The where and the who of HRM decision-making: HRM decentralization and devolution

Abstract

Purpose

HRM decentralization and devolution have been highlighted as key HRM processes in organizations’ quest for increased flexibility. Although they have been extensively studied in the MNC and International HRM literature, they have mainly been examined on a separate basis, and their definition and operationalization have often been confused. Thus, we first clarify the difference between the two concepts by refining the definitions by Hoogendoorn and Brewster (1992), and then empirically examine how they are related.

Design/methodology/approach

The relationship between HRM decentralization and devolution is examined by means of a survey in a large multi-country sample of multi-unit organizations.

Findings

Regarding our clarification objective, we contend that devolution has to do with who takes responsibilities for HRM (i.e. line managers or HRM professionals) while decentralization refers to where HRM responsibilities are allocated (i.e. headquarters or increasingly local units). Regarding the relationship between the two concepts, the results show that higher levels of HRM decentralization are related to higher levels of devolution, but this association is attenuated in organizations with more powerful HRM departments.

Originality/value

The study contributes to theory and practice by disentangling, at the conceptual, operational, empirical and practical levels, two different but related HRM decisions (how much to devolve and how much to decentralize HRM) that organizations must make to efficiently cope with the characteristics of their own structure and competitive environment. It highlights the role of the relative power of HRM departments in how HRM responsibilities are ultimately distributed across the organization.

Citation

Hassan, N., Trullen, J. and Valverde, M. (2024), "The where and the who of HRM decision-making: HRM decentralization and devolution", Employee Relations, Vol. 46 No. 9, pp. 76-96. https://doi.org/10.1108/ER-01-2023-0026

Introduction

To better adapt to their changing environments, organizations are gearing towards more flexible structures and operating modes, with increased levels of shared responsibility. Environmental disruptions such as the COVID-19 crisis or current trends in the labor market, such as remote working, stress the relevance of agility in HRM delivery (Adekoya et al., 2022). Within this context, decentralization of human resource management (HRM) policies and practices (Kostova et al., 2016), as well as devolution of HRM responsibilities to the line (Kurdi-Nakra et al., 2022), have become more relevant. For decades, studies have shown that organizations exhibit distinctive patterns of centralized control in HRM (Ferner et al., 2004) as well as different levels of devolution within firms (Larsen and Brewster, 2003). Both HRM decentralization and devolution processes have received considerable attention (Gooderham et al., 2015; Intindola et al., 2017; Lazarova et al., 2017), but they have been examined separately. There is a lack of research specifically addressing the linkages between them, even though both processes were connected early on in the literature by Hoogendoorn and Brewster (1992). They noted that the transfer of HRM responsibilities to middle managers often ran in parallel to larger processes of HRM decentralization within firms. A decade later, Larsen and Brewster (2003) called for further research on the relationship between the two, a call that thus-far has not been met.

A potential reason for this lack of attention to the decentralization-devolution relationship in HRM may be a certain confusion regarding the specific meaning of each phenomenon. Although both deal with the allocation of HRM responsibilities, the degree of HRM decentralization is a matter of where HRM takes place (e.g. at headquarters or at local units), while the degree of devolution indicates who is responsible for HRM within an organization (commonly either the HRM department or line managers). While this differentiation is embedded in the original definitions by Hoogendoorn and Brewster (1992), their conceptualization of the difference between the two has received modest attention in the ensuing three decades. To illustrate this confusion, Mesner-Andolšek and Štebe (2005) use both concepts interchangeably, talking about “devolution or decentralization” of the HRM function. Similarly, Davis and Luiz (2015) state that they study the devolution of the HRM function, but in fact operationalize it as HRM decentralization, looking at what decisions are made at the subsidiaries or at headquarters. In the field of public administration, devolution has been considered as another type of decentralization (Lodenstein and Dao, 2011) and, in turn, decentralization has been conceptualized as devolution, deconcentration or delegation (Feizy et al., 2015). More recently, Intindola et al.'s (2017) attempt to separate the two concepts by using “retention of power” to distinguish between HRM decentralization and devolution has not helped to this clarification.

Even when authors demonstrate a completely clear conceptualization and operationalization of these phenomena in their research, the use of the language itself sometimes gets in the way of distinguishing them from each other. For example, when Mayrhofer et al. (2019, p. 360) talk about “centralization of policy-making towards the HRM department” when discussing decreasing devolution trends across organizations, or when Nachmias et al. (2022, p. 295) refer to the “importance of decentralizing HR practices to the line” (our italics).

The first objective of this paper is thus to clarify the difference between HRM decentralization and devolution, and accordingly develop definitions for each term. Being separate phenomena, HRM decentralization and devolution may not inherently occur together. As firms decentralize HRM from the main headquarters to regional headquarters to subsidiaries or even smaller business units (a where choice), they also need to decide who will be taking over that decentralized HRM function, typically either their local HRM department (in which case, devolution would not ensue) or the line managers in the subsidiaries (in which case, both decentralization and devolution would run in parallel).

The second objective of this paper is to study the relationship between both constructs. While Hoogendoorn and Brewster (1992) already provided some descriptive data on different combinations of HRM decentralization and devolution, they did not carry out an empirical test of their association. Thus, questions remain as to whether and under what circumstances HRM decentralization is more or less likely to increase devolution. The exploration of the possible relationship between HRM decentralization and devolution is based on conceptual arguments about the connection between organizational design choices and the management of uncertainty (Donaldson, 2001).

We target a specific circumstance that may intervene in the relationship between HRM decentralization and devolution, namely the importance of the HRM department’s power (Hickson et al., 1971) as a possible boundary condition. We focus on this aspect because the allocation of decision-making rights fundamentally has a political nature (Nguyen et al., 2019), and thus it carries weight beyond any strictly efficiency reasons why organizations set their precise degree of HRM decentralization and devolution. This is especially so in any devolution decision, for which the HRM department is inherently part and parcel: On one hand, HR managers, in collaboration with top management, may be the agents establishing who will be charged with various HRM responsibilities. Therefore, HR managers might be decision-makers in determining the degree of devolution within the organization. On the other hand, the HRM department is also directly affected by the specific degree of devolution established in the organization. Intrinsically, then, the HRM department has vested interests in the determination of the degree of devolution in a company.

Therefore, this study contributes to the HRM literature by clarifying the conceptual and operational difference between HRM decentralization and devolution, and by exploring their relationship in multi-unit companies, while taking in consideration the role that HRM department power may hold in that relationship.

Defining and distinguishing between HRM decentralization and devolution

Understanding the difference between HRM decentralization and devolution is the first standing goal of this paper. In this section, we first consider each phenomenon separately and then we provide their corresponding definitions.

(De)centralization

The study of concepts and practices of centralization and decentralization of HRM has generally been embedded in the literature on MNCs (Belizón et al., 2016; Ferner et al., 2004; Lee et al., 2022; Patel et al., 2019; Smale et al., 2013) and, more recently, in the shared HRM services research (Maatman and Meijerink, 2017). Its origin lies in the literature on organizational design, with a focus on the structure of business units (Mintzberg, 1979; Zeng et al., 2018). Hence, it is concerned with where decisions and policies are made.

Centralization in general is understood as one way of integrating business functions by concentrating the decision-making authority in the business head office, because arguably has a more complete understanding of the activities commonly performed at their local units (Kim et al., 2003). One of the main purposes of centralization is the control of local units through organizational hierarchy with the aim of establishing a clear alignment with corporate strategy (Smale et al., 2013). In contrast, decentralization spreads decision-making power throughout the organizations’ units and to lower hierarchical levels (Mills et al., 1990), giving up control in exchange for higher levels of adaptation and speed in decision-making (Cray, 1984).

In the context of HRM, some of the benefits of centralization include economies of scope and scale, consistent HRM service delivery and the preservation of a global identity and internal equity, efficiency, strategic alignment, and best practice sharing. In turn, benefits of HRM decentralization would involve increased flexibility, speedy decision-making and responsiveness to local needs (Caligiuri et al., 2010; Meijerink et al., 2013). Patterns of HRM centralization and decentralization are linked to home and host countries’ industrial relations systems, employment regulation and cultural values (Edwards et al., 2013; Farndale and Paauwe, 2007; Oseghale et al., 2023).

Devolution

The idea of devolving certain HRM responsibilities to line managers as one of the central tenets of strategic HRM has received plenty of attention in the HRM literature since early on (Beer et al., 1984; Torrington and Hall, 1996). Subsequently, interest in devolution increased considerably with the downsizing, restructuring and privatization trends of the nineties, which created the need for more flexibility and a greater capacity to adapt (Intindola et al., 2017), a trend that continues nowadays in the post-COVID-19 era (Adekoya et al., 2022).

All definitions of devolution are based around the central idea of transferring HRM responsibilities from HRM specialists or department to line managers. The literature on devolution is based on the assumption that all managers play a central role in shaping HRM processes (Kehoe and Han, 2020). At first glance this seems an intuitively simple concept to grasp, as it is easy to envisage how managers may act as interpreters of the messages embedded in HRM policies (Nachmias et al., 2022). In this sense, line managers are seen as among the key actors charged with the implementation of HRM policies (Evans, 2015; Guest and Bos-Nehles, 2013; Paauwe, 2009; Townsend and Dundon, 2015) in various functional areas, such as training, selection or performance appraisal. In fact, an important proportion of the devolution literature mainly associates the HRM role of line managers as “implementers of human resources (HR) policy and practice” (Townsend et al., 2022a), with less attention specifically afforded to their role in other levels of HRM responsibilities, such as HRM policy decision-making (Gooderham et al., 2015; Suhail and Steen, 2021). Yet, more recent contributions acknowledge a more agentic role for line managers (Townsend et al., 2022b) whereby both top-down and bottom-up approaches to HRM delivery by line managers are considered (Guest, 2021; Kehoe and Han, 2020).

In this broader and more appreciative view of line manager’s roles in HRM (Bos-Nehles et al., 2017; Kim and Kehoe, 2022), the type and level of responsibilities transferred to line managers can be more diverse. Two studies attempt to clarify this matter by proposing an increasing continuum in the level or depth of responsibility passed on to line managers: First, Cascón-Pereira and Valverde (2014) empirically differentiated between simply devolving the execution of HRM tasks, decision-making power, budget or knowledge transfer. Second, Kehoe and Han (2020) described three line managers’ HRM delivery behaviors, namely implementation (akin to the earlier model’s execution), translation (somehow similar to decision-making power) and adaptation/introduction, in which line managers take over the whole responsibility for a HRM practice from the outset.

Although HRM responsibilities in the hands of line managers at any of these levels clearly belongs to the concept of devolution, the distinction among them is essential for examining the relationship between HRM (de)centralization and devolution. Indeed, to pursue the objective of linking where HRM policy decisions are being made with who is making those decisions, for the purposes of this study the focus of devolution centers on HRM responsibilities precisely at the policy decision-making level.

Definitions

The revision of the concepts and purposes of HRM decentralization and devolution shows that, while both pursue similar goals in terms of HRM efficiency and agility, both represent separate phenomena, of which the HRM literature has availed a conceptual distinction for a long time. Indeed, Hoogendoorn and Brewster (1992) defined HRM decentralization as “the allocation out to more local parts of the organizations of tasks formerly undertaken centrally” and devolution as “the allocation of tasks formerly undertaken by the personnel specialists to line managers” (p. 4). Although later contributions have adopted different interpretations (e.g. Intindola et al., 2017), we concur with Hoogendoorn and Brewster’s distinction. Accordingly, we propose to revive it, highlighting their original conceptualization, and positing that it can be further refined by reformulating two aspects:

First, the word “tasks” can imply that what is being decentralized or devolved is only the execution of HRM. Thus, we propose to substitute it with “responsibilities” to conceptually incorporate a broader range and level of actions, including any actions and decisions about HRM strategies, policies and practices. Second, the expression “formerly undertaken” implies that a change has occurred, effectively rendering both concepts ending states of a change process. However, in practice, organizations’ people management may have been devolved and/or decentralized recently, for quite some time, or since their very foundation. Thus, we propose to remove the allusion to a former/later state, as organizations’ HRM might be centralized or de-centralized, devolved or not devolved (and any degrees in between) without necessarily having reorganized their HRM architecture in the past.

From these considerations, it follows that HRM decentralization identifies where HRM decisions are made. Therefore, we define it as the degree of allocation of HRM responsibilities towards more local units rather than towards more central parts of the organization’s structure, while HRM centralization would be the allocation of HRM responsibilities towards more central parts of the organization. For example, a centralized organization might concentrate a substantial amount of HRM responsibilities in its international or national headquarters, rather than in their regional dependencies or local units, while in a decentralized one, these responsibilities will be carried out differently in each unit. Or even in non-geographically dispersed organizations, HRM might tend to have more weight at the center of the organization (for example, in a university, centralized at the presidency or rectorship), or in its component parts (for example, decentralized to its schools, faculties or departments). Thus, as the names suggest, HRM centralization and decentralization are two end points on a continuum of the organization’s location where HRM responsibilities mainly take place.

In turn, devolution informs us about who takes responsibility for HRM in a continuum between the specialized HRM department and managers from other functional areas. Therefore, we define it as the degree of allocation of HRM responsibilities to line managers. The term line manager, which may include positions as distant as team leaders to functional heads and senior executives, as well as so many others in between, such as front-line and first-line managers, department or unit managers, etc., is broad and even “blurred” (Townsend and Dundon, 2015). However, it is used here purposefully, in line with the objective of making these definitions widely encompassing. To illustrate, different companies with the same level of decentralization may have different degrees of devolution: following the above example from the education sector, in two similarly HRM centralized universities, much HRM decision-making may reside in a specialized HRM department in one organization, while in another these decisions may be in the hands of non-HRM specialists, like academic figures such as vice-rectors or vice-presidents.

Overall, these fine-tuned definitions owe the main distinction between HRM decentralization and devolution to the work of Hoogendoorn and Brewster (1992) and aim to provide a general framework in which to locate and specify the exact level of phenomena analyzed by empirical studies. Since both HRM decentralization and devolution may include HRM responsibilities at different levels, it is important to understand the realities of HRM from a multilevel perspective. To this end, the HRM literature conceptually distinguishes at least three hierarchically ordered components in HRM systems, namely, HR philosophies, policies and practices (Renkema et al., 2017). Similarly, Valverde et al. (2006) empirically distinguish between various levels of HRM decision-making including strategic, policy, operational and administrative. A specific level of HRM decentralization and devolution will occur for all these different levels. Some HRM system levels are more likely to be associated with specific degrees of decentralization and devolution: For example, it may be more plausible to set the corporate people management philosophy and strategic HRM decisions at a company’s headquarters rather than in its local units (centralization). Also, if devolved, such strategic matters might be more likely decided by executive managers than middle or front-line managers. On the contrary, day to day people management practices are more likely to be decentralized to local units and devolved to front-line managers. Nevertheless, the gamut of HRM responsibilities is wide and the layers and levels of line managers between the strategic apex of the organization and its operating core numerous (Floyd and Wooldridge, 1994). Thus, the objective behind our broad-ranging definitions is to provide a background against which managers can be aware of the multiple facets of these phenomena and researchers be prompted to specify the level of HRM responsibilities considered when studying the continuums of their location (between centralization – decentralization) and actors involved (in this case, between specialist HRM department – different levels of line managers).

Specifying the empirical objective

Availing of these definitions and related explanations, we need to choose a specific level of analysis to approach the second objective of this paper, which is to empirically test the relationship between devolution and decentralization. Accordingly, we decided to focus on responsibilities at the level of HRM policy decision-making as opposed to HRM strategy, or as opposed to execution of operational HRM tasks, which have already received more attention in the literature. By looking at the policy decision-making level, it will be possible to observe where HRM policy decisions are being made and who is making these decisions.

In particular, the study focuses on five different HRM area domains: pay and benefits, recruitment and selection, training and development, industrial relations, and workforce expansion/reduction. To offer an example, if a firm devolves and decentralizes policy level decisions about recruitment and selection, it means that line managers at the local business units or subsidiaries are charged with decisions such as whether and when new recruiting is needed, how many recruits, whether to acquire personnel from internal or external labor markets, or what recruitment methods will be used. To further illustrate, an organization that centralizes and does not devolve policy level decisions on pay and benefits would be one where the design of the compensation policy (e.g. levels of compensation for different job positions, types of employees subject to pay for performance or types of benefits to be offered) is in the hands of the HRM department at the company’s headquarters.

The relationship between HRM decentralization and devolution

Despite their different focus, both decentralization and devolution are concepts linked to choices about organizational structure and design (Mintzberg, 1979) in terms of to whom roles and decision-making rights are allocated (Cunningham and Hyman, 1995) and in terms of structural dimensions in the case of the degree of (de)centralization (Ghoshal and Nohria, 1989). Hence, any attempt at understanding how both processes are linked needs to consider arguments of organizational design.

Early organization theories linked the existence of different structural configurations to levels of task uncertainty and interdependence (Donaldson, 2001). A particular influential approach was that of Burns and Stalker (1961), who argued that decentralized structures were an important mechanism available to firms to confront uncertainty, often resulting from market or technological changes. According to them, when uncertainty is low, it is easier to routinize operations in a rule-like fashion specified from the top, with more specialized and defined roles, something typical of mechanistic structures. On the contrary, when uncertainty is high, “operations need to rely on the initiative of employees who view their role more broadly” (Donaldson, 2015, p. 609) in more organic structures. Thus, to flexibly cope with uncertainty, companies can respond by moving their HRM decision-making from the headquarters to the business units (where), and from more specialized (HR) managers to other managers (who). The international HRM literature follows a similar logic, arguing that “more complex subsidiary environments tend to be associated with less centralization (Smale et al., 2012, 2013; Lazarova et al., 2017, p. 86), as HRM decentralization is linked to increased flexibility.

Apart from coping with uncertainty, companies may also need to adapt to the local contexts of their subsidiaries when these significantly differ from that of headquarters. By decentralizing, they can be more adaptive to local context and reduce the cultural distance and heterogeneous institutional environments between different units (Birkinshaw and Pedersen, 2009; Lazarova et al., 2017). Decentralized firms are often also characterized by having lower levels of task specialization and lower levels of formalization to facilitate the necessary coordination and flexibility between units (Burns and Stalker, 1961). With lower formalization there will be less written HRM rules and procedures (Donaldson, 2001), and thus less need for a specialist HRM department and a larger role for line managers. In this context, and because line managers are in day-to-day contact with employees and have access to specific information and resources, they will be in a better position to respond with more appropriate and faster HRM decisions (McCracken et al., 2017).

To sum up, companies that need to cope with uncertainty and to adapt to different local environments may simultaneously want to decentralize HRM from headquarters to business units and to foster a reliance on the people who are in touch with daily local operations, that is, to devolve HRM to line managers. On the contrary, when HRM decisions are centralized, firms will have less of a need to rely on local managers. Drawing on these arguments, we can expect that:

H1.

Higher levels of HRM decentralization will be associated with higher levels of devolution to line managers.

Power of the HRM department

Up to this point, we have argued that increasing HRM decentralization is likely to increase the tendency to rely on line managers to make HRM decisions. However, a decentralized HRM operation could also be enacted by the HRM department itself by, for example, replicating HR structures and roles locally (e.g. local HRM professionals and HRM business partners) (Larsen and Brewster, 2003). We argue that whether HRM decentralization is indeed associated with devolution will not only depend on arguments around organizational design and the management of uncertainty, but also on the power of the HRM department vis-à-vis other units in the organization. HRM department power has traditionally been associated with the HRM department’s strategic position in the organization (e.g. Gooderham et al., 2015). According to Reichel and Lazarova (2013, p. 924), “a strategically positioned HRM department is endowed with higher status and more power compared to an HRM department that holds a purely administrative position.” Strategic position is often translated into two key components, namely, representation of the HRM department on the top management team and its involvement in formulating the corporate strategy (Brewster et al., 2015; Gooderham et al., 2015).

Within this context, it is important to observe the political mechanisms at work in the consideration of how a HRM department’s degree of power might intervene in the relationship between HRM decentralization and devolution. Particularly at the level of HRM policy decision-making, which is the focus of this study, a transfer of HRM responsibility from the HRM department to line managers may hold both promise and threat (Reichel and Lazarova, 2013). Devolution will offer promise if the HRM department uses the release of policy and administrative level tasks in the hands of the line (Sheehan, 2005) to focus on strategic issues, which, in turn, may make them more powerful. But it can also pose a threat if devolution ends up decreasing the overall decision-making responsibilities bestowed upon HRM departments, rendering them redundant and insignificant (Hall and Torrington, 1998). Current political perspectives may provide new insights on the matter by looking at the agency of HRM departments. For example, by observing how involving line managers in HRM and other symbolic actions can in fact increase the HRM department’s internal status and augment their influence in corporate strategy (Hermans and Ulrich, 2021).

Moreover, whether line managers or HRM professionals are deemed the most appropriate to make HRM decisions is likely to depend on the extent to which HRM issues are perceived to be complex or specialized, and hence more or less likely to be allocated to “experts” (Hickson et al., 1971), which has implications in terms of intraorganizational power. In this sense, it is well known that HRM departments often face a problem with the perceived substitutability of their knowledge or expertise (Reichel and Lazarova, 2013), as people management tasks are not always perceived as a true specialist activity (Galang and Ferris, 1997). This diminishes the credibility of the HRM department (Stirpe et al., 2013) and may compromise its ability to exert influence (Sheehan et al., 2014; Trullen and Valverde, 2017). If devolution processes are successful, perceptions of substitutability are likely to increase even more. On the contrary, the HRM department will be more powerful if line managers have the perception that they do not have enough HRM knowledge and information and consequently only HRM specialists can deliver HRM effectively. It follows that HRM departments, especially when they hold internal power, will be inclined to retain their role in HRM decision-making even if decentralized in local units.

Powerful departments are also likely to be more convinced of their internal worth and expertise and may find it more natural to trust their own HRM professionals on the ground, rather than passing HRM decisions onto the line. Finally, over the years, HRM personnel have been the legal, moral and traditional guardians of the HRM function (McCracken et al., 2017), making them more reluctant to allow the line to take on responsibilities that it considers legitimately theirs (Perry and Kulik, 2008) if they are powerful to do so.

Hence, the relative strength of the HRM department within the organization is likely to act as a boundary condition, attenuating the relationship between HRM decentralization and devolution. Thus, we argue that:

H2.

HRM department power will negatively moderate the relationship between HRM decentralization and devolution, so that this association will be weaker for high levels of HRM department power.

Methodology

Sample

Data for the study originate from the Cranet survey, an international research network that collects information about HRM policies and practices from organizations with more than 200 employees in different countries (Parry et al., 2021). The Cranet survey poses factual responses, that is, information about figures and neutral practices rather than opinions, which are less likely to be subjected to socially desirable responses (Chang et al., 2010). Moreover, although it collects data from a single source in each organization, it takes a key informant approach, “whereby the respondent to our questions about HRM policies is the highest-ranking HRM professional in the organization or their representative” (Parry et al., 2021, p. 276), thus obtaining the data from the most knowledgeable source possible.

In selecting the sample for this study, we excluded companies without an HRM department, since, without one, we would not be able to measure the degree of devolution. We also excluded organizations with only one site, as our focus was on multi-unit organizations to be able to take into account the degree of decentralization of HRM decisions between the headquarters and increasingly local units. This resulted in a dataset including 4,652 organizations in 35 countries.

Measures

It is important to highlight that the measures of the main phenomena of interest of this study, namely HRM decentralization and devolution, focus specifically on the level of decision-making for HRM policies rather than HRM responsibilities at other levels, such as execution of HRM tasks.

Devolution. The degree of devolution was captured by a comprehensive measure asking who bears the primary responsibility for major policy decisions concerning five areas of HRM activities: pay and benefits, recruitment and selection, training and development, industrial relations, and workforce expansion/reduction. To capture a value for the degree of devolution of HRM policy level decisions specifically, the responses were coded as 1 for “only the HRM department,” 2 for “HRM department in consultation with line management,” 3 for “line management in consultation with HRM department” and 4 for “only line management.” The answers for the five areas of HRM activities are summed to create an overall index, with larger values indicating a higher degree of devolution. This methodology is in line with Gooderham et al. (2015) and Reichel and Lazarova (2013).

HRM decentralization was measured, consistent with Lazarova et al. (2017), by a comprehensive measure asking where policies on different HRM areas (pay and benefits, recruitment, etc.) are mainly determined: (1) at international and national HQs, (2) at subsidiary, department or division and (3) at site, establishment or local offices. Again, in this case, answers for the different areas are added up for a final score, with larger values indicating a higher degree of HRM decentralization.

HR power. Based on the well-established operationalization of HRM department power to be anchored on measures of their strategic standing (Reichel and Lazarova, 2013), and following Gooderham et al. (2015) and Dany et al. (2008), HRM department power was assessed through three measures: whether the person with responsibility for HRM issues had a seat on the Board (1 = yes, 0 = no); whether the person responsible for HRM was involved in the development of the business strategy from the outset (1 = yes, 0 = no); and whether the performance of the personnel/human resources function/department was evaluated (yes = 1, no = 0). The combined measure ranged from 0 to 3.

Control variables. We controlled for organizational size -measured as the log number of employees (Mesner-Andolšek and Štebe, 2005), as there is widespread acceptance that it may influence how HRM is organized (Brewster et al., 2015). In particular, Johnston and Menguc (2007) argue that larger organizations need more expert and experienced personnel for HRM. We also controlled for the relative size of the HRM department, as more HRM personnel may imply more elaborate HRM policies and less need for devolution (Reichel and Lazarova, 2013). This was calculated as the proportion between the number of employees in the HRM department and the total number of employees, to which the logarithm was applied before running the analysis, following Lazarova et al. (2017). Third, we controlled for industry using a dichotomous measure, as differences between the service (1) and manufacturing (0) sector may affect the levels of devolution (Gooderham et al., 2015). We also considered whether the organization was a multinational (1) or national (0) firm, and whether it managed foreign (1) or mainly domestic (0) operations. Finally, organization age was measured as the log number of years in operation.

Results

Data for this study are hierarchically structured, since each organization is nested under its corresponding country. Thus, we used multilevel modeling to estimate the hypotheses with SPSS software (v27) (Peterson et al., 2012). To properly detect and estimate the slope and intercept heterogeneity and to facilitate the interpretation, variables at level 1 (i.e. organizational) were group mean centered (Enders and Tofighi, 2007).

The predictor and outcome variables were gathered from a unique source, which may raise concerns regarding potential common method bias (Podsakoff et al., 2003). However, several characteristics of the data collection process suggest that this may not be the case: As mentioned above, the informants were the most knowledgeable source about the topic of the study, and the questions dealt with factual topics rather than opinions. Additionally, our dependent and independent variables are measured in remotely separate sections of this very extensive survey (I and IV), making bias less likely (Podsakoff et al., 2003). In any case, as a precautionary measure, we examined the Harman’s single factor test and the results of the first unrotated factor of all measurement variables was 13.78%, well below 50%, supporting the argument that common method bias is not a concern in our study.

The correlation matrix is reported in Table 1. Multicollinearity among the independent and control variables was not an issue, with variance inflation factors (VIF) less than 2, well below the conventional maximum threshold value of 10 (Cohen et al., 2002).

The results of our multilevel analysis are presented in Table 2. We started by estimating the null model (with only the intercepts, no predictors). In this model, the factor variance (σ2) = 2.18 indicates how much devolution varies between countries and the residual variance (σϵ)2 = 10.05, indicating how this dependent variable fluctuates within each country (Peterson et al., 2012). The intraclass correlation coefficient (ICC) was 0.178 (p < 0.00), meaning that 17.8% of the total variance in devolution is due to differences between countries, hence the multilevel analysis is justified (Aitkin and Longford, 1986).

Next, model 1 includes control variables at level 1 (random intercept model): Organization size and HRM department size were significantly and negatively related to devolution, that is, the larger the company and the larger size of the HRM department, the less devolution; whereas MNC was positively related, that is, multinational companies tend to devolve more their HRM decisions to line managers than national companies. These effects generally remained constant throughout the analysis. Considering the Maximum Likelihood Estimation (-2LL), this model provided some improvement on model fit compared to the null model.

We also used Bayesian (BIC) and Akaike information criterion (AIC) goodness-of-fit measures in the analysis. Both measures decreased with the addition of variables in every model, indicating an increasingly better fit. To assess the relationship between HRM decentralization and devolution, it was necessary to obtain a regression equation for each country and analyze how the intersections of these equations vary. From this perspective, model 2 introduced our main predictor (HRM decentralization) with the control variables. The regression coefficient indicated a positive and significant predictive relationship between both variables. Further, according to the reduction in the -2LL value, there was a notable enhancement in the model fit when compared to Model 1. This improvement indicates that this model better captures the variability in the outcome at both the organizational and country levels. Hence, Hypothesis 1 was supported.

Hypothesis 2 predicted that HRM department power would moderate the relationship between HRM decentralization and devolution. Model 4 shows the results once the interaction term (HRM decentralization x HRM department power) was included (model 4). The effect is negative and significant, thus supporting Hypothesis 2. Fit measures in model 4 also provided some improvement compared with previous models. This result shows that indeed the positive relationship between HRM decentralization and devolution is attenuated in organizations with more powerful HRM departments, denoting that transferring HRM policy decision-making to the line may be perceived as a threat to the HRM department internal status.

In sum, we find that increasing levels of HRM decentralization are positively associated with increasing levels of devolution, but that this association is weaker in organizations where HRM departments hold more power.

Discussion

The contribution of this paper lies in the accomplishment of two connected goals: distinguishing between HRM decentralization and devolution, and investigating how these are related to each other. Regarding our clarification objective, and anchored upon the initial conceptualization by Hoogendoorn and Brewster (1992), we refine and provide full definitions of the two distinct phenomena: devolution has to do with who takes responsibility for HRM (i.e. line managers or HRM professionals), while HRM decentralization refers to where HRM responsibility is allocated (i.e. headquarters or increasingly local units). These fine-tuned definitions afford better clarity and comparability for future studies on these topics: On one hand, they help differentiating the two concepts. On the other, they encompass different levels of HRM activity, from the basic execution of administrative or operational level tasks to the political or even strategic HRM responsibilities.

To achieve our second objective, based on conceptual arguments about the relationship between organizational design choices and the management of uncertainty (Donaldson, 2001), we argued that organizations that decentralize HRM policy decision-making will also tend to engage line managers in these decisions (i.e. greater devolution) rather than keeping them in the hands of local HRM departments. The results of our robust multilevel analysis in a sample of 4,652 organizations in 35 countries show indeed a positive association between both. In this way, when local units have more autonomy to determine HRM policies, the responsibility for HRM related decisions tends to veer towards the hands of line managers. Conversely, when HRM policy decisions are taken at the headquarters, it is more likely that such decisions stay in the HRM department. While organizational design arguments (Burns and Stalker, 1961) support this positive association between HRM decentralization and devolution, they need to be complemented with power related explanations about the role of the HRM department within the organization (Sheehan et al., 2014). Our research also aimed to provide insights into how such power plays in the final shaping of the HRM decentralization-devolution relationship, acting as a boundary condition. The results show that HRM department’s power moderates the positive HRM decentralization-devolution association by attenuating it. This means that decentralized organizations with strong HRM departments retain some of their HRM policy decision-making in the hands of HRM specialists, while in decentralized companies where the HRM department holds little power, it is going to be line managers who take the lead in deciding HRM policies.

Implications for research

For researchers dealing with the phenomena of HRM (de)centralization and devolution, this study derives some recommendations. By providing and highlighting the definitions developed here from Hoogendoorn and Brewster’s (1992), we can help researchers become more sensitive to the ways in which they use both terms and analyze either or both phenomena in their work: Anchored in these definitions, we firstly recommend to conceptually separate between where HRM responsibilities are allocated in multi-unit companies from who is taking on those responsibilities. Secondly, researchers should operationalize each concept accordingly, taking two aspects into account for their fieldwork measures: On one hand, researchers should use different points or levels in the organizational structure (e.g. headquarters, regional headquarters, subsidiaries) when measuring HRM decentralization, and different actors of the HRM function when measuring devolution, namely the HRM department, line management, or various degrees of collaboration between the two. On the other hand, researchers should spell out what kinds of HRM responsibilities are being measured in their studies (for example, policy decision-making or practice execution, design or implementation, responsibilities at strategic, policy, operational or administrative level, etc.). For example, if we talk about an organization that “decentralizes and devolves performance appraisal” without further specification, this could mean that the HRM department at headquarters decides that a performance appraisal system must be in place, the HR department at the subsidiary of each country designs their own system and line managers at each unit/office/work center execute the annual evaluation of employees. But it could also describe a company where the headquarters or even the national subsidiary are not at all involved in performance appraisal and it is up to the line managers of each unit to decide whether to have a system in place, what characteristics this system will have, collect employee data and make decisions according to the performance of their employees. Thus, a more accurate description would be needed to specify that the first organization “decentralizes the design of the performance appraisal policy to the subsidiary’s HRM department and devolves the execution of this policy to line managers”, and that the second organization “decentralizes all aspects of performance appraisal to its local units and devolves them all to their line managers”. Accordingly, a study like the present one adheres to this need for precision and anchors its objective on the study of HRM decentralization and devolution at policy making level. Embedding research studies with precision is of paramount importance to understand the realities behind the measures of HRM decentralization and devolution so that the results of studies can be compared to each other.

Furthermore, academics can use the differentiations highlighted in this study not only in conceptual and operational terms, but also when writing about these topics. In this sense, we also recommend researchers not to use the noun centralization or the verb to centralize when talking about devolution to avoid confusion. Although there should be no problem in using them, as the authors ourselves have done on occasion, we suggest substituting them with expressions such as concentrating HRM decisions in the hands of HRM specialists (when not devolving), and assigning, allocating, transferring to, passing on, or ascribing HRM responsibilities to line managers, or similar expressions when talking about devolution.

Implications for practice

This paper also has practical implications for top management teams, CEOs and HR managers that need to make decisions about the HRM architecture of their organizations (Ulrich et al., 2008). Regarding the conceptual clarification provided in this study, it is important that they be aware that at least two separate (albeit related) choices need to be made for the configuration of their HRM function and HRM decision-making processes: where (headquarters, national/regional subsidiary, or increasingly local sites) and who (specialists of the HRM department or line managers). Assigning such decision-making power for major HRM policies in an intentional fashion can provide clarity at the point of making the decisions but also in terms of assigning accountability for different people management policies. In large companies, this means that the centers of HRM decision-making could be more easily reflected in company charts and job descriptions. In turn, such clarity would also allow HRM specialists to identify and systematically respond to training and/or support needs to those in charge of HRM decisions at different levels and locations of the organization, particularly when these decisions are devolved to line managers, not being specialist but generalist HRM agents.

The assignment of HRM decision-making in different parts and to different agents of the organization also brings about several associated decisions about the size, structure and composition of specialists in the HRM department. Indeed, more centralized and less devolved organizations would need a sizeable HRM department in its headquarters for HRM policy-making and design, and possibly other HRM department units in the various sites of the company to execute and administer them, affording homogeneity throughout the company; while more decentralized and devolved companies would need a smaller structure of HRM specialists but require an important investment in the provision of well-trained line managers in local sites who can make HRM decisions and arrange for their execution.

The outcomes of second hypothesis dealing with the role of the HRM department’s power in the relationship between HRM decentralization and devolution derive a more complex picture. Indeed, the fact that decentralized companies with powerful HRM departments somehow contain the HRM role of line managers indicates a certain (right or wrong) reluctance to share their power with the line. This result is in line with other studies that have observed that powerful HRM departments do not necessarily use their prominence to distribute parts of the function to other HRM agents [ex. line managers in this study, or outsourced agencies in the case of Reichel and Lazarova (2013)], but to keep them in-house and in the specialist department. However, these results seem contrary to the rhetoric that HRM departments are willing to entrust line managers with HRM responsibilities to devote more attention to strategic matters as a key tenet of the strategic HRM literature (Beer, 1997). If line managers are charged with taking decisions about the main HRM policies, the HRM department may instead fall into a more administrative role (like Tyson and Fell’s (1986) clerk of works) or at most as internal service providers. If this was the case, unsurprisingly HRM departments would not be willing to devolve policy decision-making to retain their power. Our results can thus also help HR managers become more aware of their own biases towards devolution by inviting them to reflect upon how the specific distribution of HRM decision-making between specialists and generalists reflects the needs of that particular organization or its politics and willingness to retain power on the part of the HRM department. From a broader perspective, it begs questions about the HRM profession by and large, alongside the career aspirations of HRM specialists.

Limitations and further research

This study also has several limitations that point at potential directions for future research. A first limitation lies in our operationalization of HRM decentralization and devolution as overall indexes across five different HRM subfunctions. Although this is the same approach as in previous devolution research, some scholars assert that the degree of devolution may have different patterns depending on the HRM areas analyzed (Gooderham et al., 2015; Larsen and Brewster, 2003). Similarly, different HRM subfunctions may involve different levels of centralization (Smale et al., 2013). Future research could thus examine the same relationships distinguishing by type of HRM functional area. A second limitation and corresponding opportunity for subsequent studies would be furthering the analysis with variables at country level. Indeed, HRM architecture decisions can be influenced by a variety of institutional factors (Farndale and Paauwe, 2018; Gooderham et al., 2015). One particularly relevant institutional factor has to do with employee legislation and collective bargaining more generally (Gooderham et al., 2015), as this can certainly constrain company companies’ prerogative to centralize HRM decisions, as well as the appropriateness to devolve responsibility to line managers. Thirdly, our study holds an additional methodological limitation inasmuch Cranet data is cross-sectional, single-source and focused on the level of intended or designed HRM policies (Nishii and Wright, 2008). The fact that the survey’s respondents are the highest-ranking HR professionals means that it is more difficult to gain knowledge about how HRM responsibilities materialize at lower operational levels in the organization or indeed what occurs in distant business units in decentralized organizations. However, the fact that our research focus was at the level of major HRM policies makes this risk less of a concern. In any case, future studies could avoid an excessive HR department-centric view by including perceptions of other actors such as different types of line managers or employees (Bos-Nehles and Meijerink, 2018) at different centers of the organization. In addition, longitudinal designs would be helpful to better understand how the relationship between HRM decentralization and devolution unfolds through time (Lazarova et al., 2017).

Another general limitation of the study, particularly in terms of studying power and politics in organizations, is linked to the quantitative approach to data collection, which allows to access a large number of organizations but not to obtain in-depth fine-grained nuances of the phenomena under examination. Our view of power is based on arguments around the perceived substitutability of HR knowledge (Reichel and Lazarova, 2013), whereby HR professionals may see devolution as a zero-sum game (i.e. more HRM department power means less line managers’ or the other way round). However, there is evidence in the literature that HR professionals can also increase their power by means of collaborative and networking behaviors with other managers (Sheehan et al., 2014).

Still in relation to our quantitative approach to data collection, our measures of devolution are limited since they cannot reflect the multiple realities, configurations and levels of line managers that exist in organizations. Rather, they simply distinguish between the specialists of the HRM department and line managers in general, without specifying which line managers at what hierarchies (e.g. senior managers, middle managers, front line managers, etc.) or precisely who in the HRM department is involved, and even less the specific degree of agency that each individual line manager is willing to exert in people management issues (Townsend et al., 2022b). Additionally, even if having a sense of the general weight of each actor in HRM decision-making, the survey data cannot apprehend the quality and textures of the relationship between them, whether they are more or less collaborative, etc. Thus, further studies should complement this extensive but not in-depth data with qualitative approaches that could reflect the richness of the phenomena at hand, including the responsibility for different levels of HRM responsibility in specific HRM functional areas by distinct members of the specialized HRM department and/or by various levels of line managers.

As a topic that clearly deserves fuller attention, other research avenues should provide fruitful developments: To start with, although this study has found that devolution generally follows HRM decentralization, it would be interesting to identify whether some specific companies follow the opposite pattern, that is, centralizing HRM policy in the headquarters and devolving it to line managers (in this case, probably senior line managers at the HQs), or decentralizing HRM policy to local units and leaving it the hands of the local HRM departments without much line management involvement. Moreover, gaining an insight into the rationale of why companies consciously or unconsciously make different arrangements regarding these two decisions would be an enriching line of enquiry that should be addressed through qualitative multi-informant methodological approaches. Furthermore, in the enquiry into the role of the HRM department’s power, the results of the study must be placed along the lines of the longstanding call for further research about competing perspectives in the dynamics between the HRM department’s strategic integration and devolution to the line posed since the early studies by Torrington and Hall (1996) to the most recent by Kurdi-Nakra et al. (2022).

Ultimately, these research avenues should build knowledge to address questions about the most effective balance between HRM decentralization and devolution. Do high performing companies have similar ways of allocating the where and the who of their HRM policy decision-making? It would thus be necessary to enquire about the outcomes of different levels and combinations of the two in terms of how productive different configurations might be for different actors and for overall company performance, as well as the possible tensions that they may bring about.

While these suggestions for research receive uptake, for the time being and in line with the ongoing quest and call for opening up the HRM black box (Ostroff and Bowen, 2015), we consider that this paper expands their operationalization of the “HRM construct space” (p. 200) consisting of the what, how and why dimensions, by highlighting and clarifying two novel elements, namely, the where and the who of HRM, and the exploration of their interplay. This can provide clarity for academics in future research and practical insights for HR and general managers interested in creating more flexible and resilient organizations in times of environmental disruption.

Table 1

Mean, standard deviation (SD) and correlations

	Mean (SD)	1	2	3	4	5
1. Devolution	12.03 (3.61)	1
2. Organization size	3399 (19887)	−0.058***	1
3. HR department size	43 (395)	0.030*	0.004	1
4. Age	88 (248)	−0.038**	0.532***	−0.006	1
5. Decentralization	8.01 (3.44)	0.131***	−0.059***	0.003	−0.018	1
6. Power HR function	1.57 (0.99)	−0.251***	0.082***	0.011	0.059***	−0.122***

Note(s): *p < 0.1, **p < 0.05, ***p < 0.01 (two-tailed)

Source(s): Created by authors

Table 2

Multilevel analysis of the relationship between decentralization and devolution, and the moderator effect of HR power

	Model 0	Model 1	Model 2	Model 3	Model 4
Fixed effects	Est (SE)	Est (SE)	Est (SE)	Est (SE)	Est (SE)
Intercept	11.92 (0.26)***	13.45 (0.35)***	13.23 (0.23)***	12.84 (0.24)***	12.84 (0.40)***
Organization size		−0.654 (0.10)***	−0.567 (0.11)***	−422 (0.12)***	−0.427 (0.11)***
HR department size		−1.10 (0.16)***	−0.967 (0.19)***	−873 (0.19)***	−0.872 (0.19)***
Multinationals		0.465 (0.11)***	0.445 (0.13)***	0.387 (0.13)***	0.385 (0.13)***
Industry		0.069 (0.11)	0.012 (0.13)	0.045 (0.13)	0.045 (0.13)
Age of organization		0.001 (0.14)	0.001 (0.16)	0.001 (0.16)	0.001 (0.02)
Decentralization			0.049 (0.02)**	0.046 (0.02)**	0.047 (0.02)**
HR Power				−0.429 (0.06)***	−0.427 (0.02)***
Decentralization × HR power					−0.056 (0.02)***
Random effects
ICC	0.178	0.144	0.146	0.157	1.58
(σϵ2) (Between-firms)	10.05 (0.21)***	8.82 (0.21)***	8.88 (0.25)***	8.66 (0.25)***	8.63 (0.25)***
(σ2) (Between countries)	2.18 (0.55)***	1.50 (0.38)***	1.52 (0.41)***	1.62 (0.43)***	1.62 (0.43)***
BIC	22982.66	18118.54	13034.64	12596.11	12593.82
AIC	22969.87	18106.17	13022.95	12584.47	12582.18
−2LL	22965.87	18102.17	13018.94	12580.47	12578.18

Note(s): *p < 0.1, **p < 0.05, ***p < 0.01

BIC: Bayesian information criterion; AIC: Akaike information criterion

Source(s): Created by authors