第五章 結論與未來研究方向
B.1 台北、彰化玫瑰月拍賣均價共整合分析
(a)假設模型
以台北玫瑰月拍賣均價做為相依變數,探討台北與彰化玫瑰月拍賣均價是否存在共整合 關係,其模型假設為:z(t)-z(t-1) = A(1)(z(t-1)-z(t-2)) + .... + A(p)(z(t-p+1)-z(t-p)) + B.H'z(t-p) + c + u(t),其中 z(t)為包含兩變數所組成之向量,分別為台北與彰化玫瑰月拍賣均價,H'z(t-p)為 其誤差修正項,c 則為常數項,u(t)為兩向量之誤差項。此處落後期數先以 VAR 模型以 AIC 法所選取12 期落後為其落後期數。資料時間自 1997 年 1 月至 2003 年 11 月,共 83 個樣本。
此處已判別過兩時間數列變數經過二階差分後為同條件程序,因此可進行後續共整合分析。
表B-1 兩變數共整合檢定之假設模型
(b)最大特徵值檢定
Johansen 提出兩種概似比檢定統計量來檢定共整合向量的個數,分別為跡檢定與最大特 徵值檢定兩種方法。其中最大特徵值檢定之虛無假設與對立假設如下,
H0:共整合向量=r H1:共整合向量=r+1
最大特徵值檢定先假設變數間不存在共整合關係,即 r=0,若拒絕該假設則依次增加向 量個數再行檢定,直到完全無法拒絕假設為止,檢定結果若存在有一個或多個顯著的特性根,
LR test (Lambda-max test) of the null hypothesis that there are r cointegrated vectors against the alternative that there are r + 1 cointegrated vectors
Table 1: No restriction on intercept.
C.f. Johansen & Juselius (1990), Table A1 critical values conclusions:
r test statistic 20% 10% 5% 20% 10% 5%
0 8.4 10.1 12.1 14.0 accept accept accept 1 1.8 1.7 2.8 4.0 reject accept accept LR test (Lambda-max test) of the null hypothesis that there are r cointegrated vectors against the alternative that there are r + 1 cointegrated vectors
Table 2: Restrictions on intercept, but not imposed.
C.f. Johansen & Juselius (1990), Table A2 critical values conclusions:
r test statistic 20% 10% 5% 20% 10% 5%
0 8.4 10.7 12.8 14.6 accept accept accept 1 1.8 4.9 6.7 8.1 accept accept accept
則表示相關變數之間具有長期共整合的均衡關係。經由上述兩種統計量,可決定共整向量
r
的個數,以判斷變數間是否具有共整合關係。於最大特徵值檢定中,Table 1 與 Table 2 分別為截距項無限制與截距項存在共整合限制 之假設,其分別於20、10、5%臨界值情況下所得之結果。
表B-2 最大特徵值檢定輸出結果
(c) 跡檢定
跡檢定之虛無假設與對立假設如下,
H0:共整合向量≦r H1: 共整合向量>r
檢定統計量λmax
(
Q:qq+1)
=−Tλn(1−λˆr+1)跡檢定先假設變數間不存在共整合關係,即共整合向量=0,若拒絕該假設則依次增加向 量個數再行檢定,直到完全無法拒絕假設為止,檢定結果若存在有一個或多個顯著的特性根,
則表示相關變數之間具有長期共整合的均衡關係。經由上述兩種統計量,可決定共整向量
r
的個數,以判斷變數間是否具有共整合關係。於跡檢定中,Table 1 與 Table 2 分別為截距項無限制與截距項存在共整合限制之假設,
其分別於20、10、5%臨界值情況下所得之結果。Easy Reg 並將檢定結果列示於表末,表(b) 末顯示台北與彰化玫瑰月拍賣均價經線性組合後,為一個同條件程序之時間數列,代表兩個
LR test (trace test) of the null hypothesis that there are at most r cointegrated vectors against the alternative that there are 2 cointegrated vectors
Table 1: No restriction on intercept.
C.f. Johansen & Juselius (1990), Table A1 critical values conclusions:
r test statistic 20% 10% 5% 20% 10% 5%
1 1.8 1.7 2.8 4.0 reject accept accept 0 10.2 11.2 13.3 15.2 accept accept accept
LR test (trace test) of the null hypothesis that there are at most r cointegrated vectors against the alternative that there are 2 cointegrated vectors
Table 2: Restrictions on intercept, but not imposed.
C.f. Johansen & Juselius (1990), Table A2 critical values conclusions:
r test statistic 20% 10% 5% 20% 10% 5%
1 1.8 4.9 6.7 8.1 accept accept accept 0 10.2 13.0 15.6 17.8 accept accept accept
Conclusion: r =2
N.B.: This means that z(t) is (trend) stationary!
時間數列間存在共整合關係,
表B-3 跡檢定輸出結果
(b)與(c)顯示無論使用跡檢定法或最大特徵值檢定法,均存在共整合向量,因此,台北與 彰化玫瑰月拍賣均價存在亦步亦趨的共整關係。
z(t,1) = DIF1[DIF1[台北玫瑰拍賣均價 z(t,2) = DIF1[DIF1[彰化玫瑰拍賣均價
Dependent variables:
Y(1) = DIF1[DIF1[台北玫瑰拍賣均價 VAR(p) model:
z(t) = A(1)z(t-1) + ... +A(p)z(t-p) + B.d(t) + u(t), where d(t) is a vector of deterministic variables:
d(t)=1
Chosen VAR(p) order: p = 12
Zeros in matrices A(1) A(12): (1 = nonzero)
OLS estimation results:
y(1) = DIF1[DIF1[台北玫瑰拍賣均價
Explanatory variables OLS estimate t-value [p-value]
x(1,1) = LAG1[DIF1[DIF1[台北玫瑰.. -15.42355E-01 -3.10 [0.00196]
x(1,2) = LAG1[DIF1[DIF1[彰化玫瑰.. 18.76816E-02 0.35 [0.72438]
x(1,3) = LAG2[DIF1[DIF1[台北玫瑰.. -19.71723E-01 -2.31 [0.02111]
x(1,4) = LAG2[DIF1[DIF1[彰化玫瑰.. 60.23025E-02 0.66 [0.51147]