I think at least it depends on what kind of analysis you conduct and the
quality of your model. A rule of thumb for parameter estimation is that the
precision is in the order of square root of the sample size. So in this case
, it would not reduce the sample size significantly.
