Developing a Hedonic Regression Model for VCRs in the U.S. CPI

William Thompson

Background

The Bureau of Labor Statistics (BLS) is continuing with its research effort to improve the estimation of quality change for selected items in the Consumer Price Index (CPI). Some experts believe that the CPI overstates the rate of inflation for products that are rapidly improving in quality. Unfortunately a measure of quality-adjusted price change faced by consumers cannot be directly observed in the marketplace. Hedonic regression analysis, however, offers one way to estimate price values for individual characteristics of an item. The estimated price values can then be used to differentiate between pure price change and price change due to quality differences between old and replacement items in the CPI sample. If we are successful in the modeling effort, we can account for a greater number of price changes realized by consumers even when new product introductions are marked by technological improvements.

The inclusion of the video cassette recorder (VCR) in this research effort is linked to the perception that there may be some quality bias in the price index for this item. Also, the success of earlier hedonic research for VCRs suggests that VCRs should be included in the continuing effort to develop hedonic regression models for the purpose of adjusting for quality change in the CPI index.

The VCR was introduced twenty-five years ago and has become a staple in American households. The market place for VCRs is populated by a relatively large number of brands and a vast number of models with numerous features and characteristics. The purpose of this research is to identify VCR features and characteristics that are valued by the consumer and determine how they contribute to the market price of the product. Using hedonic regression analysis, parameter estimates are developed that serve as an indication of the relative price value of specific features or characteristics in the overall price of the VCR. The VCR data consist of current CPI data, as well as special pricing observations to augment the CPI sample and add robustness to the research effort. The data were collected by CPI representatives, and information on the items was supplemented by secondary sources, including Internet sites, consumer magazines, and manufacturer advertisements.

The Modeling Process

The sample of VCR prices used in monthly CPI estimation was insufficient for regression modeling purposes. Based on current CPI sampling procedures, CPI statisticians designed a supplemental sample for hedonic modeling purposes only. The new sample added 296 observations in addition to the 243 observations from the CPI sample. The final sample included 217 specially collected observations and 243 observations from the CPI sample. The field data collectors were unable to collect data for 26 percent of the supplemental sample. Observations with inconsistent or incomplete data and for which no manufacturer model number could be found were dropped from the data set.

To collect data for VCRs CPI data collectors use data collection forms known as checklists, which outline the product characteristics judged to be most important. When the CPI representative completes the data collection form we get a description indicating all the pertinent characteristics and the price of the item. (See Attachment A for the RA031 Cluster 01A, Video Cassette Players/Recorders data collection form.) Collecting data for electronic goods is not a simple task. Manufacturers and retailers use confusing terminology and often use different names for the same feature. Consumers are left to ask sales personnel for assistance to decipher technical information. Some of these difficulties spill over into CPI data collection efforts. Where there were inconsistencies or incompleteness in the data collected for this research effort, the manufacturer model numbers obtained by CPI data collectors were matched to specifications provided by the manufacturer Internet sites.

Although most VCRs are indeed capable of recording as well playback, there are units on the market which are playback only. There were a total of thirteen player-only units in the sample. There was an unsuccessful attempt to include unit type (player or play/recorder) in the model; however the t-statistic did not come close to a reasonable level of acceptance. The mean price for play only units was $126.50 compared to a mean price of $195.26 for play/record units. These play-only observations were deleted from the final data set, as it appears that the pricing structure for player-only units is different than that of units that have recording ability.

Tape format is dominated by the Video Home System (VHS), VHS High Quality (VHS-HQ) or Super VHS (S-VHS); the latter offers the best among analog formats, with 50 percent more picture sharpness than standard VHS. The beta system, which was the early competitor to VHS, is used primarily for commercial applications. The 8mm players are relegated to use by those with 8mm camcorders that do not have converter with which to watch their 8mm camcorder tapes on VHS. Finally, Digital VHS (D-VHS) first introduced in 1998 and designed to be a partner with DirecTVSatellite System or the Dish Network, has not yet achieved mainstream acceptance.

As expected, the sample proved to be skewed: While VHS had observations split between VHS and S-VHS, there were no observations for beta tape, two observations for 8 mm and one observation for digital. The digital and 8 mm observations were deleted from the sample to allow focus on the more popular VHS and Super VHS formats. There seems to be no discernible difference between VHS and VHS-HQ. Manufacturer reporting of this information was inconsistent, and it became clear that consumers would have difficulty in discerning if the units they were purchasing were equipped with HQ. Instead VHS and VHS-HQ variables were combined, so by default, tape format was either VHS or Super VHS. Since VHS made up an overwhelming majority of the sample it was chosen as the base variable for this category.

Brand was dominated by four companies: Sony, Panasonic, JVC and RCA, which combined made up 61 percent of the sample. Of the other 21 brands named, no other brand made up more than five percent of the sample. Sony and JVC (which combined made up 31 percent of the sample) were included in the model along with Zenith (five percent) and Proscan ( two percent). Thus, 38 percent of the brands by volume of sales were explicitly included as variables in the model. By implication brands not included in the model were grouped into a miscellaneous "Other" category and designated as a variable. The direction and magnitude for each of the included specifications met a priori expectations.

There is a direct relationship between the number of heads a VCR has and the quality of the picture during playback, thus number of heads is a fundamental attribute of the VCR. Most of the observations consisted of units with four heads, approximately 93 percent of the sample. VCRs with two heads comprised approximately six percent and VCRs with six heads made less than one percent. Although only a small percentage of the sample had two or six heads, the corresponding variables proved to be statistically significant and met expectations in terms of magnitude.

Hi-fi stereo units made up 81 percent of the sample, while mono-sound comprised 16 percent and surround sound two percent of the sample. With stereo sound as the base variable in the category, the mono-sound variable displayed the expected direction and magnitude and proved to be significant. The results for the surround sound variable were not as anticipated and did not prove to be significant.

The variable dual tape well represents the feature of dual tape capability, which allows the consumer to record tapes in addition to television shows. This feature is associated exclusively with the GO Video brand. Despite only a few observations, this variable proved to be large in terms of magnitude and significance. The GO Video brand has invested heavily not only in the technology but also the legal concerns of exclusively manufacturing a "dual well" machine. Thus, the dual-well VCR is a high-end type of VCR.

Other features that were included as variables in the model were flying erase head which permits smoother editing of tapes, the capability to advance through commercials or omit commercials when recording, and "VCR plus " recording, which allows the consumer to program the VCR to record a particular show using a five digit code.

A disappointing result of the modeling process was that the lines of horizontal output, a measure of the quality of the picture produced, (the more lines the higher the quality), could not be included in the model. Initially it was anticipated that the lines of horizontal output would be fruitful; however, due to the technical nature of the information and the inconsistency in its availability there was a low rate of successful data capture. So while lines of horizontal output may be an important technical characteristic, it remains unclear if and how consumers value this characteristic, especially if they are unable to determine the lines of output for their unit. A couple of other characteristics that may have suffered similar plights are Quasi S-VHS playback, and multiple format conversion. Some common characteristics that did not prove significant include automatic rewind, automatic head cleaner, child lock, and automatic clock set. Remote and manufacturer warranty also did not prove to be significant.

Country of origin was a new collection initiative. Japan made up most of the sample at 42 percent, while 31 percent of the sample came from other countries in the Far East. The remaining sample came from Mexico (seven percent) and the U.S. A. (three percent), while in 17 percent of the observations the country of origin was not available. While country of origin was given considerable attention the modeling results offered no compelling reason to include country of origin as a variable.

Statistical results of the model are presented in the table below:

The R-squared of .7581 says that almost 76 percent of the variation in the dependent variable is explained by the independent variables. This result compares to the Liegey and Shepler hedonic model which had an R-squared value of 95 percent.(1) (See Attachment C.) Although the hedonic models are similar in terms of types of variables included in the model, the Liegey and Shepler model is clearly more encompassing in terms of features characteristics, brand name and type of player. Although a comparison between the two hedonic models may provide some insights and topics for debate, any comparison should tempered by an understanding of the differences in approach for these two modeling efforts.

The model developed in this research was based on consumer retail offer prices and characteristics collected by CPI representatives. The Liegey and Shepler research was based exclusively on data obtained from the Consumer Digest 1998 Annual Buying Guide. In addition to the fact that these data are less current that the data used here, there is a theoretical concern that the list prices such as those listed in the Consumer Digest do not reflect prices that the consumer faces in the marketplace.(2) Using list prices that do not equal consumer retail offer prices may give a higher R-squared value but provide for a less accurate hedonic model. If this concern is true then it would be difficult to compare a hedonic model based on consumer retail offer prices and a hedonic model based on list prices. Also complicating comparisons between the two approaches is that the Liegey and Shepler model is unable to account for the price effects of the type of business, the region of the country or size of a city.

Comparing Quality Adjusted and Published Indexes

Simulating an index with quality adjusted VCR substitutions will illustrate the impact of using the VCR regression model in calculating the CPI. The published index (without quality adjustments) is recreated with CPI historical data. This recreation is then compared to the simulated index that includes quality-adjusted substitutions. The simulation of the published and quality adjusted indexes covers the time period from May 1999 to December 1999. Index price changes for the U.S. city level, such as those examined in this study, were obtained by summing price changes over all (elementary) index areas using aggregation weights derived from the Consumer Expenditure Survey.

VCRs are included in the Other video equipment item stratum (RA031) along with camcorders, DVD players, satellite video products and other miscellaneous video products. VCRs have an estimated 46 percent of the weight within Other video equipment. During the time period from December 1997 to December 1999 the Other video equipment index decreased 26.5 percent. The average monthly decline was 1.1 percent. The Other video equipment index was redefined for the January 1998 CPI revision so a longer term comparison is not available. Prior to the revision, VCRs were included in the Video products other than televisions index. This index decreased 38.7 percent from December 1988 to December 1997 — an average monthly decline of 0.4 percent.

For the time period June 1999 to December 1999 Other video equipment had a total of 1589 quotes priced. Over this time period there were 224 substitutions for an average substitution rate of 14 percent. Of the 224 substitutions, 99 substitutes, or 44 percent, were deemed to be 'comparable', meaning that 56 percent of the substitutes were coded as non comparable and were included for index calculation using the 'class-mean imputation' method.(3)

For VCRs alone there were a total of 130 substitutions during the June to December time period. In the published index 51 percent of the VCR substitutions directly compared the price of the price of the new item to the price of the previous item. There was one observation coded as quality adjusted. The quality adjustment for this observation was not based on parameter estimates from a hedonic model; the quality adjusted price was determined by the analyst from information received from the representative. Finally, price changes for the remainder of the substitutions were imputed using the class-mean imputation method.

For the purpose of calculating the quality adjusted index all the substitutions were reevaluated. One of the benefits of using a hedonic model in evaluating substitutions is that the analyst has an opportunity to review price data and item characteristics with a statistical tool, thus enabling him/her to render judgments based on statistics rather than expert judgment alone. A direct result of this benefit was that the method of price change for 64 percent (up from 51 percent for the published index) of the substitutions used the direct comparison between the price of the new item and the price of the old item. Twenty-seven percent of the substitutions were deemed to have changes in quality that could be adjusted using the hedonic model results. Only nine percent of the substitutions continued to use the class mean imputation method. The quality adjusted index had an overall comparability ratio of 91 percent compared to 52 percent for the published index. The use of the hedonic model to reevaluate substitutions impacts the index in two ways: 1) via substitutions that are quality adjusted; and 2) through substitutions that with additional information provided by the model can be directly compared rather than use the class mean imputation method. From the table below the mean price change for directly comparable quotes went from a negative 3.72 percent to a negative 1.00-percent.

Graphic presentations of the results of the RA031 index simulation, with and without hedonic quality adjustments, are shown in Attachment B. The first graph shows a line graph of index levels for the published and quality adjusted indexes from May 1999 to December 1999. The second graph illustrates the differences for one-month changes between the published and quality adjusted indexes.

From May 1999 to December 1999 the quality adjusted index decreased 6.7 percent compared to a 7.7-percent decrease for the published index.(4) The differences, quality adjusted minus published, for the one-month changes ranged from a low -0.72 percentage point to a high of 1.33 percentage point. The average difference for the seven month time period was 0.15 percent. These results show that the CPI may have been overstating the rate of price decline for item category of Other video equipment. That is, if the CPI actually had used the regression model for VCRs, the index would have had a smaller decline than the CPI actually shows.

The results of this study are not surprising given that Liegey and Shepler had a similar although smaller difference (0.1 percent difference over a 12 month study) between the published index and the quality adjusted index. Additionally, Kokoski, Waehrer, and Rozaklis found similar results (1.7 percent difference over a 12 month study) for some audio products.(5) Just as the Liegey and Shepler research was based on data from non-CPI sources, the Kokoski, Waehrer and Rozaklis research was based on point-of-sale data purchased from NPD, a private firm specializing in the collection and marketing of such data for sale. NPD data more closely resembles CPI data than the Consumer Digest's list price data, but differs in that NPD data covers a significantly wider product range than CPI data. Although there are differences in the source of the data the index simulation results can not be totally discounted.

It is clear the VCR is no longer a new, high tech consumer good, since 93 million U.S households own at least one VCR, according to industry sources. As Liegey and Shepler point out, the issue for the VCR index is not to capture falling prices that a high-tech consumer good faces early in it's product cycle; rather, the VCR index faces the quality and price change issues as they relate to a mature product.(6)

It is unclear what the future holds for the VCR. It is likely that the VCR will continue to be popular in the near term. While someday in the future the VCR or at least the VHS format will give way to some type of digital format, recent VCR sales have hit record numbers. 1998 was a banner year for VCR sales, up eight percent from the previous year, and the second consecutive record breaking year.(7) If the DVD does develop the technology to record and continues its downward price trend then it is very likely that the DVD player will make the VCR obsolete at some point in the future.

The CPI will begin to use hedonic regression models for VCRs beginning in April 2000.

Given the technological evolution of the consumer electronic market, in which VCRs exist, it is likely that this model will not prove to be completely stable. CPI intends to test the stability of the model using new data in the near future.

Notes

ATTACHMENT A

BUREAU OF LABOR STATISTICS             U.S. DEPARTMENT OF LABOR
CONSUMER PRICE INDEX - ELI CHECKLIST______________________________________________________
 collection             outlet                       quote            arranging
 period: __ __ __ __    number: __ __ __ __ __ __ __ code: __ __ __     code:  __ __ __ __
 _________________________________________________________________________________________
 ELI No./                                                                     cluster
   title  RA031 OTHER VIDEO EQUIPMENT                                            code  01A
 item availability:    1-AVAILABLE    2-ELI NOT SOLD     3-INIT INCOMPLETE
 purpose of checklist: 1-INIT   2-INIT COMPL  3-SPEC CORR   4-SUB   5-REINIT   6-CHECK REV
 _________________________________________________________________________________________
           CURRENT PERIOD                    |      SALES TAX
                                             |
           price _ _ _ _ _ _ . _ _ _         |      included:        YES    NO
                                             |
           type of price:  REG   SALE        |
                                             |
                                            |____________________________________________
             |
 YEAR-ROUND  |  in-season:  JAN  FEB  MAR  APR  MAY  JUN  JUL  AUG  SEP  OCT  NOV  DEC
 ____________|____________________________________________________________________________
 respondent:                                   location:
 _________________________________________________________________________________________
 field message:
 _________________________________________________________________________________________
  CLUSTER 01A - VIDEO CASSETTE PLAYERS/RECORDERS
 
  TYPE                                        MANUFACTURER'S MODEL NUMBER
    A1  Player Only
    A2  Player/Recorder                        D99  ______________________________
 
  TAPE FORMAT                                 NUMBER OF VIDEO HEADS
    B1  VHS                                     E1  Two
    B2  VHS HQ                                  E2  Three
    B3  Super VHS (S-VHS)                       E3  Four
    B4  Beta                                    E4  Five
    B5  Extended Definition Beta (ED Beta)      E5  Six
    B6  8mm                                    E99  Other,
    B7  Hi 8mm
    B8  Digital VHS (D-VHS)                         ______________________________
    B9  Digital Video (DV)
   B99  Other,                                LINES OF HORIZONTAL OUTPUT
 
        ______________________________         F99  ______________________________
 
  BRAND                                       AUDIO CAPABILITIES
    C1  Emerson                                 G1  Surround Sound
    C2  Fisher                                  G2  Hi-Fi Stereo
    C3  GE                                      G3  Monaural
    C4  Hitachi                                H99  Other audio capability,
    C5  JVC
    C6  Mitsubishi                                  ______________________________
    C7  Panasonic                              I99  Other audio capability,
    C8  Philips/Magnavox
    C9  Proscan                                     ______________________________
   C10  RCA
   C11  Samsung
   C12  Sanyo
   C13  Sharp
   C14  Sony
   C15  Toshiba
   C16  Zenith
   C99  Other,
         ______________________________

RA031-01A - VIDEO CASSETTE PLAYERS/RECORDERS - CONTINUED
 
  GENERAL FEATURES                            OTHER FEATURES
    J1  Dual well deck, both record and        AG1  Front panel Audio/Video inputs
          playback                            AH99  Other,
    J2  Dual well deck, one record and
          one playback                              ______________________________
    K1  Quasi S-VHS Playback                  AI99  Other,
    L1  Multiple format conversion
    M1  Flying Erase Head                           ______________________________
    N1  Automatic Rewind                      AJ99  Other,
    P1  Automatic Head Cleaner
    Q1  Cable and/or satellite dish                 ______________________________
          channel control
   R99  Other general feature,                WARRANTY
                                               AK1  One year parts and labor
        ______________________________         AK2  Two years parts and labor
                                               AK3  Three years parts and labor
 PROGRAMMING FEATURES                         AK99  Warranty, other type,
    S1  On screen program guide
    T1  Commercial Skip                             ______________________________
    U1  Commercial Advance
    V1  Skip Search                           COUNTRY OF ORIGIN
    W1  Index Plus                             AL1  Not Available
    X1  VCR Plus + Recording                   AL2  United States
    Y1  Child Lock                             AL3  Japan
   AA1  Automatic Clock Set                   AL99  Other,
   AB1  Movie Advance
  AC99  Other programming feature,                  ______________________________
 
        ______________________________        OTHER PRICE FACTORS
 
 REMOTE CONTROL                               BA99  ______________________________
   AD1  Controls VCR only
   AD2  Controls same brand TV                BB99  ______________________________
   AD3  Controls other brand TVs
   AE1  Illuminated Remote                    BC99  ______________________________
  AF99  Other remote control feature,
                                             ** OTHER CLARIFYING INFORMATION
        ______________________________
                                              CA99  ______________________________
 
                                              CB99  ______________________________
 
                                              CC99  ______________________________
 
 
 
 
 
 
 
 
 
 
 
 
 
 _________________________________________________________________________________________
 BLS 3400B (Rev. February 1995)                                         RA031 page 3 of 16
                                                                        Revised April 1999

DEFINITION OF TERMS
 
 CLUSTER 01A-VIDEO CASSETTE PLAYERS/RECORDERS
 
 8mm - Eight millimeter is a type of video tape.
 
 AUTOMATIC CLOCK SET - The VCR reads instructions from a broadcast signal that
   allows the VCR to set its own clock.
 
 AUTOMATIC HEAD CLEANER - The machine is able to clean its own video heads.
 
 AUTOMATIC REWIND - The tape rewinds automatically after play is complete.
 
 BETA - Type of tape.  Rare in consumer applications.  Do not price a
   professional model.
 
 CABLE AND/OR SATELLITE DISH CHANNEL CONTROL - When the VCR is programmed to
   record over a cable or satellite system, it is able to set the cable or
   satellite system to the correct channel for recording.
 
 CHILD LOCK - Freezes control of VCR except for timer functions.
 
 COMMERCIAL ADVANCE - The VCR marks sections of commercials electronically and
   fast forwards through them automatically during playback.
 
 COMMERCIAL SKIP - The VCR omits commercials while recording.
 
 DIGITAL VHS (D-VHS) - The video is stored in digital format on the VHS tape
   rather than as an analog tape.
 
 DIGITAL VIDEO (DV) - A particular standard for storing digital video information
   used in VCRs and Camcorders.
 
 DUAL WELL DECK - There are two places in the chassis of the VCR (wells) to input
   tapes.  Some types may permit recording in either of the wells.
 
 EXTENDED DEFINITION BETA (ED BETA) - A system used in the Beta tape format which
   provides superior video quality versus "standard" definition Beta.
 
 FLYING HEAD ERASE - A special head mounted on the same spinning headwheel as
   the recording heads that allows the VCR to create edited tapes with smoother
   transitions than would be possible otherwise.
 
 FRONT PANEL AUDIO/VIDEO INPUT - There is a place on the front panel of the VCR
   where video input from another source such as a video camera or video game
   can be connected.
 
 HI-BAND 8mm (HI8 8mm) - A system used in eight millimeter tape format which
   provides superior video quality versus "standard" eight millimeter.
 
 HiFi STEREO - The VCR will decode and output two channels of audio information
   (stereo).
 
 INDEX PLUS - May create an on-screen listing of taped programs along with date
   recorded and running time.  Fast forwards to a selected program and plays it.
 

ATTACHMENT B

 

ATTACHMENT C

Liegey and Shepler VCR Model

Variable name	Parameter Estimate	T-statistic	Variable name	Parameter Estimate	T-statistic
Intercept	5.12849	179.013	Number of heads:
Type:			Two video heads	-0.10864	-3.449
VHS Player	-0.13886	-3.053	Four video heads	Base
VHS	Base		Six video heads	0.48777	6.279
Quasi SVHS	0.09705	1.877	Other:
SVHS	0.71605	16.697	Auto rewind	0.12569	5.895
Dual deck VCR	0.89027	17.211	Cable channel changer	0.12856	5.540
Multi format converter	1.38890	12.955	Child lock	0.03977	2.002
Brand:			Commercial advance	0.17805	5.889
Marantz	0.59280	7.445	Commercial skip	0.11848	4.748
Sony/Toshiba/Mitsubishi	0.22754	7.559	Flying head edit feature	0.27088	7.539
Proscan	0.15815	2.337	Automatic head cleaner	0.03369	1.503
Other brands not listed	Base		Hi-fi stereo	0.19133	9.333
Funai/Phillips/Symphonic	-0.17155	-5.839	Index plus	0.25869	2.356
Samsung	-0.23819	-5.757	Skip search	0.17100	6.988
			Starsightâ	0.56336	9.431
			VCR plus	0.12219	5.537
N = 176 R2 = 0.9581 Adjusted R2 = 0.9514 F statistic = 143.746

 

Last Modified Date: October 16, 2001

Recommend this page using:

Facebook

Twitter

LinkedIn