There are many, many approaches, each having its pros and cons when it comes to simplicity of reading, easiness of modifying data, easiness of modifying data schema (refactoring), easiness of parsing data, type of data you want to model.
If you were I would experiment. I would pickup the simplest approach (properties) to see if fits your problem. Here's the list of more approaches.
Cons: schema-less - you need to write your own parser if you want to model more complex data structures (e.g. subproperties or multi-dimensional data)
Pros: models well tabular data
Cons: may not be sufficient for more complex data structures
Pros: self-descriptive, models well tree-like data
Cons: Very verbose
Pros: All pros of XML + binding for refactoring and editing from IDE
Cons: like in XML
Pros: Good trade-off between self-descriptiveness and conciness
Pros: Centralized access, though everything above can be centralized (+versioned) using any versioning system
Cons: Hard to refactor, when you want to change schema, requires opening SQL client each time you want to modify it
Beans in your programming language
Pros: Tightly bound to your test code but still not hard-coded, easy to refactor together with your test scripts
Cons: Sometimes to verbose