High-Performance Work Systems: Evidence, Implementation, and Unintended Consequences

A high-performance work system is a set of employment practices intended to raise employees' skills, their motivation to use those skills, and their opportunity to apply them to problems that managers cannot solve alone. Typical lists include careful selection, substantial training, teams, information sharing, pay linked to group or organizational results, and reduced status distinctions. Article 18 explains how they are grouped into ability, motivation and opportunity; this article asks what the evidence says they do, and to whom.

Two features of the concept shape everything that follows. The first is that there is no agreed list. Mark Huselid (1995) built his measure from ten practices used in an earlier survey plus three of his own; John Paul MacDuffie (1995) combined HR policies with production practices; later studies used other combinations, and the labels "high-commitment" and "high-involvement" cover overlapping sets. Comparing studies therefore often means comparing different things under one name (Combs et al., 2006). The second is that the central claim concerns systems: the practices are expected to reinforce one another, so that their combined effect exceeds the sum of their separate effects. That claim of complementarity is the concept's main idea and, as later sections show, one of its less securely established parts.

MacDuffie's (1995) study of automotive assembly plants is usually cited for a single word, "bundles," but its argument was more specific than the word suggests. Using a 1989-90 survey of 62 plants in the International Assembly Plant Study, supported by MIT's International Motor Vehicle Program, he argued that innovative HR practices improve performance only when three conditions hold at once: employees have knowledge and skills that managers lack, they are motivated to apply them through discretionary effort, and the production strategy can be achieved only if they do (p. 199). Remove any one condition and the practices have little to work on: skill without motivation adds nothing, and skilled, motivated workers in a system with no use for their ideas have nowhere to direct them.

The production side of the argument is the part most often forgotten. Alongside HR policies and work organization, MacDuffie measured how far plants relied on buffers: repair space, vehicles held between paint and assembly, and parts inventory. In mass-production plants, buffers absorb disruptions so that the line keeps moving. In flexible, or lean, plants they are deliberately cut, which makes every defect visible and gives workers both a reason and a need to solve problems at their source. Flexible production, he wrote, "transfers the ability to cope with contingencies from the technical system" into "the human resource system" (p. 202).

The results supported both of MacDuffie's hypotheses, and he described them as "the strongest statistical evidence to date of a positive relationship between innovative human resource practices and economic performance" (p. 218). He was equally explicit about what the data could not show. Critics argued that lean plants gained productivity by "sweating" workers through a faster pace, peer pressure and the stress of a system without buffers, which they called "management by stress" (Parker and Slaughter, 1988, among others). Because work pace was not measured, MacDuffie wrote, he could not evaluate whether flexible production "leads inevitably to 'speedup'" or means "working smarter" (p. 218). The founding study of HR bundles left the question this article is about explicitly open.

Huselid's (1995) survey of 968 US firms is best known for its dollar estimates, which Article 15 discussed. Three less-quoted features matter here. The first is the content of his "system." Thirteen practices, reported by the senior HR professional in each firm as the proportion of employees covered, formed two factors, "employee skills and organizational structures" and "employee motivation," with reliabilities of .67 and .66, and together the two factors accounted for about 30 percent of the variation in the thirteen items (Table 1). The practices were related, but loosely, which is a modest basis for treating them as a coherent system.

The second feature is that fit, the idea much strategic HR theory depended on, found little support. Huselid reported "only modest evidence" of an effect for internal fit "and little evidence for external fit" with competitive strategy (p. 667). He suggested that simple adoption might matter more than alignment, then cautioned against accepting that until fit could be measured better. The third concerns employees. He wrote that "presumably" some of the value created would accrue to employees through higher wages and benefits, a claim his data, which held no employee outcomes beyond turnover, could not test. He also acknowledged that part of the practices' association with profits "remains unaccounted for" by turnover and productivity, which is the black box in its original form.

The study that set out to measure the worker side directly was Eileen Appelbaum, Thomas Bailey, Peter Berg and Arne Kalleberg's Manufacturing Advantage (2000). It combined plant performance data and interviews with surveys of employees in steel, apparel, and medical electronic instruments and imaging (ERIC record ED456338). Article 18 discusses its place in the origin of the ability-motivation-opportunity framework; what matters here is its claim that participatory work systems benefit employees and plants at the same time. In their review of the wider literature, Michael Handel and David Levine (2006) reported that Appelbaum and colleagues found wage gains of a similar magnitude to the premium of just under 4 percent that Paul Osterman found for blue-collar manufacturing workers, and listed the study among those showing positive effects on job satisfaction. This "mutual gains" position is the benchmark against which the critics' evidence has to be read.

By the mid-2000s there were enough studies to pool. James Combs, Yongmei Liu, Angela Hall and David Ketchen (2006) meta-analysed 92 studies and estimated the overall correlation between high-performance work practices and organizational performance at .20. The relationship was stronger when researchers measured systems of practices than single practices, and stronger among manufacturers than service organizations, but it did not differ between operational and financial measures of performance.

The figure needs careful reading in both directions. A correlation of .20 is not small in organizational research, where performance has many causes and every measure is noisy. Yet two variables correlated at .20 share only about 4 percent of their variance, and a meta-analysis averages associations; it cannot turn them into causal effects. The authors noted that differences in samples, designs, practices and measures had "led extant findings to vary dramatically." The stronger result in manufacturing is also consistent with MacDuffie's argument that the practices pay most where a production system needs worker problem-solving, which makes the setting of a study part of its result.

The causal bias can run either way. Huselid (1995, p. 640) observed that if better-performing firms are more likely to adopt the practices, cross-sectional estimates overstate their effect, while if weaker firms adopt them as a remedy, the estimates understate it; Article 15 examined evidence for the first pattern. The association is well established. Its size and direction as a cause are not.

The "black box" refers to everything between the adoption of a set of practices and an organization's results. The founding studies observed the two ends. Between them, management adopts practices, often as policies reported in a survey by one senior respondent; supervisors implement them with varying conviction and skill; and employees interpret them and respond with more or less commitment, effort, learning and strain.

Each step can make an observed association mislead. At adoption, surveys record whether a practice exists and how many employees it covers, not how seriously it is applied. MacDuffie (1995, p. 218) noted that survey data "cannot identify the nuances of team functioning": some teams rotate jobs and solve production problems, others mainly discuss schedules. Handel and Levine (2006) concluded that small and inconsistent findings across studies might reflect "the seriousness of implementation (with many workplaces making few real changes)," the quality of studies and measures, or real variation in effects, and that the evidence could not yet separate these. Even the bundling idea has not always survived close testing. From a careful establishment-level study, Peter Cappelli and David Neumark (2001, p. 760) concluded that "there is not consistent evidence of statistically significant effects of bundles of work practices" (quoted in Handel and Levine, 2006).

The interpretation step turned out to matter a great deal. Lisa Nishii, David Lepak and Benjamin Schneider (2008) studied a US supermarket chain, using surveys of 4,208 employees and 1,010 department managers in 362 departments across 95 unionized stores. Employees under the same HR practices attributed them to different motives. Where they believed the practices were meant to improve service quality and employee well-being, they reported higher commitment and satisfaction, and these attitudes, once shared within departments, were related to citizenship behaviour and customer satisfaction. Where they saw cost-cutting and getting as much as possible out of staff, the association was negative, though weaker. A team target or a suggestion scheme therefore has no fixed meaning: employees can experience it as an investment or as a squeeze, depending partly on what they conclude about management's purpose.

Article 16 set out the universalist, contingency and configurational views of HR and performance. HPWS research is usually filed as universalist, but the founding studies sit awkwardly there.

MacDuffie's conclusion was universalist within an industry and configurational in its mechanism. He reported that, at least for auto assembly plants, flexible production "consistently leads to better performance" than mass production (p. 218), which contradicts a fit hypothesis in which either system would perform well if its HR practices matched it. Yet his HR practices paid off in combination with lean buffers and team-based work, and the bundle worked because of the production logic it served. Nothing in his data shows that the same practices, removed from that logic, would help. Huselid's weak fit results pointed towards universal adoption, but he attributed them partly to "the preliminary nature of the measures of fit" (p. 668). And the meta-analytic finding that effects are larger in manufacturing than in services suggests that context affects at least how much the practices pay.

One way to reconcile these findings is to separate levels. As a principle, the claim that work organized to use employees' knowledge outperforms work organized to exclude it has broad support across settings. As a design, meaning which pay scheme, which team structure, how much training and how far to cut buffers, the evidence offers little support for a single template, and the founding study shows practices drawing their value from a particular production system. This distinction is this article's interpretation of the studies cited, not a finding that any of them tested.

The claim that HPWS benefits employer and employee alike is better described as one of three positions, each with evidence behind it.

Mutual gains. The first position holds that the practices make jobs more skilled, autonomous and better paid, so employees share in the gains. Beyond Appelbaum and colleagues' study, Handel and Levine's (2006) review supports a moderate version: involvement practices usually raise job satisfaction and on average raise wages slightly. They concluded that the findings "do not indicate that management-by-stress is typical," while noting that effects on injury rates were less clear and that measurement error and selection bias had not been ruled out.

Conflicting outcomes. The second position, rooted in labour process theory, holds that the performance gains come from extracting more effort: work intensification, the offloading of task controls onto employees, and greater job strain. Harvie Ramsay, Dora Scholarios and Bill Harley (2000) tested both positions with the 1998 British Workplace Employee Relations Survey (WERS98), asking whether the practice-performance link ran through commitment or through intensification and strain. In their words, both models "were found wanting." Paul Osterman (2000) had reached a sceptical conclusion from US establishment data: "there is very little evidence that HPWOs [high performance work organizations] have delivered on their promise of 'mutual gains'" (quoted in Handel and Levine, 2006). John Godard's (2004) review went further, concluding that, compared with what has long been thought of as good management practice, claims of superior performance "may be unwarranted," and that the implications for workers and unions "are at best uncertain."

Dependent on design. The third position, consistent with the studies that follow, treats intensification as a risk whose size depends on how a system is built. Brigitte Kroon, Karina Van De Voorde and Marc Van Veldhoven (2009) surveyed 393 employees in 86 Dutch organizations, with HR managers reporting the practices. They found a slightly positive relationship between high-performance practices and burnout that was completely mediated by job demands, and read this as support for a critical "employee exploitation" perspective. Jaclyn Jensen, Pankaj Patel and Jake Messersmith (2013), studying 1,592 government workers in 87 departments in Wales, found that HPWS implemented with low job control (discretion over how employees carry out their responsibilities) was associated with anxiety and role overload, and through them with intentions to quit. Peter Boxall and Keith Macky (2014), using a New Zealand national population survey, separated two things usually introduced together. Greater autonomy and participation in decisions had positive or neutral effects, predicting higher job satisfaction and better work-life balance, with no relationship to stress or fatigue. Higher work intensity increased fatigue and stress and undermined work-life balance.

A review by Van De Voorde, Paauwe and Van Veldhoven (2012) of 36 quantitative studies published between 1995 and May 2010 helps reconcile the positions. Well-being understood as happiness, such as job satisfaction, or as relationships, such as trust, generally moved together with organizational performance, as the mutual-gains view predicts. Health-related well-being, such as strain and exhaustion, appeared to function as a conflicting outcome. Both camps can therefore be right about different outcomes in the same workplace: employees may be more satisfied with more interesting work and more exhausted by it.

The table summarizes the sources as described in this article; for the last four studies, the descriptions rest on the published abstracts. Its right-hand column shows how much of the early literature took the employee side on trust.

The third position is the most useful for managers, because it turns an ideological argument into design choices. The studies above suggest four. These are recommendations drawn from the evidence, not rules that the studies tested as a package.

1. Pair new demands with real discretion. Jensen and colleagues' results imply that anxiety and overload concentrate where responsibilities are devolved without control over methods and pace. A team expected to solve quality problems needs the authority to stop work, change standard methods and call for support.
2. Cut buffers at the pace that capability grows. In MacDuffie's logic, lean buffers expose problems so that skilled workers can solve them. Cutting them before workers can deal with what is exposed turns visibility into pressure.
3. Measure the employee side from the start. Health outcomes can worsen while satisfaction and output improve, so overtime, absence, injuries and reported strain belong beside productivity and quality, with a baseline taken first.
4. Consider what employees will conclude about purpose. Nishii and colleagues' findings suggest that practices introduced alongside job cuts, or whose gains are visibly kept by the employer, are likely to be read as cost reduction whatever their design.

The underlying tension remains. Discretionary effort is effort the employer cannot require, and a system built to draw it out can drift from asking for ideas to expecting extra effort as a matter of course.

The industry context (real and dated). Within a decade Bangladesh's motorcycle industry moved from imports to local manufacturing. The Ministry of Industries' Motorcycle Industry Development Policy 2018 set targets of raising production to at least 5 lakh (500,000) units by 2021 and 10 lakh by 2027, and of raising direct and indirect employment in the sector from 5 lakh to 15 lakh by 2027 (section 2.4). It also called for stronger training for engineers, workers and managers alike (section 4.4). Duties strongly favour local manufacture over imported finished bikes, and sales peaked at over 6.4 lakh units in FY2021-22 before falling to 3.84 lakh in FY2023-24, a seven-year low (The Business Standard, 2 September 2024). They then recovered: sales in the first half of 2025 reached 257,632 units against 202,330 a year earlier (The Daily Star, 29 December 2025).

For HPWS, two features matter: demand swings sharply, so plants face both slack periods and surges, and national policy treats skills as a constraint, which makes training-heavy work systems attractive. The law also sets limits that an intensification strategy would run into. Under the Bangladesh Labour Act 2006, an adult worker may ordinarily work eight hours a day and 48 hours a week; with overtime, the limits are ten hours a day, 60 hours in any week and an average of 56 hours a week over a year (sections 100 and 102, English text hosted by the ILO; later amendments were not reviewed). Establishments with 50 or more workers must form a participation committee whose statutory functions include improving working conditions and also productivity and quality (sections 205 and 206; Articles 13 and 14 discuss these committees in practice).

The company, people and figures below are fictional and were constructed for teaching. Any resemblance to a real manufacturer is unintended.

Atrai Motors Limited assembles 100 to 150cc commuter motorcycles under licence from a foreign brand at a plant north of Dhaka. It employs 460 production workers on two assembly lines and an engine sub-assembly section, on one shift with frequent overtime. Work follows the mass-production pattern MacDuffie described. Workers are hired for experience in a similar job, trained for two days and assigned to one station. Inspectors check finished bikes at the end of the line, a repair bay holds up to 60 motorcycles awaiting rework, and two days of engine stock ensure the main lines never wait. Final inspection finds about 9 defects per 100 bikes, and annual labour turnover is 28 percent.

With sales recovering, the new operations director, Farhana Kabir, proposes an "Atrai Production System": teams of eight with an elected leader, quality checks at each station, authority for any worker to stop the line, a suggestion scheme, 60 hours of training a year, a quarterly team bonus, a single canteen for managers and workers, and cuts in the engine buffer to half a day and in the repair bay to 15 places. The finance director supports it because it promises more output without hiring. The HR manager, Tanvir Hossain, raises three concerns: the plan would start just as dealers ask for 20 percent more bikes before Eid, the draft bonus is based on units completed per shift, and the participation committee has not been consulted. The board asks for a recommendation.

What the exercise is designed to show. The choice turns on sequence and design: which practices go together, when buffers are cut, what the bonus rewards, who has a say, and what is measured. Options (a) and (b) fail in opposite directions: one adds demands faster than capability and control, the other adds practices without the production changes that give them a purpose.

Rate each statement for a unit where high-performance practices are being considered or have recently been introduced. Totals range from 10 to 50. The statements and score bands are this article's teaching tool, based on the evidence discussed above. They are not a validated instrument.

Reading the result. Statements 1, 2 and 8 concern complementarity; 3, 4 and 7 how employees are likely to read the change; 5, 9 and 10 whether the organization would notice gains made at employees' expense; 6 concerns implementation. A high total with low scores on 5, 9 and 10 describes a plan that may work and cannot tell whether it has. If an operations director and an HR manager complete it separately, the items on which they disagree make a useful agenda.

Three decades after MacDuffie and Huselid, the field has firm evidence of an association between high-performance practices and organizational results and a much weaker grasp of the mechanism. Designs that could settle the causal question remain rare, and because the content of an HPWS varies from study to study, "the evidence" describes a family of related practices, not one intervention.

The welfare question has been narrowed rather than answered. Research suggests that involvement and intensification can be separated, but employers often introduce them together, and why is poorly understood. Godard (2004) traced the disappointing results to the employment relationship itself and argued for legal and institutional reform. That raises a question this evidence cannot answer for Bangladesh, since almost all of it comes from North America, Western Europe and Australasia. Whether practices would strike the same balance between satisfaction and strain on a Gazipur assembly line as in a Welsh public agency has to be tested, not assumed. For a manager, the practical decision is which version of an HPWS to build: one that widens discretion as it raises demands, or one that raises demands and calls the result involvement. Article 20 turns to how an HR scorecard can be designed to show which of the two an organization has actually built.

Select an organization that uses or is considering high-performance work practices: your employer (which you may disguise), or a company described in published case material, annual reports or credible press coverage.

Prepare a report of 2,500 to 3,000 words that:

1. maps the organization's current practices, including how work and buffers or slack are organized as well as HR policies, and identifies where practices reinforce or undermine one another;
2. appraises what performance gains can reasonably be expected, using the evidence on effect size, setting and causality, and stating your assumptions;
3. identifies the main risks of work intensification and strain, naming the mechanisms and the employee groups most exposed, with reference to job demands and job control;
4. designs an implementation plan that specifies sequence, a pilot or comparison, the measures to be tracked (including employee health outcomes), and the evidence that would lead management to slow down or redesign;
5. assesses the legal and institutional context, such as working-time law and employee representation, and how it constrains or supports the design; and
6. makes a recommendation and states the conditions under which it would be wrong.

Use at least five academic sources, including at least three from this article's reference list. Do not present an association between practices and performance as proof of effect without stating the comparison that supports it. Where information is unavailable, say so and explain how the gap limits your conclusions.

Assessment rubric (100 marks)

1. Appelbaum, E., Bailey, T., Berg, P., & Kalleberg, A. L. (2000). Manufacturing advantage: Why high-performance work systems pay off. ILR Press (Cornell University Press), for the Economic Policy Institute. ERIC record: https://eric.ed.gov/?id=ED456338
2. Boxall, P., & Macky, K. (2014). High-involvement work processes, work intensification and employee well-being. Work, Employment and Society, 28(6), 963–984. https://doi.org/10.1177/0950017013512714
3. Combs, J., Liu, Y., Hall, A., & Ketchen, D. (2006). How much do high-performance work practices matter? A meta-analysis of their effects on organizational performance. Personnel Psychology, 59(3), 501–528. https://doi.org/10.1111/j.1744-6570.2006.00045.x
4. Godard, J. (2004). A critical assessment of the high-performance paradigm. British Journal of Industrial Relations, 42(2), 349–378. https://doi.org/10.1111/j.1467-8543.2004.00318.x
5. Huselid, M. A. (1995). The impact of human resource management practices on turnover, productivity, and corporate financial performance. Academy of Management Journal, 38(3), 635–672. https://doi.org/10.2307/256741 (author-hosted copy: https://www.markhuselid.com/pdfs/articles/1995_AMJ_HPWS_Paper.pdf)
6. Jensen, J. M., Patel, P. C., & Messersmith, J. G. (2013). High-performance work systems and job control: Consequences for anxiety, role overload, and turnover intentions. Journal of Management, 39(6), 1699–1724. https://doi.org/10.1177/0149206311419663
7. Kroon, B., Van De Voorde, K., & Van Veldhoven, M. (2009). Cross-level effects of high-performance work practices on burnout: Two counteracting mediating mechanisms compared. Personnel Review, 38(5), 509–525. https://doi.org/10.1108/00483480910978027
8. MacDuffie, J. P. (1995). Human resource bundles and manufacturing performance: Organizational logic and flexible production systems in the world auto industry. Industrial and Labor Relations Review, 48(2), 197–221. https://doi.org/10.2307/2524483 (copy on the author's faculty page: https://faculty.wharton.upenn.edu/wp-content/uploads/2012/05/ILRR-1995.pdf)
9. Nishii, L. H., Lepak, D. P., & Schneider, B. (2008). Employee attributions of the "why" of HR practices: Their effects on employee attitudes and behaviors, and customer satisfaction. Personnel Psychology, 61(3), 503–545. https://doi.org/10.1111/j.1744-6570.2008.00121.x (details checked against CAHRS Working Paper 08-03, Cornell University)
10. Ramsay, H., Scholarios, D., & Harley, B. (2000). Employees and high-performance work systems: Testing inside the black box. British Journal of Industrial Relations, 38(4), 501–531. https://doi.org/10.1111/1467-8543.00178

Other sources cited

• Cappelli, P., & Neumark, D. (2001). Do "high-performance" work practices improve establishment-level outcomes? Industrial and Labor Relations Review, 54(4), 737–775. Quoted as in Handel and Levine (2006).
• Handel, M. J., & Levine, D. I. (2006). The effects of new work practices on workers. In E. E. Lawler III & J. O'Toole (Eds.), America at work (pp. 73–85). Palgrave Macmillan. https://doi.org/10.1057/9781403983596_5 (text read: University of California working-paper version, https://escholarship.org/uc/item/2fm3x01d)
• Osterman, P. (2000). Work reorganization in an era of restructuring: Trends in diffusion and effects on employee welfare. Industrial and Labor Relations Review, 53(2), 179–196. https://doi.org/10.1177/001979390005300201 (quoted as in Handel and Levine, 2006)
• Parker, M., & Slaughter, J. (1988). Choosing sides: Unions and the team concept. South End Press. Cited as in MacDuffie (1995).
• Van De Voorde, K., Paauwe, J., & Van Veldhoven, M. (2012). Employee well-being and the HRM–organizational performance relationship: A review of quantitative studies. International Journal of Management Reviews, 14(4), 391–407. https://doi.org/10.1111/j.1468-2370.2011.00322.x

Bangladesh policy, legal and news sources

• Ministry of Industries, Government of the People's Republic of Bangladesh. Motor Cycle Shilpa Unnayan Niti 2018 [Motorcycle Industry Development Policy 2018] (in Bangla), sections 2.4 and 4.4: https://objectstorage.ap-dcc-gazipur-1.oraclecloud15.com/n/axvjbnqprylg/b/V2Ministry/o/office-moind/2024/12/cd49bfd1568147ebb7db0eafd2e6c220.pdf
• Bangladesh Labour Act, 2006 (XLII of 2006), sections 100, 102, 205 and 206. English text hosted by the International Labour Organization: https://www.ilo.org/media/43266/download
• Babu, M. U. "Motorcycle sales tumble to seven-year low." The Business Standard, 2 September 2024. https://www.tbsnews.net/bangladesh/transport/motorcycle-sales-tumble-seven-year-low-931301
• Haq, R. A. "How Bangladesh became a motorcycle nation." The Daily Star, 29 December 2025. https://www.thedailystar.net/supplements/long-ride/news/how-bangladesh-became-motorcycle-nation-4068976

Verification notes: MacDuffie (1995), Huselid (1995), the working-paper versions of Nishii et al. (2008) and Handel and Levine (2006), the Labour Act and the 2018 policy were read in full. For Combs et al. (2006), Jensen et al. (2013), Kroon et al. (2009), Van De Voorde et al. (2012), Boxall and Macky (2014) and Godard (2004), the published abstracts were checked; for Ramsay et al. (2000), the abstract in the University of Strathclyde repository. Findings attributed to Appelbaum et al. (2000) about workers are reported as described by Handel and Levine, and the exact numbers of plants and employees surveyed are omitted because secondary accounts differ. The quotations from Osterman (2000) and Cappelli and Neumark (2001) are taken from Handel and Levine. The Atrai Motors case, the decision options and commentary, the diagnostic, the study summary table and the implementation recommendations in Section 7 are original teaching material for this series.

Next in The Human Capital MBA series: Article 20, on designing a strategic HR scorecard.