1、<p>　　中文5694字，3500單詞，18900英文字符</p><p><b>　　畢業(yè)論文外文翻譯</b></p><p>　　外文題目：Causality between Export and Growth: Evidence from South Asian Countries

2、 </p><p>　　出 處： MPRA Paper No. 21027, posted 28. February 2010 / 13:01 </p><p>　　作 者： Eusuf, M Abu and Ahmed, Mansur </p><p>　　Ca

3、usality between Export and Growth: Evidence from South Asian Countries</p><p>　　M. Abu Eusuf and Mansur Ahmed Abstract</p><p>　　Strong economic growth accompanied with robust export performance

4、leads many people to conclude that export sector of a country has pivotal role in the economic growth of that country. Empirical evidence on export growth nexus has been mixed and inconclusive. This study examined whethe

5、r there was any time series support for such export-led growth hypothesis for South Asian Countries. Engle-Granger's Error Correction Model (ECM) was used to test the Granger causality between export and output. Th&l

6、t;/p><p>　　Key Words: Export-led growth hypothesis, Granger causality test, Unit Root Tests, Error Correction Model.</p><p>　　Introduction:</p><p>　　The export-led growth hypothesis (E

7、LGH) postulates that export expansion is one of the main determinants of economic growth. It holds that the overall growth of countries can be generated not only by increasing the amounts of labor and capital within the

8、economy, but also by expanding exports. Economists behind export-led growth hypothesis consider exports can perform as an “engine of growth". This type of advocacy has been generated from the following reasons: Firs

9、t, expansion in demand for the</p><p>　　A number of studies including Jung and Marshall (1985), Chow (1987), Darrat (1987), Hsiao (1987), Bahmani-Oskooee et al. (1991), Kugler (1991), Dodaro (1993), Van den

10、Berg & Schmidt (1994), Greenaway and Sapsford (1994), and Islam (1998) had adopted time series analysis for exploring the causal liaison between export growth and output growth. Using the Granger (1969), Sims (1972),

11、 and Hsiao (1987) causality procedures, these studies were failed to provide a uniform conclusion about the export-le</p><p>　　However, very few empirical studies have been done in the recent past to investi

12、gate the export- led growth (ELG) hypothesis for South Asian countries (Jung and Marshall 1985; Bahamani, Oskooe and Alse 1993, Dodaro 1993; Khan and Saqib 1993; Chandra 2000, 2002 and Begum and Shamsuddin (1998). The av

13、ailable evidence in relation to export-led growth in South Asia appears rather mixed. In case of India, Chandra (2000, 2002) found bidirectional causal relationship between export growth and GDP growt</p><p>

14、;　　The motivation for undertaking this study is thus threefold. First, by covering the entire South Asian region, it fills an important gap in the literature. Second, it tries to confirm the validity or otherwise of the

15、mixed results obtained in the empirical literature for South Asian as well as other countries. The causality directions between economic growth and exports have very crucial policy implications. Therefore, this study is

16、conducted to investigate the relationship between output and ex</p><p>　　to examine the short run and long run causality relationship between output and exports; and</p><p>　　to suggest some pol

17、icy implications .</p><p>　　Methodology:</p><p>　　Cointegration, Error-Correction Modeling and Granger Causality Tests</p><p>　　Before cointegration is applied, it is essential to t

18、est a time series for stationarity. A time series is stationary (in the sense of weak stationarity) if its mean, variance and covariance remain constant overtime. At a formal level, stationarity can be tested by determin

19、ing whether the data contain a unit root. This can be done by the Dickey and Fuller (1979), Augmented Dickey-Fuller (ADF) and Phillips and Perron (1988) tests. The ADF test is used here for testing for stationarity as we

20、ll as f</p><p>　　LGDP=θ+ηLX+u.......................................................................................... (5)</p><p>　　LX=δ+λLGDP+e...........................................

21、........................................... (6)</p><p>　　If LGDP and LX are both I(1), then for them to be cointegrated u and e should be stationary or I(0). To check whether there is valid long-run/cointegr

22、ating relationship among the variables, we need to test the stationarity of residuals (i.e. linear combination of variables) employing the ADF test, which is given in (7). The ADF test statistics is the t-ratio on the te

23、rm .The critical values for the test is given by McKinnon (1991).</p><p>　　?Et=ρEt-1+γ?Et-1+vt ........................................................................ (7)</p><p>　　Where ? is

24、 the first difference, Et is the residual from cointegrating regressions and vt is the white noise.</p><p>　　Once it is established that two variables are cointegrated, the next issue is that of which variab

25、le “causes” the other. Before the advent of cointegration and error-correction modeling, the standard Granger tests were used widely to determine the direction of causality. However, as noted earlier, the standard Grange

26、r method is likely to be misleading if variables are cointegrated since the standard tests do not contain an error-correction term. The error-correction representation of the Granger</p><p><b>　　p1q1&

27、lt;/b></p><p><b>　　?LGDPt</b></p><p>　　= C1 + ∑φ1m ?LX t ?m + ∑?1m ?LGDPt ?m + λ1ut ?1</p><p>　　................................ (8)</p><p><b>　

28、　m=1</b></p><p><b>　　m=1</b></p><p>　　?LX t= C2 + ∑φ2m ?LX t ?m + ∑?2m ?LGDPt ?m + λ2 et ?1.............................. .... (9)</p><p>　　Where the error-corre

29、ction terms ut-1 and et-1 are stationary residuals from the cointegrating equations. By introducing error-correction terms in the above equations, an additional channel is opened up through which causality is tested. For

30、 example, in equation (8), growth of real exports (?LX) is said to Granger cause real income growth (?LGDP) either when the coefficients of lagged ?LX are positive and jointly significant through the F-test or if λ1 is s

31、ignificant or both. If income growth ca</p><p>　　In the above analysis, the inclusion of the error-correction terms makes it possible to distinguish between short-term and long-term causality. The lagged cha

32、nges in the independent variables represent the short-run causal impact whilst the significance of the error-correction term gives the information on long-run causality.</p><p>　　Before implementing the Gran

33、ger Causality test one has to chose the order of lag (i.e. pi and qi, where i=1, 2) appropriately. There is evidence that the causality tests are often sensitive to the choice of the lag lengths. In the literature there

34、exist a number of suggested methods for choosing the lag orders. Here "simple to general" recommended by Engle and Granger (1987) has been followed. They favored starting with fewer lags and then testing for ad

35、ded lags. The idea is that if non-autocorr</p><p>　　Data Issues</p><p>　　The study uses annual time-series data from the IMF’s International Financial Statistics (1995, 1999, 2004, 2006). Export

36、s are in domestic currencies deflated by unit export values where available and the consumer price index if the unit export value index is unavailable. Only in the case of Srilanka, unit value indices of exports is avail

37、able continously; therefore, the consumer price indices are used instead for the other countries. All indices take 2000 as the base period (2000=100).</p><p>　　Jung and Marshall (1985) also make use of the e

38、xport price index or the consumer price index and point out that there are difficulties with both price indices. The consumer price index fails to pick up changes in terms of trade while the export price index is frequen

39、tly not a constant basket index but a unit value index, the composition of which varies. They also find that where exports are deflated by the consumer price index, the results are less favourable to the export-led growt

40、h hypothesis.</p><p>　　Furthermore, all of the series are transformed into log form. Log transformation can reduce the problem such as heteroscedasticity because it compresses the scale in which the variable

41、s are measured, thereby reducing a tenfold difference between two values to a twofold difference (Gujarati 1995). The period for each country is different and is dictated by the availability of data. The longest period i

42、s for India (1965-2005) ,Nepal (1965-2005), Sri Lanka (1965-2005), Pakistan (1965-2005), followe</p><p>　　Before we proceed further let us note a few methodological problems of the present study arising out

43、of data availability. Firstly, it can be argued that export-growth models based on a bivariate framework may be misspecified as besides exports other important variables such as terms of trade are omitted. If the objecti

44、ve is to have a comprehensive study for the entire region, in view of the data limitations, there is no escape from the bivariate methodology adopted here.</p><p>　　Finally, as noted in the previous section,

45、 the unit value index as well as the consumer price index is both problematic, but in the absence of any better alternative, one is left with no option but to use them. In the literature, therefore, a variety of indices

46、including the consumer price index, the unit value index for exports and the GDP deflator have been used, sometimes all within the same study.</p><p>　　The Results</p><p>　　3.1. Time Series Prop

47、erties of the variables</p><p>　　First, the Augmented Dickey-Fuller (ADF) tests, the popular tests for unit roots, have been performed for variables based on the following equation:</p><p>　　?Y=

48、 α + (ψ ? 1)Yt ?1 + ηT + γ?Yt ?1 + et ..................................................... (9)</p><p>　　The ADF test for unit root is based on equation (9) with the null hypothesis of (Ψ-1)=0 ( i.e. the Yt

49、is non-stationary) against the alternative of (Ψ-1)<0 (i.e. Yt is stationary).The t-test on the estimated coefficient of Yt-1 provides the ADF test for the presence of a unit root. Since the data are annual in nature,

50、 following the usual practice of unit root test we have used only one lag in equation. This is done to ensure that the error process in the estimating equation is residually uncorre</p><p>　　Tests for Co-int

51、egration: The Engle-Granger Procedure</p><p>　　Since pre-testing suggests all variables in our model are to be I(1), we compute what is known as the first step of Engle–Granger procedure. To check whether th

52、ere is valid long-run/cointegrating relationship among the variables, we need to test the stationarity of residuals (i.e. linear combination of variables) employing the ADF test. The ADF test statistics is the t-ratio on

53、 the term .The critical values for the test is given by McKinnon (1991).</p><p>　　Granger Causality Tests</p><p>　　As seen earlier, there are two ways in which causality can express itself: thro

54、ugh the F-test of joint significance of the lagged differenced terms, and through the error-correction term. The results are reported in Table 3. It can be seen that in both the cases of Bangladesh F-statistics for x→y a

55、nd y→x are insignificant at 95 percent level of confidence. Thus, the data suggest that there is no short-term causality in either direction. If one looks at the error-correction terms, they appear ins</p><p&g

56、t;　　Following the same approach in each of the given cases, it can be seen that in only three countries, India, Nepal and Maldives are cases of growth-led exports. This is in contrast with, in case of India, Chandra (200

57、0, 2002) found that there is a two-way relationship between export growth and GDP growth, but this relationship is short term in nature as real exports and real GDP do not exhibit cointegration or a long-term relationshi

58、p. But we have found unidirectional causality runs from growth t</p><p>　　The results for all South Asian countries show that the evidence for export-led growth in South Asia appears rather mixed. The mixed

59、nature of results, irrespective of the period taken, is hardly surprising for the countries of South Asia. Firstly, lack of conclusive evidence in favor of export-led growth reflects the inward-looking nature of these ec

60、onomies, where, for most of the relevant periods, trade was not considered even a handmaiden of growth, let alone an engine of growth. Thus lack of</p><p>　　Conclusion and Policy Implications</p><

61、p>　　This paper has studied the possibility of Granger causality between the logarithms of real exports and real GDP in seven South Asian countries for different time period. The study findings suggest that real export

62、s and real GDP are cointegrated only in Bangladesh, Pakistan and Nepal. While Pakistan, Srilanka and Bhutan are cases of export-led growth either short-term or long-term, India, Nepal, and Maldives show the opposite resu

63、lt of growth-led exports. In the Bangladesh’s case, the problem is </p><p><b>　　譯 文：</b></p><p>　　出口和經(jīng)濟增長間的因果關(guān)系研究—來自南亞國家的證據(jù)</p><p><b>　　摘要</b></p>&

64、lt;p>　　強勁的經(jīng)濟增長往往伴隨著良好的的出口實績讓許多人得出這樣的結(jié)論:出口領(lǐng)域在一個國家的經(jīng)濟增長中起關(guān)鍵作用。然而現(xiàn)有的實驗證據(jù)形形色色，沒有定論。本課題試圖驗證是否有任何時間序列支撐著針對于南亞國家的出口導向型經(jīng)濟增長假設(shè)。Engle-Granger的誤差修正模型(ECM) 是用來測試出口和產(chǎn)量的Granger因果關(guān)系的。這項研究已經(jīng)得出相當多的結(jié)果,但并沒有找到任何確鑿的證據(jù)來支持針對于南亞國家的出口導向型經(jīng)濟增長假

65、設(shè)。對于巴基斯坦、斯里蘭卡、不丹來說出口帶動了經(jīng)濟增長；而對于印度、尼泊爾和馬爾代夫的的調(diào)查卻顯示了完全相反的結(jié)果：經(jīng)濟增長帶動了出口。在一個國家,即孟加拉國，數(shù)據(jù)不能顯示出口與經(jīng)濟增長間誰是誰的因果關(guān)系。</p><p>　　關(guān)鍵詞:出口導向型經(jīng)濟增長假說，格蘭杰因果關(guān)系測試，單位根檢驗，誤差修正模型</p><p><b>　　簡介</b></p>

66、<p>　　出口導向型經(jīng)濟增長假說將出口增長假設(shè)為經(jīng)濟增長的主要決定因素。該假說認為一個國家的全面發(fā)展認為國家整體的成長不僅要通過增加勞動力和資本數(shù)量,也要通過擴大出口來促進。支持出口導向型經(jīng)濟增長的經(jīng)濟學家認為出口就像 “經(jīng)濟增長的引擎”。這種類型的倡導歸因于以下理由:第一,通過出口增長來擴大國家對外輸出的需求的增長有利于擴大小型開放經(jīng)濟的經(jīng)濟規(guī)模。第二，出口擴大將能緩解外匯緊張狀況，從而有利于滿足國內(nèi)進口輸入的要求，反過來

67、又能促進出口的擴大。 第三，出口擴大將促進出口產(chǎn)品生產(chǎn)的專業(yè)化水平，反過來將可能提高出口部門的生產(chǎn)力水平和普遍工藝水平。同時，這還有可能使資源分配從相對低效率的、非貿(mào)易部門向更高的多產(chǎn)的出口部門轉(zhuǎn)變。而生產(chǎn)力的改變又將促進產(chǎn)量的增長。第四，外向型的貿(mào)易政策將有利于接觸更先進的科學技術(shù)，更好的管理方法，從收獲中學習，將獲得更有效收獲。因此，國際貿(mào)易和發(fā)展理論認為歸因于出口導向型政策的出口增長積極地促進著一個國家經(jīng)濟的增長（以產(chǎn)量增長作為衡

68、量）。應該注意的是，這一理論同時也認為產(chǎn)量可以影響出口。有些專家，例如Kaldor (1967)、Lancaster (1980)和 Krug</p><p>　　出口和經(jīng)濟增長的關(guān)系的這一課題已在文獻中被做了廣泛地研究。很多早期的研究，包括Michaely (1977)，Balassa (1978)，Tyler (1981)，F(xiàn)eder (1983)， Kavoussi (1984)，Ram (1985)，Sh

69、eehey (1990)，Lopez (1991)，Edwards (1993)和Ngoc等等他們都是基于橫截面法顯著地證實了出口對經(jīng)濟增長具有意義深遠的因果影響。但是橫截面法具有自身的落后性，它設(shè)想而不是建立出口增長和GDP增長間的因果關(guān)系，同時成功的經(jīng)濟增長應該可以展示出較快的出口增長。這就使得Sheehey (1990) and Pritchett (1996)等作者根據(jù)截面研究作出的結(jié)論提出了疑問。Sheehey (1990)提

70、出，除了出口其他生產(chǎn)類別的增長和GDP增長間有類似的關(guān)系。</p><p>　　大量的研究包括Jung和 Marshall (1985)，Chow (1987)，Darrat (1987)，Hsiao (1987)，Bahmani-Oskooee (1991)，Kugler (1991)，Dodaro (1993)，Van den Berg & Schmidt (1994)，Greenaway和Sapsf

71、ord (1994)和Islam (1998)采用了時間序列分析法來探索出口增長和產(chǎn)值增長間的因果關(guān)系。但是用Granger (1969)、Sims (1972)和Hsiao (1987) 的因果關(guān)系程式來檢測，這些研究都不能提供一個統(tǒng)一的關(guān)于出口導向型假說的結(jié)論。但是時間序列研究并沒有受到輕視。盡管Granger 和 Sims 因果關(guān)系測驗只有當原始時間序列并非共和體的時候有效，而沒有一個研究證實涉及到由時間序列的變量參與的共同體的特

72、性。當時間序列被綜合在一起時，基于傳統(tǒng)時間序列模型的研究方法將會產(chǎn)生誤導。就如Granger (1988)指出，這是因為傳統(tǒng)的因果關(guān)系測試遺漏了一些“可預見性”，因此，就的出了關(guān)于因果關(guān)系的不正確的結(jié)論。此外，這所有的研究都是在二手</p><p>　　然而，最近的幾年里對于南亞國家出口導向型經(jīng)濟增長假說的實證研究不多。Jung和 Marshall（1985），Bahamani,Oskooe 和 Alse（199

73、3）, Dodaro （1993），Khan 和Saqib（1993），Chandra （2000，2002），Begum 和Shamsuddin (1998)。關(guān)于出口導向型增長的可用的證據(jù)形形色色。就印度而言，Chandra (2000, 2002)發(fā)現(xiàn)了出口增長和GDP增長的雙向因果關(guān)系是短期的因果關(guān)系，而出口增長和GDP增長間整合的時間序列關(guān)系則沒被發(fā)現(xiàn)。就巴基斯坦而言，Bahamani-Oskooee 和 Alse (1993

74、) ，Khan 和 Saqib(1993) 做了實驗并發(fā)現(xiàn)出口增長和產(chǎn)量增長間的雙向因果關(guān)系，而Jung和Marshall(1985) 則發(fā)現(xiàn)產(chǎn)量對出口增長具有反作用。Dodaro (1993)則從任何方面都沒有發(fā)現(xiàn)這兩者之間的什么有意義的關(guān)系。而同樣的研究，在斯里蘭卡他們都沒有發(fā)現(xiàn)這兩者之間的因果關(guān)系。Abhayaratne (1996)用整合法證實了前面一個發(fā)現(xiàn)。Dodaro (1993) </p><p>

75、　　研究此課題的動機重要有三層。首先，本課題覆蓋了整個南亞地區(qū)，填補了文獻資料空白。第二，本課題試圖證明從關(guān)于南亞國家的經(jīng)驗主義文獻資料上獲得的形形色色的結(jié)果的有效性。出口和經(jīng)濟增長的因果關(guān)系具有很強的政策內(nèi)涵。所以，本課題用最近的計量經(jīng)濟學方法，Engle-Granger誤差修正模型和Granger因果關(guān)系測試理論調(diào)查了以南亞國家為例調(diào)查了產(chǎn)量和出口間的關(guān)系。我們的具體目標是：（1）驗證產(chǎn)值和出口間的短期和長期因果關(guān)系；（2）提出一些

76、政策性內(nèi)涵。</p><p><b>　　二、實證方法</b></p><p>　?。ㄒ唬﹨f(xié)整檢驗、誤差修正模型和格蘭杰因果檢驗</p><p>　　應用協(xié)整檢驗之前，至關(guān)重要的是測試一個平穩(wěn)的時間序列。如果一個時間序列的均值，方差和協(xié)方差保持不變，那么它是固定的（在這個意義上有弱平穩(wěn)性）。在正常的水平，平穩(wěn)性可以通過測試數(shù)據(jù)是否包含單位根來確

77、定。這已經(jīng)Dickey 和Fuller (1979), Augmented Dickey-Fuller (ADF)，Phillips 和Perron (1988) 證實。在這里使用ADF檢驗是為了測試平穩(wěn)性，以及一系列的整合命令。</p><p>　　使用這些變量以致于第一個差異可以解釋為增長率。如果兩個變量LX（即實際出口對數(shù)）和LGDP（實際GDP對數(shù)）集成一個序列，即I(1)，那么下一步是發(fā)現(xiàn)它們是否存在協(xié)

78、整關(guān)系。這可以通過估計以下協(xié)整方程的OLS和測試它們殘差量的平穩(wěn)性。</p><p>　　LGDP=θ+ηLX+u....................................................................................... (5)</p><p>　　LX=δ+λLGDP+e ........................

79、.............................................................. (6)</p><p>　　如果LGDP和LX都是I（1）中的，然后對它們進行協(xié)整檢驗，u和e應該是固定的或是屬于I(0)。要檢查long-run/cointegrating變量之間的關(guān)系是否是有效的，我們需要通過ADF檢驗來知道殘差的平穩(wěn)性（即變量的線性組合），這在(7)中給出。ADF檢

80、驗量是長期的T比率。對于臨界值的檢驗來自McKinnon (1991).</p><p>　　ΔEt=ρEt-1+γΔEt-1+vt ................................................................................ (7)</p><p>　　一旦確定兩個變量是協(xié)整的，接下來的問題是哪些變量之間有因果關(guān)系。在協(xié)

81、整和誤差修正模型出現(xiàn)之前，格蘭杰檢驗標準被廣泛地用來確定因果關(guān)系的方向。然而，如前面所述，如果變量是協(xié)整的，標準格蘭杰方法很可能是誤導， 因為標準測試不包含誤差修正項。用誤差修正來表示格蘭杰因果關(guān)系模型的兩個變量的制定如下：</p><p>　　誤差修正項ut-1和et-1是協(xié)整方程里的固定殘差，通過引入上述方程的誤差修正項，測試因果關(guān)系的另一途徑開辟了。例如，在方程（8）中，實際出口增長（ΔLX）表示格蘭杰檢驗

82、引起實際收入增長（ΔLGDP），通過F檢驗得知滯后系數(shù)ΔLX是積極的且共同顯著的或是λ1顯著或兩者兼而有之。如果收入的增長導致出口增長，無論滯后系數(shù)ΔLGDP是積極的或共同顯著（F檢驗）或λ2是顯著的或是兩者兼有（方程（9））。因此，誤差修正模型允許的事實是因果關(guān)系表明通過滯后變化影響獨立變量或通過誤差修正項或是通過兩者。</p><p>　　在上面的分析中，對誤差修正項的列入使它能夠區(qū)分短期和長期之間的的因果關(guān)

83、系。獨立變量中的滯后變化代表短期因果關(guān)系的影響，而誤差修正項的顯著性提供了在長期的因果關(guān)系中的信息。</p><p><b>　　(二)數(shù)據(jù)問題</b></p><p>　　本次研究所用的年度時間序列數(shù)據(jù)來自國際貨幣基金組織的國際金融統(tǒng)計（1995，1999，2004，2006）。出口指的是出口單位值可用來表示國內(nèi)貨幣貶值，而單位出口價值指數(shù)不可用來表示居民消費價格指

84、數(shù)。只有在斯里蘭卡的例子里，出口單位價值指數(shù)是連續(xù)可用的，因此，消費者價格指數(shù)用來代替其他國家。所有指數(shù)以2000年為基期（2000年=100）。</p><p>　　Jung和Marshall (1985)利用出口價格指數(shù)和居民消費價格指數(shù)指出兩種價格指數(shù)的困難。居民消費價格指數(shù)未能顯示出貿(mào)易條件的變化，出口價格指數(shù)往往不是一個固定籃子指數(shù)而是一個單位價值指數(shù)，其中的組成各不相同。他們還發(fā)現(xiàn)其中出口貶值由居民消

85、費價格指數(shù)引起，結(jié)果是不太有利于出口帶動經(jīng)濟增長的假說。這一點很重要，因此，要牢記這些因素，來解讀目前的研究成果。</p><p>　　此外，所有系列轉(zhuǎn)化為對數(shù)形式。對數(shù)變換可以減少問題例如異方差性，因為它縮小了變數(shù)衡量的規(guī)模，從而減少了兩個值兩方面十倍的差異（Gujarati（1995）)。每個國家的時期是不同的，是由數(shù)據(jù)的可用性來支配的。最長的時期是印度（1965年—2005年），尼泊爾（1965年—2005

86、年），斯里蘭卡（1965年—2005年）， 巴基斯坦（1965年—2005年），其次是孟加拉國（1980年—2005年），馬爾代夫（1980年—2005年）和不丹（1980年—2004年）。</p><p>　　在我們采取進一步研究前讓我們注意到了幾個目前研究所產(chǎn)生的數(shù)據(jù)可用性的方法論問題。首先，它可以表明出口增長模型基于二元框架可能是錯誤地指定出口之外的其他重要變量，例如貿(mào)易術(shù)語被忽略。如果目標是為了整個地區(qū)的

87、綜合研究，鑒于對數(shù)據(jù)的限制，這也沒有擺脫二元方法論在此通過。</p><p>　　最后，正如前一節(jié)所述，單位價格指數(shù)和居民消費價格指數(shù)都是存在問題的，但在沒有任何其他更好的選擇下，是別無選擇的使用它們。因此在文獻中，不同的指標包括居民消費價格指數(shù)，出口價格指數(shù)和單位GDP平減指數(shù)都已經(jīng)被使用，有時所有出現(xiàn)在同一個研究中。</p><p><b>　　三、研究結(jié)果</b>

88、;</p><p>　?。ㄒ唬r間序列變量的性質(zhì)</p><p>　　首先，ADF檢驗，單位根的顯著性檢驗，變量的執(zhí)行都基于下面的方程：</p><p>　　對單位根的ADF檢驗基于方程（9）的零假設(shè)(Ψ-1)=0（即Yt是不固定的），反對選擇(Ψ-1)<0（即Yt是固定的）。在估計系數(shù)Yt-1應用T檢驗以此提供了一個存在單位根的ADF檢驗。由于數(shù)據(jù)每年的性質(zhì)

89、，根據(jù)單位根檢驗的慣例，我們只運用一個滯后變量在方程里，這樣做是為了確保在估計方程錯誤的過程中residually是不相關(guān)的。(Ψ-1)中的T比率提供了ADF的統(tǒng)計數(shù)據(jù)?，F(xiàn)在，單位根的PP檢驗已被廣泛應用于光的事實，往往在時間序列出現(xiàn)異方差性和非正常的原始數(shù)據(jù)，而ADF檢驗不被計入考慮中。PP檢驗事實上是經(jīng)過調(diào)整后的T比率應用于方程（9）的(Ψ-1)上，似乎在協(xié)整文獻里有個共識是PP檢驗對于ADF檢驗是可取的。從ADF和PP檢驗的結(jié)果可

90、以看出整合的相關(guān)變量的順序。一個時間序列綜合的階d[通常記為I（d）]與該系列的次數(shù)d必須有差異的才能成為固定。計量經(jīng)濟學軟件Eviews3.0和Microfit 4.0版本被用于各自的測試。</p><p>　?。ǘ﹨f(xié)整檢驗：恩格爾—格蘭杰程序 </p><p>　　由于前期檢驗表明在我們的模型中全部的變量都在I（1）中，我們計算什么是恩格爾—格蘭杰程序的第一步。要確定long-ru

91、n/cointegrating變量之間的關(guān)系是否有效，我們需要采用ADF檢驗來測試殘差的平穩(wěn)性（即線性變量的組合）。ADF檢驗統(tǒng)計量是長期的T比率，測試的臨界值來自于McKinnon (1991)。</p><p>　?。ㄈ└裉m杰因果檢驗</p><p>　　正如前面所看到的，有兩種方式可以表示自己的因果關(guān)系：通過F檢驗來測試綜合意義上的滯后差分項和通過誤差修正項?？梢钥闯?，在孟加拉國的

92、F統(tǒng)計量中無論是x→y還是y→x，95%的信心水平都是微不足道的。因此，這些數(shù)據(jù)表明，在任何方向都沒有短期的因果關(guān)系。如果從誤差修正項看來，他們在孟加拉國的兩個方程中都是不重要的，這意味著不存在長期因果關(guān)系從實際收入增長到出口增長。這與前面提到的Dodaro的研究做對比，發(fā)現(xiàn)孟加拉國出口帶動經(jīng)濟增長的證據(jù)（雖然因果關(guān)系是短期性質(zhì)）。</p><p>　　對于每一個給定的情況下使用同樣的方法，可以看出，只有三個國家

93、，印度，尼泊爾和馬爾代夫是出口導向型的經(jīng)濟增長的例子。這是與印度的例子作對比，Chandra (2000, 2002)發(fā)現(xiàn)在出口增長和GDP增長之間存在雙向關(guān)系，但這種關(guān)系是短期的性質(zhì)，實際出口和實際國內(nèi)生產(chǎn)總值不表現(xiàn)出協(xié)整或一個長期的關(guān)系，但我們發(fā)現(xiàn)在印度的例子里，單向因果關(guān)系從經(jīng)濟增長到出口甚至表現(xiàn)出短期的性質(zhì)。巴基斯坦和斯里蘭卡顯示出口帶動經(jīng)濟增長的證據(jù)。在不丹例子里可以看出，雖然沒有表明短期因果關(guān)系的證據(jù)，但有一個長期的因果關(guān)系

94、在出口國內(nèi)生產(chǎn)總值之間。</p><p>　　所有南亞國家的結(jié)果表明，以出口為主導的增長在南亞國家的證據(jù)顯得參差不齊。性質(zhì)表明，不論期間采取什么方法，在南亞國家都不足為奇。首先，在出口帶動的經(jīng)濟增長方面缺乏確鑿有利的證據(jù)來反映了這些經(jīng)濟內(nèi)向型的性質(zhì)，其中，在有關(guān)時期，貿(mào)易被認為不是“增長的侍女”而是“增長的引擎”，因此，對出口帶動的增長遠沒有取得成功，只是可以預料的。其次，混合性質(zhì)的結(jié)果在本研究報告中是符合文獻的

95、主題的。事實上，正如我們所指出的，關(guān)于這個問題在早期文獻中并未有報告任何確鑿的證據(jù)，以出口為主導的經(jīng)濟增長在南亞受到普遍青睞。</p><p><b>　　四、結(jié)論和政策內(nèi)涵</b></p><p>　　本課題研究了在不同時期7個南亞國家真實的出口和真實的GDP對數(shù)間因果關(guān)系的可能性。課題研究表明真實出口和真實GDP只有在孟加拉國，巴基斯坦和尼泊爾作為共同體。而在巴基

96、斯坦、斯里蘭卡和不丹，這兩者之間既存在短期的因果關(guān)系又存在長期的因果關(guān)系。而在印度，尼泊爾和馬爾代夫則顯示了完全相反的結(jié)果經(jīng)濟增長導向出口。在孟加拉國的案例中，實際的出口增長并沒有帶來出口導向型經(jīng)濟增長。這就意味著帶來相關(guān)GDP的增長。結(jié)論是，關(guān)于這一題材的與當前現(xiàn)狀相一致的文獻形形色色，而不能給出口導向型增長的文章提供有利的證據(jù)。對于這些以內(nèi)向型計劃導向的經(jīng)濟體以來說，調(diào)查結(jié)果的多樣性并不足以為奇，他們大多將進口替代出口增長占有主要的