The Number of New Businesses Depends on What You Count
This series examines a specific disagreement in the way the United States measures business activity: the gap between the Bureau of Labor Statistics’ Business Employment Dynamics (BDM) and the Census Bureau’s Business Dynamics Statistics (BDS). The practical question is whether a reported rise in “new businesses” means that more independent firms have been created. To answer it, readers need to distinguish an establishment from a firm, a new operating location from a new legal organization, and a change in jobs from a change in the number of places where work occurs.
That distinction matters because the two official statistical views can describe the same private-sector economy quite differently. The Census working paper at the center of this series reports that, for 2023, the BLS-based BDM counted roughly 9.2 million private-sector employer establishments, while the Census-based BDS counted roughly 7.4 to 7.5 million. The paper presents the difference as roughly 1.6 million establishments, with the exact rounded presentation varying across its sections. Yet the BDS reported more total employment. A larger number of counted places did not translate into a larger counted workforce.
This is not a minor technical disagreement. It changes the apparent balance between tiny work sites and large employment centers, the measured prevalence of multi-establishment firms, and the interpretation of business dynamism. Before calling the pattern a startup boom, we need to know what the word “new” refers to and which statistical frame produced the count.
The subject is the gap between two official views
The BDM and BDS are not two headlines drawn from one perfectly shared register. BDM is built from the Quarterly Census of Employment and Wages, a state-federal system based largely on unemployment-insurance records. BDS is derived from the Census Bureau’s Longitudinal Business Database, whose Business Register relies primarily on Internal Revenue Service records supplemented by surveys, censuses, and other administrative sources. Both systems connect records over time, but they do not begin with exactly the same records or identify organizational relationships in exactly the same way.
That makes the subject of this series a relationship, not a single number. We are studying how two frames count establishments and associate them with firms, why their totals diverge, and what follows when one frame sees more small units while the other sees more large ones. The question is not which agency has made a simple arithmetic mistake. A frame can be useful for the purpose it was designed to serve while still producing a different picture from another frame with a different coverage boundary and linkage practice.
The distinction also prevents a common mistake in public discussion. “Business,” “company,” “firm,” “establishment,” and “startup” are often used as if they were interchangeable. In statistical work, they are not. The BLS defines an establishment as an economic unit that produces goods or services, usually at a single location and mainly through one activity. A firm is a legal business that may contain one establishment or several. A chain, a hospital system, or a professional-services organization can therefore be one firm with many establishments.
The establishment is the unit used to measure gross job flows in BDM. That choice is valuable: it can show where jobs are being added or lost, including changes at individual locations. But it also means that a new location can enter the data even when the parent organization is old. Counting operating units answers one question about economic change. It does not automatically answer the question of how many independent firms were born.
Consider three events that can all look like an opening at the local-unit level. A person may form a new company and hire workers at a new site. An established chain may open a branch in a new neighborhood. Or an existing organization may change how it reports a work site so that positive employment appears under a newly tracked unit. These events are economically different: the first suggests independent entry, the second suggests expansion by an incumbent, and the third may reflect a change in observation or organization. A count of openings can contain all three unless the underlying firm relationship is examined.
This is why the word “new” needs an object. A new establishment can mean new activity at a place. A new firm means a new legal or organizational business. New jobs mean additional employment, which can be created at an old site, a new branch, or a new company. Those measures can move together, but the definitions do not require them to. The value of the BDM is not weakened by this distinction; it is clarified. It is designed to show job flows through operating units, while a firm-entry question asks for another layer of evidence.
The diagram captures the discipline this series will apply. A headline is not yet an interpretation. It is the output of choices about the unit being counted, the records used to observe it, and the rules used to classify its age and size. If any of those choices changes, the meaning of the headline can change with it.
An establishment is a place of activity, not necessarily a company
The BDM’s annual measures make the distinction concrete. An opening is a unit with positive employment in March of the current year after reporting zero employment in March of the previous year. An expansion is a unit that had positive employment in both periods and increased its employment. A closing is a previously active unit with no employment in the current period, while a contraction is an active unit whose employment fell.
Gross job gains combine jobs at opening and expanding units. Gross job losses combine jobs at closing and contracting units. Net employment change is the difference between those two totals. These measures are useful precisely because they separate the movement of jobs into its underlying components. A net increase can conceal many simultaneous openings and closings; gross flows reveal that churn.
But none of those definitions says that an opening is a newly independent firm. An existing organization can open another location. A firm can reorganize reporting relationships. A service provider can add a work site without creating a new legal entity. The establishment measure records a change in the operating-unit landscape; whether that change is also firm entry requires an additional organizational judgment.
Age makes the same point in another way. BLS defines establishment age as the time since an establishment first reported positive employment. Firm age is based on the age of the oldest establishment within the firm. A firm with a long operating history can therefore contain an establishment that is less than one year old. The local unit is young; the organization is not. Calling both observations “a young startup” would erase the relationship the data are designed to preserve.
Size classifications add another layer. BLS can report age and size using establishment characteristics or firm characteristics. It also uses different size concepts, including employment in the prior March and average employment across the prior and current March. Those distinctions are not decorative metadata. They determine whether the reader is learning about the size of one location, the combined size of an owner’s establishments, or a transition between two reference dates.
The reference period matters as well. The annual comparison follows employment reported in March of one year against March of the next, so it is not a live count of every business that opened or closed on a particular day. The annual measure is therefore a defined comparison, not a raw diary of every event. Keeping that reference period visible helps prevent a reader from treating an annual age or opening category as a precise timestamp.
Even the rates for the youngest age group require care. BLS notes that a rate for a unit less than one year old can be 200% or 0% because the prior employment level is zero and the denominator is an average of prior and current employment. The unusual value is a consequence of the formula and the definition of an opening, not evidence that young establishments literally doubled employment in a conventional sense. A statistic cannot be interpreted separately from its denominator.
This vocabulary gives the startup question a sharper form. Instead of asking only, “Did the number of new businesses rise?” we can ask: Did the number of newly active establishments rise? Did employment at openings rise? Did the number of independent firms rise? Did older firms add more locations? The answers may differ, and the BDM’s definitions do not promise that they will move together.
Why the distinction changes the question
The two frames matter because measurement is not neutral about organizational scale. If one system records many small locations that another system omits or consolidates, the first can make the economy look more fragmented and the second can make it look more concentrated. Neither description can be interpreted responsibly without asking how each system constructs its business universe.
The sources support a two-part way to investigate the disagreement. First, ask about coverage: are some employers present in one frame but absent from the other? Second, ask about organizational linkage: when both frames contain the same firm, do they assign it the same number of establishments and the same single-unit or multi-unit status? The later evidence will show that both mechanisms matter, and that different industries illustrate them in different ways.
These questions also explain why the two frames should not be ranked by their totals alone. A frame that includes more small units may be more revealing for one question and less comparable for another. A frame that consolidates activity into larger units may make firm-level employment easier to interpret while obscuring the number of local work sites. The relevant test is not simply which count is higher; it is whether the frame’s coverage and linkage rules fit the claim a reader wants to make.
That is why this series will not treat the phrase “startup boom” as self-explanatory. A rise in establishment openings can be economically important even if it is not a rise in independent firm births. New locations can bring services closer to households, spread employment across communities, or show that established organizations are scaling. Conversely, a frame that counts fewer locations may still capture more employment concentrated in larger sites. These are different forms of dynamism, not merely competing labels.
The defensible starting conclusion is therefore modest but consequential: the evidence can tell us how many establishments a particular frame observes, how jobs move among those units, and how the units are linked to firms. It cannot, by itself, turn every opening into a new company. To make that judgment, we must first follow the lifecycle of an establishment and keep its age separate from the age of the firm that owns it.