国际贸易中美不平衡人民币升值英文文献及翻译
Abstract
The large US–China trade imbalance is regularly blamed on the undervalued RMB. Using an estimated model of the trade balance, we find that improvements in the trade balance are likely to be modest for plausible revaluations.
© 2007 Elsevier B.V. All rights reserved. Keywords: China; Trade balance; Revaluation JEL classification: F31; F32
1. Introduction
Bilateral trade flows between China and the US have expanded considerably over the past two 原文请找腾讯752018766优.文'论~文;网http://www.youerw.com RMB) which has been effectively pegged against the US dollar, giving Chinese manufacturers an “unfair advantage”.1
⁎ Corresponding author. Tel.: +61 8 6488 3345; fax: +61 8 6488 1016.
E-mail address: (N. Groenewold).
1 For its part, the US has long alleged that the RMB is undervalued by as much as 40%; see for example Testimony of Franklin J. Vargo before the House Committee on International Relations, United States Congress (2003). It is reported that China's foreign exchange reserve topped $853.7 billion US at the end of February 2006; see People's Daily Online (2006).
0165-1765/$ - see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.econlet.2006.12.021
128 N. Groenewold, L. He / Economics Letters 96 (2007) 127–132
The aim of this paper is not to resolve the question of whether the Chinese RMB is indeed misaligned nor, if it is, to evaluate the extent of such misalignment, nor even to consider whether bilateral trade deficits are a legitimate matter for concern. Rather, we take others’ estimates of the undervaluation and assess the likely impact on the Chins–US trade balance of the re-alignment of the RMB US dollar exchange rate. We argue that this is an interesting exercise both as a contribution to the debate about how serious the supposed misalignment is and also in light of the likely future movement of China to a more market-oriented exchange rate which should see any undervaluation disappear. We find that even large changes in the real exchange rate are likely to have modest effects on the trade deficit.
2. The model
The model we estimate is a reduced-form equation for the trade balance between China and the US derived from a simple two-country model of trade based on work of Rose and Yellen (1989). The fact that both countries have important trading relationship with third countries is likely to introduce important specification bias into the empirical model based on only two countries and we therefore extend the two-country model to account for the possible importance of a third country (the “rest of the world”).
Denoting demands for imports by country i from country j by Dij, supply of exports by country i to country j by Sij, corresponding import prices relative to domestic prices by mij, export prices relative to domestic prices by xij, and real exchange rates by qij where, in each case, i, j = u, c, w (denoting the US, China and the rest of the world), we can write the three-country version of the Rose and Yellen (1989) model as:
Duc ¼ Duc ðmuc ; muw; Yu Þ ¼ Duc ðquc : xcu ; quw: xwu; Yu Þ ð1Þ Duw ¼ Duw ðmuc ; muw; Yu Þ ¼ Duw ðquc : xcu ; quw: xwu; Yu Þ ð2Þ Dcu ¼ Dcu ðmcu ; mcw ; Yc Þ ¼ Dcu ðxuc =quc; qcw : xwc ; Yc Þ ð3Þ Dcw ¼ Dcw ðmcu ; mcw ; Yc Þ ¼ Dcw ðxuc =quc; qcw : xwc ; Yc Þ ð4Þ Suc ¼ Sucðxuc ; xuwÞ ð5Þ Suw ¼ Suwðxuc ; xuwÞ ð6Þ Scu ¼ Scuðxcu ; xcw Þ ð7Þ Scw ¼ Scw ðxcu ; xcw Þ ð8Þ Dij ¼ Sji ; i; j ¼ c; u; wði pjÞ: ð9Þ
In each case the second equality in Eqs. (1)–(4) follows from:
mij ¼ qij :xji and qij ¼ 1=qji :
N. Groenewold, L. He / Economics Letters 96 (2007) 127–132 129
The model consists of 12 equations which allow us to solve for the Dij, the Sij and the xij in terms of the qij and Yu and Yc. The US–China trade balance is defined as the ratio of US exports to imports which has the advantages that the measure is unit-free and that its magnitude is independent of scale effects (Brada et al., 1993).
Thus, the equation of interest is:
þ þ − − þ
TBuc ¼ TBuc ðquc ; quw; qcw ; Yu ; Yc Þ ð10Þ
where the indicated signs follow from standard restrictions on the demand and supply functions (1)–(8). Log-linearising and adding a random error term, we obtain the following estimating model:
lnTBuct ¼ a1 þ a2 ðlnYut Þþ a3 lnðYct Þþ a4 lnðquct Þþ a5 lnðquwt Þþ a6 lnðqcwt Þþ et ð11Þ
3. The data
Quarterly data were used for the period to 1987(1) to 2003(4). Quarterly observations, seasonally adjusted where relevant, for all variables except Yc and quc were obtained from the IMF's International Financial Statistics and Direction of Trade Statistics. Quarterly real GDP data for China are not available until 1999(1) so that annual data
were interpolated using the regression approach.2 The real exchange rate was based on annual Chinese CPI data interpolated using a moving average approach. The real exchange rates for the US and China with respect to the rest of the world were measured by real effective exchange rate indexes for each country. Finally, data for bilateral trade between the China and the US are problematic.3 There are major discrepancies between trade flows reported by China and by the US but the consensus is that US figures are more reliable that the Chinese ones and we therefore used the US-derived figures for the trade balance.
4. The results
4.1. Estimated trade-balance equations
Following Shirvani and Wilbratte (1997), we begin by estimating Eq. (11) using OLS. The resulting estimates are shown in Table 1. A trend is included to allow direct comparison to estimates based on cointegration analysis reported below.
All variables have their predicted signs and, with the exception of qcw and Yc, they are significant at
1%. In particular, the real exchange rate's coefficient displays the hypothesised sign and is significant.
Before using these estimates to draw conclusions regarding the effects of an RMB 原文请找腾讯752018766优.文'论~文;网http://www.youerw.com to have a significant trend and when lags were chosen to eliminate autocorrelation, was found to be clearly non-stationary. We therefore proceeded to test for cointegration and, using the Johansen procedure, we found evidence for a single cointegrating relationship. The
2 Cf. Chow and Lin (1976), Bahmani-Oskooee (1986) and Brada et al. (1993).
3 See Feenstra et al. (1999) and Fung and Lau (2001, 2003).
130 N. Groenewold, L. He / Economics Letters 96 (2007) 127–132
Table 1
Regression results
Variable OLS FMOLS Cointegration
Coefficient p-value Coefficient p-value Coefficient p-value
CONST 48.8203 0.000 73.5661 0.000 NA NA
TREND 0.0562 0.004 0.0941 0.000 0.0837 0.002
Yu − 5.9689 0.000 − 8.7017 0.000 − 6.2032 0.009
Yc 0.0632 0.614 0.0134 0.916 − 0.5056 0.015
quw 0.7986 0.099 0.7817 0.112 0.1995 0.757
qcw − 0.4303 0.120 − 0.6464 0.021 − 0.5967 0.106
quc 0.8298 0.003 1.0767 0.000 1.0774 0.011
R2 0.8745 NA NA
DWS 1.5185 NA NA
Note: The p-values for the OLS and FMOLS regressions are derived from a t-test while those for the cointegrating regression are derived from a Chi-squared test for the exclusion of the variable from the cointegrating vector. The intercept is omitted from the cointegrating regression since the VECM was specified with unrestricted intercepts so that the intercept for the cointegrating vector is not separately identified. Values for R2 and the Durbin–Watson statistics (DWS) are not available for the FMOLS and cointegration estimates.
cointegrating regression estimated using the Phillips–Hansen FMOLS estimator is1678