The studies testing Gibrat ´s law have incorporated different variables, adding relevant
information on the characteristics associated with employment growth, such as the innovating
activity, under the assumption that innovators experience a higher increase in employment
(Licht & Nerlinger (1998); Storey & Tether (1998); Almus & Nerlinger (1999, 2000); or
Freel (2000); the age of the firm testing if the youngest grow bigger (Reid (1995); Harhoff et
al (1998), Heshmati (2000) in an explicit way, or Almus & Nerlinger (2000) and Audretsch et
al (1999) in an implicit one); industrial technological development, under the hypothesis that
bigger growth occurs in more technologically developed industries (Almus & Nerlinger
(1999, 2000), Harhoff et al (1990), Audretsch (1995), Audretsch et al (1999), or Freel,
(2000)); and agglomeration effects (Wiseen & Huisman (2003)), supporting Gibrat’s law
rejection in urban areas, where small firms will grow faster, and, at the same time, the law’s
fulfilment in not urban areas, at least for large firms. Many of those variables have been
included for the Spanish case in Calvo (2004).
The present section test Gibrat’s law using Spanish data taking into account regional
variables. The data come from the Firms Strategic Behaviour Survey for the period 1998-
2002. A sample of 1255 firms is used: 1161 of them survive, and, consequently, 94
disappeared over the 5 year period. A typical Gibrat’s equation is estimated, where last period
employment depends on first period employment and the rest of variables (innovation, age,
legal liability, industry technological development, and regional variables). Because of
sample attrition, we use the procedure proposed by Heckman (1979), estimating by maximum
likelihood, including a probit survival equation.
3.1. Model.
In order to test Gibrat’s law we use a typical equation in which employment in the last period
(2002) is dependent on employment of the first period (1998) and the rest of variables. The
original equation is:
logSi02 = β0 + β1 logSi98 + ΣjβjXj + εi02 (1)
Where Si01 is the employment of the ith firm in 2002; Si98 is the employment of the same firm
in 1998, and Xj are the other variables.
16