Abstract
By analyzing content, this paper aims to map empirical quantitative research on International Human Resource Management. Our filters will be “when, where, what, and how.” “When” indicates the time span we use to analyze the evolution of International Human Resource Management, “where” refers to the influential journals chosen for publication; “what” covers the different topics dealt with in IHRM, and “how” is linked to the various methodologies and statistical techniques applied. Using the “when, where, what, and how” of empirical quantitative, International Human Resource Management studies allows us to identify how different topics have been investigated and so may lead us to suggest methodological refinements to improve the analysis and knowledge of topics in International Human Resource Management. It will allow us to detect trends and research gaps and point to the most prominent journals for publication and dissemination of results.
Keywords
Introduction
International Human Resource Management (IHRM) has become increasingly significant in HRM (Tarique and Schuler, 2008). Many reasons explain the growing interest in this field of research, but all of them are related to the impact of globalization on business, that forces many firms to deal with ever-growing global human resources problems (Sparrow et al., 2017). Subsequently, the scope of IHRM research has substantially evolved and expanded from traditional parent-country expatriates to many other international actors and research topics, covering all of them concerning with the wide range of people-related issues from multinational firms (Stahl et al., 2012; Welch and Björkman, 2015). However, the definitional boundaries of what constitutes IHRM are not clear. In this sense, in the discussion about topics included in IHRM, some authors offer a more broaden definition, considering comparative HRM and the (inter)cultural dimension of Human Resource Management (HRM) as topics studied under the umbrella of IHRM (Brewster et al., 2016; Pudelko et al., 2015). To offer a wide perspective of the field, we consider IHRM as all issues related to the management of people in an international context including human resource issues facing multinational companies (MNC) in different parts of their organizations; comparative analyses of HRM in different countries (Stahl et al., 2012) and cross-cultural human resource management (Brewster et al., 2016).
Along with the evolution and expansion of the studied topics, the evolution of a researching field also implies new analysis methods to employ (Bird et al., 2002). The fast growing and evolution of software has also favored both the sophistication in the analytical methods used and the accessibility of different data sources. This fact therefore, has expanded analytic capabilities and has broadened the spectrum of appropriate methodological options best suited to the study of different topics in order to provide new insights on them. As the application of appropriate data and statistical techniques is key to achieving a solid understanding of the research studies, methodology has become a key issue in IHRM research.
A great deal of research has been published in dedicated journals. Academicians must select which journal to send their research manuscripts, and they normally do it according to two main criteria (Caligiuri, 1999): the appropriateness (selection of a journal read by scholars who are interested in the field) and the prestige of their journal outlet. Although selecting a journal may be straightforward in some academic disciplines (e.g. sociology, psychology), it is much more complex in the case of hybrid fields, as it is IHRM—intersection between HRM and international business (IB). In 1999, Caligiuri identified which scholarly journals were the most prominent publication outlets of IHRM according to the two mentioned criteria, offering a list of the most influential journals in IHRM. However, the evolution of a research field also implies a likely evolution in the appropriateness and prestige of journal outlets and, therefore, it seems interesting to modernize Caligiuri’s list.
The intense development, growing, and evolution of research on IHRM suggest the need of revising the existing literature. The aim of this paper is to do so according to “when, where, what, and how” on this field. “When” indicates the range time on which we are focusing the analysis of the evolution of IHRM, “where” is regarded to the most influential journals to publish, “what” refers to the different topics dealt with in the focused field, and “how” is linked to the different methodologies and statistical techniques applied in empirical quantitative research. Therefore, this paper proposes to map the field of IHRM according to these four questions. To do so, we conduct a content analysis to obtain a systematic, quantitative description of the literature in the field: First, we set the time-span that ranges from 2010 to 2018 (“when”) as the one to be considered for our content analysis. Then, we replicate Caligiuri’s (1999) study to renew her journal ranking and to be able to select the three most influential journals in IHRM. Together with the German Journal of Human Resource Management (“where”), we do a systematic review of all the papers published in the four journals. Therefore, we examine each empirical quantitative paper on IHRM to determine the topics (“what”) and methodology used (“how”), aspect in which we focus in order to collect information regarding the method employed, to account for sample information, variables and measures as well as to gather the statistical techniques applied.
To our knowledge, two other recent studies have conducted a content analysis in IHRM: Welch and Björkman (2015) analyzed papers on IHRM focusing on thematic analysis of articles and covering a period up to 2010. Cooke et al. (2019) conducted a content analysis focused on MNC and covered the period 2000–2014. None of those studies analyzed works on cross-cultural and comparative HR topics, as we do. In addition, our work provides an up-to-date overview of the field, as we cover the 2010–2018 period, to be able to inform about the latest studies and trends in IHRM. Then, our work contributes to the academia and practitioners in two main ways: (1) showing the most prominent journals to publish IHRM research and (2) documenting publication trends of quantitative oriented IHRM studies, by offering quantitative measures about the evolution of research methodology and statistical techniques applied on empirical quantitative research in the field. Alongside with the categorization of topics, our study will help identify methodological gaps and refinements for improvement, to detect trends and research gap, and to show the best journals for publication and dissemination of results. Further, our content analysis linking “when, where, what, and how” will also help unveil how the importance given by scholar to IHRM has evolved, increasing or decreasing against other international management topics along the current decade.
When, where, what, and how
The first step in our research design consisted of selecting “when,” the time-span we wanted to cover and the current decade was selected to conduct our analysis, from 2010 to 2018. The election of the period of these last 9 years is based on the opportunity to providing a completely updated view of the latest evolution of the subject, as well as to have a long enough time period to conduct an evolution analysis. Furthermore, choosing those years for the analysis, we do not overlap with Welch and Björkman’s (2015) work, which serves as the basis to configure the “what” section.
In order to identify “where,” different strategies to select influential journals are found in the literature, among them, using a list established in the literature or academically recognized (Werner, 2002), generate list by means of different quality indicators (Podsakoff et al., 2005), and generate an ad hoc list by means of an expert panel (Caligiuri, 1999).
As established lists and quality indicators are usually classified into wide areas (e.g. management) may not be very precise when the focus is on a single sub-area in the field of management as is the case of IHRM. Then, we believe the best strategy for selecting the journals may be generating an ad hoc list for the proposal of our work.
To our knowledge, only the ranking list by Caligiuri (1999) is concentrated in the sub-field of IHRM that was generated based on the answers of 22 experts in IHRM. Following her strategy, we have renovated her list and have selected the three top journals to which we have added German Journal of Human Resource Management (GHRM). As this works tries to attend the call of the special issue “Research Paradigms in International Human Resource Management” of that journal, its inclusion in our study is justified because readers of GHRM may be very interested in knowing how the journal is covering IHRM research and whether or not it follows a similar evolution in comparison to the most prominent journals in the field.
To determine “what,” we have followed Welch and Björkman (2015) and Brewster et al. (2016). The first part of our definition of IHRM (all issues related to the management of people in an international context including human resource issues facing MNCs in different parts of their organizations) is covered using the formers’ work, which defined IHRM as the concerning with all the issues related to the management of people in the MNC context. They identified two main research stream: stream 1, international assignment management, which encompasses articles dealing with “traditional” expatriates sent on foreign assignments; and employees on “alternative” arrangements like short-term assignments, self-initiated assignments, returnee assignments, international commuter assignments, rotational assignments, and virtual assignments. Stream 2, broader HRM issues in MNCs, including HRM issues arising from global expansion (staffing policies or HR practices in foreign subsidiaries, for instance), the HR effects of corporate practices, staffing of international operations (with expatriates or local employees) as well as the purposes and roles of different employee groups. They excluded from their analysis those papers dealing with comparative analyses of HRM and employee relations in different countries and those dealing with cross-cultural management issues in a comparative, or national, sense. Then, to cover the second part of our definition of IHRM, and following Brewster et al. (2016), we propose an additional research stream to include that kind of studies. Stream 3, cross-cultural/ comparative HRM, including those issues analyzing the impact of national and organizational culture on HRM as well as exploring differences in the way of managing human resources.
The main objective of our work was analyzing the methodology and statistical techniques applied on empirical quantitative research on the different topics on IHRM, to determine “how.” Thus, we just had into account those studies, for which we focused on the sample, the measures and the analytical techniques employed, in a similar way as the method section is reported in empirical quantitative papers.
A new ranking of IHRM journals
To rank journals in IHRM, we reproduced Caligiuri’s strategy. First, based on frequently cited academic articles and IHRM textbooks, we generated a list of 35 experts in the field. As she did, to ensure that the original 35 identified were the appropriate IHRM experts, we asked them to designate four more academic colleagues “who they felt were experts in IHRM.” We received back 24 surveys (69% response rate), who nominated 70.1% colleagues from our expert list. Thirty per cent of the scholars were affiliated with US-based universities, 30% were from UK universities, 30% were from other European Universities (Germany, Ireland, France; Spain, Finland, and Denmark) and the rest were from universities in Canada and Australia.
Following Caligiuri (1999), the experts were asked to give a response to two questions. The first one was “Name five academic journals that you would reference when writing a manuscript or researching a topic in International Human Resource Management.” According to the answers received, we elaborated a list with all the nominated journals based on the percentage of experts nominating the corresponding journal. Thus, the highest possible nomination score would be 100% (nominated by every academic) and the lowest would be 4.17% (nominated by only one person). Twenty-two journals were nominated. The scores ranged from 83.33% through 4.17% (Table 1).
Journal ranking according to % of mentions.
The second question was “Rank the top five journal outlets for publications in International Human Resource Management.” These rankings were coded from 1 = good to 5 = best. For each nominated journal, a mean score was created by adding all the ranking scores for a given journal and dividing by the number of raters. Thus, the highest possible score for this question would be 5 (considered best by every academic) and the lowest would be 0.04 (considered good by only one person). The mean scores ranged from 2.96 to 0.04 (Table 2).
Journal ranking according to mean scores.
The top five journals according to percentage of nominations were (1) International Journal of Human Resource Management (IJHRM), (2) Journal of World Business (JWB), (3) Journal of International Business Studies (JIBS), (4) Human Resource Management, and (5) Human Resource Management Journal.
Results based on rating scores means (question 2) produced similar results, as the top 5 journals were the same as for question 1, however, the order changed as Human Resource Management reached a better score than International Journal of Human Resource Management.
As for the proposal of our work, we had to select three journals, we clearly chose JWB and JIBS as the appeared in the top-three positions in both rankings. However, for the third journal to be selected, we had to make a decision, as conflicting results were found regarding questions 1 and 2. The International Journal of Human Resource Management (IJHRM) was finally chosen, based on the following criteria: (1) it was in the top position in the first list, what indicates is the most referenced journal in IHRM according to our experts responses; (2) contrary to the other two selected outlets (4 stars), IJHRM is a 3-star journal, what offers a wider perspective about works in the field; and (3) it is widely considered a reference journal in IHRM, as it has been selected among the most influential journals in IHRM by other authors (Welch and Björkman, 2015).
Methodology
Data collection
Given the huge amount of papers to be checked (Figure 1), we used the following strategy: First, an initial screening of all the papers was conducted to discard those works that did not match our definition of IHRM. This initial exploration was conducted by first (R.R.) and second (M.E.F.) authors who screened half of the papers each through the title, the abstract and the key-words of the papers to confirm relevance for the study (Cooke et al., 2019). So, the selection process was thought to be conducted by one reviewer per paper as opposed to two, in order to reduce the major effort associated with this task (Staples and Niazi, 2007). To ensure the criteria for selecting papers were consistent between the two researchers, 5% of papers were randomly selected for each journal and year to be independently screened by both reviewers RR and MEF to perform an inter-rater reliability test to confirm the consistency of the selection process (Staples and Niazi, 2007). Our results indicated excellent agreement in this first step as 96% of the papers were equally classified by both reviewers and Cohen’s kappa reached a value of 0.90. This first step allowed discarding 1822 articles.
Flow chart.
The 722 selected papers passed to second stage to determine their research stream –international assignment management (stream 1), broader HRM issues in MNCs (stream 2) and cross-cultural/diversity HRM and comparative HRM (stream 3). Furthermore, we also categorized papers using sub-streams of research defined by Welch and Björkman (2015): adjustment and failure, repatriation, expatriate career, spouse/partner/family, female/gender, expatriate performance including performance appraisal, selection and cross-cultural training, and compensation were defined for stream 1. MNC context; Headquarters HRM issues, staffing policies and practices, expatriate roles and purposes; subsidiary HRM; transfer of knowledge/HR practices; and HRM operation modes were established for stream 2. In addition, we considered the sub-streams cross-cultural/diversity HRM and comparative HRM for stream 3. We also indicated whether or not they were empirical and quantitative and to code them in that case.
Article categorization involved, thus, three rounds: First, sorting into research streams; second, thematic analysis within these streams; and third, full coding of methods for empirical and quantitative (or mixed) papers. Each article was independently analyzed and coded by two reviewers who entered data into corresponding excel files. We are cognizant that not all articles are easily classified. In fact, some single studies were difficult to be assigned to just one of the streams as some overlaps happened. To minimize the number of articles in possible conflict or disagreement we took different actions: (1) based on Welch and Björkman (2015), we followed a clear set of classification criteria from the beginning, (2) in case of possible overlaps, we focused on the most important aim of each study to categorize in one of the streams, (3) each article was scanned for key words using an agreed a priori coding system, and (4) we checked files to ensure accuracy and agreement and differences were discussed with a third researcher to reach inter-coder agreement, which provides a measure of consistency.
Each empirical quantitative (and mixed) article was fully screened and was coded by year of publication, journal in which the article was published, and research methods used, aspect in which we focused. We extracted information about the level of data/unit of analysis, year of collection of data, sources of data, whether or not data was entirely cross-sectional, country of data, and sample size. Next, we analyzed the measures. We compiled the type scales used in the study, whether or not the study was devoted to validate a scale, the number of constructs and the existence of control variables. Finally, we checked for the statistical techniques applied; whether or not the study tested for mediation and/or moderation; and whether or not the paper mentioned or studied the possible threat of common method effects (if required).
Results
Weight of IHRM research
From the 2544 screened papers, 722 (28.38%) lied under the umbrella of IHRM and were classified into one of our research streams—international assignment management (stream 1), broader HRM issues in MNCs (stream 2) and cross-cultural/comparative HRM (stream 3)—and sub-streams (Table 3). By journals, there are clear differences, as IJHRM is well over the other journals regarding this percentage. This journal presents similar percentages of works for each of the three research stream. On the other side, in GHRM, nearly half of the papers on IHRM are classified in cross-cultural/ comparative HRM. Finally, JWB and JIBS offer sensible higher percentages for broader HRM issues in MNCs, what is quite reasonable as those journals are focused on IB.
Stream and sub-stream research in IHRM papers by journal.
IHRM: International Human Resource Management; GHRM: German Journal of Human Resource Management; IJHRM: International Journal of Human Resource Management; JWB: Journal of World Business; JIBS: Journal of International Business Studies; HRM: Human Resource Management; MNC: multinational companies.
Regarding the evolution in the weight of IHRM research, Figure 2 offers percentages of IHRM research by year, by 2 years and by journal. The evolution of papers on IHRM by year offers a quite peaky pattern, as it is plausible to think journals that publish a large amount of IHRM papers one year, may compensate for it the next year, reducing that amount. Also, special issues published in certain years may contribute to this fact. In order to account for this and soften the pattern, the 2-year period evolution is also presented, offering a clearer picture. Apart from GHRM, which published a special issue touching IHRM in 2013, IJHRM presents a growing tendency on IHRM research until 2016, and then initiates a slow descent. JWB maintains similar percentages until 2014, when starts decreasing and JIBS keeps stable percentages until 2016 and then, goes down. Finally, regarding GHRM, as this journal has published a smaller number of papers in comparison to the other three journals, the corresponding percentage for works specifically focusing on IHRM might be over-representing IHRM research in the journal.
Evolution of percentage of IHRM papers.
The evolution of those topics along the decade is shown in Figure 3. The percentage of each of them by year, for the reasons already mentioned, again offers a peaky pattern. Once we have softened it considering 2-year periods, we appreciate quite stable percentages especially after 2014, sensibly higher for broader HRM issues in MNCs.
Evolution IHRM research streams.
Methodology of the IHRM research
Next, we coded selected articles into non-empirical and empirical studies. For empirical studies, we used three empirical categories, that is, qualitative, quantitative, or mixed-method.
More than 79.5% (576 over 722) of studies in this review were empirical. From those, 37.3% were qualitative, 57.3% relied on quantitative research methods, while the remaining 5.4% adopted a mixed-method approach. Table 4 offers the main results for empirical methods, both by journal and by research stream.
IHRM empirical papers distribution by research stream and journal.
IHRM: International Human Resource Management; GHRM: German Journal of Human Resource Management; IJHRM: International Journal of Human Resource Management; JWB: Journal of World Business; JIBS: Journal of International Business Studies.
The evolution of empirical methods is pictured in Figure 4, that shows a predominance of quantitative methods over qualitative, and both are much wider used than mixed methods.
Evolution of IHRM empirical studies.
Data in IHRM empirical quantitative research
Three hundred and sixty one papers were empirical and quantitative/mixed and, thus, fully reviewed. Table 5 shows the different levels of data in the reviewed papers, as well as average sample sizes and average gap between data collection and publication. We have to note, however, that for computing this gap, we have only have into account 159 papers, as more that 58% of empirical quantitative studies did not report the moment of gathering data.
Level of data, average sample size, average collection-publication gap, and type of data.
MNC: multinational companies.
In brackets, standard deviation.
Two outlayers have been excluded to compute this average.
Five outlayers have been excluded to compute this average.
Five outlayers have been excluded to compute this average.
We coded levels of data into individual, unit (teams, sub-units, departments in broader HRM issues in MNCs and mainly organizations in cross-cultural/ comparative HRM); subsidiary, MNC, and country. The “Mix” category includes those studies with data at different levels (hierarchical/multilevel data). In it, the most common combinations were individual/organization and individual/country.
Most works used data gathered at the individual level (204 over 361), followed by mix-level data and data at the organization and MNC level. The vast majority of quantitative works relayed on primary data based on surveys (88.36%, 319 over 361), mostly self-reported (52.98%, 169 over 319) or provided by CEOs, managers or similar (23.20%, 74 over 319). The vast majority of works used their own survey, although some others used established ones, as CRANET, GLOBE, INTREPID, WVS surveys. Few works complemented survey data with secondary data from other sources (7.84%, 25 over 319) and only 30 works relied just on observed/objective data (e.g. private archival data from companies, company reports and public data bases such as country census data, data from Organisation for Economic Co-operation and Development (OECD), World Bank and others), remarkably in stream 2 (broader HRM issues in MNCs). Accordingly, most works were based on data from just one source/rater (61.7% 223 over 361). In the same vein, most of data were entirely cross-sectional (77.84%) and, thus, only few works counted on panel data or data gathered in different time points.
Regarding the countries involved in the collected data, IHRM research has pivoted around developed economies (mainly European Union (EU), United States, Japan; Australia, Canada) and emerging countries as China, Korea and India. The degree of concentration of studies on countries is high, as signaled by percentages in Table 6, indicating the large weight of the 15 most frequent countries on empirical quantitative studies in IHRM.
Countries studied by research stream.
Variables and measures of quantitative IHRM research
According to the previous section, as most of the analyzed works used survey data, the used of different scales becomes a critical issue. Likert-type-scales and yes/no questions are the most frequent in surveys. Particularly, scales ranging from 1 to 5 and 1 to 7 are the most popular (65.65%, 237 over 361). A few works (13.85%, 50 over 361) used non-centered Likert-type scales and others (13.57%, 49 over 361) combined the use of different Likert-type scales. The vast majority of studies used scales already validated in the literature, although a few papers had the aim of validating their own scales (23).
Constructs were widely used as a way of dealing with complex concepts appearing in the explored works (Cascio, 2012). Particularly, 73.13% (264 over 361) of the studies used them and 235 works dealt with three or more constructs. However, although reliability of the scales was addressed by all the screened studies (mainly using Cronbach’s alpha), validity of scales was disregarded in many occasions. Factor analyses (exploratory, EFA, and confirmatory, CFA) were mainly used on this regard. Figure 5 shows the evolution in the used of factor analyses.
Evolution of the use of factor analyses to test the quality of measurement model.
Regarding control variables, 262 of the works used them, although that use depended, to a high extent, on the research question and methods used. Most of the control variables were observed. When the analysis was at the individual level, most frequent controls were gender, age, position, and tenure. When the controls were at the subsidiary/MNC level, age, size, industry, sector, and financial controls were mainly used. At the country level, gross domestic product (GDP), culture, and values items were widely employed.
Analytical strategy and statistical techniques of quantitative IHRM research
In line with results in previous section, the analytical strategy of most works consisted of analyzing the quality of the measures (if needed) and then testing the different hypotheses by means of different statistical techniques in a double approach as recommended by Anderson and Gerbing (1988). Regarding the first issue, all the works reviewed provided a measure of the reliability of the scales employed, normally, Cronbach’s alpha. To check for the validity of the measures exploratory and confirmatory factor analysis were widely used (92 EFA, 123 CFA; 34 both). However, there are more works dealing with three or more constructs (235) than using at least one of those techniques (181).
Regarding the second issue, Table 7 offers a perspective of the techniques used by research stream. In “Inference tests to compare differences” methods such as t-tests, analysis of variance (ANOVA), analysis of covariance (ANCOVA), multivariate analysis of variance (MANOVA), multivariate analysis of covariance (MANCOVA), and some others are included. A big amount of works used some of those techniques (32.96%) but mainly as a secondary or additional method to complement “the main method.” Accordingly, only 11.08% of the works used these techniques as the main method. We have separately considered different regression analyses: Traditional ordinary least squares (OLS) regression analysis, hierarchical (step-wise) regression analysis and multilevel regression analysis. The interest of this distinction is offered in the discussion section. As Table 7 shows, those regression methods were, by far, the most widely used in IHRM, in the three streams of research. In structural equation models (SEM)/ partial least squares (PLS)/Path-analysis those methods using covariance structure for estimation are encompassed. As we reported, SEM, PLS, and path-analysis were the three most frequently applied in the reviewed works. In all, 13.57% of the checked works used one of these techniques. In “Bounded dependent variable analyses” those methods designed to deal with a dependent variable whose range of values is bounded or restricted are included, for example, logit, probit (binary or ordered), tobit. In our review, 12.19% works used at least one of those techniques. Finally, “Other” included a wider range of techniques less applied in the reviewed works, such us cluster analysis, experimental designs, panel models, and so on.
Statistical techniques applied overall and by research stream.
The evolution in the use of these techniques over the current decade is pictured in Figure 6. For the sake of clarity, we have considered hierarchical regression and regression analyses in a single category (as, broadly, are the same methodology) and we have omitted the category “Other.”
Evolution of statistical techniques applied.
The technique most widely used by far is regression analysis and its percentage of appearance in works is not decreasing over time. SEM/PLS/Path analysis and bounded dependent variable methods appear with similar frequency, suggesting a slow growth over time. The use of multilevel analyses has considerably increased from 2016, and, on the contrary, the use of inference test to compare differences as main method was low and is decreasing over time.
From the 361 works fully analyzed, 86 used/tested mediating relationships (Table 8). In all, 43 of those mediating relationships appeared in models using regression analysis and 41 in SEM/PLS/path-analyses and 2 in logit models. The approaches more widely applied to test mediation were the one proposed by Baron and Kenny (1986) (19); analyzing indirect effect (40) by means of bootstrapping, Sobel’s tests and other techniques and the used of competing models (8).
Works studying mediation and moderation on IHRM empirical quantitative/mixed research.
Regarding moderation, 144 works tested it, mainly in regression models (117), SEM or similar models (14), and bounded dependent variable models (13). In all cases, moderation analysis was mainly performed by introducing interaction effects (120) and by multigroup analysis (21).
Figure 7 shows the evolution of these methods. Both, mediation and specially moderation present a growing tendency. For mediation, the use of Baron and Kenny’s method is decreasing, while the analysis of indirect effect is increasing. This fact may be explained by the introduction of different macros computing indirect effects and their significance. For moderation, interaction effects are clearly used well over alternative methods, and their use keeps a stable tendency over time.
Evolution of mediation and moderation testing.
CMV on IHRM research
From the analyzed works, the 319 based on survey data were subject to common method variance (CMV). However, only 144 mentioned it. Some assessed the lack of problems regarding CMV providing some evidence by means of statistical techniques such as Harman’s one factor test (44), unmeasured latent factor test (9), or factor analyses (18). Some others reasoned about not suffering major problems regarding CMV (30) by using some of the remedies signaled by Podsakoff et al. (2003) as using several raters (22), or gathering data in different time points (5). Sometimes, these techniques are applied jointly in each paper. Finally, some others (40) commented on CMV as a limitation of the work. Particularly, it is interesting to highlight that 65 works used self-reported data, entirely cross-sectional and with just one source/rater and did not mentioned CMV. The good news is that the percentage of papers tends to diminish over time (Figure 8). In addition, the percentage of works not dealing with CMV in the best possible way is decreasing along the years, and the percentage of works specifically testing it or using data with good properties as a remedy is increasing over time.
Treatment of common method variance.
Discussion
The development of IHRM research in the last decades has generated not only an increase in the knowledge in the field but also a significant progress of new research topics, research questions, countries being researched, methods applied, and levels of analysis (Dowling et al., 2017; Pudelko et al., 2015; Stahl et al., 2012).
Among the circumstances explaining that development, we highlight two. First, the evolution of the nature of HRM from personnel management to the considerations of HRM as a key aspect for the competitive advantage of the organization (Boxall, 1994, 2015; Collings et al., 2018) and second, the entrance of the firms into the more dynamic word of IB. Mainly due to this second reason, numerous new issues have risen in HRM, given, among other things, the diversity of employees in an international context (expatriates, host-country nationals, third-country nationals and inpatriates). In the infancy of the field, the complexity of operating in different countries and employing different national categories of workers was the key issue that differentiated domestic and international human resource management, rather than any major differences between the HRM activities performed (Dowling, 1999). However, given the current globalization of business and the current diversity in the workforce even in a single country, the boundaries of IHRM are now more difficult to draw, as fields as comparative HRM and cross cultural studies are closer to domestic HR and IHRM coexists and overlaps with other research areas such as IB and global strategy and organizational behavior.
The substantial expansion of IHRM explains the results of our work, including the need to introduce a new research stream (cross-cultural/comparative HR research) from those raised by Welch and Björkman’s (2015) to account for the diversity of topics currently related to IHRM, and the specialization in the top-five most prominent journals in IHRM, just dealing with IB or HRM, that have replaced more general management journals in those positions. This could be a consequence of the mentioned overlap between IHRM and organizational behavior and IB topics.
The evolution and complexity of IHRM research have also affected the methodology, which is the focus of our review. First, we observed that among IHRM studies, the vast majority were empirical and, among them, quantitative were still predominant. This result may be explained because authors want to make a clear link between scholarly research and practice, and thus, are led by the fact that managers can only focus in factors that are easily quantifiable to make their decisions (Pudelko et al., 2015).
According to our analysis hierarchical regression/regression analyses have been, by far, the most common statistical techniques. In our review, the use authors made of the term “hierarchical regression,” as Cohen et al. (2003) proposed, is mainly linked to traditional regression models in which predictors are entered in some particular theory-based order and where changes in regression coefficients and variations in explained variance are evaluated (step-wise regression). Given that a high amount of works dealt with constructs and measurement models along with the fast advance in software packages, it is surprising that less than 14% of the screened works used a structural equation model, described as a statistical technique very appropriated for studying complex phenomena (Batista-Foguet and Coenders-Gallart, 2000) due to the advantages it offers as opposed to more traditional regression analyses (Batista-Foguet and Coenders-Gallart, 2000; Bentler, 1995; Golob, 2003; Ma et al., 2012).
Despite this concern, it is important to highlight that many quantitative studies enrich the explanation of the phenomenon under research with the inclusion of mediation and moderation analyses. Although we have not specifically investigated this, the increasing use of moderation analysis might suggest an increasing examination of phenomena with a more contextualized look, as moderation analysis tries to identify whether or not the intensity of a relationship varies according to certain (moderating) variables or factors, what allows sensing a certain departure from more universalistic conceptions. In order to test mediated and moderated relationship, according to our results, the most frequently used methods to test mediation are the one proposed by Baron and Kenny (1986) and the analysis of indirect effects significance (Judd and Kenny, 1981; Sobel, 1982). However, a fewer number of works using SEM tested mediation by Chi-square tests of competing models (e.g. Frenkel et al., 2012). Regarding moderating test, the most widely used methods were incorporating interaction effects and performing multigroup analyses.
In addition, our work analyzes other methodological aspects that help unveil current challenges in IHRM research. Regarding data collection, it is typically performed through surveys where self-reported questionnaires are the most frequent procedures. Those data collection strategies are, however, the weakest form of data collection (Teddlie and Tashakkori, 2003) because they generate several bias, particularly, common method bias (Cascio, 2012; Chang et al., 2010). Our review shows that not many studies took controls to mitigate that threat (e.g. more than one rater or source, data gathered at different time points –Podsakoff et al., 2003) and even fewer specifically analyzed the risk by means of appropriated tests (e.g. Harman’s single-factor test). Although depending on the level and the type of analysis there may be no available archival data to avoid the threat of CMV (Wall et al., 2004), it is clear that the use of richer data would help to diminish the risk. Nonetheless, despite the detected weaknesses, our results also show a substantial improvement in the use of more appropriated methods and tests in this regard, although the remaining challenge is using multiple databases and several sources to serve as the empirical basis of the quantitative analyses (Cooke et al., 2019; Pudelko et al., 2015).
In line with the use of “poor” data, particularly cross-sectional data, a new challenge emerges from our work for IHRM, and it concerns with causality analysis. If we are to build a foundation of empirical evidence that support causal inferences, then longitudinal studies are necessary (Wall and Wood, 2005). However, in practice, the use of longitudinal data is rare. These findings are is in the same vein as the ones in Cooke et al. (2019) for MNCs, who also signaled this important absence is comprehensible given the difficulty in gaining access for longitudinal studies. In this sense, despite the increasing number and sophistication of available means of asking for and collecting data nowadays, decline in respondent cooperation and attrition are one of the main difficulties researches have to handle in order to collect longitudinal data (Rindfuss et al., 2015). Accordingly, the challenge of using richer data is worthy but difficult to face, and still remains.
Regarding measures, psychological constructs are the scientific tools used to facilitate understanding of human behavior (Cascio, 2012) as they allow capturing concepts that are not tangible entities (e.g. global-mind-set, boundarylessness, self-efficacy). Accordingly, most of the screened studies used constructs. Then, some sort of conceptual framework is required to specify the meaning of the construct, to distinguish it from others, and to indicate how measures of the construct should relate to other variables (Cascio, 2012); that is, construct-related evidence of reliability and validity is required. Although all the analyzed works provided measures of reliability (basically Cronbach’s alpha), many manuscripts lacked a validity test, especially if referring to discriminant validity of the scales. This weakness of IHRM research is not new, as was already detected by Cascio (2012). Despite our evolution analysis pointing to an improvement over years, the challenge of better assessing the scales used remains.
Finally, our study also offered an overview of the countries involved in the screened studies. As different authors have signaled (e.g. Cooke et al., 2019), there is still an important bias and concentration of studies toward western and developed countries and Asian emerging economies. East European countries are increasingly appearing in those studies but Africa and Latin America have been largely neglected. Only countries as South Africa, Brazil and Mexico have received certain attention. Therefore, it would be interesting to cover this research gap in future studies.
Notwithstanding, our work has some limitations. First, due to the decision to refer to frequently cited IHRM scholars to select the experts participating in our study, the answers may be limited to voices stemming from Western countries, affecting the possible range of named publication outlets. Thus, this method might reproduce existing publication biases. In this vein, our review only considered the three journals designated by our experts as most prominent in IHRM (plus GHRM), and, accordingly, there are some significant works on IHRM that have not been screened. Second, we only checked for studies published in English, which means that certain works in GHRM have not been analyzed. Nonetheless, we are confident that our search covers a representative sample of the studies on the topic and our analysis offers a precise picture of the methods applied in empirical quantitative IHRM research.
Conclusion
Our literature review on the “when, where, what and how” of empirical quantitative IHRM research have yielded some important results and contributions. First, by reproducing Caligiuri’s (1999) strategy, we produced a new list of the most prominent scholarly journals for publishing IHRM research, finding a high agreement among experts regarding the top five journal outlets for the sub-field of IHRM: International Journal of Human Resource Management, Journal of World Business, Journal of International Business Studies, Human Resource Management, and Human Resource Management Journal.
Second, we documented publication trends as wells as trends in the application of different quantitative methods in IHRM research in three of these journals (plus GHRM) from 2010 to 2018. Regarding publication trends, we were able to show how IHRM research kept a growing tendency until 2015, when a slight decline began. The evolution of topics along those years drew a quite stable pattern, however, showing a decrease of international assignments topics and a certain increase in cross-cultural/comparative and MNC studies, being the latter the most numerous every year. Empirical quantitative studies were the most abundant in all the cases. Regarding the trends in the application of different methods in those quantitative (or mixed) IHRM studies, very clear conclusions are drown: (1) the bulk of the data employed in those studies were cross-sectional, from just one source and gathered from surveys; and, accordingly, highly subjected to CMV. It is remarkable that methodological controls and analyses are being increasingly used to address this threat. (2) Regarding variables, most of the works employed more than 1-item measures, leading to the use of constructs. However, studies not always deeply assessed the measurement model. (3) The most widely employed statistical technique is, by far, OLS (whatever the version) regression analysis, what points out that analyses have not been subjected to high degree of sophistication but, rather, have been kept simple. This fact must be nuanced by the increasing use of mediator and moderator analyses, allowing a deeper look into a research problem, despite the statistical technique may be simple.
In sum, although in our review we have detected an increasing progress in the methods used in IHRM, as we have signaled, there is still a way to go as we have found several ways of improvement and remaining challenges to be faced.
