---
title: "People as Capability or Cost: What the Evidence Shows"
description: "Does spending on training, pay and stable jobs pay back? SDG 8, Huselid, Combs, Costco and the limits of the evidence, for owners of small firms."
url: "https://staging.khoi.co.uk/en/blog/nhan-su-la-nang-luc-hay-chi-phi"
lang: en-GB
author: "Lê Quốc Khôi"
date: 2026-11-03
updated: 2026-11-03
topics: ["Strategy & leadership"]
tags: ["Human capital", "SDG 8", "Employee turnover", "High-performance work practices", "Good jobs strategy"]
---

# People as capability, not cost: what the public evidence supports

## Key takeaways

- UN Goal 8 aims at full and productive employment and decent work at country level (targets 8.2, 8.3, 8.5, 8.8). The official page does not say that paying better makes one company more profitable (United Nations Department of Economic and Social Affairs, n.d.).
- A meta-analysis of 92 studies (19,319 organisations) finds a correlation of about 0.20 between "high-performance" HR practices and organisational performance, after correcting for measurement error. The authors flag that part may run in reverse (Combs et al., 2006, pp. 513, 519–520).
- Across 968 US firms, one standard deviation more use of such practices goes with 7.05% lower turnover (relative) and $27,044 more sales per employee. The data are a snapshot; the author says longitudinal data are needed (Huselid, 1995, pp. 658, 667–668).
- On 36 steel finishing lines, a switch to a bundle of practices (teamwork, information sharing, training) added 2.5 to 3.5 percentage points of running time. Single practices alone had little or no effect (Ichniowski et al., 1997).
- Replacing an employee cost a median 21% of annual salary in 27 US case studies, excluding executives and physicians (Boushey and Glynn, 2012, p. 1). Costco reports about 94% retention of staff with a year or more of service in the US and Canada, a figure it supplies itself (Costco Wholesale Corporation, 2026, p. 6).

Sales stalled this quarter. On the cost sheet, pay is the biggest line, and the person proposing to cut hours or postpone training has a case. The person opposing says people are a capability, not a cost. Which way does the public evidence lean, and by how much?

## What does SDG 8 say about staff, and what does it leave out?

Goal 8 puts productivity and decent work side by side, but it is a goal for countries, not a cost question for one firm. The UN's page gives its title as "Promote sustained, inclusive and sustainable economic growth, full and productive employment and decent work for all" (United Nations Department of Economic and Social Affairs, n.d.). Four targets matter most to a small business:

| Target | What it says (my summary) | UN indicator |
|:--|:--|:--|
| 8.2 | Higher productivity through diversification, technological upgrading and innovation | Growth in real GDP per employed person |
| 8.3 | Decent job creation; formalisation and growth of micro, small and medium-sized enterprises | Share of informal employment |
| 8.5 | By 2030, full and productive employment and decent work for all, equal pay for work of equal value | Hourly earnings; unemployment |
| 8.8 | Protect labour rights; safe working environments | Occupational injuries; national compliance with labour rights |

Source: (United Nations Department of Economic and Social Affairs, n.d.). For Vietnam, the General Statistics Office's release of October 2026 gives 61.5% of employed people in informal work over the first nine months of 2026, and puts the share of the labour force trained to degree or certificate level at 29.9% (Cục Thống kê, 2026). The release does not call these SDG indicators; I set the first beside target 8.3 because it uses the same concept.

What the page lacks: no sentence ties a wage level or a training programme to one company's profit. SDG 8 gives you vocabulary and national yardsticks. The question "does paying more pay back?" has to come from elsewhere.

## Does spending on people go with better business results?

Yes, in the sense that the studies point the same way. Which causes which is disputed (next section). The two largest studies I read are Huselid (1995) and the meta-analysis by Combs and colleagues (2006). Both study "high-performance work practices": incentive pay, training, employee participation, selective hiring, flexible work and similar (Combs et al., 2006, p. 502).

| | Huselid (1995) | Combs and colleagues (2006) |
|:--|:--|:--|
| Sample | 968 publicly held US firms with over 100 employees; 28% response from 3,452 firms (Huselid, 1995, p. 644) | 92 studies, 19,319 organisations (Combs et al., 2006, p. 509) |
| Result | Per standard deviation of practices: turnover 7.05% lower (relative); per employee, sales $27,044, profit $3,814 and market value $18,641 higher (Huselid, 1995, p. 667). The sales figure is nearly 16% of mean sales (Huselid, 1995, p. 658) | Correlation 0.15, or 0.20 corrected for measurement error; the figures after that are corrected too: whole systems of practices 0.28 against 0.14 for single practices; manufacturing 0.30, services 0.17 (Combs et al., 2006, p. 513) |

Combs and colleagues also report training alone at 0.15 and pay level at 0.18 (corrected) (Combs et al., 2006, p. 516). Do not compare the two columns directly. Huselid estimates a 1.30-point fall in turnover, which is 7.05% of the mean (Huselid, 1995, p. 656). Combs and colleagues instead take Huselid's mean and standard deviation for turnover, apply their own 0.20 effect, and arrive at a 4.4-point fall (18.4% to 14.0%) (Combs et al., 2006, p. 518). They are different calculations.

## How solid is that evidence?

Solid on direction, weak on cause. Four points.

First, Huselid's data are a snapshot. He cautions that firm conclusions about causality need data over time on both practices and performance, which did not yet exist (Huselid, 1995, p. 668). His two attempts to reduce reverse causality, using next year's profits and two-stage estimation, kept the same direction (Huselid, 1995, pp. 666–667).

Second, Combs and colleagues raise the chance that profitable firms use their slack to adopt more HR practices. They cite a longitudinal study (Wright and colleagues, 2005) where the effect on financial results shrinks sharply once earlier performance is controlled (Combs et al., 2006, pp. 519–520). I did not read that paper.

Third, the steel study by Ichniowski, Shaw and Prennushi (1997) gets closer to cause because it follows each line over time, so a line can be compared with itself before and after it changed how it worked: 36 finishing lines at 17 companies, 2,190 monthly observations, mean running time 91.9% (Ichniowski et al., 1997). Lines that moved from the traditional approach (narrow jobs, close supervision, training only by picking things up on the job) to a bundle with limited teamwork and labour-management communication gained 2.5 percentage points of running time; those that moved to a bundle with deep team involvement and extensive training gained 3.5 (Ichniowski et al., 1997). Single practices had little or no effect, and pay level was not statistically significant once added to the model (Ichniowski et al., 1997). The limit: one industry, one process (Ichniowski et al., 1997), and this is still observational data, not an experiment. The authors also record why many older lines did not adopt: transition costs and distrust. One manager reports that even after financial figures were shared with workers, some still suspected the company kept two sets of books (Ichniowski et al., 1997).

Fourth, not every practice shows an effect. In the meta-analysis, performance appraisal, teams and information sharing have not yet been shown to exceed zero (Combs et al., 2006, p. 518), and the data could not test whether fit with strategy helps (Combs et al., 2006, p. 503).

## What does a public company say about pay and retention? The case of Costco

Costco says in its 10-K that it does not seek to minimise wages and benefits, and ties that to turnover, productivity and satisfaction. That is the company's belief, not the result of a test. The report covers the year ended 30 August 2026, was filed on 7 October 2026, and counts 355,000 employees (Costco Wholesale Corporation, 2026, p. 6).

| What the report says | Page |
|:--|:--|
| Pay and benefits are the largest expense after the cost of merchandise, and are "carefully monitored" | 6 |
| The aim is "reducing employee turnover, increasing productivity, and enhancing employee satisfaction", which it says requires pay "better than the industry average for much of our workforce" | 24 |
| About 94% retention of staff with at least a year's service, US and Canada, 2026; the company calls this high for its industry | 6 |
| Net income $9,226 million on net sales of $297,247 million, about 3.1% (my division); the company writes that small cost changes can move net income substantially | 24, 25 |

Source for all rows: (Costco Wholesale Corporation, 2026).

Zeynep Ton, an MIT Sloan professor of the practice who works on operations, argues that pay has to come with job design. She studied Costco, Mercadona, Trader Joe's and QuikTrip and concluded that the supposed trade-off between investing in staff and low prices is false (Ton, 2012; Ton, 2023). Her four operating choices: focus and simplify; standardise routine work and let frontline staff decide; cross-train; operate with slack. Firms that treat staff as a cost to minimise, she writes, pay the going market rate and live with high turnover; in a good-jobs system it is a cost you cannot tolerate (Ton, 2023).

Three limits. One, a 10-K is written by the company and states its own belief. Two, Note 10 of the same report lists employee lawsuits about pay and breaks. The Reyes case (California, filed November 2023) alleges, among other things, unpaid minimum wage and overtime and missed meal and rest breaks; the report says the company has moved to dismiss the second amended complaint. The Madera case (Washington State, January 2026) alleges similar failures; the company reports an agreement to resolve it for an immaterial amount, subject to court approval (Costco Wholesale Corporation, 2026, pp. 60–61). These are the plaintiffs' allegations as the company reports them, and I did not read the court files. Three, Ton studied firms she herself describes as already in a virtuous cycle, and she co-founded the Good Jobs Institute, which helps companies apply the approach (Ton, 2023). I have not read her underlying data.

## How much does a departure cost, and does a raise keep people?

A departure is costly, but a flat raise does not clearly keep people. Boushey and Glynn (2012) pooled 30 cases from 11 papers published between 1992 and 2007; the places named in the report's table are all in the US. Excluding executives and physicians, the median replacement cost was 21% of annual salary; for jobs paying $30,000 or less, 16%; estimates ran from 5.8% to 213% (Boushey and Glynn, 2012, pp. 1–2). Costs include hiring, training, the weeks a new person is slower and makes errors, and extra work for those who stay; only 2 of the 11 papers counted indirect costs (Boushey and Glynn, 2012, pp. 5–6). Where a paper gave no salary, the authors assigned the occupation's mean wage (Boushey and Glynn, 2012, p. 9).

Dube, Giuliano and Leonard (2018, an NBER working paper) used data from a large US retailer in the late 1990s, where a pay formula created arbitrary gaps between co-workers. Quits responded strongly to gaps, especially when others got bigger raises, which suggests fairness matters. When everyone got the same raise, separations were less sensitive: the nine-month elasticity is −2.3, with a 95% interval that rules out only values more negative than −11.1 (Dube et al., 2018). My reading: in that low-wage group a flat raise kept people only a little, and imprecisely, while gaps within a team mattered a great deal.

## What can a small-firm owner do with this?

Borrow the questions, then measure your own numbers. No source above says how much to raise pay or how many training hours to give in a 30-person firm in Vietnam. The table follows the cost groups Boushey and Glynn list (Boushey and Glynn, 2012, p. 5); the third column is work for your own books.

| Cost | Question | Number to pull from your firm |
|:--|:--|:--|
| Hiring | What does each vacancy cost to fill? | Cost per hire, last 12 months |
| Training and the gap | Hours of coaching, overtime or temporary cover while the seat is empty? | Coaching hours times hourly pay, plus cover cost |
| Not yet at full output | How many weeks is a newcomer slower or error-prone? | Error and complaint rates for staff under three months |
| Those who stay | Does anyone leave after a colleague goes? | Quits within 90 days of each departure |

A hypothetical example shows the arithmetic. A firm of 30 has 6 people leave in a year; average pay is 9 million dong a month, or 108 million a year. At the 16% ratio for the lowest pay group in the US report, each replacement costs about 17.3 million dong, roughly 104 million in total. The firm, the numbers and the use of that ratio are my assumptions; I do not know whether 16% carries over to Vietnam, so treat this as an illustration, not an estimate.

This suits you if turnover really disrupts your operation and you hold twelve months of data. It does not suit you if you need a firm figure for the return on training: the studies give averages and direction only. If you are weighing cuts, the post on [layoffs and the psychological contract](https://staging.khoi.co.uk/en/blog/cat-giam-nhan-su-hop-dong-tam-ly) covers what those who stay see, and the one on [Theory E and Theory O through the Asda case](https://staging.khoi.co.uk/en/blog/thuyet-e-thuyet-o-asda) sets shareholder value against organisational capability.

## Going deeper: where do the ideas come from, and where does the evidence stop?

Two lines of thought meet here. The resource-based view: Huselid draws on Wright and McMahan's argument that people can be a source of lasting advantage when they are valuable, rare, hard to imitate and not replaceable (Huselid, 1995, pp. 636–637). And work motivation: [Herzberg's two-factor theory](https://staging.khoi.co.uk/en/framework/herzberg-two-factor) separates what dissatisfies people (pay, conditions) from what motivates them (interesting work, challenge, responsibility) (Herzberg, 1968). My reading is that the steel result, where work organisation mattered and pay did not, sounds like Herzberg's spirit. That is inference, not a test.

What the sources do not show: none is a Vietnamese firm. Huselid has firms over a hundred employees, Ichniowski one stage of steelmaking, Dube a US retailer of the late 1990s, Boushey and Glynn cases from before 2008. Combs and colleagues note that sampling and measurement error explain only 37% of the variance, so the effect varies with context (Combs et al., 2006, p. 513). I have no source on the consequences of cutting hours, the other half of the opening question, and I read no critique of Herzberg's method.

What did I read? In full: Huselid, Combs and colleagues, the Cục Thống kê release, Ton (2023) as an excerpt on MIT Sloan's site, and the body of Boushey and Glynn. Partly: Ichniowski and colleagues (JSTOR scan: abstract, sample, practice systems, fixed-effects results, wage check, non-adoption, conclusion; not the quality section or appendix closely); Dube and colleagues (NBER Working Paper 24906, August 2018: abstract, introduction, Table 4, conclusion; not the *American Economic Review* version, so figures are unchecked against it); Costco's 10-K (Item 1, one risk factor, MD&A highlights, Note 10); the UN Goal 8 page (all targets, not resolution A/RES/70/1). Abstract or summary only: Ton (2012) and Herzberg (1968, the Harvard Business Review page for the 2003 reprint), as the rest is paywalled. Costco and statistics figures are as of 3 November 2026.

## Frequently asked questions

### Does SDG 8 oblige a Vietnamese company to raise pay or train staff?

Not on the UN page I read. The targets are goals for countries and are measured with national statistics, such as the share of informal employment or national compliance with labour rights (United Nations Department of Economic and Social Affairs, n.d.). This post does not cover Vietnamese labour law; for minimum wage, contracts and working hours, ask a lawyer or the labour authority where you operate.

### Can I apply an American study's figures to a 30-person firm in Vietnam?

Not directly, because none of the sources in this post measures a Vietnamese firm. Boushey and Glynn, for example, found 17 case studies at average pay of $75,000 a year or less with replacement costs between 10% and 30% of annual salary (Boushey and Glynn, 2012, p. 2), but those are US pay levels and a US labour market. For a 30-person firm, borrow the question and measure from your own books: how often you replace people, what each replacement costs, and how long a newcomer takes to reach full output.

### Herzberg said pay does not motivate. Is a raise therefore wasted?

I read only Harvard Business Review's summary of his article, so I will not go far. By that summary, environmental factors such as low pay can make people unhappy, but even when managed brilliantly they do not make anyone work much harder or smarter (Herzberg, 1968). The studies in this post measure other things: quits, output and company-wide results. The two need not clash, but I found no source that tests them side by side.

## References

- Boushey, H. and Glynn, S.J. (2012) *There are significant business costs to replacing employees*. Washington, DC: Center for American Progress. Available at: https://cdn.americanprogress.org/wp-content/uploads/2012/11/CostofTurnover.pdf (Accessed: 2026-11-03).
- Combs, J. et al. (2006) ‘How much do high-performance work practices matter? A meta-analysis of their effects on organizational performance’, *Personnel Psychology*, 59(3), pp. 501–528. https://doi.org/10.1111/j.1744-6570.2006.00045.x
- Costco Wholesale Corporation (2026) *Annual report on Form 10-K for the fiscal year ended August 30, 2026*. Washington, DC: U.S. Securities and Exchange Commission. Available at: https://www.sec.gov/Archives/edgar/data/909832/000090983226000093/cost-20260830.htm (Accessed: 2026-11-03).
- Cục Thống kê (2026) *Thông cáo báo chí về tình hình lao động, việc làm quý III và 9 tháng năm 2026*. Website Cục Thống kê (General Statistics Office of Vietnam). Available at: https://www.nso.gov.vn/tin-tuc-thong-ke/2026/10/thong-cao-bao-chi-ve-tinh-hinh-lao-dong-viec-lam-quy-iii-va-9-thang-nam-2026/ (Accessed: 2026-11-03).
- Dube, A., Giuliano, L. and Leonard, J. (2018) *Fairness and frictions: the impact of unequal raises on quit behavior (NBER Working Paper 24906)*. Cambridge, MA: National Bureau of Economic Research (NBER Working Paper Series). https://doi.org/10.3386/w24906
- Herzberg, F. (1968) ‘One more time: how do you motivate employees?’, *Harvard Business Review*, 46(1), pp. 53–62. Available at: https://hbr.org/2003/01/one-more-time-how-do-you-motivate-employees.
- Huselid, M.A. (1995) ‘The impact of human resource management practices on turnover, productivity, and corporate financial performance’, *Academy of Management Journal*, 38(3), pp. 635–672. https://doi.org/10.5465/256741
- Ichniowski, C., Shaw, K. and Prennushi, G. (1997) ‘The effects of human resource management practices on productivity: a study of steel finishing lines’, *American Economic Review*, 87(3), pp. 291–313. Available at: https://www.jstor.org/stable/2951347 (Accessed: 2026-11-03).
- Ton, Z. (2012) ‘Why "good jobs" are good for retailers’, *Harvard Business Review*, 90(1–2), pp. 124–131. Available at: https://hbr.org/2012/01/why-good-jobs-are-good-for-retailers (Accessed: 2026-11-03).
- Ton, Z. (2023) *The 4 essential choices in a good jobs system (excerpt from The case for good jobs, Harvard Business Review Press)*. MIT Sloan Ideas Made to Matter. Available at: https://mitsloan.mit.edu/ideas-made-to-matter/4-essential-choices-a-good-jobs-system (Accessed: 2026-11-03).
- United Nations Department of Economic and Social Affairs (n.d.) *Goal 8: promote sustained, inclusive and sustainable economic growth, full and productive employment and decent work for all*. Sustainable Development Goals Knowledge Platform. Available at: https://sdgs.un.org/goals/goal8 (Accessed: 2026-11-03).

---

Canonical HTML version: https://staging.khoi.co.uk/en/blog/nhan-su-la-nang-luc-hay-chi-phi