R E S E A R C H A R T I C L E
Empirically exploring the veracity of the new stakeholder perspective in strategy:
Documenting workforce rents
Jeroen Neckebrouck1 | David Kryscynski2
1Assistant Professor of Entrepreneurship, IESE Business School, Barcelona, Spain
2Associate Professor of Human Resource Management, Rutgers University, Piscataway, New Jersey, USA
Correspondence
David Kryscynski, Associate Professor of Human Resource Management, Rutgers University, 94 Rockafeller Road, Piscataway, NJ 08854, USA.
Email:[email protected]
Abstract
Research Summary: Without compelling empirical proxies for economic profits, we may need to reconsider the decades of empirical research purporting to inform our theories of competitive advantage. The new stake- holder perspective suggests that stakeholders may cap- ture significant shares of the firm's economic profits that should be incorporated into these proxies. In this article, we propose a novel empirical approach to mea- suring stakeholder rents and then apply our approach to measure workforce rents across the population of all Belgian firms employing workers from 2008 to 2016.
Our results demonstrate substantial variance in work- force rents among firms, with some firms allowing most of the economic profits they generate to flow to the workforce. We discuss the implications of our findings in detail and lay out a pathway for future research.
Managerial Summary: This article examines the extent to which companies pay their workforces above (below) what the labor market demands as a way of exploring how much of the company's economic profits go to stakeholders other than shareholders. We demon- strate a wide range of over (under) payments to
This is an open access article under the terms of theCreative Commons Attribution-NonCommercialLicense, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.
© 2024 The Authors.Strategic Management Journalpublished by John Wiley & Sons Ltd.
Strat Mgmt J.2024;45:2159–2188. wileyonlinelibrary.com/journal/smj 2159
workforces in a large sample of Belgian firms from 2008 to 2016. One of the important contributions of our paper is developing a method to determine over (under) payments for the workforce, but our method can also be applied to other stakeholders. We hope our work provides an empirical approach for others to explore how stakeholders capture portions of the economic profits that companies create.
K E Y W O R D S
human capital, measuring economic profits, new stakeholder theory, value appropriation, value creation
1 | I N T R O D U C T I O N
Without compelling empirical proxies for economic profits, we may need to reconsider the decades of empirical research purporting to inform our theories of competitive advantage. The
“new stakeholder (SH) perspective” in the resource-based view (RBV) suggests that to truly understand competitive advantages we must account for the economic profits created by and then appropriated by the nexus of SHs that contribute to the firm's economic activities (Barney, 2018, 2020; Jourdan et al., 2021; Kaplan, 2020; McGahan, 2021; Stoelhorst, 2023).
Scholars embracing this perspective point out that much of our strategy theorizing focuses on understanding which firms create more economic profits than competitors, but much of our strategy empirical work examines differences in shareholder value capture across firms (Bowman & Ambrosini,2000; Coff, 1999). Recent work suggests the importance of more care- fully exploring both value creation and value appropriation by various SHs (Garcia-Castro &
Aguilera,2015; Lieberman et al.,2017; Lieberman et al.,2018). This is particularly important if SHs capture rents—that is, payments above their opportunity costs (Coff, 1999). Inasmuch as different SHs capture rents then a true estimate of a firm's economic profits is likely the sum of rents captured by all the firm's internal SHs (Barney,2018; Coff,1999). It thus seems critical to empirically explore SH rents to better understand when and why some firms generate more economic profits than others.
Unfortunately, however, prior strategy research exploring SH rents is primarily theoretical.
For example, Coff (1999) theorizes that the workforce may be in a privileged position to bargain for rents, Barney (2018) argues that non-shareholder SHs must be able to capture rents as a pre- condition for the firm to create any economic profits, Jourdan et al. (2021) argue that the amount of value SHs create by engaging with the firm is likely a function of how much of that value they are able to capture, and Stoelhorst (2023) builds a theoretical model explaining how payments to different SHs emerge as a result of how they participate in value creation. Coff (1999) even suggests that some firms may have competitive advantages by creating more eco- nomic profits than competitors, but that these competitive advantages may not appear in con- ventional accounting performance measures because much of those profits are captured by the workforce. Barney (2020) further argues that strategy research in the RBV tradition needs a new empirical measure of firm performance that in some way accounts for the rents captured
by SHs, which may not be directly tracked in traditional accounting measures of performance.
While these theoretical arguments for why we must more fully account for SH rents are com- pelling, we lack empirical work to verify the extent to which these SH rents are substantive. Put differently, it is not clear whether these SH rents are substantively important empirical realities for firms or just a theoretical exercise in the new SH perspective.
In this article, we develop a general approach to measuring SH rents based on Coff's (1999) definition of rents as the payments SHs receive above their opportunity costs. We argue that a rea- sonable empirical proxy for opportunity costs is what we call“market-appropriate payments”—
that is, what the market would pay a particular SH for their contributions. In other words, we assess information in the market to determine the payments an SH could reasonably expect from the market if they were to leave the focal firm, using these payments to estimate the SH's opportu- nity cost associated with staying in the focal firm. We then calculate SH rents as the difference between observed real payments to an SH and the estimated market-appropriate payments to that SH. This novel approach to estimating SH rents allows us to examine variance in rents flowing to SHs and, therefore, provides a way to determine the extent to which this variance is substantive for a population of firms. In other words, our approach shifts the new SH perspective in the RBV tradition from a theoretical discussion to an empirical investigation.
We demonstrate our empirical approach by estimating workforce rents in a sample of 14,312 Belgian firms from 2008 through 2016. To illustrate the importance of our workforce rents esti- mate we report a set of comparisons between workforce rents and other commonly used firm per- formance indicators such as net accounting profits, gross added value, shareholder returns, and so forth, and present a set of charts that help to illustrate the variance in workforce rents relative to several of these performance metrics. For example, finding that the standard deviation in our workforce rents estimate is approximately 86% of average net accounting profits in our sample suggests that workforce rents are substantively important. We also examine a host of potential challenges to our empirical proxy to demonstrate its robustness. We infer from our results that there is meaningful variance in workforce rents across firms and that workforces in some firms may capture substantial portions of the economic profits the firm creates.
Accordingly, our work lends empirical credence to the theoretical issues raised by the new SH perspective in the RBV tradition (Barney,2018,2020; McGahan,2021) as well as the grow- ing value creation and appropriation (VCA) perspective (Garcia-Castro & Aguilera, 2015;
Lieberman et al.,2017; Lieberman et al.,2018). Our findings illustrate that failing to account for workforce rents, as a special case of SH rents, may lead to significant underestimations of the true economic profits a firm generates. Accordingly, any empirical analyses purporting to draw inferences about competitive advantages from performance measures that do not incorporate these rents may be inaccurate at best, but also potentially misleading. Our empirical findings suggest a need to reconsider some of our classic empirical findings in strategy with a more robust accounting for at least workforce rents, and potentially for other forms of SH rents as well. Our work also contributes by providing a general approach to estimating SH rents that may be applicable for multiple SHs in multiple contexts. In addition, our work has several important implications for strategic human capital research.
2 | W H Y S H R E N T S M A T T E R
To motivate our research question it is first useful to explain why SH rents matter. Following Coff (1999), Stoelhorst (2023) and others (e.g., Blair & Stout,1999), we conceptualize the firm as
a nexus of contracts between SHs that is primarily designed to solve the team production prob- lem (McGahan,2021; Stoelhorst,2023). For our purposes, we narrowly define SHs as those who make a material contribution to the ongoing operations of the firm1(e.g., employees, managers, suppliers, and so forth). We note that while firms tend to transact in monetary currencies, indi- viduals transact in utils. Accordingly, SH rents are the total utility (both monetary and non- monetary) that an SH receives in the focal firm above and beyond their opportunity costs, or their next best total utility option at another firm (Coff,1999; Garcia-Castro & Aguilera,2015).
The next best option for any given SH could be a single firm that uniquely values the SH's resources, or a group of firms who similarly value the SH's resources at some market appropri- ate level. Like Coff (1999) and Stoelhorst (2023), we assume that SHs stay and contribute as long as the total utility they receive from the focal firm is at least as great as their opportunity cost (i.e., when the rents that they receive are positive). Any economic profits generated by the nexus of SHs due to team production, co-specialized assets, firm-specific investments, and so forth, can be distributed to SHs as rents.
Conceptualizing the firm as a nexus of contracts is important because the traditional defini- tion of a competitive advantage in the RBV is when a firm generates more economic profit than the breakeven competitor (Hoopes et al.,2003; Molloy & Barney,2015; Peteraf & Barney,2003).
When economic profits are defined as the difference between a firm's revenues and the total of SHs' opportunity costs (Brandenburger & Stuart, 1996; Garcia-Castro & Aguilera, 2015;
Stoelhorst, 2023), any rents that an individual SH might capture are technically part of the economic profits created by the firm, even if these rents are due to co-specialized assets, firm- specific investments, owner benevolence, and so forth. In fact, as Barney (2018) argues, the exis- tence of these economic profits is likely the direct result of firm-specific combinations of resources that various SHs bring to the table. Thus, the total economic profits a firm generates is equal to the sum of SH rents (Barney, 2018; Coff, 1999; Peteraf & Barney, 2003;
Stoelhorst,2023).
Strategy scholars embracing this “new SH theory” perspective explicitly assume that SH rents exist and suggest that we must find more compelling ways to account for these rents (Barney,2018,2020; Jourdan et al.,2021; Kaplan,2020; McGahan,2021; Stoelhorst, 2023). As mentioned previously, however, these discussions are primarily theoretical and leave open the question of whether SH rents are substantive in practice. We certainly do not mean to imply that no scholars have done empirical work to explore SH value creation and capture. To be sure, there is an entire field of SH research that examines different empirical proxies (such as ESG and KLD scores) that may represent value captured by different SH groups. These approaches, however, do not seek to explicitly proxy for SHrents—that is, the extent to which SHs receive payments above their opportunity costs.
One important example of empirical work that does exist in this vein is that of Lieberman et al. (2018) and Lieberman et al. (2017), who explore the VCA framework. While an important step in documenting the processes through which value is created and appropriated, this work seeks to document how firms' gains are distributed to different SH groups—that is, to what extent do different groups experience increases (decreases) in their value capture as the firm generates more (less) value over time. Their model elucidates insights regarding how value is appropriated by different groups but does not explicitly model opportunity costs and, therefore,
1We note that our arguments may also hold for more distal stakeholders, but exploring this is beyond the scope of the present paper. We reserve the term shareholders for a particular class of stakeholders that contributes financial capital and has formal ownership in the firm.
does not capture rents (Garcia-Castro & Aguilera,2015). Simply put, the model explains where gross revenues go, but does not clarify which SHs capture rents.2
By empirically exploring an issue that until now has remained largely theoretical, the cur- rent paper addresses an important topic in the RBV tradition by explicitly modeling SH rents:
Research Question. To what extent are SH rents substantively important, and, if substantive, what implications do these rents have for modern strategy theory?
3 | D E V E L O P I N G A N E M P I R I C A L P R O X Y F O R S H R E N T S While SH rents seem a central concept in the RBV tradition, we are not aware of empirical work that explicitly models these rents. The closest we find is the employee rent-sharing litera- ture in labor economics (Card et al.,2018), but this vast literature focuses on examining general labor market-level patterns of workforce rent-sharing rather than examining firm-level variance in rents. As mentioned above, we define SH rents as the difference between an SH's total utility and opportunity costs (Coff,1999):
SH Rentsi,k=SH Total Utilityi,k−SH Opportunity Costi,k ð1Þ
where SH Total Utility captures the sum of monetary and nonmonetary utility the SH receives from the firm, SH Opportunity Cost captures the SH's next best total utility option in the mar- ket, the subscript idenotes the individual SH and the subscript k denotes the firm. Thus, SH rents represent the total utility above and beyond what the SH could reasonably receive elsewhere.
In many contexts, researchers can observe various forms of SH utility. For example, we may be able to observe wages and/or fully loaded labor costs for workforces, we may be able to observe transaction prices for suppliers, and so forth. It is much harder to observe opportunity costs, however. Some SHs may have private information about the value of their resources for one or more specific firms in the market. Consider, for example, a research scientist with exper- tise in a particular area that only very few companies are working on. While most companies might only pay the researcher an average“market appropriate”salary, the small set of compa- nies working on this issue might be willing to pay much higher than market average wages to hire this scientist. If this scientist knows about this private value to each of these firms, then the scientist may be able to evaluate the utility she would receive at each of these firms and parse out quite accurately the total utility of the second-best option. Accordingly, this scientist's per- ceived next best option is not market average total utility but, instead, the total utility available at one of these few firms that uniquely values the scientist's human capital. This unique
2For example, some have identified the“great resignation”and/or“great renegotiation”of the COVID-19 pandemic— that is, a time period when broad swaths of employees were quitting, some were opting out of work altogether, and many were identifying that the pay available in the current labor market was simply not worth the risks of returning to work. The VCA model might find that the shares of value creation to labor increased during the pandemic as employees collectively bargained for value. But, if the overall payments to employees increased in the broad market, then employee outside options also increased. Accordingly, these higher payments may be simply adjustments to the new“market- appropriate”payments rather than sharing rents with the workforce. Our approach seeks to specifically examine the rents rather than examining shares of the overall value.
second-best option is likely very difficult (or impossible) for outsiders to observe and is likely not the most common experience of employees in the labor market.3
For most SHs, however, it is not entirely clear what the second-best option means in a real market. SHs' resources may be valued differently at different firms for a host of idiosyncratic reasons, firms may have different sources of both monetary and nonmonetary utility available to SHs, and firms may vary in their need for certain SHs at different times. Accordingly, it may be difficult for individual SHs to evaluate what their second-best option may be. Accurate assessments would require an SH to assess reasonable options available in the market, accu- rately evaluate and compare various sources of utility available at these other options compared to the focal firm, and accurately assess their likelihood of successfully switching to the new firm (i.e., evaluating accurately whether the new firm is willing to do business with them). The cog- nitive burden and information availability requirement make such an accurate assessment pro- hibitively costly for the typical boundedly rational individual SH. In this sense, “those that provide resources to a co-specialized bundle may not fully understand the opportunity costs of doing so”(Barney, 2018, p. 3316). If it is this difficult for individual SHs to know their own opportunity costs, then it seems even less likely that outside researchers might be able to accu- rately assess these alternatives.
In the absence of a clear observable indicator of what an SH's true next best option might be, we suspect that SHs create their own approximations of their next best options from available mar- ket information. A supplier might not be able to determine perfectly what the next best firm will pay for their products but may be able to observe the general market price. An employee may not be able to parse out exactly what total utility they will get from any specific firm (without an explicit job offer) but may be able to use market data from employee review platforms, conversa- tions with peers, informal assessment of nonmonetary factors, and so forth, to determine their market-appropriate total utility. Research faculty, for example, may have a hard time knowing what university might truly be the next best fit, may not know whether that university may have lines open, the extent to which they will connect with other faculty, and may not be able to deter- mine what compensation that university will offer. However, research faculty can likely determine what general market level total utility options are at universities in their similar tier for faculty at their similar rank and comparable productivity. Thus, in practice, we argue that SHs are likely to use their assessment of market-appropriate total utility as their own internal proxy for their next best options in the market. If so, an SH'smarket-appropriate total utilityis the level of payments and other nonmonetary benefits that an SH could reasonably expect to receive in the market when leaving the focal firm. Thus, SH rents can be approximated by the difference between the total util- ity they receive in the focal firm and their market appropriate total utility as follows:
SH Rentsi,k=SH Total Utilityi,k−SH Market Appropriate Total Utilityi,k ð2Þ
This is important because opportunity costs are generally not observable to the researcher and imperfectly observable to the SH while market-appropriate total utility is somewhat observ- able to the SH and estimable by the researcher. To estimate an SH's market appropriate total utility, we need to consider all factors that would predict what an SH should expect to receive
3The extent to which individuals rely on private information about dyad-specific second-best options versus general information available in the market to assess outside options is ultimately an empirical question that is beyond the data available in the present study. For empirical convenience we assume that assessing dyad-specific options is too difficult for the typical employee such that most stakeholders rely on general market information.
from a typical firm in that particular market. There are at least two categories of factors to con- sider. First,SH characteristicsare observable SH-level factors that firms believe affect the value creation potential of the SH. We assume that firms will pay more for SHs with greater value cre- ation potential and pay less for SHs with lesser value creation potential. If the SH is a supplier then these factors may include the quality of the supplier's products, the reputation for quality, and so forth. If the SH is an employee then these factors may include the employee's education level, prior work experience, general mental ability, and so forth. Second,market characteristics are factors that might affect what typical firms may need to pay for SHs independent of their SH characteristics. These factors may include the number and density of firms in the market, the ratio of SHs to firms in the market, the overall munificence of the market, and so forth. An SH with certain characteristics may demand very different payments in different markets based on local market conditions. Thus:
Market Appropriate Total Utility=f Market Characteristics,Stakeholder Characteristicsð Þ ð3Þ
Note that Equation (3) doesnotinclude factors that are specific to the SH's current firm such as the firm's philosophy toward SH payments, the firm's future aspirations for SH support, or the firm's ability to bargain for shares of economic profits vis-à-vis the SH. Equation3also does notinclude firm-SH dyad specific factors in the focal firm such as co-specialization between the SH and the firm, or the SHs' unique ability to bargain with the current firm for payments. These factors likely affectactual current payments to an SH in a specific firm, but they should not affect market-appropriate payments to the SH. In other words, these firm-specific and dyad- specific factors may affect what aspecificfirm might pay to aspecificSH, but they do not affect outside options and, therefore, they should not affect market appropriate payments.
Consider, for example, an SH that makes a highly firm-specific investment that leads to unique value creation in the focal firm. This SH creates some value that no other firm-SH combi- nation could create. Classic theory may predict that the SH and firm will share the extra value created because this is a bi-lateral monopoly, and both parties are essential (Campbell et al.,2012;
Lazear,2009). We fully agree that it may be economically rational for both shareholders and SHs to share the resulting rents to preserve their unique value creation arrangement. It is important to note, however, that this rent sharing arrangement is precisely that—the sharing of rents above and beyond each partys' opportunity costs. Rent sharing in a bi-lateral monopoly like this does not change either SHs' outside options. Thus, it is important that Equation3focuses on market and SH characteristics that arenotfirm- or dyad-specific with the current firm.
If we have high-fidelity measures of (1) the actual total utility that SHs receive, (2) market characteristics, and (3) SH characteristics, we can construct a model predicting SHs' actual total utility only using these market and SH characteristics. The predicted values from such a model will then be equivalent to the SH's market appropriate total utility, and any unexplained vari- ance in such a prediction model will represent the extent to which that specific firm's payments to that specific SH deviates from the predicted market appropriate total utility. This unexplained variance, then, represents the rents that the specific SH captures from that specific firm. We elaborate on these arguments in detail in Online Appendix1. A practical representa- tion of this approach is as follows:
Stakeholder Total Utilityi,k¼αþβMCX Market Characteristicsk þβSCX Stakeholder Characteristicsiþεi,k
ð4Þ
where the subscriptidenotes the SH, the subscriptkdenotes the firm,βMCrepresents the vector of coefficients for the market characteristics,βSC represents the vector of coefficients for the SH characteristics, andεi,krepresents the unexplained residual, or the SH rents. If we were to esti- mate Equation (4) using ordinary least squares (OLS) regression the mean value of the residuals is zero, so thatbεis best interpreted as a mean-centered estimate of SH rents.4Our core empiri- cal proposition, therefore, is as follows:
Empirical Proposition 1. The residual of a well-specified empirical model predicting actual SH total utility using (1) high-fidelity measures of SH payments, (2) high-fidelity measures of market characteristics and (3) high-fidelity measures of SH characteristics will provide a high-quality proxy for mean-centered SH rents.
4 | E M P I R I C A L I L L U S T R A T I O N : T H E C A S E O F W O R K F O R C E R E N T S
We illustrate our approach to estimating SH rents by focusing on a special case: workforce rents. Our data have rich workforce and labor market measures that allow us to develop a high- fidelity proxy for market-appropriate monetary payments and, therefore, for mean-centered monetary workforce rents. We assume that market-appropriate workforce payments are deter- mined by workforce and labor market characteristics only. Accordingly, we regress total pay- ments to the workforce on a set of labor market and workforce factors, and use the residual of this regression as our proxy for mean-centered workforce rents.
We recognize that our results require several empirical assumptions and decisions that could directly affect our interpretation. Accordingly, we describe in our main analysis the empirical path we believe is most appropriate given our research goals, but then follow the rec- ommendations of King et al. (2021) to analyze a more complete set of empirical choices. In our additional analysis section, we lay out an epistemic map of our choices and then show the results conditional on different choices in the map.
4.1 | Sample and data collection
We leverage data from Bureau van Dijk, one of Europe's leading electronic publishers of busi- ness information. We gathered data for the population of Belgian firms from 2008 to 2016.
Belgium represents a unique“laboratory”in which to examine our research question as all Bel- gian firms that employ staff— irrespective of their size and age—have a legal obligation to annually file detailed information about their workforce, along with their financial accounts, in a predefined format with the Belgian National Bank. Belgian firms are required to report more than 50 employee-related variables on a yearly basis: number of people employed, personnel movements, gender, occupation, and so forth. We started our data collection in 2008 because it was not until that year that firms were also required to publish information on employees'
4In Online Appendix1, we explain in detail why this approach cannot parse out the mean level of workforce rents without additional information, but since our goal is to explore the between-firm variance in workforce rents this mean- centered estimate is sufficient for our exploration.
education level. To be part of our sample, we first required firms to have at least 10 employees (or, more accurately, full-time equivalents [FTEs]) during at least 1 year of the 2008–2016 period (Neckebrouck et al., 2018). We applied this cutoff because the Belgian National Bank allows smaller firms to submit more limited information about their workforces. Our final data set consists of 74,455 firm-years of data from 14,312 firms.
4.2 | Dependent variable
We measured workforce payments as the natural log of total labor cost an employer pays each year divided by the total number of employees (measured in €000 s per FTE). Total labor cost includes salaries and other costs insofar as they are part of an employment contract—for example, direct social benefits, overtime, benefits in kind (car, mobile phone, etc.) and bonuses granted as part of an employment contract. Social security taxes that employers pay directly to national social security funds that do not show up in employees' gross income are excluded. These total labor costs cannot capture many of the inherently nonmonetary aspects of total utility such as the firm's reputation, confidence in company leadership, and so forth.
4.3 | Market-appropriate payment predictors
Following the logic from Equations (1)–(4), we sought to hold constant overall workforce characteristics (factors that determine what a typical firm will pay for this workforce) and labor market characteristics (factors that determine what a typical firm will pay in this mar- ket). We provide a full list of our proxies and descriptions in Table1. We note that our work- force and market characteristics proxies are incomplete. With these proxies, we could have observationally equivalent workforces that are nevertheless different in their actual value cre- ation potential in the firm. Our Online Appendix1 explores the conceptual implications of incomplete SH and/or market characteristics in our proxies and our additional analysis section explores several ways to rule out concerns about incomplete proxies in our main analysis.
5 | R E S U L T S
Table2presents the means, standard deviations, and correlations of the variables.
5.1 | Main model
Table 3 shows the results from OLS regressions for market-appropriate workforce payments.
Note that we are not testing the effect of any variable but, instead, trying to develop a reason- able proxy for market-appropriate workforce payments. The adjustedr-squared in the model is .65 indicating that the market-appropriate payment predictors explain about 65% of the observed variance in actual workforce payments.
T A B L E 1 Predictors table with descriptions.
Predictor
category Predictor Description and inclusion logic Firm-level
workforce characteristics
Occupation Percentage of workforce in management, blue collar, and white collar positions (white collar excluded as reference group) Education level Percentage of workforce with primary education, secondary
education, tertiary education and university education (primary education excluded as reference group)
Gender Percentage of workforce who are identified as male. While the gender pay gap has decreased over time (Gayle & Golan,2012), studies still consistently reveal that women tend to be paid less than men in comparable jobs (see Joshi et al. (2015) for a recent meta-analysis)
Hours worked Average number of hours worked per employee in a firm-year (in thousands). Included because of the general relationship between working hours and productivity (Collewet &
Sauermann,2017; Feldstein,1967; Pencavel,2015)
Contract type Percentage of workforce on permanent contract (excluded as reference group), fixed term contracts (those with a specified end), replacement contracts (those designed specifically to replace employees on leave), and task-specific contracts (contracts for specific work outcomes)
Labor market conditions
Regional
unemployment rate
Percentage of unemployed individuals in the active labor force in the province where the firm is located at the end of each year, data collected from the National Bank of Belgium. Regional unemployment rates are generally considered a useful reflection of labor market conditions and typically serve as a proxy for the probability of receiving a wage offer (Hausknecht &
Trevor,2011; Makarius & Stevens,2019) Regional voluntary
turnover
Percentage of employees in a province who voluntarily left their employers in a given year
Geographic concentration
Location quotients measure the industry specialization of regions (e.g., Delgado et al.,2010). When industry specialization is high, employees may believe there is a higher probability they will be offered a job and/or will receive a premium for their industry-level experience. Location quotients are calculated by the following formula, Location quotient=(Ei,r/Ei)/(Er/E), whereby Ei,ris employment in industry i in region r, Eiis total employment in industry i, Eris employment in region r and E is total employment. A value of one means that region r has the same share of employment in industry i as the rest of the country. A value greater than one means that region r has a higher share of employment in industry i than the rest of the country.
Region level firm density
Natural log of number of firms operating in the region
Border region Four dummy variables indicating whether the firm is located in a two-digit postcode area that borders Netherlands, Germany,
T A B L E 1 (Continued) Predictor
category Predictor Description and inclusion logic
Luxembourg, or France. These control for potential differences in the outside employer options of employees working in border regions in Belgium (Broersma et al.,2022)
Region-level firm size
Average size of the firms active in the region of the focal firm, excluding the focal firm, measured as the natural log of total FTEs. We include this because studies have consistently found a positive relation between firm size and employee wages, controlling for worker's characteristics (Brown & Medoff,1989;
Burdett & Mortensen,1998; Michelacci & Quadrini,2009; Oi &
Idson,1999; Zabojnik & Bernhardt,2001) Region-level firm
age
The average age of the firms (natural log of years since
founding) active in the region of a focal firm, excluding the focal firm, measured as years since founding. We include this because of the relationship between firm age and employee
compensation (Brown & Medoff,2003).
Region-level capital intensity
Average capital intensity—that is, log ratio of the firm's fixed assets to the number of its employees in FTEs—of all firms active the focal firm's region, excluding the focal firm. We include this because of the relationship between capital intensity and employee wages (see Brown & Medoff,2003; Oi &
Idson,1999) Region-level asset
tangibility
Average asset tangibility—that is, the ratio of fixed assets to the book value of total assets—of all firms in the focal firm's region, excluding the focal firm. We include this because of the negative relationship between asset tangibility and compensation levels of a firm (Aggarwal & Samwick,1999; Akyol &
Verwijmeren,2013) Industry-level
market munificence
Measured as the average growth in industry sales (in nominal value) over the prior 5 years, as employees are more likely to receive wage offers when working in munificent industries (Keats & Hitt,1988). Industry categorizations were based on three-digit NACE codes (European Industry Classification system).
Industry market uncertainty
Because environmental uncertainty may influence employees' likelihood to alter employer, we also addedMarket uncertainty and measured it as the instability of industry (three-digit NACE code) sales over the prior 5 years (Bergh & Lawless,1998; Keats
& Hitt,1988). Specifically, market uncertainty was measured by regressing industry sales in the five prior years against time, and subsequently dividing the standard errors of the regression slope coefficients by the mean value of industry sales over this period (Dess & Beard,1984).
Industry level firm density
Because of the relationship between industry competition and wages (Blanchard & Giavazzi,2003), we also includedfirms in industrymeasured as the natural log of the number of firms in the focal firm's industry (three-digit NACE code).
5.2 | Calculating and interpreting workforce rents
As mentioned previously, we estimate workforce rents by subtracting the estimated market- appropriate workforce payments for each firm-year observation5from actual workforce payments, which is equivalent to the residual from the regression described above. This residual is a mean- centered proxy for workforce rents. We find that the standard deviation, indicating how much work- force rents vary across observations in our sample, is 6432 EUR per FTE. For reference, the average payment per employee in our firms is 24,860 EUR. This means that our rents standard deviation is about 26% of the size of the mean employee payment in our sample—that is, a firm that is one SD above the mean in workforce rents pays the workforce 26% more than a firm at the mean level. Fur- ther, the standard deviation of our workforce rents proxy is about 10% of the mean gross added value (revenues minus cost of goods sold) for all firms in our sample (i.e., 66,101 EUR per FTE), and 86%
of the mean net profits (of 7450 EUR per FTE). From a workforce perspective, these differences in rents seem substantial. A histogram of the workforce rents proxy is shown in Figure1.
To further illustrate the practical importance of the variance of our workforce rents proxy, Figure2 shows a series of scatter plots of our mean-centered workforce rents proxy with key financial metrics—that is, revenues, gross added value, EBITDA, and net profits. We note sev- eral important observations about these scatters. First, there is a broad scatter without an obvi- ous visual correlation between mean-centered workforce rents and these financial metrics.6 This means that the rents captured by the workforce may not directly correlate with the amount of revenues, gross added value, EBITDA or net profits a firm reports. Second, we add a handful of benchmark lines to the charts to aid interpretation. These are not regression lines, but simply visual cues to help interpret what observations along these lines mean. In the scatter of work- force rents and revenues, for example, the steepest solid line represents a slope of 0.05—that is, any observation along that line has 5 euro in mean-centered rents for every 100 euro in
T A B L E 1 (Continued) Predictor
category Predictor Description and inclusion logic Industry fixed
effects
Capture remaining industry-level (three-digit NACE code) heterogeneity.
region-year fixed effects
Capture remaining unobservable regional heterogeneity and unobservable macroeconomic effects, including (a) time variant factors not accounted for elsewhere, such as recessions or tax regime changes, (b) time invariant factors that are specific to the region but not accounted for elsewhere, such as proximity to other non-Belgium labor markets and/or local immigration, and (c) any unique interactions between time and region such as situations in which an economic downturn or a temporary influx of refugees disproportionally affects some regions relative to others.
5We are using point estimates here, but there are confidence intervals around these point estimates that deserve consideration. We address this in our additional analyses.
6The correlations between workforce rents and revenues (0.09), workforce rents and gross added value (0.20), workforce rents and EBITDA (0.13) and workforce rents and net profits (0.10) are modest.
TABLE2Summarystatisticsandcorrelations. MeanSD12345678 1Workforcepayments3.430.35 2Management0.010.040.15 3Bluecollar0.510.36−0.53−0.14 4Secondaryeducation0.560.36−0.25−0.060.24 5Tertiaryeducation0.170.230.440.09−0.53−0.44 6Universityeducation0.070.140.500.15−0.46−0.370.28 7Gender0.660.300.05−0.040.480.10−0.23−0.16 8Hoursworked1.560.250.400.02−0.23−0.050.120.110.01 9Fixedtermcontracts0.050.10−0.20−0.010.000.07−0.06−0.03−0.150.04 10Replacementcontracts0.000.03−0.050.02−0.16−0.020.120.01−0.32−0.09 11Task-specificcontracts0.010.02−0.01−0.01−0.010.000.010.000.020.01 12Regionalunemploymentrate0.080.050.040.08−0.170.020.090.18−0.12−0.04 13Regionalvoluntaryturnover0.170.02−0.040.02−0.060.010.010.05−0.050.04 14Locationquotients1.281.280.040.010.010.010.000.040.02−0.04 15#Firmsinregion5.500.850.11−0.01−0.04−0.110.030.01−0.010.09 16Regional-levelfirmsize3.680.100.290.04−0.27−0.170.180.23−0.150.09 17Regional-levelfirmage2.900.180.10−0.01−0.04−0.070.060.04−0.10−0.03 18Regional-levelassettangibility0.260.03−0.28−0.070.300.14−0.19−0.260.13−0.10 19Regional-levelcapitalintensity3.170.16−0.09−0.050.220.02−0.13−0.160.130.01 20Marketmunificence1.070.16−0.050.01−0.13−0.030.060.03−0.22−0.01 21Marketuncertainty0.110.210.030.000.02−0.01−0.010.000.060.04 22#Firmsinindustry5.921.13−0.05−0.03−0.040.07−0.04−0.130.000.04
TABLE2(Continued) 9101112131415161718 10Replacementcontracts0.09 11Task-specificcontracts0.000.00 12Regionalunemploymentrate0.120.070.05 13Regionalvoluntaryturnover0.050.020.020.33 14Locationquotients−0.01−0.020.000.000.09 15#Firmsinregion−0.15−0.07−0.01−0.10−0.17−0.13 16Regional-levelfirmsize−0.060.020.000.26−0.06−0.040.35 17Regional-levelfirmage0.020.09−0.01−0.02−0.25−0.01−0.020.42 18Regional-levelassettangibility0.060.010.00−0.41−0.030.04−0.36−0.830.02 19Regional-levelcapitalintensity−0.07−0.07−0.04−0.49−0.060.010.01−0.42−0.170.40 20Marketmunificence0.070.120.000.060.11−0.030.00−0.08−0.18−0.02 21Marketuncertainty−0.02−0.030.000.00−0.060.030.000.020.090.00 22#Firmsinindustry−0.020.05−0.02−0.050.00−0.210.02−0.03−0.060.00 192021 17Marketmunificence−0.09 18Marketuncertainty0.04−0.15 19#Firmsinindustry0.010.03−0.06 Note:N=74,455.Correlationslargerthanj0.01jaresignificantat1%level.SD=standarddeviation.
revenues. Since Figure2 shows mean-centered rents, these lines likely underestimate the real ratio of workforce rents to each key metric. While the solid lines represent mean-centered work- force rents, the dashed lines represent how the solid lines would shift if we applied an over- simplified assumption that there can be no negative rents in our estimates (we discuss this in more detail in our additional analysis section below). In this case, we would shift every observa- tion upward by the absolute value of the largest negative workforce rent estimate in our sample.7
T A B L E 3 Model for market-appropriate workforce payments.
DV=workforce payments
Coef. SE p
Workforce Management 0.354 0.030 .000
Blue collar −0.160 0.014 .000
Secondary education 0.023 0.003 .000
Tertiary education 0.249 0.006 .000
University education 0.676 0.010 .000
Gender 0.271 0.007 .000
Hours worked 0.371 0.005 .000
Fixed term contracts −0.285 0.009 .000
Replacement contracts −0.230 0.040 .000
Task-specific contracts −0.171 0.027 .000
Labor market Regional unemployment rate −0.201 0.207 .332
Regional voluntary turnover −0.149 0.198 .451
Location quotient 0.033 0.001 .000
#Firms in region 0.011 0.105 .915
Regional-level firm size −0.231 0.112 .040
Regional-level firm age 0.056 0.121 .640
Regional-level asset tangibility −0.198 0.495 .689
Regional-level capital intensity −0.012 0.061 .840
Market munificence −0.004 0.014 .794
Market uncertainty −0.003 0.005 .521
#Firms in industry −0.047 0.012 .000
Border region dummies Yes
Fixed effects Industry fixed effects Yes
Region-year fixed effects Yes
Constant 4.258 1.149 .000
AdjustedR-squared .650
p .000
Note:N=74,455. Coef.=Coefficient.SE=standard error. Independent variables were winsorized at the 0.5th and 99.5th percentiles to mitigate the impact of outliers; results were robust to not winsorizing.
7In practice, and as we will discuss, there are likely some firms offering negative rents, so that the sloped lines for absolute rents would lie between the solid and dashed line.
F I G U R E 1 Histogram of mean-centered workforce rents.
F I G U R E 2 Scatter of mean-centered workforce rents and selected financial measures (revenues, gross added value, EBITDA, net accounting profits). The solid sloped lines are visual benchmarks for mean-centered workforce rents. The dashed sloped lines are visual benchmarks for workforce rents shifted upward so that no workforce rents are negative.
In the net profits scatter, more than 34% of our observations lie above the 0.5 slope line, meaning that more than 34% of firms in our sample reportat least 5 euro in mean-centered workforce rents for every 10 euro in net profits. We conclude from these figures that variance in workforce rents seems substantively important for this population of firms. We emphasize, again, that this represents some workforces getting paid substantially more than market appro- priate payments based on workforce and market characteristics, not just that some workforces get paid more than others.
6 | A D D I T I O N A L A N A L Y S I S
While we believe that our empirical illustrations above suggest that workforce rents are sub- stantive and worthy of additional study in strategy research, it is useful to explore in more detail what our results mean and the extent to which we can be confident that we have actually indi- cated workforce rents.
6.1 | Interpreting the magnitude of workforce rents in absolute terms We have been careful in our discussion so far to focus on the variance in workforce rents with our mean-centered estimate, and we believe that simply documenting the substantive between- firm variance in these rents is important for management theory. At the same time, however, it may be informative to explore the magnitude of these rents in absolute terms. We do so in two ways. First, we explore what the mean workforce rents value would be under different assump- tions about the percentage of firms that offer“negative”rents to the workforce. Theoretically, it is unclear what negative rents might mean. If a firm offered an employee total utility that was lower than the employee's opportunity cost then a rational employee will leave. Thus, one very extreme assumption might be that no firm in our sample offers negative rents—that is, we would shift all mean centered rents observations upward by the same amount so that there are no“negative rents”in the sample. When we apply this logic, mean workforce rents increases to 19,507 EUR per FTE. Since the assumption that no individual employee receives negative rents is unlikely in practice, however, this value may be unreasonably inflated. Therefore, we relax this extreme assumption to allow different percentages of firms to offer negative rents (in 1%
increments) and show in Figure3 how our mean workforce rents estimate changes based on the assumption. For instance, when we assume that workforce rents are negative in as many as 25% of firms in our sample on average across years (a percentage that we suspect is much higher than what likely occurs in practice), the mean of our workforce rents proxy drops to 3174 EUR per FTE. While this value is only 16% of our extreme estimate, it still represents about 43% of average net profits for firms in the sample.
Second, we leverage basic information about the national minimum wage in Belgium to try to estimate a conservative mean workforce rents value. Using data on historical legal minimum wages in Belgium from Eurostat,8we calculate the absolute legal minimum total labor cost for any given firm-year in our sample as follows: FTE×minimum wage. If our predicted market appropriate wage is accurate, then it should be higher than this calculated minimum wage in all cases (note that our estimate includes other nonwage labor expenses). In our case, 11.7% of
8The national minimum wage in Belgium increased from 15,715 EUR per FTE in 2008 to 18,201 EUR per FTE in 2016 Eurostat.
our estimated market appropriate wages are lower than this minimum wage estimate. We take these observations and calculate the distance between our estimated market appropriate pay- ment and the minimum wage-based payment, then average these together. This average, then, is the average amount we would need to shift our market appropriate payments estimate so that the new estimated market appropriate payments is above the minimum wage requirement for all firms. When we do this, mean workforce rents becomes 3013 EUR per FTE. Thus, while we cannot offer a precise estimate of mean workforce rents in our population, it seems reasonable that even conservative approximations are large for these firms.
6.2 | Exploring alternative methodological choices
We explain above why we have made our methodological choices given the data constraints we have and the research question we explore, but we recognize that these empirical choices may have a substantial impact on our results and our interpretations (King et al.,2021). In Table4, we document our key research choices (and alternatives) to illustrate various possible forking paths in our research. While it is not practically feasible to report a complete analysis of all potential paths (with 24 market-appropriate payment predictors in our main model we obtain over 16.7 million paths [2^24] when just considering the inclusion/exclusion of each payment predictor), we created 12 clusters of choices and ran our model for each potential choice set (10,208 potential paths in total). Thereby, we also considered several empirical choices that are further discussed below. For instance, we considered how our empirical estimates are influenced by the inclusion of different clusters of firm-specific factors (choices 6–9 in Table 4) or by the inclusion of labor productivity (choice 10) as additional predictors for market-appropriate payments in our models (see sections below for more details).
We report the expansive set of results from this analysis in an online interactive chart9 where the x-axis indicates different model configurations. Here, the reader may see in detail how different choices affect our estimate of the standard deviation in rents relative to the result from our main model (depicted as a dashed horizontal line for reference). Overall, we find that
F I G U R E 3 Workforce rents as a function of percentage of firms that offer negative rents.
9The url for the online interactive chart is:https://bireporting.rutgers.edu/views/EpistemicMap/WorkforceRents.
workforce rents are substantively important across all models. The lowest standard deviation we find is 5654 EUR/FTE, or 88% of the value reported in our main analysis above.
6.3 | Attributing unexplained variance to workforce rents
We hope it is clear that our proxy for rents is obtained from the residuals of a regression of actual payments on factors that theoretically influence market-appropriate workforce
T A B L E 4 Assumption space for epistemic mapping.
Choice cluster
Choices in main model predicting market-appropriate workforce
paymentsa Options for epistemic mapping
Market-appropriate payment predictorsb 1. Workforce
characteristics
Included Include/exclude
2. Labor-market characteristics— industry level
Included Include/exclude
3. Labor-market characteristics— geographical level
Included Include/exclude
Fixed effectsb 4. Industry fixed effects
Included Include/exclude
5. Region and year fixed effects
Region-year fixed effects None/region fixed effects/year fixed effects/region and year fixed effects/
region-year fixed effects Additional market-appropriate payment predictorsc
6. Owner philosophy Excluded Include/exclude
7. Firm-specific stakeholder strategies
Excluded Include/exclude
8. Firm-specific investments
Excluded Include/exclude
9. Workforce bargaining power
Excluded Include/exclude
10 Labor productivity Excluded Include/exclude
Other specifications 11. Sample, industry selection
All industries Including/excluding highly regulated industries
12. Dependent variable workforce payments
Logged Logged/unlogged
aCfr. Table3.
bDetailed variable descriptions are provided in Table1.
cDetailed variable descriptions are provided in Online Appendix2.
payments. If we have a sufficiently convincing measure of actual workforce payments and include a sufficiently robust set of workforce and market characteristics then it seems reason- able to attribute this residual to workforce rents because the core concept of interest is precisely the difference between theactual paymentthe firm makes to the workforce and the theoretical market appropriate paymentfor that workforce in that labor market. In this section, we explore the extent to which attributing the residual to rents seems appropriate.
6.3.1 | Potential errors in predicted payments
Our main empirical approach uses the point estimate for predicted market-appropriate pay- ments (the residual is the difference between the actual workforce payment and the point esti- mate of the market-appropriate payment). This is potentially concerning since each predicted payment has a confidence interval and the “real” market-appropriate payment likely falls within that interval. To explore the sensitivity of our findings to this observation we constructed conservative market appropriate payments estimates for each firm using the end of the 95%
confidence interval that would minimize the workforce rents estimate. This approach makes the gap between the actual payment and the market-appropriate payment, that is, the work- force rent proxy, as small as possible. The standard deviation of this extreme form version of workforce rents is 4742 EUR per FTE, which amounts to 74% of the standard deviation of the original measure.
6.3.2 | Unobserved workforce characteristics
The reader may note that while we have controlled for many human capital measures at the firm level, there are a host of potential workforce characteristics that we could not measure and therefore could not include in our analysis. The concern here is that what we are calling a workforce rent might actually be the variance in workforce payments that should be attributed to some element of workforce quality that our empirical proxies are not able to capture. We explore this possibility in three ways.
First, it is important to understand the variance in unobserved workforce value creation potential that would be required to ameliorate the workforce rents proxy we have created. The 1st to 99th percentile range of workforce rents we observe after accounting for all observable workforce human capital characteristics is about 32,569 EUR per FTE per year, and recall that the average workforce payment in the overall sample is 24,860 EUR per FTE per year. This means that the observed range in workforce rents is larger than the average total workforce payments in our sample. Accordingly, after accounting for all observable workforce characteris- tics in our data, there would have to remain some unobserved element of workforce quality that explains differences between observationally equivalent firms that are of a larger magnitude than the base payment these firms make to their workforces. This seems a heavy burden to place upon any remaining unobserved workforce characteristics.
Second, we sought to rule out an unobserved workforce quality concern more directly by including labor productivity as an additional factor in our models. We did not include labor pro- ductivity in our main analyses because market wages are theoretically only determined by fac- tors that directly capture a workforce's outside options. In contrast, labor productivity may be influenced by firm-level factors that do not directly relate to the workforce's profit creation