Transcript Document

Quality adjustment procedures in HICP
Prepared by
Oskars Alksnis, Central Statistical Bureau of Latvia
EU Twinning Project
Forwarding Armenian Statistics Through Twinning
AM09/ENP-PCA/TP/04
Component E: Harmonized Consumer Price Index
Activity E.3: Development of technical methodologies
Yerevan 5-9 September 2011
1
20
08
_
20 01
08
_
20 03
08
_
20 05
08
_
20 07
08
_
20 09
08
_
20 11
09
_
20 01
09
_
20 03
09
_
20 05
09
_
20 07
09
_
20 09
09
_
20 11
10
_
20 01
10
_
20 03
10
_
20 05
10
_
20 07
10
_
20 09
10
_
20 11
11
_
20 01
11
_
20 03
11
_
20 05
11
_0
7
P rice in d ex
100
90
80
70
60
50
150
75
100
P rice, L V L
Price and Quality changes
Av e rag e p rice s an d p rice in d e x (2007 X II = 100) fo r d ig ital came ras
250
95
200
85
index
pric es
65
50
55
0
2
Quality adjustment – EU legislation and
recommendations
• Legislation
•COMMISSION REGULATION (EC) No 1334/2007 of 14
November 2007 amending Regulation (EC) No 1749/96 on
initial implementing measures for Council regulation (EC)
No 2494/95 concerning harmonised indices of consumer
prices
 Article 5: Minimum standards for replacements and
quality adjustment
• CENEX handbook
•http://www.destatis.de/jetspeed/portal/cms/Sites/destatis/Internet/DE/Content/P
ublikationen/StatistikundWissenschaft/Band13__Handbook1030813099004,propert
y=file.pdf
3
Quality adjustment methods
• Most commonly used (known) methods for QA:
 direct comparison
 overlap
 bridged overlap
 option cost
 (supported) judgmental adjustment
 monthly chaining and resampling
 hedonic methods
 link to show no changes
4
Quality adjustment – recommended methods
• Methods adapted or recommended at EU level (A or B methods out
of the ABC ranking):
 clothing and footwear – hedonic, direct comparison, bridged
overlap or class mean imputation
 books
• long-selling market – direct comparison, bridged overlap
• rapidly changing market – direct comparison
• cars
 new cars – option cost or bridged overlap
 used cars – supported judgmental adjustment or bridged
overlap
• TVs, washing machines – direct comparison, bridged overlap
5
Quality adjustment – recommended
methods (2)
cont.
• computers
 desktops – option cost, bridged overlap
 notebooks - bridged overlap
• software
 applications and games – direct comparison
• other durables – direct comparison or bridged overlap,
• hedonic regression as alternative to above mentioned methods
• other methods: overlap, quantity adjustment, judgmental
adjustment, monthly resampling and chaining
6
Used cars in Latvian HICP
• Two groups of used cars in the HICP basket:
 3-5 year old cars
 9-11 year old cars
• Information on the weights from National Accounts,
adjusted for latest changes in the market development
• Detailed information on the breakdown by car models
from Road Traffic Safety Directorate (CSDD)
• Major part of used cars come from abroad, yielding
significant consumption weight
7
Used cars in Latvian HICP (2)
W e ig h ts o f th e c a r s in th e H IC P b a s k e t, %
4.5
4
3.5
3
Us ed
ca rs
2.5
2
New
1.5
ca rs
1
0.5
0
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
8
Used cars in Latvian HICP (3)
F irs t tim e re g is te re d c a rs in L a tv ia 2 0 1 0 (C S D D d a ta )
2000
1800
1600
1200
1000
800
600
400
200
Y e a r o f p r o d u c tio n / f ir s t p u r c h a s e
20
09
20
07
20
05
20
03
20
01
19
99
19
97
19
95
19
93
19
91
19
89
19
87
0
19
85
C ount
1400
Used cars -> 9-11 year old cars
• 9-11 year old cars
 data source – advertisements from different websites
 offer prices
 many with many price comparison
 sample structure (consumption segments)
• medium size cars (Ford Focus, VW Golf, Opel Astra, etc.)
• large size cars (Opel Vectra, BMW 5 series, Mazda 626,
etc.)
• minivens (Opel Zafira, VW Sharan)
Used cars -> 9-11 year old cars (2)
Used cars -> 9-11 year old cars (3)
• Comparability, Replacements and Quality adjustment
 no reliable technical information is available
 adjustment for the depreciation or age is carried out
 additional stratification by type and capacity of engine
10 y ears old c ar
10 y ears old c ar
date of produc tion
date of s ec ond
and/or f irs t
purc has e
purc has e
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
Used cars -> 9-11 year old cars (4)
• Comparability, Replacements and Quality adjustment
 Example: Volkswagen Golf 10 and 11 year old cars
(registration years 2000 and 2001)
 Year is divided in 12 parts (months)
• In February 2011 price for car 2000 registration year is 2100 LVL
• In February 2011 price for car 2001 registration year is 2300 LVL
• average price is adjusted for age as follows:
• 2100 x 10 /12 + 2300 x 2 /12 = 2133
• In March 2011 price for car 2000 registration year is 2080 LVL
• In March 2011 price for car 2001 registration year is 2270 LVL
• average price is adjusted for age as follows:
• 2080 x 9 /12 + 2270 x 3 /12 = 2128
Used cars -> 3-5 year old cars
• 3-5 year old cars
 data source – car sellers / dealers (websites and
outlets)
 sample structure (consumption segments)
• medium size cars (Mazda 3, VW Golf 5)
• large size cars (Volvo V70, VW Passat, Toyota Avensis, Audi
A4)
• sport utility cars (Toyota RAV4, BMW X5)
• minivens (Oper Zafira, VW Touran)
14
Used cars -> 3-5 year old cars (2)
• Comparability, Replacements and Quality adjustment
 Usual price collection and comparison like with like
 If the selected product is not available, replacement is
chosen within the framework of defined specification
 better technical information normally is available
 adjustments for the age and mileage is carried out
 CENEX recommendations serve as basis for sampling,
replacement strategy and quality adjustment
15
Used cars -> 3-5 year old cars (3)
• Changes in the age or in the mileage of the car should
be adjusted by supported expert judgement.
• For each considered primary model in each age class
one depreciation rate for age and one for mileage has to
be calculated on a yearly basis
• Approximately 10 price observations (5 pairs) is
necessary for each depreciation rate
• Important! When calculate depreciation rate for age,
the mileage of the observation pairs should be very
similar (nearly constant) whereas the age in months can
vary within the interval of the considered age class and
vice versa
16
Used cars -> 3-5 year old cars (4)
• “delta” can be interpreted to be the absolute monetary value of 1
month of age for the considered primary model in the considered age
class
• The absolute monetary value of an additional mileage of 1000
kilometres for the considered primary model in the considered age
class
17
Used cars -> 3-5 year old cars (5)
• Limitations in Latvian case:
- some observations, which were collected for the
age sample had also be applied for the mileage
sample (and the other way round) - if they were
appropriate;
- restrictions to the sub-models had been relaxed:
observations came from the same primary models but
could be of different sub-models;
- limitation to the age class was removed: one
depreciation rate for age or mileage was calculated for
all age classes of one primary model.
18
Used cars -> 3-5 year old cars (6)
s u b -m o d e l
a g e (m o n th s )
m ile a g e (th k m )
p ric e (L s )
1 .6
3
52
6 88 7
1 .6
10
53
6 46 6
1 .6
6
1 22
6 15 0
1 .6
29
1 02
4 42 8
1 .9
19
1 83
6 22 0
1 .9
34
1 82
4 77 9
1 .9
19
1 91
6 00 0
1 .9
24
1 90
5 57 0
1 6887  6466
6150  4428
6220  4779
6000  5570
  (



)  79 . 27 LVL
4
10  3
29  6
34  19
24  19
19
Used cars -> 3-5 year old cars (7)
s u b -m o d e l
a g e (m o n th s )
m ile a g e (th k m )
p ric e (L s )
1 .6
4
77
8363
1 .6
6
122
6150
1 .6
8
29
8082
1 .6
7
125
5412
1 .6
28
50
5615
1 .6
29
102
4428
1 .9
19
183
6220
1 .9
19
191
6000
1 .9
24
94
6100
1 .9
23
100
5997
1 .9
29
95
6501
1 .9
34
182
4779
1 8363 6150 8082  5412 5615  4428 6220  6000 6100  5997 6501 4779





)  27.38 (LVL)
6
122  77
125  29
102  50
191183
100  94
182  95
 (
20
Used cars -> 3-5 year old cars (8)
• Example
Previously observed model in 3-year old age class (VW
Golf 5 Trendline 1.6l, diesel engine, price 8625 LVL
(Latvian lats), mileage 27 900km, first registration
11.2007 = age 28 months) disappeared from the market
and new similar representative model was chosen with
price of 6255 LVL but with mileage 39 200 km and first
registration 01.2007 = age 38 months.
The observed price in March 2010 had to be adjusted for
10 months and 11 300 kilometres:
21
Used cars -> 3-5 year old cars (9)
• Example
• Adjusted replacement price = 6255 + 79.27 x (38 –
28) + 27.38 x (39.2 – 27.9) =
= 6255 + 1102 = 7357 (LVL)
• Quality adjusted price change: 7357 / 8625 x 100 =
85.3 or -14.7%
22
Computers in Latvian HICP
• Information processing equipment in the HICP basket since
1999
• in 1999 two items, in 2011 seven items in the HICP basket
• weights (in graph)
W e ig h t s o f t h e IT e q u ip m e n t in t h e H IC P b a s k e t , %
0 .8
0 .7
0 .6
0 .5
0 .4
0 .3
0 .2
0 .1
0
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
23
Computers in Latvian HICP (2)
• Most important items: desktop computers and notebooks
A v e r a g e p r ic e s f o r c o mp u te r s ( L V L )
800
700
600
500
De s kto p s
400
No te b o o k
300
s
200
100
2011_07
2011_05
2011_03
2011_01
2010_11
2010_09
2010_07
2010_05
2010_03
2010_01
2009_11
2009_09
2009_07
2009_05
2009_03
2009_01
2008_11
2008_09
2008_07
2008_05
2008_03
2008_01
0
Pr ic e d e v e lo p me n t f o r c o mp u te r s ( De c .2 0 0 7 =1 0 0 )
110
100
90
De s kto p s
80
No te b o o k
70
s
60
2011_07
2011_05
2011_03
2011_01
2010_11
2010_09
2010_07
2010_05
2010_03
2010_01
2009_11
2009_09
2009_07
2009_05
2009_03
2009_01
2008_11
2008_09
2008_07
2008_05
2008_03
2008_01
50
24
Computers in Latvian HICP (3)
• Quality adjustment methods proposed by CENEX
research:
 bridged overlap for notebooks and
 option pricing for desktops.
• Alternative approach: development of the hedonic
model and applying hedonic re-pricing model both for
desktops and notebooks.
• If replaced and replacement models essentially
equivalent: direct comparison.
25
Desktops
• Option pricing or hedonic
• Latvian approach – option pricing (in general: follow the
CENEX recommendations)
• Recommended characteristics for observation:
 Processor type (e.g. Celeron, Pentium IV, . . .).
 Processor speed (in MHz).
 Hard disk capacity (in GB).
 Working memory (in MB/GB).
 Operating system (Windows XP, Linux, none, . . .).
 Other software (MS Office, . . .).
 Type of monitor, including size (TFT or not, . . .).
 Type of drive (CD-rewriter, DVD player/rewriter, . . .).
26
Desktops (2)
• Adjustment procedure:
Observed price of the replacement
+
Option prices of the characteristics in which the replaced model
differs from the replacement
–
Option prices of these characteristics of the replacement
=
Quality adjusted price of the replacement
27
Desktops (3)
• Example
• old (replaced) model 400 LVL
• new (replacement) model 450 LVL
• both desktops essentially equivalent except for their hard
disk. The old model had 100 GB capacity, while new model
200 GB. When bought separately in the current month, a 200
GB hard disk would cost 100 LVL, whereas a 100 GB hard
disk would cost 40 LVL
• adjusted price for replacement model : 450 + 40 - 100 =
390 LVL
• Price change, adjusted for quality change: 390 / 400 = 2,5%
28
Desktops (4)
• Latvian approach
• the same idea, whereas different mathematical interpretation
• calculate the share of price difference between old and new
model, which is attributed to quality change (100 – 40) / (450
– 400) = 120%
• adjusted price for replaced (old) model: (450 * 400)/(450 ((450 - 400) * 120 / 100)) = 461,5 LVL
• Price change, adjusted for quality change: 450 / 461,5 = 2,5%
• Price changes without quality adjustment: 450 / 400 =
12,5%
29
Notebooks
• One of recommended methods: Bridged overlap
• A bridge is needed to reflect and apply the price change of
all unchanged (matched) notebooks to the new model.
30
Notebooks (2)
• Example
E le m e n ta ry
Ag g re g a te in d e x fo r
P ro d u c t1 (m o n th ly
B a s e p ric e
N o te b o o k s P ric e 1
P ric e 2
P ric e 3
P ric e 4
P ric e 5
R e p la c e m e n t6
Av e ra g e (AM )
A ve ra g e o f
c h a in in g v e rs io n :
(D e c e m b e r;t J a n u a ry
F e b ru a ry
M a rc h
0)
(t2 )
(t3 )
(t1 )
t5 /t4 *t4 /t3 *t3 /t2 *t2 /t1
Ap ril (t4 )
540
580
620
500
600
540
580
610
490
600
510
580
610
490
600
540
500
610
490
600
5 6 8 .0 0
5 6 4 .0 0
5 5 8 .0 0
5 4 8 .0 0
1 ;3 ;4 ;5 (A M )
Av e ra g e (AM )
M a y (t5 )
530
500
610
490
600
5 8 2 .6 1
5 4 6 .0 0
540
5 5 7 .5 0
5 6 2 .5 2
5 5 5 .0 0
5 6 0 .0 0
*t1 /t0 O R t5 /t4 *t4 /t0 )
610
490
580
580
9 5 .6 9 5 6 9 8 8 6
Im p u te d p re vio u s p e rio d 's p ric e
R e p la c e m e n t
fo r 6 ->
p ric e
5 8 0 / (5 6 0 / 5 6 2 .5 2 )
31
Thank you for your attention!
32