Ceris-CNR, W.P. N° 10/1997
at the changes in output shares between 1987 and 1993 and if one compares the number of country
entries and exits. Analogously, with the exception of German firms, EU leading firms have decreased
the degree of diversification in their home country. This empirical fact is suggested by the descriptive
statistics on output shares and it is confirmed by the reduction in the number of industries between
1987 and 1993. Overall these results point out that: i) there is some evidence in favour of the “return
to core” hypothesis, at least as far as home diversification is concerned7 and ii) the prediction of a
reduction in multinational activity in primary industry, motivated by the possibility of exploiting plantlevel scale economies is not supported by the data.
However, for more convincing empirical evidence, a more rigorous approach to hypothesis
testing is required. The starting assumption of the empirical work presented in this section is that
European integration and, more in general, “globalisation” are associated with the toughening of
price/product competition and that this association is stronger in industries where non-tariff barriers
were important before the completion of the SEM program, i.e. the so-called forty “most affected”
industries.
With this perspective in mind, we present four sets of exploratory econometric estimates
where dependent variables are respectively: i) number of entries in foreign countries - primary
industry (FC-PI), ii) number of exits from foreign countries - primary industry, iii) number of entries
in home country - secondary industries (HC-SI), iv) number of exits from home country-secondary
industries. Since by construction all dependent variables can assume only non-negative discrete
values, we make use of the Poisson regression model for estimation purposes8. We then assume that
dependent variables Y1, Y2,...., Yn have independent Poisson distributions with parameters, λ1, λ2,
...,λn, respectively. Hence,
Prob (Yi=yi) = λyie-λi/yi!
(1)
6
The somewhat unexpected exit figure in the home country primary industry (second row) is due to four firms
which apparently abandoned the industry which accounted for their largest EU share of output in 1987.
7
From our previous empirical work, it emerges that it is more difficult to explain “home” diversification than
“foreign” diversification . This is turn might suggest that “home” diversification has often no strong industrial
logic.
8
In the Poisson model the variance is assumed to be equal to the mean. For all equations we also estimated
negative binomial models where overdispersion can be tested. For all reported equations the hypothesis of no
overdispersion is not rejected by the data.
11