The structural resources available for exporting were

The structural resources available for exporting were measured by there being an
export department. The evaluation of the competitive intensity was made according
to the work of Cadogan et al. ( 2012 ), being represented as the extent of rivalry
among different players in an industry. This is a first order reflective construct and
was measured through a five-point Likert type scale, allowing managerial percep-
tions of the variables analysed to be picked up. Finally, export performance is
considered a second-order formative construct composed of two very different
dimensions, growth in sales and satisfaction. Following Cadogan et al. ( 2002 ),
the qualitative dimension of export performance was evaluated through export
managers’ perceived satisfaction over the achievement of five objectives in the
last 3 years: growth of international sales, firm image and notoriety in foreign
markets, profitability of the export business, market share and international expan-
sion. The quantitative dimension was measured through growth in export sales in
the last 3 years (Cavusgil and Zou 1994 ; Navarro et al. 2010 ).
8 A. Navarro-Garcı´a and M. Peris-Ortiz

1.5 Results
1.5.1 Evaluation of Measurement Model
Two different stages were carried out to interpret and analyze the model proposed
using PLS (Barclay et al. 1995 ): (1) the evaluation of the measurement model; and
(2) the analysis of the structural model. This sequence ensures that the measurement
scales proposed are valid and reliable before testing the hypotheses. For the
reflective scales, the factor loadings were all above the recommended 0.7 score
(Carmines and Zeller 1979 ). The composite reliability and average variance
extracted (AVE) values also exceeded the recommended values of 0.7 and 0.5,
respectively (Fornell and Larcker 1981 ). Thus, the results support the convergent
validity of the reflective scales considered in this study (Table 1.1 ). Then, to ensure
the discriminant validity, the authors confirmed that the squared correlations
between each pair of constructs did not exceed the AVE (Barclay et al. 1995 ). It
was also checked that the inter-correlations between constructs were significantly
different from 1, which provided additional evidence of the discriminant validity. In
addition, none of the correlations between constructs reaches the 0.5 score
(Table 1.2 ).
1.5.2 Testing Hypotheses: Parameters of the Structural
Model
After having ensured the convergent and discriminant validity of the measurement
model, the relationships between the different variables were tested. The different
statistical parameters were calculated using the bootstrap method (1,000 subsam-
ples) (Table 1.3 ). Although many researchers opt for computing 500 subsamples in
their studies, and this is enough, in the current work it was decided to use 1,000 to
reduce the randomness (Davidson and Mackinnon 2000 ). The hypothesis tests
considered the sign and significance of the t-values in each relation ( β coefficient).