The Shock: How a Founder’s Loss Travels Through a Team
This chapter follows the timing and specificity of the employee response to sudden founder loss. The key question is not only whether turnover rises, but when it rises, for whom, and whether the same pattern appears after the death of a non-founder employee. Those distinctions help us judge whether the founder is functioning as an organizational and relational center, while keeping grief and other alternative explanations in view.
The strongest Census comparison treats premature founder death as a sudden shock. It finds a post-loss increase in turnover relative to similar startups, with no detectable divergence in the measured turnover paths before the event. But the timing carries additional information: the increase is strongest in the first two years and then declines. A shock that is immediate and concentrated is different from a slow deterioration that happens to culminate in a founder’s exit.
Why timing changes the interpretation
An event study traces an outcome around a focal date rather than looking only at an average before-and-after difference. Here, the focal date is the year of the founder’s premature death. Years before the event serve as a check on whether the treated and comparison startups were already moving apart. Years after the event show whether any difference appears immediately, persists, or fades.
The pre-loss estimates are statistically indistinguishable from zero. That does not mean the affected and comparison startups were identical in every way. It means that, on the measured turnover outcome, there is no visible evidence that their trajectories were separating before the founder loss. This matters because an already widening gap would make a later difference easier to attribute to an underlying decline rather than to the event itself.
After the loss, the effect appears quickly. The researchers report that the increase in employee turnover is strongest in the first two post-treatment years and then diminishes. The pattern is consistent with an event that changes employees’ immediate assessment of the workplace: who will make decisions, whether the mission still holds, whether relationships will survive, and whether the startup remains a place where their own skills fit.
The decline after the first two years should not be treated as proof that the organization has fully recovered. Several processes could produce it. Some employees may adapt to a successor or to a new way of working. The people most attached to the founder may leave early, reducing the pool that remains exposed. New employees who join after the loss may never have had the opportunity to build a founder-based relationship. Or the startup may shrink, close, or change in ways that alter who is still observed. The event study establishes the timing of the response; it does not by itself select among these mechanisms.
That timing also helps separate founder loss from a general upward trend in startup turnover. If turnover simply rose as firms aged, we would expect a more gradual pattern that also appeared in the comparison group. A concentrated post-loss movement, after similar pre-event paths, is more consistent with a discrete organizational shock. It does not eliminate every competing explanation, but it changes which explanations are plausible.
Timing also changes the population being observed. Employees present before the loss have had an opportunity to build a relationship with the founder, while employees hired afterward do not. A post-loss workforce can therefore look more stable partly because the most exposed employees have already departed and partly because later hires entered under a different leadership arrangement. The declining estimate is not a clean measure of healing; it is a mixture of adaptation, selection, replacement, and the startup’s survival.
This is why the event-study should be read as a sequence rather than as a single number. The pre-event years ask whether treated and comparison firms were moving similarly. The first post-event years ask whether the founder loss changed the path. Later years ask whether the difference persists, fades, or becomes harder to interpret as the workforce and organization change. Each stage answers a different question about the same shock.
Why the loss of a founder may travel farther
The most restrained interpretation is that a founder occupies a position with several overlapping forms of centrality. They may be the person who embodies the mission to employees, the source of difficult decisions, the bridge between groups, or the holder of relationships with investors, customers, and early recruits. The administrative records do not measure each of these functions. They do, however, show a response that is larger after founder loss than after the comparison loss of a non-founder employee.
This is different from saying that every founder is the best manager, or that every employee has a personal bond with them. Organizational centrality can exist without universal affection. A founder may be a bottleneck, a translator, or a coordinating reference point even for employees who find the person frustrating. When that reference point disappears, people must reassess how work gets done and whether the organization they joined still exists in a meaningful sense.
The change can travel through several layers. At the individual level, an employee may wonder whether their role still fits the new direction. At the team level, colleagues may no longer know who can settle an ambiguous priority or protect a long-standing norm. At the organizational level, customers, investors, and prospective hires may interpret the change differently, adding uncertainty to the internal transition. These are plausible pathways, not variables directly observed in the Census design. Their value is to explain why a founder’s departure could affect employees who did not work with the founder every day.
The comparison with non-founder deaths is designed to test a particularly important alternative: perhaps any sudden death creates enough grief and disruption to raise turnover. The authors examine startups that lose a non-founder employee who joined after the first year of the firm and compare them with similar startups. The resulting sample contains roughly 17,000 startups. In the main specifications, the loss of a non-founder does not produce a statistically or economically meaningful subsequent increase in employee turnover.
The event-study comparison is not perfectly flat. There is a relatively small increase—about 7% relative to the control mean—in the year of the non-founder loss, but no meaningful effect in the subsequent years. Against the larger and more persistent founder-loss pattern, this is evidence that the organizational position of the person who leaves matters. The result narrows the claim that grief alone explains the founder estimate: if any sudden death had the same effect, the non-founder comparison should show a similar cascade.
Still, the placebo does not measure emotion, and it does not make the founder and non-founder events interchangeable. A non-founder who joined after the first year may have had a different role, tenure, network, or degree of authority. The comparison therefore supports a founder-specific component without identifying exactly what that component is. It cannot rule out that employees experience grief after either event, only that the measurable turnover response appears uniquely large and durable after founder loss in this setting.
The comparison is strongest as a statement about organizational position. A founder is present at the firm’s beginning, when roles, norms, and stories are being formed. Their relationships may be woven into the startup’s hiring, strategy, and external credibility. A later employee can be highly valued without occupying that same origin point. The result does not say that non-founders are unimportant; it says that importance has different forms, and the form associated with founding appears to have a distinctive retention consequence when suddenly removed.
That distinction also prevents a misleading interpretation of the small non-founder effect. The study does not show that employees are emotionally unaffected by losing a colleague. It shows that the observed, longer-lasting turnover response is not equally large for the comparison event. Evidence of a founder-specific difference is therefore compatible with some general grief and with additional effects that the data cannot see.
A sudden loss is not a succession plan
The distinction between sudden loss and planned departure is more than a methodological footnote. A death removes a person and creates an emotional event at the same time. A planned succession can unfold through announcement, overlap, delegation, and negotiation. Employees may have time to decide whether the successor understands the mission, whether their relationships will remain useful, and whether the new strategy offers them a place. Some will still leave, but their decisions may reflect a different mixture of disappointment, opportunity, and adaptation.
A founder dismissed by investors is different again. The founder may leave because the board wants a new operating model, because the company needs a different type of leadership, or because the founder’s relationship with employees has already become a liability. In that setting, the turnover response could be smaller if employees welcome the change, or larger if the dismissal signals deeper conflict. A founder who gradually withdraws may generate yet another pattern, with employees responding before the formal departure date.
The practical implication is not to pretend that these events are interchangeable. It is to ask what has already changed before the official transition date. If employees have stopped receiving decisions through the founder, if key relationships have moved to other leaders, or if routines have become genuinely distributed, a planned exit may carry less of the sudden-loss risk. If the founder remains the only person who can interpret the mission or connect major parts of the team, advance notice alone may not remove the underlying dependence. These are diagnostic questions, not effects tested by the study.
The Census result should therefore guide questions rather than supply a universal forecast. It tells a startup that a founder can be part of the retention system and that sudden removal carries a measurable risk. It does not tell the startup how many employees will leave after a planned transition, whether keeping the founder would improve performance, or whether the turnover would be beneficial because it clears space for a new strategy. Those judgments require information the study does not contain: the founder’s actual role, the quality of employee relationships, the strength of distributed routines, and the credibility of the successor.
This is also why the event’s early concentration matters for practical preparation. If the first two years carry the largest disruption, a transition cannot rely on the hope that formal structures will eventually absorb the change. The startup may need to understand before departure which decisions, relationships, and knowledge flows are concentrated in the founder. That is not a claim that every function can be documented or transferred. It is a recognition that the highest-risk period may arrive before adaptation has had time to occur.
The same timing creates a measurement trap. A startup that survives the first year with limited turnover may appear stable, even though the employees who were most attached to the founder have already left. A startup that records a burst of exits may be experiencing damage, healthy renewal, or both. Turnover is an outcome that deserves explanation, not a complete evaluation of the organization. The founder-loss evidence identifies a vulnerability; it does not define the ideal amount of employee movement.
For a decision maker, the relevant follow-up is not simply “How many people left?” It is “Which capabilities and relationships left with them, and which new capabilities became possible?” A founder-dependent startup may lose people who carried tacit knowledge, customer trust, or coordination capacity. It may also gain employees who were previously constrained by the founder’s style. The study establishes the risk of a retention shock, but judging its net organizational effect requires evidence about performance and renewal that this design does not provide.
What the comparison lets us say
The evidence supports three linked conclusions. First, a sudden founder loss changes employee turnover relative to similar startups, and the absence of visible pre-loss divergence makes a pre-existing turnover trend a less satisfactory explanation. Second, the effect is concentrated soon after the event, which is consistent with a sudden change in attachment and coordination rather than only a slow maturation trend. Third, the smaller and shorter response after a non-founder loss suggests that founder position adds something beyond the disruption of losing a colleague.
The evidence does not support a single, fully observed mechanism. Grief may still matter. The data do not show what founders did formally or informally, why each employee moved, or whether the founder’s departure improved or damaged the startup’s long-run performance. The comparison is strongest for unexpected founder loss and weaker as a direct guide to ordinary succession.
The next puzzle follows directly from this one. If the founder effect is a shock to accumulated relationships, we should expect it to vary with how much time employees have spent working with the founder and with the kinds of similarity or alignment that make collaboration easier. The question is no longer only whether the team moves after a founder leaves. It is which employees are most exposed—and why that exposure might become greater as the startup matures.