exported product (PSE) are functions of some observable characteristics of the firm and
of the entrepreneur as:
yi = f (xi) (3)
Both these percentages are bounded from below by 0 and above by 100. The most
appropriate econometric formulation of the relationship in equation (3) is a tobit model
as:
Г0
У, =Ixiβ + εi
100
if 0 ≥ xiβ +εi
if 0 < xiβ +εi < 1
if 1 ≤ xiβ +εi
(4)
where the yi is either the variable PMIR or the variable PSE in turn, while the vector xi
includes network characteristics, and enterprise and entrepreneurial characteristics of
the firm. The marginal effects of the tobit model presented in equation (4), given
censoring from below at 0 and from above by 100 is given by:
'r[y Xi] = β[φ(z,00 )-φ(z■ 0 )]
∂xi
z0 =- xiβσ (5)
z100 = (100 - xiβ)/σ
where Φ(.) denotes the cumulative normal distribution function and σ the variance,
while Φ(z100)- Φ(z0) represents the probability of observing a noncensored
observation.
Furthermore, we assume that the percentage of material inputs used by a firm and
produced by local firms affects various dimensions of performance. Due to the dummy
nature of the variables showing performance, an appropriate econometric formulation is
a logit model as:
β'xi
ei
= Λ(β'xi)
(6)
Pr ob (У = 1, Positive Change of Performance Indicator) =-----—
1 + eβxi
where У can be in turn the variables indicating performance (PEREMP, PERPM,
PERTS, PERINV), xi is a vector of factors influencing performance and including,
among others, the percentage of material inputs produced by local firms or the
percentage of exported product, β is a vector of parameters to be estimated by the
model and Λ(.) indicates the logistic cumulative distribution function. The log-
likelihood function for the logit model in equation (6) is estimated as:
14