ﬁnite subset is ﬁnite, holds in X; by transfer, the sentence ∀B ∈ *FP(C)∀A∈ *P(C) [[ ∀x∈ *C[x∈ A⇒ x ∈ B]] =⇒ [A∈ *FP(C)]], (2.17) holds in *X. Thus,A∈ *FP(C), so A is hyperﬁnite.
2.7 Internal Deﬁnition Principle Deﬁnition 2.7.1 A superstructure embedding * from X to *X satisﬁes the internal deﬁnition principle if, for every term J in *L, and every interpretation I of *L in *X, I(J) is internal or undeﬁned.
2.8. NONSTANDARD EXTENSIONS 35
2.8 Nonstandard Extensions
Deﬁnition 2.8.1 A nonstandard extension of a superstructure X is a superstructure embedding *:X→Ywhich is saturated (Deﬁnition 1.10.1) and satisﬁes the Transfer Principle (Deﬁnition 2.6.1) and the Internal Deﬁnition Principle (Deﬁnition 2.7.1).
Theorem 2.8.2 Suppose * : X→Yis a nonstandard extension, and N ⊂X. Then * N\N
= ∅.n∈ *N\N ⇒ n>m for every m ∈ N. Proof: Given m ∈ N, letAm = {n ∈ *N : n>m }. Am isinternal by the Internal Deﬁnition Principle; Am ⊃{n ∈ N :n>m }, soAm
= ∅. Givenm1,...,mk ∈ N with k ∈ N, ∩k i=1Ami = Amax{m1,...,mk}
= ∅. By Saturation ∩m∈NAm
= ∅. Take any n ∈∩ m∈NAm; thenn ∈ *N, but n>mfor everym ∈ N, son ∈ *N\N. The sentence ∀n∈ N ∀m∈ N [[mn]] (2.18) holds in X; by Transfer, the sentence ∀n ∈ *N ∀m∈ *N [[mn]] (2.19) holds in *X. Now suppose n is any element of *N\N, and ﬁx m ∈ N. Since * is a superstructure embedding, m ∈ *N. Since n ∈ *N\N, m
= n. Suppose nmfor every m ∈ N.
36CHAPTER 2. NONSTANDARDANALYSISREGULAR
Theorem 2.8.3 Suppose * : X→Yis a nonstandard extension. If B ∈Xand n ∈ *N \ N, then there exists a hyperﬁnite set D with |D| mfor each m ∈ N by Theorem 2.8.2. Let Λ = B, Aλ = {D ∈ *FP(B):* λ ∈D, |D| = ∅ by saturation; if D is any element of ∩λ∈ΛAλ, thenD ∈ *FP(B), so D is hyperﬁnite, |D| 2.9 The External Language
We shall deﬁne a language ˆ L which permits us to refer to external objects, including the set of ordinary natural numbers N, viewed as a subset of its nonstandard extension *N, or the set of all nonstandard real numbers inﬁnitely close to a given real number. We can use all the usual axioms of conventional mathematics to make arguments involving formulas in ˆ L; however, the Transfer Principle cannot beapplied to formulas in ˆ L. Deﬁnition 2.9.1 The external language ˆ L is deﬁned in ex-actly the same way as the language L, except that the setof constant symbols is {Cy : z ∈ Y}and the set of func-tion symbols is {Cf : f is a function from x to y for some x,y ∈Y}. Aninterpretation of ˆ L in Y is a function from {v1,v2,...} to Y. Given a formula F in ˆ L and such an in-terpretation I, the truth value I(F) is deﬁned exactly as in Deﬁnition 2.4.3. If F is a sentence of ˆ L, we say thatF holdsin Y if I(F)=t for some interpretation of I(F) inY, and F fails in Y otherwise.
2.10. THE CONSERVATION PRINCIPLE 37
2.10 The Conservation Principle
The usual axiomatization of set theory is due to Zermelo and Frankel. The Zermelo-Frankel axioms, with the addition of the Axiom of Choice, are collectively referred to as ZFC. Theorem 2.10.1 (Los,Robinson,Luxemburg) Let X be a superstructure in a model of ZFC. Then there exists a nonstandard extension * : X→Y . Proof: See Appendix A. Corollary 2.10.2 (The Conservation Principle) Let A be a set of axioms containing the axioms of ZFC. Suppose F is a sentence in L which can be deduced from A and the assumption that there exists a nonstandard extension * : X→Y. Then F has a proof using the axioms of Aalone.
38CHAPTER 2. NONSTANDARDANALYSISREGULAR
Chapter 3
Euclidean, Metric and Topological Spaces
In this Chapter, we explore the nonstandard formulation of the basic results in Euclidean, Metric and Topological Spaces. The results are due for the most part to Robinson (1966) and Luxemburg (1969). The results stated here are of considerable use in applications of nonstandard analysis to economics. In addition, the proofs given here illustrate how the properties of nonstandard extensions are used in writing proofs. We form a superstructure by taking X0 to be the union of the point sets of all the spaces under consideration, and suppose that * :X→Yis a nonstandard extension.
3.1 Monads Deﬁnition 3.1.1 Suppose (X,T) is a topological space. If x ∈ X, themonad of x, denoted μ(x), is ∩x∈T∈T*T. If y ∈ *X and y ∈ μ(x), we write y x (read “y is inﬁnitely close to x”). Deﬁnition 3.1.2 Suppose (X,d) is a metric space. If x ∈
39
40 CHAPTER 3. REAL ANALYSIS
*X, themonad of x, denoted μ(x), is {y ∈ *X :* d(x,y) 0 }. Ifx,y ∈ *X and y ∈ μ(x), we write y x (read “y is inﬁnitely close to x”).
Proposition 3.1.3 Suppose (X,d) is a metric space, and x ∈ X. Then the monad of x (viewing X as a metric space) equals the monad of x (viewing X as a topological space).
Proof: Suppose y is inthe metricmonad of x. Then * d(x,y) 0. Suppose x∈ T ∈T. Then there exists δ ∈ R++ such that the formula d(z,x) <δ⇒ z ∈ T (3.1) holds in X. By Transfer, *d(z,x) <δ⇒ z ∈ *T (3.2) holds in *X. Since this holds for each T satisfying x ∈ T ∈T, y is in the topological monad of x. Conversely, suppose y is in the topological monad of x. Choose δ ∈ R++ and let T = {z ∈ X : d(z,x) <δ }. x ∈T ∈T, soy ∈ *T. Therefore, *d(x,y) <δfor each δ ∈ R++,so * d(x,y) 0. Thus, y is in the metric monad of x. Remark 3.1.4 The topological monad of x can be deﬁned for an arbitrary x ∈ *X, not just for x ∈ X. However, it is not very well behaved.
Proposition 3.1.5 (Overspill) Suppose (X,T ) is a topological space, x ∈ X, andA is an internal subset of *X. 1. If x ∈ A ⊂ μ(x), there exists S ∈ *T with A ⊂ S ⊂ μ(x). 2. If A ⊃ μ(x), thenA ⊃ *T for some T satisfying x ∈ T ∈T.
3.1. MONADS 41
3. If (X,T) is a T1 space and μ(x) is internal, then μ(x)= {x}∈T. Proof: Let T = {T ∈T: x ∈ T}. (1)Suppose A ⊂μ(x). By Theorem 1.13.4, there exists a hyperﬁnite set S⊂*T such that T ∈T implies *T ∈S . Let S = {T ∈S: T ⊃ A}; S is internal by the Internal Deﬁnition Principle, and hyperﬁnite by Proposition 1.13.5. Let S = ∩T∈ST. ThenA ⊂ S ⊂ μ(x). Since T is closed under ﬁnite intersections, *T is closed under hyperﬁnite intersections, by Transfer. Therefore S ∈ *T. (2) Suppose A ⊃ μ(x). Given T ∈T , letAT =* T \A.A T is internal by the Internal DeﬁnitionPrinciple. ∩T∈TAT =( ∩T∈T*T)\A = μ(x)\A = ∅. By Saturation, there existT 1,...,Tn (n ∈ N) such that ∩n i=1ATi = ∅, soA ⊃∩ n i=1*Ti = *(∩n i=1Ti). Since x ∈∩ n i=1Ti ∈T, the proof of (2) is complete. (3) Suppose μ(x) is internal and (X,T ) is aT1 space. By (2), there exists T ∈ *T such that μ(x) ⊂ *T ⊂ μ(x), so μ(x)=* T. If y ∈ X, y
= x, then there exists S ∈Twith x ∈ S and y
∈ S. By Transfer, y
∈ *S, soy
∈ μ(x)=* T. By Transfer, y
∈ T. Sincey is an arbitrary element of X diﬀerent from x, T = {x}. Then μ(x)=* T = {x} byTransfer.
Proposition 3.1.6 (Overspill) Suppose A is an internal subset of *N. 1. If A ⊃ *N \ N, thenA ⊃{n,n +1,...} for some n ∈ N. 2. If A ⊃N, thenA ⊃{ 1,2,...,n} for some n ∈ *N\N. Proof: We could prove this as a corollary of Proposition 3.1.5 by considering X = N∩{∞}with the one-point compactiﬁcation metric. However, we shall present a direct proof.
42 CHAPTER 3. REAL ANALYSIS
Every nonempty subset of N has a ﬁrst element; by transfer, every nonempty internal subset of *N has a ﬁrst element. (1) Let B = {n ∈ *N : ∀m ∈ *N[m ≥ n ⇒ m ∈ A]}.B is internal by the Internal Deﬁnition Principle. Let n be the ﬁrst element of B. SinceA ⊃ *N\N, B ⊃ *N\N, so n ∈ N. (2) Let B = {n ∈ *N : ∀m ∈ *N[m ≤ n ⇒ m ∈ A]}. Band * N\B are internal by the Internal Deﬁnition Principle.If * N\ B = ∅, we are done. Otherwise, let n be the ﬁrst element of *N\B. SinceA⊃ N, B ⊃ N, son ∈ *N\N. Proposition 3.1.7 Suppose (X,T) is a topological space. Then X is Hausdorﬀ if and only if for every every x,y ∈ X with x
= y, μ(x)∩μ(y)=∅. Proof: Suppose X is Hausdorﬀ. If x,y ∈ X and x
= y, we may ﬁnd S,T with x ∈ S ∈T, y ∈ T ∈T, andS ∩T = ∅. *S∩*T = *(S∩T)=∅, by Transfer. μ(x)∩μ(y) ⊂ *S∩*T = ∅. Comversely, suppose μ(x) ∩ μ(y)=∅. By Proposition 3.1.5, we may ﬁnd S,T ∈ *T with x ∈ S ⊂ μ(x) andy ∈ T ⊂μ(y), and thus S ∩T =∅. Thus, the sentence ∃S ∈ *T∃ T ∈ *T [x∈ S ∧ y ∈ T ∧ S ∩T = ∅] (3.3) holds in *X. By Transfer, the sentence ∃S ∈T ∃T ∈T [x∈ S ∧ y ∈ T ∧ S ∩T = ∅] (3.4) holds in X, soX is Hausdorﬀ. Deﬁnition 3.1.8 If (X,T) is a Hausdorﬀ space, y ∈ *X, and y ∈ μ(x), we write x = ◦y (read “x is the standard part of y”).
3.1. MONADS 43
Proposition 3.1.9 Suppose {xn : n ∈ *N} is an internal sequence of elements of *R. Then the standard sequence {◦xn : n ∈ N} converges to x ∈ R if and only if there exists n0 ∈ *N\N such that xn x for every n ≤ n0,n∈ *N\N. Proof: Suppose ◦xn converges to x. Fix δ ∈ R++. Thereexists nδ ∈ N such that n ≥ nδ,n∈ N implies |◦xn −x| <δ/ 2; thus, |xn −x| <δ. Thus, given δ ∈ R++ and k ∈ N, let Aδk = {n ∈ *N : n ≥ k and |xm −x| <δfor each m ∈ {k,...,n}}. For any ﬁnite collection {(δ1,k1),...(δn,kn)} with ki ≥ nδi, ∩n i=1Aδiki
= ∅. Let Λ = {(δ,k) ∈ R++ ×N : k ≥ nδ}. By Saturation, ∩(δ,k)∈ΛAδk
= ∅. Choosen 0 ∈∩ (δ,k)∈ΛAδk. Thenn0 ∈ *N\N; givenn ∈ *N\N with n ≤ n0, |xn −x| 0. Conversely, suppose there exists n0 ∈ *N\N such thatn ∈ *N, n ≤ n0 implies xn x. Given δ ∈ R++, letA = {n ∈ *N : |xn − x| < δ/ 2}∪{ n ∈ *N : n>n 0}.A is internal and contains *N \ N. By Proposition 3.1.6, A ⊃{n,n +1,...} for some n ∈ N. Thus,|◦xm −x| <δfor m ∈ N satisfying m ≥ n. Therefore ◦xn converges to x. Proposition 3.1.10 Suppose {xn : n ∈ N} is a sequence of elements of R. Thenxn → x ∈ R if and only if xn x for every n ∈ *N\N. Proof: Suppose xn → x. Given ∈ R++, there existsn 0 ∈ N such that the sentence ∀n ∈ N[n≥ n0 ⇒|xn −x| <] (3.5) holds in X. By Transfer, the sentence ∀n ∈ *N[n ≥n0 ⇒|xn −x| <] (3.6) holds in *X. Ifn ∈ *N\N, then|xn −x| <; since is an arbitrary element of R++, xn x. Conversely, suppose xn x for all n ∈ *N \ N. Forn ∈ N, ◦xn = xn, soxn →x by Proposition 3.1.9.
44 CHAPTER 3. REAL ANALYSIS
3.2 Open and Closed Sets Proposition 3.2.1 Suppose (X,T) is a topological space. Then A ⊂ X is open if and only if μ(x) ⊂ *A for every x ∈ A. Proof: If A is open and x∈ A, thenμ(x) ⊂*A by Deﬁnition 3.1.1. Conversely, suppose μ(x) ⊂ *A for every x ∈ A. By Proposition 3.1.5, we may ﬁnd S ∈ *T with x ∈ S ⊂ μ(x). Thus, the sentence ∃S ∈ *T x ∈ S ⊂*A (3.7) holds in *X, so the sentence ∃S ∈T x ∈ S ⊂A (3.8) holds in X by Transfer. Thus, A is open. Proposition 3.2.2 Suppose (X,T) is a topological space. Then A ⊂ X is closed if and only if y ∈ *A implies x ∈ A for every x ∈ X such that y ∈ μ(x). Proof: Let B = X \A. Suppose A is closed. If y ∈ *A and y ∈ μ(x) withx ∈ X\A , thenx∈ B. SinceA is closed, B is open. Since y ∈ μ(x),y ∈ *B by Proposition 3.2.1. *B ∩*A = *(B ∩A)=∅, by Transfer. Thus, y
∈ *A, a contradiction. Conversely, suppose y ∈ *A implies x ∈ A for everyx ∈ X such that y ∈ μ(x). Suppose x ∈ B. Then we must have y ∈ *X \*A =*B for every y ∈ μ(x). Accordingly, B is open by Proposition 3.2.1, so A is closed. Proposition 3.2.3 Suppose(X,T) is a topological space, and A ⊂ *X is internal. Then {x ∈ X : ∃y ∈ A [y ∈ μ(x)]} is closed.
3.3. COMPACTNESS 45
Proof: Let C = {x∈ X : ∃y ∈ A [y ∈ μ(x)]}; we shall show that B = X \C is open. Let D =*X \A; D is internal by the Internal Deﬁnition Principle. If x ∈ B, thenD ⊃ μ(x), so D ⊃*T for some T satisfying x∈ T ∈T, by Proposition 3.1.5. If y ∈ T, thenμ(y) ⊂ *T, soD ⊃ μ(y), so y ∈ B. Thus, B is open, so C is closed.
3.3 Compactness Deﬁnition 3.3.1 Let (X,T) be a topological space and y ∈ *X. We sayy is nearstandard if there exists x∈ X such that y x. We letns(*X) denote the set of nearstandard points in *X. Theorem 3.3.2 Let (X,T ) be a topological space. Then (X,T ) is compact if and only if every y ∈ *X is nearstandard. Proof: Suppose (X,T) is compact, and there is some y ∈ *X which is not nearstandard. Then for every x ∈ X, there exists Tx with x ∈ Tx ∈Tand y
∈ *Tx. {Tx : x ∈ X} is thus an open cover of X; let{Tx1,...,Txn} be a ﬁnite subcover (so n ∈ N). Since * is a superstructure embedding, ∪n i=1*Txi = *(∪n i=1Txi)=*X, soy
∈ *X, a contradiction. Conversely, suppose that every y ∈ *X is nearstandard.Let {Tλ : λ ∈ Λ} be an open cover of X. Let Cλ = X \ Tλ. If there is no ﬁnite subcover, then for every collection {λ1,...,λn} with n ∈ N, ∩n i=1Cλi
= ∅. ∩n i=1*Cλi = *(∩n i=1Cλi)
= ∅.|Λ|≤|X 1|, so by saturation, C = ∩λ∈Λ*Cλ
= ∅. Choose anyy ∈ C. Givenx ∈ X, there exists λ such that x ∈ Tλ. Sincey ∈ C ⊂ *Cλ, y
∈ *Tλ, soy
x. Sincex is an arbitrary element of X, y is not nearstandard, a contradiction. Thus, {Tλ : λ ∈ Λ} has a ﬁnite subcover, so X is compact.
46 CHAPTER 3. REAL ANALYSIS
Deﬁnition 3.3.3 x∈ *R is said to be ﬁnite if there is some n ∈ N such that x ≤n. Proposition 3.3.4 Suppose y ∈ *R, andy is ﬁnite. Then y is nearstandard.
Proof: Let A = {z ∈ R : zx−δ. Butzyby the deﬁnitions ofA and x. Therefore x−δ x); 2. transitivity (for all x,y,z ∈ X, x y,y z ⇒ x z); 3. continuity ({(x,y)∈ X2 : xy} is open).
3.3. COMPACTNESS 47
Then X contains a maximal element with respect to , i.e. there is some x∈ X such that there is no z ∈ X with z x. Proof: By Theorem 1.13.4, there exists a hyperﬁnite set A such that T ∈T⇒∃ x ∈ A [x ∈ *T]. Since is irreﬂexiveand transitive, any ﬁnite set B ⊂ X contains a maximal element with respect to . By Transfer, any hyperﬁnite set contains a maximal element with respect to *. Let y be such a maximal element of A. SinceX is compact, there exists x ∈ X such that y x by Theorem 3.3.2. Suppose z ∈ X and z x. Then there exists S,T withx ∈ T ∈Tand z ∈ S ∈Tsuch that v w for each v ∈ S and w ∈ T. By transfer, v*w for each v ∈ *S and each w ∈ *T. But there existsv ∈ *S ∩A, and sov*y,a contradiction. Thus, x is maximal in X with respect to .
Proposition 3.3.7 Suppose(X,T) is a regular topological space, and A ⊂ *X is internal. Suppose further that every y ∈ A is nearstandard. Then {x ∈ X : ∃y ∈ A [y ∈ μ(x)]} is compact. Proof: Let C = {x ∈ X : ∃y ∈ A [y ∈ μ(x)]}. Suppose {Cλ : λ ∈ Λ} is a collection of relatively closed subsets of C, with∩λ∈ΛCλ = ∅, but ∩n i=1Cλi
= ∅ for every ﬁnite collection {λ1,...,λn}; C is closed by Proposition 3.2.3, so Cλ is closed in X. Given x ∈ C with x
∈ Cλ, we mayﬁnd sets Sλx,Tλx ∈Tsuch that Cλ ⊂ Sλx, x ∈ Tλx, andS λx ∩ Tλx = ∅. Let Λ = {(λ,x):x
∈ Cλ}. Given any ﬁnite collection {(λ1,x1),...,(λn,xn)}⊂Λ , ∩n i=1Sλixi is an open set; because it contains ∩n i=1Cλi
= ∅, it is not empty.Choose c ∈∩ n i=1Cλi. Then c ∈ C, so there exists a ∈ Awith a ∈ μ(c). ∩n i=1*Sλixi =*∩n i=1 Sλixi ⊃ μ(c) by Proposi-tion 3.2.1. Therefore, A∩(∩n i=1*Sλixi)
= ∅. By saturation,A ∩ (∩(λ,x)∈Λ*Sλixi)
= ∅; choosey ∈ A ∩ (∩(λ,x)∈Λ*Sλixi).
48 CHAPTER 3. REAL ANALYSIS
Since y ∈ A, y is nearstandard, so y ∈ μ(x) for somex ∈ X. By the deﬁnition of C, x ∈ C. Since∩λ∈ΛCλ = ∅, there exists λ ∈ Λ withx
∈ λ. Since * Tλx ⊃ μ(x), *Sλx ∩*Tλx =*( Sλx ∩Tλx)=∅, we gety
∈ μ(x), a contradiction. Therefore, C is compact.
3.4 Products Proposition 3.4.1 Let (Xλ,Tλ) be a family of topological spaces, and let (X,T) be the product topological space. Then *X = {y : y is an internal function from *Λ to∪λ∈*Λ *Xλ and ∀λ∈ *Λ yλ ∈ *Xλ}. (3.11) Given x∈ X, μ(x)={y ∈ *X :∀λ∈ Λ yλ xλ}. (3.12) Proof: The formal deﬁnition of the product is X = {f ∈F(Λ,∪λ∈ΛXλ):∀λ∈ Λ f(λ) ∈ Xλ} (3.13) where F(A,B) denotes the set of all functions from A to B. By the Transfer Principle, *X = {f ∈ *(F(Λ,∪λ∈ΛXλ)) :∀λ∈ *Λ f(λ) ∈ *Xλ}(3.14) = {y : *Λ→∪ λ∈*Λ*Xλ : y is internal,∀λ∈ *Λ yλ ∈ *Xλ}. (3.15) Suppose y ∈ μ(x) withx∈ X. Fixλ∈ Λ. Given T ∈T λ with xλ ∈ T, letS = {z ∈ X : zλ ∈ T}.S∈T and x ∈ S, so y ∈ *S. Therefore, yλ ∈ *T, soyλ xλ. Conversely, suppose y ∈ *X and yλ xλ for all λ ∈ Λ. Ifx ∈ T ∈T, then there exist λ1,...,λn ∈ Λ withn ∈ N and
3.5. CONTINUITY 49
Tλi ∈T i such that if S = {z ∈ X : zλi ∈ Tλi(1 ≤ i ≤ n)},then x∈ S ⊂ T. But * S = {z ∈ *X : zλi ∈ *Tλi(1≤i ≤ n}) by Transfer, so y ∈ *S ⊂*T. Therefore y x. Theorem 3.4.2 (Tychonoﬀ) Let (Xλ,Tλ) (λ ∈ Λ) be a family of topological spaces, and (X,T) the product topological space. If (Xλ,Tλ) is compact for each λ ∈ Λ, then(X,T ) is compact. Proof: Suppose y ∈ *X. For each λ ∈ Λ, there existsx λ ∈ Xλ such that yλ xλ. By the Axiom of Choice, this deﬁnes an element x∈ X such that yλ xλ for each λ∈ Λ. Therefore, y x by Proposition 3.4.1. Thus, every y ∈ *X is nearstandard, so X is compact by Theorem 3.3.2.
3.5 Continuity Proposition 3.5.1 Suppose (X,S) and (Y,T) are topological spaces and f : X →Y . Thenf is continuous if and only if *f(μ(x)) ⊂μ(f(x)) for every x∈ X. Proof: Suppose f is continuous. If y = f(x) andy ∈ T ∈T, then S = f−1(T)∈S. Hence, the sentence ∀z ∈ Sf (z) ∈ Tholds in X. By Transfer, the sentence ∀z ∈ *S *f(z) ∈ *Tholds in * X. Ifz ∈ μ(x), then z ∈ *S, so * f(z) ∈ *T. Since this holds for every T satisfying f(x) ∈ T ∈T,* f(z) ∈μ (f(x)). Thus, *f(μ(x)) ⊂μ(f(x)). Conversely, suppose *f(μ(x)) ⊂μ(f(x)) for every x ∈ X. Choose T such that f(x) ∈ T ∈T. By Proposition 3.1.5, we may ﬁnd S ∈ *S such that x ∈ S ⊂ μ(x). Accordingly, the sentence ∃S ∈ *S [x∈ S ∧ *f(S) ⊂ *T] (3.16) holds in *X; by Transfer, the sentence ∃S ∈S [x∈ S ∧ f(S) ⊂T] (3.17)
50 CHAPTER 3. REAL ANALYSIS
holds in X, sof is continuous. Corollary 3.5.2 If f : R→ R, thenf is continuous if and only if y x∈ R implies *f(y) f(x). Deﬁnition 3.5.3 Suppose (X,S) and (Y,T) are topological spaces with (Y,T) Hausdorﬀ, and suppose f :*X → *Y is internal. f is said to be S-continuous if f(x) is nearstandard and f(μ(x)) ⊂ μ(◦f(x)) for every x∈ X. Deﬁnition 3.5.4 A topological space (X,T) isregular if it is Hausdorﬀ and, given x ∈ X and C ⊂ X with x
∈ C and C closed, there exist S,T ∈Twith x ∈ S, C ⊂ T, and S ∩T = ∅. Proposition 3.5.5 Suppose (X,S) and (Y,T) are topological spaces with (Y,T) regular, and f : *X → *Y is Scontinuous. Deﬁne ◦f : X → Y by (◦f)(x)=◦(f(x)) for each x ∈ X. Then◦f is a continuous function. Proof: Because f(x) is nearstandard for each x ∈ X, there exists y ∈ Y such that f(x)∈ μ(y); since (Y,T) is Hausdorﬀ, this y is unique by Proposition 3.1.7. Thus, the formula for ◦f deﬁnes a function. Suppose x ∈ X, y = ◦f(x). If y ∈ V ∈T, thenX \ V is closed. Since (Y,T) is regular, we may ﬁnd S,T ∈ T with y ∈ S, X \ V ⊂ T, andS ∩ T = ∅. Since f is S-continuous, f−1(*S) ⊃ μ(x). f−1(*S) is internal by the Internal Deﬁnition Principle, so it contains *W for some W satisfying x ∈ W ∈S, by Proposition 3.1.5. If w ∈ W, then w ∈ *W, sof(w) ∈ *S. If ◦f(w)
∈ V , then◦f(w) ∈X \V ⊂ T. SinceT ∈T, f(w) ∈ *T by Proposition 3.2.1. But *S ∩*T = *(S ∩T)=∅, sof(w)
∈ *S, a contradiction which shows ◦f(w) ∈ V for w ∈ W. Thus,◦f is continuous.
3.5. CONTINUITY 51
Remark 3.5.6 Inthe proof of Proposition 3.5.5, one is tempted to consider the function g = *( ◦f) and apply Proposition 3.5.1. However, since ◦f is constructed using the nonstandard extension, using the properties of f propels us into a second nonstandard extension *(*X), creating more problems than we solve. Hence, the argument must proceed without invoking the nonstandard characterization of continuity presented in Proposition 3.5.1.
Deﬁnition 3.5.7 Suppose (X,T ) is a topological space and (Y,d) is a metric space. A function f : X → Y is bounded if supx,y∈X d(f(x),f(y)) <∞.( C(X,Y ), ¯ d) denotes the metric space of bounded continuous functions from X to Y , where ¯ d(f,g) = sup x∈X d(f(x),g(x)).
Theorem 3.5.8 (Nonstandard Ascoli’s Theorem) Let (X,T ) be a compact topological space and (Y,d) a metric space. If f is an S-continuous function from *X to *Y , then *¯ d(f,◦f) 0, i.e. f is nearstandard as an element of *(C(X,Y ), ¯ d), and◦f is its standard part.
Proof: By Proposition 3.5.5, ◦f is a continuous function from (X,T) to (Y,d). Let g = ◦f. Givenz ∈ *X, z ∈ μ(x) for some x ∈ X. Transferring the triangle inequality, *d(*g(z),f(z)) ≤ *d(*g(z),g(x)) + *d(g(x),f(x)) + *d(f(x),f(z)). (3.18)
The ﬁrst term is inﬁnitesimal by Propositions 3.5.1 and 3.5.5; the second by the deﬁnition of g = ◦f; and the third because f is S-continuous. Therefore *d(g(z),f(z)) 0 for every z ∈ *X. Therefore, *¯ d(f,g) 52 CHAPTER 3. REAL ANALYSIS
Corollary 3.5.9 (Ascoli) Suppose A ⊂ C([0,1],R) is closed, bounded and equicontinuous. Then A is compact.
Proof: Given ∈ R++, there exists δ ∈ R++ and M ∈ R such that the sentence
∀f ∈ A∀x,y ∈ [0,1] [[|f(x)| 3.6 Diﬀerentiation Deﬁnition 3.6.1 Suppose x,y ∈ *R. We write y = o(x) if there is some δ 0 such that |y|≤δ|x| and y = O(x) if there is some m ∈ N such that |y|≤M|x|. Proposition 3.6.2 Suppose f : Rm → Rn and x ∈ Rm. Then f is diﬀerentiable at x if and only if there exists a linear function J : Rm → Rn such that *f(y)=f(x)+ *J(y−x)+o(y−x) for all y x.
3.7. RIEMANN INTEGRATION 53
Proof: Let L be the set of all linear maps from Rm to Rn. Suppose f is diﬀerentiable at x. Then there exists J ∈ L such that for each ∈ R++, there exists δ ∈ R++ such that the sentence ∀y ∈ Rm[|y−x| <δ⇒| f(y)−f(x)−J(y−x)|≤|y−x|] (3.21) holds in X; by Transfer, the setence ∀y ∈ *Rm [|y−x|<δ⇒| f(y)−f(x)−*J(y−x)|≤|y−x|] (3.22) holds in *X. Therefore, if y x, then|f(y)−f(x)−*J(y− x)|≤|y − x|. Since is an arbitrary element of R++, |f(y)−f(x)−*J(y−x)| = o(|y−x|) for all y x. Conversely, suppose that there exists J ∈ L such thaty x implies |f(y)−f(x) −*J(y −x)| = o(|y −x|). Fix ∈ R++. Then the sentence ∃δ ∈ *R++ [|y−x|<δ ⇒| f(y)−f(x)−*J(y−x)|≤|y−x|] (3.23) holds in *X. By Transfer, the sentence ∃δ ∈ R++ [|y−x|<δ ⇒| f(y)−f(x)−J(y−x)|≤|y−x|] (3.24) holds in X. Since is an arbitrary element of R++, f is diﬀerentiable at x.
3.7 Riemann Integration Theorem 3.7.1 Suppose [a,b] ⊂ R and f :[a,b] → R is continuous. Given n ∈ *N\N,
b a f(t)dt = ◦ 1 n n k=1 *f(a + k(b−a) n )
. (3.25)
54 CHAPTER 3. REAL ANALYSIS
Proof: Let In = 1 nn k=1 f(a + k(b−a) n ) forn ∈ N. By Transfer, In = 1 nn k=1 *f(a+k(b−a) n ) forn ∈ *N. Sincef is continuous, In →b a f(t)dt. By Proposition 3.1.10, *In b t=a f(t)dt for all n ∈ *N\N.
3.8 Diﬀerential Equations
Nonstandard analysis permits the construction of solutions of ordinary diﬀerential equations by means of a hyperﬁnite polygonal approximation; the standard part of the polygonal approximation is a solution of the diﬀerential equation. Construction 3.8.1 Suppose F : Rk ×[0,1] → Rk is continuous, there exists M ∈ N such that |F(x,t)|≤M for all (x,t) ∈ Rk ×[0,1], and y0 ∈ Rk. Choosen ∈ *N\N. By the Transfer Principle, we can deﬁne inductively z0 n = y0z k+1 n = zk n+ 1 n*Fzk n, k n (3.26) and then extendz to a function with domain *[0,1] by linear interpolation z(t) = ([nt]+1−nt)z [nt] n
+(nt−[nt])z [nt]+1 n
(3.27) where [nt] denotes the greatest (nonstandard) integer less than or equal to nt. Let y(t)=◦(z(t)) for z ∈ [0,1]. (3.28) Theorem 3.8.2 With the notation in Construction 3.8.1, z is S-continuous and y is a solution of the ordinary diﬀerential equation y(0) = yo y (t)=F(y(t),t). (3.29)
3.8. DIFFERENTIAL EQUATIONS 55
Proof: Given r,s ∈ *[0,1] with r s, |z(r) − z(s)|≤ M|r −s| 0, so z is S-continuous. By Theorem 3.5.8, y is continuous and there exists δ 0 such that |z(t)−*y(t)|≤δ for all t∈ *[0,1]. Then y(t)−y0 z(t)−z(0) [nt]−1 k=0 zk+1 n −zk n =[nt]−1 k=0 1 n*Fzk n, k n [nt]−1 k=0 1 n*F*yk n, k n t 0 F(y(s),s)ds (3.30) by Theorem 3.7.1. Sincey(t)−y0 andt 0 F(y(s),s)ds are both standard, they are equal. By the Fundamental Theorem of Calculus, y (t)=F(y(t),t) for all t∈ [0,1], so y is a solution of the ordinary diﬀerential equation 3.29.
56 CHAPTER 3. REAL ANALYSIS
Chapter 4
Loeb Measure
The Loeb measure was developed originally by Peter Loeb (1975) to solve a problem in potential theory. Loeb’s construction allows one to convert nonstandard summations on hyperﬁnite spaces to measures in the usual standard sense. It has been used very widely in probability theory, and is an important tool in nonstandard mathematical economics.
4.1 Existence of Loeb Measure
Deﬁnition 4.1.1 An internal probability space is a triple (A,A,ν) where 1. A is an internal set, 2. A⊂(*P)(A) is an internalσ-algebra, i.e. (a) A∈A ; (b) B ∈Aimplies A\B ∈A; and (c) If {Bn : n ∈ *N} is an internal sequence with Bn ∈A, then∩n∈*NBn ∈Aand ∪n∈*NBn ∈A ; and
57
58 CHAPTER 4. LOEB MEASURE
3. ν : A→*[0,1] is an internal *-countably additive probability measure, i.e.
(a) ν(A) = 1; and (b) if{Bn : n ∈ *N}is an internal sequence and Bn∩ Bm = ∅ whenever n
= m, thenν(∪n∈*NBn)= n∈*N ν(Bn).Remark 4.1.2 The Loeb measure construction also works if we merely assume that A is closed under ﬁnite unions and ν is ﬁnitely additive. We shall be primarily interested in hyperﬁnite spaces, in which integration is just summation.
Deﬁnition 4.1.3 An internal probability space is hyperﬁnite if A is a hyperﬁnite set,A = (*P)(A) (i.e. Ais the class of all internal subsets of A), and there is an internal set of probability weights {λa : a ∈ A} such that ν(B)=a∈B λa for all B ∈A . Example 4.1.4 The canonical exampleof a hyperﬁnite probability space is (A,A,ν), where A = {1,...,n} for some n ∈ *N\N, A = (*P)(A), and ν(B)=|B| n for all B ∈A . Construction 4.1.5 (Loeb Measure) Suppose (A,A,ν) is an internal probability space. The Loeb measure construction creates an (external) probability measure in the standard sense, deﬁned on an (external) σ-algebra of (internal and external) subsets of the internal set A. Deﬁne ¯ A = {B ⊂A :∀∈ R++ ∃C ∈A∃D ∈A [C ⊂B ⊂D, ν(D\C) < ]} (4.1) and ¯ ν(B) = inf {◦ν(D):B ⊂D ∈A} = sup{◦ν(C):C ⊂B, C ∈A} (4.2) for B ∈ ¯ A.¯ ν is called the Loeb measure generated by ν.
4.1. EXISTENCE OF LOEB MEASURE 59
Theorem 4.1.6 (Loeb) Suppose (A,A,ν) is an internal probability space. Then 1. ¯ A is a σ-algebra, ¯A⊃A ; 2. ¯ ν is a countably additive probability measure; 3. ¯ A is complete with respect to ¯ ν; 4. ¯ ν(B)=◦ν(B) for every B ∈A; and 5. for each B ∈ ¯ A, there exists A ∈Asuch that ¯ ν((A\B)∪(B \A)) = 0. (4.3) Proof:
1. First, we note that part 4 is obvious. 2. Second, we prove part 5. If B ∈ ¯ A, then for each n ∈ N, we may ﬁnd Cn,Dn ∈Awith C1 ⊂ C2 ⊂ ...⊂ B...D 2 ⊂ D1 and ν(Dn \ Cn) < 1/n. By Theorem 1.10.3, we can extend Cn andDn to internal sequences in A. {n ∈ *N :[ Cm ⊂ Cm+1 ⊂ Dm+1 ⊂ Dm ∧ν (Dm \Cm) < 1/m] (1≤m ≤n)} is internal and contains N, so it contains some n ∈ *N\N by Proposition 3.1.6. Cn ∈A. Ifm ∈ N,¯ ν((Cn \B)∪(B \Cn)) ≤¯ ν((Cn\Cm)∪(Dm\Cn)) ≤ ¯ ν(Dm\Cm) < 1/m. Sincemis an arbitrary element of N,¯ ν((Cn\B)∪(B\Cn)) = 0. 3. Next, we establish parts 1 and 2. (a) Suppose B,B ∈ ¯ A. Fix ∈ R++, and ﬁnd C ⊂ B ⊂ D, C ⊂ B ⊂ D with C,C ,D,D ∈Aandν (D\C) < / 2,ν(D \C ) < /2. C\D ⊂ B\B ⊂D \C and ((D\C )\(C\D )) ⊂((D\C)∪(D \C )), so ν((D \C )\(C \D )) < /2+/2=. Thus, B\B ∈ ¯ A.
60 CHAPTER 4. LOEB MEASURE
(b) Nowsuppose B1,B2,...∈ ¯ Aand let B = ∪n∈NBn.By considering B n = Bn \∪ n−1 i=1 Bi, we may assume without loss of generality that the Bi’s are disjoint. Given ∈ R++, we may ﬁnd Cn ⊂Bn ⊂ Dn with Cn,Dn ∈Asatisfying ν(Dn \ Cn) < /2n+1. ExtendCn and Dn to internal sequences with Cn,Dn ∈Aby Theorem 1.10.3. If we let αn = ◦(ν(∪n i=1Ci)), then αn is a nondecreasing sequence in [0,1], so it converges to some α ∈ R. By Proposition 3.1.9, we may ﬁnd some n0 ∈ *N\N such that ν(∪n i=1Ci) α for n ∈ *N\N, n ≤n0.Moreover, {n : ν(∪n i=1Di) ≤ν(∪n i=1Ci)+/2}is aninternal set which contains N, so by Proposition 3.1.6, it contains all n≤ n1 for some n1 ∈ *N\N. Taking n = min {n0,n1}, we see there exists n ∈* N\N such that ν(∪n i=1Ci) α and ν(∪n i=1Di) ≤ν (∪n i=1Ci)+/2. Moreover, we can ﬁnd m ∈ N such that αm >α− /2. Let C = ∪m i=1Ci and D = ∪n i=1Di. ThenC,D ∈A , C ⊂ B ⊂ D,and ◦(ν(D \ C)) <, soν(D \ C) <. There-fore, B ∈A . α − /2 < ◦(ν(C)) ≤ ¯ ν(B) ≤◦ (ν(D)) ≤ α+/2. |¯ ν(Bn)−(αn−αn−1)| < / 2n+1, so |n∈N ¯ ν(Bn)−α| < /2. Since is arbitrary, ¯ ν(B)=n∈N ¯ ν(Bn), so ¯ ν is countably additive. 4. Finally, we establish part 3. Suppose B ⊂ B with B ∈ ¯ A and ¯ ν(B ) = 0. Then given ∈ R++, thereexists D ∈Asuch that B ⊂ D and ◦ν(D) <. Then ∅⊂B ⊂ D, soB ∈ ¯ A, so ¯ A is complete with respectto ¯ ν.
4.2. LEBESGUE MEASURE 61
4.2 Lebesgue Measure
In this section, we present a construction of Lebesgue measure in terms of the Loeb measure on a natural hyperﬁnite probability space. Construction 4.2.1 (Anderson) Fix n ∈ *N\N, and let A = {k n : k ∈ *N, 1 ≤ k ≤ n}. LetA = (* P)(A), the set of all internal subsets of A, andν(B)=|B| n for B ∈A . Let (A, ¯ A, ¯ ν) be the Loeb probability space generated by (A,A,ν). Let ◦ denote the restriction of the standard part map to A, i.e. ◦(a)=◦a for a ∈ A, and let st −1 denote the inverse image of st, i.e. st−1(C)={a ∈ A : ◦a ∈ C} for C ⊂ [0,1]. Let C = {C ⊂ [0,1] : st−1(C) ∈ ¯ A} andμ (C)=¯ ν(st−1(C)) for all C ∈C . Theorem 4.2.2 (Anderson, Henson) ([0,1],C,μ) is the Lebesgue measure space on [0,1]. Proof: Since ¯ A is a σ-algebra, so is C; since ¯ ν is countably additive, so is μ. Consider the closed interval [a,b] ⊂ [0,1]. Then st−1([a,b]) =∩m∈N*(a− 1 m,b+ 1 m)∩A. *( a− 1 m,b+ 1 m)∩A ∈Aby the Internal Deﬁnition Principle, so st−1([a,b])∈ ¯ A. μ([a,b]) = ¯ ν(st−1([a,b])) = limm→∞◦|*(a− 1 m,b+ 1 m)∩A| n = limm→∞ b−a + 2 m = b−a. Thus,μ([a,b]) = μ([a,b]).Let B be the class of Borel sets. Since B is the smallestσ -algebra containing the closed intervals, C⊃B. Since μ and Lesbesgue measure are countably additive, and agree on closed intervals, they agree on B. C is complete with respect to μ because ¯ A is complete with respect to ¯ ν. Therefore, C contains the class of Lebesgue measurable sets and μ agrees with Lebesgue measure on that class. Finally, we show thatC iscontained inthe class of Lebesgue measurable sets. 1 Suppose C ∈C. Given ∈ R++, there ex1The proof of this part given here is due to Edward Fisher.
62 CHAPTER 4. LOEB MEASURE
ist B,D∈Awith B ⊂st−1(C)⊂ D and ◦ν(D)−◦ν(B) < . Let ˆ B = {◦b : b ∈ B}, ˆ D = [0 ,1] \{ ◦a : a ∈ A − D}. Then ˆ B ⊂ C ⊂ ˆ D. ˆ B and [0,1]\ ˆ D are closed by Proposition 3.2.3, so ˆ D is open; thus, ˆ B, ˆ D ∈B . μ( ˆ D)− μ( ˆ B)= ¯ ν(st−1( ˆ D))−¯ ν(st−1( ˆ B)) < ¯ ν(D)−¯ ν(B) (since st −1( ˆ B)⊃ Band st −1( ˆ D) ⊂ D)=◦ν(D)−◦ν(B) <. Since the Lebesgue measure space is complete, C is Lebesgue measurable.
4.3 Representation of Radon Measures
Deﬁnition 4.3.1 A Radon probability space is a probability space (X,B,μ) where (X,T) is a Hausdorﬀ space, B is the σ-algebra of Borel sets (i.e. the smallest σ-algebra containing T), and μ(B) = sup {μ(C):C ⊂ B, C compact} = inf{μ(T):B ⊂T, T ∈T} (4.4) for every B ∈B . Example 4.3.2 Let (X,d) be any complete separable metric space,Bthe σ-algebra of Borel sets. Then any probability measure μ on B is Radon (see Billingsley (1968)). Theorem 4.3.3 (Anderson) Let (X,B,μ) be a Radon probability space, and ¯ B is the completion of B with respectto μ. Then there is a hyperﬁnite probability space (A,A,ν) and a function S : A → X such that B ∈ ¯ B if and only ifS −1(B)∈ ¯ A. For every B ∈ ¯ B, μ(B)=¯ ν(S−1(B)). Proof: The proof is similar to the proof of Theorem 4.2.2; for details, see Anderson (1982).
4.4. LIFTING THEOREMS 63
4.4 Lifting Theorems
In this section, we state a number of “Lifting Theorems” relating integration theory in Loeb probability spaces to the internal integration theory in internal measure spaces. Note that, in the case of a hyperﬁnite measure space, the internal integration theory reduces to ﬁnite summations. In particular, feasibility conditions in hyperﬁnite exchange economies can be formulated as conditions on summations, in exactly the same way as they are formulatedin ﬁnite exchange economies. However, the Loeb measure construction can be used to convert a hyperﬁnite exchange economy to an economy with a measure space of agents in the sense introduced by Aumann. The lifting theorems provide the link between the two constructions. Deﬁnition 4.4.1 Suppose (A,A,ν) is an internal probability space and (X,T) is a topological space. A function f : A → *X is said to be ν-measurable if it is internal and f−1(T)∈Afor every T ∈ *T. Theorem 4.4.2 (Loeb, Moore, Anderson) Let (A,A,ν) be an internal probability space. 1. If (X,T) is a regular topological space, f : A → *X is ν−measurable, andf(a) is nearstandard for ¯ ν-almost all A ∈A, then◦f : A →X is ¯ ν-measurable. 2. If (X,T ) is a Hausdorﬀ topological space with a countable base of open sets and F : A→ X is ¯ ν-measurable, then there exists a ν-measurable function f : A → *X such that ◦f(a)=F(a)¯ v-almost surely.
Proof: This will be provided in the monograph. For now, see Anderson (1982).
64 CHAPTER 4. LOEB MEASURE
Deﬁnition 4.4.3 Suppose (A,A,ν) is an internal probability space and f : A → *R is ν-measurable. Suppose A ∈ *Xn. LetI be the function which assigns to every pair ((B,B,μ),g), where (B,B,μ) is a standard probability space, B ∈X n, andg is a μ-measurable real-valued function, the integral I((B,B,μ),g)=B gdμ. Theinternal integral of f, denotedA fdν, is deﬁned byA fdν = (* I)((A,A,ν),f). Example 4.4.4 Let (A,A,ν) be as in the construction of Lebesgue measure (Construction 4.2.1). If f : A → *R is any internal function, then f is ν-measurable by the Internal Deﬁnition Principle. Moreover,
A fdν = 1 n a∈A f(a) (4.5) by the Transfer Principle.
Deﬁnition 4.4.5 Suppose (A,A,ν) is an internal probability space and f : A →*R. We sayf is S-integrable if 1. f is a ν-measurable function; 2. ◦A|f|dν < ∞; 3. B ∈A,ν (B) 0⇒B |f|dν 0. Theorem 4.4.6 (Loeb, Moore, Anderson) Let (A,A,ν) be an internal probability space. 1. If f : A→ *R is ν-measurable andA|f(a)|dν is ﬁnite, then ◦f is integrable with respect to ¯ ν and A|◦f|d¯ ν ≤ ◦A|f|dν; 2. If f : A → *R is S-integrable, then ◦f is integrable with respect to ¯ ν; moreover, ◦A fdν = A ◦fd¯ ν;
4.5. WEAK CONVERGENCE 65
3. If F : A → R is integrable with respect to ¯ ν, then there exists an S-integrable function f : A → *R such that ◦f = F ¯ ν-almost surely. Moreover, ◦A fdν = A ◦fd¯ ν. Proof: Parts 2 and 3 are proved in Anderson(1976). To see part 1, suppose m ∈ N, and deﬁne fm(a) = min{|f(a)|,m}. fm is obviously S-integrable, so by (2), ◦fm is integrable and ◦A fmdν =A ◦fmd¯ ν. By the deﬁnition of the standard integral, A ◦|f|d¯ ν = limm→∞A ◦fmd¯ ν = limm→∞◦A fmdν; note that this last sequence is increasing and bounded above by ◦A|f|dν < ∞. Therefore ◦f is integrable andA|◦f|d¯ ν ≤ ◦A|f|dν. Deﬁnition 4.4.7 Suppose An is a sequence of ﬁnite sets. A sequence of functions fn : An → Rk + is said to be uniformly integrable if, for every sequence of sets En ⊂ An satisfying |En|/|An|→0, 1 |An| a∈En fn(a)→ 0. (4.6) Proposition 4.4.8 Suppose {An : n ∈ N} is a sequence of ﬁnite sets and fn : An → Rk +. Then {fn : n ∈ N} is uniformly integrable if and only if for all n ∈ *N, fn is Sintegrable with respect to the normalized counting measure νn(B)=|B|/|An| for each internal B ⊂An.
Proof: See Anderson (1982), Theorem 6.5.
4.5 Weak Convergence
Deﬁnition 4.5.1 A sequence of probability measures μn on a complete separable metric space (X,d) is said toconverge
66 CHAPTER 4. LOEB MEASURE
weakly to a probability measure μ (written μn ⇒ μ) if
X Fdμn →
X Fdμ (4.7) for every bounded continuous function F : X → R. The standard theory of weak convergence of probability measures is developed in Billingsley (1968). Because the theory is widely used in the large economies literature (see for example Hildenbrand (1974,1982) and Mas-Colell (1985)), it is useful to have the following nonstandard characterization in terms of the Loeb measure. Theorem 4.5.2 (Anderson, Rashid) 2 Suppose νn (n ∈ N) is a sequence of Borel probability measures on a complete separable metric space (X,d). Letμn(B)=*νn(st−1(B)) for each Borel set B ⊂X deﬁne a Borel measure on X for each n ∈ *N. Thenνn converges weakly if and only if 1. *νn(ns(*X)) = 1 for all n ∈ *N\N; and 2. μn = μm for all n,m∈ *N\N. In this case, the weak limit is the common value μn for n ∈ *N\N. Proof: 1. Suppose νn ⇒ ν. (a) Since ν is a probability measure and X is complete separable metric, {νn : n ∈ N} is tight, i.e. given ∈ R+, there exists K compact such 2The result holds for spaces much more general than the complete separable metric spaces considered here. See Anderson and Rashid (1978) for details.
4.5. WEAK CONVERGENCE 67
that νn(K) > 1 − for all n ∈ N (Billingsley (1968)). By transfer, *νn(*K) > 1 − for all n ∈ *N\N.* K ⊂ ns(*X) by Theorem 3.3.2, so* νn(ns(*X)) = 1. (b) Given F : X → R, F bounded and continuous, X Fdνn → X Fdν by assumption. Therefore, X *Fdνn X Fdν for all n ∈ *N\N by Proposition 3.1.10. Then
X Fdμn =
ns(*X) F(◦x)dνn (4.8) =
ns(*X) ◦(*F(x))dνn (4.9) by Proposition 3.5.1 =
*X ◦(*F(x))dνn
*X *F(x)dνn (4.10) by Theorem 4.4.6
X Fdν. (4.11) Therefore,X Fdμn =X Fdν for all n ∈ *N\N. Since this holds for every bounded continuous F, μn = ν by Billingsley (1968). In particular, if m,n∈ *N\N, thenμn = μm. 2. Suppose νn(ns(*X)) = 1 for all n ∈ *N\N and μn = μm for all n,m∈ *N\N. Letν be the common value of μn for n ∈ *N\N.
*X *Fdνn
*X ◦(*F(x))dνn (4.12) by Theorem 4.4.6 =ns(*X) ◦(*F(x))dνn =ns(*X) F(◦x))dνn =X Fdμn =X Fdν. (4.13)
68 CHAPTER 4. LOEB MEASURE Thus, for n ∈ *N \ N, *X *Fdνn X Fdν. Then X Fdνn →X Fdν as n →∞by Proposition 3.1.10. Therefore νn ⇒ν.
Chapter 5
Large Economies
The subject of large economies was introduced brieﬂy in Section 1.1.1. We are interested in studying properties of sequences of exchange economies χn : An → P ×Rk +, where An is a set of agents, Rk + the commodity space, and P the space of preferencerelations on Rk +. In this section, we examine price decentralization issues for the core, the bargaining set, the value, and the set of Pareto optima, as well as the question of existence of approximate Walrasian equilibria. We begin by studying the properties of hyperﬁnite exchange economies. The Transfer Principle then gives a very simple derivation of analogous properties for sequences of ﬁnite economies. Much of the work on large economies using nonstandard analysis concerns the core. For this reason, we have chosen to devote considerable attention to the core, in order to illustrate the use of the nonstandard methodology, and to contrast that methodology to measure-theoretic methods. In Section 5.5, we focus on the following issues:
1. The properties of the cores of hyperﬁnite exchange economies, in Theorem 5.5.2.
69
70 CHAPTER 5. LARGE ECONOMIES
2. The use of the Transfer Principle to give very simple derivations of asymptotic results about the cores of large ﬁnite economies from results about hyperﬁnite economies, in Theorem 5.5.10.
3. The close relationship between hyperﬁnite economies and Aumann continuum economies, which are linked using the Loeb measure construction, in Proposition 5.5.6.
4. The ability of the nonstandard methodology to capture behavior of large ﬁnite economies whichis not captured in Aumann continuum economies, in Remark 5.5.3 and Examples 5.5.5, 5.5.7, 5.5.8 and 5.5.9.
5.1 Preferences Given x,y ∈ Rk +, xi denotes the ith component of x; x>y means xi ≥ yi for all i and x
= y; x >> ymeans xi >y i for all i. If 1≤ p ≤∞ , ||x||p =( k i=1|xi|p)1/p; ||x||∞ = max{|xi|: i =1,...,k}. Deﬁnition 5.1.1 A preference is a binary relation on Rk +. Let P denote the set of preferences. A preference 1. is transitive if ∀x,y,z ∈ Rk +[xy,y z ⇒ x z]; 2. is continuous if{(x,y)∈ Rk +×Rk + : x y}is relativelyopen in Rk + ×Rk +; 3. is (a) monotonic if ∀x,y ∈ Rk +[x >> y⇒ xy]; (b) strongly monotonic if∀x,y ∈ Rk +[x>y⇒ xy];
5.1. PREFERENCES 71
4. is irreﬂexive if ∀x∈ Rk +[x
x]; 5. is (a) convex if ∀x∈ Rk +, {y : y x} is convex; (b) strongly convex if ∀x,y ∈ Rk +,x
= y ⇒ [x + y 2 x∨ x + y 2 y]; (5.1)
6. satisﬁes free disposal if ∀x,y,z ∈ Rk +,[[x>y ]∧[y z]⇒ x z]. (5.2) Let Pc denote the space of continuous preferences.
Deﬁnition 5.1.2 We deﬁne a metric on Pc as follows: Let d1 be the one-point compactiﬁcation metric on R2k + ∪{∞} . Given any compact metric space (X,d), the Hausdorﬀ metric dH is deﬁned on the space of closed sets of X by dH(B,C) = inf{δ :[∀x∈ B ∃y ∈ Cd (x,y) <δ ] ∧[∀y ∈ C ∃x∈ Bd (x,y) <δ ]}. (5.3) Let d2 be the Hausdorﬀ metric (d1)H. Given∈ Pc, deﬁne C = {(x,y)∈ R2k + : x
y}∪{∞}. Then deﬁne d(, )=d2(C,C). (5.4) Proposition 5.1.3 (Brown, Robinson, Rashid) If ∈ Pc, μ()={ ∈ *Pc :∀x,y ∈ Rk + [x y ⇔ μ(x) μ(y)]}(5.5) where μ(x),μ(y) are taken with respect to the Euclidean metric on Rk +. (Pc,d) is compact.
72 CHAPTER 5. LARGE ECONOMIES
Proof: 1. Recall that the one-point compactiﬁcation metric induces the usual Euclidean topology on Rk +, so that if x ∈ Rk +, thed1-monad of x, μd1(x) coincides with the Euclidean monad μ(x). Suppose ∈Pc. We will show that equation 5.5 holds. (a) Suppose ∈ *Pc, ∈ μ(). Fix x,y ∈ Rk +. We show x y if and only if μ(x) μ(y). i. Suppose x y. If there existu ∈ μ(x), v ∈ μ(y) withu
v, then (u,v)∈ *C, so there exist (w,z) ∈ *C such that *d1((u,v),(w,z)) ≤ *d2( ,) 0. *d1((w,z),(x,y)) ≤ *d1((w,z),(u,v)) + *d1((u,v),(x,y)) 0, so w ∈ μ(x), z ∈ μ(y). Since (w,z) ∈ *C, w*
z. Sincex y and ∈ Pc, μ(x)*μ(y), so w*z, a contradiction which shows μ(x) μ(y). ii. If x
y, then (x,y) ∈ C, so there exists (u,v)∈ *C such that *d1((u,v),(x,y)) 0. Therefore u ∈ μ(x) andv ∈ μ(y), so μ(x)
μ(y). (b) Conversely, suppose for every x,y ∈ Rk +, x y ⇔μ (x) μ(y). We will show that every w ∈ *C is inﬁnitely close to some z ∈ *C, and vice versa. i. Suppose w ∈ *C. We will show there exists z ∈ *Cwith *d1(w,z) 0. We consider two cases: A. Suppose w ∈ μd1(∞). In this case ∞∈ *Cand *d1(w,∞) 0. B. w =(u,v) ∈ μd1(x,y) for somex,y ∈ Rk +. In this case,u
v, soμ(x)
μ(y),so x
y, so (x,y)∈ C.
5.1. PREFERENCES 73
Accordingly, for every w in *C, there exists z ∈ *Csuch that *d1(w,z) 0. ii. Suppose w ∈ *C. We will show there exists z ∈ *C with *d1(w,z) 0. Again there are two cases. A. The case w ∈ μd1(∞) is handled as in item 1(b)iA above. B. Suppose w =(u,v) ∈ μd1(x,y) for some x,y ∈ Rk +. In this case,u*
v, soμ(x)*
μ (y), so x
y (since is continuous), so( x,y)∈ C. Therefore,
{n ∈ *N :[∀x∈ B ∃y ∈ Cd (x,y) < 1/n] ∧ [∀y ∈ C ∃x∈ Bd (x,y) < 1/n]} (5.6) contains N. The set is internal by the Internal Deﬁnition Principle. Hence, it includes some inﬁnite n by Proposition 3.1.6, so *d(, )= *d2(*C,*C) 0. Therefore, ∈ μ(). We have thus veriﬁed equation 5.5. 2. It remains to show that (Pc,d) is compact. Given ∈ *Pc, deﬁne by x y ⇔ μ(x) μ(y). If x y, then μ(x) μ(y). Let B = {(u,v):u v}. B is internal and contains μ(x,y), so it contains *T for some open set T with (x,y)∈ T. If (w,z) ∈ T, thenμ(w,z) ⊂*T, so μ(w) μ(z), so w z. Thus,∈Pc. By equation 5.5, ∈ μ(). By Theorem 3.3.2, (Pc,d) is compact.
74 CHAPTER 5. LARGE ECONOMIES
5.2 Hyperﬁnite Economies
Deﬁnition 5.2.1 A hyperﬁnite exchange economy is an internal function χ : A→ *(P ×Rk +), where A is a hyperﬁnite set. We deﬁne the endowment e(a) and preference a of a by (a,e(a)) = χ(a).
5.3 Loeb Measure Economies Deﬁnition 5.3.1 Let (A,B,μ) be a standard probability space. An Aumann continuum economy is a function χ : A → Pc ×Rk + such that 1. χ is measurable;
2. e(a) is integrable. Construction 5.3.2 Suppose χ : A → *(Pc ×Rk +) is a hyperﬁnite exchange economy. Let A denote the set of all internal subsets of A, andν(B)=|B| |A| for B ∈A. Let (A, ¯ A, ¯ ν) be the Loeb measure space generated by (A,A,ν). Deﬁne ◦χ : A→ Pc ×Rk + by ◦χ(a)=( ◦a,◦e(a)). Theorem 5.3.3 (Rashid) If χ : A → *(Pc ×Rk +) is a hyperﬁnite exchange economy with n = |A|inﬁnite and 1 na∈A e(a)is ﬁnite, then ◦χ as deﬁned in Construction 5.3.2 is an Aumann continuum economy. A ◦e(a)d¯ ν ≤ ◦(1 na∈A e(a)),with equality if e is S-integrable. Proof: Since Pc is compact by Proposition 5.1.3, a is nearstandard for all a ∈ A. ◦ν({a : ||e(a)||∞ ≥ M}) ≤ ◦(a∈A||e(a)||∞ Mn ) ≤ ◦(k||a∈A e(a)||∞ Mn ), so ¯ ν({a : e(a) is ﬁnite }) = 1. Thus,◦χ(a) is deﬁned for ¯ ν-almost all a ∈ A; it ismeasurable by Theorem 4.4.2. A ◦e(a)d¯ ν ≤ ◦(1 na∈A e(a))(with equality in case e is S-integrable) by Theorem 4.4.6.
5.4. BUDGET, SUPPORT AND DEMAND GAPS 75
5.4 Budget, Support and Demand Gaps Deﬁnition 5.4.1 Let Δ = {p ∈ Rk : ||p||1 =1 },Δ + = Δ∩Rk +, and Δ++ =Δ∩Rk++. Deﬁnition 5.4.2 Deﬁne D,Q :Δ×P ×Rk + →P (Rk +) by D(p,,e)={x∈ Rk + : p•x ≤p•e,y x ⇒p•y>p•e}, (5.7) Q(p,,e)={x∈ Rk + : p•x ≤p•e,y x ⇒p•y ≥ p•e}. (5.8) D and Q are called the demand set and the quasidemand set respectively.
Deﬁnition 5.4.3 Deﬁne φB : Rk + × Δ× Rk + → R+, φS :R k + ×Δ×P → R+, andφ : Rk + ×Δ×P ×Rk + → R+ by φB(x,p,e)=|p•(x−e)|; (5.9) φS(x,p,) = sup {p•(x−y):y x};and (5.10) φ(x,p,,e)=φB(x,p,e)+φS(x,p,). (5.11) φB, φS and φ are referred to as the budget gap, thesupport gap, and thedemand gap respectively. Proposition 5.4.4 Suppose x,e∈ *Rk + are ﬁnite, ∈*Pc,and p∈ *Δ. 1. If *φS(x,p,) 0, then◦x ∈ Q(◦p,◦,◦x). If in addition ◦p∈ Δ++ and 0
0, then◦x∈ D(◦p,◦,◦x). 2. If *φ(x,p,,e) 0, then◦x ∈ Q(◦p,◦,◦e). If in addition ◦p∈ Δ++, then◦x ∈ D(◦p,◦,◦e). Proof:
76 CHAPTER 5. LARGE ECONOMIES
1. Suppose the hypotheses of (1) are satisﬁed. If y ∈ Rk + and y◦◦x, theny μ(◦x) by Proposition 5.1.3, so y x. Therefore, ◦p•y p•y ≥ p•x−φS(p,x,) p•x ◦p•◦x, so◦p•y ≥◦p•◦x, hence ◦x∈ Q(◦p,◦,◦x). If ◦p∈ Δ++, we show that ◦x ∈ D(◦p,◦,◦x) by considering two cases: (a) If ◦p•◦x = 0, then◦x = 0. Since 0
0, 0◦
0, so D(◦p,◦,◦x)={0}. Therefore ◦x∈ D(◦p,◦,◦x). (b) If ◦p•◦x>0, suppose y ∈ Rk +, y◦◦x and ◦p •y = ◦p•◦x. Since ◦ is continuous, we may ﬁndw ∈ Rk + with ◦p•◦w<◦p•y = ◦p•◦x with w◦◦x. By Proposition 5.1.3, w x, soφS(x,p,)
0, a contradiction. Hence ◦x∈ D(◦p,◦,◦x). 2. If the hypotheses of (2) are satisﬁed, then (1) holds and in addition ◦p•◦x = ◦(p•x)=◦(p•e)=◦p•◦e, so the conclusions of (2) follow from those of (1).
5.5 Core Deﬁnition 5.5.1 Suppose χ : A → P ×Rk + is a ﬁnite exchange economy or an Aumann continuum economy. The Core, the set ofWalrasian allocations, and the set ofquasiWalrasian allocations, ofχ, denoted C(χ), W(χ) andQ(χ) respectively, are as deﬁned in Chapter 18 of this Handbook Hildenbrand (1982). In case χ is a ﬁnite exchange economy, C(χ),W(χ), andQ(χ) are deﬁned by the following sentences: C(χ)={f ∈F(A,Rk +): a∈A f(a)= a∈A e(a) ∧∀S ∈P(A)∀g ∈F(S,Rk +)[ a∈S g(a)= a∈S e(a)
5.5. CORE 77
⇒ [S = ∅∨∃a∈ Sg(a)
a f(a)]]} (5.12)
W(χ)={f ∈F(A,Rk +): a∈A
f(a)= a∈A
e(a)
∧∃ p∈ Δ∀a ∈ Af(a) ∈ D(p,a,e(a))} (5.13)
and
Q(χ)={f ∈F(A,Rk +): a∈A
f(a)= a∈A
e(a)
∧∃ p∈ Δ∀a∈ Af (a) ∈ Q(p,a,e(a))}. (5.14) Given δ ∈ R++, deﬁne Wδ(χ)={f ∈F(A,Rk +): 1 |A|| a∈A f(a)−e(a)|<δ ∧∃ p∈ Δ ∀a∈ Af(a) ∈ D(p,a,e(a))}. (5.15) Because C, Q, andW are deﬁned by sentences, if χ is a hyperﬁnite exchange economy, we can form *C(χ), *W(χ), and *Q(χ); each is internal by the Internal Deﬁnition Principle. Deﬁne W 0(χ)=∩δ∈R++*Wδ(χ). (5.16) Theorem 5.5.2 (Brown, Robinson, Khan, Rashid, Anderson) Let χ : A → *(Pc ×Rk +) be a hyperﬁnite exchange economy.
1. If (a) n ∈ *N\N; (b) for each a ∈ A, a i. is *-monotonic; ii. satisﬁes *-free disposal;
78 CHAPTER 5. LARGE ECONOMIES
(c) ◦(1 na∈A e(a))∈ Rk++.(d) e(a)/n 0 for all a ∈ A. then for every f ∈ *C(χ), there exists p ∈ *Δ+ such that ◦f(a) ∈ Q(◦p,◦a,◦e(a)) for ¯ ν-almost all a ∈ A.If ◦p ∈ Δ++ and for each a ∈ A, 0
0, then◦f(a) ∈D (◦p,◦a,◦e(a)) for ¯ ν-almost all a ∈ A. 2. If the assumptions in 1 hold and in addition for each commodity i, ¯ ν({a ∈ A : ◦a is strongly monotonic,◦ e(a)i > 0}) > 0, then◦p ∈ Δ++ and hence ◦f(a) ∈ D(◦p,◦a,◦e(a)) for ¯ ν-almost all a. 3. If the assumptions in (1) and (2) hold and in addition e is S-integrable, then f is S-integrable and (◦p,◦f) ∈ W(◦χ). 4. If the assumptions in (1) hold and in addition (a) ◦ a is strongly convex for ¯ ν-almost all a∈ A; (b) for each commodity i, ¯ ν({a : ◦e(a)i > 0}) > 0; (c) a is *-irreﬂexive, *-convex, and *-strongly convex for all a∈ A; then f(a) *D(p,a,e(a)) for ¯ ν-almost all a∈ A. 5. If the assumptions in (1) and (4) hold and, in addition, e is S-integrable, then there exists g ∈W 0(χ) such that 1 n a∈A |f(a)−g(a)| 0. (5.17)
Proof:
5.5. CORE 79
1. Suppose χ satisﬁes the assumptions in part (1) of the Theorem. By Anderson (1978) (see also Dierker(1975)) and the Transfer Principle, there exists p ∈ *Δ+ such that 1 n a∈A *φ(f(a),p,e(a))≤ 6k maxa∈A||e(a)||∞ n 0 (5.18) since maxa∈A|e(a)|/n 0. 1 na∈A f(a)=1 na∈A e(a)is ﬁnite, so f(a) ande(a) are ﬁnite for ¯ ν-almost all a ∈ A. ◦f(a) ∈ Q(◦p,◦a,e(a)) by Proposition 5.4.4. If ◦p ∈ Δ++, then◦f(a) ∈ D(◦p,◦a,e(a)) by Proposition 5.4.4.
2. Suppose in addition that for each commodity i,¯ ν({a∈A : ◦a is strongly monotonic, ◦e(a)i > 0}) > 0. We will show that ◦p ∈ Δ++ by deriving a contradiction. If ◦p
∈ Δ++, we may assume without loss of generality that ◦p1 =0 ,◦p2 > 0. By assumption (1)(c), |e(a)| isﬁnite for ¯ ν-almost all a ∈ A. Let S = {a∈ A : ◦a is strongly monotonic, ◦e(a)2 > 0, ◦f(a) ∈ Q(◦p,◦a,◦e(a))}. (5.19) ¯ ν(S) > 0 by the conclusion of (1) and the additional assumption in (2), so in particular S
= ∅. Suppose a ∈ S. Then ◦p •◦e(a) ≥ ◦p2 •◦e(a)2 > 0. Let x =◦ f(a)+(1,0,0,...,0). Since◦a is strongly monotonic,x ◦a◦f(a). ◦p•x = ◦p•◦f(a) ≤ ◦p•◦e(a). There are two cases to consider. (a) ◦p•x<◦p•◦e(a): Then ◦f(a)
∈ Q(◦p,◦a,◦e(a)), a contradiction. (b) ◦p•x = ◦p•◦e(a) > 0. Since ◦a is continuous, there exists δ ∈ R++ such that y ∈ Rk +,|y−x|<δ
80 CHAPTER 5. LARGE ECONOMIES
implies y◦a◦f(a). We may ﬁnd y ∈ Rk + such that |y−x| <δand ◦p•y<◦p•x = ◦p•◦e(a), so ◦f(a)
∈ Q(◦p,◦a,◦e(a)), again a contradiction. Consequently, ◦p ∈ Δ++, so◦f(a) ∈ D(◦p,◦a,◦e(a)) by the conclusion of (1). 3. We show ﬁrst that f is S-integrable. Suppose S ⊂ A is internal and ν(S) 0. 1 n|| a∈S f(a)||∞ ≤ 1 nmini pi a∈S p•f(a) ≤ 1 nmini pi a∈S p•e(a)+*φB(f(a),p,e(a))
1 nmini pi a∈S p•e(a) 0 (5.20) since e is S-integrable. Thus, f is S-integrable. By Theorem 4.4.6,
a∈A ◦fd¯ ν = ◦ 1 n a∈A f(a)
= ◦ 1 n a∈A e(a)
=
a∈A ◦ed¯ ν, (5.21) and so (◦p,◦f) ∈W (◦χ). 4. (a) Suppose ◦a is strongly convex. We show ﬁrst that ◦a is strongly monotonic. Suppose x,y ∈ Rk + and x>y. Letz =2 x−y. Thenz+y 2 = x. Since z
= y, eitherx◦ax or x◦ay. If x◦ax,then x a x by Proposition 5.1.3, which contradicts irreﬂexivity. Therefore, we must have x◦ay, so◦a is strongly monotonic. Consequently,
5.5. CORE 81
the assumptions in (4) imply the assumptions in (2), so ◦p ∈ Δ++ and ◦f(a) ∈ D(◦p,◦a,◦e(a)) for ¯ ν-almost all a. (b) Suppose a ∈ A. Transferring Theorem 1 of Anderson (1981), *D(p,a,e(a)) contains exactlyone element. Deﬁne g(a)=* D(p,a,e(a)). For ¯ νalmost all a ∈ A, we havep• f(a) p •e(a) inf{p•x : x a f(a)} and e(a) is ﬁnite; consider any such a ∈ A. We will show that f(a) g(a). We consider two cases: i. If e(a) 0, then p•f(a) 0 p•g(a). Since ◦p >>0, f(a) 0 g(a), so f(a) g(a). ii. If e(a)
0, then p•e(a)
0. If f(a)
g(a), then either
◦f(a)+◦g(a) 2
◦a◦f(a) (5.22)
or
◦f(a)+◦g(a) 2
◦a◦g(a). (5.23) A. If equation 5.22 holds, then since ◦a is continuous and p•e(a)
0, we can ﬁnd w ∈ Rk + with p • w

p •e(a), such that w◦a◦f(a). By Proposition 5.1.3, w a f(a), which contradicts inf{p•x : x a f(a)} p•e(a). B. If equation 5.23 holds, we may ﬁnd w ∈ Rk + with p • w

p •e (a), such that w◦a◦g(a). By Proposition 5.1.3, w a g(a), which contradicts g(a)=*D(p,a,e(a)). Accordingly, f(a) g(a).
82 CHAPTER 5. LARGE ECONOMIES
Therefore, we have f(a) g(a) for ¯ ν-almost all a∈ A. 5. Suppose the assumptions in (1) and (4) hold and in addition e is S-integrable. The assumptions in (4) have been shown to imply the assumptions in (2), so f is Sintegrable by (3). As in (4), let g(a)=*D(p,a,e(a)). An easier version of the argument in (3) proves that g is S-integrable. Therefore
1 n a∈A
|f(a)−g(a)|
A ◦|f(a)−g(a)|d¯ ν = 0 (5.24)
by Theorem 4.4.6. Therefore
1 n|a∈A g(a)−e(a)| = 1 n|a∈A g(a)−f(a)| ≤ 1 na∈A|g(a)−f(a)| 0, (5.25) so g ∈W 0(χ).
Remark 5.5.3 Theorem 5.5.2 reveals some signiﬁcant differences between the hyperﬁnite and continuum formulations of large economies.
1. One can introduce atoms into both the hyperﬁnite and continuum (as in Shitovitz (1973,1974)) formulations. However, as noted by Hildenbrand on page 846 of this Handbook, this leads to problems in interpreting the preferences in the continuum formulation. In essence, the consumption set of a trader represented by an atom cannot be Rk +; it must allow consumptions inﬁnitely
5.5. CORE 83
large compared to those of other traders. In asymptotic analogues of the theorems, key assumptions1 are required to hold under rescalings of preferences; the economic content is then unclear, except in the special case of homothetic preferences. In the nonstandard formulation, this problem does not arise. Preferences over the nonstandard orthant *Rk + are rich enough to deal with atoms, although we do not cover this case in Theorem 5.5.2.
2. Now, let us compare how the nonstandard and continuum formulations treat the atomless case. In the continuum formulation, the endowment map is required to be integrable with respect to the underlying population measure. One could of course consider an endowment measure which is singular with respect to the underlying population measure. In this case, however, the representation of preferences becomes problematic. Speciﬁcally, if one considers a consumption measure μ which is singular with respect to the population measure, then μ has no Radon-Nikodym derivative with respect to the population measure, so one cannot identify the consumption of individual agents as elements of Rk +. Moreover, an allocation measure μ may allocate a coalition consumption which is inﬁnitely large compared to the consumption allocated that coalition by another measure μ . As in the case with atoms, the consumption space over which preferences need be deﬁned must be larger than Rk +. Asymptotic formulations require assumptions about rescaled preferences which are hard to interpret except in the case of ho
1Forexample, strong monotonicity,inconjunctionwithcompactness conditionsinherent inthe measure-theoretic formulationof convergence for sequences of economies, becomes a uniform monotonicitycondition.
84 CHAPTER 5. LARGE ECONOMIES
mothetic preferences. In the nonstandard framework, replacing the assumption that e is S-integrable with the much weaker assumption that e(a)/|A| 0 for all a ∈ A poses no technical problems. Part (1) of Theorem 5.5.2 analyzes precisely what happens in that case, while part (3) indicates how the result is strengthened if we assume that e is S-integrable and ◦a is strongly monotonic for a set of agents of positive ¯ ν-measure. Example 5.5.4 provides an example of a hyperﬁnite economy satisfying the hypotheses of part (1), but not those of part (3).
3. Suppose that the endowment map e is S-integrable, which corresponds to the integrability of endowment inherent in the deﬁnition of the continuum economy. In a continuum economy, allocations (including core allocations) are by deﬁnition required to be integrable. In the hyperﬁnite context, allocations may fail to be S-integrable. If f ∈ *C(χ) is not S-integrable, then
A ◦fd¯ν<◦ 1 n a∈A f(a)
= ◦ 1 n a∈A e(a)
=
A ◦ed¯ ν, (5.26) so ◦f does not correspond to an allocation of the associated Loeb measure economy. In Example 5.5.5, we present an example due to Manelli of a hyperﬁnite economy χ with a (non S-integrable) core allocation f such that 1 na∈A *φB(f(a),p,e(a))
0. However, core equivalence holds in the associated continuum economy χ, in the sense that g ∈C (◦χ) implies g ∈Q (◦χ). Indeed, Proposition 5.5.6 shows that, in the absence of monotonicity assumptions, any S-integrable core allocation f is close to an element of the core of ◦χ. In other words, the integrability condition in the deﬁni
5.5. CORE 85
tion of the continuum core is revealed by the hyperﬁnite formulation to be a strong endogenous assumption.
4. In Example 5.5.8, we present Manelli’s example of an economy χ with endowment e and core allocation f, both of which are S-integrable, such that
1 n a∈A
*φB(f(a),p,e(a)) 0 (5.27)
for some p ∈ *Δ but there is no p ∈ *Δ such that 1 na∈A *φ(f(a),p,a,e(a)) 0. Core equivalenceholds in the associated continuum economy χ, in the sense that g ∈C (◦χ) implies g ∈Q (◦χ). Indeed, ◦f ∈C (◦χ), so ◦f ∈Q (◦χ). In the example, the commodity bundles which show p is not an approximate supporting price for f are inﬁnite; they thus pose no barrier to the veriﬁcation of the support condition in the continuum economy.
5. The condition
*φ(f(a),p,a,e(a)) 0 (5.28) in the hyperﬁnite formulation implies the condition φ(◦f(a),◦p,◦ a,◦e(a)) = 0 (5.29) in the Loeb continuum economy which in turn implies ◦f(a) ∈ Q(◦p,◦a,◦e(a)). (5.30) In the presence of strong convexity, equation 5.28 implies that f(a) *Q(p,a,e(a)); (5.31)
86 CHAPTER 5. LARGE ECONOMIES
without strong convexity, equation 5.31 may fail, as shown by Example 5.5.9. The formulas 5.28 and 5.31 are nearly internal; using the Transfer Principle, we show in Theorem 5.5.10 that strong convexity of preferences implies a stronger form of convergence for sequence of ﬁnite economies. However, strong convexity is not needed to deduce formula 5.30 (which corresponds to the conclusion of Aumann’s Equivalence Theorem) from 5.29. Thus, in the continuum economy, convexity plays no role in the theorem. Since formula 5.30 is far from internal, it is not amenable to application of the Transfer Principle. Thus, the conclusion of Aumann’s Theorem does not reﬂect the behavior of sequences of ﬁnite economies, in the sense that it does not capture the implications of convexity for the form of convergence.
Example 5.5.4 (Tenant Farmers) In this example, we construct a hyperﬁnite economy in which the endowments are not S-integrable. Core convergence of the associated sequence of ﬁnite economies follows from Theorem 5.5.10; however, the sequence does not satisfy the hypotheses of Hildenbrand (1974) or Trockel (1976). 1. We consider a hyperﬁnite economy χ : A→ *(P×Rk +), where A = {1,...,n2} for some n ∈ *N\N. For all a ∈ A, the preference of a is given by a utility function u(x,y)=2√2x1/2 + y. The endowment is given by e(a)= (n +1 ,1) if a =1,...n (1,1) if a = n +1,...,n2 (5.32)
Think of the ﬁrst commodity as land, while the second commodity is food. The holdings of land are heavily
5.5. CORE 87
concentrated among the agents 1,...,n, a small fraction of the total population. Land is useful as an input to the production of food; however, the marginal product of land diminishes rapidly as the size of the plot worked by a given individual increases.
2. There is a unique Walrasian equilibrium, with p = 1 2, 1 2and allocation f(a)= (2,n) if a =1,...,n (2,0) if a = n +1,...,n2. (5.33)
Thus, the “tenant farmers” n +1,...,n2 purchase the right to use land with their endowment of food; they then feed themselves from the food they are able to produce on their rented plot of land.
3. By part (4) of Theorem 5.5.2, g ∈ (*C)(χ) ⇒ g(a) (2,0) for ¯ v-almost all a ∈ A, so that almost all of the tenant farmers receive allocations inﬁnitely close to their Walrasian consumption. A slight reﬁnement of Theorem 5.5.2 in Anderson (1981) shows that ◦⎛ ⎝ 1 |A| n2 a =n+1 g(a)⎞ ⎠= (2 ,0) (5.34) and ◦ 1 |A| n a=1 g(a)
= (0 ,1). (5.35) Thus, the per capita consumption allocated to the two classes (landowners and tenant farmers) is inﬁnitely close to the Walrasian consumptions of those classes.
88 CHAPTER 5. LARGE ECONOMIES
4. In the associated sequence of ﬁnite economies, if gn ∈ C(χn), one concludes by transfer that ⎛ ⎝ 1 |An| n2 a =n+1 gn(a)⎞ ⎠→(2,0) (5.36) and 1 |An| n a=1 gn(a)
→ (0,1). (5.37) 5. If one forms the associated continuum economy ◦χ via the Loeb measure construction, one gets
A ◦e(a)d¯ ν = (1 ,1)
(2,1) = 1 |A| a∈A e(a). (5.38) In other words, the measure-theoretic economy ◦χ has less aggregate endowment than the hyperﬁnite economy χ. In ◦χ, the unique Walrasian equilibrium has price √2 1+√2, 1 1+√2and consumption (1,1) almost surely. Thus, the continuum economy does not capture the behavior of the sequence χn of ﬁnite economies. Trockel (1976) proposed a solution involvingrescaling the weight assigned to the agents inthe sequence of ﬁnite economies. However, the example violates Trockel’s hypotheses, since the preferences do not converge under Trockel’s rescaling to a strongly monotone preference as he requires. We conclude that the assumption that endowments are integrable inthe continuum model represents a serious restriction on the ability of the continuum to capture the behavior of large ﬁnite economies.
Example 5.5.5 2 (Manelli)
2Examples 5.5.5,5.5.8,and5.5.9were originallygiveninthe context of a sequence of ﬁnite exchange economies.
5.5. CORE 89
1. We consider a hyperﬁnite exchange economy χ : A → *(Pc×R2 +). A ={1,...,n+2} with n ∈ *N\N. The endowment map is e(1) = e(2) = 0, e(a) = (1 ,1) (a = 3,...,n+2). LetV denote the cone {0}∪{x∈ R2++ : 0.5 < x1 x2 < 2}. Consider the allocation f(1) = (n,0),f(2) = (0, n 2),
f(a) = (0 , 1 2)( a =3,...,n).
(5.39)
The preferences have the property that
xa f(a) ⇐⇒ x−f(a) ∈ *V. (5.40) It is not hard to see that there are internal complete, transitive preferences that satisfy equation 5.40. In addition, we can choose a so that ◦a is locally nonsatiated for each a ∈ A. 2. It is not hard to verify that f ∈ *C(χ). However, f is not approximable by a core allocation of ◦χ. Indeed,
A ◦fd¯ ν = (0 , 1 2 )
= (1 ,1) =
A ◦ed¯ ν, (5.41) so ◦f is not even an allocation of ◦χ.
3. Given p∈ *Δ+,
1 n +2 a∈A
φB(f(a),p,e(a))
=
n|p1|+ n 2|p2|+ n|p1 + p2 2 | n +2 ≥
n 2(n + 2)
0. (5.42)
90 CHAPTER 5. LARGE ECONOMIES
4. ◦χ is an Aumann continuum economy with locally nonsatiated preferences. As Hildenbrand notes on page 845 of Chapter 18 of this Handbook, a careful examination of the original proof of Aumann’s Equivalence Theorem shows that C(◦χ)⊂Q (◦χ). In particular, g ∈C (◦χ)⇒∃ p∈ Δ
A φ(g(a),p,◦a,◦e(a))d¯ ν =0 . (5.43) Comparing equations 5.42 and 5.43, one sees that the decentralization properties of *C(χ) are totally diﬀerent from those of C(◦χ). By the Transfer Principle, one can construct a sequence of ﬁnite economies whose cores have the decentralization properties exhibited by *C(χ) rather than those exhibited by the Aumann continuum economy C(◦χ). 5. In Proposition 5.5.6, we show that if h is S-integrable and h ∈ *C(χ), then ◦h ∈C (◦χ); hence, h ∈C (◦χ)⇒∃ p∈ Δ
A φ(h(a),p,◦a,◦e(a))d¯ ν =0 (5.44) by item (4). Consequently, the properties of the internal core are signiﬁcantly diﬀerent from those of the set of S-integrable core allocations. By the Transfer Principle, one can construct a sequence of ﬁnite economies whose core allocations have the decentralization properties exhibited by *C(χ). Consequently, the restriction to integrable allocations inherent in the deﬁnition of the core in the Aumann economy is thus a strong endogenous assumption which prevents the Aumann economy from capturing the properties of certain sequences of ﬁnite economies. Proposition 5.5.6 (Brown, Robinson, Rashid) Suppose χ : A →*(Pc ×Rk +) is a hyperﬁnite exchange economy. If e
5.5. CORE 91
andf are S-integrable, and f ∈ *C(χ), then◦f ∈C (◦χ). Proof: A ◦fd¯ ν = ◦(1 na∈A f(a)) = ◦(1 na∈A e(a)) = A ◦ed¯ ν by Theorem 4.4.6. Thus, ◦f is an allocation of ◦χ. Suppose ◦f
∈C (◦χ). Then there exists S ∈Awith ¯ ν(S) > 0 and an integrable function g : S → Rk + such that S gd¯ ν = S ed¯ ν and g(a)◦a◦f(a) for ¯ ν-almost all a ∈ S.By Theorem 4.1.6, there exists T ∈Asuch that ¯ ν((S \T ) ∪ (T \ S)) = 0. Deﬁne g(a) = 0 fora ∈ T \ S. By Theorem 4.4.6, there is an S-integrable function G : T → *Rk + such that G(a) g(a) for ¯ ν-almost all a ∈ T . LetJ = {j ∈{ 1,...,k} :S ejd¯ ν =0 }. We can chooseG suchthat G(a)j = 0 for alla ∈ T , j ∈ J. LetT = {a ∈ T :G (a) a f(a)}. T ∈Aby the Internal Deﬁnition Principle; moreover, ν(T \T) 0. Given m ∈ N, letTm = {a ∈ T : y ∈ *Rk +,|y−G(a)| < 1 m ⇒ y a f(a)}. ThenTm ∈Abythe Internal Deﬁnition Principle and ¯ ν(∪m∈NTm)=¯ ν(T) by Propositions 3.1.5 and 5.1.3, and the fact that ◦g(a)◦a◦f(a) for ¯ ν-almost all a ∈ S. SinceG is S-integrable, there existsm ∈ N such that 1 n a∈Tm G(a)j ≥ 1 na∈T G(a)j 2 (5.45) for j ∈{ 1,...,k}\J. LetH(a)=G(a) ifa ∈ T \Tm. Fora ∈ Tm, deﬁne H(a)j = a∈T e(a)j |Tm| if j ∈ J (5.46) and H(a)j = 1−b∈T G(b)j −e(b)j b∈Tm G(b)j
G(a)j if j
∈ J. (5.47) Then H(a)=G(a) a f(a) fora ∈ T \ Tm. For a ∈ Tm,H (a) ∈ *Rk +, and|H(a)−G(a)| 0, so H(a) a f(a). Thus,
92 CHAPTER 5. LARGE ECONOMIES
H(a) a f(a) for all a ∈ T. An easy calculation shows that a∈T H(a)=a∈T e(a), so f
∈ *C(χ), a contradiction which completes the proof. Example 5.5.7 In this example, we show that the converse to Proposition 5.5.6 does not hold. Speciﬁcally, we construct an S-integrable allocation f such that ◦f ∈C (χ) but f
∈ *C(χ). In a sense, this example is merely a failure of lower hemicontinuity on the part of the core, a well-known phenomenon. Its importance lies in showing that the topology on Pc is inappropriate for the study of economies where large consumptions could matter. We consider a hyperﬁnite exchange economy χ : A →*(Pc ×R2 +). A = {1,...,n+1}with n ∈ *N\N. The endowment map is e(a) = (1 ,1) for all a ∈ A. Let p = (1− 1 n, 1 n). The preferences have the property that x1 (1,1) ⇐⇒[x >>(1,1)]∨[x >>(0, n 2 )]; x a (1,1) ⇐⇒ p•x>p•(1,1) (a =2,...,n+ 1) . (5.48) Consider the allocation f = e. f is Pareto dominated in χ by the allocation g(1) =1 n, 3n+3 4 ;g (a)= 1+1 n − 1 n2 , n+1 4n (a =2,...,n+ 1) . (5.49) Note however that ◦f (which equals f) is a Walrasian allocation of ◦χ with Walrasian price (1,0). One cannot block ◦f by ◦g precisely because g is not S-integrable. Accordingly, the restriction to integrable blocking allocations inherent in the deﬁnition of the core in the Aumann continuum economy is a signiﬁcant endogenous assumption.
Example 5.5.8 (Manelli)
5.5. CORE 93
1. We consider a hyperﬁnite exchange economy χ : A → *(Pc ×R2 +). A = {1,...,2n} with n ∈ *N\N. The endowment map is e(a) = (1 ,1) for all a ∈ A. LetV denote the cone {0}∪{x ∈ R2++ :0 .5 < x1 x2 < 2}.Consider the allocation
f(a) = (0 , 1 2), (a =1,...,n); f(a) = (2 , 3 2), (a = n +1,...,2n).
(5.50)
The preferences have the property that
xa f(a) ⇐⇒ x−f(a) ∈ *V (a =2,...,2n); x1 f(1) ⇐⇒ [[x−f(1) ∈ *V ]∨[x >>(n,0)]] (5.51) It is not hard to see that there are internal complete, transitive preferences that satisfy equation 5.51. In addition, we can choose a so that ◦a is locally nonsatiated for each a ∈ A. 2. It is not hard to verify that f ∈ *C(χ). Moreover, e and f are S-integrable, so ◦f ∈C (◦χ) by Proposition 5.5.6. As in item 4 of Example 5.5.5, there exists p ∈ Δ such that
A φ(◦f(a),p,◦a,◦e(a))d¯ ν =0 . (5.52) Indeed, it is easy to see that p = ±1 3,−2 3. Conse-quently,
1 |A| a∈A
*φB(f(a),p,e(a))
A φB(◦f(a),p,◦e(a)) = 0 (5.53)
94 CHAPTER 5. LARGE ECONOMIES by Theorem 4.4.6.3 However, with p =±1 3,−2 3, 1 |A| a∈A *φS(f(a),p,a)=−∞. (5.54) Comparing equations 5.54 and 5.52, one sees that the decentralization properties of f are quite diﬀerent from those of ◦f. By the Transfer Principle, one can construct a sequence of ﬁnite economies whose cores have the decentralization properties exhibited by f rather than those exhibited by ◦f.
Example 5.5.9 (Anderson, Mas-Colell) We consider a hyperﬁnite exchange economy χ : A → *(Pc ×R2 +). A = {1,...,n} with n ∈ *N\N. Fix a transcendental number ξ ∈ [0,1]. The endowment map is e(a) = (1 +ξa)(1,1) for all a ∈ A. Letδ = min{|a∈A ha(1 + ξa)| : h internal, ha ∈ {−1,0,1}, ha not all 0}. Sinceξ istranscendental, δ ∈ *R++. One can construct a homothetic preference∈*Pc such that (1 2, 3 2) (1,1) and (3 2, 1 2) (1,1), but such that f = e ∈ *C(χ); the idea is to make the region around (1 2, 3 2) and (3 2, 1 2) which is preferred to (1,1) very small.4 For any price q ∈ *Δ, ◦|f(a)−*D(q,a,e(a))|≥ 1 √2 for all a ∈ A. However, ◦f ∈ W(◦χ), in fact ◦f ∈ D((1 2, 1 2),◦a,◦e(a)) for all a ∈ A. As a consequence, the Aummann continuum economy fails to distinguish between the equivalence conditions in equation 5.28 (which says that the demand gap of the core allocation is small) and 5.31 (which says that the core allocation is close to the demand set). In particular, convexity plays no role in Aumann’s equivalence theorem, while it signiﬁcantly alters the form of the equivalence theorem for hyperﬁnite
3It is also easy to verify equation 5.53 by direct reference to the hyperﬁnite economy χ. 4See Anderson and Mas-Colell (1988) for details.
5.5. CORE 95
economies; by the Transfer Principle, convexity signiﬁcantly alters the form of core convergence for sequences of large ﬁnite economies.
Theorem 5.5.10 Brown, Robinson, Khan, Rashid, Anderson Let χn : An → (Pc × Rk +) be a sequence of ﬁnite exchange economies.
1. If (a) |An|→∞ ; (b) for each n ∈ N and a ∈ An, a i. is monotonic; ii. satisﬁes free disposal; (c) i. lim1 na∈A e(a) << ∞;ii. lim1 na∈A e(a) >> 0; and(d) maxa∈An |e(a)| |An| →0; then for every sequence fn ∈C (χn), there exists a sequence pn ∈ Δ+ such that 1 |An| a∈An φ(fn(a),pn,a,e(a))→ 0. (5.55)
2. If the assumptions in (1) hold and in addition there is an compact set K of strongly monotonic preferences and δ ∈ R++ such that for each commodity i and each n ∈ N, |{a∈ An :a∈ K, e(a)i ≥δ}| |An| ≥ δ, (5.56) then there is a compact set D ⊂Δ++ and n0 ∈ N such that pn ∈ D for all n ≥n0.
96 CHAPTER 5. LARGE ECONOMIES
3. If the assumptions in (1) and (2) hold and in addition the endowment sequence {en : n ∈ N} is uniformly integrable, then the sequence {fn : n ∈ N} is uniformly integrable.
4. If the assumptions in (1) hold and in addition (a) for all γ ∈ R++, there is an compact set K of strongly convex preferences such that for all n ∈ N, |{a∈ An :a∈ K}| |An| > 1−γ; (5.57) (b) there is a δ ∈ R++ such that, for each commodity i, |{a∈ An : e(a)i ≥ δ}| |An| ≥ δ; (5.58) (c) a is irreﬂexive, convex, and strongly convex for all n ∈ N and all a ∈ An; then for each ∈ R++, |{a∈ An :|fn(a)−D(pn,a,e(a))|>}| |An| →0.(5.59)
5. If the assumptions in (1) and (4) hold and, in addition, e is S-integrable, then there exists a sequence n → 0 and gn ∈W n(χn) such that 1 |An| a∈An |fn(a)−gn(a)|→0. (5.60)
Proof:
5.5. CORE 97
1. This follows immediately from Anderson (1978); see also Dierker(1975). The proof giveninAnderson (1978) was originally discovered by translating nonstandard proofs of part (1) of Theorem 5.5.2 and a weaker version of part (1) of Theorem 5.5.10. Note that if n ∈ *N\N, thenχn satisﬁes the hypotheses of part (1) of Theorem 5.5.2.
2. Suppose the additional assumption in (2) holds. By Transfer, for all n ∈ *N, ν({a ∈ An : a∈ *K,e(a)i ≥ δ}) ≥ δ. Ifa∈ *K, then◦a ∈ K by Theorem 3.3.2, so ◦a is strongly monotonic. Hence, for n ∈ *N\N, χn satisﬁes the assumptions of part (2) of Theorem 5.5.2. Hence, ◦pn ∈ Δ++. Hence, for n ∈ *N \ N, ◦pn ∈ Δ++. LetM = {n ∈ N : pn
∈ Δ++}. If M is inﬁnite, then there exists n ∈ *M ∩(*N\N), a contradiction. Hence M is ﬁnite; let n0 = (maxM)+1. Let D = {◦pn : n ∈ *N,n≥ n0}. D is compact by Proposition 3.3.7, D ⊂ Δ++, andpn ∈ D for all n ≥ 0, n ∈ N.
3. Suppose that the sequence en is uniformly integrable. Then for n ∈ *N\N, en is S-integrable by Proposition 4.4.8. By part (3) of Theorem 5.5.2, fn is S-integrable for n ∈ *N. Then the sequence {fn : n ∈ N} is uniformly integrable by Proposition 4.4.8.
4. Fix ∈ R++. It is easy to see that the assumptions in (4) imply that the assumptions of part (4) of Theorem 5.5.2 hold for n ∈ *N\N. Thus, forn ∈ *N\N, νn({a∈ An :|fn(a)−*D(pn,a,e(a))|>}) 0. (5.61)
98 CHAPTER 5. LARGE ECONOMIES
By Proposition 3.1.9, for n ∈ N, νn({a∈ An : fn(a)−D(pn,a,e(a)) > }) →0.(5.62) 5. For n ∈ N, choosepn and gn ∈ D(pn,a,e(a)) to minimize 1 |An|a∈An |fn(a) − gn(a)|. If n ∈ *N \ N, then χn satisﬁes the hypotheses of part (5) of Theorem 5.5.2, so 1 |An| a∈An |fn(a)−gn(a)| 0. (5.63) By Proposition 3.1.10,
n =
1 |An| a∈An
|fn(a)−gn(a)|→0. (5.64)
Then gn ∈W n(χn), which completes the proof.
5.6 Approximate Equilibria
This section will give a discussion of Khan (1975), Khan and Rashid (1982), and Anderson, Khan and Rashid (1982).
5.7 Pareto Optima
This section will give a discussion of Khan and Rashid (1975) and Anderson (1988).
5.8 Bargaining Set
This section will give a discussion of Geanakoplos (1978).
5.9. VALUE 99
5.9 Value
This section will contain a discussion of Brown and Loeb (1976).
5.10 “Strong” Core Theorems
This section will contain a discussion of Anderson (1985) and Hoover (1989).
100 CHAPTER 5. LARGE ECONOMIES
Chapter 6
Continuum of Random Variables
6.1 The Problem
In modelling a variety of economic situations, it is desirable to have a continuuum of independent identically distributed random variables, and to be able to assert that, with probability one, the distribution of outcomes of those random variables equals the theoretical distribution; in other words, there is individual uncertainty but no aggregate uncertainty. Some applications include Lucas and Prescott (1974), Diamond and Dybvig (1983), Bewley (1986), and Faust (1988); see Feldman and Gilles (1985) for other references. There is no diﬃculty in deﬁning a continuum of independent, identically distributed random variables. Suppose (Ω0,C0,ρ0) is a probability space, and X :Ω 0 → R a ran-dom variable with distribution function F. Let (Ω ,B,ρ)= t∈[0,1](Ω0,B0,ρ0), and deﬁne Xt(ω)=X(ωt). Then the family {Xt : t ∈ [0,1]} is a continuum of independent random variables with distribution F.
101
102CHAPTER 6. CONTINUUMOF RANDOM VARIABLES
The problem arises in the attempt to formulate the statement that there isno aggregate uncertainty. Suppose ([0,1],B,μ) is the Lebesgue measure space. Given ω ∈ Ω, the empirical distribution function should be deﬁned as Fω(r)=μ({t ∈ [0,1] : Xt(ω) ≤ r}. Unfortunately, {t ∈ [0,1] : Xt(ω) ≤ r} need not be measurable, so the empirical distribution function need not be deﬁned.
Judd (1985) considered a slightly diﬀerent construction of (Ω,B,μ) due to Kolmogorov. In it, he shows that {ω : Fω is deﬁned} is a non-measurable set with outer measure 1 and inner measure 0. Thus, one can ﬁnd an extension μ of the Kolmogorov measure μ such that Fω is deﬁned for μ -almost all ω. However, {ω : Fω = F} is not measurable with respect to μ ; in fact, it hasμ outer measure 1 and μ inner measure 0. Thus, one can ﬁnd an extension μ of μ with the property that μ ({ω : Fω = F}) = 1. However, the extensions to μ and μ are arbitrary, leaving the status of economic predictions from such models unclear.
A variety of standard constructions have been proposed to alleviate the problem (Feldman and Gilles(1985), Uhlig (1988), and Green (1989)). Much earlier, Keisler gave a broad generalization of the Law of Large Numbers for hyperﬁnite collections of random variables on Loeb measure spaces (Theorem 4.11 of Keisler (1977)). Since Loeb measure spaces are standard probability spaces in the usual sense, this provides a solution of the continuum of random variables problem. In section 6.2, we provide a simpliﬁed version of Keisler’s result. In section 6.3, we describe a nontatonnement price adjustment model due to Keisler (1979, 1986, 1990, 1992, 1996); in Keisler’s model, individual uncertainty over trading times in a hyperﬁnite exchange economy results in no aggregate uncertainty.
6.2. LOEB SPACE CONSTRUCTION 103
6.2 Loeb Space Construction
Loeb probability spaces are standard probability spaces in the usual sense, but they have many special properties. In the following construction, the internal algebra is guaranteed to be rich enough to ensure that the measurability problems outlined in Section 6.1 never arise. The construction also satisﬁes an additional uniformity condition highlightedin Green (1989), since the conclusion holds on every subinterval of the set of traders (conclusion 2b of Theorem 6.2.2, below). Construction 6.2.1 Let (A,A,ν) be as in the construction of Lebesgue measure (Construction 4.2.1). Suppose Y : A → *R is ν-measurable, and ◦|Y (a)| <∞for ¯ ν-almost all a∈ A. Deﬁne Ω = a∈A A, R = (* P)(Ω), ρ(B)=|B| |Ω| for B ∈R , Ya(ω)=Y (ωa), and Xa(ω)=◦Y (ωa).
Theorem 6.2.2 (Keisler) Consider Construction 6.2.1. Let F be the distribution function for ◦Y , i.e. F(r)=¯ ν({a∈A : ◦Y (a) ≤r}). 1. Xa is ¯ ρ-measurable, and has distribution function F, for all a∈ A; 2. for ¯ ρ-almost all ω ∈ Ω, for all r ∈ R, for all s,t ∈ [0,1] satisfying s 1. Xa is ¯ ρ-measurable by Theorem 4.4.2. For all a ∈ A, the distribution function of Xa equals F, the distribution function of ◦Y .
104CHAPTER 6. CONTINUUMOF RANDOM VARIABLES
2. If r ∈ R, letCr = {a ∈ A : ◦Y (a) ≤ r}. Since Cr = ∩∞ m=1{a ∈ A : Y (a) ≤ r + 1 m}, ◦ν({a ∈ A : Y (a) ≤ r + 1 m}) → ¯ ν(Cr) asm →∞. Therefore,there exists m ∈ *N\N such that ν({a ∈ A : Y (a) ≤r + 1 m}) ¯ ν(Cr), by Proposition 3.1.9; let Dr = {a ∈A : Y (a) ≤r + 1 m}. Let W = {(r,s,t) ∈ R × [0,1] × [0,1] : s = ∅. Choose g ∈∩ V∈FP(R)GV and deﬁne Dr = g(r) for all r ∈ T1. For (r,s,t) ∈ T, letk be the greatest element of *N less than or equal to n(t − s). Let Ast = A ∩ *[s,t]. nν(Y −1(Dr) ∩ Ast) has a *-binomial distribution B(k,ν(Dr)), so it has mean kν(Dr) and standard deviationkν(Dr)(1−ν(Dr)) < √k by Feller (1957) and the Transfer Principle. Thus, ν(Y−1(Dr)∩Ast) has mean kν(Dr) n and standard deviation less than √k n ≤ 1 √n. Therefore, ρω : ν(Y−1(Dr)∩Ast)− kν(Dr) n >n −1 4≤ 1 √n (6.1)
by Chebycheﬀ’s Inequality (Feller (1957)). Therefore ρω :∃(r,s,t)∈ T, ν(Y−1(Dr)∩Ast)− kν(Dr) n >n −1 4≤ n1/4 1 √n 0. (6.2)
6.3. PRICE ADJUSTMENT MODEL 105
Therefore
¯ ρ({ω :∀(r,s,t)∈ W ν(Y−1(Dr)∩Ast)=( t−s)F(r)})=1 ,
(6.3)
so
¯ ρ({ω :∀(r,s,t)∈ W¯ ν({a∈ Ast : ◦Y (a)≤ r}) ≥(t−s)F(r)})=1 . (6.4) Similarly,
¯ ρ({ω :∀(r,s,t)∈ W¯ ν({a∈ Ast : ◦Y (a)≤ r}) ≤(t−s)F(r)})=1 , (6.5)
so
¯ ρ({ω :∀(r,s,t)∈ W,¯ ν({a∈ Ast : ◦Y (a)≤ r})=( t−s)F(r)})=1 . (6.6)
Remark 6.2.3 Let G be an arbitrary distribution function. There exists a random variable Z deﬁned on the Lebesgue measure space [0,1] with distribution function F. Deﬁne Z : A → R by Z (a)=Z(◦a). Z has distribution function G by Theorem 4.2.2. There exists a ν-measurable function Y : A → *R such that ◦Y (a)=Z (a) almost surely by Theorem 4.4.2; the distribution function of ◦Y is G. Thus, Construction 6.2.1 allows us to produce a continuum of independent random variables with any desired distribution.
6.3 Price Adjustment Model
In the tatonnement story of the determination of equilibrium prices, it is assumed that a ﬁctitious Walrasian auctioneer
106CHAPTER 6. CONTINUUMOF RANDOM VARIABLES
announces a price vector, determines the market excess demand at that price, and then adjusts the price in a way which hopefully leads eventually to equilibrium. No trade is allowed until the equilibrium price is reached. This story has two critical ﬂaws:
1. If no trade is allowed at the non-equilibrium prices called out by the auctioneer, why should individual agents bother to communicate their excess demands to the auctioneer? If they do not convey their excess demands, how does the auctioneer determine what the social excess demand is? The more we require the auctioneer to know, the less the tattonement story ﬁlls the role of providing foundations for a theory of decentralization by prices.
2. Convergence to the equilibrium price requires a countable number of steps. If there is a technological lower bound on the length of time needed for the auctioneer to elicit the excess demand information, equilibrium cannot be reached in ﬁnite time. Thus, no trade occurs in ﬁnite time.
Thus, it is highly desirable to replace the tatonnement story with a model which allows trade out of equilibrium, and in which the information required to adjust prices is kept to a minimum. Keisler has developed such a model using a hyperﬁnite exchange economy in which agents are chosen to trade randomly. remainder of this section, we sketch Keisler’s result, listing the principal assumptions and conclusions in a special case. For a complete statement, see Keisler(1979,1986,1990,1992,1996).1
1In the monograph, we intend to expand this section to include a formal statement of Keisler’s result in the special case considered here.
6.3. PRICE ADJUSTMENT MODEL 107
The set of agents is A = {1,...,n} and the time line is T =j n : j ∈ *N,j≤ n2for n ∈ *N\N. There arek ∈ N commodities. There is a central market. At each time t ∈ T, one agent is chosen at random to go to the market and trade. Thus, the underlying probability space is (Ω,B,ν), as described below. 1. Ω = AT. Thus, an element ω ∈ Ω is an internal function from T to A. Ifω(t)=a, then agent a is chosen to go to market at time t. 2. B is the set of all internal subsets of Ω. 3. ν(B)=|B| |A| for B ∈B . D(p,I,a) denotes the demand of agent a with income I and price vector p. The prevailing price in the market is set initially at an arbitrary price p(1 n). The market has an initial inventory I(1 n)=(n,...,n). Each agent begins with an endowment f(a,0). The prevailing price at time t, denoted p(t), the market inventory at time t, denoted I(t), and the commodity bundle of agent a at time t, denoted f(a,t) are determined inductively by the formula
f(a,t)= ⎧ ⎪ ⎨ ⎪ ⎩ f(a,t− 1 n) if ω(t)
= a D(p(t),p(t)•f(a,t− 1 n),a) ift>0 & ω(t)=a pt + 1 n = p(t)+λ f(ω(t),t)−f(ω(t),t− 1 n)¬ It + 1 n= I(t)+ f(ω(t),t)−f(ω(t),t− 1 n)¬. (6.7)
We hope to replace Keisler’s parametrization assumption (discussed brieﬂy below) with a compactness condition.
108CHAPTER 6. CONTINUUMOF RANDOM VARIABLES
In other words, in each time period, the agent who trades purchases his/her demand given the prevailing price and his/her holding from previous trades, the net trade of these agents is taken from the market inventory, and the price is adjusted proportionately to the net trade of this agent. The parameters λ and n are chosen so that λ = n−c for some c ∈ (0,1), so ◦(λn)=∞,λlog(λn) 0 (6.8) (in particular λ 0). The inventory parameter 0, but is not too small. The demand functions of the agents are assumed to be parametrizable in a certain fashion. This parametrization assumption appears to be a form of compactness condition on the demand functions. An economy with a ﬁnite (in the standard sense) number of types of agents with C1 demands satisfying a global Lipschitz condition will satisfy the parametrization assumption. It is also assumed that the initial endowment f(a,0) is uniformly bounded by some M ∈ N and that p(0) is ﬁnite. The evolution of the economy is a random process, depending on the realization of the random variable ω which determines which agents trade at each time. We can associate a deterministic price adjustment process deﬁned by the diﬀerential equation
q(0) = p(0) q (t)=◦1 na∈A[D(p(t),p(t)•f(a,0),a)−f(a,0)]. (6.9) We assume that the solution of equation 6.9 is exponentially stable, with limit p, a Walrasian equilibrium price, for every initial value in a neighborhood of p(0). Keisler shows that for ¯ ν-almost all ω ∈ Ω, the followingproperties hold.
6.3. PRICE ADJUSTMENT MODEL 109
1. p(t) *q(λnt) for all t ∈ T with ◦t<∞. Note that since q(t) → p as t →∞and λn is inﬁnite, p(t) p for ﬁnite t∈ T satisfying ◦(λnt)=∞. Thus, (a) the path followed by the price is, up to an inﬁnitesimal, deterministic; and (b) the price becomes inﬁnitely close to the Walrasian equilibriumprice p in inﬁnitesimal time and stays inﬁnitely close to p for all ﬁnite times. 2. I(t) 0 for all t ∈ T. Thus, an initial market inventory which is inﬁnitesimal compared to the number of agents suﬃces to ensure that the trades desired by the agents are feasible when the agents come to market. 3. For almost all agents a, there exists t(a) ∈ T with ◦t(a) < ∞ such that f(a,t) D(p,p • f(a,0),a) for all t ≥ t(a). Thus, almost all agents trade at a price inﬁnitely close to the Walrasian price p, and they consume their demands at p. An inﬁnitesimal proportion of the trade takes place at prices outside the monad of the Walrasian price p.
110CHAPTER 6. CONTINUUMOF RANDOM VARIABLES
Chapter 7
Noncooperative Game Theory
111
112CHAPTER 7. NONCOOPERATIVE GAME THEORY
Chapter 8
Stochastic Processes
113
114 CHAPTER 8. STOCHASTIC PROCESSES
Chapter 9
Translating Nonstandard Proofs
Because hyperﬁnite economies possess both the continuous properties of measure-theoretic economies(via the Loeb measure construction) and the discrete properties of large ﬁnite economies (via the Transfer Principle), they provide a tool for convertingmeasure-theoretic proofs intoelementary ones. The strategy for doing this involves taking a measure-based argument, and interpreting it for Loeb measure economies; the interpretation typically involves the use of formulas with iterated applications of external constructs. One can then proceed on a step-by-step basis to replace the external constructs with internal ones; each time one does this, the conclusion of the theorem is typically strengthened. In most cases, the process terminates with one or more external constructs still present, and no tractable internal arguments to replace them. However, it is occasionally possible to replace all the external constructs; if one succeeds in doing this, the internal proof is (with *’s deleted) a valid standard proof which is elementary in the sense that measure theory is not used.
115
116 CHAPTER 9. TRANSLATION
An extended discussion of translation techniques is given in Rashid (1987). Given the limitations of space, we will limit ourselves to listing a few examples.
1. The elementary proof of a core convergence result in Anderson (1978) was obtained by applying the translation process to a nonstandard version of the KannaiBewley-Grodal-Hildenbrandapproach to core limittheorems using weak convergence reported in Hildenbrand (1974). Much of the groundwork for the translation was laid in Brown and Robinson (1974,1975) who ﬁrst developednonstandard exchange economies, and in Khan (1974b) and Rashid (1979). A critical phase in the translation was carriedout by Khan and Rashid (1976), who showed that one can dispense with the assumption that almost all agents in the hyperﬁnite economy have preferences which are nearstandard in the space of monotone preferences.
2. Anderson, Khan and Rashid (1982) presents an elementary proof of the existence of approximate Walrasian equilibria (in the sense that per capita market excess demand is small) in which the bound on the excess demand is independent of compactness conditions on preferencessuch as uniform monotonicity. The proof is a translation of the nonstandard proof in Khan and Rashid (1982). As in item 1, a key to the successful completion of the translation was the discovery that one preferences in the hyperﬁnite economy need not be nearstandard in the space of monotone preferences.
3. Anderson (1985,1988) proved that “strong” versions of core convergence theorems and the second welfare theorem hold with probabilityone in sequencesof economies obtained by sampling agents’ characteristics from a
117
probability distribution, even with nonconvex preferences; here, “strong” means that agent’s consumptions are close to their demand sets. The proofs are highly external, and so would appear poor candidates for translation. However, they required checking certain conditions for standard prices only; since the set of standard prices can be embedded in a hyperﬁnite set by Theorem 1.13.4, this suggested strongly that the key was to consider only ﬁnitely many prices at a time. Hoover (1989) recently succeeded in giving a standard proof by carrying out the translation. Hoover’s proof is elementary in the sense that it uses little measure theory beyond the bare bones necessary to deﬁne sequences of economies obtained by sampling from a probability distribution.
118 CHAPTER 9. TRANSLATION
Chapter 10
Further Reading
There are a number of other applications of nonstandard analysis in economics which space did not permit us to discuss in Anderson (1990). We hope to cover some of them in the monograph, but for now we limit ourselves to the following listing of references:
1. Richter (1971) and Blume, Brandenburger and Dekel (1991a,1991b) on the representation of preferences;
2. Lewis (1977), Brown and Lewis (1981), and Stroyan (1983) on inﬁnite time horizon models;
3. Geanakoplos and Brown (1982) on overlapping generations models;
4. Muench and Walker (1981) and Emmons (1984) on public goods economies; and
5. Simon and Stinchcombe (1989) on equilibrium reﬁnements in noncooperative games.
There are a number of approachable books giving an introduction to nonstandard analysis. We particularly recom
119
120 CHAPTER 10. FURTHER READING
mend Hurd and Loeb (1985), which gives a thorough nonstandard development of the principal elements of real analysis. Keisler (1976), an instructor’s manual to accompany a calculus text based on inﬁnitesimals, is very useful for those who ﬁnd mathematical logic intimidating. An entirely different approach to nonstandard analysis is given in Nelson (1977). Rashid (1987) provides a broader survey of the applications of nonstandard analysis to the large economies literature, and gives an extended discussion of techniques for translating nonstandard proofs intoelementarystandard proofs. There is an extensive literature on stochastic processes, including Brownian motion and stochastic integration, based on the Loeb measure. Since stochastic processes play an important role in ﬁnance, this is a potentially fertile area for future applications of nonstandard analysis to economics. See Anderson (1976), Keisler (1984), and Albeverio, Fenstad, Høegh-Krohn and Lindstrøm (1986).
Appendix A
Proof of the Existence of Nonstandard Extensions
This appendix will provide a complete proof of the existence of nonstandard extensions as deﬁned in Chapter 2. æ
121
122 APPENDIX A. EXISTENCE PROOF
Bibliography
[1] Albeverio, Sergio, Jens Erik Fenstad, Raphael HøeghKrohn, and Tom Lindstrøm (1986), Nonstandard Methods in Stochastic Analysis and Mathematical Physics. Orlando: Academic Press.
[2] Anderson, Robert M. (1976), “A Nonstandard Representation for Brownian Motion and Itˆo Integration”, Israel Journal of Mathematics, 25:15-46.
[3] Anderson, Robert M. (1978), “An Elementary Core Equivalence Theorem”, Econometrica, 46:1483-1487.
[4] Anderson, Robert M. (1981), “Core Theory with Strongly Convex Preferences”, Econometrica, 49:14571468.
[5] Anderson, Robert M. (1982), “Star-ﬁnite Representations of Measure Spaces”, Transactions of the American Mathematical Society, 217:667-687.
[6] Anderson, Robert M. (1985), “Strong Core Theorems with Nonconvex Preferences”, Econometrica, 53:12831294.
[7] Anderson, Robert M. (1988), “The Second Welfare Theorem with Nonconvex Preferences”, Econometrica, 56:361-382.
123
124 BIBLIOGRAPHY
[8] Anderson, Robert M. (1992), “Nonstandard Analysis with Applications to Economics”, in Werner Hildenbrand and Hugo Sonnenschein (eds.), Handbook of Mathematical Economics, Volume IV. Amsterdam: North-Holland.
[9] Anderson, Robert M. and Andreu Mas-Colell (1988), “An Example of Pareto Optima and Core Allocations Far from Agents’ Demand Correspondences”, appendix to Anderson (1988), Econometrica, 56:379-381.
[10] Anderson, Robert M., M. Ali Khan and Salim Rashid (1982), “Approximate Equilibria with Bounds Independent of Preferences”, Review of Economic Studies, 44:473-475.
[11] Anderson, Robert M. and Salim Rashid (1978), “A Nonstandard Characterization of Weak Convergence”, Proceedings of the American Mathematical Society, 69:327332.
[12] Aumann, Robert J. (1964), “Markets with a Continuum of Traders”, Econometrica, 32:39-50.
[13] Bewley, Truman F. (1973), “Edgeworth’s Conjecture”, Econometrica, 41:425-454.
[14] Bewley, Truman F. (1986), “Stationary Monetary Equilibrium with a Continuum of Independently Fluctuating Consumers”, in Werner Hildenbrand and Andreu Mas-Colell (eds.), Contributions to Mathematical Economics: In Honor of G´erard Debreu, 79-102. Amsterdam: North-Holland.
[15] Billingsley, Patrick (1968), Convergence of Probability Measures. New York: John Wiley and Sons.
BIBLIOGRAPHY 125
[16] Blume, Lawrence, Adam Brandenburger and Eddie Dekel (1991a), “Lexicographic Probabilities and Choice under Uncertainty”, Econometrica 59:61-80.
[17] Blume, Lawrence, Adam Brandenburger and Eddie Dekel (1991b), “Equilibrium Reﬁnements and Lexicographic Probabilities”, Econometrica 59:81-98.
[18] Bourbaki, N. (1970), Th´eorie des Ensembles. Paris: Hermann.
[19] Brown, Donald J. (1976), “Existence of a Competitive Equilibrium in a Nonstandard Exchange Economy”, Econometrica 44:537-546.
[20] Brown, Donald J. and M. Ali Khan (1980), “An Extension of the Brown-Robinson Equivalence Theorem”, Applied Mathematics and Computation, 6:167-175.
[21] Brown, Donald J. and Lucinda M. Lewis (1981), “Myopic Economic Agents”, Econometrica, 49:359-368.
[22] Brown, Donald J. and Peter A. Loeb (1976), “The Values of Nonstandard Exchange Economies”, Israel Journal of Mathematics, 25:71-86.
[23] Brown, Donald J. and Abraham Robinson (1974), “The Cores of Large Standard Exchange Economies”, Journal of Economic Theory, 9:245-254.
[24] Brown, Donald J. and Abraham Robinson (1975), “Nonstandard Exchange Economies”, Econometrica, 43:41-55.
[25] Davis, Martin (1977), Applied Nonstandard Analysis. New York: Wiley.
126 BIBLIOGRAPHY
[26] Diamond, D. W. and P. H. Dybvig (1983), “Bank Runs, Deposit Insurance and Liquidity”, Journal of Political Economy 91:401-419.
[27] Dierker, Egbert (1975), “Gains and Losses at Core Allocations”, Journal of Mathematical Economics, 2:119128.
[28] Emmons, David W. (1984), “Existence of Lindahl Equliibria in Measure-Theoretic Economies without Ordered Preferences”, Journal of Economic Theory 34:342-359.
[29] Emmons, David W. and Nicholas C. Yannelis (1985), “On Perfectly Competitive Economies: Loeb Economies”, in C. D. Aliprantis, O. Burkinshaw and N. J. Rothman (eds.), Advances in Equilibrium Theory, Lecture Notes in Economics and Mathematical Systems, 244:145-172.
[30] Faust, Jon (1988), “Theoretical and Empirical Asset Price Anomalies”, Ph.D. Dissertation, Department of Economics, University of California at Berkeley.
[31] Feldman, Mark and Christian Gilles (1985), “An Expository Note on Individual Risk without Aggregate Uncertainty”, Journal of Economic Theory 35:26-32.
[32] Feller, William (1957), An Introduction to Probability Theory and Its Applications Vol. I, second edition. New York: John Wiley and Sons.
[33] Geanakoplos, John (1978), “The Bargaining Set and Nonstandard Analysis”, preprint, Department of Economics, Harvard University.
BIBLIOGRAPHY 127
[34] Geanakoplos, John and Donald J. Brown (1982), “Understanding Overlapping Generations Economies as a Lack of Market Clearing at Inﬁnity”, preprint, Department of Economics, Yale University.
[35] Green, Edward J. (1989), “Individual-Level Randomness in a Nonatomic Population”, Working Paper #227, Department of Economics, University of Pittsburgh.
[36] Hildenbrand, Werner (1974), Core and Equilibria of a Large Economy. Princeton: Princeton University Press.
[37] Hildenbrand, Werner (1982), “Core of an Economy”, in Kenneth J. Arrow and Michael D. Intriligator (eds.), Handbook of Mathematical Economics, Volume II, Amsterdam: North-Holland Publishing Company, 831-877.
[38] Hoover, Douglas N. (1989), private communication.
[39] Hurd, Albert E. and Peter A. Loeb (1985), An Introduction to Nonstandard Real Analysis. New York: Academic Press.
[40] Judd, Kenneth L. (1985), “The Law of Large Numbers with a Continuum of IID Random Variables”, Journal of Economic Theory 35:19-25.
[41] Keisler, H. Jerome (1976), “Foundations of Inﬁnitesimal Calculus”, Boston: Prindle, Weber and Schmidt.
[42] Keisler, H. Jerome (1977), “Hyperﬁnite Model Theory”, in R. O. Gandy and J. M. E. Hyland (eds.),Logic Colloquium 1976, 5-110. Amsterdam: North-Holland.
[43] Keisler, H. Jerome (1979), “A Price Adjustment Model with Inﬁnitesimal Traders”, preprint, Department of Mathematics, University of Wisconsin.
128 BIBLIOGRAPHY
[44] Keisler, H. Jerome (1984), “An Inﬁnitesimal Approach to Stochastic Analysis”, Memoirs of the American Mathematical Society, 297.
[45] Keisler, H. Jerome (1986), “A Price Adjustment Model with Inﬁnitesimal Traders”, in Hugo Sonnenschein (ed.), Models of Economic Dynamics, Lecture Notes in Economics and Mathematical Systems, 264. Berlin: Springer-Verlag.
[46] Keisler, H. Jerome (1990), “Decentralized Markets with Fast Price Adjustment”, preprint, Department of Mathematics, University of Wisconsin–Madison.
[47] Keisler, H. Jerome (1992), “A Law of Large Numbers withFast Price Adjustment.”Transactions of the American Mathematical Society, 332:1-51.
[48] Keisler, H. Jerome (1996), “Getting to a Competitive Equilibrium,” Econometrica, 64:29-49.
[49] Khan, M. Ali (1974a), “Some Remarks on the Core of a ‘Large’ Economy”, Econometrica 42:633-642.
[50] Khan, M. Ali (1974b), “Some Equivalence Theorems”, Review of Economic Studies, 41:549-565.
[51] Khan, M. Ali (1975), “Some Approximate Equilibria”, Journal of Mathematical Economics, 2:63-86
[52] Khan, M. Ali (1976), “Oligopoly in Markets with a Continuum of Traders: An Asymptotic Interpretation”, Journal of Economic Theory, 12:273-297.
[53] Khan, M. Ali and Salim Rashid (1975), “Nonconvexity and Pareto Optimality in Large Markets”, International Economic Review 16:222-245.
BIBLIOGRAPHY 129
[54] Khan, M. Ali and Salim Rashid (1976), “Limit Theorems on Cores with Costs of Coalition Formation”, preprint, Johns Hopkins University.
[55] Khan, M. Ali and Salim Rashid (1982), “Approximate Equilibria in Markets with Indivisible Commodities”, Journal of Economic Theory 28:82-101.
[56] Lewis, Alain A. (1985), “Hyperﬁnite Von Neumann Games”, Mathematical Social Sciences, 9:189-194.
[57] Lewis, Lucinda M. (1977), “Essays on Purely Competitive Intertemporal Exchange”, Ph.D. Dissertation, Yale University.
[58] Loeb, Peter A. (1975), “Conversion from Nonstandard to Standard Measure Spaces and Applications in Potential Theory”, Transactions of the American Mathematical Society, 211:113-122.
[59] Loeb, Peter A. (1979), “Weak Limits of Measures and the Standard Part Map”, Proceedings of the American Mathematical Society, 77:128-135.
[60] Lucas, Robert E. Jr. and Edward C. Prescott (1974), “Equilibrium Search and Unemployment”, Journal of Economic Theory 7:188-209.
[61] Luxemburg, W. A. J. (1969), “A General Theory of Monads”, in W. A. J. Luxemburg (ed.), Applications of Model Theory to Algebra, Analysis and Probability, New York: Holt, Rinehart and Winston.
[62] Manelli, Alejandro (1991), “Monotonic Preferences and Core Equivalence”, Econometrica (forthcoming)****.
130 BIBLIOGRAPHY
[63] Mas-Colell, Andreu (1985), The Theory of General Economic Equilibrium: A Diﬀerentiable Approach. Cambridge: Cambridge University Press.
[64] Muench, Thomas and Mark Walker (1979), “Samuelson’s Conjecture: Decentralized Provision and Financing of Public Goods”, in Jean-Jacques Laﬀont (ed.), Aggregation and Revelation of Preferences, Studies in Public Economics 2. Amsterdam: North-Holland.
[65] Nelson, Edward (1977), “Internal Set Theory: a New Approach to Nonstandard Analysis”, Bulletin of the American Mathematical Society, 83:1165-1198.
[66] Rashid, Salim (1978), “Existence of Equilibrium in Inﬁnite Economies with Production”, Econometrica, 46: 1155-1163.
[67] Rashid, Salim (1979), “The Relationship Between Measure-Theoretic and Non-standard Exchange Economies”, Journal of Mathematical Economics 6:195202.
[68] Rashid, Salim (1987), Economies with Many Agents: An Approach Using Nonstandard Analysis. Baltimore: Johns Hopkins University Press.
[69] Richter, Marcel K. (1971), “Rational Choice”, in John S. Chipman, Leonid Hurwicz, Marcel K. Richter, and Hugo F. Sonnenschein (eds.), Preferences, Utility, and Demand. New York: Harcourt Brace Jovanovich, 29-58.
[70] Robinson, Abraham (1966), Non-standard Analysis. Amsterdam: North-Holland Publishing Company.
[71] Royden, H. L. (1968), Real Analysis. New York: Macmillan Publishing Co.
BIBLIOGRAPHY 131
[72] Rudin, Walter (1976), Principles of Mathematical Analysis, Third Edition. New York: McGraw Hill.
[73] Shitovitz, Benyamin(1973), “Oligopoly in Markets with a Continuum of Traders”, Econometrica 41:467-501.
[74] Shitovitz, Benyamin (1974), “On some Problems Arising in Markets with some Large Traders and a Continuum of Small Traders”, Journal of Economic Theory 8:458-470.
[75] Simon, Leo K. and Maxwell B. Stinchcombe (1989), “Equilibrium Reﬁnement in Games with Large Strategy Spaces”, preprint, Department of Economics, University of California at San Diego.
[76] Maxwell B. Stinchcombe (1992), “When Approximate Results are Enough: The Use of Nonstandard Versions of Inﬁnite Sets in Economics”, preprint, Department of Economics, University of California at San Diego.
[77] Stroyan, K. D. (1983), “Myopic Utility Functions on Sequential Economies”, Journal of Mathematical Economics, 11:267-276.
[78] Stroyan, K. D. and W. A. J. Luxemburg (1976), Introduction to the Theory of Inﬁnitesimals. New York: Academic Press.
[79] Stutzer, Michael J. (1987) “Individual Risk without Aggregated Uncertainty: A Nonstandard View”, preprint, Federal Reserve Bank of Minneapolis.
[80] Trockel, Walter (1976), “A Limit Theorem on the Core”, Journal of Mathematical Economics, 3:247-264.
132 BIBLIOGRAPHY
[81] Uhlig, Harald (1988), “A Law of Large Numbers for Large Economies”, preprint, Institute for Empirical

Inﬁnitesimal Methods in Mathematical Economics

你可能感兴趣的:(Inﬁnitesimal Methods in Mathematical Economics)