Details on the State Service Performance Index (SSPI) are here. Read that first to provide context for this post.
This post is a follow-up to that post addressing the potential criticism that the weights I have applied represent ideological skewing.
Yes. All weights are ideological. The moment you combine multiple domains into a composite index, you are making value judgments. If you weight infrastructure heavily, you are signaling that economic productivity matters more. If you weight education and health heavily, you are signaling that equity and human development matter more. If you weight public safety heavily, you are signaling that order is foundational.
There is no neutral composite.
That is precisely why the SSPI does not present a single sacred ranking. It presents:
- An efficiency-focused weighting.
- An equity-focused weighting.
- A sensitivity analysis showing how much rankings change when weights move.
If a state’s ranking swings wildly depending on weight choice, that tells you something important: its performance is domain-specific and contested. If a state ranks consistently high (or low) across weighting schemes, that tells you something even more important: its institutional structure is robust (or structurally weak). In other words the presence of weighting does not invalidate the model. It makes the value assumptions explicit.
What is ideological is pretending that raw statistics are neutral. They aren’t—raw NAEP scores embed demographic composition; raw road condition grades ignore traffic demand; raw crime rates ignore crime load complexity; raw life expectancy ignores income structure. Those are hidden weightings.
The SSPI makes them visible.
I have one additional note. I won’t pretend that I did not use LLM AI as a “force multiplier” in creating the SSPI; I did. It was fun, instructive, and it allowed me to do an enormous amount of work in a relatively short period of time. It’s a demonstration of what can and will be done going forward using AI.
