Services Development and Comparative Advantage in Manufacturing

III. Empirical analysis

III.1 Empirical strategy

In our empirical analysis, we use RCA to measure the export competitiveness of individual manufacturing sectors. We will explain later in this paper how we modify the conventional definition of RCA after stating our specification.

To test Hypothesis 1, we estimate the effect of services development $D$ on manufacturing export performance (RCA), and analyze how this effect depends on service input intensity as measured by the ratio of embodied domestic services in total final demand to manufacturing valueadded (or simply $SII$ ); see a later subsection for more details). Our baseline regression specification is:

(1) $R C A_{i s t}=\beta_{0}+\beta_{1} D_{i t}+\beta_{2} S I I_{i}+\beta_{3} D_{i t} * S I I_{i s t}+Z \boldsymbol{\gamma}+a_{i}+a_{s}+a_{t}+e_{i s t}$

where subscripts

$i, s$ , and

$t$ refer to country, manufacturing sector and year respectively; SII may measure a benchmark country's or each country's own services input intensity, being averaged over time or time-varying;

$\boldsymbol{Z}$ is a vector for other control variables;

$a_{i}, a_{s}, a_{t}$ are the country, manufacturing sector, and year fixed effects; and

$e_{i s t}$ is an error term. As a robustness check, we also use time-varying country and sector fixed effects (i.e., Country*Year and Sector*Year).

Hypothesis 1 suggests a positive $\beta_{3} . \beta_{1}$ can be negative because a more developed services sector (a higher D) in a country could imply a higher services export RCA which in turn could lead to a lower manufacturing export RCA.

Our second hypothesis suggests that the effect of D on RCA depends not only on SII, but also on the access to foreign services markets. To capture the relative importance of foreign services inputs compared to domestic services inputs, we measure access to foreign services markets by the share of embodied foreign services in total embodied (domestic and foreign) services in a manufacturing sector of a country (forsh). To ease the interpretation of the results, we run regressions using the subsample for only the manufacturing sectors with high services input intensity because services development and services inputs are less relevant when a sector uses little services as inputs. We also run the same regressions for all of the other sectors with low $SII$ to show how the results differ. The specification of the regressions is similar to equation (1), except that we replace $SII$ with forsh as follows:

(2) $R C A_{i s t}=\theta_{0}+\theta_{1}$ forsh $_{i s t}+\theta_{2} D_{i t}+\theta_{3} D_{i t} * forsh_{i s t}$

$+Z \boldsymbol{\gamma}+a_{i}+a_{s}+a_{t}+e_{i s t}$

According to Hypothesis 2, coefficient

$\theta_{1}$ is expected to be positive, while

$\theta_{3}$ should be negative.