4 ms·

This should have been the most predictable result ever. If only the top 20% of students are taking the test, as used to be the case, then the average score is

This should have been the most predictable result ever. If only the top 20% of students are taking the test, as used to be the case, then the average score is going to be quite high. If you expand that pool with weaker students, it's going to pull the average down. If you make everyone take the SAT, then you are going to wind up with a lot of really low scores dragging that average down hard. They run into this all the time with mandated standardized testing. In smaller school districts, just the relative number of special ed students from class to class can create so much variance that year-to-year comparisons are useless.

Nasrudith 7y ago
Yeah that has been used so many times with the "old test questions prove we are getting stupider!" trope circulating.
rowanG077 7y ago
There are easy statistical methods to still get useable data from something like this. Average is just the average.
- thrower123 7y ago
  Oh, of course. But statistical illiteracy hurts us badly here. Even just distinguishing between median and mean is beyond the abilities of most popular reporting. As it is, this article is almost worthless. The raw data would be interesting to go through.
thaumasiotes 7y ago
The official numbers are wildly against you on this. Here are the numbers of testees, total and those scoring 700+ on the math section, for recent years and then in 5-year intervals back to 1999: year takers math 700+ 2018 2,136,539 202,088 2017 1,715,481 120,043 2016 1,637,589 117,067 2015 1,698,521 121,057 2014 1,672,395 123,781 2009 1,530,128 97,296 2004 1,519,870 94,985 1999 1,302,903 71,214 (numbers pulled from https://research.collegeboard.org/programs/sat/data/archived https://research.collegeboard.org/programs/sat/data/archived ) In 1999, 1.3 million students took the SAT and 5.5% of them scored 700+ on the math section. Just last year, 2.1 million took the SAT and 9.5% scored 700+ on the math section! Or year-over-year, we can see that the marginal 421,058 students taking the SAT in 2018 compared to 2017 earned a marginal 82,045 700+ math scores, which at 19.5% is well over double the rate at which testees earned those scores in 2017. This makes it clear that the additional students are mostly long-overlooked geniuses. (In fact, what's happening is that the college board is making it much easier to get a high score. A 720 math in 2018 is a much lower score than a 700 math in 2004. Sometime between 2004 and 2009, they stopped even providing the information of how many students scored 750+ on each section; 700+ is the finest resolution they're willing to provide now.)

Nasrudith 7y ago

Yeah that has been used so many times with the "old test questions prove we are getting stupider!" trope circulating.

rowanG077 7y ago

There are easy statistical methods to still get useable data from something like this. Average is just the average.

thrower123 7y ago
Oh, of course. But statistical illiteracy hurts us badly here. Even just distinguishing between median and mean is beyond the abilities of most popular reporting. As it is, this article is almost worthless. The raw data would be interesting to go through.

thaumasiotes 7y ago

The official numbers are wildly against you on this. Here are the numbers of testees, total and those scoring 700+ on the math section, for recent years and then in 5-year intervals back to 1999: year takers math 700+ 2018 2,136,539 202,088 2017 1,715,481 120,043 2016 1,637,589 117,067 2015 1,698,521 121,057 2014 1,672,395 123,781 2009 1,530,128 97,296 2004 1,519,870 94,985 1999 1,302,903 71,214 (numbers pulled from https://research.collegeboard.org/programs/sat/data/archived https://research.collegeboard.org/programs/sat/data/archived ) In 1999, 1.3 million students took the SAT and 5.5% of them scored 700+ on the math section. Just last year, 2.1 million took the SAT and 9.5% scored 700+ on the math section! Or year-over-year, we can see that the marginal 421,058 students taking the SAT in 2018 compared to 2017 earned a marginal 82,045 700+ math scores, which at 19.5% is well over double the rate at which testees earned those scores in 2017. This makes it clear that the additional students are mostly long-overlooked geniuses. (In fact, what's happening is that the college board is making it much easier to get a high score. A 720 math in 2018 is a much lower score than a 700 math in 2004. Sometime between 2004 and 2009, they stopped even providing the information of how many students scored 750+ on each section; 700+ is the finest resolution they're willing to provide now.)