Notes on scanner data price collection in

Notes on scanner data & price collection in the CPI Muhanad Sammar Statistics Sweden muhanad. sammar@scb. se 1

Swedish price collection in (the) field Outlet sample Product sample Price collection • Manually by price collectors Collection Period • 3 weeks 2

Findings - CPI Missing prices Time-consuming Monotonically for price collectors 3

Eurostats regulations - Commission regulation (EC) No 2602/2000 “(3) Prices used in the HICP should be purchaser prices actually paid by households to purchase individual goods and services in monetary transactions, including any taxes less subsidies on the products, after reductions for discounts for bulk or off-peak purchases from standard prices or charges…” 4

National Accounts In the National accounts the transaction prices should be registered Scanner data measures transaction prices Better basis than manually collected prices for offered products 5

Scanner data and EAN-barcode Identifies uniquely an article, package or service The EAN-number is international Article Info - supplier, brand, width, depth, weight, labeling, etc 6

Scanner data EAN-code - not fruit & vegetables, meat & fish Three deliveries each month Variables • Turnover per store and EAN-code • Number of packages sold • Price (=Turnover/Number of packages sold) 7

Four ways to use scanner data Replace manually collected prices with scanner data As auxiliary information Compute index from a census based on all products For auditing and quality control 8

The risks of using scanner data Deposits for beverages Package sizes Delivery problems Missing values • Small volumes • Significant price increase 9

A pilot study - Scanner Data (SD) and Manually Collected Prices (MCP) in comparison Matching categories SD and MCP. Equal prices SD and MCP. Unequal prices SD, but not MCP Not SD and not MCP, but not SD All (49 600 products) 75 11 2 2 10 10

A pilot study - Monthly price index 2009 for most of the everyday commodities Month January February March April May June July August September October November MCP index 100. 24 100. 73 101. 02 100. 94 101. 14 101. 67 102. 11 102. 12 102. 35 102. 54 102. 68 Difference SD-MCP – 0. 09 – 0. 14 – 0. 17 – 0. 11 – 0. 12 – 0. 13 – 0. 18 – 0. 17 – 0. 22 S. E. * 0. 10 0. 11 0. 10 0. 14 0. 16 0. 20 0. 18 0. 19 0. 16 0. 15 0. 17 * Standard error of MCP (manually collected prices) index 11

Auditing in one selected outlet - price discrepancies for 20 of the products Prices were erratic for 15 of the products For 3 products the store had not updated the shelf price with the latest price information In 2 cases, unclear descriptions of the products 12

Decisions to make Minimize sources of error • Improved education of price collectors • Increase the number of products • Increase the number of outlets Average price; a week or a three-week period Transaction price • Substitution effect 13

Concluding remarks Today's method is not optimal Scanner data is compatible with EU’s regulatory framework Scanner data allows adequate price data In the long run; may be more cost-effective As part of quality assurance, Statistics Sweden will continue to engage and develop routines that minimize price collector errors CPI is an important measure and must be computed on accurate and reliable prices 14

THANK YOU FOR YOUR ATTENTION 15