Haute Lumière

Commerce · V · MMXXVI · daylight

La Bourse  /  Volume V  /  Nº V.E2

For the Corporation

Volume V — Labour, Value, Flourishing · Extension II of III Nine movements, one committee.

---

THE PLATE

---

THE LETTER

You sit on, or report to, a committee that will decide compensation, workforce planning, health spending and what the board is told about its own people. Five times a year somebody brings you a number about all four, and the numbers are not equally good.

This chapter sorts Volume V's findings by the only criterion a committee can actually apply: whether the claim can be verified at the headcount the company has. Not whether it is plausible. Not whether the vendor is reputable. Whether the endpoint is detectable — and that is a computation, it takes two inputs, and almost nobody performs it before signing.

The result is uncomfortable in a specific and useful way, and it is set out in the Arithmetic. The two claims a board is most readily sold are the two the company cannot check, by factors of 10.05 and 6.40. The endpoint it clears most comfortably is the one nobody sells.

The load-bearing case is Chapter V.09's, and it is handled here without softening. A workplace wellness programme that returns 5.60 : 1 on the vendor's own arithmetic returns 1.12 : 1 once you ask what the gap between the two groups was before the programme existed, and 0.00 : 1 at the point estimate of the two large randomised trials. That is not a scandal and nobody lied. It is a selection effect with a closed form, and this chapter gives you the form, the money, and what to do with the paper on your desk this quarter.

You will also find, because a committee is entitled to both halves, what the company should stop — eight things, each with the arithmetic that retires it — and what it should start, which is nine things, of which four are free and one of them is the single highest-return administrative decision in the volume.

And one honest negative that this house's own argument does not dissolve: Chapter V.06's cost of an unsellable claim. It is 1.46 points of hurdle and 2.70 points of growth a year, it compounds to 64.2 per cent over twenty years, and any paper that proposes employee ownership at scale to a growth business without naming it will be correctly refused.

— The Editors

---

DISCOVERY

What is already working, and it is more than the reputation

Begin where corporate practice is at its best, because the good version is better than the deck and almost nobody on a committee has been shown it.

A randomised inspection, and a saving nobody could have selected into. California allocated workplace safety inspections by lottery, and David Levine, Michael Toffel and Matthew Johnson read the lottery as the experiment it was. Inspected establishments had 9.4 per cent fewer injuries and 26.0 per cent lower workers' compensation costs over the following four years. And the result everyone expects and nobody finds: no detectable difference in employment, sales, credit rating or survival. The safety did not cost the jobs. That is a randomised finding, about money, at establishment scale, and it is the cleanest thing in the whole corporate-health literature.

A programme that published its own two halves. RAND's seven-year evaluation of PepsiCo's Healthy Living reported 1.46 : 1 overall and then, honourably, reported it split: 3.78 : 1 on disease management for people with a diagnosed condition, saving $136.00 per member per month — $1,632.00 a year — against 0.48 : 1 on lifestyle management for healthy people's habits. The two differ by a factor of 7.88 and point opposite ways. A firm that funds the first half and declines the second is acting on the best evidence in the field, and the split was published by the evaluator rather than extracted from them.

A financial incentive that worked, at a price you can put in a budget. Kevin Volpp and colleagues randomised 878 General Electric employees and found cessation at nine to twelve months of 14.7 per cent against 5.0 per cent — 9.7 points, a number needed to treat of 10.31. That is a workplace buying a behavioural change at a stated price with a control group attached, and it is the opposite of a null.

A field where the reversal was published by the people who would lose from it. Between the meta-analysis that produced the 3.27 : 1 medical figure and today, the field ran two large randomised trials — Illinois, 12,459 university employees; BJ's Wholesale Club, 160 worksites and 32,974 employees — and found close to nothing on cost. One of the authors of the original meta-analysis is a co-author of the trial that found it. A field that publishes its own reversal has handed a committee something worth more than the original number, and the rest of this chapter is spending it.

And the corporate cases where the flourishing argument simply won. Alex Edmans tracked the hundred best companies to work for in America from 1984 to 2009 and found a four-factor alpha of about 3.5 per cent a year, roughly 2.1 per cent industry-adjusted — 2.45 times the market compounded, or 1.72 times on the adjusted figure. Wayne Cascio's comparison put Costco's median wage at $17.00 an hour against a competitor's $10.11, a premium of 68.2 per cent, with turnover at 17 per cent across all staff and 6 per cent after the first year, and sales per employee of $795,000 against $516,000 — a ratio of 1.54. Neither of those is a values argument. Both are line items.

The pattern, and it is the one this chapter runs on. In every good case the outcome was measured by somebody who could not choose who was in the programme — a lottery, a randomisation, a state inspection schedule, an administrative claim run, a stock price. Where that is true the field produces numbers a committee can sign. Where it is not, it produces the slide.

---

THE ARITHMETIC

The measurability frontier, the decomposition, and the cost this volume cannot talk away

First, the cut.

Every claim a committee is asked to believe has a sample size attached to it whether or not anyone computed one. The two-arm formula takes two inputs — the standard deviation of the endpoint and the effect you want to detect — and returns the number of employees required:

  z at 2.5 per cent in each tail                  1.959964
  z at 80 per cent power                          0.841621
  (z_a + z_b)                                     2.801585
  two-arm constant  2 (z_a + z_b)^2              15.697759

Run it on the six endpoints a board is actually offered, at a worked employer of 5,000 people.

The engagement row is computed differently and is worth showing, because it is the row a committee never questions. To detect a true-score correlation of 0.22 at 80 per cent power you need ((z_a + z_b) / Fisher z)^2 + 3 business units, and the Fisher z of 0.22 is 0.223656:

  business units needed                                160
  people, at 200 to a unit                          32,000

A firm of five thousand is 6.40 times short of being able to verify its own engagement-to-performance claim, and 10.05 times short on medical spending. It clears voluntary turnover by 1.59, absence by 1.11, the wellbeing cross-section by 1.14, and — over four years rather than one — injury frequency by 1.23.

Now put the two orderings side by side. Rank the six by how readily a board is sold the claim, which is this chapter's own assertion and is labelled as one; rank them again by how many employees it takes to check. Kendall's tau between the two rankings, on fifteen pairs, thirteen concordant and two discordant:

  Kendall tau                                     0.7333

A positive tau, and a large one. The more readily a claim is bought, the more employees it takes to check it — and the two hardest claims in the list are the two most reliably purchased. That is the whole of the corporate people budget's pathology in one coefficient, and it is not anybody's dishonesty. It is what happens when the cheapest claim to make is also the most expensive claim to verify, and nobody in the room is asked to compute the denominator.

The operational consequence is a single sentence and it belongs in the procurement policy: no people programme is funded whose primary endpoint requires more employees than the company has.

Second, the decomposition, without flinching.

Take the worked employer: 5,000 people, a programme at $150.00 a head — $750,000 a year — with participation at 56 per cent, so 2,800 participants and 2,200 others. The evaluation compares the two groups afterwards:

  AFTER    participants  $5,100      non-participants  $6,600
  the gap the study reports                            $1,500
  naive saving  1,500 x 2,800                      $4,200,000
  naive return                                       5.60 : 1

Now ask the question the method cannot ask.

  BEFORE   participants  $4,900      non-participants  $6,100
  the gap that already existed                         $1,200
  causal effect  1,500 - 1,200                           $300
  saving  300 x 2,800                                $840,000
  return                                             1.12 : 1

The naive figure is 5.00 times the difference-in-differences figure. And the general form makes the bracket arithmetic rather than opinion: naive / true = 1 + (baseline gap / causal effect), so a pre-existing gap two, three or four times the size of the effect overstates by 3.0, 4.0 and 5.0 times respectively. Three to five times is not a range somebody chose. It is what a selection effect of that size produces.

At the randomised point estimate the effect is $0, the return is 0.00 : 1, and the year's net is −$750,000. On a three-year contract, −$2,250,000.

Say the same thing a second way, because this is the form that ends the conversation in a committee. The published 3.27 : 1 on medical cost, applied to a $150 programme, asserts a saving of $490.50 per employee. The employer's own share of a single-coverage premium is $8,435 less a worker contribution of $1,401, so $7,034. The ratio therefore asserts that a hundred and fifty dollar programme removes:

  490.50 / 7,034                                      6.97 %

Seven per cent of the employer's entire medical cost, for $150 a head. Once it is stated that way, the randomised confidence intervals — tight enough to exclude 78.0 per cent of the previously published savings estimates — stop being a surprise.

And there is a second engine that runs even on people who did nothing. Target the highest-spending decile. With a population mean of $6,000, a standard deviation of $12,000 and a year-to-year correlation of 0.35, the expected value of a standard normal above the ninetieth percentile is 1.7550:

  the targeted decile, this year                     $27,060
  the same decile, next year, untouched              $13,371
  apparent saving                                    $13,689   =  50.6 %

Half the spending disappears by arithmetic. A programme that targets high spenders and measures itself against their own prior year will report a fifty per cent saving having done nothing at all, and it will do so every year.

Third — the honest negative, and it is this volume's own argument that pays it.

Chapter V.06 priced what it costs to hold a claim nobody can sell. A conventional cost of equity of 8.50 per cent, against a discount for lack of marketability of 25.0 per cent over 8 years — an annualised illiquidity premium of 3.66 points — gives a cost of equity of 12.16 per cent. Carried into a weighted average cost of capital at 40 per cent equity:

  conventional WACC                                   5.65 %
  the same firm, unsellable                           7.11 %
  the hurdle gap                                      1.46 points

Every project returning between 5.65 and 7.11 per cent creates value in the listed structure and destroys it in the employee-owned one. Same machine, same country, same customers. And the growth arithmetic is worse, because the conventional firm can issue equity and the employee-owned one can only admit members:

  conventional sustainable growth                    10.20 % a year
  the same firm, unsellable                           7.50 %
  the growth gap                                      2.70 points

  after ten years, the conventional base is larger by  28.2 %
  after twenty                                         64.2 %
  after thirty                                        110.5 %

A committee asked to move a quarter of the equity into a trust is being asked to accept that, and a paper that does not name it will be refused by the first person who models it. This volume's arithmetic does not dissolve that cost. It prices it, which is a different and more useful service, and Chapter V.06's Member Capital Bridge and Chapter V.03's three rules are what a firm does about it. For a mature business generating more cash than it can reinvest, 2.70 points of foregone growth is a price worth paying for the employment stability and the survival hazard; for a capital-hungry business in a growing market it is the business.

There is a second threshold worth naming in the same breath, because it also falls out of this volume's own numbers. Chapter V.09's capture ratio says that at a median tenure of 3.90 years — a hazard of 0.177730 — an employer holds 14.44 per cent of the long-run value of the health it creates, and therefore faces a hurdle multiple of 6.92. Under-investing in long-dated employee health is not a failure of corporate imagination. It is the correct answer to the arithmetic as the firm is currently constituted, and the only fix in this volume is the multi-employer pool, which requires agreement with competitors and therefore a conversation with counsel before it is a conversation with a treasurer.

---

DREAM

What becomes ordinary

In the company that has absorbed this, the people section of the board pack is the shortest it has ever been and the most believed.

Every proposal about people arrives with three lines on its front page: the endpoint, the number of employees required to detect the claimed effect, and the control group. The control group exists, because the company worked out some years ago that a staggered rollout is a free experiment and an unstaggered one is a free anecdote, and the wave order is now assigned at random as a matter of scheduling policy rather than of research ambition.

Nobody is embarrassed about the endpoints the company cannot measure. The list is published — medical spending at 50,233, engagement to performance at 32,000 — and it is used as a buying rule rather than as an apology. When a vendor's paper claims one of them, the conversation is over in ninety seconds and the vendor is asked to re-price against turnover, which the company clears by 1.59 and holds people to.

The capital paper template carries one extra line: expected duration of benefit, share retained by this entity, and who holds the rest. It is the capture question, asked automatically, by everyone, forever. When the answer is 14.44 per cent the next question is not whether to proceed but who the other party is, and about a third of the time the answer is the stop-loss carrier, the pension scheme or the trade — and those conversations were simply not available while everybody was pretending the firm owned the whole asset.

Compensation is benchmarked against the skill and not against the sector, and the finance function computes δ once a year and publishes it with its method attached. The number goes in the annual report beside the charitable donations because that is what it is: a donation made by the workforce. The first computation was uncomfortable and the method was attacked, which is how everyone knew it was measuring something.

The annual statement carries two numbers: the value of a participant's stake, and that stake as a share of their total retirement savings against a ceiling the plan holds itself to. The match on the savings plan is paid in cash. Nobody experiences the ceiling as a restriction; it reads the way a seatbelt reads.

And the monthly pack carries two columns, one flow and one stock, on the same page, with a written threshold that fires on a divergence and opens a named three-week enquiry. The threshold was published cold, before anybody knew which way it would fire first, and the first time it fired nobody moved it.

---

DESIGN

What to stop, and what to start

Eight things to stop, each retired by its own arithmetic.

Stop buying the medical-spend ROI. You cannot detect it — 50,233 employees, 10.05 times what the firm has — the randomised point estimate is 0.00 : 1, and a three-year contract at $150 a head across 5,000 people is −$2,250,000 of certain expenditure against an effect the best available trials could not find.

Stop the quartile contrast in the engagement deck. A top-versus-bottom quartile comparison is any correlation with its tails cut off. The mean of a standard normal's top quartile sits 1.2711 standard deviations above the overall mean, so the gap is 2.5422 standard deviations of the predictor; multiplied by an r of 0.22 that is 0.559 standard deviations. Every celebrated headline in the field is that number in a coat. Say 0.559 and you will still be persuasive, and you will survive the first analyst who checks.

Stop quoting a same-source correlation uncorrected. Where the predictor and the outcome come from the same person on the same form, method variance inflates the correlation by about 1.26. A published 0.40 becomes 0.317, and the variance it explains falls from 0.1600 to 0.1008 — 37.0 per cent of the explained variance was the form. It takes four seconds.

Stop adding presenteeism into a single total. At a salary of $60,000, a reported impairment of 5 per cent and a claimed relative reduction of 10 per cent, the same programme returns 1.00 : 1, 2.00 : 1 or 3.00 : 1 depending on a multiplier nobody defends in the appendix — a spread of 3.00 times. Absence is read from payroll and claims from the claim run. Report the three separately or report the hardest one alone.

Stop paying the savings-plan match in company stock. Chapter V.03 priced a stake at 35.4 cents on the dollar for a participant at 63.4 per cent concentration, and priced the unexercised diversification election at 32.1 cents — $41,867 for the worked participant, costing nothing and giving up no expected return. A match paid in shares is asking the employee to fund the ownership out of their own diversification, which is the one structure the evidence condemns without qualification.

Stop benchmarking mission-sector pay against the mission sector. The differential is already inside the benchmark, the answer comes back near zero, and the instrument has certified the thing it was built to find.

Stop leaving a published price unindexed. California's non-economic damages cap was set at $250,000 in 1975 and left alone; it is worth $45,951 in the money it was legislated in — 18.4 per cent of its original value. A number that is not indexed is a number somebody is lowering every year without ever having to vote for it, and that applies to a life price, a floor, a threshold and a cap alike.

Stop rolling anything out in a fixed order. This is the free one and it is the last on the stop list because it is really the first on the start list.

Nine things to start, four of them free.

Randomise the wave order. Almost every programme is deployed in waves for reasons of capacity. Assign the order at random and a scheduling constraint becomes a randomised controlled trial at a marginal cost of zero. The entire literature this chapter had to correct would not exist if this had been standard practice.

Put the measurability line in the procurement policy. One sentence. It retires more bad spending than any review.

Put the capture line in the capital paper template. Duration of benefit, share retained, who holds the rest. At 3.90 years of median tenure the share is 14.44 per cent and the hurdle multiple is 6.92.

Compute δ in finance and publish it. Chapter V.08's conservative 5.5 per cent and aggressive 17.0 per cent, against a replacement cost of 16.1 per cent of salary, give break-evens of 34.16 and 105.59 points of annual turnover, while the separation elasticities deliver 8.79 to 20.68 — short by 13.48 to 25.37 points. Paying the differential back does not close on churn alone, and the board paper must say so on the first page. Publish the number anyway; an unmeasured subsidy grows and a measured one gets argued about.

Write the concentration covenant into the plan document. At an ownership premium of 1.0, 1.5 and 2.0 points a year, Chapter V.03's ceiling is 7.63, 11.44 and 15.26 per cent of a participant's total retirement savings. Choose one, minute the premium you believe justifies it, and enforce it by redirecting future contributions rather than by forced sale.

Add the care exposure line and fund the employer's share. Chapter V.04's employer saving is verifiable on the firm's own P&L with no appeal to anything: a 6-point fall in the annual separation rate among parents of under-fives at a replacement cost of 21.4 per cent of a $70,000 salary is $898.80, plus 2 avoided absence days at $269 — $1,437 a year. At 12 per cent of 5,000 staff that is 600 people and $862,080. It is 10.78 per cent of the $13,333 cost of a quality place, and it is the only share in the whole structure that requires nobody to believe anything.

Pass the life-price resolution and add the cost-per-statistical-life-saved line. A guard-rail programme costing $2,400,000 that avoids 6.0 statistical deaths implies $400,000 per life, which is 33.00 times below the $13,200,000 the transport department writes into its own appraisals. The value of the line is that it makes the easy cases obvious, which is where most of the delay in safety spending actually lives.

Gate the divergence threshold. Two series doing nothing move in opposite directions half the time, so a trigger on sign alone fires twice running 25.00 per cent of the time. Require both moves to clear one standard deviation and the probability is 0.2534 per cent — an improvement of 98.64 times, bought with one sentence of drafting.

Buy the panel, not the cross-section. The wellbeing endpoint costs 4,392 employees as a cross-section and 879 as a panel — a factor of 5.00 — and the firm clears the panel by 5.69. At $27.95 per respondent per year, an honest, protocol-frozen series costs $24,569.06 annually, and the panel identifies within-person change, which is the comparison that survives the ordinal problem.

And the one rewrite that carries a real facility. Chapter V.02's working-time instrument converts structural overtime into straight time. Two hundred workers at $28.00 an hour, five overtime hours a week over forty-eight weeks:

  overtime hours                                48,000 h/yr
  overtime at time-and-a-half               $2,016,000
  straight-time equivalent                  $1,344,000
  the premium being paid                      $672,000
  heads to absorb it                                26
  the same hours, loaded at 1.32            $1,774,080
  wage-line saving — from payroll only        $241,920

Chapter V.02 added an injury saving of $100,800 at a 0.20 factor and a turnover saving of $51,300 at a 0.15 factor, reaching $394,020 and a return of 76.9 per cent on $512,650 of facility, verification and administration. Now do what a remuneration committee will do, and strike both soft lines out entirely:

  return on the wage line alone                      47.19 %
  against a WACC of                                   9 %
  clears by                                          38.19 points

  break-even fall in output per worker, all lines      4.10 %
  break-even on the wage line alone                    2.52 %

The instrument survives with the two contestable lines deleted, and the bet it rests on shrinks from 4.10 per cent to 2.52 per cent. Take that version to the committee. You will be asked about the other two lines, and the honest answer is that they are real and you have not counted them.

---

DESTINY

How it holds when the sponsor is promoted

Three fastenings, and each has a named way of failing.

It holds when the randomised rollout is a rule about scheduling. If the allocation is a clever thing an analyst did once, it dies with them. If it is one line in the deployment standard, it survives everyone and produces a control group every year without anybody advocating for it. The failure mode: a capacity emergency, a single unstaggered rollout, and a precedent.

It holds when the endpoints are owned by the functions that already report them. Turnover to the people function, claims to finance, injuries to operations. A people metric that lives in a wellbeing team's own report dies when the team is reorganised; one that lives in payroll is read every month for other reasons. The failure mode: a new dashboard that consolidates them, at which point all three become one team's property and the softest of them sets the tone.

It holds when a liability has been booked. Chapter V.08's fastening is the least sentimental in the volume and it applies to every instrument here: once the differential is an accrual, reversing the policy means writing back a liability, and writing back a liability requires an explanation to an auditor. Irreversibility is not willpower; it is double-entry. The failure mode: accruing more than the company can fund, which converts an honest measurement into a covenant breach and teaches everybody never to measure it again. Accrue what is affordable and publish the whole δ regardless.

And the failure this chapter exists to prevent. A genuine wellbeing effect is oversold as a cost effect; the cost effect does not arrive, because at 50,233 employees it could never have been seen; and the programme is cut for failing at something it never claimed. That is the most expensive failure in this field and it is caused entirely by the business case, never by the programme. The discipline is small and it is total: claim the endpoint you can detect, and fund the rest out of the same budget as the coffee, for the same reason, without asking anyone to believe it returned 3.27 : 1.

---

DELIGHT

What it feels like

There is a specific pleasure in being the person who asks what the gap was before. It is not a clever question and it takes no courage — about nine seconds — and the room goes quiet in a particular way, and the conversation that follows is better than the one that was scheduled.

Then a second pleasure, and it is the one committees do not expect. Once you have stopped defending a number that cannot be defended, you can be far more generous about the things that are simply good. The supervisor training. The shift somebody needed to move. The extra week for the man whose wife is ill. The care place. None of it has to be justified by a ratio that will not survive an analyst, and it turns out that a thing you can do plainly, because it is right and it costs little, is much easier to keep than a thing you have to keep proving.

And there is the arithmetic itself. $241,920 of overtime premium converted into straight time, on the payroll ledger, with nothing soft in it; $488,508 handed back to the P&L in the year the wellness contract ended; $1,437 a head of care funding that the finance function computed and nobody had to argue about. A committee that has done this once does not go back, because the papers got shorter and the decisions got faster and nobody had to perform sincerity to get a guard rail approved.

---

OPERATIONALIZE THIS

At the level of finance

The instrument: a verified people budget — a stop-start reallocation with a measurability covenant in the procurement policy.

It is not a new facility and it asks the board for no money. It moves an existing budget from endpoints the company cannot detect to endpoints that have survived a control group, and it hands the difference back.

What is released.

What is bought, and what each one returns.

  return on the reallocated budget                    2.44 : 1
  handed back to the P&L                              $488,508
  against what it replaced                            0.00 : 1

The one line in that table a committee will interrogate, and the honest answer. The cessation value is built from Berman and colleagues' cost of a smoking employee, and their five components are $517 of absenteeism, $462 of presenteeism, $3,077 of smoking breaks, $2,056 of health care and −$296 of pension excess — $5,816 in total. Two of those are soft by this chapter's own standard. So the table counts only the two a ledger can check, $2,573, and the 87.3 additional quitters return $224,623 at 1.56 : 1. Count the breaks and the presenteeism and the same programme returns $507,737 at 3.52 : 1. The first figure is the one that goes in the paper and the second is the one that makes it safe, and saying which is which is the entire competence this chapter is teaching.

The balance-sheet treatment. Nothing here is a reclassification. The programme costs run through operating expense exactly as they did; the experience-modifier saving lands in insurance cost; the disease-management saving lands in the self-insured claim run. The only accounting conversation is the disclosure, and it is worth having: a memorandum note stating, for each funded programme, the endpoint, the evidence class it rests on, and the sample required to detect it. That note changes no recognised number and it is what makes the policy hard to quietly abandon when the sponsor is promoted.

The counterparty. Internal for the reallocation, which is why it can be approved this quarter. External at two points: the stop-loss carrier, who already holds the tail of your claim distribution and is the one party with an aligned interest over a horizon longer than 3.90 years; and the employers' liability carrier, who prices a documented reduction in claims frequency because that is their whole business.

The number that decides it. One comparison, and it belongs on the front page of every people paper the board ever sees:

      employees required to detect the primary endpoint
     ---------------------------------------------------   <=  1.0
                     employees we have

At 5,000 people this passes for turnover (3,140), the wellbeing panel (879), absence (4,521) and the wellbeing cross-section (4,392), passes for injury frequency over four years (4,060 person-cohorts a year), and fails for engagement-to-performance (32,000) and medical spending (50,233). Six endpoints, four available, and the two that are not are the two the company has been buying.

The first ninety days.

---

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

1. Which people number in our pack came from an administrative system rather than a survey, and who built it? What made it trustworthy?
2. Think of a safety or health decision we took at real cost that everyone still agrees was right. How did it get approved, and what did the sponsor have that others did not?
3. Where in this company does somebody already compute a required sample size before believing a claim — clinical, actuarial, quality, marketing — and what have they learned that the rest of us have never been told?

Dream — what becomes possible

1. If every people paper carried its endpoint, its required N and its control group on the front page, which recurring argument would simply end?
2. Imagine the annual report carries one honest number about how we pay, on the page where we would rather not put it. What would we want it to say in three years?
3. If we knew exactly what share of the health we create we actually keep, which conversation would we open first, and with whom?

Design — what we build

1. What is being rolled out in waves in the next six months, and what would it cost us to assign the order at random?
2. Which claim in our current people budget requires more employees than we have, and who would be the right person to tell the vendor?
3. What is the smallest verified saving on our own P&L that would justify funding care places this year, without anybody having to believe anything?

Destiny — how it holds

1. What would have to be in the capital paper template for the capture question to be asked automatically, by everyone, forever?
2. What is the first sign we would see that a good programme was about to be cut for failing at a target it never claimed — and who would notice first?
3. If a successor committee kept exactly one thing we build this year, which would we want it to be, and what would make that one the easiest to keep?

---

WORKS CITED

Baicker, K., Cutler, D. and Song, Z. (2010). "Workplace Wellness Programs Can Generate Savings." Health Affairs, 29(2), 304–311.

Berman, M., Crane, R., Seiber, E. and Munur, M. (2014). "Estimating the Cost of a Smoking Employee." Tobacco Control, 23(5), 428–433.

Boushey, H. and Glynn, S. J. (2012). There Are Significant Business Costs to Replacing Employees. Center for American Progress.

Caloyeras, J. P., Liu, H., Exum, E., Broderick, M. and Mattke, S. (2014). "Managing Manifest Diseases, But Not Health Risks, Saved PepsiCo Money Over Seven Years." Health Affairs, 33(1), 124–131.

Cascio, W. F. (2006). "Decency Means More than 'Always Low Prices'." Academy of Management Perspectives, 20(3), 26–37.

Cote, J. A. and Buckley, M. R. (1987). "Estimating Trait, Method, and Error Variance." Journal of Marketing Research, 24(3), 315–318.

Damodaran, A. (2012). Investment Valuation, 3rd edn. Wiley.

Edmans, A. (2011). "Does the Stock Market Fully Value Intangibles? Employee Satisfaction and Equity Prices." Journal of Financial Economics, 101(3), 621–640.

England, P., Budig, M. and Folbre, N. (2002). "Wages of Virtue: The Relative Pay of Care Work." Social Problems, 49(4), 455–473.

Harter, J. K., Schmidt, F. L. and Hayes, T. L. (2002). "Business-Unit-Level Relationship Between Employee Satisfaction, Employee Engagement, and Business Outcomes." Journal of Applied Psychology, 87(2), 268–279.

IFRS Interpretations Committee. IFRIC 2, Members' Shares in Co-operative Entities and Similar Instruments.

Internal Revenue Code, 26 U.S.C. §§ 401(a)(28)(B) and 409(h).

Jones, D., Molitor, D. and Reif, J. (2019). "What Do Workplace Wellness Programs Do? Evidence from the Illinois Workplace Wellness Study." Quarterly Journal of Economics, 134(4), 1747–1791.

Kaiser Family Foundation (2023). Employer Health Benefits Annual Survey.

Kish, L. (1965). Survey Sampling. Wiley.

Levine, D. I., Toffel, M. W. and Johnson, M. S. (2012). "Randomized Government Safety Inspections Reduce Worker Injuries with No Detectable Job Loss." Science, 336(6083), 907–911.

Loan Market Association, Asia Pacific Loan Market Association and Loan Syndications and Trading Association (2023). Sustainability-Linked Loan Principles.

Meulbroek, L. K. (2005). "Company Stock in Pension Plans: How Costly Is It?" Journal of Law and Economics, 48(2), 443–474.

Pauly, M. V., Nicholson, S., Polsky, D., et al. (2008). "Valuing Reductions in On-the-Job Illness." Health Economics, 17(4), 469–485.

Pencavel, J. (2015). "The Productivity of Working Hours." The Economic Journal, 125(589), 2052–2076.

Pérotin, V. (2016). What Do We Really Know About Worker Co-operatives? Co-operatives UK, Manchester.

Song, Z. and Baicker, K. (2019). "Effect of a Workplace Wellness Program on Employee Health and Economic Outcomes." JAMA, 321(15), 1491–1501.

Song, Z. and Baicker, K. (2021). "Health and Economic Outcomes Up to Three Years After a Workplace Wellness Program." Health Affairs, 40(6), 951–960.

US Bureau of Labor Statistics. Employee Tenure Summary; and Occupational Employment and Wage Statistics.

US Department of Transportation. Departmental Guidance on Valuation of a Statistical Life in Economic Analysis.

Volpp, K. G., Troxel, A. B., Pauly, M. V., et al. (2009). "A Randomized, Controlled Trial of Financial Incentives for Smoking Cessation." New England Journal of Medicine, 360(7), 699–709.

Note on figures. Every figure in this chapter is computed in lib/verify/V_E2.py and prints with its inputs, its units and its source; run python3 lib/verify.py V.E2. Figures carried from Chapters V.01 to V.11 are recomputed here rather than quoted. The buying order in the measurability frontier is this chapter's own assertion and is printed as STATED in the module; the Kendall tau computed on it is arithmetic performed on an assertion, and a reader who ranks the six endpoints differently will get a different tau and should. The injury base rate, the absence dispersion, the business-unit size, the survey cost per head, the prevalences of a diagnosed condition and of smoking, the safety programme's cost and the share of staff who are parents of under-fives are stated assumptions, printed as such. Where Chapter V.09's sample sizes are quoted at two decimal places of z and this module computes them by bisection, both are printed and the difference is the rounding, not a disagreement.