What a Founding Team Builds Together
The subject here is the mechanism behind founding-team stability: how a group that enters a venture together can create capabilities that later workers may not immediately reproduce. The useful distinction is between individual human capital—skills and experience a person brings—and organization-specific capital created through collaboration. Tracing that distinction lets us see why attrition can matter to revenue, productivity, and survival, while also keeping clear which parts of the story are a model, which are measured proxies, and which remain correlations.
Formation turns a group into an organization
The U.S. Census study begins with an idea that is easy to miss in a standard startup model. Entry is not only the moment when a business registers or sells its first product. It is also a formation period in which the founding team spends time and resources creating an organization. The team develops a business vision, learns how to divide work, builds relationships, and turns an uncertain opportunity into a way of operating. The paper represents this investment as organizational capital.
That capital is different from a résumé. A résumé can describe an engineer’s technical skill or a manager’s previous experience. It does not fully describe the particular knowledge that emerges when those people work together in one venture. They may discover who handles an ambiguous customer request, which supplier can be trusted when a deadline moves, or how to decide when the available data is incomplete. They may also develop a business model whose details are distributed across several people rather than written in one place. The organization becomes more capable through the interaction, not merely through the sum of its members’ prior achievements.
The model’s sequence is therefore important. A team enters and invests in formation. It receives some team-specific match quality—how well its members can work together to develop the organization. Before or during production, some members may leave. If the knowledge created during formation is not embodied in the people who remain, that departure need not reduce the firm’s productive capability very much. But if the knowledge is embodied in the team, attrition creates knowledge decay. The firm has not merely lost a worker; it has lost part of the mechanism through which its organization-specific capital is used.
The paper makes this dependence explicit with a parameter for knowledge decay from founding-team attrition. When that decay is zero, the model collapses toward a familiar view: the team pays the cost of entry, obtains a productivity draw, and later workers can implement the organization without much loss. When decay is positive, the share of the founding team that remains affects productivity. The authors then derive implications for employment, revenue, labor productivity, and exit. Lower available team capability can make it harder to cover fixed operating costs, so a firm may employ fewer people, produce less revenue per worker, or cease operating altogether.
This is a theoretical mechanism, not a direct measurement of a hidden “organizational-capital stock.” The value of the model is that it connects a personnel change to a sequence of business outcomes. It also gives a reason to expect time to matter. As a team works together, the amount of context that sits in shared routines and relationships can grow. A later departure may remove more accumulated coordination than an early departure, even if the person’s formal role looks similar. The claim is not that every venture follows this path or that all knowledge becomes irreplaceable. It is that the effect of attrition depends on how much the venture’s productive capacity is tied to the team that formed it.
The model also leaves room for imperfect substitution. The Census authors allow non-founding workers to contribute to production, but the model treats them as potentially less effective at implementing the organization-specific capital than founding-team members. That is not a claim that later hires are less capable in general. It is a claim about fit with a particular organization. A new employee may be excellent and still need to learn the relationships, norms, and decisions that the original group acquired through formation.
The fixed-cost part of the model makes the stakes clearer. A young firm has overhead that must be covered before it can operate profitably. If attrition lowers the productivity of the team and reduces the amount of work the organization can coordinate, revenue can fall faster than the firm can adjust its costs. The organization may then face a choice between operating at a smaller scale and exiting. In that sense, team-specific capital does not matter only because it makes work pleasant or efficient. It can change whether the venture’s operating model remains viable.
Consider a hypothetical venture whose first customers came through relationships held by two early members. One understands the customers’ underlying workflow; the other knows how to turn those conversations into a reliable delivery process. Neither fact may appear in a job title or an earnings record. If one person leaves, a replacement can be trained, but the remaining member must explain context while also running the business. If both the relationship and the process were documented and practiced by several people, the departure may be easier to absorb. The example illustrates the mechanism’s key condition: vulnerability depends on how much capability is shared, embodied, and transferable, not simply on whether a vacancy exists.
Time therefore matters through accumulated interaction, not through the calendar by itself. A team that has worked together longer may have more valuable shared knowledge, but it may also have built better documentation or redundancy. The model points to the first possibility; it does not tell us how much each venture invests in transfer and backup. That is why “stability matters more over time” should be treated as a claim about the growth of interdependence and embedded knowledge, not as an automatic law of firm age.
The diagram should not be read as a claim that departure always causes decline. It is the model’s conditional path: loss matters when organization-specific knowledge is embodied in the departing team member and when the remaining organization cannot immediately reproduce it. A planned transition could transfer some knowledge. A larger team could provide more redundancy. A new hire could add a capability that the original team lacked. The supplied studies do not estimate all of those alternatives. They establish why the question is economically meaningful and then test particular patterns of stability and forced separation.
How the Census study makes a team visible
To test the mechanism, the Census researchers had to decide what counts as a founding team in administrative data. Their definition is deliberately broader than “people who own the company.” For the U.S. employer startups in their sample, the founding team includes everyone with positive unemployment-insurance-covered earnings at the firm during its first year of operation. The study supplements that group with the owners of sole proprietorships, who generally do not pay themselves wages and therefore would otherwise be missing from payroll data.
This definition changes the object being studied. It includes early employees who helped turn the initial venture into an operating organization. It also avoids assuming that every important contributor held an ownership stake. For corporations, the researchers identify key founding-team members, or KFT, as the three highest earners in the first year. For sole proprietorships, the KFT group includes the business owner and the two highest-earning workers. The authors use this as an approximation for people likely to hold key leadership positions, not as a verified list of founders. A top earner may not be an original founder, and some active founders may not appear in the data.
The distinction between KFT and non-KFT is valuable precisely because it tests a narrower and a broader story. If only the loss of a top owner mattered, the evidence could be explained mainly by founder control or irreplaceable leadership. If the loss of a non-KFT member also affected outcomes, then the early team’s contribution is more distributed. That does not mean every member contributes equally. It means the boundary of relevant organizational capital may extend beyond the people most visible in the company’s public story.
The sample has meaningful scope limits. It covers non-farm employer startups tracked through the Longitudinal Business Database and linked employee–employer records from the Longitudinal Employer–Household Dynamics data. Startups are defined as firms whose oldest establishment has age zero. The analysis covers 1990 to 2015 and includes more than six million startups and more than 72 million founding-team members across the states represented in the data; state coverage becomes nearly complete by 2000. Partnerships are excluded because the available records do not capture their active or managing owners who report pass-through income through Schedule K-1. The results therefore describe a broad U.S. employer-startup population, but not every legal form or every kind of new venture.
The researchers also need a way to compare the human capital of people who join different teams. They use each member’s most recent full-quarter earnings before joining the startup as a proxy. Prior earnings can capture some combination of skill, experience, and labor-market value, and it is measured before the person joins the new firm. That timing makes it useful for descriptive comparisons and for studying heterogeneous effects. But it remains a proxy. It does not directly measure compatibility, trust, customer relationships, tacit knowledge, or the quality of the business idea that attracted the person.
With those definitions in place, the descriptive patterns are coherent. Firms with higher-human-capital key founding teams tend to start larger. Higher-human-capital key and non-key groups are associated with faster employment growth conditional on survival, and productivity growth rises with human capital for the founding team and both subgroups. Firms with higher-human-capital teams are also less likely to exit across most of the distribution. At the same time, attrition is substantial. The human capital of the members who remain rises on average because departures are concentrated toward the lower end of the initial human-capital distribution. The study also finds positive assortative matching: key and non-key members with higher human capital tend to be found together more often than a random pairing would suggest.
The first-year boundary is useful because it captures the people present when the venture becomes an operating firm, but it is not a perfect map of the instant of founding. Some people identified as founding-team members may have joined soon after the legal or economic start, and some founders may be absent from the records. That is a reason to read the study as evidence about the early operating team, not as a definitive census of entrepreneurial identity. It also explains why the broader definition is analytically valuable: organizational capital can be created during the first year by people whose formal status is employee rather than owner.
Positive assortative matching adds another complication. If high-human-capital key members tend to work with high-human-capital non-key members, then a team’s composition is partly bundled. A firm may benefit from the combination, not just from any one person’s prior earnings. Likewise, when the remaining team looks stronger after attrition, that can reflect selective departure rather than a productivity gain caused by losing people. The descriptive data reveal these patterns and help locate the mechanisms worth testing, but they do not by themselves identify which part of the team’s advantage came from selection, collaboration, retention, or the underlying business.
These observations help explain how young firms sort into different trajectories, but they do not yet prove that stability produced those trajectories. A capable person may join a venture with a better technology or a stronger market opportunity. A firm with better early performance may be more able to retain its team. Industry, location, timing, and the quality of the underlying business can affect both team composition and outcomes. Even industry-by-year controls and initial-condition controls cannot turn a descriptive relationship into an experiment.
That is why the distinction between mechanism and measurement matters. The organizational-capital model says that shared formation work may become embodied in a team. The administrative data can observe earnings, employment, revenue, productivity, survival, and whether members remain with the firm. It cannot directly observe every customer relationship, norm, or piece of shared judgment. The measured patterns are therefore evidence about the consequences and correlates of team composition, not a complete inventory of what the team has built.
The next analytical step is consequently causal: instead of asking only whether stable or high-human-capital teams perform better, ask what happens when a team member is separated from a young firm under a shock that is less likely to have been chosen by the firm. That comparison is necessary before the descriptive patterns can support a stronger claim about stability.