industry resources to be reallocated to higher productivity exporters; and the
shutdown of lower productivity firms - both exporters and more likely non-exporters
with the lowest productivity level, as predicted by some recent theoretical models
(Bernard et. al., 2003; Melitz, 2003).
Nevertheless, there has to date been little micro-based evidence for the UK that
quantifies the importance and contribution of exporting to overall UK productivity
growth and thus substantiates these associated benefits. This paucity of UK evidence
on the causes and impact of internationalisation can largely be explained by data
limitations, owing to the lack of any information about export activities in the Annual
Respondents Database (ARD) - the primary source of micro data collected by the
Office for National Statistics (c.f. Harris, 2005). There have been a limited number of
studies for the UK that have considered both whether exporters are ‘better’ than non-
exporters, and whether there is any post-entry productivity improvement to exporters
(e.g. Girma et. al., 2004; Greenaway and Kneller, 2004; Greenaway and Yu, 2004).
These analyses have used data from the FAME and OneSource databases based on
returns firms have to make to Companies House in the UK, but there are a number of
issues that arise from the use of these data, principally that the samples used in
statistical analysis are not representative of the UK population of firms and as a result
large firms are over-sampled.5
Thus the scope of the current study is to use appropriate micro data sources for the
UK to assess the extent to which productivity growth within firms may be stimulated
5 The FAME and OneSource databases are not based on samples drawn from the population of firms in
production in the UK (since only firms above a certain size have to make returns with data that are then
used in statistical analyses), and thus they contain no information on the UK population of enterprises.
Meanwhile, as these sources are based on accounting definitions of variables, they do not always relate
to the definitions assumed when estimating economic relationships, such as the production function.
There are also issues concerning how well entry and exit are captured in these data. All of these points
are returned to later in Sections III and IV.