Jump to content

Utility functions on indivisible goods

From Wikipedia, the free encyclopedia

This is an old revision of this page, as edited by Erel Segal (talk | contribs) at 13:09, 3 September 2020 (Submodularity and supermodularity). The present address (URL) is a permanent link to this revision, which may differ significantly from the current revision.

Some branches of economics and game theory deal with indivisible goods, discrete items that can be traded only as a whole. For example, in combinatorial auctions there is a finite set of items, and every agent can buy a subset of the items, but an item cannot be divided among two or more agents.

It is usually assumed that every agent assigns subjective utility to every subset of the items. This can be represented in one of two ways:

  • An ordinal utility preference relation, usually marked by . The fact that an agent prefers a set to a set is written . If the agent only weakly prefers (i.e. either prefers or is indifferent between and ) then this is written .
  • A cardinal utility function, usually denoted by . The utility an agent gets from a set is written . Cardinal utility functions are often normalized such that , where is the empty set.

A cardinal utility function implies a preference relation: implies and implies . Utility functions can have several properties.[1]

Monotonicity

Monotonicity means that an agent always (weakly) prefers to have extra items. Formally:

  • For a preference relation: implies .
  • For a utility function: implies (i.e. u is a monotone function).

Monotonicity is equivalent to the free disposal assumption: if an agent may always discard unwanted items, then extra items can never decrease the utility.

Additivity

Additive utility
0
apple 5
hat 7
apple and hat 12

Additivity (also called linearity or modularity) means that "the whole is equal to the sum of its parts." That is, the utility of a set of items is the sum of the utilities of each item separately. This property is relevant only for cardinal utility functions. It says that for every set of items,

assuming that . In other words, is an additive function. An equivalent definition is: for any sets of items and ,

An additive utility function is characteristic of independent goods. For example, an apple and a hat are considered independent: the utility a person receives from having an apple is the same whether or not he has a hat, and vice versa. A typical utility function for this case is given at the right.

Submodularity and supermodularity

Submodular utility
0
apple 5
bread 7
apple and bread 9

Submodularity means that "the whole is not more than the sum of its parts (and may be less)." Formally, for all sets and ,

In other words, is a submodular set function.

An equivalent property is diminishing marginal utility, which means that for any sets and with , and every :[2]

.

A submodular utility function is characteristic of substitute goods. For example, an apple and a bread loaf can be considered substitutes: the utility a person receives from eating an apple is smaller if he has already ate bread (and vice versa), since he is less hungry in that case. A typical utility function for this case is given at the right.

Supermodular utility
0
apple 5
knife 7
apple and knife 15

Supermodularity is the opposite of submodularity: it means that "the whole is not less than the sum of its parts (and may be more)". Formally, for all sets and ,

In other words, is a supermodular set function.

An equivalent property is increasing marginal utility, which means that for all sets and with , and every :

.

A supermoduler utility function is characteristic of complementary goods. For example, an apple and a knife can be considered complementary: the utility a person receives from an apple is larger if he already has a knife (and vice versa), since it is easier to eat an apple after cutting it with a knife. A possible utility function for this case is given at the right.

A utility function is additive if and only if it is both submodular and supermodular.

Subadditivity and superadditivity

Subadditive but not submodular
0
X or Y or Z 2
X,Y or Y,Z or Z,X 3
X,Y,Z 5

Subadditivity means that for every pair of disjoint sets

In other words, is a subadditive set function.

Assuming is non-negative, every submodular function is subadditive. However, there are non-negative subadditive functions that are not submodular. For example, assume that there are 3 identical items, , and Z, and the utility depends only on their quantity. The table on the right describes a utility function that is subadditive but not submodular, since


Superadditive but not supermodular
0
X or Y or Z 1
X,Y or Y,Z or Z,X 3
X,Y,Z 4

Superadditivity means that for every pair of disjoint sets

In other words, is a superadditive set function.

Assuming is non-positive, every supermodular function is superadditive. However, there are non-negative superadditive functions that are not supermodular. For example, assume that there are 3 identical items, , and Z, and the utility depends only on their quantity. The table on the right describes a utility function that is non-negative and superadditive but not supermodular, since

A utility function with is said to be additive if and only if it is both superadditive and subadditive.

With the typical assumption that , every submodular function is subadditive and every supermodular function is superadditive. Without any assumption on the utility from the empty set, these relations do not hold.

In particular, if a submodular function is not subadditive, then must be negative. For example, suppose there are two items, , with , and . This utility function is submodular and supermodular and non-negative except on the empty set, but is not subadditive, since

Also, if a supermodular function is not superadditive, then must be positive. Suppose instead that . This utility function is non-negative, supermodular, and submodular, but is not superadditive, since

Unit demand

Unit demand utility
0
apple 5
pear 7
apple and pear 7

Unit demand (UD) means that the agent only wants a single good. If the agent gets two or more goods, he uses the one of them that gives him the highest utility, and discards the rest. Formally:

  • For a preference relation: for every set there is a subset with cardinality , such that .
  • For a utility function: For every set :[3]

A unit-demand function is an extreme case of a submodular function. It is characteristic of goods that are pure substitutes. For example, if there are an apple and a pear, and an agent wants to eat a single fruit, then his utility function is unit-demand, as exemplified in the table at the right.

Gross substitutes

An illustration of the containment relations between common classes of utility functions.

Gross substitutes (GS) means that the agents regards the items as substitute goods or independent goods but not complementary goods. There are many formal definitions to this property, all of which are equivalent.

  • Every UD valuation is GS, but the opposite is not true.
  • Every GS valuation is submodular, but the opposite is not true.

See Gross substitutes (indivisible items) for more details.

Hence the following relations hold between the classes:

See diagram on the right.

Aggregates of utility functions

A utility function describes the happiness of an individual. Often, we need a function that describes the happiness of an entire society. Such a function is called a social welfare function, and it is usually an aggregate function of two or more utility functions. If the individual utility functions are additive, then the following is true for the aggregate functions:

Aggregate
function
Property Example
values of functions
on {a}, {b} and {a,b
}
f g h aggregate(f,g,h)
Sum Additive 1,3; 4 3,1; 4 4,4; 8
Average Additive 1,3; 4 3,1; 4 2,2; 4
Minimum Super-additive 1,3; 4 3,1; 4 1,1; 4
Maximum Sub-additive 1,3; 4 3,1; 4 3,3; 4
Median neither 1,3; 4 3,1; 4 1,1; 2 1,1; 4
1,3; 4 3,1; 4 3,3; 6 3,3; 4

See also

References

  1. ^ Gul, F.; Stacchetti, E. (1999). "Walrasian Equilibrium with Gross Substitutes". Journal of Economic Theory. 87: 95–124. doi:10.1006/jeth.1999.2531.
  2. ^ Moulin, Hervé (1991). Axioms of cooperative decision making. Cambridge England New York: Cambridge University Press. ISBN 9780521424585.
  3. ^ Koopmans, T. C.; Beckmann, M. (1957). "Assignment Problems and the Location of Economic Activities" (PDF). Econometrica. 25 (1): 53–76. doi:10.2307/1907742. JSTOR 1907742.