Rather than trying to account for the supply chain pull as you noted, I wish cities and states would standardize upon an open source formatting and presentation of the data in their accounting systems. Many competing analyses of costs to various jurisdictions can then arise from private citizens and organizations, and then we can pit them against each other historically to converge upon a small set of models that are consistently predictive. We can even then gather insightful analyses from weaving together accounting and GIS datasets, for example, that weren't possible before without creating entire new bureaucracies.
With more accurate cost management accounting models, we can then more reliably challenge organizations petitioning for specific tax concessions: we can reasonably accurately model the influx of X thousand employees to have $Y impact upon the cost accounting structure over Z years. NPV of that cost is $N. If spread out over F forecasted years of the concession you are proposing, at equivalent long-term bond rate of your corporation of P% at your longest-projected credit quality, then that comes to $T total cost to the jurisdiction over the time period.
Then the per-employee per-$10,000 wage-adjusted cost burden to the jurisdiction on a per-unit basis becomes a calculated, known amount. Tilt the unit measure towards favoring lots of employees at moderate wages instead of a few employees at extremely high wages to incentivize high monetary velocity in the local economy, lest the extremely high wages of a few employees get translated away into capital gains-structured instruments. Compare that to existing employment data, and you know if the jurisdiction is over-paying or not for the concessions. And now the jurisdictions have a metric against which to demand a surety bond: with any future adjustments to wage and/or employee count, if it falls below a certain threshhold, the jurisdiction can collect upon a proportional amount in the surety bond for non-performance. Benefits dollars don't count, because the jurisdiction is only interested in the localized monetary velocity impact upon the localized economy that comes from wages.