Geography

DTG stores data for three levels of geography: the nation, states, and counties.

This page describes the areas that exist, how area clusters are built, and what aggregation does to each measure.

The areas DTG reports

Counties and county-equivalents in the fifty states and the District of Columbia. County-equivalents include parishes, boroughs and census areas, independent cities, and Connecticut’s planning regions.

States, the fifty plus the District of Columbia.

The nation, excluding territories. A DTG national figure is the sum of the fifty states and the District of Columbia. Employment sums to within rounding; earnings and establishment counts sum very closely but not always exactly, and the differences are largest in the smallest industries.

A per-state unallocated area. Each state and the District of Columbia carries a single additional area, except Utah, which has none. It holds employment and wages reported to that state but not attributable to a specific county. It behaves as an ordinary area in the data and is included in state and national totals, but it has no geographic extent and will not appear on a map.

County definitions follow the most recent QCEW release in the vintage being queried, and the full series is restated onto them. That restatement is described on the industry data page.

Area clusters

An area cluster is a named set of areas, built by the user and aggregated at query time. A cluster may cross state lines, and may also include a state’s unallocated area.

Metropolitan areas can be selected directly rather than assembled county by county. Those definitions follow the delineation currently in use by the Occupational Employment and Wage Statistics program. A metropolitan area that a later delineation retires stays visible in the picker, greyed and not selectable, and says which area its counties went to; it is never replaced by its successor without the user choosing it.

What aggregation does to each measure

Employment, establishments and wages are summed across the counties in the cluster.

Average pay is computed as the cluster’s total wages divided by the cluster’s total employment.

Disclosure codes are derived from the components. If every county value of an industry in an area cluster was originally published by BLS, the value is marked as disclosed. If any component was originally suppressed, the value is marked as estimated.

Reliability scores are combined as an employment-weighted average of the component counties’ scores. An aggregated area cluster’s score therefore describes the cluster’s total, not its weakest component: a cluster dominated by large, fully disclosed counties will score close to 1 even when one small county within it is poorly constrained. If a particular county in a cluster matters to the analysis, read that county’s own score rather than inferring it from the cluster’s.

Reliability scores are served to two decimal places. A cluster containing a small share of estimated values can therefore display a score of 1.00 while carrying disclosure code 1. A score of 1.00 on a cluster does not mean the cluster was fully published. The disclosure code is the authoritative signal on that question.

Location quotient is computed from the aggregated values and carries no provenance of its own. See derived measures.

Stored areas and assembled clusters are not the same thing

A state in DTG is DTG’s value for that state: the value BLS published where QCEW disclosed it, and DTG’s estimate where QCEW withheld it. A cluster assembled from that state’s counties is the sum of DTG’s county values, published or estimated. These are different things, and they will not always agree.

Two differences are worth expecting.

More values will be estimated. BLS suppresses far more at county level than at state level, so a state assembled from counties will carry disclosure code 1 in many places where the stored state value carries code 0. This is not a loss of accuracy in the assembled figure; it reflects where the published record is thinner.

The total will be lower unless the unallocated area is included. Employment reported to a state but not attributable to a county sits in that state’s unallocated area. A cluster built from counties alone will fall short of the stored state value by that amount.

When the question is about a state, use the state. Assemble from counties when the cluster genuinely is a set of counties.