At the time of writing, some of the other commentors have already provided answers regarding the ease of implementation aspect of interpreters. Another important advantage of implementing an interpreter is that it usually gives you portability for free (or with modest extra works).

The nice point of figuring out optimization of interpreters is that you get the speedup on all architectures, while in a JIT you usually have to change the backend (or worse if you have to create a new backend from scratch.)

edit: I am sorry, another commenter already hinted at the portability aspect.