The offer at twenty-six
Sunday, quarter past three in the afternoon, a long lunch that has not ended. The father put the first press brake into a rented shed thirty-one years ago; today the aluminum door and window plant employs one hundred and ninety people in Villa Mercedes and he still signs off on every large purchase.
The eldest son is twenty-six. He graduated fourteen months ago and has worked in the company since he finished secondary school: first in the warehouse, then in purchasing, now at a desk next to production, with no defined title and access to everything.
He has an offer. A consumer goods multinational, plant in Córdoba, eighteen months rotating through three areas and then we will see. It pays less than he earns today.
The father listens to the end and does not say no. He says three things, and all three are good:
"Here you will learn faster, because here you touch everything. There you will be a number on an org chart."
"Here they already know you. There you start from scratch and lose two years."
And the third one, which is the one that closes the conversation:
"This is going to be yours at some point. Why would you go and learn somebody else's company?"
The son stays. Nobody made a bad decision that afternoon: the father said what anyone who built something and wants to hand it over whole would say, and the son chose to stay close. The letter ends up in a drawer of the father's desk, which is where the things you do not want to throw away end up.
(The scene is a composite reconstruction. From here on, everything carrying a number is measured and has a source.)
The sources for this article were read in the version available in August 2026.
The three facts the father states are true: the son will touch everything, they know him, and the company will be his. What did not survive measurement is none of the three: it is the conclusion drawn from the third one —if this is going to be yours, there is nothing to learn at somebody else's—, and it is the one part of that afternoon nobody argued with.
Two thousand four hundred successions, split on a single fact
Part 1 of this series ended with a measurement covering more than five thousand handovers of the top job: operating return on assets falls by at least four points between the years before and the years after. The mean of that sample was six and a half pesos a year for every hundred put into machines, inventory and sheds, so four points off that mean leaves two and a half. That is the scale of the cost: how big the hole is. Who produces it is a different measurement, and that is the one that appeared in 2025. The Part 1 count is twenty years old and nobody has refuted it; what appeared in 2025 was the location of the dividing line.
Irena Kustec, Charlotte Ostergaard and Amir Sasson went in through the administrative records of Norway's statistics office, which are not a survey of whoever felt like answering: they are the register of everyone. In there is the country's entire population of family-controlled firms, almost all of them privately held —they look a good deal more like yours than the ones that show up in the papers—, and in there are some two thousand four hundred CEO successions over ten years. With that they made the cut the literature had not made: they split the successors by where they came from when they took over, from inside the company or from outside, and the underperformance ended up entirely on one side.
On one side are the ones recruited from inside —those already working in the company when they took over—, around five in ten successions: they take over at thirty-eight, seven in ten are children of the outgoing CEO, and at least forty-five in a hundred have no record of a single job outside the company in their entire working life.
On the other are the ones who left and came back: they leave at twenty-six and return at thirty-six. They are two in ten successions and they perform on a par with a professional CEO.
The remaining three in ten are professional CEOs: the benchmark the other two are measured against.
The obvious objection remains, and it is the good one: a family has two or three candidates and the best one may not be among them. The authors went looking for it. They compared children who left against children who did not, and then narrowed the field further: firstborns against firstborns. There both groups hold the same place in the family and have the same chance of being the default choice; the only thing separating them is having left. The pattern held.
The three profiles, and which side each falls on:
Why this article rests on a working paper, and what still stands if the paper changes. The work holding up half of this article came out in 2025 in the working paper series of the European Corporate Governance Institute. It has not been peer reviewed: it is a draft put out to be criticized, and it can change. In fact it has already changed once —an earlier conference draft is in circulation, with different figures and with results the 2025 version does not carry—, and this article uses version 1046/2025 exclusively. This firm uses it anyway, for two reasons worth saying out loud. The first is that it is the only available work that separates family successors by outside experience instead of treating them as a single group, which is exactly what twenty years of literature had been doing. The second is that the rest of the evidence in this article is published and peer reviewed, and points in the same direction from another angle. If the Norwegian finding fell apart entirely tomorrow, what the next section says would still stand: that family firms choosing their boss by an automatic rule score worse than those choosing by criteria. The advice in this article —send him outside, write the criteria— does not change sign under either result. What would change is how much confidence it deserves.
What a son who never left gets calibrated against
A statistical cut explains nothing on its own. What makes this one useful is that the mechanism is recognisable without a spreadsheet, in any company, yours included.
Somebody who has worked at three companies knows three things you do not learn inside one. He knows that the way the month gets closed here is a way and not the way. He knows what it feels like to be measured against a yardstick he did not negotiate. And he knows what his work is worth in a market where nobody knows his father. The heir who never left has none of the three, and not for lack of ability: because he never had the occasion.
The other half is missing, the half almost nobody writes down and the half that falls to you. Your people cannot calibrate him either. Nobody in that company has ever assessed him without knowing whose son he is: every piece of feedback he has received in ten years came from people who report to you. The result is not that he was over-praised. It is that the data point does not exist, not for him and not for anyone, and the day he takes over there is not one clean assessment in his entire career.
It is worth naming the problem precisely, then, because the word always used —"professionalize"— points at the family name and what was measured points somewhere else. The Norwegian work does not separate family from non-family: it separates having worked outside from not having done so. That criterion does not ask about the family name, and the career manager who joined at twenty-three and never left lacks exactly the same yardstick as the owner's son.
This firm calls that the monoculture: the company where everyone who can reach the top was trained inside it. How far the backing goes: what is measured is the cut among family successors; that the monoculture weighs the same on a manager without the family name is this firm's judgment, not a data point. The word names the company and not a person, and that is the reason it is useful.
If the problem were the family name, there would be no way to fix it. Since it is the track record, there is one.
The family name does not drag the average down
A finding from a single country holds up better when something measured somewhere else points to the same place. That exists, it is published and peer reviewed, and it comes from the largest management practices survey there is, which scores each firm on eighteen concrete practices: how it sets targets, how it measures, what it does with the person who performs badly. The data is from the mid-two-thousands, the same age as the Danish measurement in Part 1.
Two warnings before the numbers. Zero is the average firm in the sample, and a whole point on that scale is the typical distance between an ordinary firm and one that is visibly better managed. And the scale measures practices, not money: this article has nothing with which to convert hundredths of a point into profitability, and it is not going to invent it.
Across seven hundred and thirty-two firms, the scale orders three situations that stack on top of each other: each one is the previous one plus a requirement.
The first: firms whose largest shareholder is a family fall just above zero, fourteen hundredths, with a margin of error that does not allow much more to be claimed than this: family ownership is not on the bad side of the table.
The second: when the CEO is also from the family, the mark barely moves —it drops one hundredth— and stays above the average. Putting a family member in the job, on its own, does not make measured management worse. This contradicts nothing above: the Norwegian work does not say the family member performs worse either. It says the one who never left performs worse.
And the third: when that CEO reached the job for being the eldest son, the mark falls a further forty-one hundredths and ends up twenty-eight hundredths below the average. Of the three moves, it is the only one measured with high precision. The technical word for that rule is primogeniture.
All three are the same family, owner of the same company, and the only thing that changes is the rule by which the person in charge gets chosen. These are associations, not proof of cause. What the scale ranks from worse to better is not families: it is ways of choosing.
One data point is missing, the one that takes that rule out of the realm of destiny: not every country chooses the same way. Among the French and British firms measured, around one in seven uses the eldest-son rule; among the German and American ones, three in a hundred. Something that varies like that between rich neighbouring countries is not a law of nature: it is a custom. The picture for this region comes later and points to the same place.
The three situations, placed on the same scale:
Nine percent fewer hours
There is something else that sets that job apart, and it shows up where nobody looks for it: the calendar. A team logged, hour by hour, the diaries of more than a thousand manufacturing CEOs across six countries. In that measurement —which compares performance, not practices— family firms perform below non-family ones. CEOs who are from the family work nine percent fewer hours than professional ones, controlling for characteristics of the manager and of the firm, and that difference accounts for around eighteen percent of that gap. It is the same measurement this firm worked with on the time of the number one. Do not stack it on top of the previous one: that one scores practices and this one compares performance. They are different yardsticks.
Neither the study nor this firm reads that nine percent as a lack of commitment: it is a fact about the design of the job. The owner-manager also has a family inside the company, an estate to administer and often a father on the board with settled opinions. The job was designed for someone who will put in all the hours, and it is held by someone who structurally cannot.
Three data points, and none of them is about the family name:
None of the three measures who owns the company. All three measure how the person running it got the job.
A succession is a capital project with no file
In a large investment, between the idea appearing and the check being signed there is definition work: the scope is studied, alternatives are compared, costs are estimated with a stated margin, and only then does the committee decide. In capital projects that has a name —front-end loading— and skipping it is the most expensive way to start. A succession is a decision of that size and the only one usually settled without compared alternatives, without written criteria and without a date. The four steps that follow are the file that is missing.
First, the date, and do not let anyone's health set it. The first thing to write down is not the name: it is the year. A provisional date, revisable, in writing and known to whoever decides. Without a date there is no process, there is expectation, and expectation manages itself, badly.
Second, the outside experience, with its floor. The profile of the ones who performed is ten years outside. This firm does not propose ten: it proposes a workable floor of between two and five years at another company, with a boss who does not know the family and a written assessment from somebody who owes nobody anything. That floor is not measured, it is this firm's judgment; what is measured is the direction.
The objection always comes up and it is sensible: if he leaves, he does not come back. The risk is real and it gets managed like any other: a return date and a return condition, written down before he goes. A son who leaves with no date is a bet; with a date, it is a development plan with two signatures.
Third, the written criteria, before looking at any candidate. This is hiring for the most important job in your company, and what the evidence says about how to hire well does not change because the candidate sleeps in your house. What whoever holds the job has to be able to do, on one page, before knowing who is going to win. And at least one candidate compared who does not carry the family name: the comparison is not there to choose somebody else, it is there to produce the criteria.
Fourth, the governance that outlives the founder.
4a. Who can tell him no. You need somebody with no financial dependence on the family and with a written mandate to say no to the successor. If there is a board, it is a director; if there is not —and in most companies of this size there is not—, the requirement is not to assemble one overnight: it is that this person exists, with the mandate in writing, even if they sit three times a year. You need a family protocol: who can work in the company, on what requirements and who assesses them. In Argentina it has had its own legal standing since 2015, and a limit worth knowing beforehand: it cannot override the heirs' forced share. That gets checked with a succession lawyer.
4b. What gets transferred and when. And since the incoming number one will put in fewer hours than a hired manager, the list of which decisions stop needing his signature. The transfer goes in parts: signing authority, the chair and ownership do not change hands on the same day.
The four stations, with who signs each one:
The two doors
The question this article has owed since the first paragraph is still there: what happens if, with all of this done, the family candidate is not the best option. Professionalization is presented to the owner as a defeat, as handing the company to a stranger. That is not what the evidence says: a hired CEO, with the family in the ownership and in the chair, is family continuity too. What changes is who executes, not who owns or who sets the direction.
That door has a trap. Across more than one thousand six hundred CEO appointments at listed family firms worldwide, the dominant flow runs toward professional management, but it happens more through internal promotion than through outside search. That work does not compare performance, so what follows is this firm's conclusion: if the cut that matters is having worked outside, promoting the person who joined young and never left changes nobody's schooling. It is the same monoculture with a different family name.
From the same work comes a data point that lowers the price of the gesture: the market pays no premium for professionalizing and imposes no penalty for putting a family member back in; the share price moves almost the same —and little— in both cases. Nobody is going to applaud your org chart; what gets paid for is the result. That comes from a blog post, not from a paper.
And a picture of the region, because by this point the objection is that things work differently here. Among the one hundred and fifty-five largest family firms in Latin America with 2014 data, Chile is the country with the fewest CEOs from the family: one in four. The explanation is given by the authors themselves and it is not cultural: Chile has the highest stock market capitalization over output in the sample and prohibits by rule that the chair of the board also be the CEO. A rule, not a temperament.
The two paths, with what each one demands:
What will be said around the table
None of this gets decided alone. It gets decided around a table with partners, siblings, an accountant of thirty years' standing and, frequently, the outgoing generation sitting opposite. Four reasonable sentences will come up there: any of us would say them, and in fact we have. Each one has a half that is true and hard to argue with, and from each one the measurement takes the other half away. None of them is answered with an opinion: they are answered with one-page documents that almost never exist.
| What will be said | What it runs into | What still stands |
|---|---|---|
| "He grew up in here. He knows the company better than anyone" | The underperformance of family successions is concentrated entirely among those who never worked outside the company (Kustec et al., 2025 — working paper) | Knowing the company is an asset. Having nothing to compare it against is the liability, and it cannot be seen from the inside. |
| "If I send him out, he won't come back" | The profile of the successor who performs on a par with a professional is exactly that of the one who left at twenty-six and came back at thirty-six (same source) | The risk that he does not come back is real. It gets managed with a written return date and return condition, not by keeping him from leaving. |
| "The problem is the family; a professional manager fixes it" | What ended up measured is not the family name, it is the outside experience; and at listed family firms, professionalizing happens more through internal promotion than through outside search, with the market paying no premium for the gesture (Barroso et al., 2026 — blog post) | What has to be bought is not a different family name: it is a different career. |
| "Now is not the time; we'll see about it later" | Family firms that choose their boss by an automatic rule score clearly worse on management practices than those that choose by criteria (Bloom and Van Reenen, 2007) | Not deciding is a selection rule too. It is the one that chooses by itself, and the day it applies there is no time left to develop anyone. |
Each sentence is half of an argument. The other half is not said out loud, and it is the half that decides.
Questions for Monday
Seven questions, and none of them needs a consulting engagement. You already have the first three, sitting where you are.
Three you already have, two that live outside your head, and two you do
- Besides you, how many employers has the person currently in line to succeed you had?
- At what age did he leave the company and at what age did he come back? If he never left, that is not a verdict on him: it is the only one of the measured factors there is still time to change.
- The last time somebody assessed his performance without knowing whose son he is — when was it?
These two have their answer outside your head:
- Ask your two longest-serving managers what they would say to your successor if he were not your son. They do not have to answer truthfully: it is enough to see how long they take.
- Where is it written what the person holding your job has to be able to do? If the answer is "in my head", that is the document that is missing and it is the cheapest of them all.
And these two are not read: they are done, and both fit on one sheet.
- Write down what happens if tomorrow you cannot come to work for six months. Who signs, who decides, who tells the bank. With names. It takes twenty minutes and almost nobody has done it.
- Make the list of the three companies where your successor could work for three years without it feeling like a betrayal. If it comes out empty, there is a pending conversation there, and it is not about him.
At that Sunday table in Villa Mercedes the father was not wrong on the facts: the son touched everything, they knew him, and the company will be his. He was wrong about one single thing, and it is not even obvious. Learning somebody else's company was not losing two years in a business that was not his: it was getting the one thing his company cannot give to anyone who stays inside it, a yardstick his father did not write.
If your successor is twenty-six, this gets fixed with a conversation and a plane ticket. If he is thirty-four and has been inside for ten years, it costs more: two or three years outside when he already has children and a settled salary, or —if that is not going to happen— a formal outside assessment, a boss who is not you for a while, and somebody with a mandate to tell him no. If he is forty-five and signing authority has already been transferred, what is left is not development: it is governance.
With this the series closes. Part 1 took apart the number quoted for forty years to announce that family businesses are doomed, and showed that the real cost sits somewhere else. This one shows that that cost has a single large variable inside it, and that the variable is a track record, not a family name. The family business is neither condemned by its form of ownership nor blessed by it: it is better served by a process than by a faith. The monoculture gets fixed by a decision taken ten years before it is needed, and that decision is being taken today by somebody, in your company, without knowing they are taking it.
Where is it written what whoever holds your job has to be able to do?
What 32sur leaves installed is one page: what whoever holds your job has to be able to do, written with no name in front of it and by the people who will decide, not by you alone. From there come the date, the development plan for each candidate and the comparison against somebody who does not carry the family name. If that page does not exist, or if the date of the succession is still being set by somebody's health, let's talk.
References
- Kustec, S., Ostergaard, C. and Sasson, A., "Underperformance in Family Successions: The Role of Outside Work Experience", ECGI Finance Working Paper 1046/2025 — the backbone of this article. Administrative data from Statistics Norway covering the population of Norwegian family-controlled firms, mostly privately held: around 2,400 CEO successions over a ten-year window. The underperformance of family successors is entirely accounted for by those recruited inside the company; family successors with outside work experience perform on a par with professional CEOs. Profile of the internal successor: takes over at 38, at least 45% have no record of any outside work experience and 70% are children of the outgoing CEO; they are around half of the successions in the sample. Profile of the successor with outside experience: leaves at 26 and comes back at 36; they are 20% of the successions. Professional CEOs are the remaining 30%. The authors rule out the small-candidate-pool hypothesis by comparing internal sons against external sons, and internal firstborns against external firstborns, and the pattern holds; they also add controls for the outgoing CEO and for his presence on the board, without the result disappearing. It is a working paper: it has not yet been through peer review. Source hygiene: a December 2020 conference draft, earlier than this version, is also in circulation, with different figures (2,500 successions, fifteen years, two thirds with no outside employment) and with additional results on professional CEOs recruited inside the company that do not appear in WP 1046/2025. This article uses version 1046/2025 exclusively and takes nothing from that draft.
- Bloom, N. and Van Reenen, J., "Measuring and Explaining Management Practices Across Firms and Countries", The Quarterly Journal of Economics, 122(4), 2007, pp. 1351-1408 — the published, peer-reviewed reinforcement of this article's thesis. Table V, p. 1384, column 5, on 732 firms. The three coefficients are nested, that is, increments that accumulate and not mutually exclusive categories: family ownership +0.138 (standard error 0.086); and also a CEO from the family −0.010 (0.113), a value the measurement cannot distinguish from zero; and also chosen by primogeniture −0.410 (0.122), the only one of the three measured with high precision. The positions on the scale, which are what Figure 2 of this article draws, come from adding them up: +0.138 · +0.128 · −0.282. In the authors' own words, firms that choose the CEO from among all family members are not worse managed than the rest; those that choose by primogeniture are. Table 1: the eldest-son rule appears at 14% of the French firms and 15% of the British firms in the sample, against 3% in Germany and 3% in the United States. Source hygiene: the NBER working paper that preceded this publication (WP 12216, 2006) carries different values in the same column —+0.304 · −0.152 · −0.450—; this article uses the ones from the published version. These are associations measured in cross section, not causal effects.
- Bandiera, O., Lemos, R., Prat, A. and Sadun, R., "Managing the Family Firm: Evidence from CEOs at Work", The Review of Financial Studies, 31(5), 2018, pp. 1605-1653 — the fact about the design of the job: CEOs from the family work 9% fewer hours than professional ones, and that difference accounts for around 18% of the performance gap between family and non-family firms. A time-use study of 1,114 manufacturing CEOs in six countries. Source hygiene: the abstract of the published version states 9%, and that is the number this article uses; some secondary summaries of the same work circulate with 8%.
- Vazquez, P., Carrera, A. and Cornejo, M., "Corporate governance in the largest family firms in Latin America", Cross Cultural & Strategic Management, 27(2), 2020 — the regional anchor: 155 family firms drawn from the ranking of the 500 largest in Latin America, with 2014-2015 data, financial firms excluded. Chile is the country with the lowest share of CEOs from the family among the largest family firms surveyed —around one in four— and the authors attribute this to two verifiable structural factors: the highest stock market capitalization over output in the sample (84.1%) and the regulatory ban on the chair of the board also serving as CEO. The sample is made up of the largest companies in the region: it does not describe the mid-sized family firm.
- Barroso, R., Kowalewski, O. and Luitel, P., "Nepotism or Stewardship? What CEO Successions in Family Firms Really Tell Investors", European Corporate Governance Institute blog post, 30 July 2026 — on more than 1,600 CEO appointments at listed family firms worldwide: 55% of successions go from one hired manager to another, 36% from a family member to a hired manager and 9% from a hired manager to a family member. The authors find no market premium for professionalizing and no penalty for appointing a family member again: both types of appointment produce approximately equal and small returns. They also point out that the dominant flow toward professional management happens more through internal promotion than through outside recruitment. It is a blog post about work in progress, not a published paper, and it is cited in this article with that label attached.
- Bennedsen, M., Nielsen, K. M., Pérez-González, F. and Wolfenzon, D., "Inside the Family Firm: The Role of Families in Succession Decisions and Performance", The Quarterly Journal of Economics, 122(2), 2007, pp. 647-691 — the measurement of the cost that closed Part 1 of this series and that this article picks up once only: 5,334 CEO successions between 1994 and 2002 at Danish limited liability companies, using the sex of the outgoing CEO's firstborn as an instrument; operating return on assets falls by at least four percentage points between the window before and the window after the transition, on a sample mean of 6.5%. The "from 6.5 to 2.5" arithmetic this article uses is done on that mean, and is there to give scale: it is not a post-transition value measured firm by firm.
All sources last checked: August 2026. Reference 1 is a working paper and may change; reference 5 is a blog post about work in progress.