‘Are Chinese Trade Flows Different?’

The answer is no, and yes.

From the paper by that title, coauthored with Yin-Wong Cheung (UCSC, CUHK) and Xingwang Qian (SUNY Buffalo State). (This is a revision of the paper discussed here.) For the “no” part:

We find that Chinese trade flows respond to economic activity and relative prices — as represented by a trade weighted exchange rate — but the relationships are not always precisely or robustly estimated. Chinese exports are generally well-behaved, rising with foreign GDP and decreasing as the Chinese renminbi (RMB) appreciates….

For the “yes” part, or “kind of”:

…However, the estimated income elasticity is sensitive to the treatment of time trends. Estimates of aggregate imports are more problematic. In many cases, Chinese aggregate imports actually rise in response to a RMB depreciation and decline with Chinese GDP. This is true even after accounting for the fact a substantial share of imports are subsequently incorporated into Chinese exports. We find that some of these counter-intuitive results are mitigated when we disaggregate the trade flows by customs type, commodity type, and the type of firm undertaking the transactions. However, for imports, we only obtain more reasonable estimates of elasticities when we allow for different import intensities for different components of aggregate demand (specifically, consumption versus investment), or when we include a relative productivity variable.

The importance of the question’s answer is clear, given this graph (familiar to Econbrowser readers):

cnyrer1.gif
Figure 1: Log trade weighted real value of Chinese yuan, 2005=0 (blue left axis), and annualized Chinese trade surplus (red line, right axis), and 12 month moving average (dark red, right axis), in bn. USD. Source: IMF, International Financial Statistics.

There are numerous results reported in the paper (ten main tables, ten appendix tables), too many to discuss in detail. However, two points with respect to imports are noteworthy. The first is the changing composition of types of firms undertaking these international transactions — changing composition implies changing aggregate price elasticities (if price sensitivities vary over firm types).

The second is that the addition of a productivity variable does wonders for the estimates. Again from the paper:

Finally, we include relative productivity as an additional regressor, following Aziz and Li (2008). The motivation is that the CPI deflated real exchange rate does not fully reflect the price of Chinese tradable output, as it includes a large nontradable components (Chinn, 2006). The proxy variable is Chinese GDP per capita relative to US output per man hour in the nonfarm business sector. We present the results of these augmented regressions over the post-WTO accession sample in Table 10.

The inclusion of the relative productivity variable yields substantially improved results. The exchange rate now has the correct sign for all aggregates and components of imports, and is statistically significant in most cases but one. Higher Chinese relative productivity decreases imports as well, which makes sense, as higher productivity is consistent with greater competitiveness. In some sense, the competitiveness variable explains even more of Chinese import behavior than the conventional variables, as the associated coefficient is significant in all cases but one.

Why this is so is explained in this paper, as well as [1] [2]. Recall, competitiveness is best measured as:

rulc = e + ulc – ulc*

where rulc is the unit labor cost deflated log value of the currency, e is the log nominal value of the currency, and ulc is the log unit labor cost, log(W/Z). W is wage per person-hour, and Z is output per person-hour.

We don’t typically observe rulc for developing countries (and even when we do in China, we do it with substantial error — see here. Rather we have:

r = e + p – p*

Where p is the log CPI, which includes lots of nontraded goods not directly relevant to international competitiveness. Including Z directly in the regression can in principle help control for the use of the “wrong” real exchange rate.

Short Run Dynamics

In the paper, we use dynamic OLS to estimate the long run (in a statistical sense) elasticities. However, we are interested in seeing how trade flows react quarter by quarter. Assuming the series are I(1), estimation of an error correction specificaiton can under the appropriate conditions uncover the short run dynamics (see here for an example, as well as a previous application to Chinese exports [3]).

Using a parsimonious specification for imports (which is the problematic series), including income, the real CNY, a dummy for the recession, and the China-US productivity differential, one obtains the expected signs. In addition, I find that about half the disequilibrium is closed in a quarter (ignoring other dynamics, the half life of a deviation is about a quarter).

Concluding Thoughts

The estimates we obtain are imprecise, and are sensitive to the treatment of trends, at least where income elasticities are concerned. But whereas many other researchers have obtained perverse price elasticities for imports, we have had more success. In my mind, then, these results confirm that exchange rate changes can have an impact on Chinese trade flows, a conclusion that would not be apparent if one merely looked at bivariate relationships.

This post originally appeared at Econbrowser and is posted with permission.