The Headline Gap: More Establishments, Fewer Employees

The BDM and BDS frames have now been separated conceptually: one is rooted in QCEW and unemployment-insurance records, the other in the Census Business Register and LBD, and both can differ in coverage and organizational linkage. This checkpoint connects those ideas to the published aggregates. The practical judgment is simple but easy to miss: a frame can report more establishments while reporting fewer employees, because the number of places and the number of people assigned to those places are different dimensions of the business landscape.

The figures also show why the gap should not be treated as a one-year curiosity. The two series were close in the early 2000s, then diverged. By 2023 they produced sharply different establishment totals and opposite patterns at the extremes of the size distribution. The numbers establish a substantial measurement difference. They do not, by themselves, identify whether coverage, linkage, or both produced it.

From rough agreement to a widening gap

The Census paper describes a long-run change in the relationship between the two published series. In the early 2000s, the BDS and BDM counts were close to one another at roughly 6.5 million private-sector establishments. That rough agreement did not continue. During the later 2000s and the 2010s, the BDS count grew more slowly, while the BDM count indicated a much larger increase in the number of establishments.

The comparison is important precisely because the divergence is not presented as a sudden statistical break in one isolated year. The paper describes a widening gap that emerged after the two frames had been near one another. That pattern changes the question an analyst should ask. It is not enough to say that one 2023 count is higher. One must ask what changed in the frames, the business population, or the organizational structures they recorded as the series moved apart.

Historical agreement is useful as a baseline, but it should not be mistaken for proof that the two frames have always measured the same population in the same way. Their totals may have been close while differences in industry coverage or organizational assignment offset one another. Later divergence may reflect an expanding difference in the business population, a change in how organizations are represented, or a combination. The time path tells us that the disagreement accumulated; it does not tell us which mechanism accumulated faster.

For March 2023, the paper reports approximately 9.2 million private-sector employer establishments in BDM/QCEW and approximately 7.4 to 7.5 million in BDS/LBD. Its abstract describes the difference as roughly 1.6 million, while the introduction and discussion of the published figure use about 9.2 million versus about 7.4 million; the rounded presentations are not identical. The responsible conclusion is therefore that the gap was on the order of 1.6 to 1.8 million establishments, not that one artificially precise decimal resolves the source’s own rounding difference.

flowchart LR A["Early 2000s\nabout 6.5 million each"] --> B["Late 2000s onward\npaths diverge"] B --> C["March 2023\nBDM about 9.2 million"] B --> D["March 2023\nBDS about 7.4–7.5 million"] C --> E["1–19 employees\nBDM 7.0902 million"] D --> F["1–19 employees\nBDS 5.5803 million"] C --> G["500+ employees\nBDM 17.5 thousand"] D --> H["500+ employees\nBDS 22.5 thousand"]

The timeline and the totals should be read together.