What is LEE(P) data?

Sources and types of LEE data

There are two main types of data sources used to create linked employer-employee (LEE) datasets. Historically, the earliest LEE datasets were based on wage surveys regularly conducted by employers. These surveys were primarily designed for statistical purposes, providing valuable insights for researchers and policymakers. One of the first examples was France’s 1966 Enquête sur la Structure des Salaires (ESS), and similar datasets emerged across Europe in the late 1970s.

Survey-based data collection offers highly detailed information about employees' working conditions. It can include job-related details such as hours worked, overtime, and bonuses, as well as employees' qualifications, skills, and workplace conditions. Surveys also allow for a broader range of questions, helping to capture economic trends more comprehensively.

However, surveys have limitations. They typically cover only a sample of firms or employees, and conducting them is costly. They also only capture employees at the time of data collection, excluding those who are unemployed, self-employed, or inactive. Additionally, participation in surveys has been declining in recent years, making it harder to collect reliable data.

To overcome these challenges, researchers have increasingly turned to administrative data sources. These datasets are constructed by linking records from tax authorities, pension systems, and other public institutions. While primarily collected for government operations, they have also become a valuable resource for economic research. Early examples of this approach appeared in the United States, where researchers used social security records to study labor market trends. Since then, many countries have developed similar databases.

Administrative data offer several advantages. They are more cost-efficient than surveys since they do not require repeated data collection. They also provide a more complete picture of the labor market, covering both employed and unemployed individuals over time. Additionally, because they come from official records, they are generally considered more reliable than self-reported survey responses.

However, administrative datasets also have limitations. They only include information recorded by public institutions, which means some details—such as certain types of bonuses or overtime pay—may be missing or difficult to separate from other income components. They often lack insights into working conditions, family background, and education history unless such data are also collected by public agencies. Furthermore, they do not cover informal employment, which plays a significant role in some economies.

Two important points should be considered. First, with the widespread adoption of digital technologies, firms increasingly use the same data for both wage surveys and administrative reporting, improving the consistency of survey-based datasets. Second, some countries conduct economy-wide surveys where all firms report workforce data. While technically classified as surveys, these datasets (such as Portugal’s Quadros de Pessoal) share key characteristics with administrative records.

Given these factors, a more useful way to classify LEE datasets is not by their data source but by whether they allow tracking individuals over time. Datasets that enable such tracking are known as longitudinal LEE data or LEE panel data (LEEP), while those that do not link individual records over time are considered cross-sectional LEE data. The latter primarily includes survey-based datasets, whereas longitudinal LEE datasets can be based on both surveys and administrative records.