Synthetic Data Generation and Data Analysis for Actuarial Assessment of Public Sector Pension Schemes Terms of Reference 1.Background and Rationale The multidimensional crisis that erupted in Lebanon in late 2019, marked by a sharp devaluation of the Lebanese currency, triple-digit inflation, severely constrained fiscal capacity, and rising unemployment, has dramatically increased the strain on the country’s social protections systems, notably its pension systems. As a result, the real value of pension benefits has significantly diminished, eroding retirees’ purchasing power and exposing deep-rooted structural deficiencies in terms of coverage, adequacy, equity, and financial sustainability. Public sector pension schemes—primarily serving civil servants and military/security forces—are under increasing fiscal pressure and face systemic imbalances. Meanwhile, the private sector’s End-of-Service Indemnity (EOSI) system has proven inadequate, prompting the recent enactment of a new legislation to establish a new pension scheme for private sector workers. Within the public sector, two main pension schemes exist: one for civil servants and another for armed and security forces. In total, the number of beneficiaries across both public pension schemes amounts to 131,000, including 93,300 retirees and 37,700 survivors. The military and security forces account for the largest share, with 90,655 beneficiaries, while the remainder is distributed among educational staff, judges, public administration employees, and other public authorities. While the design of these schemes was generally considered generous, benefits were unevenly distributed, with significant variation across different groups. The variation was driven by differences across schemes in benefit calculation formulas, eligibility conditions, and other scheme-specific parameters. Additionally, as of 2024, the active public sector workforce is estimated to include approximately 33,465 civil servants, in addition to around 120,000 military and armed forces personnel. Considering these challenges, and in the context of recommending parametric reforms, an actuarial assessment is essential to evaluate the financial sustainability and impact of proposed measures. Such an assessment will enable policymakers to better understand the demographic and financial characteristics of scheme members and assess the sustainability and equity of current provisions. The work will contribute to Lebanon’s broader pension reform agenda, which seeks to transition toward a more coherent, inclusive, and fiscally responsible system aligned with the National Social Protection Strategy. While this actuarial assessment requires high-quality, granular data, only aggregated high-level data is currently available, hence the the ILO, under the project “Supporting Social Security and Institutional Reforms towards a Strengthened Social Protection System in Lebanon,” funded by the FCDO, is seeking to engage a qualified company to generate synthetic granular datasets simulating the characteristics of the population. Additionally, available data must be reviewed, calibrated, and prepared to meet the input requirements of a basic actuarial assessment. This service contract will support the actuarial assessment process through two key assignments: Generation of synthetic individual-level data. Review, preparation, and aggregation of available data for actuarial analysis. 2.Objectives Assignment 1: For each pension scheme, generate two sets of synthetic individual-level data for active members and pensioners of the public sector pension schemes. Assignment 2: Conduct a comprehensive data analysis to assess data quality, calibrate individual data using financial aggregates, and prepare the data for actuarial modelling. 3.Assignment 1: Synthetic Data Generation Scope of Work For each pension scheme, the company will generate two synthetic datasets that need to reflect the demographic and financial characteristics of: Active Members: Age Gender Monthly salary Credited months of contribution Pensioners: Age Gender Monthly pension amount Years at retirement (or retirement age) Data Sources and Security The company will use the following datasets and data sources that will be provided by the ILO: Labour Force Surveys (2018-19 and 2022) including the distribution of surveyed scheme members by gender and age. Salary scales by grade and step (for each year from 2019 to 2024). Benchmark distribution of active members (from another country) by gender, age and service period (or number of months of contribution). Total number of active members and pensioners (for the years 2022, 2023 and 2024). Total pensions and benefits paid (for the years 2022, 2023 and 2024).   In addition, the company can consider using data from other countries to inform assumptions as applicable.   The company will maintain confidentiality and security of all sensitive data and comply with all relevant data privacy and security regulations. Deliverables For each pension scheme, two anonymized synthetic datasets (active members and pensioners) Documentation of the methodology used for data generation Summary statistics comparing synthetic data with available aggregates 4.Assignment 2: Data Analysis for Actuarial Assessment Scope of Work The company will: Review the quality of available individual and aggregate data. Calibrate individual-level data using financial and administrative aggregates. Prepare the data to meet the input requirements for a basic actuarial assessment. Maintain confidentiality and security of all sensitive data and comply with all relevant data privacy and security regulations. Tasks Data Quality Assessment Assess completeness, consistency, and validity of available data Identify and document data limitations Calibration of Individual Data Align individual-level data with financial aggregates (total contributions, pension expenditures) Adjust for inconsistencies between individual and aggregate data Data Preparation Prepare data by: Gender, age, salary, service period, and pension payable Reason for retirement (normal retirement, early retirement, disability, etc.) Type of pensioner (retiree, survivor) Prepare summary tables and visuals as per the requirements of the actuarial assessment. Deliverables Final data set that meets input requirements for a basic actuarial assessment Brief technical note summarising data quality assessment, adjustments, and summary tables and visuals. 5.Methodology The company is expected to use statistical techniques for data generation and analysis. All assumptions and methods must be clearly documented. Visualizations for the data analysis assignment should include: Frequency distributions Aggregated statistics by demographic and service characteristics Charts illustrating disaggregation by two variables where relevant 6.Qualifications and Selection Criteria The selected company must possess the following qualifications and experience: Demonstrated experience in designing and generating synthetic microdata sets using probabilistic modelling, statistical inference, or machine learning techniques. Ability to simulate complex population structures and employment-related variables, such as grade, step and career progression, using both cross-sectional and longitudinal modelling techniques. Proficiency in aligning outputs with both quantitative and qualitative sources, including national public sector aggregated indicators and statistics, financial data on wage bill and pension expenditures, labour force surveys, as well as legal and regulatory provisions relevant to pension and salary scale. Ability to produce clean, well-documented, and user-friendly datasets and outputs ready for use by actuarial teams and policy analyst. Capacity to meet project deadlines and deliverables. Demonstrated commitment to adhering to strict data security and confidentiality protocols. Familiarity with the Lebanese demographic context, public sector structure, and applicable legal and regulatory provisions, particularly those governing social protection and pension systems is considered a strong advantage. Previous experience working on social insurance or pension scheme data will be considered a strong advantage.   7.Timeline and Deliverables The assignment will commence, tentatively, on 20 August 2025 and needs to be concluded no later than 5 September 2025.   Deliverables Payment Percentage Deadline Assignment 1: For each pension scheme, two anonymized synthetic datasets (active members and pensioners) Documentation of the methodology used for data generation Summary statistics comparing synthetic data with available aggregates 50% 29 August 2025 Assignment 2: Final dataset for basic actuarial assessment Technical note 50%   5 September 2025   8.Application Process Interested companies are invited to share their technical and financial proposals with the ILO Lebanon Social Protection Team by 15 August 2025 via email to [email protected] and [email protected] (subject: “Synthetic Data Generation and Data Analysis for Actuarial Assessment of Public Sector Pension Schemes”). The technical proposal must be submitted in English and include the following: (i) a Letter of Presentation/Cover Letter, including a portfolio of previously completed similar assignments; (ii) a Curriculum Vitae (CV) highlighting relevant experience; (iii) a brief technical note outlining the proposed approach to the assignment; and (iv) at least two references with contact information, along with permission for the ILO to contact them.