Modelling replacement demand for labour using microsimulation approach – case of Slovakia

In this paper we build on an already developed labour market model (VZAM) (Tiruneh, 2012), (Lubyova, Stefanik et al., 2015), designed to project skill shortages on Slovak labour market. Within the model several relatively autonomous modules work on the supply and demand side. Existing approaches in modelling labour market development in terms of skills, differentiate between expansion demand and replacement demand for labour (CEDEFOP, Skills supply and demand in Europe, 2009, 2010 and 2011). Replacement demand captures the demand arising from transitions and implied job openings. Transitions considered are those between segments (sector/occupation) as well as into unemployment and various forms of inactivity (retirement, schooling, maternity leave and other). Replacement demand usually presents a major part of the demand created in each labour market (Kriechel and Sauermann, 2010), (Kriechel, 2013). Similar approach has been also applied in Slovakia (Radvansky, Miklosovic and Hvozdikova, 2016). The objective of the submitted paper is to explore the possibilities of switching the module on replacement demand from a semi aggregate probability model into microsimulation framework. For this purpose open-source microsimulation software LIAM 2 will be employed (http://liam2.plan.be/). Switching into a micro-simulation form could be related to several advantages in comparison to a simple probability model. It would, for example, allow us to consider individuals´ history in the probability functions on transitions and thus improving the quality of predictions. The functionality of LIAM 2 also allows us to consider more and multiple stage processes, relevant from the perspective of individual decision. The objective of the paper is to explore and consider the advantages related to switching an already existing replacement module (based on age specific simple probability functions) into a microsimulation framework. Outputs will be compared to previously obtained methods and their reliability will be discussed. The demand side of the VZAM model follows this distinction between expansion and replacement demand. Expansion demand is modelled using a dynamic CGE model (Miklosovic and Radvansky, 2015). This is complemented with a methodologically different module on replacement demand. In this module, age specific probability functions are used in order to predict transitions into retirement and outside the labour force. The objective of this paper is to switch the module on replacement demand into a micro simulation framework, in order to predict transitions between segments of the labour market (sectors/occupations), into unemployment and out of the labour force. Aggregate outputs from this module are consequently produced and imported into other modules of the VZAM model. The supply side of VZAM predicts detailed structure of the educational structure of the Slovak population combining LFS, Census and administrative data on schooling participants. The replacement module also imports aggregate information, such as the one about relative wage in segments, as well as the educational structure from other modules of the model. Within the replacement module, individual data from the Labour Force Survey (LFS) are processed and complemented with more precise information from national administrative data. In the paper we will focus purely on the outputs from the replacement module, but we will keep existing functional interlinkages with other parts of the model. The results obtained by the new (micro simulation) version of the replacement module will be extracted in a form comparable to the previous applications. Basic indicators of flows and transitions into retirement, other forms of inactivity, unemployment as well as into other segments of the labour market will be extracted. These will be produced by sector of economic activity, age group and educational level. Predictions from the new version of the module will be confronted with results from the previous version and especially the fit with real LFS microdata will be considered (with respect to LFS sample related shortcomings). The most recent (2015) round of LFS will be considered in order to assess the nowcasting potential of the designed approach.