Record Linkage Between Italian Administrative Sources and Sample Surveys - How Much Information We Can Get? Three case studies

The aim of Record Linkage is to join information referred to the same individual but stored in multiple datasets. Similarly to other data integration techniques, record linkage let a better exploitation of existing data by reducing both the respondent burden and the costs associated with the implementation of a new survey. Record Linkage techniques are commonly used in several disciplines (such as biology, economics, medical research, …) and performed by combining information gathered in two or more datasets that refer to the entire population or by linking survey data with an exhaustive source. At the same time multiple goals can be achieved by using data integration techniques. The most common of those are: duplicate record detection, building longitudinal datasets, validate new variables, study the relationship between variables collected in different datasets, estimate the unknown size of a population. Finally, both deterministic and probabilistic approach can be used. The present paper seek to highlight this complexity by illustrating the goals and the main results of three different case studies. All of these studies share: (1) the field of application (that is demography), (2) the originality of the studies since none of the corresponding research questions could be answered by using available (non-linked) data (3) one of the data source used to perform the linkage procedures. The latter is the Survey on Live Births: the individual form used to register every live birth delivered by the Resident Population includes several information on births (newborn’s name and surname, sex, date and place of birth, and citizenship), parents (name, surname, date of birth, citizenship, and marital status) and the main details of the head of the household. * Case study A: The aim of this research was to investigate the relationship between expected and actual fertility at individual level in the short run in Italy. By using deterministic record linkage procedures an ad-hoc longitudinal dataset was built and this makes possible to answer to the research question posed. The data sources used are the Birth Sample Survey (run in 2003) and the Survey on Live Births (we considered the live births registered during the period 2003-2008). * Case study B: the object of this work was to determine the relationship between socio-demographic variables and medical and health aspects related to pregnancy and delivery. In order to do so the linkage between three different data sources was needed. These are: the Certificates of Healthcare at Delivery (CeDAP), the Birth Sample Survey and the Survey on Live Births. In addition, the Record Linkage allowed us to control the quality of the information collected with the CeDAP and validate and correct selected variables. * Case study C: the purpose of this research was to study the reproductive behavior of foreigners women who live in Italy at micro level using a longitudinal approach. In order to do so the linkage between the Survey on Live Births and the Register of Residence Permits for the period 2002-2006 had been performed. Note that in this work probabilistic record linkage techniques were used