“Many OECD countries have developed and currently utilise a profiling tool which predicts the risk of Long-term unemployment (LTU). Pioneers in the development of this tool in the early 1990s included the US, Australia, Canada and the UK followed in later years by Germany, Denmark, Switzerland, Ireland, France and the Netherlands amongst others” writes Mgr. Tomas Soukup in FORECASTING OF LONG-TERM UNEMPLOYMENT AT THE INDIVIDUAL LEVEL on cejpp.eu. (Adapted Excepts by Job Market Monitor)
LTU has a negative impact both on the jobseeker and on public finances. Labour market policy therefore encourages both the preventive and proactive approaches in order to avoid negative impacts. Unfortunately, a large number of evaluation studies show that active intervention is helpful only if it is targeted according to the prevailing situation and needs of claimants. The first step in the targeting process is to determine in advance which claimant has a significant probability of becoming long-term unemployed and just how high the risk is.
The paper deals with the forecasting of long-term unemployment at the individual level. The author looks at the potential for LTU statistical forecasting in the Czech Republic concerning which the author poses the following three principal research questions: Firstly, what are the main factors which determine the length of unemployment? Secondly, how accurate can the model be if it is based solely on data from claimant registers (registers used by labour offices)? And thirdly, are there any indicators which on the one hand might enhance accuracy and on the other are easy to discover?
The results would seem to indicate that it is possible to predict the length of individual unemployment fairly accurately using register data. The ratio of correctly identified claimants is 78% which seems to be quite high, especially when compared to other available tools such as defining the risk group the accuracy of which stood at 58% of correctly identified claimants.
In order to answer the three questions, statistic model was constructed based on the Job Search Theory and tested on official register data.
Answering the first question was not easy. There are a host of indicators which must be included and which can be divided into six groups – local unemployment rate, structure of unemployment, discrimination of employers, human capital, nominal flexibility and plans and strategies. Clearly, register data provides merely a substitution for real factors. For example, registers contain a limited amount of information on human capital (education and occupation), but no information on knowledge and skills. Other information, such as the level of motivation or job search intensity, is also lacking. On the other hand, information on the length of the last period of unemployment might, at least, provide an indication of human capital and motivation. More generally, certain observable hard variables might provide a substitute for certain required unobservable soft variables.
As for the second question, it is possible to predict fairly accurately the probability of finding new job using data from official register. If the probability is divided into two groups (cut point of 50%), 78% of claimants were correctly identified. Employing a higher number of cut points provides even more accurate results (e.g. 94% of claimants with the highest probability – 91%-100% – found new jobs). Further, it would seem preferable to construct a number of local models rather than one national model.
Finally, three indicators have been identified which increase the model’s level of accuracy whilst at the same time being relatively easy to determine: 1) a subjective assessment of one’s own chances of finding a new job, 2) the promise of a new job and 3) willingness to change one’s area of work.
Read more @ FORECASTING OF LONG-TERM UNEMPLOYMENT AT THE INDIVIDUAL LEVEL by Mgr. Tomas Soukup