The BLS produced a research new vehicle index (R-CPI-U-NV) until May 2022 using a transactions dataset purchased from J.D. Power as part of our methodological improvements for the CPI. Detailed comparisons of the current CPI and this research methodology follow. The research methodology had the potential to more accurately measure changes in the cost-of-living associated with new vehicles at a similar or cheaper cost than the current index based on the CPI’s price survey, and was ultimately adopted for use in the CPI-U. We published the research index for a period of time and gather feedback before making any decisions on using this index as a replacement for the traditional new vehicles index.
The traditional survey of car dealerships relied on dealer estimated prices for a sample of hypothetical vehicle configurations based on consumer purchasing patterns that may have been several years old. For several years, we had been researching calculating price indexes for new vehicles based on transaction data and eventually released the new indexes on a research basis. The transaction data, purchased from J.D. Power, allowed us to use observed purchase prices and to reflect real-time market conditions in terms of the quantities and expenditure shares of vehicles sold. In addition to better representing the sales in the current month, the J.D. Power data provided substantially more observations and broader geographic coverage.
While the transaction data have many advantages over the survey data previously collected by the BLS, incorporating transaction data into the CPI presented several challenges. Traditional CPI methodology was not well suited for the sporadic nature of vehicle purchases and depended on fixing weights for an extended period of time rather than leveraging real-time weighting information. To address this, we developed an alternative methodology that allows us to make better use of the transaction data. This methodology is described in detail in a working paper: “A New Vehicles Transaction Price Index: Offsetting the Effects of Price Discrimination and Product Cycle Bias with a Year-Over-Year Index.”
The research found that with the transaction data, monthly, matched model price indexes declined persistently even as average prices (unit price index) rose. The matched model indexes only show the price change for the same model year vehicle. When a new model year version of a vehicle is introduced, it is offered at a relatively high price. As the months pass, the price drops as dealers offer steeper concessions and manufacturers begin to offer more rebates and incentives. Linking these price drops together results in a steadily declining price index. The traditional CPI methodology also showed these declines, but offsets them by showing the price change from a vehicle’s end of life price to the early life price of its next model year replacement. The traditional CPI method did not perform well when applied to the transaction data since indexes were very sensitive to how these offsetting price changes were weighted.
Instead of using these offsets, we focused on the price change across model years for similarly aged vehicles. This method avoids the complications of weighting within-model year price change with cross-model year offsets. Arguably, this year-over-year price measurement is more consistent with a Cost of Living Index (COLI) since a COLI should not show the price change from early adopters buying the newest version of a product and bargain hunters buying at the end of a model year. Our year-over-year index compares the price of a vehicle sold today and compares it with the price of the previous model year vehicle 12 months ago. The twelfth-root of the relative is taken to scale the price change down to a monthly frequency. These relatives are then aggregated with a Tornqvist formula, which is consistent with the superlative index formula used at the all items level for the Chained Consumer Price Index (C-CPI-U).
The resulting year-over-year index produces a measure of the trend in new vehicle pricing but misses higher frequency changes. In order to reflect short-run shocks and seasonal variation in the new vehicle market, we apply a time series filter to a monthly price index and extract a detrended component. While the monthly indexes do not represent the trend in price change well, the high-frequency components of the monthly indexes appear to accurately represent short-run fluctuations. Our research shows that long-run trends are very sensitive to index methodology but that high-frequency fluctuations are not. Moreover, both the J.D. Power and CPI new vehicle indexes display similar high-frequency patterns even though they are based on different data sources and methodologies. For our research index, we combine the year-over-year trend with the high-frequency component of a monthly index that uses a Tornqvist price index formula and model year replacement price offsets.
In January 2018, the monthly CPI new vehicle sample size was approximately 2,000 vehicles. In the traditional method, information from the Telephone Point-of-Purchase Survey (TPOPS) was used to select the dealerships surveyed for the new vehicle index. Next, a disaggregation process based on dollar volume sales was utilized to select the unique make, model, trim, and combination of options to be priced at each dealership for the index; each of those variables was associated with a unique specification on the ELI checklist.
Sampling was also used to select CPI’s geographic coverage. Instead of pricing in all small and mid-size cities, representative cities were sampled to represent all similar sized cities in a region. Additional information is available on the Consumer Price Index Geographic Revision for 2018 webpage.
The J.D. Power monthly new vehicle sample size is approximately 250,000 of all nameplates purchased from participating dealers across the country. Unlike the CPI, J.D. Power does not attempt to construct a representative sample. However, we have found that market shares in the J.D. Power data are similar to the representative sample of vehicles in the CPI. We remove vehicles specifically referenced as fleet vehicles, as well as vehicles not typically purchased for consumer use.
Unlike the CPI, the J.D. Power data does not sample geographically. The J.D. Power data are grouped into “Designated Market Areas” that correspond to all of the large cities in the current CPI sample and a large pool of small and mid-sized cities. We mapped the Designated Market Areas into the 38 old areas prior to 2018 and 32 new areas from December 2017 on. J.D. Power’s geographic indicators did not allow us to distinguish between A109 (New York City) and A111 (New Jersey-Pennsylvania suburbs), so these areas grouped together as A109 and their weight is combined in the national level index prior to 2018.
The traditional CPI methodology for pricing new vehicle purchases required a data collector to meet with a dealership respondent to select a vehicle for pricing through disaggregation. A data collector then revisited the dealer on a monthly or bimonthly basis to re-price the same make, model, and trim level while making updates to the vehicle option and base prices, transportation charges, and dealer preparation or miscellaneous charges, as well as estimated rebates and concessions. Data were collected throughout the entire month.
Since most new vehicle sales are negotiated and the exact vehicle may not have been sold, the data collector calculated the final price based on respondent estimates for the average concession and the average rebate over the past 30 days for that particular model and trim level. This relatively high level of respondent burden persistently threatened survey participation. There are also data quality issues because respondents could supply incorrect specifications or prices that required investigation and adjustment by the commodity analysts.
J.D. Power collects a large dataset of observed transaction-level new car prices and detailed vehicle information for approximately one-third of U.S. consumer class, new automobile sales. Each observation includes a transaction price as well as a set of 40 variables showing size rebates, vehicle characteristics, information on finance terms, and, sometimes, the cost of the vehicle to the dealer.
In both the current BLS and J.D. Power samples, prices are collected without sales tax; the sales tax is applied by BLS.
Price data for sampled vehicles were transmitted to our national office throughout the month on a flow basis. Commodity analysts reviewed individual observations, such as unusual price movements, and description updates and item replacement, and made adjustments (including quality adjustments) as necessary.
The CPI price relative was the estimate of the change in the estimated price for an observation from the previous period to the current period.
Once the data are received from J.D. Power, we remove fleet and commercial vehicles leaving consumer purchased vehicles. Then the data are filtered to find specific vehicle combinations (squishVINs=make/model/trim, etc.) that are new to the sample. A squishVIN is a shortened version of the vehicle identification number (VIN). The VIN is 17 digits whereas the squishVIN is made up of 10 digits from the first 11 digits of the VIN. The squishVIN is used to identify a specific vehicle configuration. Commodity analysts review the new squishVINs in order to make comparability decisions and to apply quality adjustments. The decisions and adjustments are applied to the data as they would have been under the current procedures used to produce the new vehicles component of the CPI.
With J.D. Power data, the price used is the geometric average transaction price for each vehicle type within each index area. New vehicles are reviewed in accordance to the procedures we use for the official index. This includes review for model year comparability and quality-adjustment when necessary using data received by BLS from the automobile manufacturers.
Monthly and bimonthly relatives were constructed based on comparing vehicles in month t with month t-1 or t-2 using the geomean price index formula with fixed sampling weights. Once the sales of a new model year vehicle exceeded those of its predecessor, we implemented a model year changeover and replaced the old vehicle version with the new (“50% rule”).
Index Formula: Geometric Mean
This index provides the input for the high-frequency component used in our final JDP index. Monthly relatives are constructed based on comparing vehicles in month t with month t-1 using the Tornqvist price index formula with expenditure share weights. The 50% rule is used to determine the timing of changeovers, and changeover relatives are based on the observed expenditure shares for the model years involved in the changeovers.
Index Formula: Tornqvist
This index provides the estimate of the long-run trend price change used in our final J.D. Power index.
Index Formula: Tornqvist
Model/model year combinations with no current expenditure
or no expenditure in the previous model year
are dropped and not used in calculating expenditure shares. Quality adjustments are applied to
For the research new vehicle index, basic indexes at the full sample and replicated levels combine the year-over-year (YoY) trend for each basic index that are adjusted with the latest cyclical change (monthly cycle) component from the respective basic index calculated.
|CPI||JDP (Monthly Cycle)||JDP YoY (Trend)|
50 % rule
|Continuous and substitution||Continuous and substitution, provided they have prices in t and t-1||Model year change prices (at squishVIN level) for t and t-12. Note, does not include price relatives for same model year (continuous) squishVINs|
|The price change for noncomparable model changeovers are explicitly being imputed by the ‘substitution price relative’, i.e. the price change of other (comparable and quality-adjusted) model changeovers.||The price change for noncomparable model changeovers are explicitly being imputed by the ‘substitution price relative’, i.e. the price change of other (comparable and quality-adjusted) model changeovers.||Implicit Substitution. No explicit imputation is done but this methodology is equivalent to implicit imputation. The price change for noncomparable model changeovers are effectively being implicitly imputed by the price change of other model changeovers since this index only uses cross-model year price change.|
Unavailable Price Imputation
|Cell relative. If a price is not collected, it is imputed forward assuming the price change is equivalent to other new vehicles in the same area.||None.||None.|
Average Price for a 2015 No-Brand Midsize Car w/ Navigation in March 2015 in Chicago = $37,000
Totals Sales for 2015 No-Brand Midsize Car w/ Navigation in March 2015 in Chicago = $370,000
Total Sales for all vehicles in March 2015 in Chicago = $5,000,000
Average Price for a 2014 No-Brand Midsize Car w/ Navigation in March 2014 in Chicago = $39,000
Totals Sales for 2014 No-Brand Midsize Car w/ Navigation in March 2014 in Chicago = $390,000
Total Sales for all vehicles in March 2015 in Chicago = $8,000,000
= Price relative = (37,000/39,000)^(1/12) = .996
Expenditure share in current March 2015 = $370,000/$5,000,000 = 0.074
Expenditure share in current March 2014 = $390,000/$8,000,000 = 0.049
= Tornqvist weight = .5(0.074+0.049) = 0.062
= .996^.062 = 0.99975153354
This is then aggregated with the other vehicle price relatives and weights into a Tornqvist Index for Chicago New Vehicles:
Under the traditional CPI method, an item-area price relative was calculated using the weighted geometric mean of all price changes, where the weight was an estimate of base-period expenditures. Those price relatives were used to calculate each basic item-area index, and those item-area indexes were aggregated across items and areas using a Laspeyres formula.
Under the research YoY method, a 12-month item-area price relative is calculated using a weighted Tornqvist formula, where the weights are the expenditures for each squishVIN for each month, t and t-12. The 12th root of that relative is then taken to create a monthly relative and YoY index.
The resulting price index is then adjusted using a formula to model the latest cyclical change, i.e. the application of a time series filter to the monthly JDP index to extract the cyclical change. The price relatives are then used to update the basic item index and, analogous to the current method, those item-area indexes are aggregated across items and areas using a Laspeyres formula.
We published indexes for new vehicles (cars and trucks combined), cars, and trucks in their taxed and untaxed forms for the nation as a whole and for each of the CPI index areas until May 2022. The CPI new vehicles index included motorcycles whereas the research index did not. The CPI revised its geographic structure at the beginning of 2018. We list indexes for the 38 old areas (with A111 combined with A109) prior to 2018 and 32 new areas from December 2017 on. Tax information needed for the research index was not available prior to January 2018, so taxed indexes (currently published CPI components are all taxed) are only available from January 2018. However, we provide untaxed research indexes that go back to 2008. We also include published, national level, taxed indexes for New Vehicles, Cars, and Trucks. These are published in rebased form for December 2007 = 100 and January 2018 = 100 in order to facilitate comparisons with the research indexes. Untaxed official CPI indexes are unavailable for publication. For both the research and official national level indexes, we also include standard errors.
R-CPI-U-NV data were published three business days following the monthly CPI-U data release until May 2022, when the R-CPI-U-NV methodology was adopted for use in the CPI-U. See the CPI Release Calendar for publications dates.
The Bureau is soliciting feedback on this index as a replacement for the current new vehicles index and is thus publishing it on a research basis. Our intention is to release an updated file each month after the initial release on May 16, 2019; it will be released approximately three days after the monthly CPI release. As we continue evaluation of the use of this new source, we would specifically like to know whether you agree that this proposal constitutes an improvement over the existing methodology and data source, but appreciate any additional feedback on this data. Please send your feedback on the research new vehicle data and other CPI related questions to email@example.com.
Last Modified Date: April 27, 2023