1、Prediction of US Export to China西南财经大学Southwestern University of Finance and Economics商科研究方法II 论文题目: Prediction of U.S Export to China学生姓名: 郭宏宇 所在学院: 国际商学院专 业: 国际经济与贸易学 号: 411100242014年6月Prediction of U.S Export to ChinaGuo Hong-yu(The Southwest University of Finance and Economics,International econ
2、omics and trade )Abstract: China and America is the largest trading partner. Since Chinas accession to WTO, the volume of trade between China and America have risen sharply. However, there are still some trade barriers and friction between two countries. The research based on trade between America a
3、nd China is important to the two countries even the world economy. I select the U.S export to China as dependent variable and use distributed lag model to forecast the U.S export.Key words: U.S export to China;Q-test;1. Introduction1.1Motivation (Background and Significance of Topics)America and Chi
4、na are the first two largest economic entities in the world. At the same time, china and America is also the largest trading partner. Forecasting the export and import volume in international trade is the prerequisite of a governments policy-making and guidance for a healthier international trade de
5、velopment. Since Chinas accession to WTO, the volume of trade between China and America have risen sharply. However, there are still some trade barriers and friction between two countries. The research based on trade between America and China is important to the two countries even the world economy.
6、 Here, I select the U.S export to China as dependent variable and use distributed lag model to forecast the U.S export.1.2Literature reviewMost of literatures set up ARMA models to predict future exports. And many professors point out that ARMA model is useful in short-run forecast rather than long-
7、run which may have fierce volatility. Ronald L. Coccari investigated the results of a few different models in “Alternative Models for Forecasting U. S. Exports”(Ronald L. Coccari,1998). He pointed out that One obvious conclusion from this analysis is that forecast evaluation should not be regarded a
8、s a procedure for accepting one forecasting model to the exclusion of others. Xiao Z & Gong K investigated a combined forecasting approach based on fuzzy soft sets ( Xiao Z & Gong K, 2007) , which pointed out that a combined forecasting approach based on fuzzy soft sets is a promising forecasting ap
9、proach. Diamantopoulos A and Winklhofer H. researched the technique utilization and its impact on forecast accuracy, the paper pointed that “in order to improve forecast accuracy, attention needs to be focused beyond the question of technique selection”. (Diamantopoulos A & Winklhofer H. ,2003) . Ea
10、ch alternative method usually contains a valuable piece of information that may be used by combining all available forecasts into a composite. One might simply want to use the trend and seasonal model since there was found to be such a strong trend component in U.S. exports. It also reveals the obvi
11、ous dependency of exports upon previous lagged observations (especially a three quarter lag), and thus a simple autoregressive scheme seems appropriate.2. Data2.1 Data sourceI find monthly data of export from the US to China and foreign exchange rate between the US and China in the website Federal R
12、eserve Economic Data (FRED). They are seasonally adjusted data from January 1985 to August 2013. But China is officially not a WTO (World Trade Organization) member until December 2001. So I drop the data which is previous than January 2002 and focus on the later ones. And the data is seasonally adj
13、usted.2.2 Data descriptionI plot the points of export and exchange rate and get the following graph. And we can see that there are trend both in the export and exchange rate.Then I make an Augmented Dickey-Fuller (ADF) test to see if there is unit root in the data or it is stationary. First, I test
14、the data of export.The above stata results show that export is stationary, but there is unit root in exchange rate as shown below.So it need to make difference of exchange rate to get Zt=ert-ert-1And then test if its stationary. The result is as below and its stationary.3. Model3.1 Model selectionEx
15、port is affected by change in exchange rate. As in the economic theory, when a countrys currency appreciates (rises in value relative to other currencies), the countrys goods abroad become more expensive and foreign goods in that country become cheaper (holding domestic prices constant in the two co
16、untries). So the country exports less. From 2002 to 2005, the decrease of exchange rate between the US dollar and other currencies helped US industries export more and sell more goods. The impact of exchange rate on export should be accounted for.For export, firstly I take the lag, it is still stati
17、onary. Then I remove the linear trend from log export and get the residual. Next is identify cycle in the residual. Autocorrelation graph has a trail Partial autocorrelation graph has a truncation. So it is an AR model.And its known that previous volume of export have impact on latter volume of expo
18、rt. So I combine the AR(p) model of export and distributed lag model of export on difference of exchange rate. And the number of distributed lags is q. The specific number of p and q should be identified to find the true model.3.2 Model specificationThen the author use the information criteria AIC a
19、nd BIC to get the specific number of p in the AR model. Finally, the result showed that AIC and BIC is the smallest when p equals 13As for the difference of exchange rate (z), the trend of exchange rate has been removed when make the difference. So the author just make the autoregressive of Z on its
20、 lags and then use AIC and BIC to find the number of lags. Finally, I get q equals 1 that is with the smallest AIC and BIC. Zt=+Zt-1. This is the graph of difference of exchange rate (z).Through the previous process, the true model I found is below: lnexportt=0+1t+Zt-1+1lnexportt-1+2lnexportt-2+3lne
21、xportt-3+4lnexportt-4+5lnexportt-5+6lnexportt-6+7nexportt-7+8lnexportt-8+9lnexportt-9+10lnexportt-10+11lnexportt-11+12lnexportt-12+13lnexportt-13The graph above is the regression result using classical standard error. The R2 is pretty high, which is close to 1. As for the effect of exchange rate on
22、export, I make a Granger Causality test and derive the following result. P-value is small (p=0.0210), so we can reject hypothesis of non-causality and indicate that difference of exchange rate does predictively cause ln(export) and help to predict it. 3.3 Model testFirst, the author tests if the err
23、or term of the whole model is white noise. There are two methods. The first one is Q-test.The result 0.7205 shows it is white noise.For second method, the author derives the autocorrelation of residuals from the stata. And we can make a conclusion that the error term is white noise. e from the graph
24、.Second, the author uses the White Test to see if there is heterokedasticity in the model. Then derive the following result. The P-value is 0.4583, so we cannot reject the hypothesis of homoskedasticity.4. Results and forecasts4.1 Evaluating the resultsAfter setting up the model, we can use it to ge
25、nerate the fitted values and compare the result with the actual values to make sure that the model really fit the historical data well. Basically, the forecast model is a good one known from the following graph.4.2 Making forecastThe author uses the model derived to estimate one-step-ahead to twelve
26、-step-ahead point and interval forecast. one-step-ahead forecastYt+1=0+1t+Zt+1yt+.13yt-12from two to twelve step-ahead forecast, the forecast error is not WN, it is MA(h-1).since it is correlated, the author use a new way to get better results. Yt+2=0+1t+1Zt+2Zt-1+1yt+.13yt-12Yt+12=0+1t+1Zt+.12Zt-12
27、+1yt+.13yt-12The results are as follows.5. Conclusion This paper analyze the export data from 2002m1 and establish AR(13) model with trend and distributed lags of exchange rate. Then successfully predict the export in the future year. According to our prediction, the export increase steadily in the
28、next year. The prediction conforms to the reality for the following reasons.For the first place, after the financial crisis in 2008, the global economy step into a new business cycle. In current stage, the economy of U.S. is recovering and developing at a fast speed. The good economic condition will
29、 definitely promote export as a whole. Secondly, with the trend of globalization, China is making more policies to facilitate international trade. It has launched Shanghai Pilot Free-Trade Zone on September 29, 2013. Whats more, it will continue to establish Tianjin Pilot Free-Trade Zone during next
30、 year. These open reformations definitely lower the barriers of trade. U.S can export to China more easily. And the consumption demand in China will become larger and larger. Thirdly, the lasting depreciation of dollars to yuan make U.S goods cheaper in China. This trend will continue, for the simpl
31、e reason that China, as a new-born international economy after attending WTO, is playing an increasingly important role around the world. The demand for currency yuan is surging rapidly. Finally, the main goods that U.S exports to China are high-tech products such as electronic products and financia
32、l services. Nowadays, these kinds of industries are facing unprecedented opportunities to boom. Especially in U.S, where it is abundant in high-tech human capital, these industries can grow faster and contribute a large portion to export.REFERENCEClements M P, Smith J. Evaluating the forecast densities of linear and nonlinear models: applications to output growth and unemploymentJ. Journal of Forecasting, 2000, 19(4): 255-276.Coccari R L. Alternative
copyright@ 2008-2022 冰豆网网站版权所有
经营许可证编号:鄂ICP备2022015515号-1