0% found this document useful (0 votes)
12 views15 pages

Trend Cycle

1. The document discusses decomposing an economic time series (yt) into a deterministic trend component (TDt) and a stochastic noise component (Zt) that may include cycles. 2. Zt can be integrated of order 1 (I(1)), meaning it follows a random walk, or integrated of order 0 (I(0)), meaning it is stationary. The Beveridge-Nelson and Unobserved Components Model decompositions define the stochastic trend in Zt as a random walk. 3. The Beveridge-Nelson decomposition defines the permanent component of an I(1) series (yt) as the limiting forecast as the horizon goes to infinity. It shows the BN trend follows a

Uploaded by

Gelssomina
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
12 views15 pages

Trend Cycle

1. The document discusses decomposing an economic time series (yt) into a deterministic trend component (TDt) and a stochastic noise component (Zt) that may include cycles. 2. Zt can be integrated of order 1 (I(1)), meaning it follows a random walk, or integrated of order 0 (I(0)), meaning it is stationary. The Beveridge-Nelson and Unobserved Components Model decompositions define the stochastic trend in Zt as a random walk. 3. The Beveridge-Nelson decomposition defines the permanent component of an I(1) series (yt) as the limiting forecast as the horizon goes to infinity. It shows the BN trend follows a

Uploaded by

Gelssomina
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Trend-Cycle Decompositions

Eric Zivot
April 22, 2005

1 Introduction
A convenient way of representing an economic time series yt is through the so-called
trend-cycle decomposition
yt = T Dt + Zt (1)
where T Dt represents the deterministic trend and Zt represents the stochastic, and
possibly cyclic, noise component. For simplicity, the deterministic trend takes the
form of a simple time trend
T Dt = κ + δt (2)
and the noise component has a finite order ARIMA(p,d,q) representation

φ(L)Zt = θ(L)εt , εt ∼ W N(0, σ 2 )

where φ(L) = 1 − φ1 L − · · · − φp Lp and θ (L) = 1 + θ1 L + · · · + θq Lq . It is assumed


that the polynomial φ(z) = 0 has at most one root on the complex unit circle and
θ(z) = 0 has all roots outside the unit circle.

1.1 Trend Stationary and Difference Processes


The series yt is called trend stationary if the roots of φ(z) = 0 are outside the unit
circle and is called difference stationary if φ(z) = 0 has one root on the unit circle
and the others outside the unit circle. If yt is trend stationary then φ(L) is invertible
and Zt has the stationary or Wold representation

Zt = φ(L)−1 θ(L)εt
= ψ(L)εt
P
where ψ(L) = φ(L)−1 θ(L) = ∞ k
k=0 ψ k L with ψ 0 = 1 and ψ(1) 6= 0. Alternatively, if
yt is difference stationary then φ(L) can be factored as φ(L) = (1 − L)φ∗ (L) where

1
φ∗ (z) = 0 has all p − 1 roots outside the unit circle. In this case, ∆Zt has the
stationary ARMA(p − 1, q) representation

∆Zt = φ∗ (L)−1 θ(L)εt (3)


= ψ∗ (L)εt
P
where ψ∗ (L) = φ∗ (L)−1 θ(L) = ∞ ∗ k ∗ ∗
k=0 ψ k L with ψ 0 = 1 and ψ (1) 6= 0.

Example 1 Difference stationary AR(2)

Let φ(L)Zt = εt with φ(L) = 1 − φ1 L − φ2 L2 . Assume that φ(z) = 0 has one root
equal to unity and the other root real valued with absolute value less than 1. Factor
φ(L) so that
φ(L) = (1 − φ∗ L)(1 − L) = φ∗ (L)(1 − L)
where φ∗ (L) = 1 − φ∗ L with |φ∗ | < 1. Then

φ(L)Zt = (1 − φ∗ L)(1 − L)Zt = (1 − φ∗ L)∆Zt

so that ∆Zt follows an AR(1) process.

1.2 I(1) and I(0) Processes


If the noise series Zt is difference stationary then we say that Zt is integrated of order
1 and we write Zt ∼ I(1). To see why, note that from (3) above ∆Zt = ψ∗ (L)εt = ut
where ut is stationary. It follows that Zt = Zt−1 + ut and by recursive substitution
starting at time t = 0 we have
X
t
Zt = Z0 + uk
k=1

so that Zt can be represented as the (integrated) sum of t stationary innovations


{uk }tk=1 . Moreover, since ut is stationary we say that ut is integrated of order zero,
and write ut ∼ I(0), to signify that ut cannot be written as the sum of stationary
innovations. It should be clear that an I(1) series can be converted to an I(0) series
by first differencing.
If ∆Zt is an ARMA(p,q) process then Zt is called an ARIMA(p,1,q) process. The
term ARIMA refers to an autoregressive integrated moving average process.

1.3 Impulse Response Functions from I(1) Processes


Consider an I(1) process with Wold representation ∆yt = ψ∗ (L)εt . Since ∆yt =
yt − yt−1 the level yt may be represented as

yt = yt−1 + ∆yt

2
Similarly, the level at time t + h may be represented as

yt+h = yt−1 + ∆yt + ∆yt+1 + · · · + ∆yt+h

The impulse response on the level of yt+h of a shock to εt is


∂yt+h ∂∆yt ∂∆yt+1 ∂∆yt+h
= + + ··· + (4)
∂εt ∂εt ∂εt ∂εt
∗ ∗
= 1 + ψ1 + · · · + ψh

From (4), the long-run impact of a shock to the level of yt is given by

∂yt+h X ∗

lim = ψj = ψ∗ (1).
h→∞ ∂εt
j=1

Hence, ψ∗ (1) measures the permanent effect of a shock, εt , to the level of yt .


Remarks:

1. Since ∂y t
∂εt
= 1 it follows that ψ∗ (1) can also be interpreted as the long-run effect
of a shock relative to the immediate effect of a shock.

2. ψ∗ (1) is a natural measure of the importance of a permanent shock. If ψ∗ (1) = 1


then the long-run effect of a shock is equal to the immediate effect; if ψ∗ (1) > 1
the long-run effect is greater than the immediate effect and if ψ∗ (1) < 1 the
long-run effect is less than the immediate effect.

3. If ψ∗ (1) = 0 then Zt ∼ I(0). To see this suppose Zt ∼ I(0) and has the
Wold representation Zt = ψ(L)εt with ψ(1) 6= 0. Then ∆Zt = (1 − L)Zt =
(1 − L)ψ(L)εt = ψ∗ (L)εt where ψ∗ (L) = (1 − L)ψ(L). It follows that ψ∗ (1) =
(1 − 1)ψ(1) = 0.

1.4 Forecasting from an I(1) Process


Forecasting from an I(1) process follows directly from writing yt+h as

yt+h = yt + ∆yt+1 + ∆yt+2 + · · · + ∆yt+h

Then

yt+h|t = yt + ∆yt+1|t + ∆yt+2|t + · · · + ∆yt+h|t


X
h
= yt + ∆yt+s|t
s=1

Notice that forecasting an I(1) process proceeds from the most recent observation.

3
Example 2 Forecasting from an AR(1) model for ∆yt

Let ∆yt follow an AR(1) process

∆yt − µ = φ(∆yt−1 − µ) + εt , εt ∼ W N(0, σ 2 )

where |φ| < 1. Using the chain-rule of forecasting, the h-step ahead forecast of
∆yt+h based on information at time t is

∆yt+h|t = µ + φh (∆yt − µ) (5)

Using (5), the h-step ahead forecast of yt+h is

X
h
yt+h|t = yt + [µ + φs (∆yt − µ)] (6)
s=1
X
h
= yt + hµ + (∆yt − µ) φh
s=1

2 The Trend-Cycle Decomposition with Stochas-


tic Trends
Assume that Zt ∼ I(1). Then it is possible to decompose Zt into a stochastic (random
walk) trend and a stationary, I(0), “cyclical” component:

Zt = T St + Ct (7)

The stochastic trend, T St , captures shocks that have a permanent effect on the level
of yt and the stationary component, Ct , captures shocks that only have a temporary
effect on the level of yt . Given the decomposition of Zt , the representation for yt
becomes
yt = T Dt + T St + Ct (8)
where T Dt + T St measures the overall or total trend and Ct represents the deviations
about the trend.
The decomposition of Zt in (7) is not unique. In fact, there are an infinite number
of such combinations depending on how T St and Ct are defined. Two decompositions
have been popular in the empirical literature: the Beveridge-Nelson (BN) decom-
position; and the orthogonal unobserved components (UC0) decomposition. Both
decompositions define T St as a pure random walk. They primarily differ in how
they model the serial correlation in ∆Zt . The BN decomposition uses an unrestricted
ARMA(p, q) model for ∆Zt , and the UC0 model uses a restricted ARMA(p,q) for
∆Zt .

4
2.1 The Beveridge-Nelson Decomposition
Beveridge and Nelson (1980) proposed a definition of the permanent component of
an I(1) time series yt with drift µ as the limiting forecast as horizon goes to infinity,
adjusted for the mean rate of growth over the forecast horizon,

T Dt + BNt = lim yt+h|t − δh (9)


h→∞

The stochastic part of the permanent component (11), BNt , is referred to as the BN
trend. The implied cycle at time t is then

Ct = yt − T Dt − BNt

Beveridge and Nelson showed that if ∆yt has a Wold representation

∆yt = δ + ψ∗ (L)εt (10)

then BNt follows a pure random walk without drift

BNt = BNt−1 + ψ∗ (1)εt (11)


Xt
= BN0 + ψ∗ (1) εt
j=1

The derivation of the BN trend (11) relies on the following algebraic result.
P∞
Proposition 3 Let ψ(L) = k=0 ψk Lk with ψ0 = 1. Then

e
ψ(L) = ψ(1) + (1 − L)ψ(L), (12)
X∞
ψ(1) = ψk ,
k=0
X
∞ X

e
ψ(L) = ψ ej = −
e j Lj , ψ ψk .
j=0 k=j+1

P P∞ e
In addition, if ∞ k=0 k|ψ k | < ∞ (1-summability) then k=0 |ψ k | < ∞. 1-summability
is satisfied by all covariance stationary ARMA(p,q) processes. For an algebraic proof,
see Hamilton (1993) pages 534 and 535.
Now, consider the Wold representation for ∆yt given in (10). Then by recursive
substitution
X
t

yt = y0 + δt + ψ (L) εj
j=1

5
Applying (12) to ψ∗ (L)
³ ∗
´X
t
yt ∗ e
= y0 + δt + ψ (1) + (1 − L)ψ (L) εj
j=1

X
t
= y0 + δt + ψ∗ (1) εj + e
εt − e
ε0
j=1
= T Dt + T St + Ct

where e e ∗ (L)εt , T Dt = y0 + δt, T St = Z0 + ψ∗ (1) Pt εj and Ct = e


εt = ψ εt − e
ε0 . To
j=1
∗ Pt
show that ψ (1) j=1 εj is the BN trend defined by (9), consider the series at time
t+h
X
t+h

yt+h = y0 + δ(t + h) + ψ (1) εj + e
εt+h
j=1

The forecast of yt+h at time t is

X
t

yt+h|t = y0 + δ(t + h) + ψ (1) εj + e
εt+h|t
j=1

The limiting forecast as horizon goes to infinity, adjusted for mean growth, is

X
t

lim yt+h|t − δh = y0 + δt + ψ (1) εj + lim e
εt+h|t
h→∞ h→∞
j=1

X
t

= y0 + δt + ψ (1) εj
j=1
= T Dt + BNt

as limh→∞ e
εt+h|t = 0 since e
εt+h is a mean-zero stationary process. We have just proved
∗ Pt
that ψ (1) j=1 εj is the BN trend, and that the BN trend follows a pure random
walk.

Example 4 BN decomposition from MA(1) process for ∆yt

BN decomposition of U.S. real GDP (Stock and Watson (1987)). Let yt =


ln(rgdpt ). Using postwar quarterly data from 1947:II-1985:IV, Stock and Watson
fit the following MA(1) model to the growth rate of real gdp:

∆yt = 0.008 + εt + 0.3εt−1 , εt ∼ iid(0, σ 2 ), σ


b = 0.0106

6
For the MA(1) model, the Wold representation for ∆yt has the simple form ∆yt =
δ + ψ∗ (L)εt where ψ∗ (L) = 1 + ψ∗1 L and ψ∗1 = 0.03. Straightforward calculations give
ψ∗ (1) = 1 + ψ∗1 = 1.03
∗ X∞
e
ψ0 = − ψ∗j = −ψ∗1 = −0.03
j=1
X∞
e∗ = −
ψ ψ∗j = 0, j = 1, 2 . . .
j
j=k+1

so that the trend-cycle decomposition of yt using the BN decomposition becomes


X
t
yt = (y0 + δt) + ψ∗ (1) εj + e
εt
j=1

X
t
= y0 + 0.008t + 1.3 εj − 0.3εt
j=1

so that
T Dt = y0 + 0.008t
X t
BNt = 1.3 εj
j=1
Ct = −0.3εt
∂yt ∂yt+s
Note that ∂εt
= 1 and ∂εt
= 1.3 for s > 0.

2.1.1 Computing the BN Decomposition from an Estimated ARMA for


∆yt
The naive computation of the BN decomposition requires the following steps
1. Estimation of ARMA(p,q) model for ∆yt
2. Estimation of ψ∗ (1) from estimated ARMA(p,q) model for ∆yt
P
3. Estimation of tj=1 εj using residuals from estimated ARMA(p,q) model for
∆yt
Example 5 BN decomposition from ARMA(2,2) model for ∆yt
Morley, Nelson and Zivot (2003) fit the following ARMA(2,2) model to the growth
rate of postwar quarterly real GDP over the period 1947:I - 1998:II
∆yt = 0.816 + 1.342∆yt−1 − 0.706∆yt−2 + ε̂t − 1.054ε̂t−1 + 0.519ε̂t−2
φ(L) = 1 − 1.342L + 0.706L2 , θ(L) = 1 − 1.054L + 0.519L2

7
To compute an estimate of ψ∗ (1) from the ARMA(p,q) model, solve for the wold
representation

φ(L)∆yt = θ(L)εt
⇒ ∆yt = φ(L)−1 θ(L)εt = ψ∗ (L)εt

where ψ∗ (L) = φ(L)−1 θ(L). Therefore,

ψ∗ (1) = φ(1)−1 θ(1) (13)

Using (13), the estimate of ψ∗ (1) from the ARMA(2,2) model is


1 − 1.054 + 0.519
ψ∗ (1) = = 1.276
1 − 1.342 + 0.706
The estimate of the permanent component is then

X
t
T Dt + BNt = y0 + 0.816t + 1.276 ε̂j
j=1

The previous example shows that the computation of the BN decomposition from
an estimated ARMA(p,q) model is straightforward but somewhat tedious. The fol-
lowing example shows that the computation of the BN decomposition from an AR(1)
model is simple and elegant.

Example 6 BN decomposition from AR(1) process for ∆yt

From the h−step ahead forecast for yt given in (6), it is easy to compute the BN
trend for yt :
¡ ¢
T Dt + BNt = lim yt+h|t − hδ
h→∞
X
h
= yt + (∆yt − δ) lim φh
h→∞
s=1
φ
= yt + (∆yt − δ)
1−φ
The cycle component is then

Ct = yt − T Dt − BNt
φ
= (∆yt − µ)
1−φ

Notice that the cycle inherits the behavior of the AR(1) model for ∆yt .

8
Morley (2002) shows how the BN decomposition for an AR(1) model for ∆yt may
be extended to any model for ∆yt that can be represented in state space form. In
particular, suppose ∆yt − µ is a linear combination of the elements of the m × 1 state
vector αt : £ ¤
∆yt − µ = z1 z2 · · · zm αt
where zi (i = 1, . . . , m) is the weight of the ith element of αt in determining ∆yt − µ.
Suppose further that

αt = Tαt−1 + η t , η t ∼ iid N(0, Q),

such that all of the eigenvalues of T have modulus less than unity, and T is invertible.
Then, Morley shows that
£ ¤
T Dt + BNt = yt + z1 z2 · · · zm T(Im −T)−1 at|t (14)
£ ¤
Ct = yt − T Dt − BNt = − z1 z2 · · · zm T(Im −T)−1 at|t

where at|t denotes the filtered estimate of αt from the Kalman filter recursions.

2.2 The Orthogonal Unobserved Components Model


The basic idea behind the unobserved components (UC) model is to give structural
equations for the components on the trend-cycle decomposition (1). For example,
Watson (1986) considers UC-ARIMA models of the form

yt = µt + Ct (15)
µt = α + µt−1 + εt , εt ∼ iid(0, σ 2ε )
φ(L)Ct = θ(L)η t , η t ∼ iid(0, σ 2η )

where φ(L) = 1 − φ1 L − · · · − φp Lp and θ(L) = 1 + θ1 L + · · · + θq Lq . Note that


the trend component, µt , is a random walk with drift and the cyclical component,
Ct , is an ARMA(p,q) process. As it stands, however, the parameters of the UC
model are not identified without further restrictions. Restrictions commonly used in
practice to identify all of the parameters are: (1) the roots of φ(z) = 0 are outside
the unit circle; (2) θ(L) = 1, and (3) cov(εt , η t ) = 0. These restrictions identify
Ct as a transitory autoregressive “cyclical” component, and µt as the permanent
trend component. The restriction cov(εt , η t ) = 0 states that shocks to Ct and µt are
uncorrelated. As shown in Morley, Nelson and Zivot (2003), for certain models the
assumption that cov(εt , η t ) = 0 turns out to be an over-identifying restriction.

Example 7 Clark’s (1986) JPE Model

9
Clark considered the UC-ARIMA(2,0) model

yt = µt + Ct (16)
µt = α + µt−1 + εt , εt ∼ iid(0, σ 2ε )
Ct = φ1 Ct−1 + φ2 Ct−2 + η t , η t ∼ iid(0, σ 2η )
cov(εt , η t ) = 0

The model may be put in state space form in several way. For example, one may
define the transition equation for the state vector αt = (µt , Ct , Ct−1 )0 as
        
µt 1 0 0 µt−1 α 1 0 µ ¶
 Ct  =  0 φ1 φ2   Ct−1  +  0  +  0 1  εt
ηt
Ct−1 0 1 0 Ct−2 0 0 0

and the measurement as  


µt
yt = (1, 1, 0)  Ct 
Ct−1
µ 2 ¶
σε 0
Qt =
0 σ 2η
Notice that the state vector contains the I(1) component α1,t = µt and the covariance
stationary component α02,t = (Ct , Ct−1 )0 . As a result, the distribution of the initial
state vector α0 = (µ0 , C0 , C−1 )0 cannot be determined in the usual way. This is
because, a1,0 does not have a simple stationary unconditional distribution. Since
var(µt ) = µ0 + σ 2ε · t, the usual approach is to assume

a1,0 = 0
var(α1,0 ) = κ · 106

The distribution for α2,0 may be based on its unconditional stationary distribution.
Therefore, the initial state distribution is characterized by

a0 = 0
µ ¶
κ · 106 0
P0 =
0 P2

where P2 is the unconditional variance of α2,0 .


An alternative state space representation that only treats the unobserved cycle as
a state variable is based on the transformed model

∆yt = ∆µt + ∆Ct


= α + ∆Ct + εt

10
Defining the state variables as αt = (Ct , Ct−1 )0 , the transition equation is
µ ¶ µ ¶µ ¶ µ ¶
Ct φ1 φ2 Ct−1 ηt
= +
Ct−1 1 0 Ct−2 0
The corresponding measurement equation is
∆yt = ( 1 −1 )αt + εt
Since αt is covariance stationary, the distribution of the initial state vector α0 may
be determined in the usual way.

Proposition 8 Any UC-ARIMA model of the form (15) with a specified correlation
between εt and η t is observationally equivalent to an ARMA model for ∆yt with non-
linear restrictions on the parameters. The restricted ARMA model for ∆yt is called
the reduced form of the UC-ARIMA model.

Proposition 9 (Lippi and Reichlin (1992) JME) ψ∗ (1) computed from the reduced
form ARMA model for ∆yt based on a UC-ARIMA model for yt is always less than
1.

Example 10 Random walk plus noise model

Consider the random walk plus noise model


yt = µt + η t , η t ∼ iid(0, σ 2η )
µt = µt−1 + εt εt ∼ iid(0, σ 2ε )
σ2ε
where η t and εt are independent. Define the signal-to-noise ratio q = σ2η
. The reduced
form ARMA model for ∆yt is then
∆yt = ∆µt + ∆η t = εt + η t − η t−1 .
It is straightforward to show that ∆yt follows an ARMA(0,1) process. To see this,
consider the autocovariances of ∆yt
γ ∗0 = var(∆yt ) = var(εt + η t − η t−1 ) = σ 2ε + 2σ 2η
= σ 2η (q + 2)
γ ∗1 = cov(∆yt , ∆yt−1 ) = E[(εt + η t − η t−1 )(εt−1 + η t−1 − η t−2 )]
= −σ 2η
γ ∗j = 0, j > 1
Clearly, the autocovariances for ∆yt are the same as the autocovariances for an
ARMA(0,1) model, and so ∆yt has the representation
∆yt = ς t + θς t−1 , ς t ∼ iid(0, σ 2ς )

11
where the ARMA(0,1) parameters θ and σ 2ς are nonlinearly related to the UC-ARIMA
parameters σ 2η and σ 2ε , and the error term ς t encompasses the structural shocks η t
and εt .
Notice that the reduced form ARMA(0,1) model has two parameters θ and σ 2ς ,
and the structural random walk plus noise model also has two parameters σ 2ε and σ 2η .
Since the reduced form and structural models have the same number of parameters the
order condition for identification is satisfied. If εt and η t were allowed to be correlated
then there would be three structural parameters (two variances and a covariance) and
only two reduced form parameters, and the order condition for identification would
not be satisfied. Hence, setting cov(εt , η t ) = 0 is an identifying restriction in this
model.
The mapping from UC-ARIMA parameters to the reduced form parameters is de-
termined as follows. The first order autocorrelation for the reduced form ARMA(0,1)
is
θ
ρ∗1 =
1 + θ2
and for the UC-ARIMA it is
γ∗ −1
ρ1 = 1∗ =
γ0 q+2
Setting ρ∗1 = ρ1 and solving for θ gives
p
−(q + 2) ± (q + 2)2 − 4
θ=
2
The invertible solution is
p
−(q + 2) + q2 + 4q
θ= , θ < 0.
2
Notice that the MA coefficient θ for the reduced form ARMA(0,1) model is restricted
to be negative. If q = 0, then θ = −1. Similarly, matching variances for the two
models gives
σ 2η
σ 2ς = .
−θ
Finally, note that since θ < 0 it follows that ψ∗ (1) = 1 + θ < 1.

3 Cochrane’s Variance Ratio Statistic


Cochrane (1988) considered the question of the relative importance of permanent
shocks to temporary shocks in the analysis of U.S. real GDP. Specifically, he was
interested in determining the fraction of the quarterly variation in log real GDP that
is attributable to permanent shocks. To answer this question, Cochrane proposed
the use of variance ratio statistics computed as the normalized ratio of the estimated

12
variance of the kth difference of GDP to the variance of the first difference. to be more
specific, let yt denote the log-level of real GDP, µ = E[∆yt ] and define the variance
of the kth difference as
1
Vk = var(yt+k − yt − kµ)
k
Cochrane’s variance ratio statistic is then defined as
Vk
Rk = .
V1
Example 11 Pure random walk model

Consider the pure random walk

yt = yt−1 + εt , εt ∼ iid(0, σ 2ε ).
P P Pt+k
Then yt = y0 + tj=1 εj , yt+k = y0 + t+k j=1 εt and yt+k − yt = j=t+1 εj so that
Vk = V1 = σ 2ε and Rk = 1 for all values of k.

Example 12 White noise process

Let yt = εt ∼ iid(0, σ 2ε ) be a white noise process. Then var(yt+k −yt ) = var(yt+1 −


yt ) = 2σ 2e and so Rk = k1 . Notice that limk→∞ Rk = 0.
The examples above indicate that the variance ratio statistic computed from a
difference stationary process is non-zero for any value of k whereas it converges to
zero as k gets large for a trend stationary process.
The variance ratio statistic can be rewritten in several different ways. First, note
that
1 £ ¤
Vk = E ((∆yt+1 − µ) + · · · + (∆yt+k − µ))2
k
X
k−1
k−j ∗ ∗
= γ ∗0 + 2 γ j , γ j = cov(∆yt , ∆yt−j )
j=1
k

which is a weighted average of autocovariances of ∆yt . Furthermore, it can be shown


that
X
∞ X

lim Vk = γ ∗j = γ ∗0 +2 γ ∗j ≡ V,
k→∞
j=−∞ j=1
V
lim Rk = ≡ R.
k→∞ γ ∗0

To interpret the limiting form of the variance ratio statistic we require the following
results.

13
P∞
Proposition 13 Let ∆yt = ψ∗ (L)εt , εt ∼ iid(0, σ 2ε ) and j=0 j|ψ∗j | < ∞. Then as
T →∞ √ d
T ∆yt → N(0, σ 3ε ψ∗ (1)2 )

where σ 3ε ψ∗ (1)2 is the asymptotic or long-run variance of T ∆yt .

Proposition 14 σ 3ε ψ∗ (1)2 = var(∆T St ) where T St is computed from the BN decom-


position of yt .
Pt
From the BN decomposition, T St = ψ∗ (1) j=1 εj so that ∆T St = ψ∗ (1)εt and
hence var(∆T St ) = σ 3ε ψ∗ (1)2 .
P∞
Proposition 15 σ 3ε ψ∗ (1)2 = j=−∞ γ ∗j

Combining the results in Propositions, the limiting form of the variance ratio
statistic, R, may be re-expressed as

σ 3ε ψ∗ (1)2
R =
γ ∗0
var(∆T St )
=
var(∆yt )

The expression above shows that R may be interpreted as the ratio of the variance
of the stochastic trend or permanent shock to the variance of the total change in yt .
Remarks

1. The variance ratio statistics are usually computed for various values of k and
reported in graphical form.

2. If R = 0 the series yt is trend stationary or I(0); if R > 0 then yt is difference


stationary or I(1); if R < 1 then yt is called trend reverting and if R > 1 it is
called trend averting.

3. R can be computed nonparametrically using the sample autocovariances of ∆yt .


Alternatively, parametric estimates of R can be computed from the either the
estimated BN decomposition of ∆yt or the UC-ARIMA model for yt .

4. Estimates of R based on the BN decomposition are always greater than esti-


mates based on UC-ARIMA models because for UC-ARIMA models ψ∗ (1) is
always less than 1.

5. Campbell and Mankiw (1986) derive the following relationship between R and
ψ∗ (1). Define
σ2 σ2
ρ2 = 1 − 2ε = 1 − ∗ε
σ ∆y γ0

14
which is a measure of the predictability of ∆yt . Then it is easy to show that
s
R
ψ∗ (1) =
1 − ρ2

and so R is a lower bound for ψ∗ (1).

6. There are problems with nonparametric estimates of R. First, R cannot be


computed exactly with a finite amount of data. As a practical matter, one must
choose a value of k. Choosing k too small may obscure the presence of trend
reversion that is picked up in higher order values of ρ∗j . Alternatively, choosing
k too large may produce excess or spurious trend reversion. The reason for this
result is that as k → T, Rk → 0. In other words, Rk is a downward biased
estimate of R.

4 References
1. Clark, P. K. (1987). “The Cyclical Component of U.S. Economic Activity,” The
Quarterly Journal of Economics, 102, 797-814.

2. Cochrane, J. H. (1988). “How Big is the Random Walk in GNP?” Journal of


Political Economy, 96, 797-814.

3. Lippi, M. and L. Reichlin (1992). “On Persistence of Shocks to Economic


Variables. A Common Misconception,” Journal of Monetary Economics, 29,
87-93.

4. Morley, J. (2002). A State Space Approach to Calculating the Beveridge-Nelson


Decomposition,” Economics Letters, 1, 123-127.

5. Morley, J., C.R. Nelson, and E. Zivot (2003). “Why are Beveridge-Nelson and
Unobserved-Components Decompositions of GDP so Different?”, The Review
of Economics and Statistics, 85(2), 235-243.

6. Stock, J.H. and M.W. Watson (1988). “Variable Trends in Economic Time
Series,” Journal of Economics Perspectives, 2, 147-174.

7. Watson, M.W. (1986). “Univariate Detrending Methods with Stochastic Trends,”


Journal of Monetary Economics, 18, 49-75.

15

You might also like