Methodology

How the figures on this site are sourced, cleaned, estimated and graded, and where they should not be used. Version 1.1, Q4 2026 edition, data through 2026-06-30.

Data sources and limitations

Sources

Only sources whose published terms clearly allow commercial use are used: US federal government works in the public domain (class a) and sources under an open licence that permits commercial use (class b). Sources with non-commercial terms, or with no stated terms, are excluded even where they hold more detail.

The cost records are reported to the Department of the Interior by states, tribes and federal bureaus for each well plugged with Infrastructure Investment and Jobs Act funds. Pennsylvania records are excluded. State-published copies of the same records, state well databases and state contract reports are not used, because their terms are either non-commercial or not stated.

Limitations

• One cost source. All cost records come from a single federal dataset. If it changes or is withdrawn, coverage changes with it.
• Orphan wells only. The wells are mostly old, shallow and vertical, plugged under public contracts. The figures do not describe operator-plugged wells or deep horizontal wells.
• Depth, age and condition. The cost source carries none of these. Depth has been joined for 38% of wells from the sources listed above; for Texas, Illinois, Nebraska, Indiana, Oklahoma, West Virginia and California no depth source with commercial-use terms is available. Age is known for fewer wells and condition for none. Most estimates therefore work by location and well type, and are weaker where no comparable wells have been plugged nearby.
• Averages. About half the records are contract or lease averages, which understate the spread between individual wells.
• Scope differs. Agencies differ on whether surface restoration is in the reported cost.
• Not a quote. Even a grade A estimate misses a typical held-out well by up to 25%.
• Legal status. The sources above were reviewed by legal counsel and cleared for commercial use on 2026-10-05. Sources that were not part of that review stay excluded.

Cleaning

• Records with no cost or a cost of zero are dropped. Offshore wells are dropped.
• Michigan's 14 cost records are dropped: each is a project total entered against a single well, not a cost per well.
• Well type is grouped into oil, gas, injection or disposal, and other or unknown.
• When three or more wells of one recipient report the identical cost, the value is treated as one contract or lease average. Those wells stay in the data, but count as one independent observation in grading and in the index.
• Reported cost is used as reported. It is not adjusted for scope, and what it includes varies by agency.
• Names of individuals in the source (for example, plugging witnesses) are never published.

Estimates

An estimate is the median and the 10th to 90th percentile range of reported cost per well for a defined group: a state, a well type within a state, a county, a year. The estimator returns the most specific group that carries a grade of A or B, in this order: county and well type, county, state depth band, state and well type, state.

Grading

Every published figure carries a grade based on held-out error and sample size, never on in-sample fit. For each group, the wells are split at random 80/20 five times, keeping wells that share one contract average on the same side. The median and range are computed on the 80% and tested on the 20%.

Held-out error is the median absolute percentage gap between the group median and each held-out well. The range target is 80% of held-out wells. The thresholds were fixed before the grades were computed.

Plugging Cost Index

The index is estimated by regressing the logarithm of reported cost on location, well type and period. Location is the county where the county has 20 or more wells, otherwise the rest of its state. Each independent cost observation has equal weight. The exponentiated period effects, scaled so that 2023 averages 100, are the index.

A period is graded by fitting the model on 80% of observations and measuring the gap between the fitted level and the median of the held-out 20% in that period, averaged over five splits: grade A needs 200+ independent observations in the period and a level error of 5% or less; grade B needs 60+ and 10% or less. National periods also need at least 5 states. Ungraded periods are left blank; the index is not interpolated across them.

Update schedule

Sources are re-collected monthly. Estimates and the index are recomputed after each collection and published after review. The Plugging Cost Report is issued quarterly.

Version history