Workforce Insights from Central Governments

Annex A. Technical documentation supporting the 2024 EU/OECD Survey of Public Servants

Introduction

This technical annex complements the report ‘Workforce Insights from Central Governments - Findings of the 2024 OECD/EU Survey of Public Servants’. It provides an overview of participating countries and organisations, the development and design of the survey, data collection methodology, a description of the sample, respondents anonymisation measures, data quality checks and analysis. This survey was conducted by the OECD with funding and support from the European Commission (SG REFORM).

The EU/OECD standard EU Survey of Central Government Public Servants was guided by an Advisory Group, comprised of representatives from the eight participating countries Belgium, Bulgaria, Croatia, the Netherlands, Latvia, Lithuania, The Slovak Republic, and Slovenia and The European Commission SG REFORM. The Advisory Group had Luxembourg, Denmark and Norway as observer countries and benefited from occasional participation from public employee survey experts from other OECD member countries who presented their national experiences.

The data was collected from May to June 2024 for seven of the eight countries, and from September to October for Belgium. The survey was also run in Denmark November-December and in Norway October-November 2024. The survey scope was public servants in the participating countries’ central government ministries and agencies (Annex B provides an overview of all participating organisations). The final sample size was 51 761 valid responses from the eight project countries, and 56 980 valid responses when counting the responses of Denmark and Norway who ran part of the questionnaire. The data collection was conducted through online anonymous surveys. All surveys were conducted in the country’s national language(s).

Survey preparation and implementation

Questionnaire development, translation and administration

The questionnaire was developed by the project Advisory Group (AG) comprised of members from each of the participating countries, the project officer responsible from the European Commission Reform and Investment Task Force (SG REFORM) and the project team of the OECD secretariat. The meetings of the AG also benefited from presentations and knowledge exchange with experts from other OECD countries with previous experience in running employees survey and using their results for evidence-based decision-making.

The AG developed the questionnaire through thematic discussions on their own national questionnaires as well as international best practices from other OECD countries. In the first phase, the Advisory Group (AG) identified and agreed on a set of key priority areas for the survey. In the following phase, the OECD Secretariat proposed a set of questions, drawing on examples from existing surveys as well as newly developed items, from which the AG selected and approved the final list. The final questionnaire features a range of common employee survey questions in areas such as employee engagement, leadership, management and well-being, team and organisational performance, as well as emerging topics such as innovation, use of technology and remote and hybrid work. In addition, the questionnaire includes a section of sociodemographic questions such as age, gender, managerial status and length of service. (Table A ‎A.1)

Questions were selected based on their relevance and ability to capture the key dimensions of the survey. Questions were carefully selected or constructed in clear and simple language to facilitate ease of understanding for respondents. Special attention was given to selecting or constructing questions that would translate effectively into all the languages of the participating countries. Efforts were also made to avoid questions that were overly specific to any one country, in line with the questionnaire's goal of enabling cross-country comparisons.

The sequencing of the questionnaire was randomised, meaning that the sections and their subsections after the first ‘about me section’ were presented in different orders to each respondent. The randomisation scheme was applied for three primary reasons. Firstly, to mitigate question order bias and fatigue, where respondents tend to pay more attention to the initial parts of the survey and less to the latter. Secondly, to minimise context effects, as the order of questions can influence responses to subsequent questions. For example, questions about pay satisfaction may affect responses to job satisfaction if the former are presented first and remain fresh in the respondent's mind. Lastly, randomisation was employed to ensure that those that did not complete the full survey did not disproportionately affect the number of responses per section or subsection, which would otherwise result in fewer responses for the latter sections of the questionnaire.

The questionnaire featured various types of questions, with the majority (77%) utilising a five-point Likert scale, ranging from ‘Strongly Disagree’ to ‘Strongly Agree’. This approach aimed to streamline the questionnaire by displaying questions in a matrix format, grouping related questions from the same section, to make the survey more intuitive for the respondents and improve ease of completion.

Table A ‎A.1. Survey sections and subsections

Section	Subsection
1. About me	Demographic questions i.e. organisation, gender, age group, educational attainment, contract type, managerial status full-time/part-time, tenure.
2. My personal work experience	2.1. My pay and employment conditions 2.2. My working arrangements 2.3. My well-being 2.4. My learning and development opportunities 2.5. My work experience
3. About my management and senior leadership	3.1. My immediate line manager or supervisor 3.2. My senior leadership
4. About my team and innovation	4.1. My organisations openness to innovation 4.2. My organisations use of technology 4.3. My team’s performance 4.4. My organisations performance

Note: sections were randomised after the ‘About me’ section.

The final English version of the questionnaire, can be found in Annex A. The questionnaire has reviewed by relevant internal bodies in the OECD. All surveys were conducted in the national languages of each country. The baseline English version of the questionnaire was translated into the national language(s) of the participating countries using the European Commission's e-Translation tool1. Each country was responsible for reviewing and validating the translations to ensure accuracy. For countries sharing a common language such as Belgium and the Netherlands, translation efforts were coordinated to maintain consistency and comparability. After the translated questionnaire was scripted by the OECD secretariat, the countries conducted final internal tests of the survey.

Survey scope and sampling design

The scope of the survey was carefully defined to ensure the comparability of results across countries, recognising that definitions of civil services vary slightly between countries. The survey was distributed to all public servants working in ministries and agencies at the central or federal level of government, excluding those employed by state, territorial, regional, and municipal governments. Furthermore, state-owned enterprises, public corporations, and government-owned non-profit institutions were not included in the survey. Additionally, it excluded government-employed doctors and nurses working in hospitals, teachers, police officers, judges, firefighters, and military personnel (with the exception of civilian military staff). The advisory group agreed to distribute the survey to all employees within this defined scope, making it a census sample. Table A ‎B.1 in Annex B provides an overview of all organisations participating in the survey by country. The survey was specifically designed for public servants working in these organisations, and participating countries were advised to distribute the survey link exclusively to this group of employees. However, as the survey link was distributed within large organisations, the first page of the survey included a description of those employees who should not fill out the questionnaire, in order to discourage participation of those out of scope (see instruction page in Annex C).

Survey dissemination and communication

Different communication messages were designed by the OECD project team both prior and during the data collection period to have a wider reach within the participating organisations and maximise public servant’s engagement with the survey and the response rate. These communications included email invitations, reminder emails to boost response rates, posters distributed in office spaces to raise awareness, and, in some countries, social media reminders. An official OECD contact channel was also established, providing an email address for participants to report technical issues or ask questions related to the survey.

Participating organisations in each country were strongly encouraged to send two to three reminder emails to their employees, depending on the length of the data collection period. Countries implemented a strategic schedule for reminder emails: the first reminder was sent two weeks after the survey's launch, a second midway through the survey period, and a final reminder in the last three to five days before the survey closed. The response rates of the countries and underlying organisations were monitored by the OECD secretariat to identify challenges such as stagnation in the number of daily responses and organisations with particularly low response rates, where tailored follow-up efforts were needed.

A single, standardised communication poster was developed for distribution by participating organisations (see Annex D). The poster highlighted key survey topics, the closing dates aligned with each country's data collection timeline, and contact details for inquiries, including direct support from the OECD. Additionally, the poster featured a QR code, enabling respondents to access the survey via their smartphones.

Beyond serving as a mean for accessing the survey via smartphones through a QR code, the poster played a key role in increasing survey visibility. Displayed in common work areas, it provided a constant, physical reminder that encouraged employees to participate, even outside of their usual email communications.

The poster was distributed in the official language of each country, with optional modifications requested by some countries to include contact information to a national contact point in addition to the OECD, so respondents could inquire in their national languages.

Data collection

For the majority of countries (Belgium, Bulgaria, Croatia, The Slovak Republic, and Slovenia), the OECD Secretariat collected the data through the contracted survey platform, Qualtrics. For these countries, the OECD Secretariat scripted the survey, applying the agreed survey design and sequencing. Each country subsequently tested the survey in their national language, and where necessary, the OECD made any final adjustments.

The remaining countries (the Netherlands, Latvia, and Lithuania) opted for collecting the data themselves through their own survey platform tool. To ensure comparability, these countries were provided with the Qualtrics survey script, including the sequencing and randomisation scheme, to align their survey accordingly.

The data collection of the 2024 EU Survey of Central Government Public Servant took place from mid-May until the end of June in most of the countries, allowing a survey availability period from three to six weeks. For Belgium, the data collection took place from mid-September to late October, as the initial data collection period was during their EU Council presidency. A detailed overview of the data collection period is in Table A ‎A.2.

The timing of the survey was carefully planned to be either before or after the summer holiday period to mitigate potential biases in response rates and the quality of responses. Surveys conducted during or close to holidays may face two types of biases: a positive bias, where employees may report higher satisfaction due to anticipation of upcoming leave, and a negative bias, where responses could be affected by work fatigue and eagerness for time off. Furthermore, it was important to conduct the surveys simultaneously across all participating countries to ensure data comparability. This timing prevented external factors, such as global events, and time-related influences, such as holidays, from affecting responses differently across countries, helping to minimise potential biases. The surveys were conducted online and distributed in the official language of each participating country. For Belgium, the survey was available in all three official languages: French, Dutch, and German. Each country’s advisory group members were responsible for determining the distribution strategy. In most cases, the survey was disseminated to heads of departments or human resource management (HRM) offices, which were responsible for further distribution within their respective organisations.

Table A ‎A.2. Data collection periods by country

Country	Launch date	Closing date	Survey Duration (days)
Belgium	16-Sep	31-Oct	46
Bulgaria	08-May	20-Jun	43
Croatia	25-May	14-Jun	20
Latvia	08-May	28-May	20
Lithuania	20-May	30-Jun	41
The Netherlands	05-Jun	1-Jul	27
Slovak Republic	17-May	30-Jun	44
Slovenia	16-May	14-Jun	29
Denmark	12-Dec	06-Jan	26
Norway	28-Oct	11-Nov	15

Response Rates

The total number of qualified responses across the eight countries was 51 761 valid responses from the eight project countries, and 56 980 valid responses when counting the responses of Denmark and Norway who ran part of the questionnaire. The number of qualified responses per country varied significantly, ranging from 1 771 to 12 521, depending on each country’s public service population size, which spanned from 10 580 to 73 259 employees (Table A ‎A.3).

The median time to complete the survey across the entire sample was 21 minutes and 24 seconds per respondent2. However, this duration varied between countries, from a median of 10 minutes in the Netherlands to 25 minutes in Lithuania. These differences are largely due to country-specific adjustments to the survey distributed in these countries, such as the addition or removal of questions.

Response rates ranged from approximately 9.2% to 54.5% across the participating countries, and countries with more experience in conducting similar surveys, such as Lithuania, the Netherlands, and Latvia, had stronger participation rates, likely due to the established trust between employees and the organisations distributing the surveys in these countries. Belgium, despite being among the countries with prior experience conducting employee surveys, recorded low participation. This was likely due to survey fatigue, as several organisations were running surveys simultaneously. Additionally, some organisations were reluctant to distribute the survey via employee email and opted to share it only through the workplace intranet, which may have further limited reach.

Table A ‎A.3. Response rates by country

Country	Employees receiving the survey	Completed surveys	Responses received	Qualified responses	Response rate	Median response time (minutes)
Belgium	73 259	4 610	5 484	5 117	9.2%	16
Bulgaria	60 993	11 119	12 632	11 770	19.3%	23
Croatia	23 342	3 928	4 563	4 261	18.3%	18
Latvia	20 272	7 261	7 446	7 261	35.8%	NA
Lithuania	22 974	10 891	12 580	12 521	54.5%	25
Netherlands	14 100	4 854	4 857	4 854	34.4%	10*
Slovak Republic	11 900	1 518	2 083	1 771	14.9%	21
Slovenia	10 580	3 713	4 678	4 148	39.2%	17
Denmark					24%
Norway					54.4%

* The median time was lower given that the Netherlands’ survey didn’t include all the questions from the EU Survey of Central Government Public Servant.

Note: qualified responses are those responses who submitted all demographic questions plus at least one question beyond this section.

Data cleaning and anonymisation

The integration of the cross-country data into the aggregate database required transforming each country’s dataset to ensure consistency and a uniform structure across all the responses collected. All eight participating countries followed the same survey questionnaire, with some tailored exceptions listed below. In addition, the questionnaire included four optional questions, which each country could choose to include or exclude in their survey (see Table A ‎A.4). In most participating countries, all survey questions were mandatory, with the option to select “prefer not to respond” available for respondents who wished to omit a particular question. Despite these modifications, the integrity of the survey's coding remained consistent across all countries to ensure comparability. The participating organisations for each country were determined by the initial selections made by the advisory groups, ensuring they fell within the scope of central ministries and agencies. As a result, the organisations included in the sample vary between countries. Lastly, all countries implemented the questionnaire with a randomised structure organised by sections. The demographic section was always presented first, while the final question (Q99) was not randomised, regardless of any additional questions included or removed in their respective surveys.

Table A ‎A.4. Inclusion of optional questions

Country	QO1. The best description of my work function is:	QO2. When I am in the office, my workspace is usually the following:	QO3. I am satisfied with the physical conditions at my workplace, such as:	QO4. To what extent do you trust the political leadership of your organisation?
Belgium	✓	✓	✓
Bulgaria	✓	✓	✓	✓
Croatia	✓	✓	✓	✓
Latvia	✓	✓	✓	✓
Lithuania	✓
Netherlands
Slovak Republic	✓	✓	✓	✓
Slovenia		✓	✓	✓

As for countries’ specificities, in Belgium, the variable on language was retained with the aim to see differences across French, Dutch and German speakers.

In Latvia, some questions did not provide the “prefer not to respond” option. The questions that included the “prefer not to respond” option was QH, Q21, Q27, Q39, Q40, Q41, Q43, Q44, Q58, Q70, and Q043. In addition, respondents were initially asked to identify the ministry under which their agency was affiliated, followed by a question specifying the agency. If respondents selected a ministry but chose “prefer not to respond” when asked about their specific agency, they were allocated in the ministries they initially selected, as if they worked in the ministries’ headquarters to avoid dropping these observations.

In Lithuania, certain demographic questions had some response options removed to align with the selected sample of respondents in each country. For demographic question C, “My highest educational qualification is:”, option C referring to short-cycle tertiary education (ISCED category 5) was removed. For demographic question D, “My employment status is:”, option A referring to Trainee/intern/apprentice, and option B referring to Entry-level civil service graduate scheme were removed. As with Latvia, respondents were first asked to identify the ministry their agency was affiliated with, followed by a question specifying the agency. “Prefer not to respond” selections for agencies were allocated to the ministry initially identified by the respondent.

In Slovenia, for demographic question C, “My highest educational qualification is:”, option B referring to post-secondary non-tertiary education (ISCED category 4) was removed because the Slovenian education system does not have this ISCED category in their educational system. (OECD, 2023[1])

In the Netherlands, only select parts of the complete survey were included such as 10 demographic questions from the Section 1 – About me, 21 questions from the Section 2 – About my personal work experience, 14 questions from the Section 3 – About my management and senior leadership, 26 questions from Section 4 – About my team and organisation and the final question from the survey4.

For demographic question B, “My age group is:”, the option F referring to “66 years and over” was removed and instead, option E was modified to “56 years and over”. For demographic question C, “My highest educational qualification is:”, primary and lower secondary education (ISCED category 1 and 2) where included additionally to the existent options of this questions. Therefore, to preserve consistency with the structure of the original survey, respondents in the Netherlands who selected primary or lower secondary education were recoded with empty values in the education variable to ensure comparability across the aggregate data. For demographic question D, “My employment status is:”, was removed from the questionnaire as option B referring to entry-level civil service graduate scheme was deemed too small, and in combination with other information as organisation, age, gender, years employed, and type of contract created a possible identification risk. However, since all respondents in the Dutch sample held civil servant status, all observations were coded as such for question D. For demographic question G, "I manage other employees in my organisation", the options “Yes”, “No”, and “Prefer not to respond” were provided, instead of asking for the number of employees managed. This adjustment was made to ensure anonymity, as the research instrument used did not permit collecting data on the specific number of employees managed.

The Netherlands adopted a stratified sampling method for civil servants based on organisational type rather than sampling by individual organisation. This included three categories of organisations: policymaking, regulatory/audit/oversight, and service-delivery entities. As a result, the data regarding organisations participating in the survey was only available at the ministry level, rather than at the agency level.

In Denmark, a random sampling approach was used, based on a registry of Danish government employees. Only data on employees from the central government civil service were shared with the OECD, resulting in a sample of 767 individuals (24%). The survey was distributed via Digital Post, a secure public communication channel used for correspondence between citizens and the government, unlike in other countries, where respondents received the survey through their work email. Questions on satisfaction with line managers were only asked to non-managerial staff.

In Norway, a stratified sampling approach was used. In the first stage, a random sample was drawn, including a minimum of 1 000 individuals from each of the following strata: 1) Ministries and the Office of the Auditor General, 2) Public administration (excluding ministries), 3) Research and education, and 4) Public enterprises and construction. This resulted in an initial sample of 8 163 individuals. A second sampling stage was then conducted to ensure proportional representation of managers, yielding a final sample of 4 452 individuals (54.4%).

Data cleaning/suppression

Qualified responses were defined as those from public servants who completed all demographic questions and at least one substantive question, taking five minutes or more to finish the survey. This classification means that incomplete responses—those from individuals who did not answer any questions after the demographic section—and "super speeders", who completed the survey significantly faster than average and might therefore not have fully read the questions, were excluded to ensure data quality. The threshold for identifying "super speeders" varied based to the number of questions included in each country’s survey. In countries that included the full set of questions, respondents who completed the survey in five minutes or less, representing around 25% of the median completion time for these countries, were classified as "super speeders". Lithuania, with a longer survey, had a threshold of six minutes or less, while the Netherlands, with a shorter survey, had a threshold of three minutes or less. In the case of Latvia, this specific data cleaning step was not implemented, as the time taken to complete the survey was not available. Overall, an average of 6% of responses were discarded by country due to not meeting the qualification criteria.

Anonymised databases for countries

A set of data files was prepared for the distribution of survey results among the participating countries and organisations. These included: (i) a file with the aggregate data across all eight countries, (ii) a file with aggregate data by country, and (iii) a file with data disaggregated by organisation within each country.

To ensure the anonymity of all survey respondents, a series of anonymisation procedures were applied. No personal identifiers (e.g., names, emails, IP addresses) were collected, and participation in the survey was entirely voluntary. Respondents were informed at the beginning of the questionnaire about how their data would be used and that they could withdraw from the survey at any point. For the aggregate country-level file and the organisational-level file, aggregated results were provided along with a breakdown by three demographic variables: gender, age, and the number of employees managed. At the country level, for the gender breakdown, the categories "other" and "prefer not to respond" were merged to prevent the re-identification of individuals based on multiple characteristics. At the organisational level, additional rules were in place to maintain anonymity. For organisations with 10 or fewer respondents, no results were reported at the organisational level; instead, their data was only accounted into the aggregated country-level results. For organisations with between 11 and 25 respondents, only aggregate results per question were shared. For those with over 26 responses, the breakdown by the three mentioned demographic variables were made.

For anonymity protection in the organisational level file, certain demographic groups that were too small in size were re-categorised to ensure confidentiality. In the gender category, responses of "other" and "prefer not to respond" were excluded from the crosstabs. In the age category, the youngest group, "24 years and under", and the oldest group, "66 and over", were merged with adjacent age groups. Similarly, for the "employees managed" variable, respondents were grouped into two categories: those managing no employees and those managing one or more. Additionally, on a case-by-case basis, crosstabs with fewer than 10 responses for any demographic breakdown were excluded from the files shared, though these data points were still incorporated in the overall analysis. This method helps preserve respondent confidentiality while ensuring that results are robust and reliable, particularly in protecting small demographic groups from potential identification.

Data analysis

Weighting

To address potential under- or overrepresentation in the sample, the possibility of applying weights to country samples was considered. While weighting is more straightforward in population surveys—where key demographic data, such as gender and age, are typically available through registers—this survey targeted public servants working in ministries and agencies. In some countries, such centralised registers were not available for this population.

Several challenges arose in considering weights: (1) the demographic data for the employees in our scope might not be updated or aligned with the survey's data collection period, and (2) age data, when available, was often grouped into categories that did not correspond to the age ranges used in the survey. Given these limitations, it was ultimately decided not to apply weights. The risk of introducing inaccuracies through inappropriate weighting could have diminished, rather than enhanced, the representativeness of the sample.

The Netherlands represents a special case, as they did not conduct the survey as a census but instead adopted a stratified sampling methodology, being unable to administer the survey across all central government organisations. From available registers, a stratified sample of 14 000 employees was drawn. To enhance the representativeness of the results, weights were applied based on the type of organisation (classified as: 1. ‘Policy design/regulatory drafting’, 2. ‘Inspections/oversight (audit/evaluation)’, or 3. ‘Service delivery to users/executive organisations’), as well as gender, age group, and managerial status. These weights were applied exclusively to the data from the Netherlands, as it was the only country to employ a stratified sampling methodology.

Composite indices

Some survey questions were combined into ten composite indices, grouping specific items into broader thematic dimensions or constructs. These dimensions include: Employee Engagement (two indices), Well-being, Management, Leadership, Organisational Performance, Team Performance, Innovation Climate, Learning and Development, and Job Autonomy.

All indices are based on questions using a 5-point Likert scale, ranging from "Strongly disagree" to "Strongly agree." Each index is calculated by aggregating the responses to the relevant questions and dividing by the total number of questions. To facilitate easier interpretation, the indices have been rescaled to a 0–100 range. On this scale, a response of "Strongly disagree" corresponds to 0, "Neither agree nor disagree" corresponds to 50, and "Strongly agree" corresponds to 100. An example of how the index score is calculated is provided below:

Table A ‎A.5. Example of the rescaling of indices

	Strongly disagree	Disagree	Neither agree nor disagree	Agree	Strongly agree	Score
Weight	0	25	50	75	100
Q1		✓				25
Q2	✓					0
Q3			✓			50
Q4				✓		75
Q5				✓		75
Total score of questions						225
Index score (Total/5)						45

Eight out of the ten indices only include respondents who provided answers to all items in the index. This means that any responses with missing data—either due to early dropout or the selection of 'prefer not to respond' were excluded from the analysis. The exception of this is the Learning and Development Index, here missing data was imputed for the first item: Q23 "The learning activities I completed in the last 12 months have helped me improve my performance." This item was conditional, dependent on respondents' prior indication of having participated in learning activities. Given its importance to the Learning and Development Index, missing responses for this item were replaced by the average score of other respondents within the same organisation as a proxy of the quality of the learning activities available at the organisations.

To ensure consistency across indices, some items were recoded to reverse their original response scale. This was the case for Well-being Index items Q16-18 (e.g., Q16, "I feel burned out"), which were originally negatively framed, unlike the remaining positively framed well-being items. The recoding ensures that higher scores across all items consistently reflect higher levels of well-being.

The number of observations retained for each index is provided below. As the Netherlands did not include all sections of the survey, they are only included in the indices for Employee Engagement, Management, Organisational Performance, Team Performance, and Innovation Climate.

The indices used in this report have undergone rigorous statistical validation. Cronbach’s Alpha scores for each index, presented in the , exceed 0.690, confirming strong internal consistency and the reliability of the items in measuring the same underlying constructs.

 

Employee Engagement
Q29	Overall, I am satisfied with my job
Q30	At my work I feel full of energy.
Q31	The work I do gives me a sense of accomplishment.
Q32	I am enthusiastic about my job.
Q33	I am immersed in my work.
Q34	I am willing to do extra work for my job beyond what is expected of me.
Q36	I identify with the mission of my organisation.
Q37	I would recommend my organisation as a good place to work.
Q38	I am proud to work for this organisation.

Note: Cronbachs-alpha value of 0.90, N=45 845

 

Employee Engagement (Utrecht 3)
Q30	At my work I feel full of energy. (vigor)
Q32	I am enthusiastic about my job. (dedication)
Q33	I am immersed in my work. (absorption)

Note: Cronbachs-alpha value of 0.74, N=47 587

 

Well-being
Subindex: Workload management
Q13	I achieve a good balance between my work life.
Q14	I have clear responsibilities and know what is expected of me
Q15	I can usually handle my workload well
Physical/mental well-being
Q16	I feel burned out
Q17	I often feel exhausted at the end of (..)
Q18	I sometimes work despite feeling sick
Supportive well-being environment
Q19	I would feel comfortable sharing
Q20	I feel able to support employees (..) mental health

Note: Cronbachs-alpha value of 0.69, N=40 700, Q16-18 were reversed to align with the other index question where a higher value represents higher levels of well-being.

 

Management
Q45	My line manager/supervisor plans the work well
Q46	… shares important information with me.
Q47	… trusts my judgement.
Q48	…lets me be reasonably autonomous (i.e. does not micro-manage).
Q49	… treats me with respect.
Q50	… is good at resolving conflicts
Q51	… can be counted on to help me with my work if I need support.
Q52	… encourages me to come up with new or better ways of doing things.
Q53	… provides me with helpful feedback to improve my performance.
Q54	… recognises and rewards good performance.
Q55	… provides opportunities fairly to all employees in my work unit or team (e.g., promotions, work assignments, training, etc.).
Q56	… makes decisions based on evidence and facts
Q57	… maintains high standards of honesty and integrity

Note: Cronbachs-alpha value of 0.74, N=45 527

 

Senior leadership
Q59	The senior leaders of my organisation clearly articulate the direction and priorities of the organisation.
Q60	… are aware of global events and how they might impact the organisation.
Q61	… generally manage the organisation well.
Q62	…. effectively communicate essential information to staff.
Q63	… are effectively leading change in the organisation .
Q64	… promote co-operation within the organisation .
Q65	… recognise and appreciate our work.
Q66	... trust the judgement of their employees.
Q67	… uphold and defend public service values (e.g. honesty, integrity, impartiality, etc.)
Q68	… maintain high standards of honesty and integrity.
Q69	… provide evidence-based advice to political leaders, even if this advice goes against the political position.

Note: Cronbachs-alpha value of 0.74, N=34 957

 

Organisational performance
Q87	My organisation is quick to respond when changes need to be made
Q88	… uses its resources efficiently.
Q89	… has rules and procedures in place that make it difficult and complicated to work effectively and efficiently.
Q90	… takes measuring and monitoring performance seriously.
Q91	… dedicates sufficient attention to reducing its environmental footprint (saving energy, reducing waste, promoting sustainable travel, recycling, etc.)
Q92	… is open with the public (shares information publicly, engages with the public, etc.)
Q93	… serves citizens and users well.
Q94	… makes decisions based on facts and evidence
Q97	…. Is successful in achieving its mission and goals.
Q98	… is ready and able to take on new and emerging challenges.
Q75	… takes full advantage of technology to promote innovation and new ways of working.

Note: Cronbachs-alpha value of 0.89, N=44 270 Q89 was reversed to align with the other index question where a higher value represents higher levels of organisational performance.

 

Team performance
Q83	My team or work unit works well together.
Q84	… achieves our goals.
Q85	… produces high-quality work.
Q86	… contributes positively to the organisation’s performance.

Note: Cronbachs-alpha value of 0.91. N=48 487

 

Innovation climate
Q71	My organisation conitunually encourages me to look for new ways of improving the way things work,
Q72	… supports the idea that failure is a part of innovation.
Q73	… commits resources to develop new ideas and innovations (e.g., budget, staff, time, expert support).
Q74	… ensures that teams have the necessary skills to implement innovation initiatives.
Q75	… takes full advantage of technology to promote innovation and new ways of working.
Q76	… has sufficiently flexible legislative and regulatory frameworks to incorporate new approaches and policy solutions.
Q77	… learns from past problems and takes measures to prevent them from happening again.

Note: Cronbachs-alpha value of 0.91. N=43 486

 

Learning and development
Q23	the learning activities that i have completed in the last 12 months have helped me improve my performance
Q24	I am able to access the right learning opportunities when I need to
Q25	at my work i feel that i am growing professionally
Q26	my organisation supports mobility (temporary assignments, transfers or secondments) to help develop my career.
Q80	my organisation provides regular opportunities to improve the digital skills of employees.

Note: Cronbachs-alpha value of 0.81, N=32 410 Q23 has imputed values for those employees that selected not to have undertaken any learning courses or selected ‘prefer not to respond’ in Q22. For those the average value for the organisation they belong to (answer to QH) has been imputed.

 

Job autonomy
Q47	My line manager/supervisor trusts my judgement
Q48	My line manager/supervisor lets me be reasonable autonomous (i.e. does not micro-manage)
Q66	The senior leaders of my organisation trusts the judgement of their employees

Note: Cronbachs-alpha value of 0.74. N=40 144

Recoding of variables

Table A ‎E.1 in Annex E present the variables included in the analysis of the report, some variables have been slightly recoded, i.e. answer categories merged in cases where there were too few observations in some categories (i.e. the ‘other’ category of gender) or order of answer categories changed to facilitate the communication of the results.

Regression analysis

All non-descriptive results presented in the aggregate report are based on various regression models. Models ae developed based on the analytical framework described in the Chapter 1 of this report. The data is analysed as mixed-effects models to account for the hierarchical structure of the data, where individuals are nested in organisations and countries, and potential intra-group correlation. This modelling approach incorporates both fixed effects, the relationships across the entire sample, and the random effects, accounting for unobserved heterogeneity across organisations and countries. Mixed effects models are particularly appropriate to analyse the dataset, where individual-level responses are nested within larger groupings (organisations and countries in this case), and where ignoring this nesting could lead to biased estimates and underestimated standard errors. By including random intercepts, it is possible to model the variation across groups and improve the precision and generalisability of the findings.

The analysis employs both linear and logistic regression models, depending on the nature of the dependent variable. Linear regression is applied when the dependent variable is continuous, as in models where composite indices serve as outcome variables. In contrast, logistic regression, including binary, multinomial, and ordinal models, is used when the dependent variable is categorical. For instance, binary logistic regression is applied to outcomes such as turnover intention (yes/no) and ordinal logistic regression for variables measured on an ordered scale such as sick leave (categories of sick leave days). Multinomial models are employed when the outcome has more than two categories, but categories do not have a clearly defined order (outcomes for poor performers).

Prefer not to respond answers

Table A ‎E.3 in Annex E presents the share of “prefer not to respond” (PNR) answers for each survey question. This provides useful insights into both question quality and the sensitivity of certain topics covered in the survey. The questionnaire did not include a “don’t know” option, meaning that PNR responses may reflect either difficulty in answering the question or discomfort with the topic.

On average, 3.9% of responses across all questions were PNR. Only eight questions exceeded a 10% PNR rate. The three questions with the highest PNR rates each above 20%, were: “In my organisation, poor performers usually…” (26.3%), “To what extent do you trust the political leadership of your organisation?” (23.1%), and “I want to leave my organisation within the next 12 months” (21.8%). The high PNR rates for these items likely reflect the sensitive nature of the topics. Reducing PNR rates is important, as a high proportion of non-responses can affect the reliability of the results, potentially leading to over- or underestimation of satisfaction or dissatisfaction. Clear communication about the anonymity of the questionnaire is essential to lowering PNR rates, helping to reassure respondents that their answers, especially on sensitive topics, will remain confidential.

References

[1] OECD (2023), “Slovenia”, in Education at a Glance 2023: OECD Indicators, OECD Publishing, Paris, https://doi.org/10.1787/80939f54-en.

Notes

← 1.  eTranslation (europa.eu)

← 2.  The median was chosen over the mean because it is less influenced by outliers, such as respondents who began the survey but completed it days later. This makes the median a more accurate and reliable measure of the survey completion time tendency.

← 3.  Please refer to Annex A to review the survey questions.

← 4.  The following questions were included: Section 1: QA-QJ; Section 2: Q1-Q5, Q27-Q42; Section 3: Q45-Q58; and Section 4: Q71-Q78, Q81a-h_6, and Q82-Q99.