No, that would be an invalid inference because he actually sampled from only half of the total population not the entire population.
so 12% of the 50% he sampled would be valid, but since he did not use the entire population, the results cannot affect all of them.