The Public-Firm Spillover: What IPOs Change
The public-firm case asks whether an improvement in an incumbent industry's information environment is followed by more business formation outside the reporting firm. The key comparison is between a broad association—industries with more public firms—and a timed event: an IPO that suddenly adds filings, coverage, and attention in a particular place. Reading the evidence through that distinction lets us judge whether public disclosure appears relevant to nearby entrepreneurs without pretending that a county-level estimate identifies any one founder's decision.
An IPO is an especially visible form of disclosure. A private company becoming public must release information about its business, risks, financial condition, and prospects. The event also attracts analyst reports, newspaper coverage, and discussion among local economic actors. The new information is not necessarily novel in every detail, and an IPO may coincide with other changes. It nevertheless provides a before-and-after marker that is sharper than simply counting how many public firms happen to exist in an industry.
From public-firm presence to an IPO comparison
The BFI study begins with a relatively broad measure called public-firm presence. It is the proportion of firms in an industry that are public, ranked into yearly quintiles. The idea is not that the number of listed companies mechanically creates entrepreneurs. Rather, a higher public-firm presence suggests a thicker information environment: more formal disclosures, more material for analysts and the business press to interpret, and more visible evidence about how an industry works.
Across county–industry–year observations, a one-quintile increase in public-firm presence is associated with roughly an 8–10% increase in new establishment births after accounting for county, industry, and year fixed effects and county-level controls. The study translates the relationship into about one additional establishment birth per year in the industry on average. In its small-establishment analysis, moving from the bottom to the top presence quintile is associated with a 12% increase among establishments with zero to five employees, and the relationship is reported as more than 40% greater for small firms than for larger establishments.
Those numbers establish a pattern, not yet a mechanism. Public-firm presence may be correlated with features that also attract new businesses: specialized workers, infrastructure, customers, capital, or a mature supplier network. Fixed effects remove stable differences among counties and industries, while controls absorb some changing population and income differences, but an association cannot show that a particular disclosure caused a particular establishment to form. The outcome also matters. A Census establishment birth is a new establishment with paid employees; it can be a branch opened by an existing company rather than a newly founded independent firm.
The size pattern is suggestive but should be read carefully. Small establishments are closer to the margin at which a person decides whether to create a business, so they may be more responsive to information that lowers search costs. A large corporation may already possess research staff, industry contacts, and financing; a small prospective entrant may benefit more from a newly public signal. Yet the result does not identify the founder, the source they used, or whether the small establishment survived. It indicates where the measured relationship is stronger, not why every small firm entered.
The IPO analysis addresses part of that problem by using a change in information availability over time. The study compares county–industry observations where an IPO has occurred with comparable observations without a recent IPO, before and after the event. A county remains in the comparison even if it differs from an IPO county in lasting ways, because county fixed effects absorb those stable differences. Year and industry effects absorb common shocks, and county-specific trends are used in some specifications to account for different underlying trajectories. The coefficient of interest is the additional change after the IPO in the affected place and, in some tests, the relevant industry.
This design is informative because the event has a recognizable public-information component. An IPO brings required corporate filings, but its local effect can also travel through newspapers, analysts, customers, suppliers, employees, and people who already understand the industry. Those channels are difficult to observe separately. That is not a reason to ignore them; it is a reason to state the treatment accurately. The study is not measuring “reading the prospectus” by a known group of entrepreneurs. It is measuring exposure to a broader information environment that the IPO makes more prominent.
The comparison asks a counterfactual question in plain terms: absent the IPO, would business formation in the affected county and industry have moved like formation in the comparison observations? County fixed effects help compare each place with itself rather than with a wholly different economy. Industry and time effects remove common changes in the market. Local trends address some pre-existing differences in trajectory. The credibility of the result therefore rests on a parallel-trends assumption, not on the idea that IPO counties and non-IPO counties were identical before the event.
That assumption is substantive. IPO counties are observably different in the source data: counties with IPO activity tend to have more establishments, higher income per capita, and larger populations. Controls and fixed effects address some of that selection, but they cannot guarantee that all factors affecting entry are captured. The most useful evidence is consequently the combination of design and pattern: no obvious divergence before the event, a post-event change, and a response concentrated in the industry most likely to use the new information.
The result is a sizeable local increase in establishment births after an IPO, reported as roughly 11–22% depending on the specification. The more revealing comparison is across industries. When the IPO firm and the new establishment are in the same industry, the increase is about 30% in the reported analysis. When they are in different industries, the effect is much smaller—about 2%—and statistically insignificant in the corresponding comparison. If an IPO simply marked a general improvement in the county, we would expect a less sharply industry-specific pattern. The stronger same-industry response is consistent with entrepreneurs using information that helps them understand their own prospective market.
The geography matters for a related reason. Formal disclosure is public in a legal sense, but the ability to interpret and act on it is not evenly distributed. A local entrepreneur may know which suppliers are constrained, which customers are dissatisfied, or which workers have relevant skills. Someone farther away may face the same filing but lack the relationships that make its implications legible. The BFI study therefore expects soft information to reinforce hard information near the public firm. Proximity is not proof that local entrepreneurs used the disclosure, but it is a prediction that follows from the way public facts become usable knowledge.
Reading the response as a sequence, not a single result
The public-firm evidence becomes more intelligible when its outcomes are arranged in time. An entrepreneur may first encounter a new report, then search for more information, then consider entrepreneurship, and only later register or operate a business. The study measures several points along that path. Its measure of entrepreneurial interest uses searches such as “start a business,” “how to incorporate,” and “small business loan.” After an IPO, an affected area becomes about 1.5 percentage points more likely to be in the top quartile of entrepreneurship-related search activity. Against a 25% baseline probability, the paper describes that as about a 6% increase. This is evidence of increased consideration, not evidence that a searcher founded a firm.
EDGAR downloads provide a different signal. After an IPO, downloads of public-company reports in the county increase by roughly the mid-20% range in the reported specifications. The study also finds a positive relation between downloads and new business registrations: its economic interpretation is that doubling EDGAR searches is associated with about a 1% increase in new business formation. The result does not tell us who downloaded the files. Downloads could include investors, employees, journalists, competitors, researchers, or entrepreneurs. Their value is therefore as a measure of information acquisition in the local environment, not as a direct record of entrepreneurial reading behavior.
The outcome closest to completed entry is the registration measure from the Startup Cartography Project. It counts newly registered corporations, limited liability companies, and partnerships at the county level. In the main IPO specifications, post-IPO registrations increase by about 4–9%, with positive associations also appearing four to six quarters later; the paper summarizes related specifications as roughly 4–10% and several additional registrations per quarter. The exact translation varies with the sample and outcome transformation, so the defensible conclusion is a persistent increase in measured registrations, not one universal effect size.
Registrations improve on establishment births in one respect: they are more directly tied to the creation of a newly registered entity and are less likely to be only a new branch of an old company. They lose information in another respect. The data do not identify the firm's industry, so they cannot show whether the registrant entered the IPO firm's market. They also do not cover every possible business form. A sole proprietor may operate without registering in the same way. That is why the two measures are complementary rather than interchangeable.
Several comparisons test whether the information interpretation fits the pattern. The first contrasts private-to-public IPOs with spinoffs from already-public parents. A spinoff can create new disclosures, but the parent may already have been reporting information about the subsidiary. A private company going public is expected to reveal more that was previously unavailable. The reported post-IPO registration coefficient is larger for private IPOs—about 0.27—than for public-company spinoffs—about 0.18. Both are positive, but the difference is consistent with a larger increment of new information producing a larger entry response.
The second comparison concerns uncertainty. The BFI study measures county-level wage-growth volatility as an indicator of economic uncertainty and finds that the positive IPO effect is more pronounced in locations with higher ex-ante uncertainty. This is the direction predicted by the information model: when potential returns are harder to assess, a credible new signal can have more value at the margin. The result should not be read as evidence that uncertain counties are inherently better places to start businesses. The underlying uncertainty may discourage entry; the claim is that information can matter more where the original decision is difficult.
The third comparison asks whether the event is followed by more economic growth in a broad sense. The study finds no significant post-IPO increase in average weekly wages, while entrepreneurship measures rise. That weakens a simple explanation in which an IPO merely arrives during a general local boom that lifts all outcomes. It does not eliminate every alternative. Wages are an imperfect measure of local growth, and an IPO can have effects not captured by the selected controls. The value of the test is narrower: it makes a pure county-wide expansion story less complete than an explanation involving disclosure and entrepreneurial response.
The same caution applies to the event-time evidence. The study reports parallel pre-IPO trends in business registrations, which is important because a pre-existing divergence would undermine the before-and-after comparison. It also uses county and industry effects, time-varying population and income controls, and county-specific trends in its stricter specifications. These choices make it less likely that the post-IPO change is simply a continuation of a visible pre-event trajectory. Still, the design does not randomize the location or timing of IPOs, and no regression can observe the private information, motives, and constraints of each new business.
The most useful conclusion is therefore layered. Public-firm presence is associated with more establishment births, but that broad measure is vulnerable to selection and outcome ambiguity. IPOs provide a sharper timing contrast, and the local, same-industry response is consistent with relevant information spillovers. Search activity and EDGAR downloads move in the direction expected before or alongside formation, while private IPOs and high-uncertainty locations show stronger responses. Registration results extend the pattern beyond possible branch expansion. Taken together, these findings make the information channel plausible and more difficult to dismiss as a generic boom.
They do not prove that disclosure is the sole cause, that every new establishment is an independent entrepreneurial venture, or that the resulting businesses are successful. An IPO can alter competition, labor demand, financing, or local expectations in ways not fully separable from information. A public report can also reveal a threat rather than an opportunity. The evidence supports a disciplined interpretation: when a new disclosure is closely tied to an industry and place, and when information acquisition and entry respond after the event, it is reasonable to investigate an information-spillover explanation. The next question is whether disclosure can do more than inform outsiders—whether it can also change the incumbent's own production and leave a competitive gap behind.