4.1. Data
To learn more about the potential impacts of the OBOR initiative on Chinese industries, this paper used the export data of countries with annual total exports exceeding 10 million US dollars to predict the relative export sophistication of China in the absence of the OBOR intervention (the observed countries are listed in
Table A1). The cross-country export data is available in the UN Comtrade database and ITC statistics. We compiled 2-digit Harmonized System (HS) classifications for different product categories and matching statistics for the corresponding sectors in China for a complete observation [
31].
Table 1 shows the statistical values of China’s exports, outputs, assets, and employees at the industrial level from 2002–2017 that were obtained from the China Entrepreneur Investment Club Statistical database. The United Nation trade data has covered all major exporters in recent years. Our sample covers the top 80 exporters in the world from 2002–2017, and the export basket complexity values (calculated based on the method of Balassa [
30]; Hausmann and Rodrik [
5,
28]) for each country are listed in the
Table A1.
Our analysis relates the values of exports at the two-digit HS level to data on sectors in China treated as a priority in conducting integration within global production networks. The sectors are classified by technological characteristics. Lall [
32] uses a four-fold technological categorization: resource-based, low-tech, medium-tech, and high-tech manufacturers. Resource-based products are mainly foods and beverages, refined petroleum, leather, and rubber. Low-tech manufacturers produce textiles, toys, furniture, simple metal products, and glassware. Medium-tech manufacturers utilize complex but not fast-moving technologies to produce machinery, chemicals, and simple electronics. High-tech industries include those that produce complex chemicals, pharmaceuticals, automobiles, aircrafts, and ships. China has high growth rates in each category. Nonetheless, the share of low-tech manufacturers in the Chinese export basket is declining steadily, largely compensated for by a rise in high-tech manufacturers. The descriptive statistics show that the production level of high-tech manufacturers has already exceeded those of medium- and low-tech manufacturers since 2016.
We included the export sophistication of China in logarithmic form, and use it to reflect its manufacturing upgrades at the industrial level (
Table 1). In general, the statistic shows that there has been a rapid and sustained technological upgrading of China’s export basket.
4.2. Tests for Significance of the OBOR Policy Intervention
We propose to evaluate the impact of the “One Belt, One Road” initiative on manufacturing industries of China. However, when domestic R&D generates knowledge spillover, the benefits of greater openness become partly indeterminate. To examine the relationship between the greater openness brought by the OBOR initiative and the Chinese manufacturer complexity, we adopted the counterfactual approach and the Panel OLS method.
In this section, we evaluate the effectiveness of the OBOR policy by adopting the Panel Data Control method of Hsiao et al. [
2], which can be seen as an endogenous control process. The counterfactual approach of Hsiao et al. [
2] shares the rationale of the difference-in-differences approach. We compare the Chinese aggregate manufacturer complexity before and after the OBOR initiative implementation to the Chinese aggregate manufacturer complexity without the initiative (a synthetic counterfactual) during the same time period. The effect of the economic integration policy, then, is just the difference between the outcome with the experiment and the one without; in this manner, we focus on the impacts of the policy on the upgrading of manufacturing industries in China.
Since we cannot simultaneously detect the competitiveness of Chinese manufacturing industries under the OBOR initiative and without the initiative, it is necessary to construct the counterfactual of a China that has not experienced the policy. Applying the approach of Hsiao et al. [
2], we use other countries’ exports sophistication to predict what would have happened to China had it not been subject to the OBOR policy intervention. In their method, the cross-sectional dependence is attributed to the presence of common factors that drive all the relevant cross-sectional units.
Let
denote the manufactures’ competitiveness of China at year
t under the OBOR policy intervention and
denote the manufactures’ competitiveness of China at year
t without the policy. The treatment effect of the OBOR policy for Chinese manufacturing competitiveness and export sophistication at year
t is
The observations for other countries could be used to identify the number of common factors,
K, and estimate
by the maximum likelihood procedure [
33]. In our case, the panel control method has been adopted to predict
by
in lieu of
[
2]. We assume the idiosyncratic components are uncorrelated across the selected countries, which means the OBOR policy intervention on China has no bearing on the idiosyncratic components of the selected countries. Hence, the contamination in the control group synthetic procedure could be avoided.
Instead of using the Akaike Information Criterion model employed in Hsiao et al. [
2]’s approach, we adopted the elastic-net selection method for selecting the most relevant cross-sectional economies with which to construct the counterfactuals of China [
34]. With an improved penalization on coefficients, the elastic-net selection method can screen out the most relevant predictors for the control group synthesis [
35]. Although there are numerous countries available for the control selection, the elastic-net selection method only selected out the 10 most relevant predictors for synthesizing the counterfactuals.
Table 2 lists the 10 candidates for the control group that were selected by the elastic-net method of Zou et al. [
34]. The computation package of lasso in STATA is available at
http://fmwww.bc.edu/RePEc/bocode/l.
According to Chen et al. [
14], the participants in the OBOR initiative and China’s Asian neighbors should be excluded as control units without the treatment. Our selected countries for synthesizing the counterfactual include Argentina, Italy, Canada, Mexico, Egypt, Spain, France, Switzerland, Germany, and the UK, which coincidentally satisfy the presumed selection criteria, given that they are geographically remote from and have significant cultural differences with China.
The counterfactual has been synthesized by using observations for these 10 countries in 12 recent years (
N = 10, T = 12).
Table 2 lists the OLS estimated weights of the selected countries for the synthesis of the control group. With a
of 0.943, the counterfactuals before the OBOR policy was implemented could be accurately approximated. All of the estimated parameters are not statistically significant, exactly satisfying the requirement for conducting a quasi-natural experiment that there be no consistent and continuous links between the treated group and the control group.
The actual and hypothetical export performance paths for the period 2001–2017 are plotted in
Figure 3.
Table 3 shows that the differences between the actual export sophistication of China and the counterfactual ones are very small. The synthetic control group has been approximated well in the period from 2001–2012 since the errors of fitting are within a tolerable range.
The treatment effects listed in
Table 4 suggest that the export sophistication of China has risen substantially compared to the export sophistication the country would have had in the event that there had been no OBOR policy intervention. This finding supports the view that the complexity of Chinese manufacturing industries has increased by up to 50% through greater economic integration.
The counterfactual export sophistication of China has been approximated during the time period of 2001–2012. The empirical analysis indicates that there has been a certain and positive impact brought by the OBOR policy. Since the red-dashed line and the blue-solid line coincide with each other until the OBOR policy intervention was implemented (
Figure 3), adequate synthesis results have been achieved. The blue-solid line (the actual export sophistication) separates from the red-dashed line (the counterfactual export sophistication). Specifically, the actual line rises well above the counterfactual one, regardless of the former’s decline after 2017, clearly there has been a certain and positive impact brought by the OBOR policy, in addition to showing that the actual relative export basket sophistication of China was valued at 6.12 trillion US
$ in 2017. These significant intervention effects are not surprising given the “greater integration to re-industrialize” concept proposed by Xi’s administration. It is generally recognized that further opening and actively integrating with other industrialized countries would lead China to greater prosperity.