Haute Lumière

Commerce · V.08 · MMXXVI · daylight

La Bourse  /  Volume V  /  Nº V.08

Meaning as an Economic Input

Volume V — Labour, Value, Flourishing

---

THE PLATE

---

THE LETTER

There is a version of this chapter that is a poster. It has a photograph of a sunrise on it, it says that people want their work to matter, and it is true, and it is worth nothing, because everybody already knows it and nobody can price it.

This is the other version. The claim here is narrow and it is arithmetical: meaning behaves like an input. It is supplied by the employer at some cost, it raises output by a measurable amount, it is valued by the worker at a measurable amount, and the two amounts are not the same — which is where all the interesting economics lives. Treat it as an input and you can ask input questions of it. What does a unit cost. What is the marginal return. Who captures the surplus. At what point does more of it stop helping.

Every one of those questions has an answer in the literature, and several of the answers are uncomfortable. The most uncomfortable is the one this chapter is built around, so it may as well be said in the first hundred words: the best evidence that meaningful work is valuable is the evidence that people accept less money to do it — and that same evidence is the mechanism by which the most meaningful sectors of the economy pay the worst. One coefficient, two readings. The finding that dignifies care work is the finding that underpays it.

You will not be asked to choose between those readings. You will be asked to compute the coefficient, look at both faces of it, and then decide what an employer who does not wish to collect it should do instead — which turns out to be a specific instrument with a specific accounting treatment, and not a sentiment.

A note on the evidence, because this is the chapter of the edition most exposed to motivated reading. Where a study asked a person about their own motivation and then asked the same person about their own performance, the effect is discounted here, and the size of the discount is computed. Where a study measured output at the till, the effect stands as measured. Those two literatures disagree by roughly a factor of two, and the smaller number is the one this chapter uses.

— The Editors

---

DISCOVERY

What is already working

Start where it is already being done well, and note what the good cases have in common: in every one of them, meaning was supplied deliberately, at a stated cost, and the return was measured against something other than a questionnaire.

Adam Grant's fundraisers, at a large public university. Callers in a scholarship call centre were arranged to spend five minutes with a student whose scholarship their work had funded. Five minutes. One conversation. In the month that followed, the group who had met the student earned 142 per cent more weekly pledges than the control group and brought in 171 per cent more weekly revenue — multipliers of 2.42 and 2.71 on a line that the organisation had been trying to move for years with scripts and incentives. The control group did not change.

The same researcher ran the mechanism in the other direction with lifeguards, who read brief accounts of rescues that had actually happened at pools like theirs. Hours worked rose 43 per cent; helping behaviour rose 21 per cent. The intervention in both cases was information about a beneficiary, it cost essentially nothing, and it worked on output rather than on mood.

Buurtzorg, in the Netherlands. Neighbourhood nursing, founded in 2006 by Jos de Blok on a structural bet rather than a motivational one: self-managing teams of no more than twelve nurses, holding their own caseload, their own rota and their own client relationships, with a very thin central office and no supervisory layer above the team. The Commonwealth Fund case study and the KPMG evaluation both report the same headline — Buurtzorg delivers its outcomes using roughly 40 per cent fewer hours of care per client than comparable providers, which is 1.67 times the client-outcome per hour of nursing. It has also been voted the best employer in the country repeatedly. The design gave nurses back the judgement their profession is made of, and the hours fell out of the model as a consequence rather than a target.

Costco, in the American warehouse trade. Wayne Cascio's comparison against the nearest competitor is the case to carry, because it is the case of an employer who declined to collect the discount. Costco paid a median 17.00 dollars an hour against the competitor's 10.11 — a premium of 68.2 per cent. Annual turnover ran at 17 per cent across all staff and 6 per cent after the first year, in a sector where the figure is routinely several times that. And sales per employee were 795,000 dollars against 516,000: a ratio of 1.54. The firm that paid more per head sold half again as much per head.

Mayo Clinic's academic physicians. Shanafelt and colleagues asked a question nobody had asked with a dose-response design: what share of a doctor's working time is spent on the activity that doctor personally finds most meaningful? Below a threshold of 20 per cent of working time, burnout ran at 53.8 per cent. At or above it, 29.9 per cent — a difference of 23.9 percentage points and a relative risk of 1.80.

And then the detail that matters more than the headline: above 20 per cent, there is no further benefit. The curve plateaus. More meaning, past the threshold, buys nothing.

That is the shape of a real input and it is why this case closes the movement. A slogan does not have a threshold. A slogan does not have diminishing returns, a ceiling, or a point beyond which more of it is waste. Meaning has all three, which is precisely what licenses the rest of this chapter to treat it as a factor of production rather than as a mood.

---

THE ARITHMETIC

What works, what does not, and what one coefficient is doing on both sides of the table

First, the engagement literature, at its honest size.

The largest business-unit meta-analysis of engagement and performance — Harter, Schmidt and Hayes, across 7,939 business units in 36 companies — reports a true-score correlation of 0.22 between engagement and a composite of business outcomes. That figure is credible precisely because the performance data comes from the company's own systems rather than from the same people who filled in the survey.

Square it. An r of 0.22 explains 4.84 per cent of the variance in business-unit performance. Which leaves 95.16 per cent explained by something else.

That is not a debunking. Four point eight per cent of the variance in performance, from a construct you can move at low cost, is a large commercial prize. It is a debunking only of the presentation, and the presentation is worth one paragraph because you will meet it in every deck you are ever shown.

Engagement research is nearly always reported as a contrast between the top and the bottom quartile: 23 per cent higher profitability, 18 per cent higher productivity, 81 per cent lower absenteeism, 64 per cent fewer safety incidents, 10 per cent higher customer loyalty. Those are real numbers from the source. But a top-versus-bottom-quartile contrast is what any correlation looks like when you cut its tails off and compare them. For a normal distribution, the mean of the top quartile sits 1.2711 standard deviations above the overall mean, so the gap between top and bottom quartile is 2.5422 standard deviations of the predictor. Multiply by r:

  quartile gap  =  0.22 × 2.5422  =  0.559 standard deviations

Every celebrated quartile headline in this field is that number wearing a coat. Half a standard deviation is a genuinely useful effect and it is not a transformation of the firm. Say the true thing and you will still be persuasive, and you will survive the first analyst who checks.

Second, the discount on everything measured by questionnaire.

Where both the predictor and the outcome come from the same person on the same form, some of the correlation is the form. Cote and Buckley, across 70 studies, put method variance at 26.3 per cent of measured variance; Doty and Glick put the resulting inflation of a same-source correlation at about 26 per cent. So take a typical single-source finding — meaningful work correlated at 0.40 with self-rated performance — and correct it:

  corrected r   =  0.40 / 1.26          =  0.317
  variance      =  0.1600  ->  0.1008
  method's share of the explained variance  =  37.0 %

The correlation falls by about a quarter. The variance it explains falls by 37.0 per cent, because the square is where the damage lands. This is the single most useful correction in the chapter and it takes four seconds.

It also explains a discrepancy you would otherwise have to argue about. The self-report literature — Christian, Garza and Slaughter — puts engagement against task performance at 0.43. The independent-outcome literature puts it at 0.22. The ratio is 1.95. The field that asks people to grade themselves reports an effect almost exactly twice the size of the field that reads the till.

Third, what meaning does to effort, measured in units rather than opinions.

Ariely, Kamenica and Prelec built the cleanest experiment in this literature. Participants assembled models for a piece rate that started at 2.00 dollars and fell by 0.11 dollars per unit, so each person revealed the wage at which they stopped. In one condition the completed work was set aside; in the other it was dismantled in front of them, in view, immediately. Output: 10.6 units against 7.2, a difference of 47.2 per cent.

Now read the same experiment as a wage. The marginal rate at which each group quit is the rate at the unit where they stopped:

  meaningful condition   2.00 − 0.11 × 10.6  =  $0.834
  dismantled condition   2.00 − 0.11 ×  7.2  =  $1.208
  the discount                                 $0.374  =  31.0 %

The people whose work was preserved did 47.2 per cent more of it for a reservation wage 31.0 per cent lower. Those two sentences are one measurement. Everything that follows is about who gets to keep the difference.

Fourth, self-determination theory, sized honestly. Deci, Koestner and Ryan's meta-analysis is the most-cited evidence that pay can displace intrinsic motivation, and the effect is real and replicated: expected tangible rewards reduce free-choice persistence at d = −0.34, while verbal recognition raises it at d = +0.33. Convert the first to a correlation — r = 0.168 — and it explains 2.81 per cent of the variance. Cerasoli, Nicklin and Ford, across 183 studies and 212,468 participants, find intrinsic motivation predicting the quality of work and incentives predicting its quantity, with the two interfering when pay is tied tightly to output. Crowding-out is a reason to design pay carefully. It has never been a reason to pay less, and it is quoted as one constantly.

Fifth — the hardest number in the chapter. What people pay for meaning.

A compensating differential is the wage a worker gives up for a non-wage feature of a job. It is the oldest idea in labour economics and one of the hardest to measure, and the chapter's whole argument rests on it, so here is the estimate built in the open.

Start with what a raw gap is not. Against the May 2023 US median annual wage of 48,060 dollars, home health and personal care aides earn 33,530, which is 30.2 per cent below; childcare workers 30,370, or 36.8 per cent below; nursing assistants 38,130, or 20.7 per cent below; preschool teachers 37,130, or 22.7 per cent below. None of that is the differential. It is education, hours, experience and occupational crowding first, and only then anything to do with meaning.

The differential itself has to be estimated with the observable characteristics held down. Four independent attempts, each measuring a slightly different object:

The teacher figure repays a moment: the weekly wage penalty was 26.4 per cent and the total compensation penalty 17.0 per cent, the 9.4-point difference being benefits. Quote the wage penalty alone and you have overstated by nine points, which is how this number is usually quoted.

Sorted, the three career-job estimates are 5.5, 8.0 and 17.0 per cent: a median of 8.0 and a mean of 10.17. This chapter carries the conservative end, 5.5 per cent, as δ, and carries 17.0 per cent alongside it as the aggressive case. The 44 per cent figure is deliberately excluded from δ — it is a short online task with no career at stake, and it is an upper bound on a different object. Naming what an estimate cannot cover is part of the estimate.

And here is the cut.

Write the worker's decision. Let W_c be the wage available in a comparable job without the meaning, δ the fraction of wage forgone, and m the money-value the worker places on the meaning:

  the worker accepts   iff   m  ≥  δ · W_c
  the employer saves                δ · W_c

They are the same term. The worker's minimum valuation of meaning is identically the employer's maximum saving from supplying it. There is no version of the finding "people value meaningful work" that is not also the finding "an employer can pay less for it," because they are one equation read from opposite sides of a table.

And the equilibrium is worse than the equation. Where meaningful jobs are rationed and applicants are plentiful — which describes care, teaching, conservation, the arts and most of the non-profit sector — competition among workers drives δ upward until δ · W_c approaches m. At that limit the worker keeps none of the value of the meaning. It has all been competed into the wage, and the sector looks exactly as it looks: full of people who will tell you the work is worth it, paid as though the work were worth less.

Now the honest negatives, and there are two.

The first is the break-even. Take a 600-person home-care agency at the median wage of 33,530 dollars. At δ = 5.5 per cent, the skill-matched benchmark is 35,481.48 dollars, the differential is 1,951.48 per worker, and across the agency it is 1,170,888.89 dollars a year — an unrecorded transfer from the workforce to the mission that appears nowhere in the accounts.

Suppose the agency gives it back. How much less turnover must follow before the giving-back has paid for itself? Replacement cost runs at 16.1 per cent of annual salary for low-wage roles, and nursing-home staff turnover runs at a mean of 128 per cent and a median of 94 per cent annually:

  break-even turnover reduction  =  δ / c  =  5.5 / 16.1  =  34.16 points
  94 %  ->  59.84 %

Thirty-four points, from 94 per cent to 59.84 per cent — a workforce that still replaces itself roughly every twenty months. The bar is not excellence; the bar is merely bad. But the firm-level separation elasticities in the monopsony literature run about 1.7 to 4.0, which turn a 5.5 per cent raise into 8.79 to 20.68 points of turnover recovered — short of break-even by 25.37 to 13.48 points. Paying the differential back does not pay for itself on churn alone. The residual has to be found in quality, in client retention or in price, and a chapter that pretended otherwise would be the poster.

At the aggressive δ of 17.0 per cent it is worse and cleanly so: the break-even is 105.59 points of turnover against the 94 that exist. Eliminating every single separation would not repay a teacher-sized differential. That is the threshold below which this argument fails, stated as a number.

The second honest negative is about the number itself. Compensating differentials are notoriously hard to estimate, and estimated wrong-signed more often than not, because the workers who take pleasant jobs also tend to be the workers with the better outside options — so unobserved ability loads onto the amenity and the coefficient comes back saying that good conditions pay more. Brown established the problem; Hwang, Reed and Hubbard showed how badly unobserved productivity heterogeneity biases these estimates. The hardest number in this chapter is also the one most likely to be wrong, and the correct response is to carry a range, name the method, and never let a single point estimate into a pay decision.

---

DREAM

What becomes ordinary

In the organisation that has taken this seriously, the pay committee and the mission committee do not share a paper.

That is the whole of it, and everything else follows. Compensation is benchmarked against the skill, not against the sector — a nurse is priced against what that nurse's training and hours command anywhere, and the fact that the work is beloved is not an input to the calculation, because it was established years ago that treating it as one is a way of charging staff for the privilege of caring.

The differential is still there. It is simply visible. Once a year the finance function computes δ the way it computes any other variance: median pay against a skill-matched external benchmark, published, with its method attached. The number appears in the annual report beside the charitable donations, because it is one — a donation made by the workforce, and it is described in exactly those words. Nobody finds this awkward. A number that is named stops being free, and a subsidy that has a line in the accounts acquires a committee that argues about its size.

Meaning itself is supplied on purpose and costed like any other input. Contact with the beneficiary is scheduled rather than hoped for, because the effect of five minutes has been measured and nobody is willing to leave a return of that size to chance. Time on the work a person finds most meaningful is tracked against the 20 per cent threshold, and when it drops below, that is a staffing finding, not a wellbeing one.

Nobody claims more for it than it does. The engagement number in the pack is reported as a correlation with its variance stated, and when a consultant arrives with a quartile contrast somebody in the room converts it to standard deviations before the second slide. This is not scepticism as a posture. It is the ordinary competence of an organisation that intends to still believe its own numbers in five years.

And the people doing the work can say what the arrangement is. Not that they are valued — that they are paid X, that the market rate for their skill is Y, that the difference is δ, that δ is accruing to them in an account with a vesting date, and that if they leave before that date they know exactly what they are leaving. The meaning is real and it is not the payment. Having both sentences in the open is the entire dream.

---

DESIGN

The structure that gets there

Four moves, in order. Each is available to a single operator without a change in law, ownership or anybody's convictions.

One. Separate the two pricing questions, structurally.

Meaning may never enter the model that sets pay. In practice this means the compensation benchmark is drawn from the skill-matched labour market rather than the sector: what does this bundle of training, licensure, hours and physical demand earn in the whole economy, not in the part of it that shares your mission? Where a sector benchmark is the only data available, use it and state the contamination, because a sector benchmark in a mission sector has the differential already inside it and will reproduce it forever.

The governance form is blunt and it works: the committee that approves pay does not receive mission materials with the paper, and the paper carries the external benchmark on the first page.

Two. Compute δ annually and publish it.

  δ  =  1  −  (median internal pay  /  skill-matched external benchmark)

One number per job family, one method note, published internally at minimum. The purpose is not shame; it is that an unmeasured subsidy grows and a measured one gets argued about. Expect the first computation to be uncomfortable and expect the method to be attacked; both are signs it is measuring something. Hold a range rather than a point, for the reason the Arithmetic gave.

Three. Turn the differential into a claim rather than a donation.

If staff are funding the mission by δ, the honest structure is that they hold an asset for it. That is the instrument in the next movement, and it is the only part of this design that requires finance to do anything.

Four. Supply meaning deliberately, and stop at the threshold.

The supply side is cheap and it is nearly always under-provided:

• Beneficiary contact, scheduled. Not a story in a newsletter — five minutes with a person, on the rota, quarterly. The measured effects were 142 and 171 per cent, on revenue, in a call centre.
• Task significance made legible. Lifeguards moved 43 per cent on hours from reading what had happened at pools like theirs.
• Judgement returned to the team. Buurtzorg's 40 per cent is not a motivation programme; it is the removal of a supervisory layer that had been spending nurses' judgement on its own behalf.
• Time on the meaningful activity, protected to 20 per cent — and no further, because past the plateau you are spending money for nothing and the budget is better used on pay.

Sequence matters. Do the pay separation first. An organisation that improves the supply of meaning before it has fixed the pricing question has, in the strictest arithmetical sense, just increased the amount it can extract.

---

DESTINY

How it holds when nobody is pushing

The structure holds on three fastenings, and it fails on three, and they are the same three.

It holds because δ is in the reporting pack, computed by the finance function rather than by the people who benefit from the answer. A number owned by human resources drifts toward a narrative. A number owned by finance drifts toward a method, and a method survives the departure of its author.

It holds because the accrual is a liability. This is the fastening that does the most work and it is the least sentimental: once the differential is booked as an obligation, reversing the policy means writing back a liability, and writing back a liability requires an explanation to an auditor. Irreversibility here is not willpower. It is double-entry.

It holds because the workforce can compute it. The single most robust governance mechanism available is a number that several hundred people know how to check.

Now the failure modes, named. It fails when δ is computed against a sector benchmark, because the answer comes back near zero and everybody relaxes — the differential is inside the benchmark, and the instrument has certified the thing it was built to find. It fails when the accrual is set at a level the organisation cannot fund, which converts an honest measurement into a covenant breach and teaches everybody involved never to measure it again; the discipline is to accrue what is affordable and publish the whole δ regardless of what is accrued. And it fails — most often — when meaning is improved first and pay second, because the organisation discovers that its people will accept the improvement in place of the money, and at that point the mechanism has been pointed backwards and is working perfectly.

---

DELIGHT

What it feels like

The pleasure of this one is unusual because it arrives as relief rather than as pride.

For the worker it is the end of a particular kind of private arithmetic — the one that runs at three in the morning and asks whether staying is devotion or foolishness, with no figure available on either side. Put δ on a page and the question stops being a character test. It becomes a decision about money, taken by an adult, with the number in front of them. People who have had that number handed to them describe the feeling as being taken seriously, which is a different and better thing than being appreciated.

For whoever computes it there is the specific satisfaction of an instrument that reads a room correctly for the first time. The sector has been saying we do this for love for a century, and love has been doing the accounting. Now love does the work and the accounts do the accounting, and each is better at its job.

And there is the small pleasure of a threshold. Twenty per cent. Not more, not a philosophy of work, not a transformation — a fifth of a working week spent on the thing a person is actually for, with a measured effect and a measured ceiling. Something that finite can be arranged by Thursday.

---

OPERATIONALIZE THIS

At the level of finance

The instrument: the Mission Differential Account. A book-entry deferred compensation obligation, funded from the measured wage differential, vesting to the people who supplied it.

The structure. Each year the firm computes δ against a skill-matched external benchmark, per job family. It accrues δ × W × N as a long-term employee benefit obligation, credited pro rata to each employee's account, vesting after a stated period. On vesting it is paid in cash, in equity, or into a retirement vehicle, at the employee's election. On departure before vesting, the employee takes the vested portion and the balance returns to the pool rather than to the firm — a forfeiture that funds other people's accounts and not the P&L, which is the clause that stops the instrument becoming a retention handcuff dressed as a gift.

The worked case. The 600-person agency above:

Note the arithmetic that catches people out: δ is 5.5 per cent of the benchmark, which is 5.82 per cent of the current wage bill. Quote the wrong denominator in a board paper and you will be corrected in public.

The balance-sheet treatment. An accrued compensation obligation — other long-term employee benefits under IAS 19 or deferred compensation under ASC

1. Not a provision and not a contingency: it is owed, it is measurable, and it

vests. The P&L charge lands in staff costs, which is where it belongs, because that is what it is. Your auditors will want the benchmark method documented and consistently applied; give them that in year one and the treatment is uncontroversial thereafter.

The counterparty. Internal first — the obligation sits on the firm's own balance sheet. At scale, or where the covenant position is tight, settle it into an employee benefit trust with an independent trustee, which converts an internal promise into an external claim and is worth doing before anybody asks.

The first ninety days.

The number that decides it. Not δ. This one:

              δ
   ─────────────────────   compared with the turnover reduction
    replacement cost c      your sector can actually deliver

At δ = 5.5 per cent and c = 16.1 per cent, the break-even is 34.16 points of annual turnover, and the elasticities deliver 8.79 to 20.68. The instrument does not close on churn alone, and the board paper must say so on the first page. Close the residual on the things the same literature supports — clinical quality, client retention, error rates, the price the sector's best providers can command — or present it honestly as a distribution of value that the firm has decided to make. Both are defensible positions. Only one of them is defensible after an analyst has run the numbers themselves, and it is whichever one you said first.

---

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

1. When has someone here met the person their work was for, and what changed in the weeks afterwards? Describe the change in something we count.
2. Which team in this organisation has the most judgement over its own work, and what does it get right that the others do not?
3. Think of the part of your week you would protect if everything else were cut. What proportion of your time is it now — and when was it most?

Dream — what becomes possible

1. If our people could each state the wage they are forgoing to work here, and state it without embarrassment, what conversations would become possible that are not possible now?
2. Imagine our annual report carries one number about how we pay, on the page where we would rather not put it. What is the number, and what would we want it to say in three years?
3. If five minutes of beneficiary contact moved our most important line the way it moved a call centre's, what would we schedule on Monday?

Design — what we build

1. Which of our pay benchmarks has our own sector inside it, and where would we get a clean one?
2. What would it take to give one team the judgement Buurtzorg gives a team of twelve — and which team would we choose first?
3. Where are we already above the 20 per cent threshold, and what is protecting it there that we could copy?

Destiny — how it holds

1. Who would compute δ if the person who built it left, and what would they need in writing to get the same answer?
2. What would be the first sign that we were improving meaning and quietly charging for it — and who in this room would see it first?
3. What would we want to still be true about how this organisation pays, ten years after everyone here has moved on?

---

WORKS CITED

Allegretto, S. (2023). Teacher Pay Penalty Still Looms Large: Trends in Teacher Wages and Compensation Through 2022. Economic Policy Institute.

Ariely, D., Kamenica, E. and Prelec, D. (2008). "Man's Search for Meaning: The Case of Legos." Journal of Economic Behavior & Organization, 67(3–4), 671–677.

Brown, C. (1980). "Equalizing Differences in the Labor Market." Quarterly Journal of Economics, 94(1), 113–134.

Burbano, V. C. (2016). "Social Responsibility Messages and Worker Wage Requirements: Field Experimental Evidence from Online Labor Marketplaces." Organization Science, 27(4), 1010–1028.

Cascio, W. F. (2006). "Decency Means More than 'Always Low Prices': A Comparison of Costco to Wal-Mart's Sam's Club." Academy of Management Perspectives, 20(3), 26–37.

Cerasoli, C. P., Nicklin, J. M. and Ford, M. T. (2014). "Intrinsic Motivation and Extrinsic Incentives Jointly Predict Performance: A 40-Year Meta-Analysis." Psychological Bulletin, 140(4), 980–1008.

Christian, M. S., Garza, A. S. and Slaughter, J. E. (2011). "Work Engagement: A Quantitative Review and Test of Its Relations with Task and Contextual Performance." Personnel Psychology, 64(1), 89–136.

Cote, J. A. and Buckley, M. R. (1987). "Estimating Trait, Method, and Error Variance: Generalizing Across 70 Construct Validation Studies." Journal of Marketing Research, 24(3), 315–318.

Deci, E. L., Koestner, R. and Ryan, R. M. (1999). "A Meta-Analytic Review of Experiments Examining the Effects of Extrinsic Rewards on Intrinsic Motivation." Psychological Bulletin, 125(6), 627–668.

Doty, D. H. and Glick, W. H. (1998). "Common Methods Bias: Does Common Methods Variance Really Bias Results?" Organizational Research Methods, 1(4), 374–406.

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

Folbre, N. (2012). For Love and Money: Care Provision in the United States. Russell Sage Foundation.

Gandhi, A., Yu, H. and Grabowski, D. C. (2021). "High Nursing Staff Turnover in Nursing Homes Offers Important Quality Information." Health Affairs, 40(3), 384–391.

Grant, A. M. (2008). "The Significance of Task Significance: Job Performance Effects, Relational Mechanisms, and Boundary Conditions." Journal of Applied Psychology, 93(1), 108–124.

Grant, A. M., Campbell, E. M., Chen, G., Cottone, K., Lapedis, D. and Lee, K. (2007). "Impact and the Art of Motivation Maintenance: The Effects of Contact with Beneficiaries on Persistence Behavior." Organizational Behavior and Human Decision Processes, 103(1), 53–67.

Gray, B. H., Sarnak, D. O. and Burgers, J. S. (2015). Home Care by Self-Governing Nursing Teams: The Netherlands' Buurtzorg Model. The Commonwealth Fund.

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

Hwang, H., Reed, W. R. and Hubbard, C. (1992). "Compensating Wage Differentials and Unobserved Productivity." Journal of Political Economy, 100(4), 835–858.

Judge, T. A., Thoresen, C. J., Bono, J. E. and Patton, G. K. (2001). "The Job Satisfaction–Job Performance Relationship: A Qualitative and Quantitative Review." Psychological Bulletin, 127(3), 376–407.

Mas, A. and Pallais, A. (2017). "Valuing Alternative Work Arrangements." American Economic Review, 107(12), 3722–3759.

Podsakoff, P. M., MacKenzie, S. B., Lee, J.-Y. and Podsakoff, N. P. (2003). "Common Method Biases in Behavioral Research." Journal of Applied Psychology, 88(5), 879–903.

Rosen, S. (1986). "The Theory of Equalizing Differences." In Ashenfelter, O. and Layard, R. (eds), Handbook of Labor Economics, Vol. 1. North-Holland.

Ruffini, K. (2024). "Worker Earnings, Service Quality, and Firm Profitability: Evidence from Nursing Homes and Minimum Wage Reforms." Review of Economics and Statistics, 106(1), 68–83.

Shanafelt, T. D., West, C. P., Sloan, J. A., Novotny, P. J., Poland, G. A., Menaker, R., Rummans, T. A. and Dyrbye, L. N. (2009). "Career Fit and Burnout Among Academic Faculty." Archives of Internal Medicine, 169(10), 990–995.

Sokolova, A. and Sorensen, T. (2021). "Monopsony in Labor Markets: A Meta-Analysis." ILR Review, 74(1), 27–55.

US Bureau of Labor Statistics (2024). Occupational Employment and Wage Statistics, May 2023. Washington, DC.

Note on figures. Every number above is computed in lib/verify/V_08.py and printed with its inputs and its source: run python3 lib/verify.py V.08. Where this chapter carries a range rather than a point — δ above all — the range is the finding, and a single point estimate of a compensating differential should not be allowed into a pay decision.