The independent variables included in the equations are defined in the Appendix. But the four
regional variables introduced in the equations deserve a special mention: The first one
(PIBPC) proxies the wealth of the region. So, we will expect a positive sign associated to this
variable on both equations; the second one, EMPINDSH, should have a positive sign on its
estimator, since we could suppose a better behaviour of firms on industrialised regions;
PARPYME is related to the relevance of small firms on regions employment. If size and
growth have a negative relationship at macroeconomic regional level, as we supposed at the
introduction, then the sign of the estimator should be positive; finally, IDPIB proxies the
innovation effort of Spanish regions. We could assume a positive sign associate to this
variable on growth and survival if the second hypothesis of the introduction is true.
3.2. Data.
The data used in this section come from the Firms Strategic Behaviour Survey (ESEE). This is
a survey of Spanish manufacturing firms that began in 1990 and is conducted annually for
about 2000 firms16. We use a sample of 1255 firms17: for 1161 there is full information for the
whole period, and 94 can be considered as firms that have exited the market (closed).
Some interesting results can be extracted from the comparison of survival and disappearing
firms for the different definitions18: first, surviving firms are bigger than disappearing for all
the definitions; secondly, the human capital employment definition implies smaller statistics,
as we would expect, since this definition weights employees depending on the degree
obtained; finally, there are not big differences between non weighted and weighted employees
(PERS1 and PERTOT variables). This means that most of the employment of those firms is
composed by employees with full working hours, as we can also trust if we look at the small
differences between ASATC and PERS1.
16 The questionnaire and general results of ESEE can be found in www.funep.es/PIE/ESEE/esee1.htm.
17 The original size of the sample is 1283, but there are problems of information in any of the variables related to
28 firms. That is why Table 2’ statistics use the first sample.
18 The statistics of the employment variables are upon demand.
18