Track Record · Case Study

California is about to spend billions cleaning up freight. Which way actually saves the most lives?

The headline finding flips one of the most-cited “co-benefits” of California freight electrification at the level most cost-benefit studies actually compute it. The standard reduced-complexity air-quality model says cutting NOx in the LA Basin lowers ozone. The chemistry says the opposite — LA is a VOC-limited basin, well-known to atmospheric chemists since the 1990s but routinely glossed over in fast cost-benefit work. We measured the dollar exposure in our own cascade at ~$10.4B. Once that correction is in, only one of the candidate portfolios still comes out ahead under every uncertainty we stress-tested. What follows is the work behind that finding: 57 chained investigations, a ~700,000-person real cohort, and a five-level atmospheric chemistry ladder that pinned the sign and size of the ozone channel.

By Michael Key · ORCID

The decision behind the spending

California has billions of dollars flowing through cap-and-trade revenue, AB 617 community-air funds, and CEC clean-energy programs aimed at the same problem: freight pollution that kills people, mostly in disadvantaged communities along the I-710 and the San Joaquin Valley.

The default story is that electrifying trucks and buildings is a clear win — the diesel goes away, the air gets cleaner, the health benefits are obvious. The reality is messier. Which trucks first? Buildings or vehicles or biomass plants? How fast? At what cost to ratepayers? Where do the benefits actually land?

The numbers behind those answers come from a stack of models, each with its own assumptions. Some of those assumptions are quietly wrong in ways that move the answer by billions of dollars. Some of them produce a single point estimate where the honest answer is a distribution. And nobody’s decision should rest on a study that hasn’t been pressure-tested.

This study is that pressure test.

Eight focus areas

The study is long, so it’s split into eight focus areas. Each one answers a question the decision-maker actually has to settle before spending the money. Every page leads with what we found, walks through how we got there, and points to the underlying investigations in the appendix.

What changed at each step

The answer flipped twice on the way through. That’s the most useful thing the study did — not the final number, but the stages where the early answer stopped being defensible.

  1. Start. A standard emissions inventory said the $2 B accelerated transport push (T2) avoids 87 deaths a year by 2035. Clear win.
  2. Add dose-response uncertainty. The published health-effects numbers carry a wide confidence interval. Once we propagated it, the baseline scenario (just enforcing existing rules) was actually the better choice in 55.8% of plausible draws.
  3. Add regime-aware ozone chemistry. The standard linear air-quality model (used by most fast cost-benefit work) assumes cutting NOx always lowers ozone. In a VOC-limited basin like LA it doesn’t — well-known chemistry, but the linear shortcut routinely glosses over it. Once we ran the regime-aware version, a ~$10.4B “co-benefit” the cheaper analysis would have credited to freight electrification went away.
  4. Stack every uncertainty at once. Six axes of uncertainty across 114,688 Saltelli draws. Most of the candidate portfolios stop coming out ahead. One survives: the $4 B all-in option (D), with positive net benefit in 99.99% of draws.
  5. Stress-test what we missed. A wildfire-PV side analysis found California’s solar fleet loses about $49 M a year to smoke (with panel soiling, not irradiance, as the dominant channel). A separate sensor-VOI analysis returned a clean negative: at canonical decision conditions, more monitors don’t change the answer.

What this can’t tell you

The honest list, before you cite any number above:

What the analysis supports

With the caveats above:

Explore the data

Interactive Tool Decision Dashboard Pick a budget scale and a health-effects anchor. See which portfolio survives. Sortable across six robustness lenses with the close call on the runner-up surfaced directly. Twelve portfolios, three dose-response anchors, two budget scales.

Sources and methods

Built on peer-reviewed dose-response anchors (Di et al. 2017, Krewski et al. 2009), an open-source reduced-complexity air-quality model (Tessum et al. 2017’s ISRM), CARB’s AB 617 disadvantaged-community list, EPA AQS monitoring data, GFED5 wildfire emissions, Childs et al. wildfire PM2.5, and CMIP6 climate projections. The dose-response work is grounded in a ~700,000-person NHIS-linked Medicare cohort.

All 57 chained investigations are listed in the investigation appendix — the eight focus areas above link to the most load-bearing pages; the remaining 14 are supporting analyses (validation rungs, cross-checks, foundational methodology) summarized there with one line each.