Trust and signals in workplace organization: evidence from job autonomy differentials between immigrant groups

Abstract

1. Introduction

The role of trust in governing workplace organization is widely recognized (Arrow, 1974; Granovetter, 1985). Fukuyama (1995, p. 31), for instance, finds that trust fosters flexibility in the workplace and the allocation of greater responsibility to lower levels of the organization. Several studies subsequently link societal differences in trust norms to average firm size and, particularly, the granting of autonomy to employees (La Porta et al., 1997; Bloom et al., 2012; Van Hoorn, 2017). Job autonomy thereby matters because it allows for more specialization and higher productivity as a result of expert employees leveraging their tacit knowledge to organize their tasks in the way they deem best. However, we lack a full understanding of the deep processes that affect the link between trust and key features of workplace organization such as job autonomy.

This paper seeks to unpack the black box of trust governing exchange in the context of workplace organization, specifically the decision of how much autonomy to grant to an employee. In market settings, trust is understood to involve both a willingness to be vulnerable to the actions of the other party and an assessment of the counterparty’s trustworthiness (Arrow, 1972; Granovetter, 1985). Not all (potential) exchange partners are equally trustworthy, and drawing on a variety of signals allows actors to make an informed assessment of the risk of transacting with a particular party (Coleman, 1990; Hardin, 2002; Gambetta and Hamill, 2005). In workplace settings, principals that outsource tasks to agents are similarly vulnerable to agents’ actions. In general, decentralization and outsourcing of tasks are desirable for efficiency reasons. Trust issues, however, prevent principals from simply granting complete autonomy to their agents and reaping the full benefits of specialization through the division of labour. Hence, employers need to differentiate between those employees that they can trust more and offer higher degrees of job autonomy to and those employees that they can trust less and need to monitor and control more closely. The concrete aim of this paper is to uncover some real-world specifics of the process and factors guiding employers in deciding on how much autonomy to grant to different employees.¹

The literature with most relevance to this issue involves studies that consider employers differentiating between employees in the context of recruitment decisions. Key insights from this literature concern the role of signalling (Spence, 1974) and statistical discrimination (Arrow, 1973). Faced with limited information by which to judge potential employees, employers rely on employee signals as well as other observable characteristics to make rational inferences about underlying intangible traits and dispositions (Altonji and Pierret, 2001). Oft-mentioned signals are educational credentials (Chatterji et al., 2003), but (involuntary) membership to a particular social group (e.g., blacks versus whites or males versus females) can also be a powerful basis for statistical discrimination.² Also relevant is the literature on trust in (quasi-)experimental games, specifically studies of the effect of (unintentional) signals on the amount of trust that an individual trustor places in certain trustees. Fershtman and Gneezy (2001), for instance, report that among two groups of Israeli Jews, those from European and American descent were trusted more than their fellow citizens from Asian and African descent were. Other work finds that even a simple signal such as counterparties’ physical appearance can affect trustors’ behaviour in experimental trust games (van ’t Wout and Sanfey, 2008; Eckel and Petrie, 2011). Experimental results by McEvily et al. (2012) further indicate that laboratory trust decisions are shaped by trustors’ perceptions of trustees, which, in turn, are based on observable background characteristics. Finally, the analysis of self-reported trust in Uganda by De Luca and Verpoorten (2015) finds evidence of distrust towards specific individuals based on these individuals’ association with violent events.

Following the above body of research, the specific feature of the trust-autonomy nexus considered in this paper is how group membership, as emphasized by theories of statistical discrimination (Arrow, 1973), can go on to generate job autonomy differentials between individuals with otherwise similar qualities and features. Because individual employees’ trustworthiness (or other relevant qualities) are only partially observable, I expect that employers use group membership—and specific traits of these groups—as a signal of trustworthiness. The implications is that employers are more/less willing to grant autonomy to employees belonging to some social groups than to others, even when these employees do not deviate on the most important observable individual-level characteristics, for example, their education level. Trustworthiness is often considered in terms of reputation building and repeated interactions that allow trust between two parties to develop over time (Granovetter, 1985; Dasgupta, 1988). My interest is not in specifics of the relationship between selected principals and agents, however, but in broad patterns of group differentials in job autonomy that testify to the real-world process and factors that guide employers’ decisions of how much autonomy to grant to specific employee groups.

Reflective of this interest, the chosen research context for my analysis is a cross-national sample of immigrants from different birth countries living in various host countries, which is close to ideal for my purpose because ethnonational identity is easily recognizable and widely considered a chief criterion for social categorization and group classification (Barth, 1969; Jenkins, 1994). My empirical evidence subsequently concerns the following two distinct but related issues. The first is the broad issue of the presence of variation in job autonomy between groups of migrants with different ethnonational identities not accounted for by (easily observable) individual-level traits. The second issue concerns specific group-level traits, i.e., features of migrants’ birth countries, that employers might draw on to differentiate between migrant groups when deciding on the amount of autonomy to grant to their employees. For this latter issue, I focus on the role of the positive or negative image concerning honesty and reliability that different birth countries may have, which I operationalize by considering the level of corruption in a migrant’s country of birth. The two hypotheses that I test are as follows. The first is non-directional: 

The second hypothesis is more specific: 

The empirical analysis provides robust support for these hypotheses, indicating substantial variation in job autonomy between migrants from different birth countries and a strong negative relationship between birth-country corruption and job autonomy. These results are robust to controlling for differences in the main (observable) individual-level factors logically affecting individuals’ job autonomy, for instance, education level, as well as, for example, differences in preferences. Hence, it seems unlikely that these results are driven by variables concerning individual-level traits that are (directly) observed by employers but not adequately controlled for in the empirical analysis. Overall, the evidence strongly supports the idea that country of birth, and birth-country corruption in particular, are taken as a signal of individual migrants’ trustworthiness and end up affecting how much job autonomy migrant employees from different birth countries have in their host countries.

The key contribution of this paper is to present real-world evidence on an important feature underlying the process of trust governing exchange in the context of workplace organization. In doing so, this paper helps extend and bring together a set of disparate literatures. While employers’ reliance on employee signals and statistical inferences when making employment decisions is widely recognized, these decisions are typically limited to recruitment and selection. Moving beyond initial recruitment decisions, this paper shows the relevance of signalling and statistical discrimination also in the post-recruitment managerial treatment of distinct groups of employees. Similarly, while prior research has found that various traits of the counterparty can affect an individual’s trusting behaviour, this evidence remains limited to decisions made in laboratory settings. This paper, in contrast, has sought to consider how employers’ consideration of observable signals of (un)trustworthiness pans out in the real world, giving rise to systematic patterns of job autonomy differentials between groups of migrants with different ethnonational identities. Finally, though not the main concern of the present paper, the evidence of the effect of birth country on individuals’ managerial treatment testifies to the importance of statistical discrimination for the extent to which individual migrants are able to integrate successfully in the workplace. As one’s country of birth is strictly beyond one’s control, how a migrant’s birth country scores on various indicators may be one of the most significant barriers that a migrant faces in achieving professional success in his/her host country.

2. Theoretical background and empirical context

2.1 Job autonomy and signals of trustworthiness

This paper’s interest in job autonomy as a key feature of workplace organization resonates with the long-standing literature relating the organization of the workplace to possible efficiency gains due to specialization that traces back to Adam Smith’s famous pin factory. A straightforward definition of job autonomy is ‘the condition or quality of being self-governing or free from excessive external control’ (Jermier and Michaels, 2001, p. 1006). When it comes to job autonomy, the ultimate challenge that employers face in deciding how much autonomy to grant to different employees is to strike a balance between the costs and benefits of different amounts of autonomy versus the intensity of monitoring and control. Monitoring and control thereby have direct costs in terms of taking up some of the firm’s resources, for instance, managerial attention, but also indirect costs. More importantly, however, there is an essential connection between the costs and benefits of control on the one hand and the costs and benefits of autonomy on the other.

The classic understanding of the specialization benefits of employee autonomy comprises two elements. The first is that employees are specialists that have gained unique knowledge on how to perform their production tasks most efficiently. The second is that the specific knowledge or skills that employees have accumulated are typically tacit (or at least only partly codifiable), so that leveraging this knowledge requires that employees are granted freedom to perform their jobs in the way they deem best. As non-specialists, managers or employers should refrain from prescribing employees how they ought to do their job, as the former’s lack of relevant knowledge results in a production process that is less efficient than a production process that is organized by specialist employees themselves. Part of the costs of monitoring and control is thus that they prevent the reaping of efficiency gains from specialization. In contrast, the costs of autonomy are that lack of monitoring and control gives employees more opportunity to shirk. If employees have complete autonomy, there is no formal mechanism that ensures that employees act in the best interest of their employer or prevents employees from pursuing their own interests at the expense of their employers’ interests. The benefits of monitoring and control are that they help reduce employee shirking.

Trustworthiness matters because it changes the balance between the costs and benefits of autonomy versus control. If a principal can trust the agent to look after the principal’s interests and not to shirk, there is simply less need for monitoring and control so that autonomy can increase, allowing for more efficiency gains from specialization (Gur and Bjørnskov, 2017; Van Hoorn, 2017). Vice versa, the costs of autonomy are higher—and, hence, the gains from control higher—in case an employee cannot be relied upon to work diligently, absent any formal mechanism for ensuring cooperation. The degree to which specific employees are honest and can be relied upon is not typically well known to an employer, however. Trustworthiness is characteristically difficult to observe, meaning that employers need to rely on signals such as group membership and information on specific group-level traits to make inferences about individual employees’ trustworthiness. These inferences result in differentiation between employees and are taken into account in the decision of how much autonomy to grant to specific employees. In practice, there can be many signals or traits that employers can draw on to infer trustworthiness. Hence, my generic proposition that individuals recognized to belong to social groups with positive/negative images concerning their honesty and reliability are deemed more/less trustworthy and have more/less job autonomy.

2.2 Empirical context

Empirical testing of the above proposition requires a research context that involves multiple social groups as well as the possibility of identifying a specific group-level trait that would allow employers to infer differences in trustworthiness and adapt their decision on how much autonomy to grant to a particular employee accordingly. As stated, the chosen research context for my empirical analysis is immigrants originating from different countries of birth (immigrants with different ethnonational identities). The reason for choosing this particular research context is threefold. First, ethnonational background is widely recognized as a chief criterion for social categorization, meaning it is common for an individual to classify other individuals into distinct social groups delineated by, say, nationality or individuals’ country of birth (Barth, 1969; Jenkins, 1994). Second, it is common for people to hold a stereotypical image of particular countries (Madon et al., 2001; Schneider, 2005).³ Finally, the content of such national stereotypes can be traced back to specific country characteristics, for which ample secondary data are available, for instance, countries’ economic status (Lee and Fiske, 2006).

Following my use of a migrant sample, I operationalize the idea of different social groups in terms of individuals’ ethnonational background, specifically their country of birth. My first hypothesis derives directly from the generic proposition presented above and concerns assessments of trustworthiness and employers’ use of statistical discrimination broadly. Given that observable individual-level traits only go so far in informing employers about an employee’s trustworthiness, I deem it likely that group membership, notably ethnonational identity, is a factor that employers take into account when deciding how much autonomy to grant to a particular employee. If so, I expect to see systematic differences in how much autonomy individuals from different countries of birth have at their jobs. Hence, the hypothesis (H1) that, controlling for observable individual-level characteristics, there are significant differences in job autonomy between groups of migrants from different birth countries. Importantly, this is a non-directional hypothesis that is only meant to show the real-world relevance of statistical discrimination and the signalling power of group membership, over and above the various observable individual-level traits that employers are likely to take into account when deciding how much job autonomy to grant to a particular employee.

My second concern is with specific group-level traits that employers may take as a signal of individual employees’ trustworthiness. Beyond group membership itself, there are many group traits that employers can, in principle, draw on to make inferences about employees belonging to a specific group. However, when it comes to assessing trustworthiness, the degree to which corruption is institutionalized in an individual’s birth country would seem a most salient and easily recognizable trait to base one’s inference on.⁴ I therefore expect that differences in the level of corruption in migrants’ birth countries give rise to systematic differences in inferred trustworthiness and thus give rise to clear patterns of job autonomy differentials between various migrant groups. Hence, the hypothesis (H2) that the higher the level of corruption in a migrant’s birth country, the less job autonomy this migrant has in his/her host country.

I test this hypothesis as well as the hypothesis on variation in job autonomy between migrants from different birth countries below. First, however, I discuss some specifics of my empirical method and the data that I use to estimate my empirical models.

3. Method and data

3.1 Method

The most important difference between eq. (2) and eq. (1) is the adding of a predictor variable at the birth-country level. This variable, C_b, denotes the degree to which corruption is institutionalized in the individual’s birth country, while X_ib is again a set of (individual-level) control variables and ε_ib is a random disturbance term. The evidence supports my second hypothesis if the coefficient for birth-country corruption (β₁) is statistically significantly negative. Importantly, when estimating eqs (1) and (2), I take into account that my data have a hierarchical structure with individuals nested in birth countries. For eq. (2) in particular, this means that I use robust standard errors that are clustered at the birth-country level. Finally, the birth-country subsamples in my analysis can have highly unequal size, with some subsamples containing very few individual observations. Hence, to assess the robustness of my baseline results, I also check whether I obtain similar results when considering only birth-country subsamples with pre-specified minimum numbers of individual observations.

3.2 Data source and sample

Data for my analysis come from seven waves of the European Social Survey or ESS (European Social Survey, 2016), supplemented with cross-country data from various sources. The ESS is a bi-annual survey of nationally representative samples from more than 30 countries, mostly in Europe but also covering such countries as Israel, Turkey, and Russia. Following my interest in inferences based on group membership, I use only a portion of all respondents in the ESS, namely those respondents that currently live in a country other than their birth country. I identify these individuals using the answer to the ESS item asking respondents whether they were born in their current country of residence. Foreign-borns from a particular birth country living in a particular host country are typically migrants, which have been extensively studied using data from the ESS. I further focus on actual employees, meaning that I do not consider respondents that are self-employed or working for their own (family) business, as these individuals typically are themselves employers rather than employees. Finally, I do not consider the subset of individuals that are not living in their birth country, but at the same time are living in the birth country of their parents. An example would be a child born abroad during an extended holiday. Excluding respondents with missing data on the variables considered in the analysis leaves a main sample of about 10,800 individuals from approximately 170 birth countries. Table A1 in the online appendix (this and all other appendix tables can be found in the Supplementary material) presents descriptive statistics for a selection of variables used in the analysis, while Table A2 presents an overview of the birth countries in the sample, sorted by average level of job autonomy. More information about the ESS is available from the survey’s website, http://www.europeansocialsurvey.org (last accessed 14 February 2018).

3.3 Variables

3.3.1 Dependent variable

The dependent variable in my analysis is the amount of autonomy an individual experiences at his/her job. As a key feature of workplace organization, much effort has been devoted to measuring job autonomy, particularly by business and management scholars. The standard approach is to use surveys, asking respondents to rate their own autonomy (Hackman and Oldham, 1975). I use this standard approach, which has the advantage that it does not suffer problems deriving from the fact that formal autonomy or authority of the type that can be measured by external observers need not match real autonomy (i.e., the type of autonomy that employees actually experience in their daily work activities) (Aghion and Tirole, 1997). The wording of the job autonomy item included in the ESS is as follows: ‘Please say how much the management at your work allows/allowed you to decide how your own daily work is/was organised.’ And the accompanying answering scale ranges from 0 (‘I have/had no influence’) to 10 (‘I have/had complete control’). To keep things simple and facilitate interpretation of the results, for my main analyses, I assume that job autonomy is measured on a cardinal scale. However, results (available on request) are similar when I treat the job autonomy measure as an ordinal indicator and estimate ordered probit or ordered logit models instead. Importantly, subjective job autonomy indicators of the type included in the ESS, and the ESS survey item in particular, have been widely validated (Morgeson and Humphrey, 2006; Van Hoorn, 2016a). Table A3 in the appendix presents some stylized evidence on the construct validity of the ESS job autonomy item, particularly on the relationship between measured job autonomy and other features of an individual’s job. If the ESS job autonomy measure is valid, we expect, for instance, a positive relation between the non-routineness of work and job autonomy and clear differences in job autonomy between the self-employed and ordinary employees, which is confirmed by the evidence.

3.3.2 Main independent variables

Testing my hypotheses involves two different but related independent variables. The first, generic variable concerns individuals’ country of birth. If respondents indicated that they were born in their country of residence, the ESS followed up with a question asking in which country the respondent was born. Hence, I am able to identify migrants’ country of birth. As mentioned, the resulting sample comprises individuals from some 170 different countries of birth. My second main independent variable concerns the degree to which corruption is institutionalized in the birth countries of the migrants in my sample. The specific measure that I use is the Control of corruption measure from the Worldwide Governance Indicators (WGI) project (World Bank, 2016), which captures ‘perceptions of the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, as well as capture of the state by elites and private interests’ (Kaufmann et al., 2009, p. 6). I recode this measure so that higher scores indicate higher levels of corruption in the individual’s country of birth. The WGI project has collected cross-country data bi-annually since 1996 and annually since 2002. A priori, however, it is not clear that one particular year of observation is more representative of former residents of a country that have migrated abroad than another year. I thus calculate the average corruption level in the birth country over the years covered by Waves 1–7 of the ESS: 2002, 2004, 2006, 2008, 2010, 2012, and 2014. However, to rule out that my results are sensitive to the period chosen, I repeat my baseline analyses replacing this measure of the degree to which corruption is institutionalized in migrants’ birth country with a measure of birth-country corruption based on averages over the years prior to 2002 (1996, 1998, and 2000). As a second alternative, I construct a measure of the cultural norm of corruption that exists in the birth country based on Fisman and Miguel’s (2007) data on unpaid parking tickets by UN diplomats in New York. My measure of the cultural norm of corruption in a country is simply the average number of unpaid parking violations per country diplomat per year. Meanwhile, a generic challenge to considering birth-country differences in corruption is that this measure is not, in fact, very informative of individual migrants’ trustworthiness. Above, I already noted that for the validity of my empirical analysis it does not really matter whether birth-country corruption is an accurate signal of trustworthiness—the reason is that my only concern is whether corruption appears to be used as a signal of trustworthiness and not whether corruption truly reveals something about migrants’ actual trustworthiness. Nevertheless, as a further robustness check, I also estimate models using two alternative birth-country features that employers may draw on to make inferences about individual migrants’ trustworthiness. The first of these is the quality of law enforcement and the judicial system in the birth countries of the individuals in my sample. The specific indicator that I use is the Rule of law measure, also from the WGI project, which captures, among others, ‘perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence’ (Kaufmann et al., 2009, p. 6). As before, I calculate averages for the years 2002–2014, where a higher Rule of law score indicates higher-quality law enforcement and less crime and violence. The second alternative to corruption concerns corporate ethics in an individual’s birth country. Compared to standard corruption measures, corporate ethics is more concerned with firm behaviour and workplace practices and less with public officials, meaning that birth-country differences in corporate ethics might be a more relevant signal for employers than birth-country corruption is. The specific measure that I use is the corporate ethics index developed by Kaufmann (2004).

3.3.3 Standard control variables

The analysis in this paper concerns birth country and birth-country corruption as possible signals that employers use to make inferences about individual migrants’ trustworthiness and decide how much job autonomy to grant to these individuals. Note, though, that not all individual-level traits that are relevant determinants of how much autonomy a particular employee is granted are, in fact, unobservable. Education, for instance, seems a relevant individual-level trait that employers are able to observe and likely to take into account when deciding how much autonomy to grant to a specific employee. Moreover, employers might be able to observe some individual-level traits that are relevant for job autonomy and also happen to correlate with employees’ birth country or the level of corruption in these employees’ countries of birth. If so, any relationship between individuals’ birth country and/or birth-country corruption uncovered by my analysis would be spurious, in the sense that it does not really testify to any signalling effect of either birth country or birth-country corruption. Again, education would be a prime example, as it seems likely that there is systematic variation in the education level of migrants from different birth countries. To address this possible omitted variable bias, my analysis includes a range of control variables. Importantly, though, these variables are not considered for the purpose of giving a complete account of all the factors predicting job autonomy. Instead, these variables are included as a way of making sure that all relevant factors that can reasonably be expected to be observable for employers are, in fact, considered. If all observable individual-level traits are adequately controlled for, any remaining effect of birth country or birth-country corruption can be validly interpreted as representing a signalling effect that employers use to infer trustworthiness and guide their decision of how much autonomy to grant to a particular employee. Stated differently, the reason that group membership is expected to matter for job autonomy is precisely that there will be some relevant individual-level traits that neither researchers nor employers can observe but that can be proxied on the basis of group membership, specifically ethnonational identity.

That being said, the most basic set of control variables that I include are year/wave dummies, standard demographic characteristics (age, age squared, and sex) and dummies for employment status (whether the individual is currently in paid work, unemployed and looking for work, unemployed and not looking for work, retired, etc.). Inclusion of host-country dummies further controls for direct effects of host-country environment on job autonomy and helps rule out that any relationship between birth country and/or birth-country corruption on the one hand and job autonomy on the other is due to a sorting effect of individuals from specific birth countries (e.g., from relatively corrupt birth countries) migrating to host countries with relatively low levels of job autonomy. In addition, host-country dummies control for any country-specific job autonomy effects due to the presence of certain migrant networks or the general attitude towards foreigners in the host country. At the employer level, I further control for the size of the firm or establishment for which the individual is working using dummy variables concerning the number of employees (10 to 24; 25 to 99; 100 to 499; 500 or more; under 10 employees is the base category). Other factors that I control for are at the level of the individual employee and concern differences in the total number of hours normally worked in a week and differences in educational background, specifically the total number of years of full-time education that an individual has and the highest education level that he/she has achieved as measured by the International Standard Classification of Education (e.g., ‘less than lower secondary’ or ‘higher tertiary’). Given that my sample consists of people that are living abroad as migrants, I further control for time spent in the host country (five years or less; six to 10 years; 11 to 20 years; more than 20 years) and the language spoken at home, specifically whether the language of the host country is also the main language spoken at home or not (1 = yes) (cf. Van Hoorn, 2016b). A final individual-level control variable that I consider is individuals’ income. Following the standard signalling perspective and theories of statistical discrimination, there is a concern that birth country or birth-country corruption do not inform employers about individual migrants’ trustworthiness but about some other unobservable factor that would be taken into account by any employer deciding on how much autonomy to grant to a particular employee. The most prominent such unobserved factor would be an employee’s ability, which, in turn, may affect job autonomy because the specialization and efficiency benefits of delegation and autonomy are greater, the more skilled the employee is. I therefore include income as a proxy for ability, as it is known to/observed by the individual’s employer. If the relationship between birth country and/or birth-country corruption and job autonomy continues to hold with personal income rank controlled for, it seems unlikely that unobserved systematic differences in ability or skills between individuals from different birth countries account for this relationship.⁵ The ESS asks respondents about the total net income of their household and classifies their answer on a scale depicting different income brackets. In its first three waves, the ESS used a 12-point scale, while the later waves used a 10-point scale. To ensure that measured income is comparable over time and across respondents from different countries, I recode answers to create a measure of rank income. Hence, for each respondent, I calculate his/her income percentile relative to respondents from the same country surveyed in the same year/wave. To complete my standard set of control variables, I include three measures of characteristics of the birth-country/host-country dyad to which an individual belongs. These variables are meant to capture a specific relationship that exists between individuals’ birth country and their host country that would affect job autonomy or factors that may affect the likelihood of a migrant from a given birth country to migrate to a given host country (cf. Van Hoorn, 2016b). The first dyadic measure is a dummy variable that captures whether the host country has the same official language as the birth country has (1 = yes). The second measure is a dummy variable that captures whether the host country is a former colonizer of the birth country (1 = yes). The third measure is a dummy variable that captures whether the home and the host country are contiguous (1 = yes). Data for these dyadic measures come from the famous CEPII GeoDist database (Mayer and Zignago, 2011).

3.3.4 Alternative explanations and additional control variables⁶

In addition to the above set of standard control variables, for my robustness checks I consider additional control variables that speak to possible alternative explanations for an effect of birth country and/or birth-country corruption on migrant employees’ job autonomy. The first type of alternative explanations thereby concerns mostly individual-level variables and differences that could have the potential to bias the results of both the analysis of birth-country variation in job autonomy (H1) and the analysis of the effect of birth-country corruption on migrant employees’ job autonomy (H2). The second type, in contrast, concerns only the signalling value of birth-country corruption and the possibility that birth-country features other than corruption are driving observed variation in job autonomy between employees from different birth countries.

Starting with the first type of alternative explanations, there is a concern that the measure of job autonomy is biased because of the particular way in which individuals born (and socialized) in certain countries perceive the world. Specifically, there may be social conventions or cultural norms that cause individuals to exhibit an upward or downward bias when asked to evaluate and score their own lives. To address this contingency, I include self-reported happiness as an additional control variable, as this variable seems most influenced by such differences in response style. If differences in response style, specifically the tendency to be overly negative or overly positive when making a subjective evaluation, truly explain the relationship between birth country and birth-country corruption on the one hand and job autonomy on the other, I expect this relationship to vanish when self-reported happiness is controlled for. If the effects of birth country and birth-country corruption on job autonomy remain with self-reported happiness controlled for, it seems unlikely that the found effects are spurious, driven by systematic differences in response style.⁷ I measure self-reported happiness using the ESS item asking individuals: ‘Taking all things together, how happy would you say you are?’ Respondents can answer on a Likert-type scale that ranges from 0, ‘extremely unhappy’, to 10, ‘extremely happy’.

A second alternative explanation for birth country and birth-country corruption affecting job autonomy involves employees’ preferences for job autonomy. So far, I have grounded my analysis in the idea that employers rely on specific observable signals of trustworthiness when deciding how much job autonomy to grant to an employee, particularly to employees from different birth countries. This perspective is one-sided, however, in the sense that it neglects the possibility that individuals from different birth countries may have different value preferences, meaning that individuals born in certain countries attach much less value to job autonomy than individuals from other birth countries do. If migrants from some birth countries simply care less about job autonomy, the effect of birth country on job autonomy could be spurious, reflecting differences in preferences rather than a process of employers relying on signals of trustworthiness to determine how much autonomy to grant to specific employees. I address the potential of a preference-based bias in the effects of birth country and birth-country corruption on job autonomy in two ways, one involving measures of differences in preferences at the country level and one involving measures of differences in preferences at the individual level. For the former, I identify two culture dimensions in Hofstede’s (2001) seminal framework of national culture that seem likely to affect people’s preference for job autonomy. These two dimensions are individualism and uncertainty avoidance, which can be defined as the degree to which individuals in society are supposed to look after themselves only or remain integrated into groups and the extent to which individuals in society are programmed to feel comfortable or uncomfortable in unstructured situations respectively (Hofstede, 2001, pp. xix–xx). I thereby expect that more individualist societies value job autonomy more and therefore have higher levels of job autonomy and vice versa for uncertainty avoidance. Data for the two culture measures come from Hofstede (2001). For the latter, I consider differences in personal values. Values speak to people’s deepest motivations, the importance they attach to certain objectives compared to other objectives, and provide cross-situational guidance to individuals when selecting between alternative courses of action or states of affairs (Schwartz, 1992). The personal values that I consider derive from the framework of universal human values developed by Shalom Schwartz and collaborators (e.g., Schwartz, 1992), which is the standard values framework in psychology. The Schwartz values framework identifies 10 basic human values that cover the complete spectrum of human motivations. These 10 values are Power, Achievement, Hedonism, Stimulation, Self-direction, Universalism, Benevolence, Tradition, Conformity, and Security. The description of these basic values is that Power refers to social status and prestige and control or dominance over people and resources; Achievement refers to personal success through demonstrating competence according to social standards; Hedonism refers to pleasure and sensuous gratification sought for oneself; Stimulation refers to excitement, novelty, and challenge in life; Self-direction refers to creating, exploring, and independent thought and action-choosing; Universalism refers to understanding, appreciation, and protection of and tolerance for the welfare of all people and nature; Benevolence refers to preservation and/or enhancement of the well-being of people with whom one has frequent personal contact; Tradition refers to respect for, commitment to, and acceptance of the customs and ideas that traditional culture or religion provide the self; Conformity refers to restraint of actions, inclinations, and/or impulses that are likely to upset or harm others and violate social expectations or norms; and Security refers to safety, harmony, and stability of society, of relationships, and of self (Schwartz et al., 2001, p. 521). Details on the ESS items—21 in total—and the procedure used to measure these 10 basic values are presented in Schwartz et al. (2001) and are available on request as well. As an alternative to the robustness checks involving basic human values, I also consider two attitudinal items from the ESS measuring the importance that individuals attach to seeking adventures and having exciting lives and to making their own decisions and being free. These two items are part of the 21 items used to construct the Schwartz values measures, but I consider them separately because they seem to have the most direct bearing on the preference for job autonomy.

A third alternative explanation for an effect of birth country and birth-country corruption on migrant employees’ job autonomy involves migrants’ tendency to work in specific occupations. My analysis focuses on the direct effect of inferred trustworthiness on the amount of autonomy that employers grant to different employees. Alternatively, inferred trustworthiness may influence recruitment, as when only employees that are deemed trustworthy are hired to work in high-autonomy jobs, while employees that are deemed untrustworthy end up working in low-autonomy jobs. If so, birth country and birth-country corruption may affect individuals’ autonomy via inferred trustworthiness, but through the channel of recruitment, which is a different channel than is the focus of this paper. Controlling for average job autonomy in individuals’ occupations allows me to check whether the observed relationship between birth country and job autonomy is perhaps due to such an indirect selection effect or due to the direct effect emphasized in this paper.⁸ Across its seven waves, the ESS has recorded respondents’ occupation using different revisions of the International Standard Classification of Occupations (ISCO). My measure of average job autonomy involves averaging across each four-digit occupation thus included and combining the resulting scores in a single measure of average job autonomy in one’s occupation.

Moving on to alternative explanations for the signalling value of birth-country corruption, I consider selected birth-country features in addition to birth-country corruption as the specific group-level trait underlying the broader effect of group membership (i.e., birth country) on job autonomy. A first specific concern and possible alternative explanation is that birth-country corruption correlates with the quality of education in the birth country, which, in turn, may affect how much autonomy an employer is willing to grant to particular individuals. If so, the corruption-autonomy relationship found in my baseline analysis could be spurious, reflecting the signalling effect of birth-country quality of education rather than of birth-country corruption. However, if the relationship between birth-country corruption and job autonomy remains with quality of education in the birth country controlled for, it seems likely that birth-country corruption indeed acts as an observable signal of (un)trustworthiness that employers rely on when deciding how much autonomy to grant to specific groups of employees. To measure the quality of the educational system in different countries, I use indicators of student performance in the areas of Mathematics, Reading, and Science collected by the Programme for International Student Assessment (PISA) at the OECD. These data are not available for the majority of birth countries in my sample. However, using data collected in the 2012 round of the PISA (OECD, 2013), I am able to retain 59 birth countries. Because country mean scores on Mathematics, Reading, and Science are strongly correlated (r > 0.95), I add the measures of performance in Mathematics, Reading, and Science to construct a single country index of average student performance. For completeness, however, I also present results for models that include separate country scores on Mathematics, Reading, and Science.

Akin to the potential confounding effect of birth-country differences in educational quality, a second alternative explanation for the apparent signalling value of birth-country corruption involves the level of educational attainment in the birth country. Even though my baseline analysis controls for educational differences at the individual level (years of education and education level), host-country employers may base their decision of how much autonomy to grant to an employee from a particular birth country more on the level of educational attainment in the birth country than on the level of corruption. Moreover, to the extent that corruption and country-level educational attainment are correlated, the relationship between birth-country corruption and job autonomy may be spurious. I measure educational attainment using the measure of average years of schooling of the population of age 25 and older available from Barro and Lee (2013). Data are available at five-year intervals, and I take the average for the years 2005 and 2010 to match the 2002–2104 period covered by the ESS data.

A third concern and possible alternative explanation involves the level of economic development in the birth country, which is likely correlated with birth-country corruption and could plausibly act as a valuable signal to employers (cf. Lee and Fiske’s [2006] work on the content of stereotypes). Hence, as a robustness check I control for differences in birth-country GDP per capita. Data come from the World Bank World Development Indicators database (World Bank, 2017), and the measure that I use is calculated as the average GDP per capita during the years 2002–2014.

A final alternative explanation for the apparent signalling value of birth-country corruption involves the potential role of visas and visa restrictions. Dependent on their country of birth, migrants face different regulations concerning employment in different host countries. Migrants from some birth countries do not face any formal barriers in gaining employment in an occupation of their choosing in a given host country. Migrants from other birth countries, however, may have a hard time finding paid employment in these same host countries, assuming they are allowed to work in these countries to begin with. To deal with this issue, I repeat my baseline analysis focusing on migrants that are both born in and currently living in one of the 28 EU member states, as these countries are known for having relatively free movement of people within their borders. Alternatively, I repeat my baseline analysis focusing on migrants that are born in and currently living in countries that are a member of the Schengen area.⁹

4. Results

4.1 Baseline results

Table 1 presents the results of my baseline analysis. Confirming H1, results reveal that a statistically significant amount of total variation in job autonomy occurs between individuals from different birth countries (Model 1). Similarly, results indicate a strong, statistically highly significant negative correlation between birth-country corruption and the amount of job autonomy an individual migrant has, which confirms H2 (Model 3). The empirical evidence therefore strongly supports the idea that country of birth and birth-country corruption are taken as a signal of individual migrants’ trustworthiness, which, in turn, affects how much job autonomy migrant employees are granted in their host countries.

Inclusion of some additional control variables, particularly income, lowers the amount of variation between birth countries (Model 2) and the coefficient for birth-country corruption (Model 4). Both the effect of birth country and of birth-country corruption remain sizeable and highly statistically significant, however. As I have standardized my coefficients, effect sizes can easily be gauged by looking at the estimated coefficients. On this count, birth country and birth-country corruption seem quite important, although some individual-level variables are clearly more important predictors of job autonomy, for instance income. As Models 1 and 2 do not show actual job autonomy differentials between employees from different birth countries, Table A4 in the appendix depicts the 10 birth countries with the lowest and the highest average job autonomy (cf. Table A2).

4.2 Robustness checks

4.2.1 Alternative explanations

Although the models estimated above control for such factors as years of education, hours worked, and income, the main concern with my baseline results is the possibility of an omitted variable bias and alternative explanations of the found effects of birth country and birth-country corruption on job autonomy. Tables A5a and A5b in the appendix present the results for various models that add control variables to rule out alternative explanations for both the effect of birth country (H1) and the effect of birth-country corruption (H2). Although most of the factors identified as possible confounders, notably birth-country culture, personal values, and occupation (Models A2a, A3a, and A6a in Table A5a and Models A2b, A3b, and A6b in Table A5b), are indeed important predictors of job autonomy, both the effect of birth country and of birth-country corruption remain. Hence, even though controlling for, say, happiness might be inappropriate from a theoretical perspective (cf. notes 5–7), my baseline results survive also this very stringent way of assessing the potential spuriousness of and alternative explanations for the found effects of birth country and birth-country corruption on job autonomy. Effect sizes are smaller than before (cf. Table 1), but this is as expected and of course also consistent with theoretical arguments concerning the direct and indirect effects of birth country on, among others, migrants’ happiness (cf. notes 5–7).

Considering alternative explanations for the signalling value of birth-country corruption only, in all cases I find that the estimated coefficient for birth-country corruption remains sizeable and statistically highly significant (Table 2). Hence, there is no indication that the baseline result for the effect of birth-country corruption on job autonomy is spurious. Interestingly, the estimated coefficient for some of the added control variables may be smaller than expected, for instance, for birth-country quality of education. There is a simple, generic explanation for small effect sizes, however, which is that a sizeable effect size requires that employers are able to take into account how employees score on a particular variable, which, in turn, requires that this variable is easily observable by these same employers. The more difficult it is for employees to observe a particular (group-level or individual-level) trait of an employee, the less impact this trait will have on how much autonomy the employee is granted.

Finally, results are similar to my baseline results when estimating models based on samples involving only migrants born in and currently living in EU member states or in countries that are a member of the Schengen area (Table A6 in the appendix). Hence, there is no indication that visa restrictions or similar types of regulations for foreign-born workers are a source of bias in my analysis.

4.2.2 Minimum number of individual observations per birth-country subsample

As mentioned, the birth-country subsamples in my analysis are of differing size and some birth-country subsamples comprise relatively few individual observations. To deal with this issue and address potential biases, I first repeat my baseline analysis using bootstrapping procedures to obtain estimates for my standard errors. Results obtained using bootstrapping are almost identical to my baseline results (details available on request). However, as the main check of the potential sensitivity of my results to the small size of some of the birth-country subsamples, I repeat my baseline analysis excluding birth countries that do not have a certain minimum number of individual observations (cf. Table A2 in the appendix). Scenarios with different minimums (2, 10, or 50; see Table A7 in the appendix) all render results that are almost identical to my baseline results (cf. Table 1). Hence, there is no indication that the baseline results are biased for including relatively small birth-country subsamples.

4.2.3 Alternative measures of main independent variable

For my last robustness check, I test whether my baseline results concerning the effect of birth-country corruption on job autonomy are sensitive to the use of corruption as a group-level trait that employers take into account when deciding how much job autonomy to grant to individuals from specific birth countries. First, I substitute the measure of birth-country corruption used for the baseline analysis with an alternative measure of birth-country corruption based on data covering the period 1996–2000 instead of 2002–2014. Results are largely the same as before (Model A21 in Table A8). Similarly, the relationship between birth-country corruption and migrants’ job autonomy is robust to using the measure of the cultural norm of corruption in the birth country based on Fisman and Miguel’s (2007) country averages of unpaid diplomatic parking violations instead of corruption as a proxy for migrants’ perceived honesty and reliability (Model A22 in Table A8). The size of the estimated coefficient is smaller than before, but this is as expected given that birth-country societal norms are likely much less observable by host-country employers than actual corruption is. Finally, considering other birth-country characteristics as observable signals of individual migrants’ trustworthiness confirms the basic expectation of employers taking into account group-level traits when deciding how much autonomy to grant to different groups of employees (Models A23 and A14 in Table A8). In particular, using measures of birth-country Rule of law or Corporate ethics obtains the same relationship between birth-country image and individuals’ job autonomy, although the sign of the coefficients is of course reversed compared to the measure of corruption used to estimate my earlier models.

5. Discussion and conclusion

This paper has sought to unpack the black box of trust shaping workplace organization. Although the role of trust in providing governance to workplace organization is widely recognized, for practical reasons, in quantitative work this idea has been watered down to testing relationships between a measure of trust as the independent variable and some feature of workplace organization as the dependent variable. Notably, there are several studies finding that stronger societal trust norms increase the amount of autonomy that employers grant to their employees, which is an understandable simplification from a practical perspective. As is, we lack empirical insight on the deeper process underlying employers’ ultimate trust or autonomy decision.

Trust involves not only a willingness to be vulnerable to the actions of the other party but also an assessment of the counterparty’s trustworthiness. For my unpacking of the trust-organization nexus, I have analyzed this latter feature of the decision to trust, using job autonomy as the dependent variable. When it comes to trustworthiness and job autonomy, the key challenge that employers face is to make an informed decision about which employees can be trusted with higher amounts of autonomy and which employees need to be monitored and controlled more closely. Prior work, particularly theories of signalling and statistical discrimination, has argued the power of signals in informing decision makers. Similarly, laboratory studies have found that even simple informational cues can affect trustors’ behaviour in quasi-experimental trust games. Building on these bodies of research, I proposed that employers take into account employees’ membership of particular social groups as a way of inferring trustworthiness and, ultimately, deciding on how much job autonomy to grant to specific individual employees. I tested this proposition empirically in the context of migrants originating from different birth countries and using both the country of birth itself and the degree to which corruption has been institutionalized in these birth countries as an observable signal of individuals’ (un)trustworthiness. Specifically, I considered whether there is systematic variation in job autonomy granted to migrants from different birth countries and whether migrants from birth countries in which corruption is more pervasive, on average, have less job autonomy compared to immigrants from birth countries in which corruption is less pervasive. In a cross-national sample comprising some 10,800 migrants from approximately 170 birth countries, I found strong support for these hypotheses. Extensive robustness checks ruled out alternative explanations for these findings, for instance, systematic birth-country differences in the preference for job autonomy. Overall, this paper contributes important real-world evidence on process-oriented features of trust governing exchange in the context of workplace organization. In addition, the evidence presented in this paper demonstrates how birth country can be an important determinant of how migrant employees are treated and their ability to integrate successfully in their places of work.

These contributions notwithstanding, there are also several limitations to the analysis presented in this paper. First, this paper has not studied individual employers and their actual trust and autonomy decisions. Rather, this paper has focused on outcomes of trust decisions and patterns in the data consistent with a particular process by which employers decide how much autonomy to grant to specific employees. Accordingly, the analysis remains indirect, which, in turn, leaves more room for confounding influences than, for instance, a laboratory experiment would leave. I do not think that confounding influences do, in fact, bias my results, given the various alternative explanations considered and the extent to which my analysis controls for observable individual-level traits likely affecting how much autonomy employers grant to a specific employee. Moreover, the indirect approach has some clear advantages over laboratory experiments that focus on individual decision makers, as the results provide direct evidence on individuals’ real-world experiences and, as such, do not suffer low external and ecological validity. Laboratory experiments could be helpful, however, for probing deeper in the process of inferring trustworthiness, including analyses of the weight that employers assign to different employee signals and group-level traits.

A second limitation is that the social group—i.e., migrants from different birth countries—and the group-level trait empirically analyzed in this paper have been rather narrow. Although I cannot see any reason why the underlying mechanism of relying on group membership and group-level traits to infer trustworthiness would not generalize to other social groups, a logical avenue for future research is to extent the analysis presented here to consider other types of groups in society and other salient group-level traits.

Finally, it has been beyond the scope of the present analysis to link the evidence on group-level traits shaping employers’ decisions of how much autonomy to grant to specific employees to organizational performance. A generic concern with statistical discrimination is that it can lead to suboptimal allocation decisions, since a consequence of considering groups as a whole is that specific qualities of some individual employees remain underappreciated. For job autonomy we expect the same outcome of some individual employees being granted less autonomy than would be optimal in terms of maximizing the net sum of efficiency gains due to specialization minus the costs of shirking. However, future work is needed to assess the actual performance consequences of biased managerial treatment on the count of group-level inferences concerning individual employees’ trustworthiness. This paper, then, provides a stepping stone towards studying this and other important features of trust as a provider of governance in the context of workplace organization.

Supplementary material

Supplementary material is available online at the OUP website. This comprises the online Appendix, the Data, and the replication files.

Footnotes

Acknowledgments

I am grateful for helpful suggestions from the editor, Francis Teal, and two anonymous referees in particular.

References