In Relational Database terms efficiency revolves around avoiding the duplication of data. The 3 tables that you show in your question seem reasonable provided:
- You can have a Locality without having a Country
- You can have a Country without having a Locality
This is actually quite a viable design, since in some parts of the world they may not define locations as you describe. (There are a lot of unique decisions made around the planet.) Of course you can generate an artificial name or description for your tracking purpose should that be necessary.
For example: There is no named locality, you could create a Locality like: "Island NW of Salty Creek"
If you intend for each Locality to belong to a single Country then you might benefit from a small change to add a CountryLocality table:
- PK UN LocalityID
- UN LocalityName
- PK UN CountryID
- UN CountryName
- PK UN CountryLocalityID
- FK CountryID
- FK LocalityID
- FK CountryLocalityID
This is definitely a design and usage issue for you to decide on since the CountryLocality table adds some overhead. It depends on how strong the Country-Locality connection should be and whether this extra step is worth it for your application.