Detecting learning‐by‐exporting effects on firms' productivity distribution by accounting for heterogeneous macrofactors and panel attrition

• Author: Maria R. Ferrante,Marzia Freo
• Date: 01 September 2019
• Published date: 01 September 2019
• DOI: http://doi.org/10.1111/twec.12807

World Econ. 2019;42:2745–2773. wileyonlinelibrary.com/journal/twec

|

2745

© 2019 John Wiley & Sons Ltd

Received: 27 April 2017

|

Revised: 2 April 2019

|

Accepted: 11 April 2019

DOI: 10.1111/twec.12807

ORIGINAL ARTICLE

Detecting learning‐by‐exporting effects on firms'

productivity distribution by accounting for

heterogeneous macrofactors and panel attrition

Maria R.Ferrante1

|

MarziaFreo2

1Department of Economics,University of Bologna, Bologna, Italy

2Department of Statistical Sciences,University of Bologna, Bologna, Italy

KEYWORDS

macroeconomic cycle, panel selection bias, postentry effect, quantile effect, total factor productivity distribution

JEL CLASSIFICATION

F14; D24; C21; C23

1

|

INTRODUCTION

The seminal empirical work of Bernard and Jensen (1995) shows that exporting firms exhibit higher

productivity than non‐exporting firms within the same industry; since then, many researchers have

investigated the causal relationship between exporting and productivity at firm level. A wide stream

of empirical literature supports the finding that the most productive firms undergo a self‐selection

process to enter foreign markets. By contrast, evidence regarding the hypothesis that firms experience

an increase in productivity during the period following their entrance into international markets (the

so‐called learning‐by‐exporting, LBE, effect) is “mixed.” For detailed reviews on empirical findings

on the LBE hypothesis, see International Study Group on Exports and Productivity (ISGEP, 2008),

Manjón, Máñez, Rochina‐Barrachina, and Sanchis‐Llopis (2013), Martins and Yang (2009), Silva,

Afonso, and Africano (2012), Yang and Mallick (2014) and Wagner (2007, 2012b).

To provide additional details regarding this controversial picture, two main approaches have been

adopted. The first approach is from ISGEP (2008) that coordinates a large applied study. This group

uses comparable microdata for 14 countries along with a common specification for the empirical

models and finds strong evidence in favour of the self‐selection of more productive firms into interna-

tional markets but insufficient evidence in favour of the LBE hypothesis. The exporter premiums are

found to differ considerably across countries even when empirical models are identically specified.

Furthermore, ISGEP notes that “the number of export starters that can be monitored with the data

set for [the] study is too small for most countries to offer a solid basis for empirical comparisons,”

2746

|

FERRANTE ANd FREO

and there is a shortage of observations against finding solid results (pp. 617). Regarding the second

approach, two meta‐analyses have been run to investigate the mechanisms through which LBE may

be explained. Martins and Yang (2009) explore topics related to sampling and methodological hetero-

geneity and find that the LBE effects are higher in developing than in developed economies, higher

in the first year that firms start exporting and lower when the self‐selection of better firms into inter-

national markets is accounted for. More recently, Yang and Mallick (2014) integrate the analysis in

Martins and Yang by enlarging the previous meta‐analysis to several country‐specific macroeconomic

dimensions, such as external demand, relative prices, gross domestic product (GDP) growth, inflation,

trade openness, foreign direct investment inflows, financial reform and crises. The main findings are

that the bigger the external demand and the higher the competitiveness, the higher LBE effects are

detected, and that higher returns from overseas productions and periods of financial crisis lower the

estimated LBE effects.

This paper is a further contribution that aims to verify the presence of the LBE effect on productiv-

ity growth going back to the firm level. The novelty lies in designing an ad hoc counterfactual study

which, in addition to microlevel factors, allows explicit accounting for heterogeneous cycles of inter-

national and domestic macrofactors and for the attrition bias that currently affects longitudinal studies.

In our opinion, macroeconomic factors, sample attrition and pre‐entry heterogeneity at microlevel

have to be managed jointly with causality because they act as confounding factors at the same time

the causality, if any, intervenes. Neglecting these aspects will provide biased estimates of LBE effect.

Our contribution is in explicitly considering these confounding factors in the estimation process. At

the best of our knowledge, the literature on LBE attributes importance to these aspects but it does not

deal with them jointly.

Our research is based on the following choices. First, differences between exporters and nonexport-

ers are analysed from the perspective of the evaluation literature by referring to counterfactual meth-

odologies. Second, the LBE effect focus is enlarged to cover the whole probability distribution of the

net premium by examining differences in the whole distribution of the outcome, that is by evaluating

the effect on different quartiles, rather than the effect on average.

As an outcome variable, we consider the growth in total factor productivity (TFP), a measure

widely used as a performance benchmark to rank producers or to evaluate their rate of performance

over time. The LBE effect is evaluated by estimating whether the TFP of firms starting to export in-

creases faster, after their entrance into international markets, compared with the TFP of other firms,

displaying an export premium on TFP. The premium is defined as the percentage of the differential in

TFP growth between a starter exporter in the period subsequent to the firms' entrance into the inter-

national market, the so‐called postentry period, and control groups of non‐starter exporter firms. We

originally distinguish two groups of control firms, domestics and incumbents, and investigate whether

firms that move from the status of non‐exporter to that of exporter (starters) experience an increase in

their TFP during the period following their entry into the export market with respect to (a) firms oper-

ating only in the domestic market (domestics) and (b) firms that entered international markets before

the considered period and continue to export for the whole period (incumbents).

Concerning the macroeconomic heterogeneity, the standard assumption that all the firms are

subject to the same macroeconomic circumstances, known as the “common macrotrend” restriction

(Blundell & Costa Dias, 2000), is relaxed in this paper for two reasons. First, domestic firms have to

mainly exist in the internal macroeconomic environment, and starter firms, when they begin to export,

and incumbents are subjected to internal and international macroeconomic factors. Second, as we

show later, by following our design, we identify cohorts of firms; we then align firms according to the

period they begin to export, and we run the study by comparing firm observations generated during

and affected by different time circumstances. In both cases, the assumption of a common macrotrend