Robustness Checks

6.1 IV Estimation

A potential endogeneity problem may exist in our empirical framework in that unobserved factors that are correlated with processing exports also influence import distance. For example, locations that are considered ideal for processing activities due to their proximity to the destination market may simultaneously be favorable because of their closeness to suppliers in the destination market. In order to account for this potential issue, we adopted an IV approach in which we used “supplier access” as an instrument for import distance ImDisti. Supplier access is an economic geography concept proposed by Redding and Venables (2004) to measure a location's access to foreign sources of input supply. To measure supplier access, the authors developed a two-stage least square (2SLS) procedure to estimate for each location an appropriately distance-weighted measure of the location of its import supply. This approach was adapted by Ma (2006) to PRC data.

To estimate supplier access, in stage one, we estimate a gravity equation of PRC processing imports by province with year and economy dummies. The estimated coefficients from the gravity equation are then used in stage 2 to calculate the supplier access variable by applying the following equation:

where it SA^ˆ_it denotes province i's supplier access in year t, ˆλ_1t and ˆλ_2t are the coefficients on internal and external distances, respectively, and ηˆ_jt denotes the economy dummy.

Column 6 of Table 4 shows the estimated results using the IV method. Our instrument passes the Stock-Yogo weak identification test at the 10% level and is correctly specified according to the Hansen test. Moreover, the IV estimation using the full sample yields results that are similar to those in the benchmark specification provided in column 3. Specifically, there is strong evidence for Hypothesis 2 in that the coefficient on East_j∗lnExDist_i is negative and statistically significant, while the coefficient on East_j∗lnExDist_i is positive and statistically significant. However, contrary to the benchmark results, the coefficients on lnExDist_it, lnImDist_it, and Eastj are insignificant.

6.2 County-Level Estimation

Another potential estimation issue arises from the presence of industrial clustering in particular regions and the level of aggregation in our analysis. Low-import-distance and highexport- distance provinces may be specialized in different industries than high-importdistance and low-export-distance provinces. In line with this, there might be structural differences between the coastal and internal provinces in their share of processing trade. As is shown in Figure 5 [ PDF 82.2KB | 1 page ], in 2005, about 97% of processing exports and 98% of processing imports were conducted by the 10 coastal provinces listed in the figure. Guangdong alone accounted for approximately 43% of the total processing trade in that year.

Ideally, we should control for locational differences between provinces and across industries by disaggregating our analysis at the industry and county level. However, we are limited by the lack of available data. Specifically, it is not possible to disaggregate the analysis at the industry level because we do not have the necessary input-output information regarding the combination of inputs that are used to produce specific exports. For example, semiconductors can be used to produce both cars and computers. The processing trade data identifies the value of semiconductors that are imported by a certain location but not in which industry they are put to use. Furthermore, conducting a full analysis at the county level is not possible because we are limited by the available number of explanatory variables; most notably, county-level information on GDP per capita, population, and wages is not available.

To at least partially address these estimation issues we re-estimated equation (16) at the county level with the inclusion of county fixed effects to take into account the unobserved heterogeneity across counties. Moreover, we included year dummies and interaction variables between county and year to capture changes over time. We restricted the analysis to the counties in the coastal provinces to control for a potential structural difference between processing activities in coastal and internal provinces. The results presented in column 1 of Table 5 [ PDF 52.5KB | 1 page ] continue to support Hypothesis 2. In particular, an increase in export distance leads to a larger decrease in processing exports destined for the East Asian economies. By contrast, processing exports shipped to East Asia are less sensitive to import distance than those exported to non-Asian OECD countries.

We further disaggregated the county-level analysis for the coastal region to differentiate between processing exports by foreign-invested enterprises (FIEs) and domestic firms. We present the results of this analysis in columns 2 and 3 of Table 5. In line with Hypothesis 2, we find that for both domestic firms and FIEs, processing exports to East Asian economies are more sensitive to export distance and less sensitive to import distance. But the coefficient on the interaction between East and ImDist_i is not significant for FIEs. A comparison of the coefficients in columns 2 and 3 of Table 5 suggests that processing exports by foreign firms are more sensitive to GDP per capita and export distance. Furthermore, processing exports shipped to East Asia by foreign firms are more sensitive to export distance and less sensitive to import distance than those shipped by domestic firms. This latter result may suggest that FIEs are primarily used by Western global production networks to process goods destined for larger and richer East Asian consumer markets; domestic firms are used by Eastern production networks to process inputs from neighboring East Asian economies for goods destined for Western markets. Overall, the results from estimating at the county level for coastal provinces provide support for Hypothesis 2.

Download this Paper [ PDF 340.3KB| 28 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.