As regards regional levels, we use here very simple bivanate regressions (tables 11 and 12) between growth rates of real GDP and 01 employment/unemployment level growth, and (ii) ER, UR and LTUR dynamics.
In these models, the size of the GDP coefficient is remarkable especially for the unemployment indicators, suggesting that a 1 % growth in GDP would be associated with a reduction in UR of about 2% and of more than 3% in LTUR (of their initial levels, e.
The following results (statistically significant) arise in models 5 and 6 considering, as explicative variables, changes in the main institutional variables: (i) changes in ALMP have a positive relationship with changes in ER and a negative one with changes in UR and LTUR; (ii) the opposite relationship (with respect to ALMP) occurs for changes in passive labour policies; (iii) changes in the degree of product market regulation have a negative relationship with changes in ER and a positive one with changes in UR and LTUR; (iv) the opposite relationship (with respect to PMRI) occurs for changes in EPL; (v) as regards changes in bargaining indicators (40), growth of centralisation is positively associated with ER growth and negatively with that of UR and LTUR
The introduction of a quadratic term explores the possibility of a U-shaped relationship (in dynamic terms) for the degree of centralisation, which does not emerge as statistically significant (41); last, (vi) changes in union density have an inverse relationship with changes in ER and a positive and significant one with changes in LTUR.
In the last two models (7-8), we consider changes in labour costs in addition to the main institutional variables, with the following results: (i) the signs of the parameters of the institutional variables are not affected by the introduction of the new variable; (ii) changes in labour costs have a negative relationship with ER and a positive relationship with changes in UR, whereas effects on changes in LTUR are not statistically significant.
Among the economic and structural variables presumed to influence ER (table 18), a negative impact is played by LTUR and population density.
Specialisation in business services is the only significant variable among the industry indicators; GDP and HOURS_corr are not significant; population density decreases long-term unemployment, and stronger ACT_POL are associated with lower LTUR The positive and significant relationships with PMRI and union density are also confirmed for LTUR; EPL is never significant.
5% on average; UR and LTUR decreased in 12 out of 15 countries, by -24.
C) in the whole period considered, clear beta convergence emerges for ER, UR and LTUR (in ER terms, this means convergence towards the main quantitative objective of the European Employment Strategy); sigma convergence also occurred in the whole period for ER and started later for UR and LTUR (since 2000 and 2001, respectively), the latter two showing much higher dispersion than the former;
G) transition matrixes highlighted a widespread lower persistence probabilities for UR and LTUR with respect to ER;
H) lowess beta convergence dynamics emerged for all three indicators; sigma convergence occurred for ER (throughout the period) and for UR and LTUR (since 2001), the latter two showing much higher dispersion than the former.
6 of Herefordshire, Worcestershire and Warwickshire, UK) and LTUR (13.