@article{Beatty:205898,
      recid = {205898},
      author = {Beatty, Timothy and Snyder, Samantha},
      title = {Regional Fresh Fruit and Vegetable Price Indices Using  Supermarket Scanner Data},
      address = {2015},
      number = {330-2016-13507},
      series = {Poster},
      year = {2015},
      abstract = {The Consumer Price Index (CPI) as constructed by the  Bureau of Labor Statistics is one of the most widely-used  measures of both aggregate and category specific price  indices in the US. Within the broad CPI, price indices for  several food categories and individual items are also  calculated. For example, there is an index for aggregated  fresh fruits and vegetables, while there are also separate  indices for apples, bananas, tomatoes and onions, among  others. In general, there has been little innovation in  terms of CPI methodology over the past several decades.  This research combines leading-edge techniques in index  number theory with a strong practical application to the  food CPI, a highly cited economic indicator with real  implications on policy.  Specifically, this study will help  to detail and quantify the range in prices of fresh fruits  and vegetables across major metropolitan areas in the US.  As focus is placed on American diet quality, with special  attention paid to low-income households, food assistance  programs feature heavily in policy recommendations aimed at  improving household purchase and consumption decisions.  This research sheds light on how regional fresh fruit and  vegetable price differences might impact household  purchasing power given nationally fixed benefit levels for  food assistance program participants.

While the  construction of the US’ CPI has remained largely unchanged,  statistical agencies in a few European countries have  recently augmented traditional CPI construction approaches  to take advantage of improvements in data collection  capabilities, namely the proliferation of store scanner  data and the adoption of chained indices. The use of store  scanner data presents a host of challenges, however, most  notably contributing to chain drift in resulting indices  (Ivancic et al. 2011). Chain drift, common in indices built  with high-frequency data, is characterized by indices that  do not return to a value of one even when prices return to  those in the base period. To address this issue, Ivancic et  al. (2011) propose a rolling window GEKS method that serves  as a sort of hybrid between traditional fixed base indices  and the more flexible chained indices. 

A second challenge  arises when trying to perform multilateral price  comparisons. Using a panel of price indices (multiple  regions, measured monthly, over several years) adds an  additional dimension to the typical index construction  focused on temporal variation only. Spatial comparisons are  especially confounding because, unlike time, there is no  natural ordering amongst regions. Hill (2004) proposes a  number of potential approaches to multilateral index  construction and comparison and the Minimum Spanning Tree  and the standard Geary Khamis methods are most amenable to  the challenges presented by high frequency price  measurements like those contained in store scanner  data.

Adopting both of these important innovations in  index number theory potentially represents an important  contribution to the field. From a more applied standpoint,  the price indices constructed in this paper fill important  holes left by the CPI and apply the use of store scanner  data and the rolling window GEKS method to the US market  for fruits and vegetables. First, we have extended the  measurement of the fresh fruit and vegetable subcategories  to include monthly observations at the Metropolitan  Statistical Area (MSA) level. The CPI currently only  reports a price level for ‘Food’ at the MSA level. Second,  we have adopted the methods recommended in Hill(2004) to  enable price level comparisons across regions. As  constructed, multilateral, cross-MSA comparisons are not  possible with the CPI as each MSA’s index is constructed  independently of all the others.

To perform our price  analysis, we use store scanner data collected by IRI. The  data includes food at home purchases recorded at the  store-UPC level made at a variety of retailer types  including grocery store, supercenter and convenience store  formats. Purchase quantity and total expenditure are  reported weekly at the store level over the five-year  period between 2008 and 2012. Along with the quantity and  expenditure level for each UPC, store location is also  recorded. Using purchase data on fresh fruits and  vegetables as well as store location, we are able to  construct price indices for the 26 MSAs covered at least  quarterly by the CPI.

Because we are able to perform both  inter- and intraregional price level comparisons across the  five years studied, our results may be particularly useful  in informing policy. The Supplemental Nutrition Assistance  Program (SNAP) provides low-income households with monthly  benefits earmarked for purchases of food. In general,  benefit levels are determined based on the size of the  household and are uniform nationwide. As a targeted  response to concerns about diet quality, especially among  low-income and low-education households, there are some  efforts to link SNAP program participation with nutrition  education and/or limit the types of foods that are eligible  to be purchased with SNAP dollars (Shenkin and Jacobson  2010). While there is some evidence of significant regional  food price variation, especially among fruits and  vegetables (Sturm and Datar, 2011), more work needs to be  done. This research enables a closer look at regional price  differences in fresh fruits and vegetables and could shed  light on whether uniform benefit levels are making  purchases of fresh fruits and vegetables relatively more  difficult for low-income households in some regions of the  country.},
      url = {http://ageconsearch.umn.edu/record/205898},
      doi = {https://doi.org/10.22004/ag.econ.205898},
}