Numeric Precision - Storing an Integer
The default length of a numeric variable is 8 bytes. Numeric variables in SAS are stored in floating point format (essentially 1 byte for an exponent and 7 bytes for the mantissa). The exact representation varies from machine to machine. On all machines I know of, the storage representation is called "double-precision floating point." On most unix machines this type of storage (8 bytes) yields 16 decimal digits of precision. If you store a number in less bytes, you loose some precision.
The table below shows the largest integer such that all integers less than or equal to it can be exactly represented by a floating point number with the given number of bytes.

Length in bytes	Largest Integer Represented Exactly	Exponential
3	8,192	2**13
4	2,097,152	2**21
5	~536 million	2**29
6	~137 billion	2**37
7	~35 trillion	2**45
8	~9,007 trillion	2**53

A floating-point number must be stored in at least 3 bytes when using SAS on a PC.