3.1 Datasets

The British WERS 2004 and 2011 (Department for Business Innovation and Skills 2013) and the French REPONSE 2005 and 2011 (DARES 2013) were carried out at similar times and cover a period when both economies were close to their archetypal models of liberal economy and state capitalism (before the 2017 reforms under Macron that liberalized the labour market in France) and operating within a comparable international context as members of the European Union. WERS 2011 provides the most up-to-date British data, whereas REPONSE was conducted again in 2017 and 2023. We use the 2017 wave to assess the robustness of our estimates.⁴

Both surveys employed similar methodologies, gathering data from managers on HR practices at the workplace and demographic information about the site and any parent enterprise, along with linked data on job characteristics and working conditions from randomly sampled employees in those workplaces.

Though not explicitly harmonized, the surveys were developed loosely in parallel, containing many comparable data items. They offer larger national samples than prominent harmonized cross-national surveys (such as the European Working Conditions Survey, the European Social Survey and the European Company Survey). Their linked employer-employee data also enable detailed information on individual employees’ working conditions to be linked with rich information on the characteristics, ownership patterns and management practices of the workplaces in which they are employed.

Our analysis focuses on private sector workplaces with 11 or more employees, covering all sectors except agriculture, mining, public administration and defence, activities of households as employers and extraterritorial organizations (i.e. NACE Rev.2, sections A, B, O, T and U); this is the lowest common denominator in terms of the coverage of WERS and REPONSE. This accounts for almost three quarters (74 per cent) of all private sector employment – and around 55 per cent of all employment – in each of the two economies (Amossé et al. 2016: table A.5).⁵

REPONSE randomly selected 9 employees in every workplace for its employee survey, whereas WERS surveyed all employees in workplaces with 11–24 employees and randomly selected 25 employees in larger workplaces. However, since only employees with 15 or more months of tenure were eligible for the REPONSE employee questionnaire, we omit the observations of employees with less than one year of tenure from WERS for comparability.⁶ Employees with more than 12 months’ (or 15 months’) tenure account for more than four fifths of all private sector employees in workplaces with 11 or more employees (87 per cent in France and 83 per cent in Britain).

Our estimation sample is further restricted to workplaces with at least two employee observations, to identify workplace fixed effects. This yields data for 10,645 employees in 1,055 workplaces in WERS 2011 (from a total of 11,581 employee observations in 1,602 workplaces) and for 10,456 employees in 2,985 workplaces in REPONSE 2011 (from a total of 11,244 employee observations in 3,947 workplaces). Our WERS estimation sample thus has greater workplace attrition, primarily due to the higher share of workplaces with no employee responses, but a higher density of employee observations per workplace by virtue of the sample design.⁷

We use survey weights at both the employee and workplace levels, as appropriate, to ensure representativeness, and cluster standard errors at the workplace level to account for the nesting of the employees within workplaces.

3.2 Identifying HWLT workplaces

HWLT workplaces are defined as those offering: (i) high job tenure, conditional on a given set of employee attributes; and (ii) high wages, conditional on those same attributes and on employee tenure. To estimate the workplace component of employees’ wage and tenure profiles, we estimate equations 1 and 2 on pooled data for Britain and France:

$Tenur{e}_{ij}={\beta }_0+{\beta }_{1}Gende{r}_{ij}+{\beta }_{2}Ag{e}_{ij}+{\beta }_{3}Educatio{n}_{ij}+{{\rm Y}}_j+{\epsilon }_{ij}$    (Eq. 1)

$Wag{e}_{ij}={\beta }_0+{\beta }_{1}Gende{r}_{ij}+{\beta }_{2}Ag{e}_{ij}+{\beta }_{3}Educatio{n}_{ij}+{\beta }_{4}Tenur{e}_{ij}+ {\Phi }_j+{\epsilon }_{ij}$    (Eq. 2)

where i indexes employees and j indexes workplaces; Gender indicates the employee’s gender (male/female); Age indicates the employee’s age (with nine age groups, ranging from 16–17 to 65+); Education indicates the employee’s educational attainment (with six levels based on the ISCED (1997) scale, ranging from level 0/1 to level 5A Long/6); Υ_j is the workplace fixed effect on Tenure and Φ _j is the workplace fixed effect on Wage. Both equations are estimated via ordinary least squares (OLS). We take the workplace fixed effects Υ_j (and Φ_j) from these equations to indicate the extent to which, for a given workplace, tenure (or wages) deviate from what a worker could expect to receive elsewhere based on their individual characteristics.⁸

The pooled sample allows us to compare the positions of British and French workplaces within the distribution of Υ_j (and, equivalently, the distribution of Φ_j). We use the median value of Υ_j as a threshold to distinguish between “long-tenure” and “short-tenure” workplaces, and the median value of Φ_j to distinguish between “high-wage” and “low-wage” workplaces. We are then able to identify the shares of workplaces in Britain that lie above or below these thresholds and compare them with the corresponding shares in France. Although our choice of the median is arbitrary to some extent, we consider it the natural choice when seeking to divide a distribution into two groups (long/short or high/low), as it splits the sample into two populations of equal size. HWLT workplaces are defined as those that lie above the median in both distributions.⁹

In WERS, Tenure is measured on an ordinal, four-point scale (1–2 years; 2–5 years; 5–10 years; 10+ years), so OLS is not strictly the most appropriate estimation method. However, comparisons show that linear and ordered models yield similar results (Dickerson, Hole and Munford 2014; Riedl and Geishecker 2014); consequently, we use OLS throughout. Although REPONSE provides a continuous measure of tenure, we transform this into the four-point scale described above for comparability. If we use the continuous measure of tenure in REPONSE, we obtain a workplace fixed effect that is highly correlated with that based on the ordinal variable (r = 0.93). Descriptive statistics show that, among workers with more than one year of tenure (see above), tenure is typically longer in France than in Britain: around half (47 per cent) of sampled employees in France have tenure of 10+ years, compared with 30 per cent in Britain (see table A1 in supplementary online appendix A).¹⁰

Wage is the employee’s weekly or monthly wage divided by the usual number of hours worked.¹¹ For France, we match the REPONSE employee sample to individual records in an administrative database recording gross wages and hours worked (déclarations annuelles de données sociales). For Britain, wages are observed directly in the WERS employee questionnaire, although they are recorded in intervals, so we obtain point estimates for gross hourly wages by dividing the upper and lower bounds of the interval by a continuous measure of hours worked and then taking the midpoint.¹² The lower and upper tails of the WERS wage distribution are longer than indicated by the Structure of Earnings Survey and Labour Force Survey for Britain; we therefore winsorized the WERS wage distribution at the 5th and 95th percentiles. No such winsorization was applied to the REPONSE wage data, which aligned well with other sources for France. The resulting wage distribution is skewed to the right in both countries, and normally distributed after taking logs. We convert the WERS wage measure to euros using the prevailing exchange rate at the time of the 2011 surveys. Wage dispersion – measured by the interquartile range – is higher in Britain (0.77 log points) than in France (0.55 log points) (see table A1 in supplementary online appendix A).

3.3 HWLT workplaces: Connecting wages and tenure

Figure A1 in supplementary online appendix A shows the distribution of workplace fixed effects (Υ_j and Φ_j) for each country. Within the pooled distribution of Υ_j, 34 per cent of workplaces in Britain, compared with 66 per cent of workplaces in France, lie above the median (see tables A2 and A3 in supplementary online appendix A). Turning to the pooled distribution of Φ_j, 45 per cent of workplaces in Britain lie above the median, compared with 55 per cent of workplaces in France.¹³ We define HWLT workplaces as those which are above the median on both dimensions. In the pooled sample, Υ_j and Φ_j are approximately uncorrelated, and 27 per cent of workplaces are above the median on both dimensions. However, the situation differs between the two countries: Υ_j and Φ_j are more highly correlated in France than in Britain. The share of HWLT workplaces is greater in France (38 per cent) than in Britain (16 per cent), and the share of employees is also correspondingly higher (40 per cent, compared with 16 per cent in Britain) (see table 1). As expected, HWLT workplaces are thus more prevalent in France than in Britain.¹⁴

Our analysis focuses on 2011 (the latest year with available data for Britain). However, data for both countries in 2004 and French data for 2017 show that the prevalence of HWLT workplaces in Britain is very similar in 2004 and 2011 (see table B1 in supplementary online appendix B). In France, the share is slightly lower in 2011 than in 2004, but then stable between 2011 and 2017. These estimates support the hypothesis that the results presented elsewhere in this article are relevant beyond the period (2011) featured in most of our analysis.

3.4 Characterization of HWLT workplaces

In the remainder of the article, we describe HWLT workplaces and discuss the implications of these results. We begin by examining which employer and employee demographic characteristics are common to HWLT workplaces across the two countries: does a specific type of workplace provide high wages and long tenure, and do these workplaces tend to employ a particular type of worker? We then explore the HR practices that tend to coincide with the provision of high wages and long tenure.

Our aim is to highlight the potentially specific and homogeneous characteristics of HWLT workplaces. We primarily use a bivariate approach to contrast the average characteristics of HWLT workplaces with those of other workplaces. Our main objective is not to pinpoint the unique characteristics that define HWLT establishments “all other things being equal”. Instead, we aim for a broader understanding of the full range of establishment characteristics and HR strategies that are typical of HWLT establishments. A bivariate analysis also enables us to take full advantage of our rich dataset. Due to variations in the wording of questions or level of measurement (employer versus employee) across countries, certain factors cannot always be comprehensively analysed (or clustered) together in a multivariate analysis.

4.1 Firm characteristics

Consistent with the long-standing literature on the firm-size wage effect (Oi and Idson 1999), and the fact that long tenure relies on a degree of firm longevity, HWLT workplaces in both France and Britain tend to have larger workforces and to be older than other workplaces (see table 2). In Britain, around 1 in 7 HWLT workplaces (14 per cent) have 100 or more employees, compared with around 1 in 12 non-HWLT workplaces (8 per cent), and the ratio of mean workforce size is around 1.5:1 (76 employees versus 51). In France, the pattern is broadly similar, though the average HWLT workplace in France has twice as many employees as other workplaces (mean; 82 employees versus 43). In both countries, approximately three quarters of HWLT workplaces are at least 20 years old; in Britain, this is the case for less than half of all other workplaces, and in France for only two thirds. These associations remain apparent in a multivariate regression combining all employer characteristics (see table A4 in supplementary online appendix A).

We expect HWLT workplaces to operate in high value-added industries with large and stable markets, fuelling stable employment and high pay levels. In both countries, workplaces in the manufacturing/utilities and finance sectors are over-represented within the HWLT group. The construction and other business services sectors also have some degree of over-representation, but the difference is only statistically significant in France. On the other hand, in both countries, wholesale/retail and public administration/education/health are under-represented in the HWLT group (as are other personnel services in Britain, and transport/communications in France).¹⁵ These sectoral patterns are not directly replicated across bivariate and multivariate analysis, but they are broadly similar (see table 2, as well as table A4 in supplementary online appendix A).

In line with the comparative literature on capital ownership and HR practices (Conway et al. 2008; Forth and Rebérioux 2016), we find that, in France, HWLT workplaces are more likely to form part of larger organizations and listed firms, but less likely to form part of family-owned firms (see table 2 for bivariate analyses, confirmed by multivariate regressions in table A4 in supplementary online appendix A). The results are less straightforward for Britain. HWLT workplaces are less likely to be family-owned, but the difference is not significant. The link with listed status is weaker than in France and, again, not significant. Last, we find that HWLT workplaces are less likely to form part of larger organizations in Britain, but the difference is not significant in a multivariate analysis.

In both countries, HWLT workplaces tend to operate in national or international markets, rather than in local or regional markets, and in more mature markets (see table 2). The latter result remains after controlling for other factors (see table A4 in supplementary online appendix A). Turning to the nature of production, the available measures of capital intensity are different across the two surveys – labour costs as a share of revenue and the extent of computer use in Britain, and the extent of computer-controlled and digital systems in France. However, in both countries, the association with HWLT is positive.

Overall, the profile of HWLT workplaces is quite similar in the two countries, with the partial exception of ownership characteristics. The profile of the two economies is somewhat different across the demographic characteristics considered in table 2 (Britain having a higher share of smaller and younger workplaces, for instance). In a pooled regression, the gap of 22 percentage points between Britain and France reduces to 13 percentage points after controlling for (equalizing) those characteristics.¹⁶ Differences in the workplace age profile and industry profile of the two economies are the main contributors to this reduction. Differences in the productive structure of the two economies therefore account for around one third of the differing prevalence of HWLT workplaces across the two countries, but the remainder is accounted for by other factors. We return to this issue later, when discussing the institutional environment in each country.

4.2 Employee characteristics

One expects workers in high-paid, stable jobs to be those with the most market power (those with the best outside options, whom employers will wish to retain and develop). On the employer side, we also expect employers to offer better wages and greater job stability to skilled workers, who are a key determinant for profitability (Doeringer and Piore 1971), and as a means to encourage commitment and effort in jobs where employers have limited direct control (Akerlof and Yellen 1986).

An analysis of the socio-economic profile of workers in HWLT workplaces is presented in table 3, focusing on employees’ education, occupation, age and gender.¹⁷ In both countries, HWLT workplaces are more likely than other workplaces to have a workforce that is highly educated and working in occupations requiring high levels of skill. The workforce in HWLT workplaces is also older and more male-dominated than in other workplaces. The broad similarity of the results for France and Britain indicates a similar logic in both countries. There is a slight difference, however, in that HWLT workplaces in France place even greater emphasis on more-educated employees and higher-skilled jobs.

The bivariate associations are confirmed in multivariate analyses (see table A5 in supplementary online appendix A). The Pseudo-R² of these estimations is very similar across the two countries, indicating that these employee characteristics are of similar importance in each country in explaining the probability of being in an HWLT workplace.

Overall, the analyses of firm and employee profiles show that HWLT workplaces have a number of distinct characteristics, which are fairly similar in both countries. This suggests the existence of a specific employment model. To explore this question further, we examine the HR practices used in HWLT workplaces.

5.1 Incentives based on internal mobility

Workforce stability can be promoted by offering stable contracts or expectations of future rewards. We find that the shares of permanent contracts and full-time workers are higher in HWLT workplaces than in other workplaces, and that the differences are statistically significant in both countries (see table 4). Information on internal mobility is provided in the lower part of table 4. In Britain, the WERS management questionnaire asks whether vacancies are “filled internally by preference”; this is the case in 42 per cent of HWLT workplaces, compared with only 31 per cent of other workplaces. The difference is not statistically significant from zero (p = 0.130) but would reach statistical significance if the WERS sample were the same size as the REPONSE sample. For France, the employee questionnaire yields 2 complementary pieces of information: the share of employees promoted in the past 3 years, and the share expecting a high likelihood of promotion or pay rise in the next 12 months. On both indicators, HWLT workplaces in France tend to offer greater chances of promotion than other workplaces, with the differences being statistically significant.

HWLT workplaces thus appear to offer stabler employment conditions across multiple dimensions, allowing employees to evolve once recruited, by providing them with full-time jobs with open-ended contracts and opportunities for internal promotion.

5.2 Incentives based on skill development opportunities

In addition to employee stability and internal promotion opportunities, we expect HWLT workplaces to provide their employees with opportunities for skill enhancement. In both countries, employees in such workplaces are more likely to receive off-the-job training (see table 5, panel 1). In France, employees in HWLT workplaces are also more likely to offer jobs with opportunities to learn new things. These differences are all statistically significant. In Britain, despite the higher levels of training provision in HWLT workplaces, employees are less likely to feel that their managers encourage them to develop their skills. Overall, however, there are several indications that HWLT workplaces stimulate human capital development among their employees, particularly in France.

5.3 Incentives based on pay policy

The coexistence of high wages and high tenure suggests a tenure-based pay system. To assess whether this is typical of HWLT workplaces, we regress the employee’s log hourly wage on job tenure and interact this with the HWLT workplace indicator, after controlling for the employee’s gender, age and educational attainment. The results show that wages rise with tenure in all workplaces, but that the gradient is steeper in HWLT workplaces (see table A6 in supplementary online appendix A). The interaction term becomes statistically significant in both countries once an employee reaches ten years of tenure.

In recent decades, one distinctive feature of pay policies has been the diffusion of individual and group-level incentive pay systems (e.g. Lemieux, MacLeod and Parent 2009). Collective bonus or profit-sharing schemes are one means by which an employer may seek to align the employee’s incentives with the goals of the firm. Our empirical results show that such schemes are more frequent in HWLT workplaces (see table 5, panel 2); the difference is only statistically significant for France, but it would reach statistical significance for Britain if the WERS sample were the same size as the REPONSE sample. Shareholding and share ownership plans are another means of aligning employers’ and employees’ interests, situating them in a longer-term perspective. Here, the association with HWLT workplace status is apparent only in France. Schemes linking pay to individual performance are common in both Britain and France, but they are associated with HWLT workplaces only in France. Overall, while HWLT workplaces in France often use incentive pay schemes, the same cannot be said of such workplaces in Britain.

5.4 Incentives based on voice mechanisms

Institutional economists (Hirschman 1970; Freeman and Medoff 1984) have long argued that unions – and possibly other forms of voice – can reduce labour exits by providing employees with opportunities to influence their work environment through collective discussion and negotiation. The empirical literature tends to confirm that voice mechanisms are associated with lower labour turnover at the micro level (Bryson and Forth 2010; Amossé and Forth 2016). We therefore expect voice mechanisms to be more prevalent in HWLT workplaces, and our estimations confirm this (see table 5, panel 3): union membership density is higher, and on-site employee representatives, consultative committees and collective bargaining are more common in such workplaces. There is also a positive association between HWLT workplace status and the presence of non-union representatives, at least in France. In that country, the difference between the two workplace profiles is similar to that observed in Britain, but is statistically significant due to the larger sample size. Union and non-union voice mechanisms thus appear to be complementary in HWLT workplaces. By contrast, HWLT workplaces do not appear to differ markedly in their use of direct voice mechanisms. In France, they are more likely to use departmental meetings, but the prevalence of employee attitude surveys and suggestion schemes does not differ by workplace status in either country. HWLT workplaces are thus characterized by greater use of voice, but this is primarily channelled through (union or non-union) representatives rather than through direct mechanisms. This aligns with the notion that different forms of voice have a specific “fit” with particular industries, economic strategies, forms of work organization and management styles (Storper and Salais 1997).

5.5 Work organization

In his review of the HR management performance literature, Wood (2020) distinguishes between two forms of employee involvement: employee participation in decision-making on broad issues (via direct or indirect representation mechanisms); and employee participation in defining the contours of the job (e.g. through teamwork or problem-solving groups). Regarding the latter, the final panel in table 5 shows that employees in HWLT workplaces are more likely to work in teams and have access to problem solving groups, even though the difference is not significant in Britain (and would be significant for teamwork only if the WERS sample were the same size as the REPONSE sample). Here again, British HWLT workplaces display similar tendencies to those observed in French HWL workplaces, but the patterns are less pronounced.

6.1 Drawing the contours of an HR model?

Taken together, the set of HR practices used in HWLT workplaces form a consistent profile combining complementary practices: the promotion of human capital (taking into account job stability, internal promotion opportunities and increasing qualifications), the implementation of incentive methods (in terms of remuneration) and opportunities to involve employees in the decision-making process (with multiple voice mechanisms). This profile echoes the HR model of HPWS studied since the 1990s (Huselid 1995; Applebaum et al. 2000), which emphasizes the bundling of complementary practices designed to promote employees’ knowledge, skills and abilities, motivate them to perform at their best, and provide opportunities for them to contribute and make a difference (Boxall and Purcell 2003). The correspondence is striking in the French case, but more muted in Britain given the weaker relationships between the HWLT classification and the use of incentive pay and employee involvement practices. Such country differences are coherent with earlier findings showing that HPWS are less developed in Britain than in France (Conway et al. 2008), especially in respect of incentive pay (Marsden and Belfield 2010).

To complement this picture, we conduct a multivariate estimation of the likelihood of being an HWLT workplace, using all reviewed HR practices as explanatory variables (see table A7 in supplementary online appendix A). For methodological reasons, our analysis is limited to practices that are measured in the management questionnaire of each survey.¹⁸ The results should not be interpreted causally; instead, our aim is to identify practices that are distinctively characteristic of HWLT workplaces in each country. The results are similar to those discussed in the previous sections. The only notable difference – observed in both France and Britain – is that HWLT workplaces are now negatively associated with direct voice mechanisms. Most HR practices remain significantly associated with being an HWLT workplace, even when controlling for other factors, indicating that they are all distinctive features of the employment pattern we are mapping out.

6.2 Similarities with ILMs

In addition to HPWS, HWLT workplaces also resemble another classic management model: the ILM. Our empirical portrait of HWLT workplaces shows common features with the ILM literature from Doeringer and Piore (1971) to Osterman (2024): workers are often highly skilled men, who are offered full-time, permanent positions with training, tenure-pay systems and internal promotion opportunities. Higher levels of union membership and the presence of union representatives are also associated with the ILM model. Moreover, the fact that wage-setting practices in HWLT workplaces play a key role in tenure, and that wages exceed those that workers could earn elsewhere, is consistent with the core principle of ILMs (as set out by Doeringer and Piore 1971; Osterman 2024), namely, that wages are determined internally and are disconnected from market forces.

On the whole, employee profiles in HWLT workplaces share clear similarities with those proposed under the traditional conception of ILMs (Doeringer and Piore 1971). However, ILM firms were traditionally known for their presence in manufacturing industries and for integrating low-skilled workers and enabling them to progress internally, whereas the HWLT workplaces clearly focus on highly qualified workers and are quite prevalent in finance and business services. This change in ILMs would align with the significant shifts in productive systems in recent decades, particularly deindustrialization and the restructuring of manufacturing sectors, as well as the growing segregation of top earners described by Godechot et al. (2024).

6.3 France versus Britain: One model or two different models?

We can draw on the similarities with ILM and HPWS models to discuss the comparative aspect of our analysis: do French and British HWLT workplaces illustrate different models? This would echo the opposition in the labour market segmentation literature between ILMs (characteristic of France and the US) and occupational labour markets (characteristic of Britain) (Eyraud, Marsden and Silvestre 1990). Alternatively, we might consider that the French and British situations represent two ways of implementing the same model: each country illustrates how the model adapts to the prevailing economic and institutional context (Grimshaw and Rubery 1998; Doellgast and Marsden 2019; Amossé et al. 2025). Given that the differences highlighted between Britain and France primarily concern the extent of certain practices (e.g. incentive pay or participative forms of work organization), rather than the use of different practices, we hypothesize that there is a single model of HWLT workplaces. The main difference is that the HWLT profile is more clearly apparent in France. This may be due to the smaller sample size for Britain, but it could also suggest a less distinctive profile. The latter hypothesis would be consistent with earlier literature describing ILMs as typical of France (Eyraud, Marsden and Silvestre 1990). Further research would be needed to establish whether an employment model similar to that of the occupational labour market remains typical of the British labour market.¹⁹

How can the national context thus explain country differences? The arguments developed by Marsden and Belfield (2010) and Amossé et al. (2025) are compelling here. First, employer and employee networks are denser in France (Amossé et al. 2025). Employers often belong to employer organizations, especially in large firms, which tend to have large HR departments. HWLT workplaces often form part of larger firms. Marsden and Belfield (2010) argue that such networks help to disseminate new HR practices, which might explain why HWLT workplaces are more widespread in France. Second, Amossé et al. (2025) argue that differences in corporate governance and financing of firms are also crucial in understanding national models. They argue that British managers tend to be more autonomous in their decision-making and less dependent on other organizations. This may also slow down the spread and/or standardization of practices. Finally, differences are likely to stem from different regulations and legal standards (Amossé et al. 2025). The French State plays a more central role in the labour market, and employment protection legislation is stricter in France, which may lead employers to find new ways to manage and retain workers. Marsden and Belfield (2010) have highlighted how employment protection legislation can discourage the use of dismissal threats and encourage alternative HR policies, such as the use of incentive pay. But regulations and legal standards do not refer to stricter redundancy rules alone. For example, in France, there exist a distinct set of laws encouraging (fiscally) the use of performance-related pay schemes and even making it compulsory for the largest companies to use certain remuneration tools (in a profit-sharing logic). All in all, these characteristics of the French and British national models may be the reasons why HWLT workplaces are less widespread in Britain, and why their characteristics are less distinctive when they exist.