Eurobubble at work: wellbeing and (dis)satisfaction - survey results
In December 2024, we ran a survey across the Eurobubble.
We explore job satisfaction and well-being at work across more than 40 different parameters.
Nearly 700 people answered.
As a volunteer side project, we never really had time to crunch the data.
Almost 2 years later, we put the data in Claude (Opus 5 - high mode) because we still don’t have the time, nor, frankly, the competencies to produce a full quanti and quali data analysis out of the blue.
The Eurobubble
at Work
A survey of work wellbeing and job satisfaction across Brussels' European affairs community. People here find their work meaningful and their colleagues good. What they do not find is management, recovery time, or a reason to stay in the job they have.
Who answered, and who stopped answering
684 people opened the survey. 683 answered the first question. 501 reached the last rating item, and 472 completed the background block. The shape of that decline is itself a finding.
Response counts fall almost monotonically down the questionnaire — this is a long instrument (42 rating items plus 15 background questions) and people left steadily rather than in a single cliff. Two places show sharper steps: around the mental-health and HR support items early on, and again at the demographic block, where salary (n = 323) and nationality (n = 255) are the least-answered questions in the whole survey.
The important question is not how many left but who. Comparing people who reached the satisfaction item against those who dropped out, on the early questions they both answered, the drop-outs were consistently slightly more positive — significantly so on manager support (6.12 vs 5.23, Cohen's d = 0.32, p = 0.001), and in the same direction on HR support and mental-health access. Nobody dropped out because they were too happy to continue; but the people who stayed to the end were, on average, the more disaffected ones.
Attrition here is not random with respect to the outcome. Because the more supported respondents disproportionately left, the completed-response sample tilts negative. On top of that, this is a self-selected, self-administered survey — people with something to say about their workplace are the ones who answer a workplace wellbeing survey. Treat the direction and structure of the findings as robust, and the absolute levels as an upper bound on how bad things are, not a population estimate.
Composition of the sample
The respondents are recognisably the Brussels EU-affairs workforce, skewed the way that world's under-40 cohort skews: predominantly women, predominantly in their late twenties and thirties, drawn from 63 nationalities, and overwhelmingly working from Brussels itself (399 of 435 who answered). Civil society and NGOs are the largest single sector, followed by consultancies and the EU institutions.
The item ladder: what is good and what is broken
All 42 items, rescaled so that a high score always means a good outcome. The pattern is unusually consistent: the things people bring to work are strong, and the things the organisation is supposed to provide are weak.
Six items were negatively worded (loneliness, stress, harm to physical health, rumination after hours, bullying, and intent to leave). These are reverse-scored throughout this report so that, everywhere, higher is better. A reverse-scored item is marked R in the chart below.
Read from the bottom of the ladder up. The weakest scores in the entire survey are not about pay or workload in the abstract; they are about infrastructure and recovery. Half the sample (50.4%) scores 3 or below on having appropriate access to mental health support through their job. Almost exactly as many (49.6%) score 3 or below on being able to talk openly to HR. 42.1% score 3 or below on opportunities for growth or promotion. Over half (52%) cannot stop thinking about work outside working hours.
At the top of the ladder sit three things employers did not have to build: colleagues are good (7.50), relations are cooperative rather than competitive (7.11), and the job is not physically hard (8.47 — the highest score in the survey, and largely an artefact of desk work). Pride in the job (6.88) and autonomy (6.91) are the strongest genuinely organisational items.
Bullying and harassment
The bullying item is worth isolating because it was asked as an intensity rating rather than a yes/no, which makes prevalence claims delicate. 38.9% of the 519 people who answered chose the absolute floor of 1 — the only unambiguous reading of "this has not happened to me." Everyone else placed themselves somewhere above it. 34.9% scored 5 or higher; 22.5% scored 7 or higher.
Whatever threshold you prefer, roughly a fifth to a third of this workforce is reporting a substantial experience of bullying, harassment or aggression in their current job. The rate is higher for women (3.62 vs 3.01, d = 0.22, p = 0.027) and highest of all among respondents who declined to state their gender (5.00, n = 26) — a small group, but one whose combination of non-disclosure and high scores is itself worth noting.
The survey measures four things, not forty-two
Exploratory factor analysis of the 37 non-outcome items recovers a clean four-factor structure that explains 51.5% of common variance. Each factor forms a reliable scale, and the four together behave like a usable instrument.
The correlation matrix is emphatically factorable: Kaiser–Meyer–Olkin = 0.93, Bartlett's test of sphericity χ² = 11,564 (p < 0.001), n = 492 complete cases. Horn's parallel analysis against 500 permuted datasets retains four factors — the fifth observed eigenvalue (1.33) falls below its 95th-percentile null (1.39). Principal axis factoring with varimax rotation gives the structure below.
The four dimensions are substantively distinct and internally consistent:
Three points about this structure matter for how the survey should be read and how the next wave should be designed.
First, the four factors are not equally healthy. Meaning & development is the strongest (6.48) and Workload & recovery the weakest (5.33). Within the same person, the gap between what the job means and what it costs runs at +1.12 scale points in favour of meaning (paired t = 11.6, p < 0.001, n = 529), and 68% of respondents show a positive "meaning premium" over the average of their other three dimensions. That gap is the structural signature of this labour market.
Second, bullying loads on the recognition factor, not on a factor of its own. Its highest loading (+0.57) is with feeling valued, manager support, role clarity and cooperative relations. Statistically, in this dataset, harassment is not a freestanding pathology — it lives in the same latent dimension as ordinary managerial neglect. That is a substantive claim about the phenomenon, and it is testable: a survey with more harassment items should be able to confirm or break it.
Third, four items fail to load on any factor at all (no loading ≥ 0.40): workplace conditions, access to mental health support, the work-from-home policy, and "my tasks are not physically hard." The last of these has a communality of 0.13 — it measures desk work, not wellbeing, and should be cut or reframed. The most interesting of the four is mental health support: it is the second-weakest item in the entire survey, and it is statistically independent of everything else about the job. Whether an organisation provides it appears to be unrelated to whether that organisation is any good in other respects. Loneliness is a borderline case — it loads on the recognition factor at 0.50 but has a communality of only 0.27, meaning most of what it measures is not shared with the rest of the instrument, despite recurring often in the free text.
What actually drives satisfaction — and what drives leaving
The four dimensions explain 71% of the variance in overall job satisfaction. Recognition dominates. Money is, by some distance, the least important of the four — until you switch the outcome from satisfaction to quitting.
Regressing standardised overall satisfaction on the four scale scores gives R² = 0.713 (adjusted 0.711, n = 518), with no multicollinearity problem (all VIF ≤ 2.9). Because the predictors are correlated, the raw coefficients understate shared contribution; the chart below therefore decomposes R² by exact Shapley values (the LMG method), which apportions the explained variance across all 24 orderings of the four predictors.
Material rewards contribute 10% of the explained variance and their standardised coefficient is not significant (β = 0.053, p = 0.059) once recognition, meaning and workload are held constant. This is the single most counter-intuitive result in the dataset, and it is not a fluke of the scale construction: at the item level, a LASSO regression across all 37 individual driver items (n = 491, R² = 0.763, 15 items shrunk to exactly zero) puts pride in the job, use of skills, manager support, growth prospects and feeling valued in the top five, with "salary in line with expectations" ninth.
Switch the outcome to P(intent to leave ≥ 8) and material rewards become significant: odds ratio 0.757 per standard deviation (95% CI 0.59–0.97, p = 0.028), while workload & recovery — a strong satisfaction driver — drops out entirely (OR 0.82, p = 0.12). Logistic model n = 513, base rate 58.9%, pseudo-R² = 0.226. In plain terms: people are not made happy by being paid more, but they are made less likely to walk. Recognition remains the strongest predictor of both (OR 0.478).
Adding gender, management status and job tenure to the satisfaction model raises R² by 0.0009 (F-test p = 0.698, n = 470). Who you are explains essentially nothing about your job satisfaction once we know what your job is actually like. Every demographic difference reported later in this document therefore works through experienced conditions, not around them — a useful discipline when reading the group comparisons.
The dividing line is contract, not sector
Across eight sectors, wellbeing differences are statistically indistinguishable from noise. Across contract types, they are enormous. The Eurobubble does not have a bad-sector problem; it has a status-hierarchy problem.
Testing the wellbeing index across sectors returns Kruskal–Wallis H = 7.76, p = 0.355, ε² = 0.002 (k = 8, N = 488) — that is, sector membership accounts for two-tenths of one percent of the variance. Overall satisfaction and intent to leave are likewise flat (p = 0.59 and 0.75). Across the full battery of 70 group-by-scale omnibus tests, only 13 survive Benjamini–Hochberg correction, and not one of them is a sector effect.
Now the same test across contract types, on the material rewards scale: H = 74.1, p = 1.4 × 10⁻¹⁴, ε² = 0.150. That is a seventy-five-fold larger effect than sector, and it is the largest structural effect in the dataset that is not itself an outcome variable.
The ladder runs from permanent EU officials (8.11) down through contract agents (6.63), permanent private-sector CDI holders (6.00), fixed-term CDD (5.14) and consultants (4.65) to trainees and interns (3.37). Median FTE-adjusted net pay runs the same way: €5,650 → €4,700 → €3,000 → €2,500 → €2,850 → €1,350.
Trainees and interns are the worst-paid group by a wide margin (median €1,350, material rewards 3.37) — and they report the highest workload & recovery score (6.20), above-average recognition (6.08), and satisfaction (5.81) above the sample mean. They also report the highest intent to leave in the survey (8.26). This is not contentment; it is a group whose relationship to the job has not yet been damaged by it, and who already know they are on their way out. It also means the aggregate figures in this report are held up slightly by people who are only briefly present.
What pay actually buys
Reported salary correlates strongly with feeling fairly rewarded, weakly with wanting to leave, and — once you get above the bottom quartile — not at all with being satisfied.
323 respondents disclosed a net monthly salary (47% of the sample; the lowest response rate of any question). FTE-adjusted, the median is €2,850, with the 10th percentile at €2,000 and the 90th at €5,000. Log salary correlates ρ = 0.46 with the material rewards scale — a useful validity check that the scale is measuring what it claims — but only ρ = 0.086 with overall satisfaction (p = 0.12, i.e. not significant) and ρ = −0.15 with intent to leave (p = 0.008).
Split into quartiles, the shape becomes clear. Material rewards rise steeply and monotonically across the pay distribution, from 4.28 to 7.02. Satisfaction rises from 5.02 to 5.76 between the first and third quartiles — and then falls back to 5.57 in the top quartile. Workload & recovery is almost perfectly flat across all four (5.18 → 5.39): the best-paid people in this sample get no more recovery than the worst-paid.
Intent to leave does respond to pay, and does so most sharply at the bottom: 8.11 in the lowest quartile against 6.60 in the highest. The money is doing retention work at the bottom of the distribution and very little at the top — consistent with the logistic model in section 4, and consistent with a labour market where the binding constraint for junior staff is the cost of living in Brussels rather than the attractiveness of the alternative.
Age tells a compatible story: material rewards climb steadily with age (4.47 at 18–25 to 6.69 at 50–59) while workload & recovery does not improve at all — indeed the 40–49 group reports the worst recovery in the sample (4.81), the cohort most likely to be combining senior workloads with dependent children.
Gender: the same job, a different bill
Women and men in this sample rate their jobs almost identically on every composite scale. They differ on what the job costs them physically, and on what it pays.
There is no significant gender difference in overall satisfaction (5.51 vs 5.77), in the wellbeing index (5.82 vs 6.04), in recognition, in meaning, or in intent to leave. If the analysis stopped at the composite scales, the honest conclusion would be "no gender effect."
It does not stop there. On the three items that measure the bodily and psychological cost of the work, women score significantly worse: harm to physical health (d = −0.32, p = 0.003), stress and anxiety (d = −0.22, p = 0.037), and experience of bullying or harassment (d = −0.22, p = 0.027). Negative d here means women score lower on the reverse-scored item — that is, more harm, more stress, more harassment. Three further items sit just outside conventional significance in the same direction: ability to disconnect, growth prospects, and workload & recovery.
On pay, median FTE-adjusted net salary is €2,800 for women against €3,000 for men — a raw gap of 6.7% (Mann–Whitney p = 0.008, nF = 215, nM = 95). Regressing log salary on gender with controls for age band, total experience, sector, tenure and management status, the gap does not close: the adjusted coefficient implies 8.1% (p = 0.053, n = 308, R² = 0.43). The point estimate is larger after controls than before, which is what you expect when women in the sample are, if anything, slightly better positioned on the observables than their pay reflects.
These are uncorrected tests on a self-selected sample and the salary model rests on 308 people. The finding to take forward is not the precise percentage but the pattern: identical assessments of the job, divergent assessments of its physical toll, and a pay gap that survives the obvious controls.
What people wrote
151 respondents left a free-text comment, a median of 225 characters long. Coded thematically, they do not merely echo the quantitative findings — they explain the mechanism behind the largest one.
Commenters are marginally less satisfied than non-commenters (5.19 vs 5.59) but the difference is not significant (p = 0.17), so the open text is not simply a complaints box. Themes were coded by dictionary and verified by reading; a comment can carry several.
58% of all comments raise management capability. This is not a close race — it is nearly twice the next theme, and it is by far the most frequent thing anyone volunteered. Crucially, the specific complaint is remarkably consistent: not that managers are malicious, but that they were never trained, were promoted for technical output, and have no incentive to manage. Read against section 4 — where recognition & support climate explains 41.6% of the variance in satisfaction — the qualitative and quantitative evidence converge on the same lever.
The theme that co-travels with the worst scores is not workload. Respondents who raise bullying, harassment or toxicity (19% of comments) average 3.52 on satisfaction — nearly two points below the sample mean — and 8.00 on intent to leave. Respondents who raise recognition average 8.09 on intent to leave, the highest of any theme. Those who raise workload average 5.26 on satisfaction, only slightly below the mean: heavy work, on its own, is survivable. Heavy work without recognition is not.
The management theme
“None of the managers or supervisors I had since I have started in EU affairs went through management training. The only time I had a manager actually properly trained on project management/team management was when I worked for a short stint as a customer call center operator. This is insane.”
Trade & business associations · satisfaction 7 · intent to leave 1
“It seems that the only way of fulfilling seniority in the Eurobubble is becoming a project manager. Most of these senior project managers, though, have near-zero managerial skills and refuse to take any training to improve. They seem to think it comes naturally with experience.”
Civil society / NGO · satisfaction 1 · intent to leave 10
The status hierarchy
41% of comments touch the two-tier employment structure — trainee, intern, interim, CAST, contract agent, external, freelancer, consultant. This is the qualitative counterpart to the largest structural effect in the quantitative data (section 5), and the comments make clear that the grievance is not only about money but about differential treatment: access to training, to decision-making, to being taken seriously.
“The disparity between freelancers, CAST, AST, AD, Temporary contractants, etc. is visible and suffocating. Especially in EU Institutions… where grade and type of contract creates a gap in the way people are treated and access to training courses.”
EU institutions · satisfaction 8 · intent to leave 8
“The Brussels bubble tends to hire young people like me to justify underpaying you for the responsibilities they give you.”
Trade & business associations · satisfaction 7 · intent to leave 7
Meaning, and its limits
Meaning & development is the highest-scoring dimension in the survey, and the free text shows both why that matters and why it is not a solution. Several respondents describe meaning as the thing that keeps them in an unsustainable job — the mechanism by which a workforce absorbs conditions it would otherwise refuse. Others push back on the premise entirely.
“We have limited resources and work in infinite. Setting boundaries is relatively easy but enforcing them is hard as we all want to do more, leading to burnouts in the team.”
Civil society / NGO · satisfaction 8 · intent to leave 1
“I care about what I do and feel valued by my employer, but feel trapped in an unsustainable loop.”
Civil society / NGO · satisfaction 5 · intent to leave 9
“My job is fine, but it's a second-order bullshit job in a bullshit organisation.”
EU institutions · satisfaction 5 · intent to leave 7
“Also need to be wary of having an inflated sense of purpose. If you write briefings and book flights, you are not a decision maker or in a critical function. Many people need a dose of reality and humility.”
In-house lobbyist · satisfaction 9 · intent to leave 1
HR as a non-functioning institution
19% of comments name HR, and almost none of them positively. The quantitative item agrees: HR support scores 4.13, the third-weakest item in the survey, with 49.6% of respondents at 3 or below. Two structural explanations recur in the text — small organisations that have no HR function at all, and organisations where HR is understood to report to the person you would need to complain about.
“There is zero confidentiality at work. Anything we say to HR goes straight to the big boss, and this is unspoken, internal knowledge.”
Civil society / NGO · satisfaction 7 · intent to leave 10
“The structural reasons for my burnout have not changed. That is, the organisation I work for is still under-staffed… with a workload that is too large, and therefore a certain amount of stress that is not fully avoidable despite my best efforts.”
Civil society / NGO · satisfaction 2 · intent to leave 10
Two things the survey did not ask about
The open text surfaces two hazards the instrument has no item for. The first is client and donor pressure as an external psychosocial risk: 17% of comments raise it, and consultants in particular describe the commissioning institution — not their own employer — as the source of the pressure. One respondent objected to the survey design on exactly this point, arguing that a study of wellbeing in this sector that omits the client relationship is missing a large part of the hazard. On the evidence here, they have a case.
The second is the expat condition: a workforce drawn from 63 nationalities, most of it living far from family, for whom remote-work-from-abroad policy is not a perk but the mechanism that makes the job liveable. It appears in 7% of comments and is entangled with the work-from-home item, which the factor analysis found does not load cleanly anywhere.
“As long as people are eager for jobs, there will always be someone willing to fill the position, and the cycle repeats.”
Consulting firms · satisfaction 2 · intent to leave 2
“I'm really lucky with my job, my colleagues, and the general atmosphere at work. Dream job. Happy camper.”
Civil society / NGO · satisfaction 10 · intent to leave 1
Conclusions
Six findings that the data will support, stated at the strength the evidence allows.
- This is a recognition crisis, not a pay crisis. Recognition & support climate accounts for 41.6% of the explained variance in job satisfaction; material rewards account for 10% and are not statistically significant once the other three dimensions are controlled. The single most-mentioned theme in the free text — untrained managers promoted for technical output — is the mechanism. If there is one lever in this dataset, it is the competence and behaviour of line managers, and it is a lever almost nobody in this sector is currently pulling.
- Sector is a red herring; contract status is the real cleavage. Eight sectors are statistically indistinguishable on wellbeing (ε² = 0.002). Six contract types differ enormously on material rewards (ε² = 0.150), with permanent EU officials at 8.11 and trainees at 3.37. The Brussels wellbeing conversation is usually framed as "NGOs versus consultancies versus the institutions." The data say that framing is wrong, and that the meaningful line runs horizontally through every organisation instead.
- Meaning is doing load-bearing work, and that is a risk rather than a reassurance. Meaning & development is the strongest dimension (6.48) and workload & recovery the weakest (5.33); the within-person gap is +1.12 points and 68% of respondents show it. A workforce that draws on mission to absorb conditions it would otherwise reject is stable right up until the mission stops compensating — and the free text contains many people describing precisely that moment.
- Wanting to leave the job is not wanting to leave the field. 58.9% intend to look for a new job within six months, yet intent to leave and commitment to European affairs are essentially uncorrelated (ρ = −0.08, p = 0.07): 42% of the people most determined to leave their current job are also the most determined to stay in the sector. What this workforce is generating is not an exodus but very high internal churn — organisations losing people to each other, at full replacement cost, while the talent pool stays put.
- The support infrastructure is largely absent. Half the sample cannot access mental health support through work (50.4% at ≤3) and half cannot speak openly to HR (49.6% at ≤3). These are the second- and third-weakest items in the survey. Where formal channels do not function, the manager becomes the only route — which returns the problem to conclusion 1 and explains why manager quality carries so much weight here.
- Gender differences appear in the body, not in the assessment. Women and men rate the job identically on every composite scale, and diverge significantly on harm to physical health (|d| = 0.32), stress (|d| = 0.22), and experience of bullying (|d| = 0.22), all in the direction of women bearing more, alongside a pay gap that widens rather than closes under controls (8.1%, p = 0.053). An analysis that looked only at satisfaction would have concluded there is no gender story here. There is; it is just not in the satisfaction question.
Limitations
What this dataset cannot support, in descending order of how much it should change your reading.
Self-selection with directional attrition. There is no sampling frame and no response rate, so none of these figures is a population estimate for the Brussels EU-affairs workforce. Worse than ordinary self-selection, the attrition is demonstrably non-random with respect to the outcome: drop-outs were more positive on manager support (d = 0.32) and directionally more positive on HR and mental-health items. Levels are biased pessimistic. Relationships between variables are much more robust to this than levels are, which is why this report leans on structure and effect sizes rather than headline percentages.
Common-method variance. Every variable in the driver models — predictors and outcomes alike — is self-reported by the same person at the same moment on the same response scale. This inflates R² substantially. The 0.713 should be read as "these four dimensions organise how people talk about their jobs," not as "71% of job satisfaction is caused by these four things." The same caution applies with double force to the perception-evolved variable, which is essentially a retrospective satisfaction rating and correlates with everything.
No causal identification. The design is cross-sectional. Every relationship here is an association. The contract-type finding in particular is open to selection: people who can command a permanent EU post may differ systematically from people who cannot, in ways that also affect wellbeing.
Clustering is ignored. Respondents are nested within organisations — at least 12 came from a single institution — but the organisation field was optional (154 responses) and inconsistently written. All standard errors here treat respondents as independent, which almost certainly makes them too small. Group differences with p-values near 0.05 (the gender pay gap at 0.053, material rewards at 0.059) should be treated as suggestive rather than established.
Instrument problems. Several items are double-barrelled ("I feel supported by HR, and I can talk to them openly" conflates two different failures). Two use awkward negation ("my tasks are not psychologically stressful"), which is known to inflate careless-responding error. Four items have communalities below 0.30 and one — "not physically hard," at 0.13 — is measuring desk work rather than wellbeing. Most consequentially, bullying was measured as a 1–10 intensity rather than as an incidence question, so no clean prevalence figure can be extracted from it; the 35% and 22.5% quoted above are threshold choices, not measurements.
Salary missingness. 53% of respondents did not disclose a salary, and non-disclosure is unlikely to be random with respect to either pay or satisfaction. The gender pay gap rests on 308 people. FTE adjustment assumes the reported part-time fractions are accurate.
Multiple comparisons. 70 omnibus group tests were run with Benjamini–Hochberg correction, of which 13 survive. The item-level gender tests reported in section 7 are uncorrected — with 12 tests at α = 0.05, roughly one false positive is expected, and the three significant results should be read as a pattern rather than three independent discoveries.
Qualitative reach. 151 comments from 684 respondents is a good rate for an optional field, but it is 22% of the sample and it cannot be treated as representative. Theme frequencies describe what commenters chose to raise, not what the workforce thinks.
Statistical power on the nulls. The sector null is informative — with N = 488 across 8 groups it rules out large sector effects — but it does not rule out small ones, and it says nothing about variation between organisations within a sector, which the free text suggests is where the real differences live.
Open questions for the next wave
Eight things this dataset raises and cannot answer, with what it would take to answer them.
- Is the meaning premium protective or corrosive? Cross-sectionally, people with high meaning and poor conditions report slightly lower intent to leave (7.36 vs 8.02, p = 0.10) — meaning appears to be holding people in place. Whether that ends in recovery or in burnout is a longitudinal question. Panel the same respondents at 12 months and test whether a high meaning-to-conditions gap predicts subsequent burnout, sick leave, or exit.
- Does bullying really belong to the recognition dimension? Here it loads at +0.57 with feeling valued and manager support rather than forming its own factor. That would be an important claim about how harassment operates in this sector — but it may be an artefact of having only one harassment item. Add a proper battery (frequency, type, perpetrator, reporting, outcome) and re-run the factor analysis.
- Why does satisfaction peak in the third salary quartile and fall in the fourth? The reversal (5.76 → 5.57) is not large and may be noise. If it is real, the candidate explanations — responsibility load, seniority isolation, the 40–49 recovery trough — are separable with a bigger sample and a direct question about hours.
- Is the contract hierarchy causal? Ask respondents whether their contract status changed in the last 12 months, and compare the wellbeing of movers before and after. That converts a cross-sectional ranking into something much closer to an effect.
- How much of the variance is between organisations rather than between sectors? Sector explains 0.2%. The free text insists that the employer, not the sector, is what determines whether the job is survivable. Collect organisation identity properly (structured, optional, with a confidentiality guarantee) and fit a multilevel model. The intraclass correlation would be one of the most actionable numbers this survey could produce.
- Is bad management a training problem or an incentive problem? Respondents consistently describe managers promoted for technical output and never trained. Two distinguishable remedies follow — mandatory management training, or separating technical and managerial career tracks. Ask whether the respondent's manager has had management training, and whether their organisation offers a senior individual-contributor path.
- What is the churn actually costing? With 58.9% intending to look and field commitment uncorrelated with that intention, this sector appears to be recycling the same people between employers. Ask how many EU-affairs employers a respondent has had, and how long each lasted. That turns an intention measure into an estimate of real turnover.
- Should the client relationship be inside the instrument? 17% of commenters raised client or donor pressure unprompted, and consultants located the hazard in the commissioning institution rather than their own employer. A survey covering both consultants and the institutions that commission them is unusually well placed to measure a psychosocial risk that flows between organisations — and no current item touches it.
Cut or reframe "my tasks are not physically hard" (communality 0.13). Split the double-barrelled HR item into "HR is available to me" and "I would trust HR with a problem" — the free text suggests these fail separately and for different reasons. Replace the bullying intensity scale with an incidence question plus follow-ups, so the next wave can report a prevalence figure that means something.
