Q&A: API scores online?

Yesterday I found myself answering some questions about school API data asked by a parent in the midst of the school enrollment process. Seems relevant for this venue, so here's a reprise of the conversation:

Back in August, the Chron published a stand-alone section with SF's API scores. I happened to have recycled that paper. I've been searching online at both sfgate and the district site, to no avail. Has anyone by chance found that info on the web somewhere?

You can get them from the links on this blog post: 2006 SFUSD API data

In particular, the summary spreadsheet makes it easy to see the data for each school over the whole 7 year span. I also publish the detailed data that shows all of the subgroup data as well as the data that goes into the "SCI" that is used to conjure up the "similar schools" ranks.

Thanks for the info, I had a look at it all and have no idea what all the numbers mean, Is there somewhere that tells you what all the abbreviations are and what the numbers really tell you?

The short answer is, they mean very little. Relying on standardized testing data is a really rotten way to assess schools for your child. It is no substitute for the up- close, in-person experience of visiting with and talking to members of a school community.

That said...

The state provides a lot of documentation of the data at this site:
Academic Performance Index (API)

and this one that documents *all* the acronyms:
API Data Files - Academic Performance Index

Let's keep it relatively simple and look at the summary data which tabulate the API score, the Statewide Rank, and the Similar Schools Rank.

The API score is based on all the standardized, "norm referenced" testing -- STAR testing -- that most public school students throughout the state have to take. The scores range from 200 to 1000 -- and the goal is for all schools to reach an 800 score. If they are not there yet, they are given targets and progress is monitored.

The Statewide Rank is purely based on API score. All schools are put into ten buckets of equal size based on their API. From there they get a 1-10 score. (Elementary, Middle, and High schools have separate sets of buckets. Lots of other meaningless fine print pertains.)

The Similar School Rank is supposed to adjust the ranking based on the demographic factors affecting the students at each school. If there are more language learners, higher rates of poverty, more special education kids, more parents without college degrees... the schools API scores are put on a curve.

The SCI, or "School Characteristic Index", is the key to this curve. There are reams of data that are collected about all CA students that are then thrown into a massive regression analysis that looks at how relevant the individual demographic data are when correlated to API scores. With this correlation, the demographics of each school are then used to weight or curve the actual API results.

So Clarendon ES, with high scores and high statewide ranks can appear to have mediocre results while Sanchez ES has a low rank but may be doing very well with the challenges its students face.

The cold numeric way the SCI and "Similar" schools are computed is open to skepticism. Some odd bedfellows are inevitably lumped together as "similar". It is interesting and useful to try and adjust the pure test scores based on demographics. But it is just as likely to prove the limits of testing data as it is to render more useful ways of looking at that testing data.