1 Introduction

Do you remember when you were a young undergraduate student, and you were fascinated by research conducted in the academic field that you studied during your course? Probably this experience has inspired a lot of students, all over the world, to embark on a PhD career. In this regard, between 2013 and 2017, the number of students graduating with a doctorate increased by approximately 8% across Organization for Economic Co-operating and Development (OECD) countries, reaching 276,800 students in 2017 (OECD 2019). The desire for knowledge and institutional support has led to increased participation rates in the PhD production process (Robotham 2008). But in the face of strong competition to enter academia, PhD students soon understand that the research world is not all the fun and games that they had expected. There are many obstacles in their way, starting from strong competition with their colleagues as well, insufficient resources to carry out their research, difficulties in reconciling their family obligations with their work needs - all elements that cause them to wonder ’Do I deserve this position, or am I not qualified to carry out this task?’. When they ask themselves this question, usually they are alone, without adequate support from their supervisor and colleagues, and they can start to experience a sense of anxiety which increases every day and which has a negative effect on their performance at work (Geraniou 2010; Sverdlik 2019). This sad image of the PhD student has been depicted in many reports from all over the world (The Economist 2012; Philips and Heywood-Roos 2015; Schillebeeckx et al. 2013) in which the mental health of non-tenured university researchers is closely examined. For instance, in a pioneering study, (Levecque et al. 2017) found that a sizeable group of PhD students in the Netherlands experienced psychological distress, or were at risk of having or developing a common psychiatric disorder. Most prevalent were feelings of unhappiness and depression, sleeping problems due to worries, inability to overcome difficulties, and inability to enjoy day-to-day activities. Low mental health among non-tenured researchers has a considerable negative impact for research institutions and teams (Lee et al. 2015), as it creates serious financial costs for the institution and has an adverse effect on the efficacy of the larger research teams to which the individual researchers belong (Goh et al. 2015, 2016, see for example). The mental health problems described in the literature affect the well-being of PhD students, and this condition has broader repercussions for their institutions and for academia as a whole (Satinsky et al. 2021). Furthermore, several studies of PhD students’ careers indicate that the dropout levels range from 30 to 50 percent, depending on the scientific discipline and country (Stubb et al. 2012).High turnover among PhD makes the academic world less attractive for new potential candidates threatening the quality of academic research (Lievens and Highhouse 2003). In view of the negative consequences of low psychological well-being among non-tenured researchers, it is important to establish what the causes are, so that effective strategies can be developed to counteract this phenomenon. On the basis of existing literature, we can classify these causes into three interrelated areas: personal, work-related, and relational. Personal factors include individual (i.e psychological/mental) processes that affect academic work; work-related factors include job features that can represent an obstacle (i.e. job demands) or a support (i.e. job resources) at work; relational factors included the relationships network within the work context. In the present research we set out to test a predictive model of anxiety among non-tenured researchers that included both work-related and relational factors, intended to offer to the academic industry effective policy guidance in order to deal appropriately with this issue. We decided to concentrate on these two kinds of factors because with adequate policy collective action they can be improved. For relationship factors we considered two elements that were already well-documented in previous literature on the subject of well-being in academia (i.e. relationship with one’s supervisor and relationship with one’s colleagues). As work-related factors we examined two features that previous literature has largely overlooked, namely perception of fairness and career development. We considered that in an academic context both the perception of fairness and career development can play a central role as job resources. Perception of fairness can be defined as an unbiased and equitable treatment in the organisation, for example, in relation to the allocation and distribution of resources, and to the decision-making processes or information sharing (Colquitt and Zipay 2015). Perception of equality at work can make a significant contribution to employees’ well-being (Masagão and Ferreira 2015). Career and professional development relates to how accessible the organisation makes its criteria and the opportunities on offer within it for growth. A number of studies have demonstrated that employees present lower levels of work-related stress when career opportunities and criteria are made clearly evident (Wilson et al. 2004; Xanthopoulou et al. 2007). In such a context as academia, where merit is prioritised, the perception of equality and correctness at work and the recognition that academia facilitates professional growth can be of great benefit to researchers’ mental health. In addition, in view of the spread of COVID 19 across the world, we have examined the effect on anxiety levels of employees’ perception of safety at work. We found that after the first phase of the pandemic and the imposition of lockdown (i.e. confinement to home) as an effective strategy to contain and counteract the spread of the virus, many laboratories all over the world rapidly became operative again. The perception of safety at work among researchers who were obliged to work on site in spite of the spread of COVID is a subject that merits investigation. As stated above, PhD students are more likely than the rest of the population to develop common emotional disorders such as anxiety and depression. For this reason, several studies published in recent years give warnings about the mental health of PhD students. Physiological anxiety is the emotion that one feels in the face of a real or imagined threat, and has the aim of preparing us to confront it: it is characterised by a state of psychological and physical tension, sometimes almost terror (Steimer 2002). Anxiety, on the other hand, is pathological when it significantly disturbs one’s psychological functioning, bringing about a limitation of the individual’s ability to adapt: it is characterised by a state of uncertainty about the future, with the prevalence of unpleasant feelings (Walker et al. 1990). Anxiety can affect the quality of one’s sleep, one’s eating habits, one’s general behaviour, one’s work, social functioning, and well-being. We have chosen to investigate the presence of anxiety traits among PhD students attributable to the effect that anxiety can have on people’s work and their social functioning. In our case, the consequence is a negative effect on academic performance. To sum up: in the context of policy development to improve the quality and efficacy of research, the present study set out to test the effect of job-related factors (i.e. fairness perception, professional growth, and safety perception) and relational factors (i.e. relationships with supervisors and colleagues) on the anxiety level of non-tenured researchers. In the present research, we tried to take a step further in investigating the mental health of non-tenured researchers, by combining both ad hoc Likert questionnaires and open-ended questions. We believed, in fact, that hearing directly from researchers could represent a valuable enrichment in investigating their mental health, while also giving us an opportunity to investigate problematic areas that tend to be underestimated by current researchers in the field. These findings are then discussed to reveal their implications for policy makers.

2 Background and related works

Although they are not the focus of the present research, a wide literature has investigated personal factors having a greater impact on precarious researchers’ mental health, focusing in particular on motivation and self-efficacy. Motivation to succeed is a salient predictor of achievement and persistence in doctoral studies (e.g. Bargar and Duncan 1982; Brown and Watson 2010; Hegarty 2011; Onwuegbuzie et al. 2014). For instance, McGee et al. (2016) showed that motivation predicted the intention of engineering in continuing their PhD studies, despite the daily difficulties they can encounter. Moving to self-efficacy refers to an individual’s perceived belief in having the resources necessary to achieve desired goals (Schunk and Pajares 2009). As applied to the academic context, doctoral self-efficacy can be defined as the individuals’ confidence in performing tasks related to conducting efficient research (Forester et al. 2004). Research self-efficacy significantly predicts interest in research and the production of publications (Lambie and Vaccaro 2011), and lower levels of drop-out intentions (Litalien and Guay 2015). Doctoral students with lower levels of self-efficacy can engage in self-handicapping behaviors (i.e., sabotaging their chance of success (Jones and Berglas 1978) to avoid being perceived (or perceiving themselves) as incompetent (Schwinger and Stiensmeier 2011)). One of the most frequent self-handicapping behavior among PhD students is to find an excuse (e.g., a few times) to avoid participation in competitive grants. Moving to the work-related factors, in line with the job demands-resources theory (Bakker and Demerouti 2007, 2017, JD-R theory) we can find job demands and job resources. Job demands are “physical, psychological, social, or organizational aspects of the job that require sustained physical and/or psychological (cognitive and emotional) effort or skills and are therefore associated with certain physiological and/or psychological costs” (Bakker and Demerouti 2007). Job demands are not negative by definition; they become job stressors when meeting those demands requires a high effort that the person cannot adequately cope with. Job resources instead are “the physical, psychological, social, or organizational aspects of the job that are functional in achieving work goals; reduce job demands and the associated physiological and psychological costs; stimulate personal growth, learning, and development” (Bakker and Demerouti 2007). In the present research, we focused on three potential job resources in academia: fairness, professional growth, and safety perception at work. A large body of occupational research has shown that perception of fairness at work includes both a comparison with objective measures (e.g., comparable rates of pay and roles with the external job market and among coworkers with similar skills and experience) and subjective considerations (e.g., feeling that one’s role and pay reflect the value of one’s contribution in the workplace, (Cohen and Diamant 2019; Jawahar and Stone 2011; Till and Karren 2011)). Individuals who perceive more fairness at work are more likely to be satisfied at work, less stressed and less likely to look for alternative employment options (Erdogan et al. 2012). In the organizational context, great attention has also been devoted to career and professional growth which refers to the extent to which the organization clarifies and makes available criteria and opportunities for growth (Xanthopoulou et al. 2007). Evidence has shown that when workers perceive adequate opportunities for professional development, they feel less stressed at work with low levels of burnout (Bakker et al. 2003; Sashkin 1982; Lee RT 1996). Spreading of COVID19 has increased the attention to the importance of safety perception at work. During the last decades, governments of many countries across the world (e.g., Italy, Australia, United Kingdom) have established safety training as a legal requirement for organizations. Workers informed and trained about the risks at work can put in place all the strategies finalized to make their job safer. Crucially, a recent study conducted in Brasil by da Cunha et al. (2015) has shown that safety training of workers in food area was associated with less stress at work, confirming the importance for organizations to invest in training the workers about safety policy. Among relationship factors, low support at work from both supervisors and colleagues has long been found to affect levels of anxiety of PhD (De Lange et al. 2004; Vanroelen et al. 2008, e.g.). In a Finnish study of 383 PhD students, Stubb et al. (2011) found that 56% reported the academic community as a source of burden. According to Latona and Browne (2001), PhD would desire a relationship with their supervisor characterized by precise and timely feedback, frequent meetings that include open discussion about roles and responsibilities, a supportive and collegial relationship, and encouragement to begin working on topics of PhD interest since the beginning of their academic pathway. Numerous empirical studies have found the good fit between supervisor and supervisee predicts positive doctoral students’ emotions (Chiang 2003; Cotterall 2013; McAlpine and McKinnon 2013) and academic persistence (Ives and Rowley 2005; Leijen et al. 2016; Litalien and Guay 2015; Pauley et al. 1999) (see Sverdlik et al. (2018) for a review). The relationship with colleagues holds a critical position in academia comparable to that of supervisors. PhD students undergo a process of socialization by assuming the roles of both graduate students and professionals as they become members of the academic community (Golde 1998). Collaboration among researchers is deemed essential in a professional education model, as it optimizes the socialization of doctoral students (Bourner et al. 2001; Gardner 2010; Maxwell 2003; Usher 2002). According to Austin (2009), a successful academic model encompasses social support from colleagues, frequent and honest feedback, and an exchange of ideas that is both democratic and constructive (Shacham and Od-Cohen 2009).


Hypotheses


Basing on the above-mentioned literature, we tested the following hypotheses:

  • H1: Fairness perception (F) has a negative effect on Anxiety (AN).

  • H2: Professional and personal Growth (G) has a negative effect on Anxiety (AN).

  • H3: Safety perception at work (S) has a negative effect on Anxiety (AN).

  • H4: Positive Relationship with Colleagues(CO) has a negative effect on Anxiety (AN).

  • H5: Positive Relationship with Supervisors (SU) has a negative effect on Anxiety (AN).

Moreover, given that previous studies have shown a higher incidence of anxiety among females compared to males (Costa et al. 2001; Egloff and Schmukle 2004; Feingold 1994), we investigated whether analogous gender disparities existed in an academic setting (H6)

3 Material and methods

3.1 Participants

699 participants (398 females, 301 males, mean age = 29.60 years, SD = 4.90; range: 23-65 years) completed an online questionnaire. The survey was addressed to both PhD and researchers with a temporary contract with Italian Universities, given that both live in a non-tenured situation. Participants were recruited on a voluntary basis through social networks and institutional emails of Italian universities. We were able to have respondents from all Italian regions, by granting a representation of entire Italy. Before participating in the survey, all participants gave informed consent. No compensation was provided for participating in the study.

3.2 Procedure

The first section of the questionnaire aimed to assess demographic characteristics (i.e., gender, academic position). Then, participants responded to an open-ended question “If you could change something in the world of academic research, what’s the first thing you would change?”. From this question we expected respondents to express the most critical elements of the academic career. Open questions are useful to catch the opinion of respondents, as they can answer touching topics they care most about, and let them describe their experience. After the open-ended question, we used ad-hoc scales to measure safety perception at work, fairness perception, professional growth, relationship with supervisors and with colleagues. In the last section of the questionnaire, we measured the participants’ anxiety with State Trait Anxiety Inventory (Speilberger et al. 1983). Safety perception at work, professional growth, fairness perception, relationship with supervisor and with colleagues were assessed through the adaptation of the Italian Work Well-being questionnaire developed by the National Authority Against Corruption (ANAC 2013). Finally, participants were debriefed.

3.3 Measures

Safety perception at work was measured with 8 items adapted from the Italian Work Well-being questionnaire (ANAC 2013). Participants rated the extent to which they agreed with items such as “I have received appropriate information and training on the risks associated with my work activity and on prevention and protection measures”. Participants responded to these statements on a five-point scale from 1 (totally disagree) to 5 (totally agree). Fairness perception was measured with 3 items adapted from the Italian Work Well-being questionnaire (ANAC 2013). Participants rated the extent to which they agreed with items such as “I believe that the workload is distributed equally among the members of my working group”. Participants responded to these statements on a five-point scale from 1 (totally disagree) to 5 (totally agree). Personal and professional growth was measured with 4 items adapted from the Italian Work Well-being questionnaire (ANAC 2013). Participants rated the extent to which they agreed with items such as “My organization stimulates the development of skills and abilities”. Participants responded to these statements on a five-point scale from 1 (totally disagree) to 5 (totally agree). Relationships with supervisors and colleagues were measured with 13 items adapted from the Italian Work Well-being questionnaire (ANAC 2013), 8 items concerned about the supervisor and 5 items concerning the colleagues. Participants rated the extent to which they agreed with items such as “My supervisor is attentive to my needs”, “I feel part of a team”. Participants responded to these statements on a five-point scale from 1 (totally disagree) to 5 (totally agree). Anxiety was measured with the 20 items of the State-Trait Anxiety Inventory (Speilberger et al. 1983). The STAI-T was designed to measure a stable propensity to experience anxiety, and tendencies to perceive stressful situations as threatening. The trait scale consists of 20 statements (e.g. “I feel tense") that require individuals to rate how they generally feel on a four-point scale from 1 = never to 4 = always.

4 Results

4.1 Descriptive statistics

Figure 1 shows the boxplots of the items of questionnaire’s scales, from which it is possible to get interesting insights of the sample of respondents. It is interesting how among our predictors we observed that the median scores of fairness perception were around 3 on a 5-point Likert, with the median of the item “I think that my work-load is adequate to my retribution” that was very low (value = 2), by indicating how this topic is particularly critical in academia. Furthermore, from the boxplot emerged that participants were enough satisfied with growth and relationship with both supervisor and colleagues, given that for all these dimensions the median score was 4 on a 5-point Likert-scale. Among the predictors, the dimensions that showed the highest score was the safety perception at work (the items with a median score of 2 and 1 are reverse items) Crucially, the median scores of anxiety items were high and worrying, approximately 3 for all items on a 4-point scale (the item with a median below 3 are reverse items), indicating the presence of anxiety among our sample.

Fig. 1
figure 1

Boxplots of the dataset of items by the 5 scales used in the questionnaire

4.2 Structural equation models

The concept of Structural Equation Modeling (SEM) is a linear model framework that is designed to model regression equations with latent variables. One of the distinguishing features of SEM is its ability to encompass both measurement and structural models. Specifically, the measurement model establishes a connection between observed and latent variables, while the structural model examines the relationships between endogenous variables (which can be either latent or observed) and exogenous variables (also either latent or observed). Regression-based models are comprised of a single equation that is designed to elucidate one endogenous (dependent) variable, or alternatively, a multiequation model that examines a range of endogenous variables and their reciprocal relationships (Bollen 1989).

In our specification, we assume that there is only one endogenous latent variable (namely, anxiety) and eight exogenous latent variables. Therefore the measurement models for student \(i=1,\dots ,N\), can be written as

$$\begin{aligned} Z_{il}^{(x)}= & {} \varvec{\lambda }_{l}^{(x)'}\varvec{\theta }_i+\varepsilon _i, \quad \text {with} \quad X_{il}=c \quad \text {if } \gamma _ {l,c-1}^{(x)}\le Z_{il}^{(x)} \le \gamma _ {l,c}^{(x)} \\ Z_{ik}^{(y)}= & {} \lambda _{k}^{(y)}\eta _i+\epsilon _i, \quad \text {with} \quad Y_{ik}=c \quad \text {if } \gamma _ {k,c-1}^{(y)}\le Z_{ik}^{(y)} \le \gamma _ {k,c}^{(y)}] \end{aligned}$$

Here \(X_{l}\), for \(l=1,\dots ,L\), represent the observed categorical variables measuring the exogenous latent variables \(\theta _q\) for \(q=1,\dots ,Q\); \(Y_{k}\), for \(k=1,\dots ,K\), represent the observed categorical variables measuring the endogenous latent variable \(\eta\); \(Z_{il}^{(x)}\) and \(Z_{il}^{(y)}\) are underlying variables linked to the observed indicators through a threshold model; \(\varvec{\lambda }_l^{(x)}\) and \(\lambda ^{(y)}_k\) are the discrimination parameters; \(\varvec{\gamma }^{(x)}_l\) and \(\varvec{\gamma }^{(y)}_k\) are the vectors of the threshold parameters; finally, \(\varepsilon _i\) and \(\epsilon _i\) are normally distributed errors.

The structural equation model can be written as:

$$\begin{aligned} \eta _i = \varvec{\beta }'\varvec{\theta }_i + {\varvec{g}}'{\varvec{w}}_i + u_i \end{aligned}$$

where \(\varvec{\beta }\) and \({\varvec{g}}\) are vectors of regression coefficients, \({\varvec{w}}_i\) is the vector of observed variables and \(u_i\) is a normally distributed error.

The structural component is represented in the path diagram provided in Figure 2.

Fig. 2
figure 2

Path diagram for the structural component in the proposed SEM

4.3 Model summary of structural equation models (SEM)

To estimate the model, it was used the laavan library (Rosseel 2012) in the language and environment for statistical computing R (R Core Team 2022). The evaluation was performed through the diagonally weighted least squares DWLS estimator. The number of model parameters is 226, while the number of observations is 699. A model for ordinal data has been fitted and the variables have been assumed to be orthogonal. Of course, the fit of the model would improve by eliminating this assumption. The good of fit indices used are RMSEA, CFI and TLI. In particular, RMSEA is an absolute fit index, in that it assesses how far a hypothesized model is from a perfect model, while CFI and TLI are incremental fit indices that compare the fit of a hypothesized model with that of a baseline model, that is a model with the worst fit (Xia and Yang 2019). In this case, RMSEA is equal to 0.235, CFI and TLI are equal to 0.841 (Table 1).

For the purpose of this study, SEM technique was used to examine the relation between the variables to determine the acceptance of the research hypotheses. First of all, it is important to state that all the items considered in the model explained the respective variables (see Tables 2, 3, 4, 5, 6). This study is composed of five independent variables, which are safety, growth, relationship, fairness and gender. In total, there are five hypotheses made to study the relationship of those independent variables with anxiety. It is shown in Table 1 that all hypothesis are supported.

Table 1 Results of the SEM Regression for Anxiety
Table 2 Measurement model for the latent variable Anxiety: factor loading estimates. The question prompt the researchers to indicate how they generally feel
Table 3 Measurement model for the latent variable Safety: factor loading estimates. The question prompt the respondents to indicate their agreement with the items
Table 4 Measurement model for the latent variable Growth: factor loading estimates. The question prompt the respondents to indicate their agreement with the items
Table 5 Measurement model for the latent variable Relationship: factor loading estimates. The question prompt the respondents to indicate their agreement with the items
Table 6 Measurement model for the latent variable Anxiety: factor loading estimates. The question prompt the respondents to indicate their agreement with the items

4.4 Textual analysis

Textual questions are interesting for leaving respondents free to express their opinion, but are more challenging than traditional structured questions, as they cannot be directly processed with quantitative methods. The analysis of these questions involved the usage of Text Mining Techniques. Text Mining is a toolset of techniques that allows to extract insight from textual data. Among the several techniques available, we used the ones belonging to the Bag of Words approach. In the specific case the single answers are considered as a bag of words, without order. The loss of the order and therefore the direct capability to extract the semantic meaning, enables the application of quantitative methods able to extract helpful insights from the processed data. Figure 3, is a preliminary analysis that shows the most frequently used words in the dataset of answers to the question about what the respondents would change in the academic world. The figure shows how “work”, “funds”, “phd”, “system”, “researchers”, “career” and “publications” (translated in English) are among the most used words. This figure provides insights into what many of the answers consider as the critical flaws of the Italian academic system: the current state of the academic career and resources for funding research.

Fig. 3
figure 3

Top 20 term frequency for answers to question “If you could change something in the world of academic research, what’s the first thing you would change?”, excluding the word “ricerca” (research)

On the same question, we performed further analysis, to extract what were the main discussed topics in the questions. To perform this analysis we performed a Latent Dirichlet Allocation (LDA) (Blei et al. 2003)). This methodology considers each document in the dataset as a mixture of topics. In particular, provided the number of k topics to be extracted, the method calculates the probability of each topic to be present in the dataset, the probability of each topic in each document, and the probability of each word to occur in each topic. After several trials we determined the best number of topics to be extracted is 4, as the results provided the best interpretability. In particular in this research we used the LDA implementation of library quanteda (Benoit et al. 2018) for R. Figure 4, shows the four wordcloud according to the 4 topics extracted through the LDA. Wordcloud is a useful data visualization tool in text mining to show results from textual data. The cloud of words indicates the several words present in the set, and their different size indicates the relevance of the word in the set according to some criteria. In the figure the wordclouds show the top 30 terms for the topic: bigger is their size, higher is the probability of that word to occur in the document for that topic. The results in the figure show some interesting insight of the most discussed topics by the respondents to our research.The following interpretation of topics have been integrated with a manual examination of comments presenting the topics terms:

  • Topic 1: This topic has as most relevant terms: “researchers”, “research”, “funds”, “phd”, “projects”. This topic focuses on the issue of funds for research and expresses the request of receiving more fundings to perform research and carry on researchers activities.

  • Topic 2: This topic has as most relevant terms: “system”, “publications”, “change”, “evaluation”, “quality”, “work”, “bureaucracy”, “criteria”. This topic focuses on the desire of researchers to change the “evaluation system” of the research and publication activities in the Italian academic system.

  • Topic 3: This topic has as most relevant terms: “work”, “contracts”, “research scholarship”, “wages”, “phd”. The relevant words of this topic indicate it focuses on the economic and contractual situations of currently employed researchers in the Italian academic system.

  • Topic 4: This topic has as most relevant terms: “research”, “world”, “career”, “research”, “positions”, “non-tenured”. The relevant terms of this topic clearly indicates how this focuses on the difficulties to access the academic career and the characterization of the Italian academic system in the early stage of career of non-tenured work.

Fig. 4
figure 4

Top 20 term frequency for answers to question “If you could change something in the world of academic research, what’s the first thing you would change?”, excluding the word “ricerca” (research)

The results provided from the LDA and shown in Figure 4, provided helpful information about what are the expectations and the most relevant flaws of the academic system according to the participants of the questionnaire. The topic mainly discussed by the participants, touched on the most critical aspects of the Italian academic system, such as the lack of funding to carry out research activities, the low wages and precariousness of PhD students and researchers in the early stage of career, and the evaluation system of researchers’ activities.

5 Discussion

The present research aimed to test a predictive model of anxiety of Italian non-tenured researchers, through both closed-ended questionnaires and open-ended questions. First, our H1 that fairness perception has a negative effect on anxiety was supported. To the best of our knowledge, this is the first time that fairness perception is recognized as a predictor of anxiety among non-tenured researchers. This result is not surprising, given that studies in other organizational contexts have indicated that individuals who perceive more fairness at work are more likely to be satisfied at work, less stressed and less likely to dropout (Erdogan et al. 2012). The importance of fairness perception has been confirmed also by open-ended question, in fact; one topic was related to the evaluation system of the research and publication activities in the Italian academic system and another topic was about the difficulties to access the academic career and the characterization of the italian academic system in the early stage of career of precarious work. Said in a nutshell, Italian non-tenured researchers seem available to make sacrifices at work, but they ask for guarantees about a meritocracy system with transparent policy about the start and the development of the career in next stages. In this regard, we can not affirm that in Italy the system is not meritocratic, but if the researchers that are the main actor of the system ask for a change in the evaluation system, the accademia need to work on an efficient communication system about the policies of evaluation in Italy to make them more clear and transparent. Of course, we know that is not easy for academia to find an objective model of evaluation that can overcome the publish or perish model that characterizes academia around the world, a system that risk to rewards the “significant result” finding and not the search for knowledge that should represent the drive of academia. In this regard, are surely appreciated the creation of Movements about the replicability and the transparency of science around the world. But these movements risk being ineffective, without a cultural change that involves the European institutions to find a system that awards good research and not only significant research, with implications for fairness perception. Second, another job resource that has had a negative effect on anxiety has been the perception of professional growth, confirming our H2. This result is particularly relevant in accademia, where learning and professional growth represent the business core of work. Some topics that emerged in the open-ended question were related to professional growth: the problem of lack of funds for research, but also the problem of the economic and contractual situations of currently employed researchers in the Italian academic system. These aspects are, in fact, both essential in accademia to grant researchers at an early stage the opportunity to perform their research without worrying about the necessity to find everyday funding and economic stability. It appears clear that the precariousness represents a sad plague of academia in Italy, with PhD and post-doc contracts that are temporary contact. This economic situation determines a negative spiral, where the non-tenured researchers are paid to accomplish their research but at the same time, they need to invest time and resources in finding new job opportunities given the absence of stability with the consequences that non-tenured researchers cannot dedicate completely on the research project they are working on, until they are not employed with a permanent contract. These aspects are in line with previous literature showing that a greater access to funding has further been found to correspond with higher levels of students’ overall satisfaction with their doctoral experience and lower attrition (Ali and Kohun 2006; De Valero 2001; Gururaj et al. 2010; Leijen et al. 2016). Third, results confirmed our H3 that safety perception at work had a negative effect on anxiety. During the pandemic, the Italian government has given many guidelines to the students of primary and secondary school, with little attention devoted to the academic world, given that researchers and students of the university were considered more autonomous. Whether on one side this situation could be partially true, on the other side it has created a level of anxiety in researchers that were called to work in an emergency situation. Safety perception at work will be important also outside the pandemy, especially in some academic sectors considered more dangerous (e.g., research sector about biochemical risk). It is important for the department to communicate well about the risks and safety at work in order to make the non-tenured researchers feel they are not alone in dealing with dangers at work. Fourth, our results confirmed our H4-5 that good relationships with both supervisors and colleagues had a negative effect on anxiety, by supporting the importance of the relationships as a crucial factor in promoting the mental well-being of non-tenured researchers. Relationships with supervisors and colleagues were highly interrelated, confirming the importance to develop an efficient model of socialization in accademia that involve all stakeholders (Golde 1998; Weidman et al. 2001). In this regard it is important to put in place effective academic policies by promoting attention to personal needs by supervisors, but also a collaborative system where human resources can support each other in a positive model of socialization (Stubb et al. 2011). Making in field these strategies of socialization is difficult for accademia, in fact, from one side researchers are highly invited to collaborate each other in particular to obtain national and European fundings that promote collaboration between different labs around the world, but at the same time, the competitive context of accademia can represent an obstacle to this collaboration. A supportive model where the other is seen as a resource and not an enemy represents a challenging aim for accademia in Italy, and only a cultural change can help to promote a collaborative system within the own lab and among different labs. Investing in a collaborative system is essential to reduce anxiety among researchers and to promote their mental well-being with important practical implications for accademia as a whole. In this regard, Stubb et al. (2011)found that among 669 Finnish doctoral students, those who perceived their scholarly community as integrative, empowering, and inspiring reported better overall well-being (e.g., lower anxiety, exhaustion, isolation) as well as greater interest in their studies and better engagement with their department. An ancillary result of the present research regarded the gender difference in anxiety, with higher levels of anxiety among female non-tenured researchers than male non-tenured researchers (H6). This result is in line with literature showing a greater level of anxiety among females rather than males (Costa et al. 2001; Feingold 1994), especially where anxiety is assessed with explicit measures as in the case of the present research (Egloff and Schmukle 2004). These gender differences deserve great attention from institutions in order to promote gender equality in academic research. We have not explored the reasons for these gender differences, we can speculate that it can be due to the difficulties to combine work needs with family needs, but also the difficulties to be recognized for females as scientists, a problem that mirrors gender stereotypes that are still present nowadays. Future studies are needed to investigate whether there are gender differences affecting academic well-being, considering that anxious traits are greater in females than in males. We would like to underline, however, that levels of anxiety are high also among males, by indicating that the anxiety goes beyond gender. It is possible that with implicit measures of anxiety, gender differences would be less accentuated. Future studies could verify this possibility.

6 Conclusion

Italian non-tenured researchers live stressful situations with high levels of anxiety. The combination of standard questionnaires and open-ended questions in our study provides useful insights for decision making institutions to arrange policies in order to reduce anxiety at work. The key policy implication of our findings is to put the researchers well being as a goal on the agenda of academic policies. From our research emerged the need to create a healthier workplace by working on the creation of positive and collaborative relationships with supervisors and colleagues, by communicating in a transparent way about the procedures of access and growth in academia, and by investing on safety perception but also on funding and salary. These policies are important for academic world given that high levels of anxiety among researchers can increase the degree of turnovers with an economic impact for research institutions and the functioning of the research team (see e.g. Lee et al. 2015; Goh et al. 2015, 2016). We wish to acknowledge the potential limitations of our results. The first pertains to the interpretation of our study findings. The present cross-sectional dataset is limited in its ability to establish causality. An alternative interpretation of the findings could suggest that non-tenured researchers who experience mental health issues may be more inclined to report negative evaluations of their environmental conditions. Nevertheless, previous research within the domain of occupational health has identified a considerable number of studies that examine causal links between organizational factors and the onset of mental health problems. This body of literature provides support for the notion that the work environment may be partially responsible for the prevalence of mental health problems observed in the current study. Furthermore, the study is subject to limitations related to possible respondent auto-selection processes. Although the questionnaire was sent via institutional email, it is conceivable that non-tenured researchers who are living in problematic situations may be more motivated to complete the questionnaire. However, the large number of respondents and the concise cover story regarding the aim of the research help to mitigate this risk and give us confidence in the robustness of the results. Finally, another limitation is that we have not considered a control group. However, in the present research, we were not interested in having a comparison among non-tenured researchers and other professions, rather we were interested in testing a predictive model of anxiety which can allow identifying critical areas of intervention for institutions and university stakeholders. Our findings are in line with previous studies in the literature. future studies may consider a gender comparison, as mentioned above, and a confrontation with a group of subjects unrelated to the academic world.