Você está na página 1de 6

Comments on “Firm Size and

Participation in the International


Economy: Evidence from Bangladesh” by
Ben Shepherd

Ayako Obashi
Aoyama Gakuin University

February 6, 2020
“Trade, Global Value Chains and Small and Medium-Sized Enterprises”
Workshop @ADBI
Overview
• Using the WB Enterprise Survey data, the study examines the
export behavior of Bangladeshi manufacturing firms.

Two hypotheses:
• The export behavior of SMEs is different than larger firms.
• Data observations for export propensity/intensity (Figures 1-2).
• SMEs react to the GVC participation differently than larger firms
in determining the export behavior.
• Heckman sample selection model estimation (Table 5; Figures 3-4).
• Two measures of GVC participation: 1) dummy for imports of
intermediate goods; 2) dummy for majority foreign ownership.

Two major findings:


• Firm size is positively associated with export performance.
• The GVC participation measures are associated with a higher
propensity of exporting, although no significant (and even
negative) results for the export value.
Comment 1:
Panel data utilization
• Although approx. 1,300 manufacturing firms are
included in WB Enterprise Survey in the three-year
panel format (2007, 2011, and 2013),
• all the data observations and regression analyses
are conducted by using data for 2013 only.

• Could make use of the older data?


• e.g. for robustness checks?
• GVC participation measures as well as other firm
characteristics variables (productivity, size (employees),
capital intensity) could be lagged by one period.
Comment 2:
Heckman sample selection model estimated
without exclusion restriction
• At least one regressor in the selection equation may be
required to be excluded from the outcome equation
(Cameron and Trivedi, 2005).
• Especially in the case of single-year data set, the absence of
adequate exclusion restriction might cause the biased estimates.

• Any valid exclusion restriction?


• Trade frictions affecting (firm-specific) fixed costs of exporting but
not variable trade costs.
• ‘Foreign ownership’ dummy variable appears to be a good
candidate as firms may have better access to foreign markets if they
are part of a multinational network of firms (Christen et al., 2018).
• At least, need to ensure that the model is well-identified
without exclusion restriction. Report the estimated
coefficients for the inverse Mills ratio.
Comment 3:
Other concerns over the estimation and its results
1. Sector-specific effects
• Bangladeshi textile and garment sectors are outstanding in
terms of the integration into GVCs as export platform, unlike
other sectors and countries (WTO, 2019).

2. Coefficients for the GVC participation


• More interpretation is need as to why the coefficients for the
two GVC participation measures are estimated to be
significantly positive only in the selection equations.
• In particular, in the third model with VA per worker, the
importer status dummy is estimated to be significantly
negative in the outcome equation, in contrast to the
selection equation.
• At least, shouldn’t ‘Foreign ownership’ dummy variable be
included only in the selection equation??
Comment 4:
Marginal effects of GVC participation by firm size
• Interested in the overall impact of GVC participation on export
value in interaction with firm size (based on employees).

To compute the marginal effects,


• Why are the lower bounds of the number of employees for
respective firm size categories used?
• Mean/median is more representative of the sample firms in each
category?
• Are all other variables (except the employees variable) held at
their means across the whole sample?
• Firm productivity is likely to be correlated with firm size. It appears to be
plausible to take the mean by firm size category?

• The scale of the vertical axis in Figures 3-4?


• What accounts for the difference in magnitude of the marginal effect
between the two measures of GVC participation?

Você também pode gostar