Good evening,
I have a question regarding the Johansen cointegration test in Stata.
I have two monthly time-series variables, i.e., Vessel Prices and Vessel Earnings for the last 30 years.
I have checked for stationarity (Dickey-fuller test and Phillips Perron for unit root test) in the log levels and log first differences. I concluded that log levels are non-stationary while log first differences are stationary.
Then I have to check their cointegration relationship. Hence, I run the Johansen cointegration test. In case that the variables are cointegrated then I should run the VECM.
I would like to point out, that aiming to double check my results, I have run all tests both in Stata and in Excel (XLSTAT).
My results in these two tools, regarding stationarity are the same.
However, when I run the cointegration test, Stata and Excel do not reach the same result.
According to my research in the literature, the variables should be cointegrated and there should be one cointegration relation after performing the Jonasen test. But, in Stata I can not conclude to this result.
Below and/ or attached you can find for your reference, relevant outputs both from Stata and excel.
Have I made any mistake in my commands in Stata? Why can I not confirm also in Stata that my variables are cointegrated? Is it possible to have the expected result in excel and not in Stata? Can anyone assist me?
STATA:
. import excel "C:\Users\user11\Desktop\Data.xlsx", sheet("data") firstrow
. gen time=_n
. tsset time
time variable: time, 1 to 368
delta: 1 unit
. gen logEarningsCape=ln(yTCRateCapesizemonth)
. gen logPricesCape=ln(ySPCapesize)
. gen d_LogEarningsCape=d.logEarningsCape
(1 missing value generated)
. gen d_LogPricesCape=d.logPricesCape
(1 missing value generated)
EXCEL (XLSTAT OUPTUP-only Johansen Test):
I have a question regarding the Johansen cointegration test in Stata.
I have two monthly time-series variables, i.e., Vessel Prices and Vessel Earnings for the last 30 years.
I have checked for stationarity (Dickey-fuller test and Phillips Perron for unit root test) in the log levels and log first differences. I concluded that log levels are non-stationary while log first differences are stationary.
Then I have to check their cointegration relationship. Hence, I run the Johansen cointegration test. In case that the variables are cointegrated then I should run the VECM.
I would like to point out, that aiming to double check my results, I have run all tests both in Stata and in Excel (XLSTAT).
My results in these two tools, regarding stationarity are the same.
However, when I run the cointegration test, Stata and Excel do not reach the same result.
According to my research in the literature, the variables should be cointegrated and there should be one cointegration relation after performing the Jonasen test. But, in Stata I can not conclude to this result.
Below and/ or attached you can find for your reference, relevant outputs both from Stata and excel.
Have I made any mistake in my commands in Stata? Why can I not confirm also in Stata that my variables are cointegrated? Is it possible to have the expected result in excel and not in Stata? Can anyone assist me?
STATA:
. import excel "C:\Users\user11\Desktop\Data.xlsx", sheet("data") firstrow
. gen time=_n
. tsset time
time variable: time, 1 to 368
delta: 1 unit
. gen logEarningsCape=ln(yTCRateCapesizemonth)
. gen logPricesCape=ln(ySPCapesize)
. gen d_LogEarningsCape=d.logEarningsCape
(1 missing value generated)
. gen d_LogPricesCape=d.logPricesCape
(1 missing value generated)
EXCEL (XLSTAT OUPTUP-only Johansen Test):
| VAR order estimation: | |||||
| Number of lags | AIC | HQ | BIC | FPE | |
| 1 | -9,902 | -9,885 | -9,859 | 0,000 | |
| 2 | -10,156 | -10,121 | -10,070 | 0,000 | |
| 3 | -10,146 | -10,095 | -10,017 | 0,000 | |
| 4 | -10,134 | -10,066 | -9,962 | 0,000 | |
| 5 | -10,114 | -10,029 | -9,900 | 0,000 | |
| The VAR order estimate according to BIC is 2. | |||||
| Lambda max test: | |||||
| H0 (Nbr. of cointegrating equations) | Eigenvalue | Statistic | Critical value | p-value | |
| None | 0,070 | 26,411 | 15,892 | 0,001 | |
| At most 1 | 0,021 | 7,800 | 9,164 | 0,090 | |
| Lambda max test indicates 1 cointegrating relation(s) at the 0,05 level. | |||||
| Trace test: | |||||
| H0 (Nbr. of cointegrating equations) | Eigenvalue | Statistic | Critical value | p-value | |
| None | 0,070 | 34,211 | 20,262 | 0,000 | |
| At most 1 | 0,021 | 7,800 | 9,164 | 0,090 | |
| Trace test indicates 1 cointegrating relation(s) at the 0,05 level. | |||||
| Adjustment coefficients (alpha): | |||||
| logEarningsCape | 0,010 | -0,019 | |||
| logPricesCape | -0,010 | -0,006 | |||
| Cointegration coefficients (beta): | |||||
| logEarningsCape | -4,282 | -0,681 | |||
| logPricesCape | 6,346 | 3,640 | |||
| Intercept | -54,213 | -54,650 | |||
| Normalized to beta'.S11.beta = Id. | |||||
