Nonparametric cointegration analysis



7. Cointegrating systems with unconstrained drift

7.1. Non-seasonal drift

Until so far all our derivations were based on the assumption that the drift parameter
vector μ is orthogonal to the cointegrating vectors. Cf. Assumption 3. The problem is that
without Assumption 3 the result of Lemma 2 no longer holds, due to the fact that

Σn=11cos(2kπt/n) = n/2, although Σnt=1cos(2kπt/n) = 0, so that, with Fk(x) = cos(2kπx), the result
of Lemma 2 now becomes:

RM(Ft√n - 1rμ.-'

RM (Fk) n
                k

(          __________'

1CD (1) Yk ftFk(x )2dx

`    Fk(1)R,TD,Z

and consequently, part (15) of Lemma 4 becomes:


D

RC-A R
m


(

ɪ _

RqT-rC(DE XkXkTC(1)TR,-t

k- 1

m _

RrT μ∑ XkTC (1)R

F               k1


'

RTCd)∑ Xt μ tr
-,

Rr μμtR


Clearly, this will render all our test results invalid. However, a minor change of the functions
Fk will cure the problem:


Lemma 8. Let Fn,k(x) = cos[2kπ(nx - ½)/n] and Fk(x) = cos(2kπx). Then


nn

Fnk (t (t/n ι = tLk(τ/.)( t/n 1 = 0,

t=1                            t-1


(38)


and


lim n sup F^nk(x) - Fk(x) = kπ .
n -∞ 0x1     ’


(39)


23




More intriguing information

1. Structural Breakpoints in Volatility in International Markets
2. Cross border cooperation –promoter of tourism development
3. RETAIL SALES: DO THEY MEAN REDUCED EXPENDITURES? GERMAN GROCERY EVIDENCE
4. 101 Proposals to reform the Stability and Growth Pact. Why so many? A Survey
5. From music student to professional: the process of transition
6. The name is absent
7. A Unified Model For Developmental Robotics
8. The name is absent
9. The quick and the dead: when reaction beats intention
10. Innovation and business performance - a provisional multi-regional analysis