Haute Lumière
Commerce · V.09 · MMXXVI · daylight
ASSESSMENT · Chapter V.09 — Health as a Return on Capital
Three instruments: a ten-point quiz, eight reflection questions, five essay prompts. The quiz checks comprehension rather than recall. The reflections are private and first-person. The essays are arguable from more than one side.
THE QUIZ — ten points
Four on recall.
1. State what the two large randomised workplace-wellness trials found, and name the one durable effect.
The Illinois Workplace Wellness Study randomised 12,459 employees — 4,834 to treatment, 7,625 to control — and found no significant effect on medical spending, health behaviours, productivity or self-reported health at one year, and none on clinical markers, spending or absenteeism at two years. The BJ's Wholesale Club trial cluster-randomised 160 worksites, 20 against 140, covering 32,974 employees, and found significantly higher self-reported exercise and weight management with no significant difference in clinical measures, spending, utilisation, absenteeism, tenure or job performance at eighteen months or at three years. One mark for the nulls, one for naming the durable effect: beliefs, and having a primary care physician. The Illinois confidence intervals exclude 78.0 per cent of the previously published estimates.
2. Write the definition of a quality-adjusted life year and of a disability-adjusted life year, and state the relation between them.
QALY = sum over health states of time × utility, with 1.00 full health, 0.00 dead, negatives permitted, discounted like any stream. DALY = YLL + YLD, where YLL is deaths × reference remaining life expectancy at age of death and YLD is prevalence × disability weight, with 0.00 full health. The mark for the relation is for saying there is not a clean one: QALYs are gained and DALYs averted, the weights are elicited by different methods on different populations, and one minus a utility is not a disability weight.
3. Name three things with a defensible return in this literature, with the figure attached to each.
Smoking cessation paid in real money: 14.7 per cent against 5.0 per cent, a difference of 9.7 points and a number needed to treat of 10.31. Disease management for the already diagnosed: 3.78 : 1, against 0.48 : 1 for lifestyle management in the same seven-year programme — a difference of 7.88 x, averaging to the quoted 1.46 : 1. Randomised safety inspection: 9.4 per cent fewer injuries, 26.0 per cent lower compensation cost, 355,000 USD per firm over four years, with no detectable job loss.
4. What is the capture ratio, and what does it do to a hurdle rate?
The share of a long-run benefit the employer retains,
r / (r + λ)withλ = ln 2 / median tenure. At 3.90 years and a real rate of 0.03 it is 14.44 per cent, so the hurdle multiple is 6.92 x — the investment must be that many times better than society needs it to be before the firm's own arithmetic approves it.
Four on application.
5. A vendor slide says "3.27 : 1." What is the first question you ask, and what is the second?
First: what was the gap between participants and non-participants before the programme started? Second: what was your control group, and how was it chosen? Full marks require noticing that the ratio asserts a 150.00 USD programme removes 490.50 USD of medical cost — 6.97 per cent of a 7,034 USD employer bill — which is the size of claim the randomised intervals exclude.
6. Your programme enrols the highest-spending decile and reports a large first-year saving. Why is that not yet evidence?
Regression to the mean. With a mean of 6,000 USD, a standard deviation of 12,000 USD and a correlation of 0.35, a top decile averaging 27,060 USD is expected to average 13,371 USD next year with no intervention — an apparent saving of 13,689 USD a head, or 50.6 per cent, from arithmetic alone.
7. Your firm has 5,000 employees. Which endpoint do you buy the programme on, and why not the other one?
Turnover. Detecting a 300 USD change in annual spending needs 25,088 per arm, 50,176 in total — 10.04 x more people than the firm has. Detecting a four-point fall in a twenty per cent turnover rate needs 1,568 per arm, 3,136 in total, which the firm clears 1.59 x over. The stronger answer adds the rule: do not buy a claim about an endpoint you could not have detected.
8. A wellbeing programme improved burnout and turnover and did not lower claims. Is it a failure?
Only against a promise it should never have been given. STAR was a genuine randomised trial that moved schedule control, burnout, distress and voluntary turnover, and did not demonstrate medical savings; Oregon moved depression by 30.0 per cent and moved blood pressure, cholesterol and glycated haemoglobin by nothing measurable at two years. The examinable point: the most expensive failure in this field is a good programme cut for missing a target its business case invented.
Two that require the arithmetic to be done.
9. A vendor serving 8,000 employees at 180.00 USD a head reports that participants averaged 4,400 USD of claims against 6,200 USD for non-participants, at 60.0 per cent participation. Your analyst finds the same two groups averaged 4,300 USD and 5,700 USD in the year before the programme. Compute the naive return, the difference-in-differences return, and the ratio between them.
Total cost 8,000 × 180.00 = 1,440,000 USD. Participants 4,800. Post gap 6,200 − 4,400 = 1,800 USD; pre gap 5,700 − 4,300 = 1,400 USD; causal effect 400 USD. Naive: 1,800 × 4,800 / 1,440,000 = 6.00 : 1. Difference-in- differences: 400 × 4,800 / 1,440,000 = 1.33 : 1. Ratio 4.50 x. Credit any method reaching the same three figures. Full marks require the closing sentence: the remaining 1.33 : 1 still assumes parallel trends, which volunteering violates.
10. A company with a median tenure of 5.00 years and a cost of capital of 9.0 per cent is offered a health investment whose benefits run for decades. Compute its capture ratio, its hurdle multiple, and the social rate of return the investment must clear.
λ = ln 2 / 5.00 = 0.13863. Capture= 0.03 / (0.03 + 0.13863) = 17.79per cent. Hurdle multiple= 1 / 0.1779 = 5.62 x. The corporate hurdle of 9.0 per cent becomes a social hurdle of9.0 / 0.1779 = 50.59per cent. The examinable conclusion is not "do not invest." It is "do not invest as this entity — find the party whose tenure over the population is longer, and invest there."
REFLECTION — eight questions, for one person and a pen
These are not for a room. Write them by hand if you can; the slowness is the point.
- What have you personally believed about the return on looking after people at work, and where did the number you believed come from? Name the slide.
- Think of a time you were healthier because of something an employer did. What was it, and would any instrument in this chapter have detected it?
- Where in your own work do you currently report a soft big number rather than a hard small one — and what would you lose by switching?
- You have about 14.44 per cent of a claim on the long-run health of the people you work with. Who holds the rest? Write their names.
- Recall a decision you made about your own health whose payback was longer than your intention to stay somewhere. What did you decide, and what did the time horizon do to the decision?
- What are you doing for the people around you that you could not justify with a ratio — and what would it take for you to stop needing to?
- Where have you accepted a measurement because it was flattering, and how long was it before somebody asked what the baseline was?
- If a programme you believed in were cut next year for failing a cost target it never claimed, what would you wish you had written down today, and to whom?
ESSAY PROMPTS — five
Each is arguable from more than one side. Each requires at least one source the chapter cites and at least one it does not.
1. The reversal, and what it cost. Katherine Baicker co-authored both the 2010 meta-analysis reporting 3.27 : 1 and the randomised trial that found no effect on spending. Argue either that this is the field working exactly as it should, or that a decade of employer spending was committed on an estimate that its own authors later could not reproduce, and that something in the publication and procurement chain should change. Use Baicker, Cutler and Song (2010) and Song and Baicker (2019), and one source on research synthesis or evidence standards that the chapter does not cite.
2. Incentives, or coercion. Financial incentives for smoking cessation work: 14.7 per cent against 5.0 per cent. Premium differentials tied to health status also move money, and Horwitz, Kelly and DiNardo argue they largely shift cost onto sicker workers rather than generating savings. Argue where the line falls between an incentive that buys a behaviour and a penalty that transfers a cost. Use Volpp et al. (2009) and Horwitz et al. (2013), and one source on health-plan regulation or employment law that the chapter does not cite.
3. The biomarker that did not move. Oregon improved depression by 30.0 per cent and financial protection dramatically, and moved no biomarker measurably at two years. Argue either that this is a reason to fund insurance on mental-health and financial-security grounds and stop promising physiological returns, or that two years is simply too short a window and the endpoint choice was wrong. Use Baicker et al. (2013), and one source on the time course of cardiovascular or metabolic outcomes that the chapter does not cite.
4. Where the health asset should sit. The capture arithmetic says an employer retains 14.44 per cent of the value of long-duration health investment at 3.90 years of median tenure, and 34.18 per cent at 12.00 years of tenure in a trade. Argue either for a sectoral or pooled vehicle as the right owner of workplace health capital, or that pooling replaces a firm's weak incentive with a committee's weaker one, and that a public payer should simply hold the whole asset. Use the Bureau of Labor Statistics tenure series, and one source on multiemployer or sectoral benefit funds — or on national health financing — that the chapter does not cite.
5. Measuring the person who is present. Presenteeism plausibly exceeds absence as a cost and is measured by asking people. Argue either that self-report is the only instrument that can see the thing at all and should be used with a stated multiplier and a sensitivity band, or that a quantity whose dollar value moves 3.00 x on an undefended parameter should be excluded from a business case entirely. Use Pauly et al. (2008) and Ospina et al. (2015), and one source on productivity measurement in a specific industry that the chapter does not cite.