What does SAS automatically do to the values of variables at the start of each DATA step?

At the beginning of each DATA step, SAS automatically sets the values of all variables to missing. This means that any new variable introduced in the DATA step that has not been assigned a value will have a missing value until it is explicitly assigned one within the step. This behavior helps ensure that calculations or operations performed on the data do not inadvertently use uninitialized values, which could lead to incorrect results or errors.

The concept of missing values is particularly important in SAS, as it treats numeric and character missing values differently, impacting data analysis and manipulations. By initializing variables to missing at the start of each DATA step, SAS provides a clean slate for data processing, helping programmers avoid potential pitfalls associated with uninitialized or garbage values.

The other options do not accurately represent how SAS handles variable initialization. For example, retaining variable values would mean that they keep their values from the previous DATA step iteration, which can lead to unintended consequences if not used carefully. Similarly, initializing with the first observation's values implies a continuity that does not occur at the start of a DATA step, as each variable begins the step in a missing state. Setting values to zero is also incorrect, as missing values and zero are treated distinctly in SAS, especially in the context of data analysis.