Improving Economic Price Statistics through the Use of Alternative Data
Sunday, Jan. 5, 2020 8:00 AM - 10:00 AM (PDT)
- Chair: Erica Groshen, Cornell University
A Nontraditional Data Approach to the CPI Gasoline Index: CPI Crowd-Sourced Motor Fuels Data Analysis Project
AbstractThe Bureau of Labor Statistics (BLS) mostly relies on manual collection of its prices for commodities and services, but momentum has increased to explore other channels and sources of price data. It has become a central theme in executing the BLS mission to employ alternative means of data collection to make our indexes more accurate, timely, and relevant. One current study is the Crowd-Sourced Motor Fuels Data Analysis project, which may lead to replacing the current Consumer Price Index (CPI) gasoline sample. These motor fuel price data are obtained via web scraping, with the site’s permission. Our interest in this data stems from improving the efficiency of collecting data for these commodities, but an equally important side effect is that using this source would greatly increase our number of data observations beyond what our current survey methodology and sampling frame can provide. Indexes were constructed using the crowd-sourced app data, as well as average prices to measure monthly price fluctuations. The results of our preliminary analysis show that the crowd-sourced data and the CPI for gasoline behave very similarly. As a source of monthly pricing information, the gasoline data from this crowd-sourced app may be a suitable replacement for the CPI field-collected data. The experimental indexes have a long-term trend similar to the official CPI for gasoline. Additionally, the larger sample size of the crowd-sourced data captures, to a greater degree, the price change variance exhibited within a geographic zone.
Implementing an Alternative Data Source to Estimate Producer Price Indexes for Selected Financial Services
AbstractThe Bureau of Labor Statistics (BLS) aims to measure the average change over time in the selling prices received by domestic producers for their marketed output through its Producer Price Index (PPI). BLS economists have difficulty finding “big” datasets that include the necessary inputs for estimating the change in prices that producers receive for the goods they produce and the services they provide. This paper describes the process BLS used for one area of financial services, where success came after several attempts and still provides opportunities for expansion. BLS is using a large, purchased database for financial services within the U.S. economy. Economists extract thousands of data points per day to compute a weighted average price for use in PPI estimates. BLS is using this source to replace directlycollected data for municipal debt securities dealing, corporate debt securities dealing, and equities securities dealing in the investment banking and securities dealing industry. For corporate bond dealing, BLS increased the number of transactions measured per day by 3,971 percent; for municipal bond dealing, by 6,640 percent. The PPI also blends data from the same database with directly-collected data to escalate merger and acquisition deal values and underwriting transaction values in the investment banking industry. For comparison, we describe BLS’ success obtaining large datasets with transaction data for a few respondents in retail trade services. The PPI relies on retail margins to estimate price change for this sector. One company provides a dataset with acquisition and sales prices from which the PPI can compute the margin information for over 30,000 products covering 15 different item groupings. For both experiences, we describe what circumstances led BLS to consider these alternative sources and collection methods for the PPI, the research and decision points, the results, and future plans.
JEL Codes: C43, E31, G12, L84
Keywords: producer price index, price level, equities, fixed income securities, financial services
Using Administrative Data to Calculate Export Price Indexes
AbstractThe BLS currently uses the administrative dataset of international trade transactions of merchandise goods as the sample frame to select representative products and companies to collect a basket of approximately 22,000 representative goods that form the basis of published Import and Export Price Indexes, Principal Federal Economic Indicators. Previous research using the administrative trade data established a prototype unit value index for homogeneous products. This research project expands the analysis to create historical monthly time series for 2012-2017 of all 127 5-digit BEA End Use monthly export unit value indexes, based on 20+ million trade records. The unit value indexes are evaluated for homogeneity and comparability to official price indexes. Among the 127 indexes, 24 unit value indexes covering 23% of export trade are deemed high quality. The impact of the historical data on the real value of exports from 2012 to 2017 is estimated. If used as deflators, real value of goods exports comprising these 24 product areas would have declined .75% more per year, and the impact on real value of all-goods exports would have been -.15 % yearly.
- E3 - Prices, Business Fluctuations, and Cycles
- O3 - Innovation; Research and Development; Technological Change; Intellectual Property Rights