Salary and Job Satisfaction among PhD Holders
Results from the 2025 Observatory Survey — Focus on Belgium
Reasons for Staying in — and Leaving — Academia
Why researchers pursue an academic career
What pushed PhD holders to leave academia
Salary among Belgian PhD Holders
Net salary distribution
Note
For context, master’s degree holders in Belgium earn approximately €3,480 net/month (estimated from Statbel’s 2022 gross figure of €5,951, source, applying 2026 tax rules), while young university graduates starting their career earn approximately €2,156 net/month (estimated from Acerta’s 2025 gross figure of €2,982, source, same method). Salary increases strongly with seniority in Belgium, and the survey sample skews young, so the entry-level figure is likely more representative here.
How salary is determined
Did your PhD lead to a more attractive salary?
Satisfaction with Salary and Benefits
By sector
| Mean satisfaction with salary and benefits by sector | ||
| Respondents working in Belgium (1 = Very dissatisfied, 5 = Very satisfied) | ||
| Sector | N | Mean satisfaction |
|---|---|---|
| Government / Public authorities | 198 | 3.97 |
| Business: Industry | 204 | 3.94 |
| University/Research center | 533 | 3.82 |
| Business: Services | 158 | 3.73 |
Association between salary satisfaction and overall job satisfaction
Beyond the objective salary amount, respondents’ subjective satisfaction with their salary and benefits is also related to their overall job satisfaction. The crosstabulation below shows a clear pattern: among respondents very dissatisfied with their salary, over half also report low overall satisfaction (levels 1-2), whereas among those very satisfied with their salary, the large majority report high overall satisfaction (levels 4-5). This association is confirmed by a Spearman correlation of 0.46 in the full sample and 0.41 among Belgian respondents, indicating a moderate, positive relationship between the two dimensions.
| Overall satisfaction by salary satisfaction for respondents working in Belgium | |||||
| Row percentages | |||||
| Overall satisfaction |
Satisfaction with salary and benefits
|
||||
|---|---|---|---|---|---|
| 1 (Very dissatisfied) | 2 | 3 | 4 | 5 (Very satisfied) | |
| 1 | 47.8% | 26.1% | 8.7% | 17.4% | 0.0% |
| 2 | 20.3% | 18.9% | 16.2% | 23.0% | 21.6% |
| 3 | 9.3% | 15.3% | 31.6% | 33.0% | 10.7% |
| 4 | 2.8% | 9.4% | 14.6% | 52.4% | 20.7% |
| 5 | 0.7% | 2.0% | 5.1% | 39.5% | 52.7% |
| Spearman correlation: ρ = 0.46 (full sample), ρ = 0.41 (Belgian subsample) | |||||
Effect of Salary on Overall Job Satisfaction
To assess the effect of salary on overall job satisfaction, we compared three modelling approaches: OLS linear regression, binary logistic regression, and ordinal logistic regression (cumulative link model, CLM). OLS was ruled out because it treats the five-point satisfaction scale as continuous and equally spaced — an assumption the data clearly violate, as confirmed by strongly non-normal residuals. Binary logistic regression was discarded because collapsing the scale into a satisfied/not-satisfied dichotomy unnecessarily discards information about degree of (dis)satisfaction. The CLM, which respects the ordered nature of the five response categories, was retained as the most appropriate model. Its proportional odds assumption was verified using the Brant test: while a few control variables show minor violations (notably contract type), the assumption holds cleanly for salary across all contrasts. However, it is worth noting that salary has a significant effect across all models considered.
Each model controlled for research intensity (low vs. high), contract type, presence of children, nationality, relationship status, research field, whether a PhD was required for the position, and age.
The analysis was conducted on the full sample and replicated on Belgian respondents only. For the Belgian subsample, the original absolute thresholds could not be retained: the lowest salary bands contained too few respondents (e.g., only 6 below €1,000/month) for stable model estimates, reflecting how few PhD holders working in Belgium in this sample earn low salaries. Salary categories were therefore redefined relative to national benchmarks (national median, mean, and 90th percentile), ensuring sufficient sample size per category while preserving a meaningful interpretive anchor.
Full sample
In the full sample, the results show a clear positive association between salary and predicted overall satisfaction. Predicted satisfaction rises from 3.3 among those earning below €1,000/month to 4.2 among those earning over €4,000/month on the five-point scale.
Belgian respondents
Among Belgian respondents, the same upward pattern holds, but with a plateau at the top. Predicted satisfaction rises steadily from 3.73 among those earning below the national median, to 3.87 around the median, and to 4.03 above the national mean. Beyond that point the gains largely disappear: those earning around the national 90th percentile (4.11) and those earning well above it (4.09) report virtually identical predicted satisfaction, with overlapping confidence intervals. In other words, for Belgian PhD holders, earning more than the national top decile does not appear to translate into additional satisfaction — most of the benefit of a higher salary is already captured at or above the national mean.
Annex: Regression Model Outputs
The tables below report the predictor coefficients from the two ordinal logistic regression (CLM) models as odds ratios with 95% confidence intervals. Threshold parameters (the cut-points between ordered response levels) are omitted for readability.
Full sample model
Because salary is ordered rather than nominal, R’s default polynomial contrasts decompose its overall effect into linear, quadratic, cubic, and quartic components (labelled .L, .Q, .C, and ^4 in the tables); the linear (.L) term is the most relevant, as it captures the dominant, monotonic trend of increasing satisfaction with salary, while the higher-order terms indicate minor deviations from strict linearity at specific points along the scale.
| CLM — Full sample | ||||
| Outcome: overall job satisfaction (1 = Very dissatisfied, 5 = Very satisfied) | ||||
| Variable | Odds Ratio | 95% CI (low) | 95% CI (high) | p-value |
|---|---|---|---|---|
| salary_5cat.L | 3.74 | 2.55 | 5.47 | < 0.001 |
| salary_5cat.Q | 0.85 | 0.61 | 1.18 | 0.336 |
| salary_5cat.C | 1.48 | 1.11 | 1.97 | 0.008 |
| salary_5cat^4 | 0.74 | 0.59 | 0.92 | 0.008 |
| Lowresearch | 1.51 | 1.19 | 1.90 | < 0.001 |
| CDIInterim | 1.32 | 0.53 | 3.27 | 0.549 |
| CDIPermanent contract | 1.66 | 1.35 | 2.05 | < 0.001 |
| childrenWithout Children | 1.09 | 0.88 | 1.35 | 0.435 |
| Nat | 1.10 | 0.89 | 1.34 | 0.377 |
| couple | 1.26 | 1.00 | 1.59 | 0.050 |
| fieldSEN | 1.10 | 0.88 | 1.37 | 0.401 |
| fieldSHS | 1.24 | 0.97 | 1.57 | 0.082 |
| Q114_phd_requiredNon | 0.89 | 0.71 | 1.10 | 0.270 |
| age | 0.98 | 0.97 | 1.00 | 0.109 |
Belgium model
| CLM — Belgium | ||||
| Outcome: overall job satisfaction (1 = Very dissatisfied, 5 = Very satisfied) | ||||
| Variable | Odds Ratio | 95% CI (low) | 95% CI (high) | p-value |
|---|---|---|---|---|
| salary_5cat_be.L | 2.12 | 1.49 | 3.01 | < 0.001 |
| salary_5cat_be.Q | 0.78 | 0.58 | 1.04 | 0.089 |
| salary_5cat_be.C | 0.91 | 0.69 | 1.20 | 0.491 |
| salary_5cat_be^4 | 1.02 | 0.81 | 1.28 | 0.861 |
| Lowresearch | 1.69 | 1.29 | 2.20 | < 0.001 |
| CDIInterim | 1.31 | 0.40 | 4.25 | 0.656 |
| CDIPermanent contract | 1.85 | 1.41 | 2.43 | < 0.001 |
| childrenWithout Children | 1.02 | 0.78 | 1.33 | 0.881 |
| Nat | 1.31 | 0.97 | 1.78 | 0.079 |
| couple | 1.32 | 0.98 | 1.78 | 0.071 |
| fieldSEN | 0.89 | 0.67 | 1.18 | 0.430 |
| fieldSHS | 1.07 | 0.79 | 1.45 | 0.665 |
| Q114_phd_requiredNon | 0.91 | 0.70 | 1.19 | 0.503 |
| age | 0.99 | 0.96 | 1.01 | 0.264 |