\
|
=
q q
q
q
I
I
A X
|
|
| |
,
,
.
q y
A
q q I | ,
JVI/QNA/L8 : 9
IMF Statistics
Department
Pro Rata Distribution and Step Problem
Benchmark-to-indicator X
q
is the level of the Benchmark to indicator
ratio presentation
X
q,
is the level of the
QNA estimate for
quarter q of year
I
q,
is the level of the
indicator in quarter
q of year
|
|
.
|
\
|
=
q
q q
I
A
I X
|
|
| | , ,
.
q y
A
q q I | ,
JVI/QNA/L8 : 10
IMF Statistics
Department
6
Pro Rata Distribution and Step Problem
Both equations are algebraically equivalent
Only the presentation differs
JVI/QNA/L8 : 11 IMF Statistics Department
Extrapolation with an Indicator
For the quarters of the current year or even the q y
most recent year, independent ANA data are
not available.
QNA estimates for these periods should be
consistent with the QNA estimates for previous
periods that are benchmarked to the annual
data.
M t i th t l i di t i d t Movements in the quarterly indicator is used to
extrapolate/update the QNA estimates
(benchmarked to the ANA) to derive the QNA
estimates for quarters with no ANA estimates.
JVI/QNA/L8 : 12 IMF Statistics Department
7
Extrapolation with an Indicator
Extrapolation presentation X
q
is the level of the Extrapolation presentation X
q,
is the level of the
QNA estimate for
quarter q of year
I
q,
is the level of the
indicator in quarter
|
|
.
|
\
|
=
+
+
I
I
X X
q
q
|
|
| |
, 4
1 ,
, 4 1 ,
.
indicator in quarter
q of year
. \
| , 4
JVI/QNA/L8 : 13
IMF Statistics
Department
Extrapolation with an Indicator
Benchmark-to-indicator X
q
is the level of the Benchmark to indicator
presentation
X
q,
is the level of the
QNA estimate for
quarter q of year
I
q,
is the level of the
indicator in quarter
|
|
.
|
\
|
=
+ +
I
X
I X q q
|
|
| |
, 4
, 4
1 , 1 ,
.
indicator in quarter
q of year
. \
| , 4
JVI/QNA/L8 : 14
IMF Statistics
Department
8
Pro Rata Distribution: Step Problem
98 0
100.0
102.0
104.0
106.0
108.0 -------------------------Back Series-------------------------
980
1000
1020
1040
1060
1080 -------Forward series-----
96.0
98.0
1997 1998 1999
960
980
Indicator (left hand scale)
QNA estimates derived using pro rata distribution (right hand scale)
JVI/QNA/L8 : 15 IMF Statistics Department
Pro Rata Distribution: Step Problem
Benchmark-to-indicator ratio Benchmark to indicator ratio
9.8
10.3
1997 1998 1999
JVI/QNA/L8 : 16 IMF Statistics Department
9
Golden rules of Benchmarking
A benchmarking procedure should satisfy two g p y
requirements at the same time:
Preserve as much as possible the short-term
movements in the quarterly source data under
the restrictions provided by the annual data
Ensure, for the forward series, that the sum of
the four quarters of the current year is as close
ibl t th k f t l d t as possible to the unknown future annual data
JVI/QNA/L8 : 17 IMF Statistics Department
Benchmarking Methods That Avoid Steps
Time series method avoiding steps Time series method avoiding steps
Various, but same purpose
Keeps the movements of the short-term benchmarked
series as proportional as possible to those in the
original series
JVI/QNA/L8 : 18 IMF Statistics Department
10
Benchmarking Methods That Avoid Steps
Proportional Denton method Proportional Denton method
Keeps the movements in the benchmarked QNA series
as proportional as possible to those in the quarterly
indicator series using a least square technique to
minimize the difference in relative adjustment to
neighboring quarters subject to annual total constraint.
JVI/QNA/L8 : 19 IMF Statistics Department
Optimal Benchmarking Solution
For the back series, the method ,
recommended in the IMF QNA manual is the
Proportional First Differences (PFD) solution
proposed by Denton (1971) which
Preserves the short-term movements in the
indicators keeping the quarterly estimates as
proportional to the indicator as possible
(avoid the step problem)
Considers the annual benchmarks as binding
constraints
JVI/QNA/L8 : 20 IMF Statistics Department
11
Denton PFD method
In mathematical terms, the benchmarked ,
series X
t
is the solution of the following
minimization problem:
{ } N t for
Y
X
Y
X
N
t t
4 ,..., 2 , 1 min
2
4
1
e
(
JVI/QNA/L8 : 21 IMF Statistics Department
{ } N T for A X t s
Y Y
T
T t
t
t
t t
,..., 2 , 1 . .
2
1
e =
e
=
Proportional Denton Method
Benchmarking without Step Problems Benchmarking without Step Problems
98.0
100.0
102.0
104.0
106.0
108.0
980
1000
1020
1040
1060
1080
------------------ Back series ------------------------------- ----Forward series ----
96.0
1997 1998 1999
960
Indicator (left hand scale)
QNA estimates derived using pro rata distribution (right hand scale)
1997-98 distributed 1999 extrapolated using Proportional Denton (right hand scale)
JVI/QNA/L8 : 22 IMF Statistics Department
12
Proportional Denton Method
Benchmark-to-Indicator Ratios
10.3
9.8
1997 1998 1999
1997-98 distributed 1999 extrapolated using Proportional Denton Annual step change
JVI/QNA/L8 : 23 IMF Statistics Department
Improving extrapolation of QNA results
Using proportional Denton method implies that Using proportional Denton method implies that
the B-I ratio for the fourth quarter of the last
benchmarked year is used to prepare the forward
QNA series
The B-I ratios for quarters with ANA data are
usually different and change smoothly y g y
Depends on the movements in the annual B-I ratios
JVI/QNA/L8 : 24 IMF Statistics Department
13
Improving extrapolation of QNA results
For the forward series, the Denton PFD ,
method results in quarter-to-quarter rates of
change that are identical to those in the
indicator series. But
When there is a bias between the movements of the
target variable and those of the indicator,
extrapolations should be adjusted to better predict
the next annual growth rate of the target variable
An adjustment (or enhancement) to the original
Denton method is proposed in the IMFs QNA manual
(chapter 6)
JVI/QNA/L8 : 25 IMF Statistics Department
Improving extrapolation of QNA results
The B-I ratios for quarters with no ANA data can The B I ratios for quarters with no ANA data can
be improved by
Forecasting annual B-I ratios
Deriving quarterly B-I ratios taking into account the
forecasted annual B-I ratio
JVI/QNA/L8 : 26 IMF Statistics Department
14
Denton PFD method with enhancement for
extrapolation
The minimization problem changes into:
{ }
{ }
4
2
1
2
1
4 min for 1, 2, , ,
t t
t
t t
N s
X X
t
I I
N s
=
+
+
(
e
(
p g
B-I forecast for year N+1
{ }
{ }
4
1
4
4 1
1
,
s.t.
for 1, 2, 3, 4
, for 1, 2, ,
t T
t T
N s
t N
t
t N
t N
X A T N
X A
w s
I I
e
+
+
= +
+
= e
= e