Learning-by-Exporting? Firm-Level Evidence for UK Manufacturing and Services Sectors

generate first-stage predicted values of the probability of exporting, with the second
stage estimation of Equation (5) including the sample selectivity correction terms
from the first-stage model. That is, if P_it is the predicted propensity score of exporting
for firm i at time t (cf. Equation 3), then the inverse Mills ratios (or selectivity terms)
from this model are give by:

ʌ ʌ

λ_0it = —^Ф(P^ )    if Export = 0;    λ_1it = ^φ(^i^t )    if Export = 1          (6)

1- Φ( P^t )                             Φ( P^t )

These selectivity terms ( λ₀ and λ₁ ) enter Equation (5) to control directly for the
correlation of the error term in the model determining TFP with the error term in the
model determining whether the firm exports or not.

Several authors (Puhani, 2000; Smith, 2004; Angrist and Krueger, 2001) point out the
problems associated with the Heckman approach, such as the need for exclusion
restrictions otherwise the model may only be identified through the nonlinearity of the
selectivity parameter included in the second stage equation. In summary, we choose to
test for the relationship between exporting and productivity using all three approaches,
viz. an IV approach, a control function approach, and a matching approach.

V. Empirical Modelling and Results

To examine the self-selection hypothesis, we estimate Equation (3) using weighted

FAME data, in a probit model to determine which firms exported at any time during
1996-2004. Because of space constraints, the results for 16 industry sub-groups are
not reported here but are available from the authors.²⁸ Our estimation results show

that larger firms are much more likely to engage in exporting; and firms with higher

See http://www.gla.ac.uk/departments/economics/staff/pdfs/Table S1 .pdf

20

<<   17   18   19   20   21   22   23   >>

More intriguing information