The tally sheet below is invented. It is the kind an analyst might keep on a buyout offer between the announcement and the deadline. The top half is already done: how many people are eligible, how many the company expects to accept, when they decide and when they go. The bottom half asks a different question, who the volunteers are, and every line is blank. It will stay blank until December. In most companies it stays blank after that, because nobody is asked to fill it in.
Buyout tally · illustrative
The count
Who said yes
On Monday, October 26, Microsoft introduced what it calls a Voluntary Exit Offer. In a memo published on the company blog, Amy Coleman, its chief people officer, described a time-limited program for eligible employees in participating teams in Canada, India, the United Kingdom and the United States, at level 70 and below. It offers financial support and career planning to people who may already be considering a future outside the company.1 The company told GeekWire that about 60,000 people are eligible, roughly a quarter of a workforce last counted at 223,000; CNBC, citing a person familiar with the matter, put the share at about 27 percent.23 The terms match Microsoft's standard severance, and there is no age, tenure or performance requirement.2 Employees decide in early December, and those who accept leave on December 31.3
Two days later, a spokesperson told Business Insider that the company expects 3,000 or fewer people to accept. Amy Hood, the chief financial officer, wrote to staff that the intent is to give employees "agency if they choose to leave the company on their own terms," and that she expects very few to take the offer.4 On the earnings call that afternoon, Hood said the quarter's cost guidance "includes an estimate for the recently announced employee voluntary exit offer."5
The backdrop is heavy investment, not distress. Microsoft reported quarterly revenue of $92.3 billion, up 19 percent, and capital spending of $45.2 billion, with more than $50 billion expected next quarter.5 It has also said plainly that the offer is not a sign it needs fewer employees and is not a replacement for layoffs.2 Harvard Business Review has asked whether the buyouts are a sign of what is to come.6 That is a fair question, but it is a question about how many. I want to ask a different one.
My argument is that a voluntary offer changes who makes the choice, and that the choice is the part to watch. In a layoff, the employer decides who goes. In a buyout, each employee decides, and the people who say yes are the ones whose prospects outside look best against what they would give up by staying. How many leave is a number a company can forecast, budget and announce. Who leaves is a selection the company has handed to the people it is selecting from, and it can only be read afterward. The research on buyouts says that selection can go either way, and which way depends on how the offer is built.
This is not an argument against voluntary exits. They are, in many ways, the more decent tool, and I make that case below. It is not a claim that Microsoft's best people will leave; nobody knows that yet, and the evidence does not predict it. It is not a suggestion that the offer is a layoff in disguise, which Microsoft denies, and it says nothing about the stock. The claim is narrower: whoever runs a buyout should plan to measure who took it, because the count alone will not say whether it worked.
What the studies found
Economists have studied buyouts at single employers for decades, usually with personnel records that let them compare the people who left with the people who stayed. Four findings matter here.
The first is that money moves people, but not predictably. John Pencavel studied the severance incentives offered to University of California faculty in the early 1990s and found that a 10 percent richer benefit raised the probability of leaving by about 7 to 8 percent. He also found that quit rates were "very difficult to forecast with accuracy," which weakens the case for buyouts as a precise tool for large reductions.7
The second is that the people who accept are not only those near retirement. Daniel Feldman and Seongsu Kim studied 812 employees in the Korean electronics industry facing downsizing. Acceptance rose with the size of the incentive, with age and with poor health, as you would expect. But among employees early and in the middle of their careers, the better performers were the most likely to accept.8
The third is that the same employer can get different results from different rounds. Tsuyoshi Tsuru examined two buyouts at a Japanese machinery maker. In 1993, the people who took the offer were rated no better than those who stayed. In 1996, among non-managerial staff in the lower grades, those who left had been rated higher than those who stayed, despite the firm's efforts to keep the people it valued. His point is structural: a firm chooses whom it dismisses, but it does not choose who accepts a buyout.9
The fourth is that adverse selection is not automatic. Kenneth Whelan, Ronald Ehrenberg, Kevin Hallock and Ronald Seeber studied a retirement incentive for Cornell University staff. Enrollment rose in units facing budget pressure and recent layoffs, and employees paid less than their characteristics predicted were more likely to take the offer. To the extent pay tracks productivity, they found no evidence that the most productive took the deal.10
Side by side, the studies do not say buyouts drain talent. They say the result depends on who is offered what, and that the employer finds out only by looking. That conclusion is less dramatic than either side of the usual argument, and more useful.
Reading the offer as a selection rule
Every design choice in a buyout is a filter. Microsoft's offer has several, and they point in different directions.
Eligibility is set by team, not by person. Business leaders decide whether their groups take part; the sales organization does not, and neither do some engineering groups and corporate functions.2 That is the main lever an employer keeps in a voluntary program, and Microsoft is using it. The selection starts with which work the company can spare, before any individual decides.
Within a participating team there is no performance screen, by design.2 That is fair, and it removes any stigma from saying yes. It also means the filter inside a team is each person's own outside option, which is the mechanism the studies describe.
The price is standard severance, without the extended healthcare that drew people to the spring program.2 A modest offer to a broad pool mostly attracts people who were already close to leaving. Some of them are people a team would miss and some are not, and the modest price is one plausible reason the expected uptake is small.
Microsoft has a comparison inside its own year. The spring Voluntary Retirement Program, its first, was limited to US employees whose age plus tenure crossed a threshold; about 8,750 were eligible and roughly 30 percent accepted. Coleman has said it reduced the July layoffs.2 That design selected by career stage. The new offer opens eligibility far more widely at a lower price, with the company expecting one in twenty eligible people or fewer to accept.4 Microsoft has said it is studying whether to make voluntary exits a regular practice, and Coleman's memo is candid about the experiment: "The Voluntary Exit Offer is new for us. We'll learn from it."21 The studies suggest what there is to learn.
In March I argued that the cut is immediate and the automation is not: a headcount is a number, while the work is a list someone has to write. A buyout adds a further step. The employer does not even choose the names on the list.
Objections, considered
Voluntary is kinder, and it may prevent a layoff. This is the strongest case for the offer, and I accept most of it. A voluntary exit respects the person making the choice, which is what Hood means by agency. It costs less trust than a forced cut, it gives people already planning to leave a supported way to go, and by Microsoft's account the spring program reduced a later round of layoffs, in which about 4,800 people lost their jobs.23 Nor is it obvious that forced cuts serve shareholders better. Across thousands of employment changes at large US companies from 1980 to 1994, Wayne Cascio, Clifford Young and James Morris found that companies that only downsized did not outperform their industries.11 None of this conflicts with the argument here. Measuring who leaves is how a humane tool stays a good one.
Good people can leave any time, so the offer changes little. Partly true. But a severance check lowers the cost of a decision someone was already weighing, and Pencavel found that the size of the check moves the probability.7 At the margin, the offer brings forward departures that might have come later, or never. Which people sit at that margin is the question.
Three thousand out of 223,000 is too small to matter. At company level, perhaps. But departures are not spread evenly. They fall in the teams that opted in, and a team of twelve that loses its two most experienced engineers feels it more than the company total suggests. The number to read is the team's, not the company's.
The studies are old, small and from elsewhere. They are. Each covers one setting: a US university system, a sample from Korean electronics, a Japanese manufacturer, another US university. They measure productivity by pay or performance ratings, which are imperfect, and none studies a large technology company. That is why the conclusion here is modest. They do not predict what Microsoft will see. They show that the outcome varies with design, and that the employers who learned something were the ones who looked.
A layoff picks the people. A buyout lets the people pick themselves.
What to decide before the offer goes out
- Set eligibility by the work you can spare. Microsoft's team-by-team approach is the right instinct. Level and tenure are proxies; the work is the thing.
- Price the offer knowing it sorts people. A richer package moves more people, and not only the ones you expected. Decide in advance which outcomes you can live with.
- Forecast the uptake, and plan for a miss. Pencavel's warning applies. Have a plan for twice as many acceptances, and for half as many.
- Compare who left with who stayed. Ratings, tenure, hard-to-rehire skills and the knowledge of why systems work as they do, measured within weeks of the deadline, the way the researchers did it.
- Keep the voluntary offer and any later cut separate. Each has its own goal and its own measure. Mixing them hides whether either one worked.
Back to the tally sheet. The gold half is finished, and it will be quoted in every article about the offer. The blue half fills in after December, one line at a time, for anyone willing to look. That is the half that says whether the offer worked.