UNIVERSITY OF AMSTERDAM
The Exchange Rate and the U.S. Imports from China
Faculty: Faculty of Economics and Business Study Program: BSc Economics and Business Student Name: Jinghan Wang
Student Number: 10227245 Supervisor: Stephanie Chan Date: 11/07/2014
Content
Abstract ... 2
1 Introduction ... 3
2 Literature Review ... 4
2.1 Exchange rate magnitude effect ... 5
2.1.1 Exchange rate effect on aggregate data ... 5
2.1.2 Exchange rate effect on disaggregate data ... 6
2.2 Exchange-rate volatility effect ... 7
2.3 The gravity model of trade ... 8
3 Methodology ... 8
4 Data ... 11
5 Empirical Results ... 15
6 Discussion ... 19
Abstract
This paper analyzes the impact of the exchange rate of the dollar against the RMB on the US imports from China. There are two regressions analysis. The first regression applies nominal quarterly data from 1987q1 to 2013q4, while the second regression uses real yearly data from 1987 to 2011. By running linear regressions of the US imports from China on the bilateral exchange rate, the US GDP, China’s GDP as well as two dummy variables, the results show negative relationship between the dollar/RMB exchange rate and the US imports from China. Besides, the US GDP is shown to be a major factor that affects the US imports in both regressions. Furthermore, unit root tests report that the data for the US GDP, China’s GDP and the US imports from China are not stationary, thus cointegration tests are conducted to investigate whether OLS regression is valid.
1 Introduction
Since 1978 when China introduced economic reforms towards market-oriented economy, the Chinese economy loomed largely on the global economic stage and China’s share of world trade has grown extremely fast. The United States, one of the most important trade partners of China, has run consistent trade deficit since 1980s. In 2013, the United States suffered a trade deficit of more than 44 billion dollars. Specially, the trade balance between the United States and China was more than 318 billion dollars towards China at the end of 2013. Despite the remarkable economic growth, the Chinese authorities have managed the Chinese exchange rate sophisticatedly for decades. Since the economic reform, the Chinese currency has been devalued considerably against the dollar until 1997. Then the Chinese government pegged the Renminbi (RMB) at 8.28 RMB per dollar until 21 July 2005 (Frankel, 2009). After that, as the US government strongly demanded an appreciation of the RMB for years, the Chinese government switched to a managed floating exchange rate regime, which set the RMB with reference to a basket of currencies (Baak, 2008). From then on the RMB started to appreciate gradually against the dollar till now.
The miracle of China’s economy sparks the debate on whether China has manipulated its currency for an unfair trade advantage. Given the growing important role of China playing in the world economy, the question of how sensitive the trade of China is to the movement of the exchange rate is relevant. Against this background, this paper aims to investigate to what extent the bilateral exchange rates between the dollar and the RMB have impact on the trade between the United States and China, more specifically, on the imports of the United States from China.
Marquez and Schindler (2007). This is because the exchange rate of the RMB has been controlled by the Chinese government before the 1990s. Also, the data is limited for the period between 1970s and 1990s. Both nominal data and real data will be estimated, however, because of data limitations, the real data are only available for the period from 1987 to 2011 in annual terms, whereas nominal quarterly data are accessible between 1987 and 2013.
The structure of this paper is as follows. In section 2, previous literature will be reviewed and divided into two groups based on their research methods. In section 3, the regression model applied in this paper will be displayed and the reason why each variable is included in the model will be explained. Section 4 elaborates the data source and some calculation for indirect data. There are also summary statistics of the data presented in tables in section 4. Next, section 5 contains the empirical results processed by STATA. Then in section 6, the empirical results will be further explained and the reason for implausible results will be discussed. Finally, conclusions and improvement for further research are presented in section 7.
2 Literature Review
Previous studies mainly focus on two aspects when examining the exchange rate effect on exports and imports between China and the US. The first group analyzes the level of aggregate as well as disaggregate data of exports and imports between the two countries resulting from the movement of bilateral exchange rate, while the second group estimates the impacts of the volatility of the exchange rate between the dollar and the RMB (RMB) on the exports and imports from the US to China.
2.1 Exchange rate magnitude effect
2.1.1 Exchange rate effect on aggregate data
Baak (2008) analyzes the exchange rate effects on the US imports from China using quarterly aggregate data for the period from the first quarter of 1995 to the second quarter of 2006. He measures the imports of US from China by estimating the cointegrating vectors and error correction models. In addition to the bilateral exchange rate between the dollar and RMB, he also includes the exchange rate of a competing country, the real GDP of the importing country and the volatility of the bilateral exchange rate as explanatory variables in his functions. Also, Baak indicates that all the variables in his functions are measured in real terms. The result of the analysis is that 1% appreciation of the dollar against the RMB boosts the imports of the United States from China by 1.7%. Another remarkable result is the coefficient of the real GDP of the importing country is positive and bigger than the coefficient of the bilateral exchange rate in absolute value, implying that the exchange rate is not the only main factor that affected the exports from China to the US.
Another study applying aggregate data is conducted by Eckaus (2004). Eckaus estimates the exchange-rate impact on the behavior of China’s export to the US using annual time series data from 1985 to 2002. He considers two dependent variables; one of them was the level of the Chinese exports to the US and the other was the share of the US imports from China. In both cases, the regressions estimates with corrections for autocorrelation and heteroskedasticity shows negative relationship between the real exchange rate and the exports from China to the United States. However, similar with that of Baak (2008), the coefficient on the logarithm of the US GDP is quite high and statistically significant, suggesting that the rapid growth of China’s exports to the US was not mainly due to the appreciation of the US dollar.
2.1.2 Exchange rate effect on disaggregate data
Marshall-Lerner condition is one of the traditional ways to estimate the effect of exchange rate on the trade balance, which relying on the price elasticities exports and imports (BAHMANI-OSKOOEE and Wang, 2007). However, when aggregate data is used, a significant price elasticity of one trading industry could be more than counteract by an insignificant elasticity of another trading industry, thereby resulting insignificant overall price elasticity (Bahmani-Oskooee and Ardalani, 2006). To identify the sensitivity of trade to the exchange rate, it is appropriate to employ disaggregate data at commodity level (BAHMANI-OSKOOEE and Wang, 2007).
Bahmani-Oskooee and Ardalani (2006) conducts cointegration analysis on monthly data of exports and imports around 66 industries in the US with the rest of the world from January 1991 to August 2002. Assessing exports and imports in value models as well as volume models, Bahmani-Oskooee and Ardalani (2006) argues that in the long run, the real depreciation of the dollar encourages export proceeds of many industries, while there is insignificant impact on most importing industries. Meanwhile, their study shows that the world income and the US income are two major explanatory variables of the exports and imports in the US.
BAHMANI-OSKOOEE and Wang (2007), using annual time series data from 1978 to 2002 for 88 industries, focuses on the exports and imports between the United States and China. Their results based on cointegration and error-correction models illustrate that the real appreciation of the dollar against the RMB stimulates the US import disbursement to China in 40 industries, whereas discourages its export earnings only in 18 industries, suggesting that the real bilateral exchange rate between the dollar and the RMB is a significant factor in trade between the US and China at industry level.
Marquez and Schindler (2007) analyzes the responsiveness of China’s trade to changes in exchange rate using monthly disaggregate data over the period of January 1997 to July 2006 in a different way. The Chinese trade data is disaggregated into two parts, the assembled goods and ordinary products, for both exports and imports. The outcomes based on ordinary least squares reveal a negative and significant exchange-rate effect on the Chinese nominal export share for both assembly and ordinary goods, thus implying that in aggregate level, a 10% real appreciation of the RMB discourages China’s exports by approximately 1% (Marquez and Schindler, 2007). The results for the response in terms of aggregate imports are negligible, because the exchange-rate effect for the ordinary-good part is positive, whereas it is negative for the assembled part in a similar range of magnitudes (Marquez and Schindler, 2007).
2.2 Exchange-rate volatility effect
In terms of the volatility effect of exchange rate, Baak (2008) finds that the uncertainty of the dollar to RMB exchange rate has negative influence on the US imports from China, but negligible influence on the exports of the US to China.
Chou (2000) estimating with error-correction model as well as ARCH model and quarterly data from the first quarter of 1981 to the last quarter of 1996, indicates that there is a significant negative long-run impact of the exchange rate variability on China’s total exports. He further divides the exports data into four categories, namely foodstuffs, mineral fuels, manufactured products and industrial materials. Specifically, the exchange-rate volatility has significant negative influence on the exports of manufactured products and the exports of mineral fuels, but insignificant influence on the exports of foodstuffs. On the contrary, the impact of exchange-rate volatility on the exports of industrial
materials is positive (Chou, 2000).
In another paper of Bahmani-Oskooee and Wang (2007), the impacts of the uncertainty of exchange rate on commodity trade between the United States and China are examined by analyzing exports and imports data in 88 industries. They discover that almost half of the industries are sensitive to the exchange-rate volatility. Moreover, Bahmani-Oskooee and Wang (2007) also indicate that although a major part of the US imports from China are negatively related to the exchange-rate uncertainty, most of its exports to China are positively related.
2.3 The gravity model of trade
When examining the relationship of trade between two countries, one of the famous models is Tinbergen’s gravity model of trade. This model is used to predict the bilateral trade flows based on the GDP of two countries and the distance between them. Mátyás (1997) gives the econometric representation of the gravity model, which takes logarithm of each variable in the gravity model of trade.
3 Methodology
Based on Tinbergen’s gravity model of trade, this paper will study the relationship between the bilateral exchange rate and the US imports from China by modifying the econometric representation of the gravity model in Mátyás (1997). Specifically, the distance term in the gravity model of trade is substituted by the exchange rate of the dollar against the RMB, and two dummy variables are added. The regression function is described as follows:
𝑙𝑛(𝐼𝑀𝑡) = 𝛽0+ 𝛽1∗ 𝑙𝑛(𝐸𝑥𝑡) + 𝛽2∗ 𝑙𝑛(𝑈𝑆𝐺𝐷𝑃𝑡) + 𝛽3∗ 𝑙𝑛(𝐶ℎ𝑖𝑛𝑎𝐺𝐷𝑃𝑡) + 𝛽4 ∗ 𝐹𝑙𝑜𝑎𝑡𝑡+ 𝛽5∗ 𝐶𝑟𝑖𝑠𝑖𝑠𝑡+ 𝑢𝑡
Where:
𝐼𝑀𝑡 is the total value of the US imports from China at time t; 𝐸𝑥𝑡 is the exchange rate of the dollar against the RMB at time t; 𝑈𝑆𝐺𝐷𝑃𝑡 is the GDP in the United States at time t;
𝐶ℎ𝑖𝑛𝑎𝐺𝐷𝑃𝑡 is the GDP in China at time t;
𝐹𝑙𝑜𝑎𝑡𝑡 is a dummy variable, and is set to 1 for the period with floating exchange rate, and 0 for the period from 1997 to 2005 when the Chinese government came up with a fixed exchange rate regime;
𝐶𝑟𝑖𝑠𝑖𝑠𝑡 is a dummy variable, which controls for the financial crisis period between 2008 and 2009, and is set to 0 to all other periods;
𝛽𝑖 are coefficients (i=0,1,2,3,4 and 5) and 𝑢𝑡 is the error term.
The reason to include Float is that the Chinese government pegged the RMB against the dollar at 8.28 RMB per dollar over the period between 1997 and 21 July 2005 (Frankel, 2009). As shown in Figure 1, the dollar experienced a drastic appreciation from 1987 to 1994. Then the dollar/RMB rate was anchored till 2005. After that, China adjusted to a new managed floating exchange rate regime and the dollar started to depreciate. Since the dollar/RMB exchange rate is fixed from 1997 to 2005, any change in the exports of China to the US explained by the bilateral exchange rate effect is not reasonable.
In addition, despite the increasing trend of trade between the US and China, the imports of the US from China declined remarkably in 2009, as depicted in Figure 2. The dummy variable Crisis is added to the regression function to investigate whether the financial crisis from 2008 to 2009 contributes to the deterioration in the imports of US from China.
Figure 1: Exchange rate of dollar against RMB from 1987 to 2013
Source: DataStream
Figure 2: Nominal trades between the United States and China between 1987 and 2013
Source: DataStream
As the first three explanatory variables as well as the dependent variable are in logarithms, the coefficient represents that 1% change in the explanatory variable is associated with a 𝛽𝑖% change in the US imports from China, i.e., 𝛽𝑖
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.190.2 0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.28 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
Exchange rate(dollar/RMB)
0.00 50,000.00 100,000.00 150,000.00 200,000.00 250,000.00 300,000.00 350,000.00 400,000.00 450,000.00 500,000.00 1… 1… 1… 1… 1… 1… 1… 1… 1… 1… 1… 1… 1… 2… 2… 2… 2… 2… 2… 2… 2… 2… 2… 2… 2… 2… 2…Trade of the US from China
Exports Imports
is the elasticity of the US imports from China with respect to each explanatory variables. In addition, the coefficients for dummy variables are interpreted in a different way. The statistical significance of Float means that the fluctuation of the exchange rate has effects on the dependent variable and the statistical significance of Crisis implies the financial crisis has affected the imports of the US from China.
Before conducting the OLS regression, the data for the US imports from China, the GDP of the US and the GDP of China should be analyzed by unit root test. If the variables contain unit roots, then it means that the variable is not
stationary and a test for cointegration is needed. All the tests mentioned above can be realized in STATA.
4 Data
In this section, the data series applied in the regressions will be described. There are two regressions discussed in this paper. Drawn from 1987 to 2013, 108 quarterly observations of each variable will be analyzed in the regression model. The data for all the variables employed in the first regression are converted into logarithms in nominal terms and can be found in Datastream, as summrized in Table 1.
Table 1:Summary of the nominal data
Variable Obs Mean Std. Dev. Min Max
lnUSimports 108 10.01243 1.279606 7.403 11.73278
lnExchangerate 108 -1.902779 0.2666043 -2.163441 -1.311929
lnUSGDP 108 16.09193 0.3790807 15.37075 16.65398
lnChinaGDP 108 14.66594 1.169781 12.25907 16.71553
standard deviation (1.28 and 1.17, respectively), comparing to those of the exchange rate and the US GDP (0.27 and 0.38, respectively). This can also be reflected by their minimum and maximum observations. In particular, China’s GDP arrived its maximum level at 16.72 in 2013q4, meanwhile, the US imports from China peaked at 11.73.
Figure 3: Scatter plots of the nominal data
In order to obtain a better view of the data, Figure 3 gives the scatter plots of variables against the time variable. While the exchange rate fluctuated over time, the data for the other three variables between 1987q1 to 2013q4 grew in general. Also, although with a smaller standard deviation (0.38), the US GDP experienced a similar trend with its imports from China, where there was a drop from 2008 to 2009. Another interesting finding is from Figure 3 (b), where
(a) (b) (c) (d) 7 8 9 10 11 12 ln (U S imp o rt s fro m C h in a ) 0 20 40 60 80 100 T -2 .2 -2 -1 .8 -1 .6 -1 .4 -1 .2 ln (Exch a n g e ra te ) 0 20 40 60 80 100 T 1 5 .5 16 1 6 .5 17 ln (U S G D P) 0 20 40 60 80 100 T 12 13 14 15 16 17 ln (C h in a G D P) 0 20 40 60 80 100 T
the exchange rate had a structual break around 1994q1. This may cause inaccurate regression results when applying OLS.
Because quarterly data for real bilateral exchange rate and China’s GDP are not available, to analyze the exchange rate effect on the US imports from China in real terms, the second regression will use real annual data from 1987 to 2011. Real historical GDP for the US and China (converted to a 2005 base year) can be found in World Bank World Development Indicator and International Financial Statistics of the IMF. The real bilateral exchange rate data is availible from IMF Data and Statistics, which is calculated from nominal exchange rates and CPIs using 2005 as base year. Indirectly, because of missing data, the approximate annual data for real imports of the US from China would be calculated based on following formula:
𝑅𝑒𝑎𝑙𝑖𝑚𝑝𝑜𝑟𝑡𝑠 = 𝑅𝑒𝑎𝑙𝑈𝑆𝐺𝐷𝑃
𝑁𝑜𝑚𝑖𝑛𝑎𝑙𝑈𝑆𝐺𝐷𝑃∗ 𝑁𝑜𝑟𝑚𝑖𝑛𝑎𝑙𝑖𝑚𝑝𝑜𝑟𝑡𝑠
Where the nominal GDP for the United States and the nominal US imports from China are drawn from DataStream.
The logarithms of the real annual data of variables in the second regression are summarised in Table 2.
Table 2: Summary of the real data
Variable Obs Mean Std. Dev. Min Max
lnRimports 25 10.48683 1.357622 7.882663 12.25501
lnrealEx 25 -2.003849 0.116658 -2.282619 -1.709698
lnRUSGDP 25 16.18036 0.2092091 15.82682 16.44355
lnRChinaGDP 25 14.08929 0.701245 12.98792 15.24805
comparing with that (1.17) of nominal data. In addition, the real GDP of China reached its highest in 2011, and the real US imports from China reached its higest point in the same year. Besides, as shown in Figure 4 (a) and (c), the real US GDP and the real US imports from China show a much more similar trend in this case, although the real US imports has a much higher standard deviation (1.36), in comparision with the standard deviation of the real US GDP (0.21). Furthermore, Figure 4 (b) discribes the annual real bilateral exchange rate varied over time, and the structure break in 1994 is a large outlier in this case.
Figure 4: Scatter plots of the real data
(a) (b) (c) (d) 8 9 10 11 12 ln R imp o rt s 0 5 10 15 20 25 t -2 .4 -2 .2 -2 -1 .8 -1 .6 ln re a lEx 0 5 10 15 20 25 t 1 5 .8 16 1 6 .2 1 6 .4 ln R U SG D P 0 5 10 15 20 25 t 13 1 3 .5 14 1 4 .5 15 1 5 .5 ln R C h in a G D P 0 5 10 15 20 25 t
5 Empirical Results
This section will present the regression results for the nominal data first, then the regression results for the real data will be described.
First of all, to examine whether the variables are generated in stationary process, unit root test is required. Phillips-Perron test for unit root is conducted on the dependent variable as well as the independent variables, the US GDP and China’s GDP. As shown in Table 3, two variables, the logarithm of the US imports from China and the logarithm of the GDP of China, fail to reject the null hypothesis that the variable contains a unit root at even 10% significant level, whilst the unit root test for the variable of the US GDP suggests that this variable is generated by a stationary process at a 5% significant level but not at a 1% significant level. In this case, the OLS regression may not be valid, and therefore a cointegration test is necessary.
Table 3: Unit root test results for nominal data
Test Statistic Z(t) 1% Critical Value 5% Critical Value 10% Critical Value MacKinnon approximate p-value for Z(t) ln(USimportsfromChina) -2.073 -3.508 -2.890 -2.580 0.2554 ln(USGDP) -2.950 -3.508 -2.890 -2.580 0.0399 ln(ChinaGDP) -1.111 -3.508 -2.890 -2.580 0.7106
To test the cointegrating relationship, the first step is to specify how many lags to include in the test and then the second step is to test for cointegration. Figure 6 (see appendix) reports the results for cointegration test. Firstly, the vecrank test indicates that there are 4 lags should be included. Then the Johansen test for cointegration shows that we reject the null hypothesis of no cointegration as well as the null hypothesis of at most one cointegrating equation, but accept the null hypothesis that there are two cointegrating relationships. This result suggests that the OLS regression process is valid.
Now, the nominal quarterly data is estimated by linear regression with robust bias correction in STATA. The data fit the regression line quite well, indicated by R-squared, which means around 99% of the variance in the US imports from China can be explained by the independent variables, namely the exchange rate, the GDP of the US, the GDP of China and two dummy variables. The p-values of the coefficients are significant at 1% significant level except for the p-value for the coefficient of Crisis. The insignificant p-value implies that period of financial crisis not has effect on the imports of US from China statistically. On the contrary, the significance of the coefficient of Float suggests that the comparing with the fixed exchange-rate period, the floating of the dollar/RMB rate has effect on the imports of the US from China. Also, the results in Table 4 shows that 1% increase in the bilateral exchange (depreciation of dollar) decreases the US imports from China by around 0.49%, while 1% increase in the US GDP and China’s GDP both increase the US imports from China (about 2.71% and 0.15%, respectively).
Table 4: The first regression results for nominal data
Linear regression Number of obs = 108 F(5,102) = 2475.67 Prob>F = 0.0000 R-squared = 0.9915 Root MSE = .12086 lnUSimports fromChina Coefficient Robust HC3 Standard Error t P>|t| [95% Conf. Interval] lnExchangerate -.4941724 .075372 -6.56 0.000 -.6436724 -.3446723 lnUSGDP 2.713811 .161697 16.78 0.000 2.393085 3.034536 lnChinaGDP .1491791 .0512387 2.91 0.004 .0475475 .2508108 Float .0985778 .0327011 3.01 0.003 .0337153 .1634402 Crisis .0836224 .0500552 1.67 0.098 -.0156619 .1829067 constant -36.85261 1.891807 -19.48 0.000 -40.605 -33.10021
Furthermore, the coefficients of the logarithms regression are the elasticities of the dependent variable with respect to the independent variables. More specifically, the elasticity of the imports of US from China with respect to the
exchange rate of dollar/RMB is -0.49, inelastic. The elasticity with respect to the US GDP is 2.71 which is elastic (greater than 1), whilst the elasticity with respect to China’s GDP is 0.15 which is inelastic (less than 1).
Even though the regression analysis for nominal data is valid and the results are statistically reasonable, we still want to investigate how does the real bilateral exchange rate between the dollar and the RMB have effect on the trade between the United States and China, specifically the US imports from China. Based on the same regression model, in the second regression, 25 real annual observations on each variables will be imported into the estimation equation. The dummy variable controled for floating exchange rate period and the dummy variable controlled for financial crisis period are still kept in this regression.
Again, before processing the OLS regresssion, unit root test is needed to see whether the data is stationary. Table 5 shows results of the Phillips-Perron test for unit roots. The MacKinnon p-value for ln(Rimports) is significant at 1% significance level, which rejects the null hypothesis that the variable contains a unit root. In contrast, the real US GDP and the real China’s GDP both have large p-values and fail to reject that the variables contain unit roots.
Table 5: Unit root test results for real data
Test Statistic Z(t) 1% Critical Value 5% Critical Value 10% Critical Value MacKinnon approximate p-value for Z(t) ln(Rimports) -3.495 -3.750 -3.000 -2.630 0.0081 ln(RUSGDP) -1.681 -3.750 -3.000 0.4408 ln(RChinaGDP) 0.546 -3.750 -3.000 0.9862
Next, the analysis moves to the cointegration test. Firstly, by implementing varsoc in STATA, the outcome suggests 4 lags to include in the following test. Then, by employing vecrank on lnRimports, lnRUSGDP and lnRChinaGDP
together with 4 lags, the cointegration test is inconclusive. Details for this cointegration test are shown in Figure 7 in Appendix.
Even though the cointegration test is inconclusive, we still run the regression on real data to see if we can gather any clue for the research question. Table 6 reports the robust OLS regression of real imports of the US from China on the real US GDP, real Chinese GDP, real exchange rate of the dollar against the RMB, all of which are in logarithms, and two dummy variables.
Table 6: The second regression results for real data
Linear regression Number of obs = 25 F(5,19) = 489.05
Prob>F = 0.0000 R-squared = 0.9933 Root MSE = .12508 lnRimports Coefficient Robust HC3
Standard Error t P>|t| [95% Conf. Interval] lnrealEx -.9107004 .3301557 -2.76 0.013 -1.601724 -.2196766 lnRUSGDP 3.793114 .9591812 3.95 0.001 1.785524 5.800703 lnRChinaGDP .7965976 .2642297 3.01 0.007 .2435586 1.349637 rFloat .0487736 .0696053 0.70 0.492 -.096912 .1944591 rCrisis -.0244198 .0906264 -0.27 0.790 -.2141031 .1652634 constant -63.96478 11.59209 -5.52 0.000 -88.22732 -39.70225
Again the data fits the regression model quite well, as the R-squared indicates that about 99% of the change in the dependent variable can be explained by the independent variables. The p-values for all OLS estimators have rised comparing with those in the previous regression using nominal time series data. The coefficient of real GDP of China and the coefficient of the real US GDP are still significant at 1% significance level, whereas the coefficient of the real exchange rate is significant at 5% significance level. In consistent with the regression results for nominal data, the US GDP influences the imports of the US the most among the factors in this model, but markedly more than that in the first regression. The coefficient for lnRUSGDP is 3.79, comparing with 2.71
for lnUSGDP in the first regression. Thus, by applying real time series data, 1% increase in the real US GDP stimulates the real imports of the US from China by around 3.79%. The sign of the coefficients for the real GDP of China and the real bilateral exchange rate are the same with those in the first regression, however, their magnitudes are much larger. Precisely, 1% change in the real GDP of China alters the US imports from China with approximate 0.80% in the same direction, comparing with only 0.15% effect in the early estimation. Likewise, 1% depreciation in the real dollar against the RMB depresses the imports of the US from China by 0.91%, while the exchange-rate effect is 0.49% for the nominal exchange rate.
Another remarkable finding is that the second OLS regression indicates that the p-values for both dummy variables are high (0.492 and 0.790, respectively) and suggests that their estimators are insignificant at even 10% significance level. Moreover, the sign of the coefficient for rFloat is positive and the same as its counterpart in the first regression, whilst the sign of the coefficient for
rCrisis is negative, which is opposite to that in previous test.
Lastly, the second regression results show that the elasticity of the real exchange rate, although still inelastic (-0.91), is much more approaching to 1. Additionally, the US imports from China responses more to the US GDP as well as China’s GDP in real terms, in comparision with those in nominal terms.
6 Discussion
First of all, the cointegration tests for the nominal quarterly data show there are two cointegrating relationships and thus OLS estimators are effective. For the real annual data, the cointegration tests are inconclusive. This may due to the fact that the yearly observations are too few (only 25 in this case) for a time series analysis. So in this case the validity of OLS estimation is questionable.
The estimation results in Section 5 shine light on some similar findings with the literature reviewed in Section 2. In the first place, by employing aggregate data, the first regression with nominal quarterly data from 1987q1 to 2013q4 as well as the second regression with real yearly data from 1987 to 2011 shows that there is a negative relationship between the US imports form China and the bilateral exchange rate. This is in consistent with the conclusions in Baak (2008) and Eckaus (2004). Additionally, when switching to real data regression, the coefficient on exchange rate is larger than that of nominal data estimation, implying a larger exchange-rate effect on the US imports.
In the second place, both regressions in this paper reveal that the US economic growth (represented by the US GDP) is an essential factor that attracts imports and outweighs the exchange rate effect, as shown by its significant coefficients in both cases (2.71 and 2.57, respectively). Bahmani-Oskooee and Ardalani (2006), Baak (2008) and Eckaus (2004) also give argument on this in their studies and suggest that the rapid growth of the US imports from China may not be the result of the appreciation of the dollar.
Figure 5: the logarithms of the US GDP from 1987 to 2011
Source: DataStream 15 15.2 15.4 15.6 15.8 16 16.2 16.4 16.6 16.8 17 17.2 17.4 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 lnrUSGDP lnUSGDP
Apart from the significant estimators of the US GDP, China’s GDP and the dollar/RMB rate in both nominal and real terms, the outcomes of dummy variables show little significance. The first dummy variable (Float) is significant at 1% significance level in the first regression, which means the research time period excluding the fixed exchange rate period (1997-2005) has influence on the US imports from China. When real annual data is applied in the second regression, rFloat becomes insignificant, thus exerts no statistic effect on the dependent variable. This may be reasonable because when real bilateral exchange rate is used, the real exchange rate becomes floating even during the period of fixed exchange rate policy, which makes the first dummy variable worthless. For the second dummy variable, the OLS estimator is insignificant in either case, suggesting no empirical difference between the financial crisis period and normal period in terms of the effect on the US imports. One of the explanations is that the dollar experienced depreciation against the RMB during the financial crisis, which helps to explain some decrease in the US imports from China. Moreover, the real US GDP declined substantially by around 382400 million dollars from 2008 to 2009 (see Figure 5). As an important factor to explain the fluctuation of the US imports, the drop of the real US GDP may also contribute to the considerable decline of the US imports in 2009.
7 Conclusion
This paper analyzes to what extent the bilateral exchange rate between of the dollar against RMB has impact on the US imports from China. Applying both nominal data and real data, the OLS regressions as well as cointegration tests are conducted. In particular, using nominal quarterly data from 1987q1 to 2013q4, the first regression shows that 1% increase in the exchange rate of the dollar against the RMB depresses the US imports from China by 0.49%. In
the second regression, real annual data from 1987 to 2011 are employed and the results suggest that 1% increase in the real bilateral exchange rate between the dollar and the RMB decreases the imports of the US from China by 0.91%.
The explanatory variables the exchange rate, the GDP of the United States and the GDP of China remain significant in both nominal and real situations. One noticeable finding is that the US GDP is an essential and important driver for the boost of the US imports from China, implying that the rapid growth of the US imports from China is not mainly because of the appreciation of dollar since the second half of 1980s. Even though the dollar started to depreciate from 2005, the US imports still keep increasing.
The first dummy variable in the regression model controls for the period when the exchange rate is not strictly fixed. In the first regression when nominal data are used, Float is significant, which means the floating of exchange rate matters for the US imports from China. In the second regression, rFloat is insignificant. This result is acceptable because applying real data makes the fixed nominal exchange rate over the period from 1997 to 2005 floating in real term. The second dummy variable in the regression model remains insignificant in both regressions, suggesting dividing the time period into financial crisis period and other period does not give statistical influence on the US imports.
As the Phillips-Perron test for unit root reveals that the data for the US GDP, China’s GDP and the imports of the US from China are not stationary, cointegration tests are processed on these data. For the nominal quarterly data, the cointegration test finds two cointegrating relations, thus the OLS estimations are valid. Nonetheless, the cointegration test for the real annual data is inconclusive. This may owe to the insufficient number of observations.
Hence in this case we can not conclude whether the OLS estimators are valid or not.
To sum up, the analysis in this paper shows negative relationship between the dollar/RMB exchange rate and the US imports from China. However, the results are not credible based on real data. To improve this, using quarterly real data may be helpful and relevant. Meanwhile, disaggregating the data into different industries may provide more empirical evidence of the bilateral exchange rate impact on the US imports from China.
Bibliography
Bahmani-Oskooee, M. and Ardalani, Z. 2006. Exchange rate sensitivity of US trade flows: evidence from industry data. Southern Economic Journal, pp. 542—559.
BAHMANI-OSKOOEE, M. and Wang, Y. 2007. UNITED STATES-CHINA TRADE AT THE COMMODITY LEVEL AND THE YUAN-DOLLAR EXCHANGE RATE. Contemporary Economic Policy, 25 (3), pp. 341--361. Brada, J. C., Kutan, A. M. and Zhou, S. 1993. China's exchange rate and the
balance of trade. Economics of planning, 26 (3), pp. 229--242.
Cheung, Y., Chinn, M. D. and Fujii, E. 2010. China's current account and exchange rate. University of Chicago Press, pp. 231--271.
Chou, W. L. 2000. Exchange rate variability and China's exports. Journal of
Comparative Economics, 28 (1), pp. 61--79.
Eckaus, R. S. 2004. should china Appreciate the Yuan. MIT department of
Economics Working Paper.
Marquez, J. and Schindler, J. 2007. Exchange-rate Effects on China's Trade*.
Review of International Economics, 15 (5), pp. 837--853.
Frankel, J. (2009). New estimation of China's exchange rate regime. Pacific
Economic Review, 14(3), pp.346--360.
Mátyás, L. (1997). Proper econometric specification of the gravity model. The
world economy, 20(3), pp.363--368.
Baak, S. (2008). The bilateral real exchange rates and trade between China and the US. China Economic Review, 19(2), pp.117--127.
Bahmani-Oskooee, M. and Wang, Y. (2007). The impact of exchange rate volatility on commodity trade between the US and China. ECONOMIC
Appendix
Figure 6: Cointegration test for nominal data
3 39 720.52265 0.02624
2 38 719.14007 0.19309 2.7652* 3.76 1 35 707.98376 0.21728 25.0778 15.41 0 30 695.24463 . 50.5560 29.68 rank parms LL eigenvalue statistic value maximum trace critical 5%
Sample: 1988q1 - 2013q4 Lags = 4 Trend: constant Number of obs = 104 Johansen tests for cointegration . vecrank lnUSimportsfromChina lnUSGDP lnChinaGDP, lag(4)
Exogenous: _cons
Endogenous: lnUSimportsfromChina lnUSGDP lnChinaGDP
4 720.523 149.1* 9 0.000 4.1e-10* -13.1062* -12.7045* -12.1146* 3 645.973 143.14 9 0.000 1.4e-09 -11.8456 -11.5366 -11.0828 2 574.401 167.77 9 0.000 4.8e-09 -10.6423 -10.426 -10.1084 1 490.518 918.12 9 0.000 2.0e-08 -9.20226 -9.07865 -8.89714 0 31.4577 .000116 -.547264 -.516361 -.470983 lag LL LR df p FPE AIC HQIC SBIC Sample: 1988q1 - 2013q4 Number of obs = 104 Selection-order criteria
. varsoc lnUSimportsfromChina lnUSGDP lnChinaGDP delta: 1 quarter
time variable: newT, 1987q1 to 2013q4 . tsset newT . format newT %tq . generate newT = tq(1987q1) +T delta: 1 unit time variable: T, 0 to 107 . tsset T
Figure 7: Cointegration test for real data 3 39 217.08856 0.35290 2 38 212.51837 0.65232 9.1404 3.76 1 35 201.42531 0.89301 31.3265 15.41 0 30 177.95745 . 78.2622 29.68 rank parms LL eigenvalue statistic value maximum trace critical 5%
Sample: 1991 - 2011 Lags = 4 Trend: constant Number of obs = 21 Johansen tests for cointegration . vecrank lnRimports lnRUSGDP lnRChinaGDP, lag (4)
Exogenous: _cons
Endogenous: lnRimports lnRUSGDP lnRChinaGDP
4 217.089 62.069* 9 0.000 1.6e-11* -16.9608* -16.5398* -15.021* 3 186.054 32.81 9 0.000 9.1e-11 -14.8623 -14.5384 -13.3701 2 169.649 29.332 9 0.001 1.5e-10 -14.157 -13.9304 -13.1125 1 154.983 221.13 9 0.000 2.5e-10 -13.6174 -13.4879 -13.0206 0 44.4178 3.9e-06 -3.94455 -3.91216 -3.79533 lag LL LR df p FPE AIC HQIC SBIC Sample: 1991 - 2011 Number of obs = 21 Selection-order criteria
