I have noticed on ACT Explore the same score have very different percentiles for each subject. For any given score Math has the highest percentile and Science the lowest with Reading and English falling in between.

So question is are the raw scores comparable as far as grade level equivalence or the percentile. For example if a kid gets 15 on both Math and Science. Does this mean evenness in capabilities or should one look more at the percentile?

I think the percentiles are more meaningful than the raw scores.

The percentiles are against 8th graders who took them. The raw scores may depend on how easy or hard the questions are to the students. There are at least some minor variance in score each year.

But bear in mind that the percentile differential is huge around raw score of 14-18 range. (45 to 85 in percentiles for 4th grade ENG as per 2013-14). So, 1 or 2 silly mistake can change quite a lot.

The rule of thumb is to use deviation-based scores (like percentiles usually are) if you want to compare performance to a group, or performance on different tasks to each other. Raw scores generally cannot be compared between any two statistically unrelated tests. The rapid change in percentiles to either side of the 50th %ile is because of the bell curve. The significance of the difference in percentiles is also relatively modest between, say, the 35th and 65th %ile, due to the "bunchiness" of the population around the mean.

Well I am actually looking to compare my DS to grade level competence. For the same score which conclusion would you draw?

1. Better in Math because of higher percentile
2. Equally capable in Science and math because of same score
3. None of the above

I would say none of the above.

I would say 3 for None of the above. I don't believe you can use the Explore scores "to compare [your] DS to grade level competence." For example, DS hit 99 percentile compared to 8th graders when he was 9, but accelerating him to 8th grade would have been laughable. He would not have been truly competent. The Explore is just 4 half-hour subject tests given in a multiple choice format. Aside from the obvious writing skills and executive function deficits, there is so much more covered by a decent curriculum in a good school. A high score on the Explore when taken out of level tells you that your child is highly capable.

Grade equivalents on norm-referenced tests are not generally reliable for determining instructional level. They tell you what score the average child of that grade-level would obtain on this specific set of tasks, but do not translate directly to the instructional level of the test-taker.

1. The median child of a certain grade-level may perform above or below the expected instructional level for that grade-level.
2. Even though the number of items correct is the same, the actual correct items that resulted in this score may differ between the test-taker and the median child of the nominal grade-equivalent.
3. Nationally norm-referenced tests are only a sampling of the skills expected in any local educational community; as they are not comprehensive, there is no measure of mastery/non-mastery for any of the (majority of) skills not assessed.

Percentiles will tell you only the rank-order comparison of your child's performance in each skill area to that of children in the norm group. They will have the same deficits as far as sampling, and a lack of correspondence between median children and instructional expectations.

If you want a better measure of instructional level, I would suggest a criterion-referenced measure, such as end-of-course or placement assessments from the curriculum actually in use in your school system.