Abstract
Using annual data for China and 88 trading partners that span the period 1995–2011, we estimate whether cross-societal cultural differences influence China’s external trade flows. Our results, obtained from the estimation of a series of multi-level mixed effect random intercepts and coefficients models, indicate that China’s aggregate exports and imports are largely unaffected by the cultural distance between China and its trading partners. Examination of disaggregate trade measures and consideration of the underlying dimensions of our composite cultural distance variable produces a largely similar result. Taken collectively, our results suggest that China’s trade is less affected by cultural distance than has been reported for other countries in similar studies.
Notes
1. In Equation (Equation1(1) ),
and
.
2. Internal distance, when k = j, is calculated as (Head and Mayer Citation2000).
3. Since we provide a detailed overview of the cultural distance variable in Section 3.2, we refrain from adding duplicate detail here.
4. Given that country i (i.e. China) is the reference country, its economic mass is subsumed into the time/year variable.
5. A panel unit root test of each of the trade flow variables, population and GDP variables rejects (at p < 0.001) the null hypothesis of that the panels contain unit root against the alternative that the panels are trend stationary. Our use of the centered measure of trend computed as the number of years prior to or since 2003, thus permits to make a meaningful inference of the predicted values of the dependent variable series at the mean of control variables.
6. We also estimate Equation (Equation4(4) ) by substituting the composite cultural distance measure with its component dimensions, dTSR and dSSE.
7. When constructing our measure of cultural distance for the years 1995–1999, we employ data from Wave 3 (conducted during 1995–1998) of the WVS. Similarly, data from Wave 4 (1999–2004) of the WVS are used to construct our cultural distance measure for years 2000 through 2005, and data from Wave 5 (2005–2009) are used when constructing our cultural distance measure for years 2006–2011.
8. Cultural distance is calculated as .
9. Given the tendency of the macroeconomic variables to increase overtime, we conduct panel unit root tests of the respective variables using two different methods: the Levin, Lin, and Chu (Citation2002) approach and the Im, Pesaran, and Shin (Citation2003) approach. The Levin, Lin, and Chu (Citation2002) approach imposes a more restrictive assumption that all panels have a common autoregressive parameter and, hence, requires a strongly balanced data-set. Thus, we restrict our data to include only the 81 countries (of the 88 in our full data-set) for which the data-set is balanced. The Im, Pesaran, and Shin (Citation2003) approach permits each panel to have its own autoregressive parameter to evaluate the null hypothesis that all panels have a unit root against the alternative that only a fraction of the panels are stationary. When employing this approach, we utilize the full unbalanced data-set.
10. While the same could be done for product categories, we restrict our analysis of the country-specific deviations to aggregate exports and imports for brevity.
11. Tadesse and White (Citation2010b) for example, report cultural distance imposes negative and statistically significant effects on bilateral trade flows of nine OECD member countries with 67 other countries.
11. Tables with detailed results obtained from using aggregate bilateral exports and imports as the dependent variable series for each of the estimations used in the robustness checks are available upon request.