The Implicit Prices of Finfish and Shellfish Attributes and Retail Promotion Strategies

MISSING IMAGE

Material Information

Title:
The Implicit Prices of Finfish and Shellfish Attributes and Retail Promotion Strategies Hedonic Analysis of Weekly Scanner Data in the U.S.
Physical Description:
1 online resource (80 p.)
Language:
english
Creator:
Gold, Glen C
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Food and Resource Economics
Committee Chair:
Larkin, Sherry L
Committee Members:
Adams, Charles M

Subjects

Subjects / Keywords:
acnielsen -- finfish -- fish -- hedonic -- labeling -- promotion -- scantrack -- seafood -- shellfish
Food and Resource Economics -- Dissertations, Academic -- UF
Genre:
Food and Resource Economics thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Using retail-level grocery store scanner data in the U.S., this study aims to estimate the implicit prices of a suite of promotional, situational, labeling and other product attributes within the frozen finfish and shellfish market. Past research has shown that “product specific” attributes (e.g., size, product form, quality and species) and “situational” attributes (e.g., natural disasters, health concerns, branding, country of origin and production method) affect the value of seafood products. Recently, various “labeling” attributes (such as certifications) have been developed to market products that are more sustainable, and at least one published study has shown these labels have received higher prices at the retail level.  Few studies have assessed the implied value of labels or marketing activities which do not require third-party certification. ACNielsen’s Scantrack data was used to estimate price premiums and discounts for promotional, situational and labeling attributes as well as product specific attributes for seafood (i.e., fish and shellfish, marine and freshwater).  In particular, this study will use data on frozen unbreaded fish and shellfish products (exclusive of crabs, oysters and shrimp) sold weekly in the U.S. from June 2007 through May 2010.  Observations related to products (i.e., defined at the 10 UPC level) that were not sold in at least one-third of the weeks or species groups that were not at least 1% of the volume of products sold were deleted. In the U.S. finfish dataset, 11 groups of fish species had some products labeled as “wild.” Of those species, 14% of the weekly observations were products with “wild” on the label and the wild label was found to command price premiums of 5.9% to 43.7%, depending on the species. In the shellfish dataset, nearly one-third of the weekly observations were products labeled as “imported” and the imported label was found to increase price 5.9% in the case of one species (lobster) but decrease price for four other species groups from 11.0% to 34.1%, depending on the species.  Across both product types (i.e., unbreaded frozen fish and shellfish), promotional activities resulted in products being sold at a discount from 15.2% to 29.7%, when 100% of sales were promotional. The retail prices for only two species groups (scallop and lobster) declined after the Deepwater Horizon oil spill, which occurred in late April 2010, while the prices for 10 others increased, but only marginally (i.e., from 3.1% to 9.6%).  Lastly, products sold under a retailers’ own label (grocery store brand) were found to reduce the price of shellfish products by an average of $3.37/lb. but increase the price of finfish products by an average of $0.37/lb.; only the latter result concurs with previous US market studies.  Collectively, it is concluded that interaction terms are necessary to properly model the seafood market using hedonic analysis, as the implicit prices of attributes must be estimated while controlling for species.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Glen C Gold.
Thesis:
Thesis (M.S.)--University of Florida, 2012.
Local:
Adviser: Larkin, Sherry L.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2013-02-28

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2012
System ID:
UFE0044311:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 THE IMPLICIT PRICES OF FINFISH AND SHELLFISH ATTRIBUTES AND RETAIL PROMOTION STRATEGIES: HEDONIC ANALYSIS OF WEEKLY SCANNER DATA IN THE U.S. By GLEN CODY GOLD A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF F LORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Glen Cody Gold

PAGE 3

3 To my family

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank my chair and supervisory c ommittee. I their wisdom and guidance, this study would not have be en possible. I acknowledge the Institute of Food and Agricultural Sciences for funding the summer internship which initiated this research, and my graduate research assistant appointment. I thank my family for believing in me, especially my and turns. This section would also not be complete unless I give my upmost thanks to my fianc e for her continue d interest in listening to my progress and discoveries during the course of this study.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 7 LIST OF FIGURES ................................ ................................ ................................ ......................... 8 ABSTRACT ................................ ................................ ................................ ................................ ..... 9 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 11 The Seafood Industry ................................ ................................ ................................ .............. 11 Background and Previous Studies ................................ ................................ .......................... 13 Hedonic Methodology ................................ ................................ ................................ ............ 16 Objective ................................ ................................ ................................ ................................ 18 Overview of Study ................................ ................................ ................................ .................. 19 2 DATA ................................ ................................ ................................ ................................ ..... 23 Overview ................................ ................................ ................................ ................................ 23 Source of Data ................................ ................................ ................................ ........................ 23 Data Description ................................ ................................ ................................ ..................... 24 Regional Aggregation ................................ ................................ ................................ ...... 24 Item and Brand Aggregation ................................ ................................ ........................... 24 Monthly and Weekly Aggregation ................................ ................................ .................. 25 Product Categories ................................ ................................ ................................ ........... 25 Data Manipulation ................................ ................................ ................................ .................. 26 Variables Created for Analysis ................................ ................................ ............................... 28 3 EMPIRICAL ANALYSIS ................................ ................................ ................................ ...... 38 Overview ................................ ................................ ................................ ................................ 38 Estimation and Model Specification ................................ ................................ ....................... 38 Shellfish Results ................................ ................................ ................................ ..................... 40 Premiums for Species ................................ ................................ ................................ ...... 41 Implied Value of Promotional A ctivities ................................ ................................ ........ 42 Price Effects f rom the Deepwater Horizon Oil Spill ................................ ....................... 42 Relative Value of Imported Shellfish ................................ ................................ .............. 43 Other Implied Prices for Shellfish ................................ ................................ ................... 43 Finfish Results ................................ ................................ ................................ ........................ 45 Premiums for Species ................................ ................................ ................................ ...... 45 Implied Value of Promotional Activities ................................ ................................ ........ 46 Price Effects f rom the Deepwater Horizon Oil Spill ................................ ....................... 46

PAGE 6

6 Relative Value of Wild Labeled Finfish ................................ ................................ .......... 47 Other Implied Prices for Finfish ................................ ................................ ...................... 48 4 CONCLUSIONS ................................ ................................ ................................ .................... 70 Summary of the Study ................................ ................................ ................................ ............ 70 Summary of the Results ................................ ................................ ................................ .......... 70 Findings ................................ ................................ ................................ ................................ .. 71 Key Results and Implications ................................ ................................ ................................ 73 Future Work ................................ ................................ ................................ ............................ 74 APPENDIX : DEVELOPMENT OF THE DEPEND ENT VARIABLE ................................ ... 77 LIST OF REFERENCES ................................ ................................ ................................ ............... 78 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ......... 80

PAGE 7

7 LIST OF TABLES Table p age 1 1 Summary of selected recent studies involving hedonic modeling of food products using scanner data ................................ ................................ ................................ .............. 20 2 1 Variables provided by ACNielsen ................................ ................................ ..................... 33 2 2 List of grocery food chains covered by ACNielsen store level data ................................ .. 34 2 3 Total U.S. seafood sales and marke t share at grocery stores that exceeded $2 million in 2010 by category ................................ ................................ ................................ ............ 36 3 1 Shellfish variable means and hedonic model results ................................ ......................... 50 3 2 Shellfish predicted price and interaction effects ................................ ................................ 52 3 3 Frequencies of name brand and store brand shellfish products by species group ............. 53 3 4 Finfish variable means and hedonic model results ................................ ............................ 54 3 5 Finfish predicted price and interaction effects ................................ ................................ ... 58 3 6 Frequencies of name brand and store brand finfish products by species group ................ 59

PAGE 8

8 LIST OF FIGURES Figure p age 1 1 Live weight of total fisheries pr oducts sold in the U.S. annually 1990 2009 ................... 21 1 2 Seafood market share by product category form in 2010 ................................ .................. 22 2 1 Proportion of promotional sales for shellfish and finfish by week, 3 year average .......... 37 3 1 Percent change of predicted price for p romotional sales by shellfish species group ......... 60 3 2 Percent change of predicted price after the Gulf of Mexico oil spill by shellfish species group ................................ ................................ ................................ ...................... 61 3 3 Percent change of predicted price for import labeling by shellfish species group ............ 62 3 4 ................................ ........ 63 3 5 ................................ ......... 64 3 6 Percent change of predicted price for promotional sales by finfish sp ecies group ............ 65 3 7 Percent change of predicted price after the Gulf of Mexico oil spill by finfish species group ................................ ................................ ................................ ................................ .. 66 3 8 Percent change of predicted price for wild labeling by finfish species group ................... 67 3 9 ................................ ........... 68 3 10 ................................ ............ 69

PAGE 9

9 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Mas ter of Science THE IMPLICIT PRICES OF FINFISH AND SHELLFISH ATTRIBUTES AND RETAIL PROMOTION STRATEGIES: HEDONIC ANALYSIS OF WEEKLY SCANNER DATA IN THE U.S. By Glen Cody Gold August 2012 Chair: Sherry Larkin Major: Food and Resource Eco nomics Using retail level grocery store scanner data in the U.S. this study aims to estimate the implicit prices of a suite of promotional, situational labeling and other product attributes within the frozen finfish and shellfish market Past research h ( e.g., size product form, quality and species (e.g., natural disasters health concerns, branding, country of origin and production method ) affect the value of seafood products Re cently, various certifications ) have been developed to market products that are more sustainable and at least one published study has shown these labels have received higher prices at the retail level. Few studies have asse ssed the implied value of labels or marketing activities which do not require third party certification was used to estimate price pre miums and discounts for promotional, situational and labeling attributes as well as product sp ecific attributes for seafood (i.e., fish and shellfish, marine and freshwater) In particular, t his study will use data on frozen unbreaded fish and shellfish products (exclusive of crabs, oysters and shrimp) sold weekly in the U.S. from June 2007 throug h May 2010. Observations related to products (i.e., defined at the

PAGE 10

10 UPC level) that were not sold in at least one third of the weeks or species groups that were not at least 1% of the volume of products sold were deleted. In the U.S. finfish dataset 11 gr oups of fish species had some products labeled as wild Of those species, 14% of the weekly observations were products with wild on the label and the wild label was found to command price premiums of 5.9% to 43.7% depending on the species In the sh ellfish dataset nearly one third of the weekly observations were products labeled as i mported 5.9% in the case of one species (lobster) but decrease price for four other species groups from 11.0% to 34. 1% depending on the species Across both product types ( i.e., unbreaded frozen fish and shellfish), promotional activities resulted in products being sold at a discount from 15.2% to 29.7% when 100% of sales were promotional The retail prices for only two species groups (scallop and lobster) declined after the Deepwater Horizon oil spill, which occurred in late April 2010, while the prices for 10 others increased, but only marginally (i.e., from 3.1% to 9.6%). Lastly, products sold under a own label (grocery store brand) were found to reduce the price of shellfish products by an average of $3.37 /lb. but increase the price of finfish products by an average of $0.37 /lb. ; only the latter result concurs with previous US market studies. Collecti vely, it is concluded that interaction terms are necessary to properly model the seafood market using hedonic analysis, as the implicit prices of attributes must be estimated while controlling for species.

PAGE 11

11 CHAPTER 1 INTRODUCTION The Seafood Industry Natu re provides an extensive selection of marine animals for harvest and consumption. Between the year 2000 and 2009 total consumption of products in the United States (U.S.) has flu ctuated around 20 billion pounds per year ( See Figure 1 1 ; NOAA, 201 0, Table 895) ) T he 2007 per capita consumption of red meat (primarily beef, pork and lamb) poultry (primarily chicken and turkey) and fish combined was 200.4 pounds About half (55%) was red meat, 36.7% was poultry and 8.2% was fish (AMI, 2009). While the consumption of red meat and poultry is relatively high, the number of species is very small compared to seafood (finfish and shellfish, marine and freshwater). In addition, raw red meat and poultry products are differentiated at the retail level by di fferent cuts ; most of the differentiation with respect to raw seafood is with the species. Species of fish and shellfish, especially when considered within broad species groups (e.g., salmon, tuna, cod or shrimp), have distinctly different attributes incl uding taste, texture, color, and ease of preparation. While these differences exist within the other types of animal based proteins, the differences are arguably greater among seafood products. The seafood industry in particular is reliant on a large sui te of species, many of which consumers can identify (at least at the broad level such as salmon, shrimp, or catfish). That said ; p ollock, hoki or catfish. Processing procedures such as breading can also disguise species In fact, a large number of breaded fish products are In contrast, some consumers are willing to pay price premiums for

PAGE 12

12 consumption of a specific species of fish especially when it is sold in unprocessed form C ostly processing procedures such as smok ing also result in more expensive (valuable) products Whereas other protein sources are farmed, and therefore available on consistent bases, many seafood products are wild caught such that supply is dependent on the cyclical behavior of natural processes including seasonal migrations and spawning. Because some products are only available during certain times of the year, ext ensive regulations have been placed on the harvesting of particular products. This has an added effect on promotional behaviors of retail outlets. Commercial harvest periods are relatively known, so retailers can anticipate supply and de mand based on the se seasons. Much research and analysis has been performed to understand what drives prices and promotional behaviors of various products within this industry. We know from past research that consumers place value on a variety of seafood product attributes and qualities including species, product form, processing, size, region and product grade/quality (Roheim et al., 2007 and Gardiner, 2007; McConne l l and Strand, 2000) Situational determinants which have been found to influence consumers purchasing beha vior typically involve health concerns (Leek and Maddock, 2000). However, s ituational determinants involve all those situations that change a ; which can include exogenous events, placing a product on sale or placing it in a different location in the store The recent Deepwater Horizon oil spill in the Gulf of Mexico is an example of such a situational determinant. Holiday traditions, such as turkey on Thanksgiving, are also situational determinants which may Different types of labeling on food products have been shown to affect consumers purchasing behavior ( e.g., Carew, 2012; Roheim et al. 201 1 ). Some labels, such as country of

PAGE 13

13 origin or nutritional information are mand ated by governments. Others provide location of production or harvest. Furthermore, some lab els symbolize a product as being certified organic or sustainable in production In the seafood industry, the mislabeling of products has been of growing concern to consumers. A 2010 study found consumers were willing to pay an average premium of $0.83 to $3.18 for grouper entrees at restaurants with a product integrity label (Ropicki, Larkin and Adams, 2010). The use of certification s and eco labeling within the seafood industry has developed widely within the last ten years. The theory that the market will reward products with these labels is highly supported by environmental groups. Many producers are considering adopting the labeling process with the thought it may hold a high return on investment. Policy makers question if the labels reflect tangible and measureable improvements in the environment, if not the level of the fish stocks directly (Roheim et al. 201 1 ). While the effectiveness of eco labeling in changing perception is still a topic of research and debate recent research has determined that such label s have resulted in higher retail prices I t must be noted that this study was in a limited market and for a single species partially du e to the limited markets in general for environmentally certified seafood (Roheim et al. 201 1 ). Regardless of the underlying environmental effects of labels, there is no doubt that marketing efforts will continue As such, research will continue to atte mpt to measure consumer preferences both before and after products have been introduced to the market. Before discussing the particular approach adopted in this study, some background and previous literature are reviewed in the following section. Backgroun d and Previous Studies The majority of past research on the value of seafood attributes has

PAGE 14

14 different attributes (characteristics) of seafood pro ducts. This methodology is necessary if new attributes are being considered such as new product forms or processing techniques are being considered. This is because there would be no existing data with which to estimate the value that consumers place on such attributes; we simply must ask buyers how they feel about the proposed attributes and what they might be willing to pay (or the discount they would be willing to accept) for a positive (negative) change in existing products. Whether or not consumers actually follow through with their statements and reveal these preferences when the new products are available in the market place is not known. This is a common criticism of these types of studies that is, that are based on historic market data and is a general concern regarding the ability to obtained unbiased information through surveys. Researchers often have no choice but to use a stated preference a pproach if they are investigating new products or attributes, which is the essence of traditional marketing. Luckily, there is a vast amount of literature on how to address several types of biases in stated preference studies. However, in cases where dat a are available, employ ing the revealed preference approach is preferred Fortunately, t echnological advances, such as the UPC label on packaging, have allowed for improved detail in the tracking of products sold at the retail level ; these data are typica of sale The companies which collect th ese data, such as ACNielsen typically provide an analysis of this data to clients for a substantial cost. Recently the se companies have been making the raw data available for purchase. The availability of th ese raw data has enabled researchers to conduct revealed preference studies Much interest lies in the results from this type of empirical analysis in part because t hey can verify or dismiss results from stated preference studies.

PAGE 15

15 Using the reported sales price with the corresponding attributes of the products sold the implicit (or implied) prices of various product attributes can be calculated (McConnell and Strand, 2000). This theory discussed further in the following section, has not been limited to the seafood market. Recent hedonic price studies of other food types include eggs (Chang, Lusk and Norw ood, 2010), wine (Carew and Florkowski, 2010), and chicken (Ahmad and Anders, 2012). Table 1 1 includes a summary of selected recent hedonic analyses of food products that have used scanner data. This increase in hedonic analysis is in part d ue to the ava ilability of retail scanner data M ore studies have been able to use hedonic analysis to calculate the revealed preferences of consumers. Scanner data provide a level of detail in retail market sales that cannot be found in other data sources, such as go vernment provided data. It is this level of detail that makes hedonic analysis of retail food sales possible. The data simply contains an abundance of product attribute information embedded within the UPC of each observation (Roheim et al 2007 ; 2010 ) Auction and home sales data have been popular sources of data used in hedonic analysis in the past. While the theory of hedonic analysis is widely accepted, guidance on the applicat io n is much less clear l form Both linear and log linear functional forms have been used in published literature. Ahmad and Anders (2012) defend the use of the log linear form as follows : Mutually advantageous exchanges of attributes by consumers may not be possible, resultin g in marginal utilities that may not be proportional in equilibrium. We therefore would expect the hedonic price function to be convex in equilibrium and a nonlinear functional form to be the appropriate specification (p. 120 ). The log linear functional form may also be more popular due to the interpretation of parameter estimates percentage (relative) changes rather than currency (absolute) changes.

PAGE 16

16 T he published hedonic literature especially those using scanner data, does not provide much advice with respect to regression techniques or econometric problems such as autocorrelation, heteroscedasticity and multi collinearity Some models are developed as more broad market analysis tools, using Ordinary Least Squares (OLS) regression and accept the fact th at better models could be developed Roheim et al. (2007) accepts multicollinearity as an issue, and suggests developing individual models for each product group (such as species of fish) to reduce this problem. This is in fact the approach used in Ro heim et al. (201 1 ) to determine the price premium associated with eco labeled Alaskan Pollock sold in the London market. While this study was narrowly focused, it was the first to show empirically that consumers pay a price premium (14%) for a seafood pro duct bearing the eco label. In reference to Steiner (2004), Ahmad and Anders employed the used of interaction terms 2012, p. 120) Advantages of this approach wo uld suggest, for example in this study, that the implicit price effect of breading a shrimp is different than breading any other species of seafood. Attributes that have been shown to affect the price of seafood include flavor and texture (Kinnucan and Wes sels, 1997), product quality (McConnell and Strand 2000 ), form, brand, species, size and processing (Roheim et al., 2007 ). A 2006 study by Leek and Maddock investigated the roles that intangible attributes such as availability, environment and other situ Hedonic Methodology This study employs techniques accepted in the literature to understand promotional, labeling and situational effects on the seafood industry. This study will i mplement the hedonic theory developed by Rosen ( 1 974 ) The basis of the hedonic model is the assumption that a

PAGE 17

17 product is composed of a variety of specific attributes or characteristics with which the consumer identifies. It has been wid e ly used in implicit price es timation models for a variety of market goods. The hedonic model attempts to estimate the value or shadow price that is contributed by each individual attribute of the product A firm or consumer may estimate the total price of a product by totaling all of the attributes corresponding shadow prices The hedonic model takes the market prices of various goods and disaggregates the total value (price) into its component parts, most ly through the use of dummy variables. In this study the use of retail sca nner data supplies the market prices of the various goods. The shadow prices are obtained from the coefficients. In a linear model, a negative parameter estimate for a n attribute means that the attribute reduced the overall price of the good. Conversely a positive parameter estimate means that the attribute increase d the overall price of the good. Ultimately, consumer preferences (i.e., willingness to pay or less for a good depending on whether the attribute is present or absent, respectively) are extrac ted for each characteristic of the product being modeled Generally, first degree hedonic models models that regress price on the attributes of the product(s) ignore production and demand effects on the price of a good ; however, changes in production a nd demand can be picked up through the use of seasonal attributes within hedonic modeling. A 1999 hedonic analysis of the apple market found price premiums associated with the summer and fall marketing seasons, and that the cyclical nature of apple prices closely followed the marketing seasons (Carew, 2000). demand exists for each attribute. Murray (1983) claims this is missing from empirical

PAGE 18

18 s model when demand estimates are included and that real world demands for attributes do have income effects just like for the product overall If so, the income effects will bias the demand parameter estimates. This study hypothesizes that P=f(X, Y) w here P is the price of a seafood product, X is a vector of promotional, labeling and situational variables and Y is a vector of product attributes. In an attempt to model products closer to the natural animal form (least processing) this analysis is condu cted with two categories of ACNielsen retail scanner data: frozen u nbreaded f ish and frozen u nbreaded r emaining (shellfish). These categories were selected since u nbreaded products have less processing in general than their breaded counterparts and theref ore should be more comparable at the species level The unbreaded product categories also contain a much larger collection of identifiable shellfish and finfish species and forms (e.g., a large portion of the breaded fish category contains products labele ). In 2010, the unbreaded products also account for the largest proportion of sales in the seafood market ( see Figure 1 2 ). Objective The objective of this paper is to determine the implicit price of a suite of promotional, labeling and situational variables that have yet to be investigated with seafood using a hedonic model of retail sales in the U.S This study will estimate pricing models for unbreaded frozen fish and shellfish. C omparisons of promotional behaviors, labeling practic es and other hypothesized pricing effects will be made between the shellfish and finfish markets in the United States Specifically, we investigate the price changes associated with different species of fish and shellfish during promotions labeled as imp orted or wild, and sold after the Deepwater Horizon Gulf Oil Spill.

PAGE 19

19 Overview of Study Promotional price changes are examined for different species of shellfish and finfish. igated. The effect on retail price of seafood after the Deepwater Horizon oil spill in the Gulf of Mexico occurred is studied The Data chapter will describe how the data were organized by ACNielsen how the data were prepared for analysis, data manipula tion and the creation of variables used in the final models. The Empirical Analysis chapter will describe model specification, estimation, and results for both the shellfish and finfish data The Summary and Conclusions chapter will discuss the results a nd conclusions. Key results are further examined and suggestions for future applications of hedonic analysis using scanner data are made.

PAGE 20

20 Table 1 1. Summary of s elected r ecent s tudies involving h edonic m odeling of f ood p roducts using s canner d ata Author s Product Method Key Findings Roheim, Gardiner and Asche (2007) Frozen fish Two UK markets; regression analysis of general frozen fish product attributes. Regions, species, brands, product form, package size and processing all have significant effects on price. Roheim, Asche and Santos (2011) Alaskan Pollock One UK market semi log regression analysis of product attributes, including labeling. Labeling, branding, processing and form all have significant effects on price. Marine Stewardship Council label o n Pollock demands an average 14.2% price premium. Ahmad and Anders (2012) Frozen Chicken and Seafood Two product genera (chicken/seafood) semi log regression analysis of attributes, including interaction terms. Species, brand, size, form and processing al l have significant effects on price. Interaction terms are necessary to identify multi dimensions of an attribute. Carew and Florkowski (2010) Burgundy Wine Two product (red/white wine) semi log regression analysis of wine attributes, including origin of production labeling. Includes quantity in the model. Origin of production labeling, alcohol content, vintage labeling and time all have an effect on wine prices. Negative relationship between quantity sold and price confirmed. Chang, Lusk and Norwood ( 2010) Retail Eggs Two market semi log regression analysis of product attributes, including labeling. Regions, production method, color, size, package, brands and seasonality have significant effects on price.

PAGE 21

21 Figure 1 1. Live w eight of t otal f isherie s p roducts s old in the U.S. a nnually (mil lbs), 1990 2009, by u se (Source: NOAA, 2010, Table 895) 0 5,000 10,000 15,000 20,000 25,000 Total Food Industrial

PAGE 22

22 Figure 1 2. Seafood m arket s hare by p roduct c ategory f orm in 2010 (Source: ACNielsen Scantrack) 40.75% 28.45% 12.15% 10.84% 7.76% 0.05% Unbreaded Shelf Stable Breaded Canned Entre Paste

PAGE 23

23 CHAPTER 2 D ATA Overview Table 2 1 provides a summary of the variables provided in the original datasets. By using nine variables provided in the original national data sets more than 30 variables are created for use in the models (i.e., finfish and shellfish) For example, q ualitative variables ar e coded into categories using dummy variables. Average market price is calculated using sales and volume figures and adjusted for general inflation This chapter first discussion the source of the data and how the original data were organized; the data were provided in very large datasets that were not in the format necessary to estimate basic hedonic models as intended. Then the use of exi st ing variables to create distinct measures of additional attributes is discussed. Lastly, the empirical models are specified and discussed. Source of Data The data was purchased from ACNielsen. ACNielsen is a global marketing research firm that is most known for the creation of Nielsen ratings, which measure television, radio and newspaper audiences. Another popular market resear ch tool provided by ACNielsen is the Homescan program in which household retail purchases are tracked by members of a sample group. Th e data used in this study are ACNielsen Scantrack data. Product sales are recorded by the register scanners at retail ou tlets and stored by ACNielsen The data were acquired from three types of retail outlets : (1) 2 million dollar grocery stores, (2) FDM, and (3) drug stores. This research used the data from the first category (grocery store chains in the U.S. ) which for seafood products is sufficient Table 2 2 provides a list of the store chains included in ACNielsen Scantrack data. Some of the stores included sell seafood products branded by their own private label. The private label

PAGE 24

24 brand s are known collectively as control brands by ACNielsen. Th ese data track products by UPC (Universal Product Code) It was discovered during this study that the number included in the dataset is technically an EAN (International Article Number) which is a superset of UPC. EAN cod es add an additional digit to the string of numbers, and indicate the country which the company selling the product is based. The code is referred to as a UPC in this study because that is a term most people are familiar with. Data Description Regional Ag gregation The 53 cities included in the data are: Albany, Atlanta, Baltimore, Birmingham, Boston, Buffalo/Rochester, Charlotte, Chicago, Cincinnati, Cleveland, Columbus, Dallas, Denver, Des Moines, Detroit, Grand Rapids, Hartford/New Haven, Hou ston, Indianapolis, Jacksonville, Kansas City, Las Vegas, Little Rock, Los Angeles, Louisville, Memphis, Miami, Milwaukee, Minneapolis, Nashville, New Orleans/Mobile, New York, Oklahoma City/Tulsa, Omaha, Orlando, Philadelphia, Phoenix, Pittsburgh, Portlan d, Raleigh/Durham, Richmond/Norfolk, Sacramento, Salt Lake City/Boise, San Antonio, San Diego, San Francisco, Seattle, St. Louis, Syracuse, Tampa, Total U.S., Washington D.C., and West Texas. Item and Brand Aggregation ACNielsen provided data aggregated by either item or brand. If item is chosen the data contain individual entries (rows) for the various products ( e.g., UPC) sold under the same brand name for a given product category. T c tl b is the value of this variable for products whic h are branded by the store which is selling them. For example, Publix, Winn Dixie and Kroger all sell a frozen salmon fillet branded as a Publix, Winn Dixie or Kroger product respectively

PAGE 25

25 brand s In the item aggregation, t he products may vary by species, size of package, or the form the fish is sold each having its own UPC These individual product sales are then aggregated from all the retail outlets in a given market. It was ve rified that each item was aggregated per market by performing a frequency procedure on the UPC variable. Each UPC had a frequency of one for each week of data. If brand aggregation is chosen the data contain only one entry (row) for all of the various products (UPCs) sold under the same brand name for a given product category. For c tl b aggregated to one entry (row) This option redu ces the richness of the data, and is not suitable for this study. However, this choice of aggregation may be suitable for studies which are interested in examining brands. Monthly and Weekly Aggregation The data for years 2006 and 2007 are aggregated on a monthly basis. The data for years 2008, 2009 and 2010 are aggregated weekly. The weekly data begins with the week ending June 17, 2007 and ends with the week ending June 12, 2010. In order to provide greater accuracy pertaining to the assignment of the situational determinants var iables only data in weekly format are used (i.e., the dataset contains 156 weeks of data) Product Categories The ACNielsen retail scanner dataset from the grocery stores is composed of 19 product categories as shown in Table 2 In 2010, the data set inc luded sales totaling 3.6 billion dollars. Shelf stable tuna is the largest product category by sales, accounting for 28% of the market share. The two product categories included in this analysis are Frozen Unbreaded Fish and Frozen Unbreaded Remaining. These product categories account for 12.3% and 1.8% of the market share,

PAGE 26

26 respectively. Figure 2 illustrates that in the aggregate, the Unbreaded Seafood product category has the largest market share in 2010. As previously explained, this study is interest ed in two of these categories : frozen unbreaded fish and frozen unbreaded remaining. The unbreaded remaining category contains shellfish such as clams, oysters and lobster s This category also includes a variety of other product types which do not fit in any of the other 19 categories, such as escargot and frogs which were removed from the dataset Although a separate category exists for crabs, a small percentage of the unbreaded remaining category included crab products. Upon investigation, these prod ucts were found to be crab and fish products combined (processed products) such as salmon pinwheels which were also removed from the dataset Data Manipulation In order to estimate a hedonic equation the dataset s had to be merged and transposed. The format and content of the original data s ets are summarized in Table 3. First, the quantitative sales data that are available for each product are the sum of sales in each market in each week. umber of units sold with and without associated promotional activities and the total dollar value and number of units sold (i.e., six spreadsheets for each year each containing one of these measures of sales). Each spreadsheet contains variables that des cribe the product (discussed below) and 52 variables (columns) that are the sales figures for each week. any advertisement, coupon and product display and/ or a price reduction of at least 5 perce nt below the suggested retail price by the producer ; in other words, include any promotional activities and not solely those associated with the concept of putting a good on sale riable (column) that indicates how each product ranked in terms of total sales dollars for the year After removing products with no sales

PAGE 27

27 during that year, the maximum value of the rank # variable is the total number of products (UPCs) sold that year in that product category. Each week of sales information was originally a separate variable. This information had to be associated with the week in which the data w ere collected (as is, the time information is embedded in the file and column titles). This was accomplished by transposing the column headings (i.e., week ending MM / DD / YY ) into variable name WEEK with values ranging from 1 to 52 in each year. Once the sheets were merged, this variable contained values from 1 156 to represent the continuous mea surement of sales across three years. All of the remaining variables (those that describe the product) had to be duplicated for each new weekly observation row. This was accomplished in SAS using the transpose procedure on the WEEK variables and selectin g the option product attribute variables. The variables provided by ACNiels e n which include product attribute information are: h l the same. However, if the brand name seen on the packaging is a subsidiary of a larger h level brand description variable. The Form variable describes the physical state, or cut, of the seafood product (e.g. f shellfish. Additional species information such as Pacific Tuna or Chilean Sea Bass, is included under Type. Size is a numerical variable representing the weight in ounces of the product, or the were r emoved from the analysis since they could not be converted to pounds for comparison

PAGE 28

28 with other products. The Type variable provides information regarding to a variety of product label s, which may reveal origin and processing information. A data input erro r was discovered in the finfish dataset by examining outliers. Three products with a package size of two ounces were removed from the dataset. One of the three was and discovered to actually be a two count p ackage. The other two, due to such an absurdly large price per pound, were assumed to be two count packages as well. It should be noted that the USDA measures a serving size of fish to three ounces, which supports this data error deletion. To reduce the that did not sell for at least one third of the time frame modeled (i.e., 52 weeks) we re removed from the dataset. This method has been previously suggested and applied by Roheim et al. (2007). Weeks with no sales information are indicated by a zero. After merging the sales sheets the data for each year and each product category (unbreaded fish and unbreaded remaining) two datasets were created: the finfish dataset and the shellfi sh dataset Variable s Create d for Analysis Each observation in the datasets contain qualitative variables that include the UPC, brand, product (species), form, size description and type; and quantitative variables on week, year, rank, size, dollars (sales), units (packages), promotio nal dollars (sales), and promotional units (packages). As explained earlier, f or a sale to qualify as promotional, it must either be recorded during a time of advertisement, with or without a price adjustment; or sold at an adjusted price (> 5%) without a dvertisement. Through calculations made with dollars, units, promotional dollars, promotional units and d. P rice is weighted by the proportion of sales made under promo tion. The price per pound is adjusted

PAGE 29

29 for inflation using the Consumer Price Index, base 2007. The dependent variable has a mean value of $ 8.8 3 per pound for shellfish and $ 6. 27 per pound for finfish. The m aximum adjusted weighted price per pound is $86 for shellfish and $148 for finfish. For a complete explanation on the development of the dependent variable, see Appendix A. S pecies group variables (e.g., salmon, tuna) were created from the product description on the dataset and are included in the mod els to correct for price influences hypothesized due to the distinct characteristics of species. Preliminary regression analysis revealed that r emoving the species groups has a dramatic negative effect on the explanatory power of the models. To further r educe noise in the data, and following previous studies using scanner data, the species grouping variable is limited to the inclusion of groupings which account for at least 1% of the market. Many of the species with less than 1% market share are accounte d for as a subgroup within a more dominant species group and, thus, were retained After these changes to the datasets, the finfish dataset contained 89,101 observations and the shellfish dataset contained 38,999 observations. The species variables in th e shellfish data are: SCAL, SQU, LOB, MUSS, CRAW, CLA and OCT which identifies the product as being a scallop, squid, lobster, mussel, crawfish, clam or octopus respectively Shrimp were not included in this dataset as ACNielsen provides all frozen unbreaded shrimp sales in a separate dataset. It should be noted that oysters were among the species dropped as the market share was less than 1%. The species variables in the finfish ile are: COD, SAL, TIL, CAT, WHIT, FLOU, POLL, PER, TUN, MAHI, SWOR, HADD, OR, HAL, SOL, SMEL and GROU which identify the product as being a cod, salmon, tilapia, catfish, whiting, founder, p ollock, perch, tuna, mahi mahi (aka dolphin or dorado), swordfish haddock, orange roughy, halibut, sole, smelt or grouper respectively

PAGE 30

30 The variable for the proportion of sales made under promotion, PRO, is a continuous number between 0 and 1. It is calculated using packages sold, to determine the proportion of pack ages sold under promotion each week A comparative chart of the 3 year average PRO variable is interesting (Figure 2 1 ). In the shellfish market, Christmas is the time of heaviest average proportion of promotional behavior. In the finfish market, Lent is the time of heaviest proportion of promotional behavior. Shellfish promotion in February peaks prior to finfish, when entrees that include lobster and other high end shellfish are popular Dumm y variables are created to capture these peaks to understand price influences during these times of promotional behavior change. This promotional variable is also interacted with the species group variables to test if promotional behavior varies by produc ts of different species group s The model time event variables are : OS, HOL_1 and HOL_2 which represent the sales made during the weeks following the Deepwater Horizon Oil Spill (which occurred in the Gulf of Mexico on April 20, 2010 and lasted into July 2010, past when this dataset stops) two weeks the two weeks prior to Christmas respectively The Oil Spill variable is interacted with the various species group variables to test if the oil spill affected species group s differently The brand variables, NB and SB identify products which are branded by name brand companies and private label companies, respectively. The inclusion of SB can be used to test whe ther US retailers own brand sell at a 10 40% discount to national name brands which was found by Halstead and Ward (1995) Roheim (2007) reports the opposite effect for the British own label brands.

PAGE 31

31 The ACNiels e n provided variable Type, which t he special attribute labeling variable s are created from The values of the variable, Type, reveal information pertaining to location of catch, imported or wild labeling or additional information specific to the species. The finfish model has a higher percentage of products without special attribute labeling ( 50% ) than the shellfish m odel ( 25% ) A r attributes on the labeling. The variable REGTYPE identifies all products which do not reveal any special attributes on the labeling, in both the shellfish and finfish models. In the shellfish model, t he variable SHEL indicates products which are identified as being in the s on the h alf s hell The variable BAY indicates products labeled It should be noted that BAY is exclusive to the scallop products and is necessary in the model to distinguish between higher valued s ea s callops The variable IMP is created to identify shellfish products that are labeled as i mported. In the shellfish m odel, 35.6 % product s were imported during the three year time horizon covered in this analysis This variable is also interacted with the species group variables to test if the imported label affected species groups differently. An analysis of companies selling the products were registered in foreign countries. This indicates that many of the products imported are actually sold by U.S. companies. In the finfish model, ALAS, PAC, ATL, BN, and SM are created to identify products labeled Alaskan, Pacific, Atlantic, boneless or smoked respectively Unfortunately, t he finfish data do not contain information on whether the product was import ed The finfish data do however, include products labeled as wild. Approximately 7% of the finfish model contains products with this label. The variable W ILD is created to identify products with this label. It is

PAGE 32

32 interacted with the species group variables to test if the wil d label affected species groups differently. The size variable, SIZE was created for those products that were identified as being measured in ounces. The size variable has a mean of 21 ounces for both shellfish and finfish. About one sixth of both market s are comprised of products with a size of ten ounces or less. This is important because these smaller package sizes in general yield a higher price per pound which was determined by examining at products with an adjusted weighted price per pound outside two standard deviations from the mean. For finfish these products cost greater than $19.47 adjusted weighted price per pound, for shellfish these products cost greater than $26.39 adjusted weighted price per pound. In the finfish market 92 percent of th ese outlier products are less than or equal to 10 ounces and in the shellfish market 85 percent of these outlier products are less than or equal to 10 ounces. Controlling for this through the use of a dummy variable was explored in previous models, and th ese smaller package sizes were found to have a significant increase in the price per pound of the products. However, in the final model this variable was dropped due to the fact that determining the relationship between size and price is not part of the r esearch question. Also, removing this variable improved the ease of interpretation. These outlier products are retained in the dataset and do not disrupt the explanatory power of the models.

PAGE 33

33 Table 2 1 Variables p rovided by ACNielsen Variable Descripti on Weekly Data by Year and Product (i.e., UPC): No Promo Sales Units Number of units sold (1) without advertisement s coupon s or display s, or (2) without price discount of at least 5% No Promo Sales Dollars Total dollar sales (1) without advertiseme nt s coupon s or display s, or (2) without price discount of at least 5% Any Promo Sales Unit Number of units sold (1) with advertisement s coupon s or display s, or (2) with price discount of at least 5% Any Promo Sales Dollars Total dollar sales (1) w ith advertisement s coupon s or display s, or (2) with price discount of at least 5% Unit Volume Total units sold ( No Promo Sales Units plus Any Promo Sales Units ) Sales Dollar Total dollar sales ( No Promo Sales Dollar s plus Any Promo Sales Dollar s) Description of Products: Universal Product Code A maximum of 12 digit number, determined to be the EAN code but without the verification digit Brand Description The product brand name. All private label store brands are identified with the value ctl b r. Table 2 2 provides a list of all the grocery store food chains included in ACNielsen Scantrack data. BE High Description The international, parent company of a particular brand. BE Low Description The national (United States) company of a particu lar brand. Form The form of the product, for example, fillet, steak, whole. Product Fish group or species of the product, for example, Tilapia, Crab, and Crawfish Style Attribute of product, specific to breaded and canned categories. For example, cr unchy, crispy. Size Numerical value for size of package (units) and r eported in either ounces or counts. Size Description Type of size count. Either ounces or count (data reported based on counts were not used) Type Attribute labeling of the product. For example, Smoked, Boneless, Skinless, Wild, Alaskan, Atlantic. A value of egular is interpreted as having no attribute labeling.

PAGE 34

34 Table 2 2. List of grocery food chains covered by ACNielsen store level data Grocery Food Chains KROGER FRYS F OOD LION PRICE CHOPPER PUBLIX SWEETBAY SAFEWAY LOWES ALBERTSONS (SV) MARSH WINN DIXIE KING SOOPER SAVE A LOT CORPORA SCHNUCK MARKETS IN PIGGLY WIGGLY CARO FOOD CITY KVAT STOP & SHOP FRESH BRANDS/PIGGL VONS PICK N SAVE (CORP HEB A&P RALPHS GROCERY BASHAS HOUCHENS/SAVE A LO HOMELAND ALBERTSONS (Cerb) DILLON/GERBES/SAV BI LO RALEYS FOOD & DRUG GIANT EAGLE INC CUB FOODS SHOP RITE/WAKEFERN DOMINICKS HYVEE SUPER FRESH INGLES TOPS SHAWS SUPERMARKETS QUALITY MEIJER PIGGLY WIGGLY/STOR GIANT (MD) WEGMANS JEWEL OSCO LUCKY STORES HARRIS TEETER INC HARVEYS SUPERMARKE HANNAFORD/SHOP N S PIGGLY WIGGLY STATER BROS MARKET BRUNOS/FOOD MAX/FO BROOKSHIRE SHOPPERS FOOD/METR WEIS WALDBAUM INC GIANT (CARLISLE) TOM THUMB FOOD 4 LESS BASHAS FOOD CITY PAT HMARK DEMOULAS/MARKET BA SAVE MART NASH FINCH/ECONO/S SMITHS BIG Y FOOD EMPORIUM STRACK & VAN TIL LOWES/PAY N SAVE RAINBOW (ROUNDY'S) PRICE CHOPPER JAY C STORE/FOOD P P & C UKROPS NIEMANN FOODS/COUN SUPERIOR SUPER WAR HARPS COPPS CO

PAGE 35

35 Table 2 2. C ontinued Grocery Food Chains FIESTA MART INC KINGS KING KULLEN SUPER ONE RANDALLS QUALITY MARKETS FARM FRESH FAMILY FARE SUPER S/MEGA G&W FOODS/FARMERS UNITED HARVEST FOODS COUNTRY MART E W JAMES & SONS S FOOD MAXX STORES DAVIDS BUEHLER FOODS I NC NOB HILL SOUTHERN FAMILY MA DIERBERGS BIG M BOYERS IGA INC REDNERS CARNIVAL FOOD STOR CITY MARKET BILO GENUARDI/MAD GROCE MORGANS HOLIDAY MA GRISTEDES KNOWLANS SUPERMKTS COBORNS/CASH WISE ROSAUERS GREERS/FOOD TIGER MINYARD SENTRY/SUPER SAVER MA RTINS SAVE A LOT FOODS LUNDS INC G U MARKETS MARVINS IGA INC PRICE RITE BEL AIR MARKETS GLENS MARKETS INC PIONEER/MET FD/ASSOC RAMEY SUPER MARKET LAWRENCE IGA TOP FOOD/HAGGENS ACME MARKETS ROUSES FRED MEYER INC HARDINGS SHOP N SAVE (SUPER MARKET BASKET/LUCK RAYS FOOD PLACE FULMER SUPERMARKET Note: These grocery food chains are the source of the data collected as ACNielsen Scantrack data if the store has sales greater than 2 million annually. Some of the retailers listed sell store brand prod ucts, known collectively in this data set as the control brand ( ACNielsen ctl br

PAGE 36

36 Table 2 3. Total U S s eafood s ales and market share at g rocery s tore s that exceeded $2 million in 2010 by c ategory Product Category 2010 Sales ($) SEAFOOD TUNA SHELF STABLE 1,031,047,472 SEAFOOD SHRIMP UNBREADED FROZEN 925,556,370 SEAFOOD FISH UNBREADED FROZEN 445,362,426 SEAFOOD FISH BREADED FROZEN 339,261,224 ENTREES SEAFOOD 1 FOOD FROZEN 235,815,831 SEAFOOD SALMON CANNED 130,154,630 SEAFOOD SHRIMP BREADED FROZEN 94,843,782 SEAFOOD S ARDINES CANNED 83,868,140 SEAFOOD REMAINING UNBREADED FROZEN 65,458,664 SEAFOOD REMAINING CANNED 49,592,763 ENTREES SEAFOOD 2 FOOD FROZEN 45,439,902 SEAFOOD CRAB UNBREADED FROZEN 40,815,006 SEAFOOD CLAMS CANNED 40,032,917 SEAFOOD OYSTERS CANNED 32,183,084 SEAFOOD CRAB CANNED 24,240,632 SEAFOOD SHRIMP CANNED 17,189,272 SEAFOOD ANCHOVIES 15,706,391 SEAFOOD REMAINING BREADED FROZEN 6,184,956 ANCHOVY PASTE 1,950,890 Total 3,624,704,352 Source: A.C. Neilson New Yor k City, NY.

PAGE 37

37 Figure 2 1. Proportion of p romotional s ales for s hellfish and f infish by w eek 3 year average 0% 5% 10% 15% 20% 25% 30% 35% 21-Jun 21-Jul 21-Aug 21-Sep 21-Oct 21-Nov 21-Dec 21-Jan 21-Feb 21-Mar 21-Apr 21-May Shellfish Finfish

PAGE 38

38 CHAPTER 3 E MPIRICAL ANALYSIS Overview Implicit price estimation and model specification methods are discussed. Results are reported for bot h finfish and shellfish models. Comparisons between the implicit prices of intangible attributes are discussed. Interaction terms are used to further understand the effects of promotional, labeling and situational attributes at the individual product spe cies group level. Because these results are specific to shellfish and finfish, the results are reported separately. Estimation and Model Specification Hedonic regression analysis is used to calculate the implicit prices of product, promotional situationa l and labeling attributes. An individual model is created for finfish and shellfish due to different values on certain variables. For example, in the shellfish model, the type variable contains information that identifies a product as imported. This var iable does not contain this information in the finfish model, so combining these data would implicitly assume that all of the finfish were domestically produced. This would be an incorrect assumption and result in mis specification of the model Initially, regression analysis wa s performed in SAS using the PROC REG procedure. Both heteroscedastici t y and autocorrelation we re identified as issues with the Ordinary Least Squares (OLS) regression which assumes the estimates are BLUE; BLUE refers to the best, l inear, unbiased estimators. Homoscedasticity is a n assumption of the OLS regression that the variances remain consistent across all values of a variable. By looking at the plots of the residuals, it was determined that this implied assumption was incorre ct since the residual variances appeared to be correlated with the SIZE variable (i.e., the size of the package in ounces). This may be due to inconsistencies in product (species) size offerings as well as

PAGE 39

39 increased price variability in smaller package si zes. For example, the smaller package size has at that size package (e.g., smaller packages had a higher share of expensive processed products such as smok ed product ). Autocorrelation was also detected as an issue with the OLS regression. The Durbin Watson (DW) statistic in both the finfish and shellfish models indicated significantly high positive autocorrelation (DW=0.15). As a result, models were estimate d that corrected for either 1 autocorrelation or heteroscedasticity. It was determined that the model correcting for autocorrelation was superior because the problem existed in all the variables, whereas heteroscedasticity was only an issue with the size v ariable. Correcting for autocorrelation increased the explanatory power of the model by increasing the number of significant parameter estimates. The following equation is the shellfish model: P it 0 + 1 ( D SQU ) + 2 ( D LOB ) + 3 ( D MUSS ) + 4 ( D CRAW ) + 5 ( D C LA ) + 6 ( D OCT ) + 7 (PRO)( D SCAL ) + 8 (PRO)( D SQU ) + 9 (PRO)( D LOB ) + 10 (PRO)( D MUSS ) + 11 (PRO)( D CRAW ) + 12 (PRO)( D CLA ) + 13 (PRO)( D OCT ) + 14 ( D OS )( D SCAL ) + 15 ( D OS )( D SQU ) + 16 ( D OS )( D LOB ) + 17 ( D OS )( D MUSS ) + 18 ( D OS )( D CRAW ) + 19 ( D OS )( D CLA ) + 20 ( D OS )( D OCT ) + 21 ( D IMP )( D SCAL ) + 22 ( D IMP )( D SQU ) + 23 ( D IMP )( D LOB ) + 24 ( D IMP )( D MUSS ) + 25 ( D IMP )( D CLA ) + 26 ( D IMP )( D OCT ) + 27 ( D HOL_1 ) + 28 ( D HOL_ 2 ) + 29 ( D SB ) + 30 ( D BAY ) + 31 ( D SHEL ) + 32 ( SIZE ) + 33 ( D PIEC it The following equation is the finfish model: P it = 0 + 1 ( D SAL ) + 2 ( D TIL ) + 3 ( D CAT ) + 4 ( D WHIT ) + 5 ( D FLOU ) + 6 ( D POLL ) + 7 ( D PER ) + 8 ( D TUN ) + 9 ( D MAHI ) + 10 ( D SWOR ) + 11 ( D HADD ) + 12 ( D OR ) + 13 ( D HAL ) + 14 ( D SOL ) + 15 ( D SMEL ) + 16 ( D GROU ) + 17 ( PRO )( D COD ) + 18 ( PRO )( D SAL ) + 19 ( PRO )( D TIL ) + 20 ( PRO )( D CAT ) + 21 ( PRO )( D WHIT ) + 22 ( PRO )( D FLOU ) + 23 (PRO)( D POLL ) + 24 (PRO)( D PER ) + 25 (PRO)( D TUN ) + 26 (PRO)( D MAHI ) + 27 (PRO)( D SWOR ) + 28 (PRO)( D HADD ) + 29 (PRO)( D OR ) + 30 (PRO)( D HAL ) + 1 GARCH parameter estimates which correct for both heteroscedasticity and autocorrelation simultaneously were exam ined. However, these estimates w ere not used because this procedure is commonly used in forecasting and square value was a concern (0.9950).

PAGE 40

40 31 (PRO)( D SOL ) + 32 (PRO)( D SMEL ) + 33 (PRO)( D GROU ) + 34 ( D OS )( D COD ) + 35 ( D OS )( D SAL ) + 36 ( D OS )( D TIL ) + 37 ( D OS )( D CAT ) + 38 ( D OS )( D WHIT ) + 39 ( D OS )( D FLOU ) + 40 ( D OS )( D POLL ) + 41 ( D OS )( D PER ) + 42 ( D OS )( D TUN ) + 43 ( D OS )( D MAHI ) + 44 ( D OS )( D SWOR ) + 45 ( D OS )( D HADD ) + 46 ( D OS )( D OR ) + 47 ( D OS )( D HAL ) + 48 ( D OS )( D SOL ) + 49 ( D OS )( D SMEL ) + 50 ( D OS )( D GROU ) + 51 ( D W )( D COD ) + 52 ( D W )( D SAL ) + 53 ( D W )( D FLOU ) + 54 ( D W )( D POLL ) + 55 ( D W )( D PER ) + 56 ( D W )( D TUN ) + 57 ( D W )( D MAHI ) + 58 ( D W )( D SWOR ) + 59 ( D W )( D OR ) + 60 ( D W )( D HAL ) + 61 ( D W )( D SOL ) + 62 ( D HOL_1 ) + 63 ( D HOL_ 2 ) + 64 ( D SB ) + 65 ( D ALAS ) + 66 ( D PAC ) + 67 ( D ATL ) + 68 ( D BN ) + 69 ( D SM ) + 70 ( SIZE ) + 71 ( D PIEC ) + 72 ( D STEA ) + 73 ( D WHOL ) + 74 ( D REMF ) + it Previous studies have examined the possibility of a non linear relationship between price and package size While this study does not claim that a linear relationship is the best fit, the models performed well (as will be discussed below) and interpretation of the coefficients is straight forward. The base product ( 0 or INT ) in the finfish m odel is a whole, name brand c od product lacking any spec ial attribute labeling and not sold under promotion or during any of the situational influence time periods. The base product parameter estimate (the intercept) is interpreted as a product with a package size equal to zero. The base product in the shellf is h model is a whole, name brand s callop lacking any special attribute s and labeling and not sold periods. The base product parameter estimate ( the intercept) is interpreted as a product wit h a package size equal to zero. The base product parameter can be adjusted to reflect a product of different package sizes by summing with the product of the SIZE parameter estimate and the number of ounces in a package (Appendix A). The explicit models specified above were estimated using PROC AUTOREG in SAS and results for the shellfish and fin fish models are shown in Tables 3 1 and 3 3 respectively. More detailed results for each model are discussed in turn. Shellfish Results The shellfish model is c omposed of 38,999 observations with a r square value of 0.7 884 The r square value indicates that 79 percent of the variation in price is explained by the

PAGE 41

41 independent variables. Twenty seven of the thirty four variables are significant at the 9 0 % level o f confidence or greater. The intercept value is 11.44 in the shellfish model This is the predicted price, without adjusting for size, for the base product cod. The predicted price, without adjusting for size, for the other species groups can be determin ed by taking the sum of the intercept value and the corresponding species group parameter estimate. The predicted price used in determining the magnitude of price effects is adjusted for a package size of 16 oz. which equates to a discount of $1.26 for t he shellfish products This eases interpretation as price is reported in dollars per pound. T he magnitude of the promotional, situational and labeling interaction term parameter estimates is calculated in relation to the predicted price for each species group. Table 3 1 contains a summary of the model results and Table 3 2 contains the resulting predicted prices and magnitudes of change for key variables in the shellfish model. Premiums for Species With the exception of crawfish, a ll of the species group variables have parameter estimates significant at the 99% level Crawfish is significant at the 95% level. The predicted price for SCAL, SQU, LOB, CRAW, MUSS, CLA M and OCT are $10.18, $5.76, $22.02, $9.52, $4.91, $6.71 and $3.36, respectively. These pr edicted prices are a national average of what may be s ee n when visiting the frozen food aisle of a grocery store. Lobster is the most highly valued species ; with a predicted price more than double th at of scallops. Scallops and crawfish both have a predi b ay scallops. Therefore, the predicted price for SCAL is assumed to be for higher valued s ea s callops. Squid, octopus and mussels are species with the lower predicted prices which is what we would expect to find These results increase ones confidence in the model.

PAGE 42

42 Implied Value of Promotional Activities The proportion of promotional sales variable, PRO, has a negative influence on price for all species groups All of the parameter estimates are significant at the 99% level The magnitude of promotional price change with respect to the species group predicted value for SCAL, SQU, LOB, CRAW, MUSS, CLA and OCT are 21%, 16%, 27%, 15%, 16%, 18% and 30%, respectively ( Fi gure 3 1 ). The results are interesting in that the marginal change of price for products sold 100% under promotion does not correlate with the product market share or underlying predicted price. For example, both lobster and octopus have the highest prom otional price cuts near 30%, yet lobster s are highly valued while octopus is of low value. Octopus products hold the smallest market share, so large price cuts may be explained by producers trying to increase market share. Many of the weekly observations had PRO values < 1. This is explained by retailers and producers implementing different promotional and non promotional strategies in different stores or regions during the same time period. Price Effects from the Deepwater Horizon Oil Spill T he oil spil l variable, OS, had both a positive and negative influence on price depending on the underlying species group Three of the seven parameter estimates are significant. The magnitude of price change after the oil spill with respect to the species group pr edicted value for SCAL, LOB and MUSS are 3.9%, 3.7% and 7.3% respectively (Figure 3 2) These results only cover six weeks following the start of the oil spill, so these price changes are initial reactions to the news. Clams, squid, crawfish, and octo pus did not have a significant price change. These four species groups are either (1) not saltwater shellfish or (2) species of lesser value. The two species with price cuts, both scallops and lobsters, are highly valued saltwater species and consumers m ay have started slowing their purchasing of the more expensive saltwater seafood species. Another thought is that these products, being of higher value, may

PAGE 43

43 have larger margins from which to make a price cut. The producers of the lesser valued products m ay not have had the room in their margins to make a 4% price cut, and may have looked towards other methods of increasing demand (e.g., promotion). Mussels are the species group of shellfish with the highest share of imported products (70%), which are mai nly from New Zealand. The price may have been increased on mussel products because consumers were aware that they were coming from waters other than the Gulf of Mexico, and were willing to pay a higher price. Relative Value of Imported Shellfish The impor ted species labeling variable has both a positive and negative influence on price. Five of the six parameter estimates are significant. The magnitude of price change for labeling the product imported with respect to the species group predicted value for SCAL, SQU, LOB, MUSS and CLA are 11%, 34%, 6%, 15% and 14% respectively (Figure 3 3) Lobsters were the only species with a positive effect for labeling seafood products as imported This may be a result of the different types of lobsters harvested outside of U.S. waters. Domestically produced lobsters may be less valued in the U.S. market. Imported squid is approximately one third cheaper than squid products not labeled as imported. Foreign production and processing of this species must be consid erably cheaper for a discount of this magnitude to show up in the retail price. The remaining three significant changes are scallops, mussels and clams. These products are all between 10 15% cheaper when labeled as imported, which may be impacting the sa les of domestically produced or processed products. Crawfish are not sold as imported and octopus labeled as imported was not significantly different than octopus without the label. Other Implied Prices for Shellfish The parameter estimate for HOL_1, whic h covers the two weeks in mid February, is not significant This means that on average, the price of products sold during this time were not

PAGE 44

44 significantly different than the price of products sold during other times of the year. Interaction terms between this holiday variable and species may reveal more about shellfish pricing during this time of the year. The parameter estimate for HOL_2 which covers the two weeks leading up to Christmas, is also not significant in the shellfish model. This variable co vered the time period of highest promotion for the shellfish model. The branding variable, SB, has a negative influence on price. Products sold under the store brand label were on average $3.37 cheaper than the name brand products. Store brand products a re generally thought to be cheaper in price, so this result supports our hypothesis and what was previously discovered in the literature pertaining to fish. S tore brand products market share is 8%. Additionally, n ot all species groups are sold as store brand products. Scallop, squid, lobster and mussel account for 69%, 4%, 24%, and 3% of the store brand products market share (Table 3 3, Figure 3 4 and Figure 3 5) The other labeling variables, BAY and SHEL have the following significant parameter estim ates: 2.33 and 1.63. b and are typically cheaper than sea scallops. The result confirms this hypothesis. SHEL is hypothesized to be negative According to Ahmad and Anders (2012), p roduct forms and processes which add convenience to the consumer have a positive influence on price. Therefore, products lacking convenience, and requiring additional processing prior to consumption, are expected to have less value. The result confirms t his hypothesis. The size variable has a negative influence. The parameter estimates is 0.0 79 in the shellfish model. For every ounce of package size, the average price per pound decreases by

PAGE 45

45 $0.08 literature, as well as in the market place. Consumers expect to pay less per pound if the purchase is made in bulk. The product form variables in the shellfish model are either WHOL (base) or PIEC As expected, consumers on average pay more for product forms which are more convenient and/or ready for consumption (Ahmad and Anders, 2012) The parameter estimates for PIEC is 1.395, supporting this claim. Finfish Results The finfish model is composed of 89,101 observations with a r square value of 0. 6470. The r square value indicates that 65% of the variation in price is explained by the independent variables included in the model. Sixty one of the seventy five variable parameter estimates are significant at the 9 0 % level Table 3 4 contains a summary of the model results and Table 3 5 contains the resulting predicted prices and magnitudes of change for key variables in the finfish model. The intercept value in the finfish model is 8.208 As in the shellfish results, the magnitude of the promotional, sit uational and labeling interaction term parameter estimates is calculated in relation to the predicted price for each species group. The predicted price used in determining the magnitude of price effects is adjusted for a package size of 16 oz. which equa tes to a discount of $0.88 for the finfish products. Premiums for Species All of the species group variables have parameter estimates significant at the 99% level besides grouper and tuna. Grouper and tuna are significant at the 90% level. The p redicted prices for COD, SAL, TIL, CAT, WHIT, FLOU, POLL, PER, TUN, MAHI, SWOR, HADD, OR, HAL, SOL, SMEL and GROU are $7.33, $8.78, $5.46, $4.80, $3.90, $5.93, $4.42, $6.24, $7.64, $6.86, $8.01, $6.22, $10.20, $12.95, $5.92, $3.58 and $6.91 respectively. The resu lts are

PAGE 46

46 what one might expect to find, with catfish, whiting, Pollock and smelt being the less valued species; and cod, salmon, tuna, swordfish, orange roughy, halibut and grouper as the more valued species. Halibut is the most valued species. As in the shellfish premiums for species, these results provide for more confidence in the model. Implied Value of Promotional Activities The proportion of promotional sales variable, PRO, has a negative influence on price for all species groups All of the paramet er estimates are significant The magnitude of promotional price change with respect to the species group predicted value for COD, SAL, TIL, CAT, WHIT, FLOU, POLL, PER, TUN, MAHI, SWOR, HADD, OR, HAL, SOL, SMEL and GROU are 20%, 20%, 22%, 18%, 12%, 21%, 18%, 21%, 25%, 28%, 24%, 19%, 21%, 17%, 17%, 14% and 29% respectively (Figure 3 6 ) A surprising similarity to the results from the shellfish model is the species with the highest promotional cut is the species with the lowest market shar e (Grouper = 30%) Again, the producers of these products may be trying to increase their market share by making large price cuts. The lowest promotional price cuts are made by species with lower predicted prices. These products may not have the room i n their margins to make large price cuts. One difference from the shellfish model is the higher valued species as a group only made price cuts between 17 and 22%. Overall, finfish promotional price cuts range from 12% to 30% and shellfish promotional price cuts range from 15% to 30%. The inclusion of interaction terms between species and promotional price change is necessary to understanding how promotions change between products. Price Effects from the Deepwater Horizon Oil Spill The oil spill var iable, OS, ha d a generally positive influence on price. Nine of the seventeen parameter estimates were statistically significant. The magnitude of the price change after the oil spill with respect to the species group predicted premiums for SAL, TIL, CAT

PAGE 47

47 FLOU, SWOR, HADD, OR, HAL and SOL of 3%, 4%, 4%, 4%, 3%, 7%, 5%, 7% and 10% respectively (Figure 3 7 ) It was hypothesized that Gulf species would experience a price cut, in order to increase demand of a product consumers may be concerned with eating. However, the results show that for Gulf species, the price did not change significantly. All of the species with significant price changes are either fresh water, or not harvested in the Gulf. The producers which rely of the Gulf of Mexico for harvest were possibly stockpiling their catch, due to the uncertainty of future harvest. Because this data only covers six weeks into the spill, a price adjustment direction may not have been determined at that time. If supply was going to get low, prices should increase ceterus paribus What is interesting is that producers of products not related to the Gulf of Mexico made increasing price adjustment rather soon after the news of the oil spill. Relative Value of Wild Labeled Fin fish The wild species labeling variable has a generally positive correlation with retail price. Eight of the seventeen parameter estimates were statistically significant. The magnitude of price change for describing the product wild with respect to the species group predicted va lue s for COD, SAL, PERC, MAHI, SWOR, OR, HAL and SOL were found to be higher by 10%, 10%, 41%, 44%, 32%, 44%, 6% and 33% respectively (Figure 3 8 ) Interestingly, many of these species such as swordfish and orange roughy are only harvested in the wild. These results show that consumers may not be aware of the production methods currently available for a particular species, and pay a higher price for the label, regardless that alternatives without the label are also harvested wild. The perch, mahi mahi, swordfish, orange roughy and sole results are surprising ; t hese products received nearly a 50% premium for including the wild descriptor on the packaging. Flounder, Pollock and tuna labeled wild were not significantly different (on a

PAGE 48

48 statistical basis ) than non wild labeled products. Catfish, tilapia, whiting, haddock, smelt and grouper did not have the wild label on any products. Other Implied Prices for Fin fish The parameter estimate for HOL_1, which covers the two weeks in mid February, is 0.029 ( approximately $0.03 per pound) and statistically significant This means that on average, products sold during this time were about three cents cheaper than during other times of the year. While this result is rather small, it was statistically differen t than the prices during other times of the year. It was hypothesized that products are cheaper during this time of year because it is the onset of Lent, a holiday in which fish is a popular substitute for red meat The promotional chart (Figure 2 1 ) sup ports the claim that this is an important time in fish marketing. This variable may be better explained by interaction terms between this holiday and the species that was beyond the scope of work for this study The parameter estimate for HOL_2 which cov ers the two weeks leading up to Christmas, was not statistically significant in the finfish model. As reasoned with the shellfish model price changes during this time may be better explained by examining interaction terms between the holiday variable and species that were beyond this study The branding variable, SB, indicated that store brands were positively correlated with price per pound Products sold under the store brand label were on average $0.37 more expensive than name brand products. This is not what we would expect to find, and is different than what is reported in the literature (Roheim et al., 2007) S tore branding is associated with higher valued species in the finfish market for example, catfish products make up 11% of name brand pro ducts and 4% of store brand products (Table 3 6, Figure 3 9 and Figure 3 10) I nteraction terms could be used to pull apart and better estimate the correlations between branding and price.

PAGE 49

49 The other labeling variables PAC ) ATL ) BN ) and SM ) have the following statistically significant correlations with price paid per pound : $ 1.31, $ 0.99, $ 0.48 and $ 8.99 respectively The parameter estimate for ( ALAS ) is not statistically significant. Atlantic l abeled products are of higher value than non labeled products. It is hypothesized that any additional information on the label should increase the price ; otherwise producers would leave that information off of the label. This is not true in the case of P acific labeled products which receive d a lower value than non lab eled products. Smoking fish products increased the retail price considerably which likely reflects that fact that smoking fish is a timely process and alters the flavor distinctively. Thi s research indicated that consumers want this attribute and paid a considerable premium for fish products that are smoked The size variable has a negative correlation with price per pound for the primary frozen unbreaded fish and shellfish products sold i n the U.S. market For every ounce of package size, the price of the product decreases by six cents (i.e., $0.055) on average which supports the strategy of discounted bulk sales. The product form variables are more extensive in the finfish model than the shellfish. As expected, consumers on average pay more for product forms which are more processed and contain less waste. The parameter estimates for PIEC, WHOL and REMFORM indicated the following prices changes: $ 0.76, $ 0.96 and $ 3.45 per pound r espectively The parameter estimate for STEA not significant ly different from the base of FILL Products that contain a variety of parts ground together (e.g., burgers and loafs) sold at a considerable price discount.

PAGE 50

50 Table 3 1. Shellfish v ariable m eans and h edonic m odel r esults Model r esults Variable Description Mean PE SE t Value p Value DV Price a djusted by CPI ( b ase 2007), $/lb. 8.825 INT Intercept 11.437 *** 0.293 38.98 <.0001 Shellfish Species Group SCAL = 1 if S callop ; = 0 otherwise 0.272 BASE SQU = 1 if S quid ; = 0 otherwise 0.218 4.420 *** 0.273 16.18 <.0001 LOB = 1 if L obster ; = 0 otherwise 0.156 11.840 *** 0.313 37.84 <.0001 CRAW = 1 if C rawfish ; = 0 otherwise 0.139 0.661 ** 0.313 2.11 0.0346 MUSS = 1 if M ussel ; = 0 otherwise 0.109 5.270 *** 0.361 14.60 <.0001 CLA M = 1 if C lam ; = 0 otherwise 0.077 3.472 *** 0.381 9.12 <.0001 OCT = 1 if O ctopus ; = 0 otherwise 0.028 6.816 *** 0.543 12.56 <.0001 Promotional Interaction Variables (PRO = Promotional Units Sold/Total Units Sold ; = 0 otherwise ) PRO_SCAL = PRO Scallop 0.054 2.118 *** 0.041 52.31 <.0001 PRO_SQU = PRO Squid 0.012 0.939 *** 0.075 12.58 <.0001 PRO_LOB = PRO Lobster 0.018 6. 030 *** 0.063 95.56 <.0001 PRO_CRAW = PRO Crawfish 0.007 1.447 *** 0.113 12.79 <.0001 PRO_MUSS = PRO Mussel 0.009 0.776 *** 0.099 7.82 <.0001 PRO_CLA M = PRO Clam 0.006 1.213 *** 0.113 10.69 <.0001 PRO_OCT = PRO Octopus 0.002 0.999 *** 0.221 4.52 <.0001 Oil Spill Interaction Variables (OS = 1 if sold after 4/20/10 ; = 0 otherwise ) OS_SCAL = OS Scallop 0.012 0.397 *** 0.102 3.92 <.0001 OS_SQU = OS Squid 0.010 0.139 0.108 1.29 0.1980 OS_LOB = OS Lobster 0.007 0.817 *** 0.137 5. 97 <.0001 OS_CRAW = OS Crawfish 0.005 0.302 0.196 1.54 0.1236 OS_MUSS = OS Mussel 0.005 0.358 ** 0.169 2.12 0.0337 OS_CLA M = OS Clam 0.004 0.191 0.198 0.96 0.3347 OS_OCT = OS Octopus 0.001 0.435 0.328 1.32 0.1852 Import Label Interaction Variab ; = 0 otherwise ) IMP_SCAL = IMP Scallop 0.080 1.124 *** 0.243 4.62 <.0001 IMP_SQU = IMP Squid 0.078 1.961 *** 0.264 7.43 <.0001

PAGE 51

51 Table 3 1. Continued Model r esults Variable Description Mean PE SE t Value p Value IMP_LOB = IMP Lobster 0.067 1.306 *** 0.314 4.16 <.0001 IMP_MUSS = IMP Mussel 0.076 0.752 ** 0.343 2.19 0.0284 IMP_CLA M = IMP Clam 0.041 0.941 ** 0.414 2.27 0.0231 IMP_OCT = IMP Octopus 0.014 0.002 0.672 0.00 0.9973 Time Even t Variables HOL_1 = 1 for 2 weeks in mid February ; = 0 otherwise 0.038 0.003 0.028 0.12 0.906 HOL_2 = 1 for 2 weeks before Christmas ; = 0 otherwise 0.038 0.013 0.028 0.45 0.6542 Branding Variables NB = 1 if national retail brand ; = 0 oth erwise 0.924 BASE SB = 1 if private label brand ; = 0 otherwise 0.076 3.372 *** 0.561 6.01 <.0001 Labeling Variables REGTYPE = 1 if no attribute labeling ; = 0 otherwise 0.269 BASE BAY b ; = 0 otherwise 0.078 2.333 ** 0.265 8.81 <.0001 SHEL s ; = 0 otherwise 1.634 *** 0.286 5.72 <.0001 Package Size Variable SIZE Size of package, ounces 21.191 0.079 *** 0.004 18.30 <.0001 Product Form Variables W HOL = 1 if w hole ; = 0 other wise 0.553 BASE P IEC = 1 if piece, ring, claw, chunk, cut; = 0 otherwise 0.447 1.395 *** 0.163 8.55 <.0001 Model Statistics Observations 38,999 Total R Square 0. 7884 Pr > ChiSq <0.0001 Durbin Watson 1. 9994 Note: Lev el of Significance = 99%***, 95%**, 90%*. Intercept = ase product of whole s callop, not imported, not on promotion, not sold after the Oil Spill or during the two holiday periods. In addition, t he base product has no attribute labeling an d is sold under a national retail brand. The package size is 0 oz.

PAGE 52

52 Table 3 2 Shellfish predicted price and interaction effects Percent c hange of p redicted p rice for p romotional, s ituational and l abeling a ttribute e ffects Species Predicted p rice PRO OS IMP SCALL $10.18 20.81% 3.90% 11.04% SQU $5.76 16.31% NS 34.05% L OB $22.02 27.39% 3.71% 5.93% C RAW $9.52 15.20% NS Not i mported M USS $4.91 15.80% 7.29% 15.32% C LAM $6.71 18.08% NS 14.03% OCT $3.36 29.69% NS NS Note: P redicted price s adjusted for size (16 oz.) NS = Parameter not significant ( p > 0.10).

PAGE 53

53 Table 3 3 Frequencies of name brand and store brand shellfish products by species group Species Name brand frequency Store brand frequency Name brand percentage St ore brand percentage Difference in percentage SCAL 11648 2808 23.7 69.2 45.5 SQU 11440 156 23.3 3.9 19.4 LOB 7332 988 14.9 24.4 9.4 CRAW 7384 15.0 15.0 MUSS 5720 104 11.6 2.6 9.1 CLAM 4108 8.4 8.4 OCT 1508 3.1 3.1 SUM 49140 4056 10 0 100 0.0 Note: Crawfish, Clam and Octopus product not sold under store branding.

PAGE 54

54 Table 3 4 Finfish v ariable m eans and h edonic m odel r esults Model Results Variable Description Mean PE SE t Value p Value DV Price a djusted by CPI ( b ase 2007), $/ lb. 6.267 INT Intercept 8.208 *** 0.124 66.43 <.0001 Fish Species Group COD =1 if Cod; = 0 otherwise 0.073 BASE SAL =1 if Salmon ; = 0 otherwise 0.171 1.442 *** 0.124 11.62 <.0001 TIL =1 if Tilapia ; = 0 otherwise 0.137 1.875 *** 0.122 15. 35 <.0001 CAT =1 if Catfish ; = 0 otherwise 0.094 2.529 *** 0.134 18.90 <.0001 WHIT =1 if Whiting ; = 0 otherwise 0.076 3.432 *** 0.133 25.76 <.0001 FLOU =1 if Flounder ; = 0 otherwise 0.072 1.407 *** 0.130 10.86 <.0001 POLL =1 if Pollock ; = 0 otherwis e 0.069 2.917 *** 0.140 20.92 <.0001 PER =1 if Perch ; = 0 otherwise 0.057 1.097 *** 0.136 8.04 <.0001 TUN =1 if Tuna ; = 0 otherwise 0.042 0.310 0.185 1.67 0.0948 MAHI =1 if Mahi Mahi ; = 0 otherwise 0.038 0.470 *** 0.168 2.79 0.0052 SWOR =1 if Sword fish ; = 0 otherwise 0.038 0.674 *** 0.220 3.07 0.0022 HADD =1 if Haddock ; = 0 otherwise 0.031 1.117 *** 0.154 7.26 <.0001 OR =1 if Orange Roughy ; = 0 otherwise 0.029 2.862 *** 0.163 17.53 <.0001 HAL =1 if Halibut ; = 0 otherwise 0.027 5.612 *** 0.198 28.33 <.0001 SOL =1 if Sole ; = 0 otherwise 0.021 1.410 *** 0.194 7.27 <.0001 SMEL =1 if Smelt ; = 0 otherwise 0.013 3.751 *** 0.325 11.55 <.0001 GROU =1 if Grouper ; = 0 otherwise 0.012 0.425 0.224 1.90 0.0576 Promotional Interaction Variables (PRO=Promo tional Units Sold/Total Units Sold; = 0 otherwise) PRO_COD = PRO Cod 0.020 1.490 *** 0.030 50.44 <.0001 PRO_SAL = PRO Salmon 0.045 1.731 *** 0.020 88.69 <.0001 PRO_TIL = PRO Tilapia 0.040 1.194 *** 0.020 60.93 <.0001 PRO_CAT = PRO Catfish 0.016 0.888 *** 0.029 30.39 <.0001 PRO_WHIT = PRO Whiting 0.013 0.454 *** 0.032 14.15 <.0001 PRO_FLOU = PRO Flounder 0.017 1.258 *** 0.029 42.93 <.0001 PRO_POLL = PRO Pollock 0.013 0.781 *** 0.035 22.54 <.0001 PRO_PERC = PRO Perch 0.014 1. 329 *** 0.033 39.78 <.0001

PAGE 55

55 Table 3 4. Continued Model Results Variable Description Mean PE SE t Value p Value PRO_TUN = PRO Tuna 0.010 1.892 *** 0.037 51.13 <.0001 PRO_MAHI = PRO Mahi Mahi 0.009 1.907 *** 0.040 47.95 <.0001 PRO_SWOR = PRO Swordfish 0.008 1.936 *** 0.043 45.45 <.0001 PRO_HADD = PRO Haddock 0.007 1.195 *** 0.041 28.94 <.0001 PRO_OR = PRO Orange Roughy 0.007 2.179 *** 0.050 43.31 <.0001 PRO_HAL = PRO Halibut 0.007 2.194 *** 0.049 44.40 <.0001 PRO_SOL = PRO Sole 0.005 1.006 *** 0.063 16.08 <.0001 PRO_SMEL = PRO Smelt 0.001 0.509 *** 0.105 4.85 <.0001 PRO_GROU = PRO Grouper 0.004 2.023 *** 0.067 30.12 <.0001 Oil Spill Interaction Variables (OS = 1 if sold after 4/20/10; = 0 otherwise) O S _COD = O S Cod 0.003 0.131 0.082 1.60 0.1104 O S _SAL = O S Salmon 0.007 0.272 *** 0.053 5.15 <.0001 O S _TIL = O S Tilapia 0.006 0.229 *** 0.058 3.94 <.0001 O S _CAT = O S Catfish 0.004 0.168 ** 0.072 2.33 0.0197 O S _WHIT = O S Whiting 0.003 0.059 0.075 0.78 0 .4375 O S _FLOU = O S Flounder 0.003 0.256 *** 0.081 3.16 0.0016 O S _POLL = O S Pollock 0.003 0.008 0.085 0.09 0.9266 O S _PERC = O S Perch 0.003 0.121 0.095 1.27 0.2025 O S _TUN = O S Tuna 0.002 0.109 0.112 0.97 0.3297 O S _MAHI = O S Mahi Mahi 0.002 0.01 0 0.121 0.08 0.9344 O S _SWOR = O S Swordfish 0.002 0.256 ** 0.127 2.02 0.0438 O S _HADD = O S Haddock 0.001 0.426 *** 0.126 3.39 0.0007 O S _OR = O S Orange Roughy 0.001 0.533 *** 0.149 3.57 0.0004 O S _HAL = O S Halibut 0.001 0.853 *** 0.136 6.29 <.0001 O S SOL = O S Sole 0.001 0.567 *** 0.178 3.19 0.0014 O S _SMEL = O S Smelt 0.001 0.233 0.178 1.31 0.1916 O S _GROU = O S Grouper 0.001 0.262 0.210 1.25 0.2124 W_COD = WILD Cod 0.012 0.763 *** 0.218 3.50 0.0005

PAGE 56

56 Table 3 4. Continued Model Results Variable Description Mean PE SE t Value p Value W_SAL = WILD Salmon 0.039 0.910 *** 0.149 6.10 <.0001 W_FLOU = WILD Flounder 0.005 0.149 0.302 0.49 0.6215 W_POLL = WI LD Pollock 0.001 0.165 0.605 0.27 0.7855 W_PERC = WILD Perch 0.001 2.637 *** 0.523 5.05 <.0001 W_TUN = WILD Tuna 0.006 0.017 0.314 0.05 0.9569 W_MAHI = WILD Mahi Mahi 0.004 3.000 *** 0.361 8.32 <.0001 W_SWOR = WILD Swordfish 0.003 2.536 *** 0.419 6.05 <.0001 W_OR = WILD Orange Roughy 0.001 4.493 *** 0.509 8.82 <.0001 W_HAL = WILD Halibut 0.005 0.763 ** 0.341 2.24 0.0254 W_SOL = WILD Sole 0.003 1.943 *** 0.428 4.54 <.0001 Time Event Variables HOL_1 = 1 for 2 weeks in mid February ; = 0 otherwise 0.038 0.029 *** 0.011 2.79 0.0053 HOL_2 = 1 for 2 weeks before Christmas ; = 0 otherwise 0.038 0.003 0.011 0.26 0.7963 Branding Variables NB = 1 if name brand; = 0 otherwise 0.796 BASE SB = 1 if private label brands ; = 0 otherwi se 0.204 0.370 ** 0.147 2.52 0.0116 Labeling Variables REGTYPE = 1 if no attribute labeling; =0 otherwise 0.496 BASE ALAS = 1 if labeled as Alaskan ; = 0 otherwise 0.066 0.038 0.125 0.30 0.7614 PAC = 1 if labeled as Pacific ; = 0 otherwise 0.05 1 1.310 *** 0.123 10.63 <.0001 ATL = 1 if labeled as Atlantic ; = 0 otherwise 0.024 0.995 *** 0.176 5.64 <.0001 BN = 1 if labeled as Boneless ; = 0 otherwise 0.264 0.482 *** 0.065 7.40 <.0001 SM = 1 if labeled as Smoked ; = 0 otherwise 0.024 8.987 *** 0.16 0 56.11 <.0001 Package Size SIZ E Size of package in ounces 20.855 0.055 *** 0.002 34.83 <.0001 Product Form FIL = 1 if Fillet; = 0 otherwise 0.774 BASE

PAGE 57

57 Table 3 4. Continued Model Results Variable Description Mean PE SE t Val ue p Value PIEC = 1 if Piece, Strip, Nugget; = 0 otherwise 0.118 0.764 *** 0.086 8.85 <.0001 STEA =1 if Steak ; = 0 otherwise 0.081 0.054 0.155 0.35 0.7253 WHOL =1 if Whole ; = 0 otherwise 0.037 0.962 *** 0.160 6.02 <.0001 REMF = 1 if Ground, Loaf, Burger ; = 0 otherwise 0.017 3.450 *** 0.193 17.86 <.0001 Model Statistics Observations 89,101 0.64 70 <0.0001 Total R Square Pr > ChiSq Durbin Watson 1.9955 Note: Level of Significance = 99%***, 95%**, 90%*. Intercept = ase produc t of fillet c od, not labeled wild, not on promotion, not sold after the Oil Spill or during the two holiday periods. In addition, t he base product has no attribute/origin labeling and is sold under a national retail brand. The package size is 0 oz.

PAGE 58

58 Ta ble 3 5 Finfish predicted price and interaction effects Percent change of predicted price for promotional, situational and labeling attribute effects Species Predicted Price PRO OS WILD COD $7.33 20.32% NS 10.40% SAL $ 8.7 8 19.72% 3.10% 10.37% TIL $ 5.4 6 21.88% 4.20% No Wild CAT $ 4.80 18.49% 3.50% No Wild WHIT $ 3.90 11.65% NS No Wild FLOU $ 5.9 3 21.23% 4.33% NS POLL $ 4.4 2 17.69% NS NS PER $ 6.2 4 21.31% NS 40.66% TUN $ 7.64 24.75% NS NS MAHI $ 6.86 27.78% NS 43.71% SWOR $ 8.0 1 24.18% 3.20 % 31.67% HADD $ 6.2 2 19.22% 6.85% No Wild OR $ 10. 20 21.37% 5.22% 44.07% HAL $ 12.9 5 16.95% 6.59% 5.90% SOL $ 5.92 16.98% 9.57% 32.81% SMEL $ 3.58 14.21% NS No Wild GROU $ 6.9 1 29.28% NS No Wild Note: Predicted price s adjusted for size (16 oz.) NS = Parameter not significant ( p > 0.10).

PAGE 59

59 Table 3 6 Frequencies of name brand and store brand fin fish products by species group Species Name brand frequency Store brand frequency Name brand percentage Store brand percentage Difference in percentage COD 5876 2028 6.8 9.22 2.4 SAL 14612 3848 17.0 17.49 0.5 TIL 13260 1560 15.4 7.09 8.3 CAT 9308 884 10.8 4.02 6.8 WHIT 7072 1144 8.2 5.20 3.0 FLOU 6292 1456 7.3 6.62 0.7 POLL 6708 728 7.8 3.31 4.5 PERC 4472 1716 5.2 7.80 2.6 TUN 3068 1508 3.6 6.86 3.3 MAHI 3172 936 3.7 4.26 0.6 SWOR 2496 1560 2.9 7.09 4.2 HADD 2288 1040 2.7 4.73 2.1 OR 2340 780 2.7 3.55 0.8 HAL 1248 1664 1.5 7.57 6.1 SOL 1508 728 1.8 3.31 1.6 SMEL 1404 1.6 1.6 GROU 832 416 1.0 1.89 0.9 SUM 85956 21996 100.0 100.0 0.0 Note: Smelt product not sold under store branding.

PAGE 60

60 Figure 3 1 Percent change of predicted price for promotional sales by s hellfish s pecies g roup -35% -30% -25% -20% -15% -10% -5% 0% SCAL SQU LOB CRAW MUSS CLAM OCT

PAGE 61

61 Figure 3 2 Percent change of predicted price after the Gulf of Mexico o il s pill by s hellfish s pecies g roup -6% -4% -2% 0% 2% 4% 6% 8% SCAL SQU LOB CRAW MUSS CLAM OCT

PAGE 62

62 Figure 3 3 Percent change of predicted price for i mport l abel ing by s hellfish s pecies g roup -40% -35% -30% -25% -20% -15% -10% -5% 0% 5% 10% SCAL SQU LOB CRAW MUSS CLAM OCT

PAGE 63

63 Figure 3 4. Shellfish n group SCAL 24% SQU 23% LOB 15% CRAW 15% MUSS 12% CLAM 8% OCT 3%

PAGE 64

64 Figure 3 5 Shellfish s tore brand products market share by sp ecies group SCAL 69% SQU 4% LOB 24% MUSS 3%

PAGE 65

65 Figure 3 6 Percent change of predicted price for promotional sales by f infish s pecies g roup -35.00% -30.00% -25.00% -20.00% -15.00% -10.00% -5.00% 0.00%

PAGE 66

66 Figure 3 7 Percent c hange of p redicted p rice after the Gulf of Mexico o il s pill by f infish s pecies g roup 0.00% 2.00% 4.00% 6.00% 8.00% 10.00% 12.00%

PAGE 67

67 Figure 3 8 Percent c hange of p redicted p rice for w ild l abeling by f infish s pecies g roup 0.00% 5.00% 10.00% 15.00% 20.00% 25.00% 30.00% 35.00% 40.00% 45.00% 50.00%

PAGE 68

68 Figure 3 9. Finfish group COD 7% SAL 17% TIL 15% CAT 11% WHIT 8% FLOU 7% POLL 8% PERC 5% TUN 4% MAHI 4% SWOR 3% HADD 3% OR 3% HAL 1% SOL 2% SMEL 2% GROU 1%

PAGE 69

69 Figure 3 10. Finfish COD 9% SAL 17% TIL 7% CAT 4% WHIT 5% FLOU 7% POLL 3% PERC 8% TUN 7% MAHI 4% SWOR 7% HADD 5% OR 4% HAL 8% SOL 3% GROU 2%

PAGE 70

70 CHAPTER 4 CONCLUSIONS Summary of the Stud y The objective of this paper was to use ACNiels e n S cantrack data to estimate price premiums and discounts associated with frozen finfish and shellfish products at the retail level in the U.S. Consumers revealed preferences for product attributes were det ermined using weekly sales data aggregated by UPC for each type of product (finfish or shellfish, exclusive of oysters, shrimp and crab). Additionally, the use of interaction terms between attributes and species of shellfish and finfish was tested as nece ssary to understand the additional dimensions of a product attribute. It was determined that interaction terms were needed to fully understand an attributes contribution to the overall price of the product. Both product and situational attributes were ex plored, specifically price effects of promotion, labeling shellfish) and the 2010 oil spill that occurred in the Gulf of Mexico. Summary of the Results The promotional, labeling and time event attributes varied significan tly between shellfish and finfish, as well as species within each product category. The holiday variables were not significant in the shellfish model. The mid February variable that the onset of Lent was significant in the fi nfish model, but has a rather small parameter estimate ( $ 0.03 / lb. ). Private label brands are significantly cheaper in the shellfish model and significantly more expensive in the finfish model reducing price by $ 3.37 per pound and increasing price by $ 0. 37 per pound, respectively C onsumers pay $1.63/lb. less for a shellfish product which contains the shell. Products which are prepared and closer to cook ready fetch a higher price Whole fish products are $0.96/lb. cheaper than fillet products, on aver age. Highly processed fish products (e.g. ground, burger, loaf) sell at a price discount of $3.45 The negative relationship

PAGE 71

71 between package size and price per pound is confirmed. For every ounce of package size, shellfish products average price per pou nd decrease by $0.08 and finfish products average price per pound decrease by $0.06. Findings By examining the promotional variable ( PRO) estimates and descriptive statistics, it is determined that the finfish market is more heavily involved in promotional behavior than the shellfish market. Both shellfish and finfish adjusted weighted price per pound decreases as proportion of promotional sales increases. This finding supports that promotional activities for these products are as traditionally assumed (i sold for a lower price), and promotional activit i es on average are those that involve price discounts. A higher percentage of finfish products are sold under promotion than shellfish products. F ollowin g the Deepwater Horizon Oil Spill, the finfish and shellfish markets reacted differently. Only the shellfish market exhibited a statistically significant price decrease for any species, scallops and lobsters which were also the two most expensive species groups. This price decrease may be explained by scallop and lobster retailers and producers attempt ing to increase demand for seafood during a time in which its safety in question. Mussels exhibited a large i ncrease in price. A large proportion (~75%) of mussels are labeled imported. This is higher than any other species group. This price increase may be explained by the foreign producers believing that their product is more valuable because it comes from waters not associated with the spill. In th e finfish market, it was observed that species which are not harvested in the Gulf of Mexico increased in price following the oil spill. Salmon, tilapia, catfish, flounder, swordfish, haddock, orange roughy, halibut and sole all exhibit ed price increases following the oil spill

PAGE 72

72 These data only encompass six weeks of sales, covering the initial reaction of the market to the oil spill. Some may debate that the oil spill had just started during this time period, and all the products on the shelf had more th an likely been harvested prior to this disaster. However, such an event acts as a reminder to consumers of past disasters, and revives past questioning of consu mers may begin to question Alaskan seafood once again because they are reminded of safety concerns due to the new oil spill far from the Alaskan fisheries. According to a review in the Journal of Environmental Health Perspectives NOAA initiate d the first fisheries closures on May 2nd and expand ed closures to nearly 84 thousand square miles by the beginning of July (Gohlke et al. 2011). Th e data end prior to the reopening of closed areas (i.e., June 23 rd ) Alternatively, the lack of a statistically signi ficant price decrease for Gulf sourced species (e.g., grouper) may also be due to the aggregation of all regional markets, or that there simply was not an effect. Finfish products wild on the packaging received a price premium in comp arison to products without this descriptor The price premium varied based on the species. The magnitude of this price premium varied from as much as 44% on orange roughy to 6% for halibut. This evidence could be used to support a movement by the indust ry to adopt a regulatory body which ensures products are being labeled correctly. These results dismiss the need for paying the high cost associated with e co l abeling certification program such as the Marine Stewardship Council (MSC). Although this varia ble encompasses products not eco label certified, the results show that the wild label fetches a price premium in excess of the MSC label (Roheim et al. 2011). It should also be noted that the products were not examined to determine if the wild labeled p roduct was also MSC certified.

PAGE 73

73 The majority of i mported products in the shellfish market we re sold at a price discount. The only product that sold at a premium in the U.S. market during the study period when labeled imported was lobster. Scallops, squi d, mussels, and clams were all sold at a price discount with These models serve as an overall representation of the United States market for selected frozen and unbreaded shellfish and finfish. Specific regional models are nec essary to reveal intricacies unable to be detected by these general models. The Simpson Paradox is a disappears when the data are aggregated for the whole populatio data apart and analyzing specific subgroups of interest could yield more explanatory results. Key Results and Implications Intangible attributes such as promotional behavior, the Gulf Oil Spill and various labeling practi ces were found to have statistically significant correlation with ( and relative magnitude of) the price consumers paid for some seafood products in the United States. It was determined that interaction terms were necessary to understand the multiple dim ensions of these attributes as suggested by Ahmad and Anderson (2012 ) Natural disasters, as well as disasters which are caused by man are highly unpredictable. The effects such disasters have on our natural resources are hard to assess, as are the tot al costs. By using interaction terms, this study is able to detect changes in the average market price of seafood products by species after the Gulf oil spill occurred. Since the oil spill, the tsunami and nuclear disaster in Japan has increased concern over highly migratory species in the Pacific Ocean. Using the methodology presented in this thesis and more current data, a similar study to assess changes in the average market price for highly migratory species in the Pacific Ocean could be developed

PAGE 74

74 As expected in the US market, the store brand shellfish products are cheaper than name brand products. Store brand products in the finfish model were o f higher value than name brand finfish products selling for an average of $0.37 more than name brand p roducts This store brand price premium is opposite the general consumer perception and findings from the US label products were found to demand a price premium in the UK market ( Roheim et al., 2007). In both the finfish and shellfish markets, t he market share of name brand and store brand products by species group indicate lesser valued species groups are less present under the store brand. The general consumer perception for imported p roducts is that they are cheaper than domestically produced alternatives This is proven to be untrue by the shellfish model. For example, imported lobsters fetched a higher price than domestically harvested lobsters. This may be due to the species of l obster imported. Domestically produced lobster may be of less value to U.S. consumers than the imported species. Labeling finfish as wild increases the value in a number of species, by a margin of nearly 50% in some cases. For example, the 14% price pr emium of Marine Stewardship Council Alaskan Pollock sold in London, as determined by Roheim et al. (2010). The discovery of such a high premium may influence the industry to adopt a wild certification program, or improve self regulation. Many of the spec ies with price premiums for wild labeling did not have a farmed fish option. Consumers may not be aware of the different production methods based on fish species. Future Work of seafood product attributes has revealed much about the potential and limitations of future

PAGE 75

75 revealed preference work. Additional analyses that could be conducted to improve the analyses presented here include: Expand on the situational variables to bet ter capture or expand on the points of interest in time. Dummy variables could be assigned to the exact weeks Lent occurs over the three year period rather than just capturing the onset. Thanksgiving is a time of low finfish promotions It may be of int erest to explore price changes of species groups during this time. The Oil Spill coefficient can be expanded on and better understood by focusing on subsets of the national dat a i.e. using regional data. Within these subset models, i t may be of more va lue to examine only those species harvested in the Gulf of Mexico Due to the richness of the entire dataset, several other avenues of research are likely to provide additional information on seafood marketing in the U.S. including the following: While ch oosing to examine a large, nationwide dataset is an ideal starting point for market analysis, restricted sample models could be used to reveal the market intricacies which are unable to be detected at such a large scale. Future work should ex plore the pos sibility of the Simpson paradox existing in relation to specific markets (either by Census region or major market areas for example) within the United States. For example, t he possibility exists that the price effects from the oil spill estimated in this model are conservative, as the data encompasses areas which may not have been concerned or were not aware (particularly at that point in time). Examining a similar model using only regions around the Gulf of Mexico may yield different results. Promotiona l behaviors may also vary based on the possibilities of recreational harvesting seasons ; r egions of which a specific species is unable to be harvested by

PAGE 76

76 the consumer (e.g. halibut in Florida) may exhibit different promotional behaviors than regions which harvest is possible. Estimation and model specification techniques as used by the Roheim et al. (2011) study of Alaskan Pollock may be applied to the examination of two species within roduced species may be possible through restricting the two included species to those which are commonly harvested in such areas. For example, models which estimate the implicit prices for Pacific Salmon and Atlantic Grouper could be compared between the Jacksonville, FL and Seattle, WA markets. To examine the value of alternative product forms for a given species, data could be pooled across the 19 files to allow for estimation of implicit prices related to more highly processed products.

PAGE 77

77 APPENDIX DEVEL OPMENT OF THE DEPEND ENT VARIABLE Non promotional price per package : (PRICE) = (non promotional sales (USD$))/ (non promotional units (packages)) Promotional price per package : (P_PRICE) = (promotional sales (USD$)) / ( promotional units (packages)) Pound s per package : (PACKLBS) = (SIZE/16) Proportion of promotional sales : ( PRO ) = (promotional units sold) / ( total units sold) Weighted price: (WPRICE) = (P_PRICE PRO) + (PRICE (1 PRO)) The Consumer Price Index variable is assigned to each observation using a time series identification variable. The time series identification variable is calculated with YEAR and WEEK. For example, if YEAR = 3 and Week = 52 then TSID = 156. The value for the Consumer Price Index when TSID = 1 is 202.416. Consumer Pric e Inde x base 2007 adjustment variable: (CPIBASE7) = (CPI/202.416) Adjusted weighted price using Consumer Price Index base 2007 : (ADJ_WPRICE) = (WPRICE/CPIBASE7) Adjusted w eighted price per pound: ( ADJ_ WPRICE_LB) = ( ADJ_ WPRICE/PACKLBS)

PAGE 78

78 LIST OF REFER ENCES Processed Food Products Canadian Journal of Agricultural Economics 60(2012):113 33. AMI Fact Sheet. April 2009. of Apple Prices and Product Quality Characteristics in British Canadian Journal of Agricultural Economics 48(2000):241 57. graphic Wine Appellations: Hedonic Pricing of Burgundy Wines in the British Columbia Wine Market Canadian Journal of Agricultural Economics 58(2010):93 108. Marine Resource Economi cs 25(2010):391 407. Egg Prices Journal of Agricultural and Resource Economics 35(2010):406 23. Jo urnal of Evolutionary Economics 18(2009):589 604. Neural Networks Applied Economics 33(2001):659 71. Gohlke, J. M., D. Doke, M. Tipre, M. Leader, and T. Fitzgerald. A Revi ew of Seafood Safety after the Deepwater Horizon Blowout. Journal of Environmental Health Perspectives 119 (2011): 1062 69. Guillotreau, P. How does the European Seafood Industry S tand after the R evolution of S almon F arming : An E conomic A nalysis of F ish P rices Journal of Marine Policy 28 (2003): 227 33. Journal of Product and Brand Management 4(1995):38 48. Jacquet, J. L. and D. Pauly Trade Secrets: Renaming and Mislabelin g of Seafood. Journal of Marine Policy 32 (2008): 309 18. Klass, G. Just Plain Data Analysis Rowman and Littlefield Publishers Maryland (2008):16. Lambert, J., L. Klieb and M. Weber Regional Influences upon the Selection of Imported Versus Domestic Se afood. Academy of Marketing Studies Journal 12( 2008 ) : 17 42.

PAGE 79

79 Leek, S., S. Maddock, and G. Foxall Situational D eterminants of Fish Consumption. British Food Journal 102( 2000 ) : 18 39. McConnell, K. E. and I. Strand Hedonic Prices for Fish: Tuna Prices in Hawaii. American Journal of Agricultural Economics 82( 2000 ) : 133 44. Hedonic Price Model Journal of Urban Economics 14(1983):327 37. Roheim, C. A., F. Asche and J. Santos. The Elusive Price Premium for Ecolabelled Products: Evidence from Seafood in the UK Market. Journal of Agricultural Economics 62( 2011 ) : 655 68 Roheim, C. A., L. Gardiner and F. Asche. Value of Brands and Other Attributes: Hedonic Analysis of Retail Frozen Fish in the UK. Marine Resource Economics 22 (2007): 239 53. Marine Resource Economics 25(2010):77 92. Rosen Prices and Implicit Markets. Journal of Political Economics 82(1974 ) :35 55. Sayrs, L. W. Pooled Time Series Analysis Sage Publications, California (1989):19. Wines on the Decline ? Econometric Evi dence Journal of Agricultural Economics 55 (2004):267 88. Thrane, M., F. Ziegler and U. Sonesson. Eco labeling of W ild caught S eafood P roducts Journal of Cleaner Production 17 (2009): 416 23. U.S. National Oceanic and Atmoshperic Administ ration, National Marine Fisheries Service, Fisheries of the United States annual, September 2010. Table 895: Fishery Products.

PAGE 80

80 BIOGRAPHICAL SKETCH Glen Gold was a University of Florida Resource Economi cs. A native Floridian, his interest in fisheries began when he was a child. Fishing, boating and enjoying the outdoors were common weekend activities for Glen growing up. He has experienced the detrimental effects of natural resource overuse first hand when the neighborhood lake was dry for several years due to over pumping of the aquifer. After completing a number of data analysis and econometrics courses, Glen was offered the opportunity to work with real market data as a research intern with the In stitute of Food and Agricultural Sciences. This summer internship developed into a Graduate Research Assistant position during his last year of studies. Glen also worked at Santa Fe College as a Graduate Assistant in the Student Life Business Office. He completed his b achelor s degree with honors in spring of 2011