Uncovering relationality from transactional data using dynamic graphs: case study from activation labour market policy in Italy

Abstract This contribution aims to show how we can use existing ‘transactional’ administrative data to rediscover the relational side of result based policy programmes, in order to improve our understanding of relational dynamics, develop innovations in labour activitation and inform future evaluation. In current orthodox statistical approaches and measures we make use of administrative data to show whether an intervention is working and is geared towards quantitative events and variables. Such analysis tends not use any (or a few) qualitative elements to support the perception of change in user behaviors or to describe the meaning and the direction of the interventions. Working with labour activation data from Italy we posed the question of whether dynamic graph analytics can be used to represent the evolution of relationships between actors involved in a policy programs. And is so how are the evolution of relationships related to programs results and outcome The paper will show how the methodology and the set of indicators made it possible to capture relevant moments of structural change in the graphs. We use dynamic graph analytics to understand the evolution of events during the life of people involved in welfare support programs. The information about the events are extracted from administrative databases . The information about some relevant relationships (employer-employee, employer-public employment service, ….) including starting and ending dates, are extracted and represented as graphs. The observation of the dynamics of the graphs are made possible by its graphical representation and by underlying statistical models, used to capture the evolution of the relationships with a set of indicators. Relational and non-relational programs are presenting different patterns, suggesting further evolution in the evaluation practices. Workers coming from a fragmented career, showing frequent short term contracts, are often accessing support programs after being fired by a restricted number of employers. The tools used to assess “the responsibility” or the accountability of the program’s results should be redefined when analytics suggests that accountability is clearly shared by many different actors acting as a network. We argue that this kind of analytics are closer to a description of complexity as relationships are the object used to describe the evolution and the effect of relational oriented service in comparison with transactional services.