In a 2000 meta-analysis, the National Reading Panel did find evidence that using phonics was a better way to teach reading, and it seems that the real problem with whole language learning was that it hurt the kids at the bottom the most. We withheld the most effective method from the kids who needed the most help! The averages, however, mostly remained the same. The more recent NAEP declines happened well after some schools decided to reintroduce phonics, and the decline in scores in both reading and mathematics indicates that there is something else going on. (Given the timing, COVID-19 seems like a likely candidate, though the downward trend was already starting in 2019.)
What was also oversimplified was the idea that we only taught phonics before we switched to whole language learning. This is false. There were various competing methods, including ‘look-say,’ which was the subject of Why Johnny Can’t Read. The story of American literacy is not just the story of phonics.
This doesn’t mean that I don’t support phonics. The story of Mississippi’s rise in literacy provides some evidence that phonics-based instruction can improve childhood literacy. But this is partial evidence, and I am now slightly less confident that phonics is the universal solution.