Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
University of Massachusetts Amherst University of Massachusetts Amherst
ScholarWorks@UMass Amherst ScholarWorks@UMass Amherst
Masters Theses 1911 - February 2014
1969
Analysis of consumer demand for selected species of fresh fish Analysis of consumer demand for selected species of fresh fish
based on data from five retail stores. based on data from five retail stores.
Robert Edward Lee University of Massachusetts Amherst
Follow this and additional works at: https://scholarworks.umass.edu/theses
Lee, Robert Edward, "Analysis of consumer demand for selected species of fresh fish based on data from five retail stores." (1969). Masters Theses 1911 - February 2014. 1706. Retrieved from https://scholarworks.umass.edu/theses/1706
This thesis is brought to you for free and open access by ScholarWorks@UMass Amherst. It has been accepted for inclusion in Masters Theses 1911 - February 2014 by an authorized administrator of ScholarWorks@UMass Amherst. For more information, please contact [email protected].
ANALYSIS OF CONSUMER DEMAND FOR SELECTED SPECIESOF FRESH FISH BASED ON DATA
FROM FIVE RETAIL STORES
A Thesis Presented
by
Robert I. Lee
Submitted to the Graduate Seho^ 1 of the
University of Massachusetts inpartial fulfillment of the requirements for the degree o:
MASTER OF SCIENCE
June 1969
ii
ANALYSIS OF CONSUMER DEMAND FOR SELECTED SPECIESOF FRESH FISfe BASED ON DATA FROM FIVE RETAIL STORES
A Thesis Presented
Robert E. Lee
Approved as to style and content by:
(montfT) (year)
iii
ACKNOWLEDGMENTS
A sincere note of appreciation is extended to those many individuals
of the case study chain, which we have agreed not to identify, who cooper-
ated completely to provide data for this study. Especial thanks are also
due to Dr. David A. Storey and Dr. James K. Kindahl for their patience and
direction. Thanks are also extended to my typist, Miss Kathy Godek, who
displayed unusual diligence in preparing the final manuscript.
iv
TABLE OF CONTENTS
PageNo.
ACKNOWLEDGEMENTS iii
TABLE OF CONTENTS iv
LIST OF TABLES vi
LIST OF FIGURES vii
CHAPTER
I INTRODUCTION 1
II REVIEW OF LITERATURE 3
III RESEARCH PROCEDURE 9
Coverage of the Study ' 9
Data Collection . 11
Analysis of the Data 12
IV DESCRIPTION OF THE CASE STUDY CHAIN AND ITS FISHMERCHANDISING PRACTICES 13
The Chain and Its Place in the Industry 13
Comparison of the Five Stores 13
Description of Fish Merchandising Practices 14
V PRELIMINARY ANALYSIS OF FRESH FISH SALES FOR FIVE
LEADING SPECIES 20
Sales in Pounds 20
Retail Prices .24
Special Promotions 27
Summary 34
VI REGRESSION ANALYSIS OF FRESH FISH SALES FOR FIVE
LEADING SPECIES 3j
Theoretical Consd derations 35
The Model 39
Results 48
Summary of Results of Regression Analysis 71
V
TABLE OF CONTENTS (continued)PageNo.
VII SUMMARY AND CONCLUSIONS 74
APPENDICIES
A MANAGER QUESTIONNAIRE 79
B FORM FOR COLLECTION OF DATA FROM INDIVIDUAL STORES 84
C SIMPLE CORRELATION TABLES 86
BIBLIOGRAPHY 97
LIST OF TABLES
Title
Carry-over of fresh fish and seafood
Location of fish display relative to normal direc-tion of traffic flow
Aggregate weekly mean pounds ordered and total reve-nue by species
Coincidence of individual species' peak quantitieswith aggregate peak quantities
Proportion of sales accounted for by the singlefeatured species
Relative importance of special and nonspecial pricesby species
Effect of specials of other species on pounds or-
dered of haddock-
Haddock adjustments for five stores
Coefficient and statistical test values for six
regressions
Model No. 1 - All variables free to enter
equation
Coefficients and statistical test values for six
regressions
J-lodel No. 2 - Significant variables free to
enter equation
Coefficients and statistical test values for six
regressions
Model. No- 3 - Haddock slope dummy variable
omitted
Coefficients, elasticities, and statistical test
values for six regressions
Model No. 4 - Haddock dummy variables omitted
Coefficients and statistical test values for six
regressions
Model No. 5 - Nonspecial data only
I
LIST OF TABLES (continued)
TableNo. Title
6-7 Number of weeks particular species were not sold
LIST OF FIGURES
FigureNo. Title
4- 1 Store manager's estimate of weekly sales distribu-
tion
5- 1 Quantity of each species and total aggregate plottedover time
5-2 Price trends of five species
5-3 Percent of five species aggregate made up of single
featured species
5- 4 Individual price-quantity ordered relationships
6- 1 Theoretical representation of the supply-demand sit-
uation
CHAPTER I
INTRODUCTION
In the early 1950 Ts Donald White in his book The New England Fish-
ing Industry [35] forecast a rather dim future for the Nevz England fish-
ing industry. At that time it faced a declining resource base, labor
problems, and growing competition from foreign countries. Through the
early 1960's White's pessimism was borne out as foreigners, with their
superior technology and cheaper labor, made large inroads into the U.S.
market. Recently, however, the introduction of stern trawling to the
Atlantic fleet and the passage of the 1964 Fishing Fleet Improvement Act
have given new promise to the industry. This act compensates fishermen
for up to 50 percent of the cost of new fishing vessels, making the adop-
tion of new technology more feasible.
Marketing problems have plagued the industry also. Historically
there has been a preference for meat in the United States which has
severely curtailed the market for fish and fish products. The pattern
of consumption of fish has been linked closely to religious factors and
custom, with tne heaviest consumption occurring on Fridays and during
the Lenten season. In 1966, the Pope removed the restrictions of the
Catholic Church on the consumption of meat, so religion may no longer
be a major factor influencing fish consumption. Whether or not the de-
mand for fish has changed because of this new ruling could be an inter-
esting topic for economic research itself, and certainly suggests the
comparison of research conducted since the ruling with that conducted
before it.
2
The possible impact of new technology which is now likely to be
adopted by the industry is also a subject for research. A question
which might L>e asked is whether or not the demand for fish justifies
new expenditures on equipment. Possibly it would be preferable to
allow the New England industry to continue declining and meet domestic
demand with the cheaper, imported product.
These are questions which are difficu.lt to answer because very lit-
tle, of a specific nature is known about the demand for fish in the United
States. Relatively little research has been conducted in this area at
any level of the marketing system, and particularly at the retail level.
Past studies have tended to use annual data with several species lumped
together..Those studies which have disaggregated by species normally
lumped fresh and frozen forms of the species together or else used only
data on the frozen form. While this certainly is of interest, the New
England fishermen cater primarily to the fresh market, since they have
a locational advantage over foreign suppliers, and the price for fresh
fish can be twice that for frozen fish. Thus, the market of most im-
mediate interest is the fresh market.
This study was intended to partially fill this informational gap.
Stated most, generally, its purpose was to improve the state of knowledge
about consumer demand for selected- species of fresh, salt-water fish.
Weekly data collected from individual supermarkets of a case study food
chain was used to construct, demand equations for the fresh form of in-
dividual species. By using weekly data, short-run relationships could
be studied instead of lon<? term trends. Factors which were of primary
interest in this study were the prices of competing species and the e
feet of promoting a species at a reduced price. From the functions
utilizing these data it was possible to calculate price and cross ela
ticies of demand. Other factors which could have been included, such
as the effects of ethnic origin, religion, and consumer income, were
beyond the scope of this study.
CHAPTER II
REVIEW OF LITERATURE
It was stated previously that relatively few studies have been con-
ducted on the consumer demand for fish. In this chapter those few which
have been conducted will be discussed along with other useful demand
studies .
-
Joseph Farrell and Harlan Lampe [11] conducted a study in 1965 to
determine the price elasticity of demand for fish. The. authors built a
model to study five market levels for haddock. Supply and demand func-
tions were estimated at the landings, wholesale, imports, cold storage,
and retail jLeve Is . Monthly data from 1954 through 1962 were analyzed by
the limited-information maximum- likelihood technique. Since monthly data
were used, the authors had no opportunity to examine weekly price move-
ments, but were able to analyze seasonal changes in demand.
At the retail level the authors found a substantial increase in de-
mand during the Lenten season, but because of high multicollinearity among
some of the independent variables, much of the statistical results was in-
conclusive „ From the retail demand equation for frozen fillets of haddock
a price elasticity was estimated. The data were separated into three cate-
gories,allowing the calculation of three distinct elasticity estimates.
From the equation describing the retail market from January through June
the price elasticity of demand was +4.93 For the period from July throuw
December the estimate was -A. 409. An equation for the whole year was es-
timated also, but the haddock quantity variable was insignificant. Of the
two elasticity estimates that were obtained, it is interesting to note
that both were highly elastic. However, for some reason, left unexplained
by the authors, the sign for the spring of the year estimate was positive.
A more recent study on consumer demand for fish has been reported by
Darrel Nash [24] of the Bureau of Commercial Fisheries. At the time of
the preliminary report on the study, time series data had been analyzed
by the single-equation least-squares regression technique. Additional
analysis was planned utilizing cross-section data from a Michigan State
University consumer panel in an attempt to obtain information about one
period in time.
From the aggregated time series data the Bureau was able to derive
demand functions for fresh and frozen flounder individually, but lumped
fresh and frozen together for haddock. For the years 1950 to- 1963, fresh
flounder had a price elasticity ranging from -4.0 to -6.0 at the mean price
and quantity. The estimate of price elasticity for frozen flounder fillets
was based on the years of 1954 to 1963. At the mean price and quantity it
was approximately -3.5. Unexpectedly, the elasticity for both forms of
flounder was higher at lower prices than at higher prices. Since Nash
was talking of 'linear functions this statement at first seemed implausible,
but apparently he was thinking of shifts in the short-run demand curves,
rather than elasticities calculated along the long-run functions. Over
the period covered by the data, price trended upward and apparently the
short-run demand curves became more inelastic.
In the haddock equation, the fresh and frozen forms of the species
were combined and treated as a single commodity. Data were available
for the years 1954 to 1964. At the mean price- and quantity for this
period the price elasticity estimate was -1.4. It will be noted that all
these estimates were negative, as normally expected for demand equations,
and greater than one. The latter factor shows demand for these individual
species to be elastic, which means that at the retail level, as the price
of the species is lowered, the quantity consumed will increase by a pro-
portionately greater amount, thus increasing total revenue.
One article particularly useful in developing the methodology for
this study was that published by Fred Nordhauser and Paul Farris [26] in
the November, 1959 issue of the Journal of Farm Economics . In their study,
weekly price and quantity data from six supermarkets in four mid-western
cities were used to estimate the short-run price elasticity for fryers.
Stores were selected from the same chain so as to have consistent pricing
and merchandising policies. When selecting stores, consideration was also
given to representing a variety of clientele - Whites, Negroes, and vari-
ous income levels.
The problem of handling advertised price specials was encountered in
their study. So as not to bias their estimates of price elasticity in
non-special weeks, the authors omitted from their analysis the weeks in
which specials 'occurred. They did notice, however, that when fryers were
offered on special one week, there x^as no noticeable effect on sales the
following week. Therefore, the increase in sales in special weeks seemed
to be a net gain in sales.
A study conducted by the National Association of Food Chains reached
similar conclusions about the effects of price specials for meat [25]= It
stated, "Changes in proportion of the total [meat, poultry and fish ton-
nage] caused by a price special do not see- to affect tonnage of the feature
item adversely in the following week. The price special can greatly in-
crease volume in the short-run, but the amount sold in the following weeks
remains near the long-run normal [25, p. 6].'" In addition, the researchers
found that f, ...a price special raises the tonnage of the featured item, but
does not depress volume of the non-featured items by an equivalent amount.
It appears from this that the added tonnage sold during a sale is largely
'plus tonnage' [25, p. 6]." Another finding of this study was that the
tonnage of a particular item may rise by as much as five or ten times the
normal volume when offered on special.
A studywhich handled the analysis of price specials differently from
Nordhauser and Farris is that conducted by Leo R. Gray [16]. He collected
weekly sales volume and poultry price data from a representative sample of
firms in the Washington, D. C. , metropolitan area. In an attempt to meas-
ure the effects of price specials, Gray employed dummy variables. One was
a normal zero-one variable to detect any shift in the level of the demand
curve as a result of price specials. The other was a uniquely constructed
variable to measure structural changes in demand. He constructed this
"slope dummy' 1 by multiplying the week's poultry price by -1 on nonsale
weeks, and +1 on sale x^eeks . He then expressed his function as:
Vt
= a0+b 1p tH-b
2vt-1+b 3
Dt;
+b4SD
t-f-u
t
where:
Vt= Weekly volume of estimated sales by all retailers in the
market area.
Pt
= Deviation from the average weekly prices for corresponding
weeks of each year,
v t„l= Deviation from the average of lagged weekly volumes for cor-
responding weeks of each year.
8
Dfc
= A dummy shift variable used to differentiate between a saleweek and a nonsale week. The sale week was given a value of1, the nonsale week 0.
SDt = The dummy slope variable. It is the product of the modified
dummy shift variable (Dlt ) and p tfor that week.
Djt
= -1-1 in a sale weekD]_
t- -1 in a nonsale week
If the coefficient of Dfc
is significant it is known that there is a
significant difference in the level of the demand curve between sale and
nonsale weeks. To be able to determine if there is a difference in slope
of the two curves, the coefficient of SD t must be tested. Should it be
significant, then it is known that the structure of demand is different
for the item when it is on sale as compared to when it is not,
er method of cons t ruct ing a s lope dummy to perform the same
task as the one just discussed, is explained by Arthur Goldberger [15].
The concept of multiplying a price by one value for a special week and by
another value for a nonspecial week is the same as that employed by Gray.
The difference is in the coefficients used. Goldberger uses 0 and 1 as
his multipliers instead of -1 and +1. If a 0 is used in a nonspecial week
and a 1 is used in a special week, then the slope dummy takes on the value
of zero on weeks where there is no price special, and a value of that week
price when there is a price special. The general result is the same as by
Gray's method, but the technique propounded by Goldberger is in more com-
mon usage
.
A recent publication worthy of mention in any study of fisheries eco-
nomics,although not used extensively in this study, is Recent Development
and Research in Fisheries Economics , edited by Frederick W. Bell and .Tared
Hazieton [3]. It contains reports of several recent studies conducted on
the economics of our fisheries
.
9
CHAPTER III
RESEARCH PROCEDURE
The procedure followed in this study could not be completely pre-
dicted, because the merchandising practices of the case study firm had
to be determined first. Once this was accomplished, full development of
the procedure followed. A summary of this procedure is given below.
Coverage of the Study
This report is a case study of a chain operating supermarkets in the
Springfield-Holyoke area of western Massachusetts. By designing the ex-
periment as a case study, rather than a sampling survey, several problems
were eliminated. One was the comparability of data collected fron differ-
ent sources. Most firms have a unique way of keeping records. Ey select-
ing stores from just one chain, the records from each store were assured
to be the same.
Another advantage of the case study was the similarity of pricing and
merchandising policies in each of the stores. These policies, while ex-
pected to be similar in different chains, would not be identical, Since
these policies would affect the demand for fish, they would have to be ac-
counted for if comparing data from different chains.
If the findings of this report are to be generalized to other chains,
it must be assumed that other chains are similar to the case study chain.
This assumption seems plausible for chains in the same geographical re-
gion, though it is not the purpose of this study to generalize in this
fashion. The express purpose here is to isolate some of the factors of
demand for fresh fish.
10
onThe particular firm selected for this study was chosen primarily
the basis of convenience. No advantages of one firm over another, rela-
tive to fish sales, were apparent. The chain which was selected operated
typical supermarkets throughout the local region, and its management dis-
played a gratifying spirit of cooperation.
It was not desired to survey all the stores of the chain, so a few
were selected on the basis of the following criteria. By surveying the
stores of the chain, it was found that there existed very few differences
among them relative to sales of fresh fish, A few sold fish through the
delicatessen counter, but most sold it through the meat case. It was felt
that stores selected for the study should represent both sales methods.
Additional criteria used in selecting the stores were their proximity to
each other, and their age. Since there were no major differences among
the stores it was decided to select stores which were close enough to each
other for all to be visited in one or two days. The age of the stores was
considered originally so that several years' data could be observed. Only
stores which had been operated for at least seven years were selected.
From the experience of another similar type study [26] it was decided that
five, stores from the chain could provide the data necessary for the study.
On the basis of all the above information, five stores were selected in
the western Massachusetts region, two of which sold fish through the deli-
catessen couiiter, and three of which sola it through the meat case.
The time period covered by the study was forty-two weeks, from July,
1967 through April, 1968. The historical records kept by the firm did not
provide ail the information necessary for the study, and July was the earl-
iest it was possible to start weekly record keeping procedures. Data
II
collection was continued until the week after Easter so as to include the
Lenten season
Five of a possible forty fish items were chosen for the study on the
bases, first of compatibility with the goals of the study, and second of
importance in sales. Since this study was concerned with fresh, salt-water
finfish, the fresh-water species and shellfish were immediately removed from
primary consideration. Of the species remaining after this division, five
were sold throughout the year and in noticeably larger volume. These, five,
haddock, swordfish, flounder, halibut, and codfish, were chosen for major
consideration.
Data Collection
Descriptive information about the operations of the stores was col-
lected by personal visits to the head office, to individual stores, and
by a manager questionnaire. The head office was visited at the beginning
of the study, and whenever problems relevant to the duties performed there
arose. Individual stores were visited at the beginning of the study, and
at approximately five week intervals thereafter. The original visit was
used to familiarize the researcher with each of the stores in the study,
and to explain to each meat manager the nature of the information he was
being asked to record. Later visits to the stores were utilized to dis-
cuss with the meat managers problems relating to their record keeping, and
to answer further questions in the resea oner's mind about store practice.;.
After several of these visits, it was deemed worthwhile to organize the
various questions which had arisen into a specific questionnaire, the pur-
pose of which was to systematically provide information about the handling
and merchandising of fish in each of the five stores [See Appendix A].
12
Numerical data were collected both from the head office and from the
individual stores. Basic information about quantities of fish purchased
by each store, broken down by species, was kept in the records of the main
office. Weekly wholesale and retail prices were also available from the
main office.
To supplement this, the meat manager in each store agreed to fill
out a weekly form recording the quantity, in pounds, of each species left
over at the end of the week. He also recorded the day and time that he
ran out of any species. This information was used to adjust the quantity
purchased figure, obtained from the head office, into the desired quantity
sold figures.
Analysis of the Data
In the early stages of analysis, the data were graphed and tabled for
simple, visual analysis. While this type of analysis will reveal some of
the more obvious relationships, it does not provide estimates of statis-
tical .significance, nor does it adequately identify multivariate relation-
ships .
A second and major stage of analysis involved the use of least-squares
multip.l e regression . This technique analyzes the effects of several inde-
pendent variables at. the same time. It also provides measurements of sta-
tistical significance.
The preliminary analysis handled aggregate data from the five stores
for each of the five individual species of fresh fish. In the regression
analysis, on the other hand, study was concentrated on fresh haddock fil-
lets. Here, the planned procedure was to first analyze each store and
then to compare the different stores by means of the Chow test.
13
CHAPTER IV
DESCRIPTION OF THE CASE STUDY CHAIN AND ITSFISH MERCHANDISING PRACTICES
The Chain and Its Place in the Industry
The chain selected for this study operated several supermarkets in
the greater Springfield area of Massachusetts. Within this region it ac-
counted for about 6.5 percent of all retail food business [17, p. 27], or
16.5 percent of total chain store business. Its stores were of average
supermarket size, each handling about $40,000 of total sales weekly, which
was also the per store mean for all the other major chains in the region
taken together [17, p. 27]
.
Fish and seafood sold by the meat department in the case study chain
accounted for about two percent of the department's sales. No comparable
figure was available for the region, but on a national basis, fresh fish
and seafood items accounted for almost three percent of meat department
sales [8, p. 32]. Relative to total store food sales, fresh fish and sea-
food accounted for approximately ,6 percent of sales in the chain and ap-
proximately .7 percent nationally.
Comparison of the Five Stores
The stores selected for this study averaged about $55,000 gross weekly
sales, slightly higher than the average for the entire chain. Most catered
to a mixed religious and ethnic clientele, but one store served a large
Jewish population. In this particular store about forty percent of the
customers were Jewish, but they accounted for an estimated fifty percent,
of the fish sales. This store and one other displayed their fish on ice
14
in the delicatessen counter. The other three stores displayed prepackaged
fish in the meat case in the same fashion as other meat items.
The peak sales periods for fish varied from store to store. Peak
sales occurred on Thursday evenings and Friday for two stores, on Thurs-
day morning and Friday for a third, and of the remaining two, one peaked
on Tuesdays and Fridays, while the other peaked on Wednesdays and Fridays
(see Figure 4-1)
.
The slack sales periods had more similarities than the peak periods,
but varied somewhat from store to store, also. On Mondays virtually no
sales were made at any of the stores, the reason being that fish was not
normally displayed in the case. Saturday sales were also quite low, ex-
cept at one store which made fifteen percent of its weekly sales on that
day (Figure 4-1). In addition to these slack periods, Tuesdays were light
for fish sales at three stores, and Wednesdays were light at two, one of
these and another. Most of the stores made some sales almost every day
except Monday, but store D T
s sales were limited strictly to Thursdays and
Fridays. A major factor in this unique pattern of sales may have been the
very difficult competitive position that this store faced. In the imme-
diate area there were three or four other large supermarkets , the closest
of which featured a large, fresh fish delicatessen display.
Description of Fish Merchandising Practices
Placement of weekly order . All fish to be sold through the meat de-
partment was delivered to the stores on Tuesday and Thursday mornings.
The meat managers placed their orders on Saturday for the following Tues-
day's delivery, and on Wednesday for Thursday's delivery. At the time the
15
Figure 4-1. store manager's estimate ofweekly sales distribution
% of weeklysales
STORE A
% of weeklysales
60
50
40
30
20
10
STORE C
E3*
1
1
13 EES
—
M W T
Day
% of weeklysales
80
70
60
50
40
30
20
10
L L
M T
% of weeklysales
60 STORE B
50
40
30
20
il10
m mM T W T F
% of weeklysales
60
50
40
30
20
10
Day
STORE E
M T W T
Day
E3
JUE3LF S
STORE D
w T
Day
16
original order was placed, neither the wholesale nor the retail prices
were normally known to the managers. They were usually informed, how-
ever, if a drab tic change was expected in the wholesale price. This in-
variably coincided with the offering of a species on a special promotion,
which fact was also made known to the managers a week in advance. In
neither case were they informed of the actual prices
-
When ordering, managers normally purchased only that amount of each
species which they estimated would be sold. To do this they worked from
the "normal 11 level of sales, purchasing a standard amount of each species,
particularly on Saturday orders. This order was then adjusted with the
second order of the week if sales of a particular species seemed to be
unusually high or low. The primary adjustments that occurred to the Sat-
urday orders resulted from the offering of a species on special. Then
the manager increased his Saturday order considerably, to the size nor-
mally used whenever that species was offered on special. Adjustments
were occasionally made for specific religious days, also. For instance,
the order would be increased for the first week of Lent, and decreased
for Roman Catholic fast days.
Carry-over from one week to th e next. Fish was carried from one week
to the next, and it was necessary to investigate the nature and extent of
this practice. It appears that fresh fist? was never carried to the follow-
ing week and sold as fresh (see Table 4-1). However, sometimes part of
Thursday's delivery was placed in the freezer immediately upon delivery.
Then, if it was not removed from the freezer for display in the meat case,
it was carried over for sale the following week as frozen fish. When this
17
was done, it was displayed separately from the fresh form of that species
and was normally priced ten cents per pound lower* Of the five stores,
only one frequently froze fresh fish and carried it over to the next week
Table 4-1. Carry-over of fresh fish and seafood.
Stores
Fresh fi
andsh carried oversold as
Frozen fish carried overand sold as
Fresh Frozen Frozen
A No Seldom Yes
B No Seldom Yes
C No Frequently Yes
D No Seldom No
E No Seldom Yes
Source: Manager questionnaire, see Appendix A..
The length of time that carried-over fish was held varied, but nor-
mally it was not kept more than one or two weeks. Two managers claimed
that they never held fish more than one week, and seldom held it at all.
The other three said that they might hold it as long as four weeks. Shrimp
was the one species that all managers agreed was more apt to be held over
for the longer periods of time.
In general, fish was ordered for a one week time period and carry-
over was the result of an overes timation of sales. However, on occasion
excessively large orders were intentionally placed. This occurred most
frequently when a manager knew that the wholesale price was going to be
unusually low, such as when a species was going to be offered on special.
As stated previously, the manager would not know the exact wholesale price
18
of an item to be featured, but he did know it would be considerably lower
than normal. One manager claimed that the meat buyer for the firm would
occasionally suggest that he purchase an extra supply of some commodity
when it was going to be cheaper.
On those occasions when the meat managers did deliberately over-order
they would order from half again as much as normal, to twice as much. The
species most frequently included in this practice were haddock fillets,
swordfish steaks, and certain shellfish items. One manager claimed that
he never deliberately ordered an amount greater than he expected to sell
that week.
" Packaging. Some packaging was done at the store and some was done
before the fish was delivered to retail stores. With the exception of
mackerel, salt-water fish were filleted, steaked, or dressed before reach-
ing the retail store. With mackerel, the heads were removed after they
were delivered. Haddock fillets, halibut fillets and ocean perch fillets
were normally wrapped before delivery, but smelts, flounder, swordfish,
codfish, clams, and halibut steaks had to be wrapped at the stores. Ex-
ceptions to this were the two stores which sold fish through the delica-
tessen counter. There, fish was displayed unwrapped on ice. At one of
the delicatessens, a large amount of fresh-water fish was sold; these
species were cleaned, and dressed or filleted in the store.
Fish display. The linear feet of the meat case allocated to fish
ranged from three to twelve feet, depending on the store (see Table 4-2).
At both stores where fish was sold through the delicatessen counter, a
full twelve foot case was used. At two of the other stores, six feet of
the meat case was allocated to fish, and at the fifth store, three feet.
19
Table 4-2. Location of fish display relative toNormal direction of traffic flow.
Linear feetStore Location in fish i
A Before major meat items 6
B Delicatessen case , before major meat items 12
C Before major meat items 6
D After major meat items 3
E Delicatessen case, before major meat i terns 12
Source: Manager Questionnaires, see Appendix A.
With one exception, fish was displayed at the beginning of the meat
case, so customers would see it before the major meat items (see Table 4-2)
Since the delicatessen case was located before the main meat case, this
meant that fish was either sold from the delicatessen case, or was dis-
played immediately after it. At the fifth store, fish was displayed after
the main meat case, but before the one with frozen sausage, turkey, and
beef patties. This also was the store that allocated only three feet of
the meat case to fish.
The location of the fish display area in the meat department did not
change on any short term basis. It may change every few years as the pat-
tern of sales changes or when a new man comes to the head office. Within
the allocated area, little changing was done either. The one major excep-
tion occur- -.a -when a species was offered on special, and then it was al-
lowed twice the amount of space it normally had.
20
CHAPTER V
PRELIMINARY ANALYSIS OF FRESH FISH SALESFOR FIVE LEADING SPECIES
This chapter is designed to acquaint the reader with the data and
present some of the more obvious relationships among the variables. It
is divided into the following major categories: sales in pounds, retail
prices, special promotions, and conclusions.
Sales in Pounds
Pounds orde red. Of the five major species considered in this study,
haddock appeared to be the most important (Table 5-1). At least this was
true when comparing the nonspecial data and the combined, special and non-
special, data. The aggregate, mean quantity ordered in normal weeks of
250 pounds was almost 100 pounds greater than for swordfish, the second
largest seller. The same situation existed when looking at the combined
data. Haddock also had the advantage in mean, weekly, total revenue, but
by a less noticeable amount. Following in importance were halibut and
flounder, and least important by far, in terms of both quantity and reve-
rtue, was codfish.
When observing just the price special data, swordfish was the most
important species. On the average, it sold greater than 100 pounds a
week more than haddock, its closest competitor, and brought an average
weekly total revenue of approximately 250 dollars more than haddock. It
showed an almost fivefold increase in sales, and a somewhat smaller in-
crease in revenue over the nonspecial weeks. The importance of price
21
specials was quite noticeable for all the species, although not all in-
creased by as great an amount as swordfish. All, with the exception of
haddock, however, did increase by at least fourfold on a quantity basis,
and at least threefold on a total revenue basis when they were offered
on special.
_
Table 5-1. Aggregate weekly mean pounds orderedand total revenue by species.
Pounds Orderedlbs.
Total Revenue
$
a
Species-
Specialweeks
1
Non-specialweeks
2
All weekscombined
3
Specialweeks
4
Non-specialweeks
5
All weekscombined
6
Haddock 619 250 322 347 198 229
Swordfish 737 163 231 590 163. 215
Halibut 466 92 163 317 62 117
Flounder 404 90 140 263 75 103
Codfish 570 39 52 336 27 35
^Total Revenue = Aggregate Mean Pounds Ordered x Mean Retail . Price.
Almost every" week, one of the above five species was featured as a
price special. It might be expected, therefore, that were all five species
aggregated, the aggregate would not fluctuate much from one week to another
However, this was not the case, as can be seen from the aggregate curve in
Figure 5-1. Tt shows considerable fluctuation, the peaks of which always
coincide with a peak of some individual species' curve. Of a total of
eleven aggregate peaks, six of these coincide with peaks of the swordfish
curve, two with the haddock curve, and one each with the flounder, codfish,
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
s
ed Figure 5-1. Aggregate Quantity of Each Species and Total Aggregate Plotted Over Time
22
WeeksJuly, '67 I Aug, *67 I Sept, '67 I Oct, '67 / Nov, '67 I Dec, '67 "1 Jan, '68 I Feb, '68 I
Mar, '68JApr, '68
23
and halibut curves (Table 5-2). It might be noted that these frequencies
follow the same general order of importance as the pounds ordered figures
of Table 5-1, column 1.
Table 5-2. Coincidence of individual species' peak quantitieswith aggregate peak quantities.
Simultaneous aggregateSpecies Number of peaks pe aks
Haddock 8 2
Swordfish 8 6
Halibut 9 1
Flounder 7 1
Codfish 1 1
The importance of price specials was further supported by the obser-
vation that the deepest troughs of the aggregate curve occur when none of
the five species were offered on special. On those occasions when there
is a trough in the aggregate curve, and also a special on some species,
the trough is not as deep as in many other instances. An example of this
is the trough on week number sixteen, as compared to weeks eleven and
twenty-one. The former trough occurs when flounder was on special, wherea
the two latter ones occur when there were no specials.
Actual pounds sold . The analysis in this chapter was based upon
"pounds ordered" data whenever quantity data was used, rather than the
actual "pounds sold" data. The reason for this stems from the fact that
in this chapter the data were aggregated by species over the five stores.
Each store had data missing, which meant that a five-store aggregate on
24
those weeks could not be calculated. Since, those occasions for each store
did not always overlap with other stores, only fifteen weeks of useful
data were available. It had been determined that the meat manager's es-
timates of how much should be sold was close enough to what was actually
sold to make the pounds ordered figures a useful approximation of pounds
sold. In Chapter 6, where the analysis was broken down by individual
stores, this problem was less acute and pounds sold figures were used.
Retail Prices
Fish prices, like other retail food prices, were administered by the
head office of the chain. The effects of this fact were immediately ap-
parent by the lack of week to week fluctuation in the prices for fresh
fish (Figure 5-2) . These retail prices were established by applying a
fairly constant markup to a mildly fluctuating wholesale price. Normally,
the retail price from the previous week was not changed unless the whole-
sale price had changed by at least three cents. The one or two cents
fluctuations which occurred frequently in the wholesale price did not
usually result in a changed retail price.
When information from both Figure 5-1 and Figure 5-2 was combined,
some evidence developed for suspecting a slightly higher level of demand
for fish in the late winter and early spring than at other periods of
time covered by this study. Obviously, the existence of seasonal shifts
in demand cannot be established with less than one year's data. However,
it is interesting to notice that the aggregate curve of Figure 5-1 is
somewhat higher after mid-January (week number 29) than before, and with
26
exception of halibut, prices tended to be somewhat higher also (Figure 5 2),
indicating that demand may have shifted upwards.
For instance, the price of haddock increased about ten cents, from
a normal range of $.69 and $.75 for the last half of 1967, to a normal
range of $.79 to $.89 and even $.99, briefly, in the first few months of
1968. Flounder prices made a similar increase from a general level of
about $.79 in the last half of 1967 to $.89 and $.99 in the early months
of 1968. Swordfish also increased, but fresh swordfish was not always
available, and it is difficult to say exactly how the fresh price would
have acted had any been available.
Fresh swordfish was available only until about the middle of Novem-
ber, after which the frozen form of the species was sold. For the dura-
tion of the study the weekly frozen price was adjusted to estimate the
fresh price. Implicit in the decision to make this adjustment was the
assumption that the demand for frozen swordfish was very similar to the
demand for fresh swordfish.
Hie increase from the actual fresh swordfish price in the last half
of 1967 to the adjusted price in early 1968 was from $.95 to $1.06. The
retail price of codfish was extremely rigid at $.65 during the summer of
1967, but rose somewhat in the fall, and eventually reached a high of $.85
in the spring. Halibut had a much more rigid price structure. Its price
rose in the middle of August, 1967, but did not change again during the
study, with the obvious exception of price specials.
"''After trying several alternative adjustment procedures, and finding
only negligible differences in the results, it was decided to multiply
each frozen price by the ratio of the last fresh price observed in ^ the
fall of 1967 to the immediately following prozen price, and use this prod-
uct as the estimate of the fresh price.
27
The increase of prices in the spring were not particularly surpris-
ing in light of the findings of other fish demand studies [11, 24]. Most
all those conducted in the past, which were dealing with more than one
year's data, found a higher level of demand in the spring. In addition,
this is the time of the year when supplies tend to be short, as it is
difficult to fish during the months of January and February, While, ail
the species had their normal early slump, haddock was particularly in
short supply, since the landings at New England ports were down 38 per-
cent at mid-March of 1968 from the same time in 1967 [34, p. 19].
Special Promotions
Price specials, as was readily apparent from the data, were by far
the most important factor affecting sales for each of the five species.
As can be seen in Figure 5-3, the proportion of total sales accounted for
by the single, specialed species was quite considerable. Overall, these
single items accounted for 53 percent of the five-species, aggregate sales
(Table 5-3)- Broken down by individual species, there was a range from
43 percent for flounder to 63 percent for haddock.
Specials were determined with the wholesaler at least a week in ad-
vance, and the chain was given a lower wholesale price on the item to be
featured. A slightly lower markup than normal was then applied to the
lower wholesale price of the featured species, which resulted in a notice-
ably lower retail price. The relative change in price from nonspecial to
special weeks varied from species to species, as can be seen from '.ha
right-hand column of Table 5-4, but in general, the special price was
about 80 percent of the normal price.
28
cu
ucd
60<u
uto
«J
CO
CD
«HCJ
CD
CO
o
QJ
a
<d4JCO
toCD
Nti3
60 COCtJ cU
*HCO CJ
cu a)*H C.CJ CO
CD
cx xjCO CD
u
> u•H ccj
<w a)mmO CD
CD
a
CD
l
CD
U
60*H
60C•HCO
-a
CD
*X3 CD
O MCD
CO cit
c
O
a)
>•H
oo o
1
29
Table. 5-3. Proportion of sales accounted for by thesingle featured species
.
Species Mean % of Aggregate3
Haddock 63Swordfish 62Halibut 44Flounder 43Codfish 44Overall 53
gj— 1 — — 1 " 1 .... —
,
These values are calculated by the following formula:
Mean % of Aggregate = Total Sales of Featured Specie sx 1Q0 ^
Five Species Aggregate Sales
They show the percent of total fish sales (for the five speciesaggregate) accounted for by the single featured species. Weeksin which no species was featured were omitted from the calcula-tions .
Table 5-4. Relative importance of special and
nonspecial prices by species.
Species
Mean specialPrice
$
Mean nonspecialPrice
$
Special Price as %
of Nonspecial Price
%
HaddockSwordfishHalibutFlounderCodfish
.56
.80
.68
.65
.59
.79
1.00.75
.83
.70
71
80
90
78
84
30
When price-quantity relationships were plotted, such as in Figure 5-4,
the importance of price specials became more obvious. For three of the
species, haddock, flounder, and cod, there was a clear break between the
special and nonspecial prices. With swordfish and halibut there was some
overlap between the two types of price. Even here though, there tended to
be two clumps of prices— one clump of high, unadvertised prices, and another
of low, advertised prices
.
With the exception of halibut, there was also a clear break quantity-
wise between the two clumps. Much larger quantities were sold when a spe-
cies was offered on special.
One thing to keep in mind when looking at these data is that the two
clumps do not show the same thing. Both are plotted as price-quantity
relationships, although the effects of other factors are also reflected in
the data. The variables from which most of these effects originate are
unknown, but it is known that the lower clump has the additional effects
of newspaper advertising and in-store promotion built into it, whereas
the higher clump does not.
Newspaper advertising for fish was merely a small part of a larger
store advertisement and informed the public that a certain species would
be sold at a lower price in the week of the advertisement. Although fish
was relatively unimportant compared to some other meats , the specials
might have drawn some customers from other stores. Likewise, in-store ad-
vertising was not expected to be a major demand shifter, but it too must
have had seme effect on attracting shoppers 1 attention to the lower price.
Promotion of this sort normally consisted of a poster above the fish dis-
play which stated which species was being featured and the price.
I
Figure 5-4.
Individual Price-Quantity Ordered Relationships
31
Retai]Price /Pound
C
100
90
80
70
60
50
40
A
Haddock
o c
©©
©© ©o ©
to €5© ©
© ©©
0 100 200 300 400 500 600
Quantity OrderedPounds
700 800 900 1000
RetailPrice /Pound
100
90
80
70
60
bO
40
© oc.'.'oo © o
© © ©c\:c:3&© ©
©o
B
Swordfish© ©
0 100 200 300 400 500 600Quantity Ordered
Pounds
'TK, 800 900 1000
Figure continued on next page
32
RetailPrice/Pound
C
J 00 L90
80
70
60
50
40
100
c
Halibut
300 400 500Quant i ty Ordered
Pounds
900 1000
RetailPrice/Pound
c Flounder
100
90
80
70
60
50 I
40
<j O CO
CO
4\ I- ,. -
0
>v
100 200- (i
300 A00 500 600 700
Quantity OrderedPounds
800 900 1000
Figure continued on next page
33
Re!tailPrice/Pound
<?
100 L90
806
~& OO Ci*
70
60
50
40
L.0 100
Flounder
200 300 400 500 600 700
Quantity OrderedPounds
800 900 1000
34
The effects of competitive species on the quantity of haddock sold
was also of some Interest in this study. It might be expected that when-
ever another species was offered on special, sales of haddock would suf-
fer. Apparently this did not occur. The figures in Table 5-5 show that,
of those observations when haddock was not offered on special3less was
ordered when no species was specialed than when one of the others was on
special. Thus, if specials on other species affected the sales of haddock
at all, it would appear that they increased them.
Table 5-5. Effect of specials of other species onpounds ordered of haddock.
Numberof weeks
Average Poundshaddock ordered
No Species on Special 15 231
Some Species on SpecialSwordfish 5 262
Halibut 9 224
Flounder 6 265
Cod 1 280
All 21 247
Summary
In summary, the primary item of importance noted in this chapter was
the strong effect of price specials on the sales of specific species of
fish. The effect of these price specials was by far the most important
factor affecting the quantity sold of any rpecies. Price fluctuations
within the normal range may have had some effect, but the exact nature
was difficult, if not impossible, to discern without using a more sophis-
ticated technique than that followed in this chapter.
35
CHAPTER VI
REGRESSION ANALYSIS OF FRESH FISH SALESFOR FIVE LEADING SPECIES
This chapter reports the theory and results of multiple regression
analysis used in an attempt to refine the relationships among the vari-
ables. Emphasis was placed on deriving a demand function for haddock.
Unlike the last chapter, separate functions were calculated for each
store. The dependent variable for these equations was expressed in ad-
justed pounds sold figures instead of the previously used pounds ordered
data.
Theoretical Considerations
Simultaneous equations . A multi-variate regression model can be ex-
pressed as either a single-equation model or a multi-equation model. While
consideration was given to the use of simultaneous equations in this study,
the single-equation least squares technique was finally adopted. Simul-
taneous equations are necessary when explaining a set of simultaneous oc-
currences . In terms of econometrics , this occurs when there is more than
one variable whose value is determined within the system, i.e., more than
one endogenous variable. There must then be a separate equation for each
of these mutually dependent variables, since each is a dependent variable.
If, as in the situation encountered here, the model can be specified with
just one ex. ^ ogenous variable there is no need for a simultaneous system
[10, p. 432].
Assumptions of regression . Certain assumptions must be recognized
when a regression model is used. First, it is assumed that the values
36
of the error term of the regression are randomly distributed with a mean
of zero. Second, it is assumed that the variance of the error term is
constant. This assumption of homoscedasticity can be stated differently
by saying that the values of the error term do not increase or decrease
as the value of any independent variable increases or decreases. A third
assumption is that the values of the error term are independent of each
other. If they are not, then there is a problem of autocorrelation.
Fourth, if linear regression is being used, it must be assumed that the
true relationship is linear.
Two additional assumptions are often made also. One of these is that
each of the independent variables and the values of the error term are in-
dependent of each other. This is most easily accomplished if the values
of the independent variables are constant from sample to sample [31, p. 26]
The second states that the error term is distributed normally. If
all the previous assumptions are met, the regression coefficients and the
standard error of estimate are efficient and unbiased estimators. This
final assumption assures the validity of the standard tests of signifi-
cance.
Demand equations are oftentimes expressed in logarithmic form, as
this greatly simplifies the task of calculating demand elasticities. When
expressed in this form, the coefficients of the variables are the elastic-
ities, so they can be read straight from the equations. Certain additional
assumptions must be made, however, to justify the use of logarithms. For
instance, it must be assumed that the true relationships of the variables
are multiplicative rather than additive. It must also be assumed that
the relationships are more stable when the variables are expressed in per-
centage terms rather than in absolute terms. In addition, the unexplained
37
residuals must be assumed to be more uniform over the range of independent
variables whan expressed in percents rather than in absolute terms [12,
p. 37].
Identification problem . A common problem with demand studies is the
identification of the demand function. Particularly in single equation
models, it is often impossible to be certain that the estimated curve is
a demand curve rather than a supply or a mongrel curve [36, p. 104]. If
it can be assumed that demand does not shift over the course of the study,
and it is known how supply shifts, then the estimated curve can be iden-
tified as a demand function.
The identification problem did not cause difficulties in this study.
Data were collected over a forty-two week period, too short to expect:
changes in tastes and preferences, population or other such demand shift-
ers to measurably affect demand. Thus, it has been assumed that demand
did not shift throughout the course of the study- In addition,, shifts
in supply were known. As described previously, the retail price was es-
tablished on a weekly basis by the head office and was not changed within
a week. Consequently, the slope of the supply curve equaled zero." indi-
cating that the weekly supply curves were horizontal straight lines
(S-l,S2, . . . ,85 in Figure 6-1) at the relevant prices. They do not extend
to infinity, however. At the established price each week, the meat man-
agers attempted to estimate the quantity that would be demanded and ordered
39
that amount. This sum of the two weekly orders set an upper limit to the
quantity made available to consumers. Thus, as shown in Figure 6-1, there
is a cutoff point for each of the supply curves (Tj , T2 , . .. , T5)
.
For each week there is also a demand curve relating prices and quan-
tities. At the specific price established on that week, the quantity
which was demanded can be read off the demand schedule. If the manager
estimated correctly when ordering the speciessthe quantity demanded 00-
incided exactly to the quantity he ordered. This situation is illustrated
by the intersection of supply and demand at (P^Q^). If the actual quan-
tity demanded fell short of the manager's expectations, the demand curve
would intersect the supply curve at some point to the left of the quantity
ordered (P2 >C>2 Q/t
^5^5) • *f tne quantity demanded was greater than
the manager's expectations, the demand curve would intersect the hypo-
thetical extension of the supply curve to the right of the quantity or-
dered at (P1 ,Q1
)
.
Connecting all these intersection points of the individual supply
and demand curves yields the curve UD. Since the supply curves are per-
fectly elastic, the curve DD can be identified as a demand curve, even
when estimated 'by a single equation model.
The Model
Functional relationships. Variables considered important and capable
of being w isured were expressed in a function as follows:
Y - f(Xi,X2> X3P,X4 ,X5
,X6,X7
,X8,X
9,D
l5Sl9D2j S 2 )
Where:
Y = The quantity, in pounds, of haddock purchased by consumers.
X3 = TimeX? - Retail price/pound of fresh haddock fillets.
40
= Retail price/pound of fresh flounder fillets.X, = Retail price/pound of swordfish steaks.X^_ = Retail price/pound of halibut
= Retail price/pound of fresh codfish fillets.Xy = Total store sales.Xg - Quantity, in pounds, of shellfish sold.Xg = Quantity, in pounds, of other finned fish sold.
D-jl- Dummy variable for haddock price specials.
D-^ = 0; No price special.= 1; Price special.
S-j_ = Slope dummy variable for haddock price specials. It equals theprice of haddock that week (X2) times the dummy variable (D^)
.
S^ = 0; No price special.S-^ - X2; Price special.
T>2 - Dummy variable for a price special on any of the other fourmajor species.
D2 = 0; No price special.D2 = 1; Price special.
S2 = Slope dummy variable for price specials of any of the fourmajor species other than haddock.
S2 = 0; No price special,S2 " Percent deviation from the mean price for the past
four weeks; Price special.
Other variables that may be substitutes for fish in the market include
other meat items and prepared frozen, fish items . However, there was no
basis for determining which of these items to select and which to omit.
In addition, statistical problems arise as the number of variables ap-
proaches the number of observations.
Des cript ion of -variables . All variables were expressed as weekly
observations. The dependent variables were expressed in pounds sold,
and were adjusted from pounds ordered figures taken from store receipts.
Weekly store forms (See Appendix B) filled out by the meat managers were
used to determine how much of each species was left over each week, or
if the store ran out of a given species. On those occasions when fish
was left over, the excess quantity was subtracted from the amount or-
dered to obtain the quantity sold. When the store ran out, the time
that it ran out, taken from this form, was combined with information
about the normal intensity of sales from the manager questionnaires to
estimate the amount of sales that should have been made. These adjusted
quantities would fall on the weekly demand curves and are analogous to
the quantities Q1 ,...,Q5in Figure 6-1. For the relevant days of the
week, primarily Fridays, but also Wednesdays and Thursdays for some
stores, the rate of sales was broken down into three hour periods (see
Appendix A, part B3) . Within these three hour periods a constant rate
of sales was assumed. Although these rates of sale were only the man-
ager's estimates, it was felt that they would provide a more realistic
estimate than that which could be obtained by assuming the rate of sales
was constant throughout the week. With this information, only a simple
mathematical calculation was required to obtain the quantity sold. The
number and mean size of adjustments for each store were quite variable
as can be seen from Table 6-1.
Table '6-1. Haddock adjustments for five stores.
No. of Weeks AverageTotal Wjth Adjustments Net Adjustment
Stores No . Weeks Runout Surplus lbs .
A 36 4 6 13.0
B 24 8 11 13.1
C 27 1 20 11-2
D 30 0 13 9-5
E 25 . 3 19 5.9
Time was introduced into these equations simply to see if the higher
level of prices in the spring, noticed in the preliminary analysis, had
any significant effect. The variahle was given values by consecutively
numbering the weeks in which data was collected.
The retail prices used as independent variables in the equations were
either the price of the species under consideration, or were the prices of
close substitute products. Swordfish, halibut, flounder and codfish were
selected as substitutes of haddock because they were displayed next to
each other in the same meat case, and they all sold in relatively large
quantities. The prices for haddock, flounder and codfish were used di-
rectly as quoted from the chain. This could be done because the fresh
form of each of these species was sold throughout the year.
Such was not the case for swordfish, however. Fresh swordfish was
available from Ju]y through mid-November. For the remainder of the study
period- only the frozen form of the species were available. In order to
use all the data it was necessary, therefore, to adjust the frozen sword-
fish price to obtain a fresh price equivalent . The procedure fol lowed
was to multiply each frozen price by an index of the last observed fresh
price divided by the first observed frozen price. Since the index yielded
a value slightly greater than one, the fresh price equivalent was slightly
higher than the frozen price, as would be expected.
Adjustments were made to the halibut price a] so, but for a different
reason. Two forms of halibut were sold at the various stores included in
this study — fillets and steaks. Depending on the particular store,
either one or the other, or both taken together were important. To have
halibut included in the study, it was necessary to combine the two forms
43
and use just one price for both. This price was calculated by taking a
weighted average of the prices for the two forms each week. The weights
used were the mean pounds sold of each form over the course of the study.
Each store was weighted separately. On weeks when one or the other form
of halibut was not sold, the mean price for that species was weighted.
Other finned fish species were also sold, but either they were avail-
able on a sporadic schedule, or else they were not sold in all the stores.
They were accounted for by the finned fish variable in the equation. Since
the same species were not sold from week to week, it was felt to be im-
proper to construct a price index for this variable. Consequently, the
total quantity of all the other finned fish species was used as a vari-
able in the equation in the hopes of explaining some of the variation in
the dependent variable. A shellfish variable was constructed in a simi-
lar manner and for the same reasons.
Total store sales were included in the equation as an indicator of
customer volume and purchasing power. It seems logical that when more
customers are in the store, more fish is likely to be purchased. Since
variation in the dependent variable caused by this factor could not be
considered random, this variable was introduced into the equation.
Four dummy variables were used to distinguish between price special
and nonspecial observations. They were used to check for changes in both
the constant term and the structure of demand. If there had been enough
observations, the same information could have been obtained by running
separate equations for the special and nonspecial situations. However,
44
in this study, there were not enough observations to run separate equa-
tions for weeks when there were specials. Some stores had as few as •
four or five price special observations to explain eight variables,
which is impossible.
The first haddock dummy variable, Dj, was used to detect any change
in the constant term resulting from the effects of promotion in price
specials. In a simple regression model, the constant term is the Y-inter-
cept and it may be easier, conceptually, to think of it in this manner.
If the coefficient of this zero-one variable is not significant when tested
by the Student's t-test, then no change is indicated in the Y-intercept
as a result of price specials. If the coefficient is significant, then a
shift is indicated. This variable then, is useful only in detecting change
in the level of a regression line. It says nothing at all about the slope
of the line. Of necessity it assumes the structure of demand (slope of the
curve) to be constant.
In order to test for changes in the structure of demand, or in terms
of simple regression, the slope of the curve, a different type of dummy
variable was constructed. The one used in this study, S-^, was discussed
in Goldberger's book [15, pp. 224-225] and is in fairly common usage. It
was constructed by multiplying the price variable, in this case the price
of haddock, by either zero or one. In this study, an appropriate zero-one
variable was already available, D]_, so it was simplest to think of multi-
plying the haddock price by this dummy. The result of this process was a
dummy variable with the values of either zero or the price of haddock.
If the coefficient of this variable is significant, it indicates a struc-
tural change in demand; if the coefficient is not significant, no change
i
in the structure is apparent.
The other two dummies in the equations x^ere based upon the same con-
cept, but were used to check for differences in either the constant term
or in the structure of demand resulting from price specials of any of the
other major four species. Ideally, two dummies would have been used for
each of these other prices, but this would have introduced too many vari-
ables into the equation. Therefore, one set of dummies was used to test
the general effects of price specials on competing species.
The only major difference in the construction of these two variables
occurred with the slope dummy. Since prices for four species, instead of
one, were involved, the variable had to be expressed in some sort of stand-
ardized units. To obtain the values for this variable, a mean price for
the species on special was calculated over the previous four weeks for
each store. Then the deviation of the current week's price from this mean
price was calculated. The actual value used as data was this deviation
expressed as a percent of the mean price. By using percentage deviations
the data were expressed relative to some base-- the general level of that
price. Its use, instead of the use of absolute deviations in cents, as-
sumed that consumers would respond more, strongly to a ten cent decrease
in price on a species that was normally priced at fifty cents a pound than
to the same decrease in price on a species that was normally priced at
eighty cents a pound.
Expected results . Some preconceptions could be formed about the re-
sults of the regressions before the equations were run on the computer.
The most: obvious was the expectation of a negative relationship between
the quantity of haddock consumed and its retail price. In addition, from
46
the results of previous studies, it was expected that demand would be
price elastic. Since the other species in the equation were competi-
tive with haddock, it was expected that their prices would be positively
related to the quantity of haddock consumed. As the price of one of
these competing items rose, consumers would be expected to switch to
relatively cheaper substitutes, such as haddock.
The signs of the shellfish and "other" finfish variables were ex-
pected to be negative, however, since these were quantity variables.
The species aggregated into these variables were also assumed to be com-
petitive with haddock, so as greater quantities of them were consumed,
less haddock was expected to be consumed.
A direct relationship was expected between total sales and the
quantity of haddock consumed, since this variable was a measure of the
volume of customers and their buying power. As either volume or buying
power increased, it seemed logical to assume that more haddock would be
purchased.
The two haddock dummy variables were expected to have positive signs
because of the. way they were expressed. Both had a zero value when there
was no price special, and either a one or greater value when there was a
special. Since their value was greater when there was a special, and the
quantity consumed was also greater when on special, a positive relation-
ship was expected. The other two dummies were just the opposite. Again,
the values were zero for the nonspecial situation, and one or greater for
the special situation, but this time the specials were for competitive
species. Consequently, when one of these species was offered on special
a greater amount of it was sold and it was expected that a smaller amount
of haddock would be sold. The sign on these dummies, then, was expected
to be negative.
Procedure of estimation. Until now, the analysis in this study has
been based on aggregate data. In this chapter, an attempt was made to
analyze each store separately and then test to see whether the separate
regressions were significantly different. The coefficients for the re-
gression equations were calculated by using a stepwise regression pro-
gram on a CDC 3600 computer at the University of Massachusetts Research
Computing Center. The particular program used was one of the Bio-Medical
programs, originally developed at UCLA, then adapted to the University of
Massachusetts facilities. With this program, independent variables are
added one at a time in a stepwise manner. At each step the variable with
the highest F-value of those variables not yet in the equation is entered.
No more are added when, of those variables left, none has an F-value above
a minimum value specified in the program. At each step the coefficients
of the regression equation are completely recalculated; they are not
merely modified from the previous step.
Since separate regressions were calculated for each store, some means
of statistically testing them for significant difference, or lack thereof,
was necessary. Chew's test for equality of sets of coefficients in two
linear regressions was selected for this purpose. Unlike other possible
tests, this one tests the whole regressions, rather than each individual
coefficient. It does, however, only test two equations at a time. Since
there are five stores, with an equation for each, a minimum of four com-
parison tests should be run to see if stores B, C, D, and £ are statis-
tically the same as store A. If one or more were not the same, it is
48
possible that the test would have to be run as many as ten times before
each store would be tested against each other store.
The compulations for the Chow test are fairly simple, and the nec-
essary information is easily obtainable from the computer output. For
the purposes of this study, the test can be described in the following
manner [18, p. 137]
:
Q2 /m+n-2k
where:
Q-j_= Sum of squared residuals from the pooled equation.
Q2~ Sum of squared residuals from two separately computed equa-
tions.
Q3 = Qx - Q2k = The number of independent variables.m = The number of observations in one sample.n = The number of observations in the other sample.
Once calculated, t?iese F-values can be compared to table values in the
normal manner to test for significance.
Equationa l f orms . Some preliminary work was done with the variables
in logarithmic form when the regressions were first run. There was some
doubt that the data fit the assumptions, however, and when it was found
that higher coefficients of determination were obtained with the data ex-
pressed in absolute terms, further experimentation with the logarithmic
foxms was discontinued.
Results
Several variations of the linear and additive form of che equations
were tried, although only a fexv- will be discussed here. They were: (1)
all variables allowed to enter the equation freely if they met the re-
quirements of the stepwise procedure., (2) price of haddock forced to
49
enter, and all the consistently insignificant variables omitted, (3) had-
dock slope dummy and all the consistently insignificant variables omitted,
(4) both haddock dummy variables plus the consistently insignificant vari-
ables omitted, and (5) only the nonspecial data used with the consistently
insignificant variables and dummies omitted. The remainder of this chap-
ter will be devoted to explaining each of these forms more fully.
Model number 1: All variables included . When all the variables were
allowed to enter the equation freely, high R !
s were obtained, as can be
seen from Table 6-2. These coefficients of determination were adjusted
for degrees of freedom and ranged from .70 to .84. They were also all
significant at a 90 percent confidence level, as measured by the F-test.
Thus, the explanatory power of the equations seems to be reasonably strong
While this is certainly of interest, the significance of individual vari-
ables is of even more importance when attempting to learn which of several
factors affect the dependent variable, and which do not.
Of the thirteen independent variables included in the equation for
store A, only the price of svordfish was significant at the 90 percent
level. Of the remaining species which were substitutes for haddock,
flounder and Halibut showed some importance, but their t-values were not
2When riot adjusted for degrees of freedom, R2 is biased upwards, i.e
it makes the equation appear to explain a greater amount of the variation
in the dependent variable than it actually does. Therefore, it was ad-
justed to tho unbiased estimator R2 , wher- R - l-(l-RZ ),(n-l) ;n = nu ,ber
n-k-1
of observations, k - number of independent variables [31, p. 80]. It was
particularly important that the adjusted coefficient of determination was
used because equations for the different stores were calculated from a
different number of observations. Therefore they had a different number
of degrees of freedom.
50
CMI
'O
CD
rHX>cd
H
, |in ni U)crt li
r-»
4.J oo 4--'
C ' C/3
i i
• Q) -I-J
to / )
rjt—i *H 4-H no j-i o rHtH p_i
CD
CO
CD
^1_cl
a) CD CO• CJ *Hd LTj •H 4-1 4-1
o o -[—
t
*H oCO 4J
cd
Mo4H CD
COCO u CD •Hcd CD cj 4-1
4-1 •H 4-1
r-J, ocd CD r—
H
o>
O CO4_)
COcu CD
4_j CD
u CD CD
rH cj XJCO •iH 4H rj
cj CO o•H CD p_, o4JCO f24
•H cd
Hcti
4J cd -id
CO j> CDcj o
x» rH CN 'H md iH U o X)cd <^ Ph cd
CO4-J • •
d tHCD
•H • rHCJ O•H & H4H4-1 r—
1
CD CD*
O XJ 4-»
CJ O
OO LT)
m oo
OA
<r inOA r-*
• •
I
CJ
CN <fO COO OA
vO CNoa
CO LT)
oa <rm cn
o OAoO rH
oovo o
* •
CM rH
CN OAiH
CO CN
to O CMCO OA
rH CO
,—
1
C\J i i 1
CT\ CO<J U) UJ < )
"1
* • • • • •
11
1
CJ
CO O mm o <j m CO OACO o CM rH
• • • •
i rH1
CM rH1
rH
CJ
OA o CNCO CO o m OArH CM CN OA CM
• • • • • •
1 rH CM
CO COCO COo \D
• • m
r-- CNUO OACN CM rH
1 1 1
d C cO o o*H H •HCO CO CO
CO CO CO
<D CD CD
U u00 4-1 00 4-1 00 4-J
CD d CD CD a CD CD a CD
0) CD Cri CD
•H rH •H rH H rHrH CJ cd rH U fd T 1 CJ cd
cd *H > cd 'H > Cd *H >•H i
H 4-1 1•H 4^ 1
4> m 4J 4-t J-J 4-)
CD CD i-l CD
rj O Cd o cd oU Ph u U u
o moo co
m mCMm rH
• •
I rH
rHO CO
CO COCO COCM O
CM COCO CMCM O
O-N
00
ao•rH
01
COCD
u00 4-»CCD
CD 0•H rH
cd
cd
>•H 4H I
4J IM 4-»
>-) CD
Cd On CM U
H mO vDo
coin
rH CMrH
CO OA• •
I
O CNCM I^-
cm <r
m mCO OA
oo o^CM COH CO
OAo
o•HCO
0)
CD
u00cd dp£i CD "J
•H rHHun]
CD
cd -hH 4H4.-1 4-J
cu ow p^ cj
CD
d•H4-J
dou
51
rrtCO
CMIf*
CM
a) !>cu
CO O £rH rH 3o 00 Q
u rH >-CU CU E
CM > eH 4-J <U 2o hJ Q
cj <U >o fx Es
rH Tj O irH rd rH BCO CO QCd
o rH >o a) eO > gH cu 3CO q
CO._Jn
4J C^•H4->
d O 0)
co pi
c-H
rd4-J co
•H •H00 4->
a O HCO rH
CD
rCCO
o
CO
aCNH
0"\
CO
CN COO^ rH
\ rH
o or-s or- rHCM
OrH VDO <i
• •
O") rH
r> com m
oCN]
I
com coO CM
00 Oo
• •
Pi
o•Hwcri
CU
u00 4_)
a) pipci cu
•HiH OfU *H*H 4-!
4-J 4-4
M CU
CO OP-! O
a)
r-l
CO
>I
00
CO
ocm
rH VD
OCM
CNJ
aCM
CM
m cm
o cm
m
0o•HCO
CO0)
CO 4-J
M cu
rH CJ
CO *(H
•H 4-1
4-J 14-1
Vj cu
CO OCQ Pm O
oCOrHCO
•
O
lOCO
orH
CO
< CMoo thm cm
• •
m
CO <f00 t-H
• •
00 rH
co
CM
ain
vO
00
r^. com cr>
O rH
I
<T rHin cmO rH
o oo orH <±
• •
rH CM
\0 orH r--
<r cm• •
CT» rH
I
oo mrH 00
OCO
CO
CO COCM u"J
CM CMoo-d- cm
I
m inr^. m
co
OrH CMO
G\ rH• •
CT> CMm
00 CMin coCM 0>
I
CO m rH CO cr»
CO CM m rH o>m CO CO O CO O
• • • • • *
rH rH
a C co o o•H •H •HCO CO CO
CO CO CO
(U CU CU
u u V-i
CO 00 4-J 00 4-»
CU <u a cu CU a CU cu a CU
3 <U pa (U rj P4 CU r3
rH *H rH •H rH •H rHcO rH CJ CO rH CJ CO r^ a CO
> CO •H > CO •H > CO •H >1
*H IW 1 H 14-11 -H m 1
4~> H-l 4-J U 14H 4-1 •u) 4-1
M CU H QJ M cu
CO o CO O cd oa p- CJ C ) a
CO
cu
rHrQCO
*HUCO
>
a<u
cu
a,cu
•H
iHO
UCU
e
CU
r£3
CJ
CO
CO
C3
O•H4J
CO
>uCU
CO
o
MHO
a)
%
<U
V-i
a)
CMP^
I
4-1
orj
o cu
•H4_) rHcO CO
>—
•
>r*H4-» CO
CO 4-J
<u *H
*-rJ
<U CU
CO •
cO •H•H
•s
aoH4-J
cO cO
0<H CU
CO
(U
o 4-1
o4-J cu
4-»
O cu
X)CU oCU 4-J
>-4
4-1 4J*H
4-1
Oo
CO 4-(
CU
CU x:U oo0/TM CJ
CU oCJ
cu
CU aX! o rc4J •H 00
4J •Hr>> CO
4-J
cu ocu4-J cu
CO CO4J cO
•n ^
rO o (U
rHCO u ,Q
CO, _j
4-J o M•H Pm cO
>O >^CO u CO
o iH4J r^J
CO CO 4-J
-a CS
V4 ?J CJ
cO rH oCX
a. X cu
d cu pi
CM
'd co dOJ cu > JCJ
CO 4-JJ
4-.'
iV) CJ pen
•H CJ CO
^ fX <U CO
M tfi CU
01 fl) £J H
iH jocu co
CM CO
P£j4J 5:
>CU
H
a<u
aua)
ao
4-)
4-J
co
cJ
co
a•H4-1
•Hrj
CO•HCOa
52
significant at the 90 percent interval. The signs of the coefficients of
flounder, swordfish and halibut were all positive as expected.
The price of haddock was negatively related to the quantity of had-
dock, but its coefficient was not significant at even a 50 percent sig-
nificance level. Some blame for this lack of significance may lie with
a problem of multicollinearity among the variables. The simple correla-
tion coefficient of the haddock price with the haddock level dummy vari-
able was -.731, and its correlation coefficient with the haddock dummy
variable was -.705 for store A (see Appendix C, Table 1). Similar inter-
correlations were observed for the other stores. As a result of multi-
collinearity the standard error of each variable involved expanded. Since
the t-statistic was used to test each variable for significance, and
t- = Partial Regression Coefficient , , , . ,——— as the standard error increased, theStandard Error
t-value decreased, making it more difficult for the variable to show sig-
nificance.
The reason for this high level of correlation is not particularly
difficult to see either, if one looks at Figure 5-4A again (p. 31).
There is no overlap between the two distinct clumps of prices shown
there. Whenever the dummies have a value of zero, the haddock price
variable has a value from the higher clump, and whenever the dummies
have nonzero values, the haddock price variable has values from the
lower clump. Since this is true, and since the lowest of the high
prices is higher than the highest of the low prices, a high degree of
correlation must result. The variables are not perfectly correlated
because of the variation of prices within each of the two clumps.
53
None of the remaining variables were significant. It should be of
interest, however, to point out a recurring problem of multicollinearity
between the haddock dummy variables. The simple correlation coefficient
between them was consistently close to one, indicating almost perfect
correlation. The cause for confusion here is not particularly difficult
to see either. Both dummy variables had a value of zero when haddock
was not offered on special, and a value of either one or the special
price when it was on special. There were only four to eight observa-
tions of price specials, depending on the store, and these were the sole
basis for any variation that did occur. On those occasions x^hen the had-
dock level dummy took on a value of one, the haddock slope dummy took on
some value within a sixteen cent range about a fifty-seven cent mean. On
all the remaining observations the dummies were perfectly correlated with
each other, both having values of zero. Evidently the amount of variation
that did occur on the four to eight price special observations was not
sufficient to distinguish between these two variables.
In the regression for store B, three of the variables were signifi-
cant — the price of haddock, the price of swordfish, and the haddock
level dummy. The signs on each of these variables conformed to expecta-
tions, with the price of haddock being negatively related to the quantity
of haddock sold, and the other two variables being positively related.
The dummy variable, with a coefficient of 45.3 as compared to -1.8 for
the price of haddock and 1.3 for the price of swordfish, exerted the
greatest influence on the dependent variable.
Multicollinearity did not arise as an important problem in the data
for this store., as the simple correlation coefficient between the price
54
of haddock and the haddock level dummy at -.652 was somewhat lower than
for the other stores. There was no problem between the two haddock dum-
my variables because the slope dummy explained so little additional vari-
ation that it was not included in the equation by the stepwise procedure.
In store C, time was the only significant variable. Since this vari-
able was not significant in the equations for any of the other stores, its
importance in the equation for store C attracted some interest. The cause
lay in the data, however, as the quantities of haddock sold in the spring
were noticeably greater than the quantities sold previously from store C,
whereas there was not such a noticeable trend of this sort for the other
stores
.
Neither time nor the price of haddock were significant enough to
enter the equation for store D. Of the remaining variables which did
enter, only the haddock slope dummy variable was significant at the 90
percent significance level.
The equation for store E had similar results. Only the haddock level
dummy was significant, while the price of codfish failed to enter the equa-
tion at all.
It is apparent from Table 6-2 that some of the variables were not
significant in any of the equations. In order to salvage some degrees
of freedom these were omitted from the following models.
Jl^l!llJlHILbclLJ- : Significant variables only . This model differed
little from the previous one, except that the insignificant variable?
from that model were excluded from this one. Specifically, these were
the quantity of shellfish, the quantity of finned fish and both of the
"other" dummy variables. Time was omitted because it was not a variable
55
of major interest in this study, and with one exception had very little
explanatory power.
The remainder of the variables were retained and used in the equa-
tions of Table 6-3. These included the prices of the five major species,
which were of particular interest for the calculation of cross elastic-
ities, the haddock dummy variables, which were highly significant when
not upset by multicollinearity , and total store sales. This last vari-
able was retained because it had shown moderate importance in various
preliminary runs
.
All the equations of Table 6-3 were significant as measured by the
F-test for the whole equation at the 90 percent significance level. As
in the previous run, each of the equations showed a high level of explan-
atory power, with R2
1
s ranging from ,70 to .88. The significance of in-
dividual variables followed a similar pattern also, although not an iden-
tical one.
In store A, the price of swordfish was significant at the 90 per-
cent level again, but now it was joined by total store sales. The had-
dock slope dummy was close to being significant, but as a result of mul-
ticollinearity, was not quite.
Store B provided a much more interesting picture. In this model,
five of the variables were significant. Only the price of flounder, total
store, sales, _^d the haddock level dummy Tailed to show significance, 0re
last of which did not enter the equation. As expected, the price of had-
dock was negatively related to the quantity of haddock sold, while the
prices of swordfish, codfish and halibut were positively related. The
haddock slope dummy was also positively related.
56
QJ
CJCO cm co rH CO rH t—\
cd P. QJ o <j- o r-"> ("T\
4-> O rH O H o o Qo 4-J cd • • • • •H CO CO CM rH rH
4-> CJ• cu in co r-- CO
CO a rQ CM r->. o\ LOp vo . i 4-1 •H LO <t rH 00 rHo n O t—
1
• • • • • •
•H Ph cd CM rH rHCO PdCO
CD
U aCO o rCj OQ) •H QJ CO O LO O rHu 4-) u •H H O CM CJ\ o vO
in 4-1 4H rH CM CM rHX o 13 • • • • • •
•H o< Pm o I CM CMCO CU ;_:>
no a)4-1 4-1 rd
rd CO o aCO a) •H <f in OCLi a 4-1 CO CO rH
o <r •H m TJ rH co COrH 4-J o M • • • •
cd Ph O iH CM rH CO rH> QJ £s
qj CO4-> uCO 4-4
oj
CO M<U QJ o> a> CO O CO \0
rH rH QJ LO CN rH O CT>
O LO rH co r-*
a cd co •H 4-1 • . • • • • •
•H •H o o rHP Ph rHCO cd Ph•H >4->
cd
4-1 a ^ OCO cd QJ O O CO CO CM CM CM
CJ a O CN CO -d- co LO rH•H Csi •H 4H CN CN LO LO
cd 4H M o X) • • • • •
cd t-! Ph CD 1rH CM
Pd 1
CO co•Hco
cu QJ
I•H T—
|
| I
a • •1 1 1
•H CN H4H4H •
0) OCOo S3 I
s- <rLO o to
T-i cd Cvl CO LOa) 4-J • • •
• CO rH LO
o Hi CO r-*
1 S o rH CO CM
<o 1 I \
cd
o•Hco
CO
0J
i-i co
CO 4-J
cu rd
p4 QJ
HrH CJ
cd -H•H 4-J
4-1 4-4
M QJ
al O
CUrd
rHcd
>I
4-»
cd
oiHCOCO
QJ
M CO
00 4J
cj rd QJ
QJ ^•H rH
rH O Cd
Cd *H•H 4H
V4
cd
co. P-^ U
>1
4-1 4-J
0)
O
co»HCO
CO
0)
>H CO
CO 4J
a) a oj
Ph qj
•rt rHH CJ cd
H >cd
•rl 4H4-> 4-4
U QJcd o
O Ph CJ
O LOO LOO r--
r-. loo
<f CM• •
I rH
co
CO LO
co<r loCM CO
'sO LOLO KOCM CM
<r coO rH
• •
I
LOCT»
CM•
OCM
§*HCO
CO
a)
M CO
'iO 4J
Q) Cd
Pi Q>
•H
cd
•H 4-1
4J 4-i
U CD
cd op p- o
o cd
H >
a
o VDO r-.
• •
r-4
o <rCO LO
O CM
CO
CM COLO 0^CO rH
• •
vOLOCM
o
O lo
LOo
CO
rd
o•HCO
CO
QJ
U CO
00 uQJ cd OJ
Pi QJ
*H rHrH O cd
cd •H >•H LH 1
4-J 4-1 4J
Vh QJ
CO OP. U
QJ
0a•H4->
rd
ou
cd
CM
P4
M CD01 CX
mmCO rC O i
iiH 4-> iHo CO
u iH0) CD
CM ^1 >I iH +J CD iO J p
aso a)
o CUt—1 x) o itH nd
m CO p
oO CD eo > §co ^ npel
(0
4-1 Pq•H
as u ua o <u
^ a
U CO
•H -H00 +J inOH
cd H2 a1
o-CO
fD
Uo
oCO
oCM
CN
CO
m asm Os
• •
CN H
CO MDas
oo co•
COCO
I
o•HwCO
CD
H CO
00 4-1
OJ aM CD
H Hcd
>•H I
4J tH 4-J
S-i CDoj O
< u
oCO
CO
oco
s£)
00
o-d- coCO >sD
ON UO
CO
oCMHUO
•
a*
o
co
cj
r-* coHm o
00
cJ
o o•H •HCO CO
CO CO
CD CD
M co M CO
50 4-> t>0 4J
CD a CD CD dP^ CD PS CD
*H H HH CJ cd occj •H r> cd •H•H UH i
-H >H4-J 4--! u mU CD M OJ
cd O cd oO O P-t u
CJm
inCM
ooCO
On
CJ
00 oh mOs O
• •
H CO
cmas o
I
CJ
asvooo
•
CM
O00
oo
O H<i- ooOn OS
CJ
h as
O H
ao oH H
CO
CO CO
CD CD
H to M CO
00 4-> CO 4J
CD CJ CD d) d& CD CD
•H H •ri
H O cd H acd •H > cd H*H M-4 l H *w4-J M-4 4~>
W CD u CD
O cd OP-l a M p^ U
cden
CD4-) oo 4J
CJ4-J CD
o Ho X)«4H cd
•H#» M
CM1
va
v£>
CO •
CD H HH XI CD
rP 4-J >cd CD
HorH
CD «+-4 4-»
CD CJ
CO CD
60 CJ
o H0 CD
enp.
m oo j=
00Os
CJ H CD
o*H 4->
4J 4->
cd Od CJ cd
cdH CO 4J
C*« cd dX cd
CD oCD
acd •H
"cd a> 00
o i •HCO
58
The equation for store C revealed only the haddock level dummy as
being significant. For store D, only the haddock slope dummy showed sig-
nificance. In the equation for store E, the haddock level dummy was
joined by total store sales, these two being the only significant vari-
ables.
Virtually no new information was gleaned from the model presented in
Table 6-3. The major issue was still that of a high level of multicol-
linearity among the haddock price and dummy variables. In an attempt to
circumvent this problem, the next model to be presented omitted one dummy
variable from those variables used in model number two.
Model number 3: Omit haddock slope dummy variable . In this model
(Table. 6-4),only the haddock level dummy was used to account for special
promotions of haddock. Either dummy could have been omitted, but the
haddock slope dummy was doing little more than describing shifts in the
level of the regression plane, so it was the one which was dropped.
The equations for all of the stores were significant, as measured
by the F-test, and all explained a relatively large portion of the vari-
ation in the dependent variable. The R f
s ranged from .70 to .83.
The haddock level dummy variable was highly significant for each of
the stores in this model, indicating that price specials had a strong
influence on increasing sales of haddock in all the stores. The remain-
ing variables used in this model reacted exactly the same as in model
number two. The variables which were significant for each of the stores
there were also significant here.
While the elimination of one dummy variable solved the greatest mul-
ticollinearxty problem, there still existed a problem of this nature betvecr
59
a)i CO CM COGJ O rH
4-J o rH o oo CC • •
H C/j CM
4-»
* CU
co o CM IT}
0 •H Mh •H LO <to M o rH * •
•H P-i cd
CO
CO
CU
uOC r£CI) a) CO cm inu o «H CO rH
in •H <4H 14H
X cj M o T3 • •
4-» Ph O l
CO 4-1 CJ
uo oM-t
CU CO
CO rH CU •H cr\ in0) a 14-4 CO rH^ cd <r CM -3"
H •H O ^ • >
Ct) O rH CM>
> C/D
4->
CO
CD4-J
H nj <u rH CMcd a m oo 0)
CT 1 MH • •
-H O o rH4~» O P^ rHto t—i Ct.
«H CO4-)
cd Mu o r^CO o 0) G CO rH
X) o o 0\ CMCM H O rH
CO M o • •
PL, cd
co
4-» •
Ci co0> CU
•h • rH e 1
o o •H 1
•H H4h4h rHCU 0) 4
o 4-»
O mcd CO4-1 •
• CO00
o•HCO
COGJ
U CO
0C 4J
cu C
rH CJ Cd
cd -h >•rH 4H I
jj a_i 4-)
a? O
rH OOO rH
O• •
CM rH
<y\ cm
o• •
rH CM
coin o>
in•
rH CO
CO iHOCO vD
rH O\D CO<t~ CM
• •
rH CMI
o
coI
o•HCO
CO
CJ
u co
to -t-»
CJ C
•H MH4J 14-4
cd oCQ O
o o>o o
CO \Din
0> rHo oCM rH
<r o<f rH
co
co vo
CM CMin rH
COinn
•
CMI
COmHCO
CO
QJ
U CO
03 4J
a) a aj
•rl rHrH a cd
>I
4-J
Cd -H-H M-l4J 14-4
m a)
cd oU Ph u
OO coo
a> coco r-.m cm
• •
I rH
<r co
CO CO• •
I H
o inco nco in
o <ro-\ orH co
m ocm co
cocoin
•
CO
o-rH
CO
CO
GJ
U CO
00 ^gj a cu
P^ GJ 3H rHrH a td
Cd 'H >*H m I
4-i IK J-»
V) GJ
cd op f^l "
r-H VDO 00O CTn
in incm <r
CO ocm in
<J- rHin oco CM
CM COCM CMCM CO
* •
I
<r mCM vO
rHinm
•
CMCO
o•HCO
CO
CU
u CO
on 4-j
gj aP^ GJ
HrH OCti "rH
•rH 4H4-J 14-H
0J
Cd oW P4 O
TJGJ
cH4J
CS
OCJ
\
id
CMI'M
CM
J-i QJ
CD Pico ,C OrH 4-> rHo C/3 jP
o qi e
HumsO i-q P
o QJ >>
c P.XJ O eHcd CO Qca
O H >-O CIJ £O 13 > £H ^ QJ ?
CO
4H •!-;
LHH-H
ano
dd ao* H
pi4
>^ ,£3
u CO
•rl •HU 4Hd O r-H
cd rHQJ
O'J
roCOoCMCM
rHGO
moo
CO rHCMco r-v
• <
LT) loO
m*
ro
CO
iooo
m oo <f
• •
<r com
g 6*H •HCO CO
CO CO
QJ QJ
u nt>0 4-J 00 4-J
CJ QJ QJ dP=J ie
rJ
rHPi ie
H a CO rH aCO •H Cd •H•H M-i ]
•H M-)
U MH 4-J !4H
14 QJ M QJ
Cd O cd oa, PQ
CMrHm
o
oo
co
m o• •
co
o•HCO
CO
a)
n60 4-)
pi QJ
Hcd
H 4H4~» m>•( QJ
cd o
a cd
CM
oCM
COCO
CO
cocm oo vo
* •
CO
doHCO
CO
a)
60 4-)
qj dPi QJ
rHCO
LHIHQJO
XCMO
oCO
CO
Xco <rCO COLO to
ioLO
a u Q Ph u
do•HCO
CO
QJ
r4
to 4->
QJ d QJ
Pi QJ CJ
H rHO cd
cd •HH <4-J
4-J MH 4-J
U QJ
cd Oa
•
cd
QJ
4J
Od4J
Oom
CM
vO
QJ
rH
cd
HQJ
•
rHQJ QJ
CO >QJ
0. rHCM
OO a>
d QJ
o x:•rH 4-J
4-J
cd 4-J
d cd
cd
rH 4->
a. dX cd
QJ CJ
•H0 14-1
cd Hd
M 00o •H
C/J
cd
61
the price of haddock and the haddock level dummy. When the haddock price
variable, was forced into the equation first, it was highly significant.
Upon the entrance of the haddock dummy variable, however, its coefficient
would diminish considerably while its standard error would diminish rela-
tively little, yielding a much smaller and sometimes insignificant t-valu
If the dummy entered first, and the price of haddock entered later, the
dummy's standard error would increase by two- or three-fold in many cases
While this was not enough to cause the dummy variable to become insignifi
cant, it did indicate a continuing problem of multicollinearity . Conse-
quently, both dummy variables x^ere omitted from the following model.
Model number 4: Omit both dummy variables . With both dummy vari-
—
o
ables omitted, the R s for the equations dropped to the 50 percent range
but all of the regressions were still significant (see Table 6-5). For
the first time, the haddock price variable was significant in each of the
equations. This was to be expected, however, since this price variable,
also accounted for the effects of price specials in this model. By elim-
inating the dummy variables from the equation and fitting all the data,
the effects of promotion were completely ignored. Inherently, it was
assumed that trie large jump in quantity consumed, between special and non
special observations could be completely explained by the change in the
price level. Logically this argument is weak, but statistically it is
sound
.
For stores B and E the additional variables which were significant
at the 90 percent significance level in this model were the same as in
the previous two models, with the exception, of course, of the dummy
variables. In the equation for store A the price of halibut became
62
CO
aoHCO
CO
CD
u
CD
•HCO
Vj
O4-1
CO
CD XJpj cu
rH i-i
4->
>4-J OCO
CD CO4J CU
rHrH -o
cd-
CJ *H•H UU cd
to >
td4-)
CO
<dT3l~* AiTo C
Or. X>
CO X)QJ cO
-H trd4-»
•Ha a •
CO m
<d o53
Q)_j
r.
CC X)4-J oCJ
CU
a•rH
4H4H<u
Ou
1
s£>
CD
HftH
rH CD CO
nj in a)4J O rHO -U CvJ
H co oi
CO
CD
o•H
P-i
'-O *HUPH
0)
CJ
<f -HH
CU
a
H o
U-l "HO rH
Cd
P3
CO
MH <4^
O X)Ou
CO
•HMh
m xjO >H
o
CO
nCU
XJ
pj
orHP»4
CD CJ
U OH ^ t)vi o rd
CTJ
cmO 00O <T
a«h o mH M NO nI" LO
m m m
fO CM CM
CO CO onCO O CM-3" rH O
• • •
rH rH rH
CJ
<r co <CN CMo
CM CO CM
VD LT, rHO CN rHO CO CO
oO rHUO ov£) CM IT)
• • a
CO ^ CMI I
COrH
ON
a -
o•Hco
CO
cu
u co
00 4-J
a) fj; cu oP^l CD P> -HH «H -U
r-H CJ cd CO
A 4-1I
rHiJ (4_
(iJ W
M CD
cd O
cm <rO ONO lo
oO N ^)N HO H• a •
CO CM CO
arH COCM CM COCM CO sO
• • •
CO CM CM
CJ
cm r% OCO \DO CN CM
a • •
CM CO CM
O M CnI
ON rH I^s
m o to
a
o CMCM CM co
• • •
CO CM
Oco
I
aoHCO
CO
cu
M CO
50 ucu apcj a) 3 -hH rH 4-»
H cj co co
H > nj
co
cu
H4-J
•HCU o
cd
•H mHCJ O
FQ £H UCU
CM CMO CMO <T
auo <r
co o CMvO h N
• • a
<r cm -<r
co<r uo -<j-
<r co ^
oCO CM CMco <rCM N N
oco
aCM
o
r^. co
CM00ON
•
oONm
l
o•HCO
CO
CU
U CO
bB 4-1
cu a cu
pd, cu DH rHrH CJ Cj
cd -H >*H M-( I
4-j m 4J
CU
nj o
rH COO COO CM
O rH COCM sf OrH rH CO
CM CO ONON CO COon co in
• • •
CM
aON CO COn lo mvo r** n
rH CM
U
CJ
CM COO r-O ON
H IN COm o cm<- »n on
o o orH CM COvO N Ch
CO O i
—
1
<J o oCO CO CO
U Pu C_)
o rH CO o o CO CO<l- CM rH <r rHrH r- CO CM CM m
a a a • a •
CM rH CM
aCO rH CJ rHo ON
CM m o O m m ON• • • •
CM rH <r1 1 i V 1
\ 1
1 1
CO
m oa
ON•
mON CO
1 i
PJ PJ
o OH •H
(0 CO CO CO CO
CU CO cu CO cu
•H CU *M cu •H4_> M CO 4-J u CO 4J
•H to 4J H 00 4-J
(U
•H
CJ cu pi CU u cur-« O
•H cu p^ a) •H4J •H rH 4-> H rH X 1
CO rH O cd CO rH a cd CO
cd cd •H > cd cd *H cd
rH 1r-l «H M-) rH
W 4-1 4-! JJ W 4J 4-1 4J wu CU M GJ
a cd Oo Cxi Pm O
cu
pi
*H+j
o
63
T3
CJ
CMco
QJ
CMOCO
CJ
00
CO
a)
COCD
co
fa
T3CO
XJ
<rCM
00
XJ
MT>
CM
•aooo
-aCMCM
CO
CM
CMC*
O QJ gH 'd CJ ^
CO
"rl »HCO 4-1 *4-|
d o iHco i—
1
CD
Sh
o4-J
to
o
co
COCO
rH O
1 1 1 1 1
1 1 1 1 1
ga a do o o o
-H fP *H .a •H £1 •H a *HCO CO CO CO CO CO CO CO CO CO
co CD CO QJ CO a) CO a1 CO CJ
Qi H CJ •H CJ •H aj •pf QJ HM 4J u 4-1 M 4-J u 4-J M 4-J
0/) 4-1 H 00 4-> •H 60 4-J •H 60 4-1 •H 60 4-J »Haj d CD O cu C CJ CJ QJ C QJ a GJ a CJ a a) d QJ O
(D •f-l CJ -H C^ CD d «H QJ •H CU—
•
«H (H 4-t •H rH 4~» . «H rH 4J •H rH 4J *iH rH J->
rH U CO CO rH CJ CO CO rH CJ CO CO rH CJ CO CO rH CJ cy c/i
CO >H > crj •H > CO CO *H > CO CO <H > CO CO •H > co
»H «4-t 1 rH •H CM H *H M-i rH •H u-i1
rH •rH 4-4 1 rH4J <4H -P XJ 4-1 W 4-J <4H 4J w 4-> 4H 4J W 4-J MH 4-» WU CU M CD H QJ QJ QJ
cO O CO C CO O CO O CO Op.. CJ CO fa o CJ fa O fa CJ W fa CJ
•
+j
-H4-J
dCO
d
-aacO
CJ
CO CJ
•HQJ4JOd d4J co
0 Q)
o &mQJ
CM4-J
4->
GJ CO
HJ^ X)CO CJ
H 4J
CO
CU rHCJ 3co a
rH»• cO
aIW
O
0O•H4JCO
dCO
rH
CJ
U 4J OCO CO 0> rH
•H CJ
UJ 4-J CU Vi
Q) CO JS ^'H 4-> 4-»
4H (/)
CO
P-t CO 4-) CX cO Q cO
QJ r- ; 3: o QJ
UJ I 'H >d "4H *HH *H 4J
Q d *H^ 60 CO
U *H OE-H n (.0
CO cj "O QJ
dCO QJ
CO
J-) CJ
o .d
64
significant in addition to the price of swordfish. The equations for
stores C and D were, the ones which changed the greatest. The prices of
flounder, swordfish and halibut were significant at store C, whereas pre-
viously none of these had shown significance. Onl^ the price of flounder
and the price cf swordfish were significant in store D, while previously
neither of these had been significant.
Since this model produced potentially useful results and did not have
any problems with multiccllinearity , it became necessary to test for auto-
correlation among the residuals. For this purpose the d-statistic was
calculated and tabulated in the last column of Table 6-5. This was used
in the Durbin-Watson test to check for autocorrelation. The results from
this test indicated positive autocorrelation did exist in each of the
equatd ons
.
The. presence of autocorrelation does not bias the least-squares esti-
mators, but it does cause their standard errors to be underestimated. Thus,
in equations with autocorrelated residuals, variables will appear to be
significant when in actuality, they are not. Consequently, it is quite
possible that some of the variables with the smaller acceptable t-values
in model number 4 were actually insignificant.
Despite this problem, both price and cross elasticities of demand
were estimated for each of the equations (see Table 6-5). In each case
the price elasticity for the price of haddock was both negative and
greater than one, indicating an elastic demand for haddock as expected.
The cross elasticities calculated for those substitutes which were sig-
nificant were always positive and greater than one. Thus, it would ap-
pear that for each one percent increase in price of a substitute species,
65
if there was any effect at all on the quantity of haddock consumed, it
would increase by an amount greater than one percent. Judging roughly
from the elasticities obtained here, changes in the price of halibut had
the greatest effect on haddock, while the effects of flounder, swordfish
and codfish all seemed to be about the same.
A theoretical problem existed with this model also, the objection
being that no account was taken of the effects of newspaper advertising
and in-store promotion separate from the effects of large price reduc-
tions. However, with the data which was available it was impossible to
separate these effects, since there were no observations of extremely
low prices without promotion, nor were there any observations of high
prices with promotion. Thus, while price specials were clearly the most
important factor affecting sales, no conclusion could be reached as to
whether this was due mainly to a price effect or a promotion effect. Be-
cause of the statistical problems all of these effects had to be gathered
into one variable, either a price variable or a special promotions vari-
able, In this model a price variable was selected.
Model numb er 5: Separate special observat ions. One way of c i rcum-
venting the problem with the dummy variables is to construct separate
equations for the special and nonspecial situations. As has been pre-
viously explained, there were not enough observations of the special
situation to do this, but there were enough nonspecial observations, by
the time some of the variables were dropped, to attempt fitting a curve
to them (Table 6-6). When this was done, only stores A, B , and C showed
any significant variables at the 90 percent level (Table 6-6). In store A
66
•
CO
oHto
CO
0)
MoOcu
M
x
CO
>-<
o
CO Hcu n
orH
cd
cd4-J *dCO
CD rH4J CO
•HH aCO CD
o cx•H CO4-J
CO pi
H o4JCO4-J
• •
LTl
cd
•
OS3
CO
4-J Ha cu01 Id•H oa•Hmma)
ou
•
vO -
CUH
H
rH CD CO
CO U 01
4-J O rHO 4J COH co to
CU 3CJ ,0
vD «h m *H>H O rHP-l cd
3J
cu co
a -hiO *H M-l
m o -aPm Oa
43CO
•H
m no
CD
a•H
p.
CU
oCO *H 4-1
^ oPm
Ol
o•HSH
CN mo
CO rHrH CNrH rH
Oo cocn co(y\ co
cn cn
o o
CN
1
oo oi—I CNIT) v£>
*
CN CN
CJ
CO<T rHCN 00
• •
CN CN
UVO coCOCO vO
• •
olO CO
CN LTl
rHI
o•
uo<rcn
I
cnrH LTl
rH CO• •
CN rH
CJ
<r oo• •
CN rH
oas iHco uo
rH O<r cn
CN
CNCO
CO
CN <J-
m in• •
i i—
i
UO G>CO v£>
o uoCN CO
rH00 <ttH CO
CO
uo
rH CT\
rH CO
rH <fco <rO rH
• •
I
o\ or-». cnrH
O CNco r>-
<T rH
coCMco
oCO
a ao o o O o•H •H •H *H •HCO CO CO CO CO
CO CO CO CO CO
cu Ql cu CU 01
u M )H
dO 4-J W) 4J 60 il W) 4J CO 4-1
0) a cu cu OJ 01 0) 0J ci QJ CU CJ 0)
etf CD 5 s CU
H rH H H •H rH *r~! H •r^ rHrH CJ CO H CJ CO iH CJ cd rH o cd rH <J CO
«i •H > rj •ri > CO »H > CO H > cd *H >cu •rH 1
4-1 *H <4-f 1 •HI
«H 4-11
u 4-t jj 4-4 4J 4J U-i 4-J 4-1 U-l 4-J 4-J U-i 4-1
o V4 cu 0) S4 0J CU M cu
u cd o CO o CO o CO O CO OCO < PM CJ P3 Ph u O P-, « Ph C) W Pm CJ
cu
a•H4-1
cou
67
XJ
Cd
CMIKS
CMC^
>CD
O 1rH 4-J iH 3o CO Q
(U OJ GCM XI > £
o a gxi o gTJ i
—
1 2cd en Q
U HO CD gO XJ > g
rl ^ (U dto h QEd
CT\ 4-J 4-1 XJa o
xj•P CO
-H -HOO W IH
c: ohcd rH3 OJ
cr xiC/j
CM
COCOo
*
CM
CO
oCM
o•HOj
CO
<D
uoo -ugj a cd
H r-t
>1
iH CJ
cd -H•H M-!
-M 4HJ-J a)
crj O< Ph o
X)CO
C)
co
0^
CMoo
U000
o•HCOCO
<u
00 -u>
oj ap4 oj
•HO cd
*H >4-1 I
4-1 4-J
oj
o
o
CJ
vOCMoCO
COCO
CO
Pi
o•HCO
«a)
M00 J->
oj c oPS qj 2
•H 1-1
Hcd
H 4-1H >
I
4J
PQ UU CD
cd oO Ph o
COas
CNJ
iH
o
oCM
s•HCO
CO
a>
00 iJ
•H rHa rj
cd *H•rl 4-1
4-J 4HV-l CU
Cd O
CU
rHmoCM
CMCM
OrH
CO
cd
CD
cd
>COu
CD
CO•H
cd
0)
a1
X!4-)
MCU4-»
cCU
cd uo
4-J
O X<0 bO
4-J CU
CM 001 tH
\o xCO-
rn
cd
H
CM
4J
Oa
CO
•HCU
a) oj
CO rH
* cd
*Hucd
a o
<u •
> £o c
• rH *HrH 4-J
CU 4-1 cd
> PJ -Icj cu a)
rH O M
4-J CD Oa p. cj
<u ou in yQ) 4-J cd
P-: Cd
j
MH > o CJ Pi
OCO
OAAT
Pi •H QJ CO rHo X CU
*H 4-J 4-J rH >P a CJ
cd 4H 4-J Pi rH§ Pi O cd o•H cd a
pi
4-J
CO rH CU 4-1 Pi
CO cx a •HQJ rH cd CJ
^1 0) cd CJ CO ^-1
to > •H cd Ql
ReOJ Pi
i '-4-4 exCD CJ • •H•rHH O rj CU o 00 CO
cd *H > o XJ • •H OJ 4-J
•H 4-) 1 H CO H <c4-J 4-4 rd Xi pi cj -o QJ
M CJ cd
cd O XJP-i u
68
the price of swordfish was the sole significant variable. In store B
,
the prices of haddock, swordfish, codfish, and halibut were all signifi-
cant, while in store C the prices of swordf ish and codfish showed signi-
ficance. As would be expected, the competitive species were positively
related to the quantity of haddock, and the price of haddock was nega-
tively related.
The equations as a whole for three of the stores were not signifi-
cant at the 90 percent level when tested by the F-statistic, The overall
regressions for stores B and C were significant at this level. It is in-
teresting to notice the range on the adjusted coefficients of determina-
tion for these equations too— from 0 for store E to . 82 for store B.
From this information it would appear that the prices of haddock and of
major competing species—swordfish, codfish, and halibut—were more capable
of explaining the variations in the quantity of haddock consumed in store B
than in any of the other stores.
The d-statistic was also calculated for each of the equations in
Table 6-6. At the five percent level, stores D and E showed no sign of
autocorrelation. Stores A, B, and C may have had positively correlated
residuals, but' the calculated d-statistic fell within the inconclusive
range of the test values.
Additional comments « While it was stated that the purpose of this
chapter was to analyze each of the stores separately, and then test them
for homogeneity by Chow's test, this became impossible by the lack of any
function common to all stores. Only store B showed any promise for ob-
taining a useful, multiple relationship. Several factors seemed to con-
tribute to this situation. The primary one was the problem of multicol-
linearity among some of the variables, most notably the haddock dummy
69
variables and the price of haddock. There were moderate levels of inter-
correlation between the prices of haddock and codfish in some of the stores
also. This relationship was true of store B too, however, so it does not
account for BT
s more desirable results..
Another one of the problems contributing to the general inconclusive-
ness of the regression analysis resulted from some species not being sold
during every week. For instance, in stores C and D, there were over forty
occasions when at least one of the four major competing species was not
sold (Table 6-7) . Ideally such weeks would be dropped from the analysis,
and the regressions would be run only on those weeks when data was avail-
able for each of the variables in the equations. Had this procedure been
followed, however, there would have been very few weeks of usable data for
many of the stores, and statistical analysis would have been impossible.
Table 6-7. Number of weeks particular species were not sold.
Total number Number of weeks the species was not sold
S tore observations Swordfish Flounder Codf ish Halibut
A 36 1 9 5 5
B 24 5 5 1 0
C 27 10 12 5 15
D 30 9 8 22 8
E 25 7 2 0 1
Sum. 32 36 33 29
70
An alternative method of handling the problem, and the one chosen
for this study was to use the mean price of a species whenever it was
not stocked. Since the least-squares technique works with deviations
from the mean, use of the mean price rather than any other price mini-
mized the bias introduced into the analysis. If the mean price was used
infrequently relative to the total number of observations, this technique
would have had little effect on the final outcome, beyond slightly reduc-
ing the explanatory power of the variables involved. On the other hand,
if it was used extensively, as it was for some stores in this study, this
technique could result in a crippling loss of explanatory power.
To be specific, it can be seen from Table 6-7 that of thirty observa-
tions for store D, only eight had any variation for codfish. During the
other twenty-two weeks no codfish was sold and the mean price was used.
Similarly for store C, many possible observations were lost because of
the frequent absence of three species in the meat case. While less crit-
ical, the same situation occurs for each of the other stores too. In
summary, uhe frequent absence of some species from the meat case must
stand as one of the major problems underlying the insignificance of some
of the variables.
It is possible, for stores D and E, that factors other than those
considered in the equations were important in affecting the sales of had-
dock. When all the dummies were dropped from the equations, the F-values
for these two stores dropped particularly low, indicating that the remain-
ing variables explained very little of the variation in the quantity of
haddock. This low level of F was even more noticeable in the equations
where only nenspecial prices were used. For the first time, the R s
71
became insignificant for these two stores. At no time were the R^ !
s in-
significant for any of the other stores.
Summary of Results of Regression Analysis
Overall the results of the regression analysis were disappointing.
The purpose of running the regressions was to isolate some of the important
variables affecting the sales of haddock and to construct a demand equa-
tion for this species* Such a demand equation, to be of any use, would
have to be general enough to apply to all five of the stores. This goal
of the research was impossible to attain, since no such general function
could be obtained. Different factors affected the sales of haddock in
different stores. However, some useful information was gleaned from the
regressions
.
The most important factor affecting sales of haddock was quite ob-
viously price specials. Whereas other factors might cause some fluctua-
tions in sales of the species, if it was not offered on special it sold
approximately some standard amount, and if it was featured as a special,
that amount increased considerably, possibly by six times. Part of this
great increase in sales resulted from the lower price, and part, it is
expected, resulted from promotion. The separate effects of these two
could not be established because of statistical difficulties.
When all of these effects were combined into the haddock price vari-
able it was found that, the relationship between the price and quantity ot
haddock was negative and price elastic for all the stores.
Of the species which were competitive with haddock, it appeared that
codfish was the least important. Swcrdfish was a fairly strong competi-
tor in most all of the stores, while halibut and flounder showed signs of
competitiveness in some of the stores.
Of the five stores, store B yielded the most useful information.
More of the variables were significant in regressions on data from this
store than from any of the others. Price specials, while highly signifi-
cant, were relatively less important for this store than for any of the
others. This point was emphasized when, in Model number 5, no price
special data were used, and the data still explained 82 percent of the
variation in the dependent variable. The explanation for these results
may derive from the large proportion of Jewish people in the store's
clientele* Since the Jewish diet includes more fish than the diets of
many other ethnic and religious groups, this store stocked more varieties
of fish on a regular basis than other stores. Consequently, the frequency
of using mean prices when fish was out of stock was reduced, thus mini-
mizing one of the most troublesome data problems. Also, since fish was
a regularly purchased item of the clientele in this store, it may be that
they v/ere more conscious of the fluctuations in price and wore more apt
to vary their purchases in response to price changes.
Another possible cause of the differences between this store and the
others might be the fact that fish was sold from the delicatessen counter
in store B. There is little evidence to support this contention, however,
since store E, which also sold fish from the delicatessen counter, yielded
extremely poor results.
In store B the prices of haddock, swordfish, codfish and halibut were
all significant, indicating that these latter three were substitutes for
haddock. Only halibut and swordfish were competitive with haddock in
store A. In store C, flounder, swordfish and halibut were competitive;
and in store D flounder and swordfish were competitive. Store E was the
other store which sold fish from the delicatessen case, but none of the
species appeared competitive with haddock-
Some work was done with both swordfish and flounder as dependent
variables, but the results were even less conclusive. The data prob-
lems discovered with haddock became even more difficult to handle with
the fewer observations available for the other species.
74
CHAPTER VII
SUMMARY AND CONCLUSIONS
The primary objective of this study was to improve the state of
knowledge about consumer demand for specific species of fresh fish. A
major part of this goal involved the construction of short-run demand
curves for these species and the derivation of elasticity estimates from
them. To a certain extent, the primary aim has been realized, but little
was accomplished in the way of constructing statistically and logically
sound demand functions.
Data were gathered on haddock, flounder, swordfish, codfish, and
halibut from five stores of a single, case study, food chain in western
Massachusetts. Other seafood items were also considered, but these were
aggregated into either a finfish or a shellfish variable. Some consid-
eration was also given to differences in sales when a species was offered
on special. Data were collected for forty-two weeks, from July 196?
through April 1968.
Overall, considering both sale and nonsale weeks for all species, it
was found that haddock was sold in the greatest quantities. On the aver-
age. it also provided the highest total revenues, although there was less
disparity here than with the quantity. A close second in both categories
was swordfish. When weeks of price specials only were considered, sword-
fish replace*! haddock as the sales leader. Following a few pounds and
many dollars behind swordfish were halibut and flounder. Fifth in both
pounds sold and revenue was codfish. Other species of fresh fish were
sold, both salt- and fresh-water, but there was a very distinct break
75
between even codfish and these species. At most stores they were sold
sporadically, and at two, not at all. Fresh-water species of fish were
particularly prominent at the store with a large Jewish clientele.
From a general type of analysis it was determined that price spe-
cials were highly important in affecting the sales of any species. When-
ever an item was featured, it sold at least three times as much, and up
to five or six times as much quantity as was sold in nonspecial weeks.
Data problems prevented a detailed breakdown of the factors causing this
large increase in sales. Factors that most likely had a major influence
were the lower price, newspaper advertising, and in-store promotion. Since
both newspaper advertising and in-store promotion were minimal, it is sus-
pected that much of the increase was caused by the lower price, but no
numerical value could be placed on any of these "factors.
A number of models were tested in an attempt to construct a haddock
demand equation. Some of these were logarithmic functions, but linear
functions were found to give better results and are the relationships
reported. Equations were constructed for individual stores, but , since
each one had a different combination of statistically significant inde-
pendent variables 3 they could not be combined to form a meaningful gen-
eral haddock demand function. In general, high R T
s were obtained for
these individual store equations as long as price special data were com-
bined with nonspecial data, and the regressions were run on these, con-
glomerate data. Attempts to separate the effects of promotion and low
price by the use of dummy variables failed because of severe multicol-
linearity problems. Regressions calculated on just the nonspecial data
had much lower R2,
s, and not always significant ones.
76
Only store B yielded acceptable results when all the independent var-
iables of the model were included in a single equation. Most importantly,
it was the only store for which both the price of haddock and the haddock
dummy variables (for price specials) were significant in the same equation.
In addition, it was the only store exhibiting significance between the
quantity of haddock sold and prices of both haddock and competing species
when no price special data was considered. Thus, it appears that the ex-
planatory power of the prices of haddock and competing species was consid-
erably more important at store B than for any of the other stores. It is
difficult to explain this result with the information available, but the
difference in clientele may be one explanation. Most of the stores served
a mixed religious and ethnic clientele, whereas 40 percent of store B's
clientele was Jewish. Another explanation may be that the data was better
from store B. A greater variety of the important species were carried at
all times in this store, which minimized the effect of rigidities built
in by more frequent data adjustments in the other stores.
In those few equations with more than one significant variable, the
signs of the coefficients were checked, and some elasticity estimates were
calculated. Whenever this was done, both the signs and the elasticities
agreed with the expected outcome. The price of haddock was negatively re-
lated to the quantity of haddock, and the prices of substitute species were
positively related.' Demand was found to have a price elasticity averaging
about -2.9, calculated at the mean price and quantity. This figure is
slightly smaller then, but compares favorably with, estimates from other
studies, which were in the -3.5 to -4.4 range.
77
The strongest competitor of haddock was swordfish. Only in store E
was this variable not important, and in store E none of the species were
significant (in Model number 4) . Halibut was competitive in three of the
stores, while flounder was competitive only in two stores. Codfish was
competitive only in store B.
Functions for the other species were not worked with extensively, as
they had greater data problems than did haddock. No results from these
equations, beyond the general importance of price, specials, were important
enough to include in the report of this study.
Recommendat ions for further study . There were two major data prob-
lems in this study. One was the lack of a sufficient number of observa-
tions to handle all the variables, particularly with the species other than
haddock. The other was an almost complete absence of overlap between spe-
cial and nonspecial prices. As a consequence of this latter problem, the
effects of promotion during price specials could not be separated from the
price effects.
The most obvious and perhaps the only way to correct these problems
is to run a controlled study where the researcher sets the price and chooses
the weeks for promotion. In this way the price could be promoted ac all
levels and the separate effects of price and promotion could be discerned.
A useful extension of the controlled study might be to run it on a
sample of stores randomly selected from the region, instead of on just a
case study chain. This, of course, will involve the close scrutiny of the.
different pricing and merchandising practices, but if properly constructed,
it could -yield a much more powerful model for understanding the factors
affecting demand.
78
It would be useful to examine, either as a part of these studies, or
perhaps as a separate study, the effects of other substitutes for fresh
fish. Of particular interest would be other meat items, if only by grou^
:
pork, chicken, beef, lamb, etc. Processed and frozen fish items might also
be interesting to survey.
This study, while not fully achieving its aim, has pointed out some
interesting relationships, and has uncovered some problems worthy of future
development. It is a start in a relatively new field of study. Hopefully,
further research will be undertaken to carry on this beginning.
MANAGER Q^SSTIOWNAI RE80
Placement of Weekly Order -
1, What day of the week do you oraer fish?
a« For Tuesday deliveryFor Thursday delivery
2 0 At the time you place your order do you know both tike whole-sale and retail prices?
a* Wholesale Yes Nob„ Retail Yes No
How do you decide how much of each species ta order?
Factors Considered ( indicate order in which mentioned)
a. Wholesale priceb 0 Retail priceCo Margin
•
d a Quant* sold last weeke. Special
( given species}f* Special (other species)
g« Special HolidaySeanon
Lj» ~ „_
4. Do orders go through the head office or directly to
Quality Seafoods ?
5. If orders go through the head office are &ey ever modifiedthsre? Yes No
Diatribution of S^les During the Week
1 0 Wh»t are the peak sales periods. during the week? { Days of
the week and hours}]
2 C What are the unusually slack sales periods? (Days and hours
r
81
3. Breakdown of weekly sales
indicate
Store Hov.rcs
Percent of
Weekly SalesDistribution of
Monday9 1Z
3-6
6-9
Tuesday9-12
12.-3
3-6
6-9
Wednesday9-12
12-33-6
6=9
Thursday9-i2
12-3
3-6
6-9
Friday9-12
12-33-6
6-9
Saturday9-12
12-33-6
6-9
4 0 How much might the above vary from week to week and for
what reasons?
j's the pattern different this year from past years
Yes No
82
if the store is closed on a holiday,, are sates transferred Soanother day or lost?
3i tile holiday
iff .
on Trgtrij fferyed^
Lost.
Friday _Other Bays
A dditSonal Comments
C . Carry^Qy g?
1« 30 fish ever carried over and sold in the next w-^ek?
tto Fresh fish sold next week as fry oh,. Yes NoHow often (no, of weeks in a yr c )
b, Fresh fish, sold &ext week as froaen= Yes N0How often (No, of weeks)
w .
c, Frozen fish Yes Noi §
How often (No* of weeks)
2 e. le fiah that ia carried over necessarily sold the folic wing week?
Yes No__
a* If aotp when would it by sold?
3„ What is the reason for carry-over, when, it is deliberate?
Wholesale- price,
Other
4, How much e;ct:«ra is ordered with a low price? (Two week's
supply instead of one? ) Fresh Froaen,
83
U Is fish del&rered pr„ -wrapped or i« it packaged inth$ store,. ?
If not pi^-v/rc\ppr.d, in what form is it delivered?
E. Digplay
1, Linear feet of meat case allocated to fish (normally)
2* Description of location within the meat case
If the above change, indicate how often and describe the change*
How often
( No, of Weeks) PesoripHon
Space More Less
Location
Reason for the changes
Spacet
Location
Within the total amount of space used by fisha is a given species
allowed more than usual when it is on special? Yes No
85
Store
Mgr.
Date
SPECIES
—
Lbs. at
Disposed 1
Of
end of weekCarriedOver
If you run out of stock during theweek indicate the hour below underthe appropriate day(s)
.
Haddock
Cooked•
Tues . Wed
.
Thurs
.
Fri . Sat.
DressedFresh filletstro^en fillets
-—. ~ —
I
Halibut
Sword fish
FreshFrozen
—.
Cod
Fresh filletsFrozen fillets
Steaks
flounder
• Fresh filletsFrozen fillets
,
All She! IfishClamsCrab meatr 1Lo d s t er s
OystersSeal lops
-
LSchrimps LS q u i d
p i a 1
All Othe.r
finned Fish
Trout,
Butterf ishSole
m
r -
:——
RETURN TO OFFICE SATURDAY NIGHT
EACH WEEK.
ATTN. OF J. FERN
L V . ..r . C L . J
PerchSalmonSalt codfish i
SmeltsWhiting
!
tot a 7 r .
87
HCO
en rH CO oH m H oH CM O rH o CNI o1 1 H
CO•H4HTJcu
o tH CO OvO CO to oCO r-t LO rH CM o
CO
Hrm>~
o
U0 O CO oCO o^0 CN CO rH o
a)
so
CN o co orH oCO o CM o
« • • •
rH
oco o
in in o• •
r rH
4-
J
•H O
co cd
3 paO'
Eh
1^
ooo•
rH
OOO
CO
•H IPCO CO m &cu *H
u rH m nd -a ndo M CO fl« rH d) 3
0) -H ,d 4-J 09 rH d rH CU
a m CO CU ci cu Cucd o •H rO a) nH > O rH cu
M U-l rl u CO M-t CU rH cu Cuo O rH o rH CO > o
Cd rH O 4-1 >^ a) rH4J CO a ^ CO 4-* 4J AS rH CO
•H *H •H O u<u cu Q) cu CU rH 4-» 4J o o u U
6 G CJ CJ o 0 RJ CI CJ <u a>
cd •rl •rH •rl •rf U cd cd ndM rl U c y cd cd 4J 4-1
cr cm p-l H cy !U 33 o o
88
U 0) p
x o EW H PO co P
rH vO vO vO in <* <± vO 00 m CM o0)
!CO o in CM H O 00 CO CO o
X! > O <r <* H H rH H CM o m LO o4-» 0)
O 1 a I 1 1 1 I 1 H
.v.
a oi >-
o ph eXI O &XJ rH 3td w P
O rHo cya >x) qjcd ,j
H a; Xi-> pi cn
Cd -rl p4
cr
4.) «H
Ci
S3
H D mcd i-i qj4-J O HO 4-1 Ct!
H CO CO
o CO H vO CO CO m m CO vO m o<T m H o vo CM CO m in oo <r CO rH rH O CO o rH o <* <r co o
i 1 1 1 1 l I H
vO rH m o 00 CM m CM oO 00 o o CO CO o m CM oO CO o o rH CM rH o CM Ox o
rH <r CO CO rH m om vO CO rH o CO CM o m oo 00 r-- O o rH CM rH o CM oI 1 1 1 1 rH
rH <f O CM oo rH CO <r orH CM CT» 00 m oCO rH CO O o rH CO CM rH o
vO vO vO 0\ CM m VO OCO CM vO rH CO O
CM o rH rH rH CM CM rH o1 i 1 1 1 1 rH
co 00 <f o OvO vO O
CM CM o o rH CM rH o
iHH
pdCO
•TS •H 1*CO CO MH
r* QJ •HCJ rG rH «4H XJ t3 xi
o r4 CO .rH CD rJ
X> QJ •H 4-' tn rH Ci rH QJ XJ XJ
XJ O -d HH CO EJ a> C3 QJQJti o XJ •H QJ •rl > O rH
rCj XJ i rJ MH H M (0 m CU rH QJ
XJ o o XJ rH O rH 0} > or*> cd rH & O cd u t>* cu rH4-1 M-l CO o CO 4J 4-^ M M rH CO
•H •H •H a VuQJ QJ CJ QJ 0) 4J 4-1 0 o rJ
cd
a CJ o CJ a 0) OJ
•H •H •H •rl •H 4J s X)V4 rl VJ u H o cd cd 4J
OtP-l P-l Ph Ph Pm H o* & Ed o
pq
CU
uo4-'
CO
Uo
X•HU4_»
cd
B
g•H
cd
Q)uuoa
cu
I-•HCO
CM
a)
cd
E-t
cuCO
o •H"-H
•H
M o*>
Cm
P CJ
H o
cd cd
oo
-3-
oo
o CO o
COco o
cm0\m OOo
to CO -3-CM in oin CO CO
• • • •
oH
OEl o•H oEh
H
•HEH
OOO
CM m oO O o•
o•
co•
o1
•
>n in oo oo• « •
i !—
1
ooo
ao
•rl
4J
cd
^)
pdCO
A3 cu <H Uo COo a tj •H rO
iw *H-u o o *d i—
!
ctf T—
f
!S o cd«w co a
cu 0) a) cu CUu o a cj o'rl H •rH -H •HM V-! M 4 iH M
Ph
90
U CD MCD {X
O4-J rHO CO
I
iH1
t—
i
vO vD m -3- CM rH rH CT> O<u O o CM H CM O o
> O LO CO o rH CM CO O CO rH o4J CU
O rJ 1 1 1 1 1 1 1 rH
O (X gXi O 0-d i—i 3
> eCD 3
cd Q
u x;•rH CD
a
cr
to
MH
-I rH0)
C/ CO
rH CD CO
r4 0)4-J OO 4-) Cd.
H co co
3
CMOO
CM
rHmo
ocoo CO
COCOCM
CM
o UOCO
mco 00
ooorH
00 m \£5 vD O rH CM orH oo CM rH CO o CM orH \0 rH o rH o rH O rH o
1 1 i 1 1 1 rH
CM CM <o o> CM co oo O00 m CM vO 00 CO CO CM o
rH rH O rH o rH o rH o1 1 1 1 1 1 1 rH
rH o CM CM CM Ox oo vD 00 m o rH oCO o rH rH o o
1
CO CO O,
1
orH
CO O \£> 00 rH m m Oo o CO rH CM CM CO CO orH CM O rH CO O rH CM o
CM rH oo> <r o m as o
rH m CO CO m o o
OOX)xjcd
oo*
rH CO
0) H 4-J
o XJ MH CO
o xj •Hxj S JH 4H •Hxj O O xj rHcd rH o cd
U-t CO a
CO cu a) cu (D
o a CJ a O•H •H •H •rH •HiH M M M y\
PM Pu Pm P-f
rCCO
•HCO CO <4H
a) •HrH m X» XJ XJCd rH CU
CO rH HI: a)
CU 1<u ,C > ou CO MH rHo rH CO
u >^r*£CO 4-J 4-J M
•H •H U orH 4J o ocd X54-J S xi
oO'
cd cd
c/ ^1
3
rHa)
>cu
cu
XJ
Q)
P.OrHCO
U0)
-C4-t
o
I
91
uQJ
UO
m
om
cd
@
E*H4Jcd
rH0>
J.;
oo
0)
rH
e•HCO
co
aiH
H
(U
«H O
QJ Oe ooH •
H
0^ tH m 00 \D OH H CO i—
1
oH CO H CN o CN o! 1 1 1 H
rH [>. CN CO oCA O as o
• <H CN CN O oCO
rH CN OS oO CO o iH oCN o o o
• • •
1
•
1
•
rH
co <tco o
CN Oo o<r o
CO oco oo
• •
1 rH
IT)
ooo
&CO
& <H
tCO CO 4-1 BQJ •H
a rH;
.
*4H njBo H co cd iH a)
QJ •H CO rH f3 rH QJ XJ rdo -a 4-1 CO CD a qj P.
cd 0 •H rQ cu *C •H > o iH QJT cJ B U 4-J •H u CO QJ rH a) ex13 o O
»>o rH CO • > o
>> cd rH O cd ,u
Ck ck
CO rHJJ
•HrCi IM O CO
It itrH CO
QJ QJ <U 03 a) rH 4-J u c o M£ CJ o a a a cd CJ a *0 01 QJ
cd «H •H «H H H p cd cd *d-H M M M U 0 0 cd cd 4J
H Pm P~i Ph H cr O* o O
92
u <u >qj
•x: o-U rHO O'J P
v D e^ > e4-J QJ DC t-4 Q
OH >O QJ B^ > EX) CD 3
4J xi•H GJ ^4-> CJ CO
ctf «H Pth
^ CO4-J »H•H M~5
U Hs ^CD CD
CO CO 00 sD CO r-» CM CM o loO DO rH <• COcm CO o rH o CO o O CO
CM
CO co oo> ooo o
ooCO
to CO cm CO o o> On] LO CM oo o rH CM CM LO LO oo cm H rH O LO LO o1 1 l 1 1 1 rH
sO CO o o oo rH to oO CM CM CO o sO CO CO CM orH CO rH rH O CO o O o o
CO o o CM O CO rH oo LO o oCM <r o CM LO rH CO o
rH CO rH rH O CO O o o o
LO rH LO OO <r oo <r CM Oo co rH O CO CM OLO rH CM o CM o CM CM O
CM OO CO o CM SO <r o OsO O co rH rH CO oo sO oCM o rH o CM O CM rH o
1 1 1 1 1 1 rH
rH <U CO CM CM OS rH LO o SO ocrj U QJ CO rH rH CO CO oo LO o4-» o rH CM CM o o <t o oO 4-J CO
H 00 l rH
qj
H
CO
•HCO CO mQJ •H
a rH l*H -a
o u CO - crj rH QJ
-u M QJ •H ,£! 4J 0"J - rHXJ CJ XJ m CO CD 5CO O •H rO QJ &
XJ s LH •H U CO LW
X* o O x) rH Otoctf H o CtJ 4J
4-> m CO a x: CO 4-1 AJ
*H H rl4» QJ QJ QJ QJ QJ rH 4J 4-J
a u CJ CJ CJ CO
a3 •H •H •rl H •H 4-> cd
u J-4 M M u ocr &4 PM CM H O*
>o >*>
1XJ3
rH QJ XJ XJQJ
> O rH a)
QJ H QJ
H COev 1n
A£ H 01
CJ oo o {J
X) XJ QJXJ XJ rj
CO CO 4-J 4-)
rd O o
93
as
CO cn 00 o•H CN o <? oU~\ CO o o
• • • • •
M l rH
CO
*H O
cd
co ovD rH a> co o CO orH O o o o H o
1 i 1 1 1 rH
00 CO Oo CO o Csl CN oCO CN rH o
• • • • •
CM H ooo o oo CO o
•
1
• •
rH
CO OCN cn O
LO o• •
! H
CN
ooo
0)
H
OoO
&CO
,C •HCO 05 4-J
0) •Ha ^3 rH MH x)o U to CO rH 0*a A! Q) 4J CO rH CJ rH 0) -a
a CO CD
o *d *Q CJ JE3 •H > o rH OJ
n S m H M CO m rH-d o o rH O rH CO > c
>, cd rH o cO 4-1 &* CO rH
it pd 4-4 CO a id w • 4-t
H it*
XL &
0!—
i
CO
4-> a) CO a) cu rH 4J 4-1 o c u Ma CJ O o o O CO 0 0 <u cu
co •H •H •H •H H 4J 2 CO TJ •dM U r4 o CO CO 4J
o* Pa pa Pa Pa N & O o
94
U CD >0) Cu g,d c S4-J rH 3O CO P
<u oj er£ > e+J CD 3a p
a a) >o p, e"doexj h P
H3
O rH >o a) E13 > |ni ^ P
4-1 *T)
•H CD
O*
4J
c3 o
p] M a)
4-» O rHO 4-J Ctf
H oo co
COoo CM
CM
CO
CO
CNoo CO
COiHO
rH vD CO CN orH iH CO CO oCN rH st co o
1 1 1 1 rH
m CN o eg CO 00 V0 lO 00 o o00 CO CO 00 O CM CO o Cr> oO <t co rH CO rH O CO H UO o
1 1 1 1 I 1 1 I H
CM CO CT\ <r o O iH oi—
1
CM CO CM O rH co CO a> oo VO o CN CM rH o rH O o1 i 1 1 1 rH
co a> 00 co m CM O oco o CM rH CO LO oo 00 o iH CM H tH rH O o
m m CO \0 <n CM o oCM CO <r 00 rH CO 00 o
CM o H rH O o CO o o• • • • •
1
• • • •
H
•H rH o 00 o H co <r oIH 00 o in co o CO orH O rH o rH CM O CO CM orH
CM o rH CO o oCO o CN CM CO oo <r o co CM o o
CO COCD •H
Oo M sn ax It
a CD H 4-1 to rHa M-* CO d CD
o a X* -H CD
.a M m *H U CO
o o rH Ocd rH o 4-J
<H CO a CO 4->
H •H4-> 0) QJ a) OJ CD rH
CD
£o a u a a CO
a•H •H •rl H •H 4~»
•H d M u M H rJ oH o* PH P4 Pi H
CO
•Hm3
>>
*d *o x>CD
rH CD x)
a CD
•H > O rH CD
M-4 CD rH CD O.rH CO > O
4-»
CD rHrH CO
H O o4J o o M
0 CD
a .c:
CO cd 4-J
o* o o
95
wQJ
o
CO
noMH
X*H
4Jcd
cd
rHcu
MS-f
OCJ
a;h&CO
cu
H
co 00 rH O•r! <f o CM rH O<4H L-O rH CM rH oxj • • • • •
1 H
xj afrl O
re cd
Si
CO <t <t rH o orH a> O m oO CM rH rH o o
1 1 l rH
o uo <t o CM oCO m CO 00 oCO o CO rH rH o
a) CO U0 \D oXJ co <* O oC CM rH o
• • •
o 1 rH
\0 oo oo* • •
1 rH
oc
ooo
0)
•rH
H
oOO
CO .
•iH
CO CO 4-1
cu •HCJ rH MH XJ x)
1O M CO cd rH " CU
xJ CU •H CO rH a rH CU XJ XJx» a id m CO CU d CU p.
cucd o XJ -H rO cu rG •H > o rHXi s M 4-4 •H u CO Uh CU rH CU
XJ o O XJ rH o rH CO > ocd rH & CO 4-J r^ ^
CU rH4-1 MH CO a CO 4-1 4J rH CO
*H *H »H CJ CJ
4~l cu cu cu CD rH 4J 4J O O M Mo CJ CJ CJ CJ cd CJ
srO XJ CU cu
s •H tH *H •H *H u cd x> x»H M M H o cd cd 4J
P^ Ph Ph P-t H cr o o
cu
V
uo
CO
u a) > O CM <r O a> o o v£> CM CO oCU i O o r-- CO CO CO m m ox: o o ("0 o H H CO <r CO oo
rHCO o 1 \ 1 1 i 1 H
OJ fc'
O t-5 P
O O t>O P, EXf O £XJ H 2
a h no cu &xj p> £x; a) 3cd jjj p
U XI•H (U ^3-U d CO
cd <H
•H •-!-;
p i-i
Pi HCO CU
3 X!C/ CO
i—
i
CU COCj CU
4J O rHO ci
H CO CO
CO O CO CM o rH in O oo oCT\ O CO CO ct\ cr, CM rH orH <t m o o CO CO CN1 CO CM m m o
1 i l 1 1 l I H
ON rH O in co m oCO CO O O co co o CT» oo co r* H rH CN O rH rH o> o
1 1 1 1 1 rH
O rH CM CT\ CO <r co CO rH oCO O CO CO o co oO o\ rH rH rH CM O rH o o
! 1 1 1 1 1 rH
CMCM00O
COCM
0\CM
minrHCM-
COCOCO
m ooo
o rH rH <!" CM o <r oCM CO O o CO CO o o
rH CM rH rH CM o <r o1 1 1 1 I 1 l rH
rH CM \D 00 o CO oCO rH CM oCM CO CM CO CO o o
H
O •aO H COXJ CU •rt 4-i
X) a X> IM CO
CO o id •H rPXl S U UH •H-o o o •o rH
rH o CO
4-1 IH CO cj ^•r(
4-J 0) CU cu CJ CU
P CJ CJ tj CJ CJ
cd •H •H •H *rt "H?H M H M
cr Cm cm CM Cm
XJW•H
i*CO CO m%cu •H P rrH XI XI
cO,
rH 0)
W.
rH CJ rH Q) xi X)
a) C3 CU P,cu0) X! «H > O rH
CO CU rH CU
o rH CO > o^ ^ CU rH
CO 4-5 rH CO
*H •H a CJ
rH 4-» o o u uCO xi XI cu CU
XJ cO X) X) £1O cO CO 4J 4J
H CD* c/ o o
BIBLIOGRAPHY
97
1. Anderson, R. L. and T. A. Bancroft, Statistical Theory in Research,New York: McGraw-Hill Book Co. Inc., 1952.
2* Baumoi, William J., Economic Theory and Operations Analysis , 2nded., Englewood Cliffs, New Jersey: Prentice Hall, 1965,
3. Bell, Frederick W. , The Economics of the New England Fishing; Indus-try
:The Role of Technological Change and Government Ai d , Re se a r ch
Report of Federal Reserve Bank of Boston, No. 31, 1966.
4. Bell, Frederick W. and Jared E. Hazleton, eds-, Recent Developmentsand Research in Fisheries Economics , Publ. for the New England Eco-
- riomic Research Foundation. Dobbs Ferry, N. Y.: Oceana Publications,Inc., 1967.
>5. Brown 5 Sidney E., Retail Sales of Broilers and Meat as Affect e d byPrice ,
Display Area, and Newspaper Advertis ing, USDA-ERS-180 , May1964." "
"
6. Bryant;, Edward C, Statistical Analysi s, New York: McGraw-HillBook Co. , Inc. , 1960.
7 . Christ , Carl F . , Econometric Models and Methods, New York : JohnWiley and Sons, Inc., 1966.
8. "Consumer Expenditures Study, 1966," Food Topics , September, 1967.
9. Croxton, Frederick E., and Dudley J. Cowdon, Applied General Statis-
tics, Englewood Cliffs, New Jersey: Prentice-Hall, Inc. , 1955.
10. Exakiel, Mordecai, and Karl R. Fox, Methods of Correlation and Re-
gression Analysis. New York: John Wiley and Sens, Inc., June 196 6
,
3rd. "ed.
11. Parrell sJoseph F. and Harlan C. Lampe, The New England Fishing In-
dustry : Function al Markets for Finned Foo d Fish II, University of
Rhode Island, Agricultural Experiment Station Bulletin 380, 1965.
12. Foota, Richard J., Analytical Tools for Studying Demand and Pric e
Structures, U. S.D.A., Agricultural Handbook #146, August 1958.
13. Fox, Kar? A., The .Analysis o f Demanv. -for Farm Products , Technical
Bulletin No. 1081, U. S.D.A.. , 1953.
14. Fox, Karl A. and James F. Cooney, Jr., Effects of Intercorrelation
Upon Multiple Correlation and Regression Measures, U. S.D.A. , AMS-
341 * April 1.954.
98
15. Goldberger, Arthur S. , Econome t ric Theory , New York: John Wiley &Sons, Inc., 1964.
16. Gray, Leo R. , Effects of Price Specials on Volume of Sales of Frying;Chickens, ERS-202, Reprinted from Agricultural Economics Research,Vol. XVI, No. 3, July 1964.
17. Griffin Report of New England, Boston, Massachusetts: Griffin Pub-lishing Company, Inc., June 1968.
18. Johnston » John, Econometric Methods , New York: McGraw-Hill Inc.,1963,
19. Klein, Lawrence R. , An Introduction to Econometrics , Englewood Cliffs,New Jersey: Prentice-Hall, Inc., 1962.
20. Klein, Lawrence R. , Textbook of Econometrics , While Plains, New York:Row, Peterson and Co. , 1953.
2.1, Leser, C. E. V., Econometric Techniques and Problems , London: CharlesGriffin & Company Ltd., 1966.
22. Li, Jerone C. R. , Introduction to Statistical Inference, Ann Arbor,Michigan: Edwards Brothers Inc. (Distributors), 1957.
23. Malinvaud, E. , Statistical Methods of Econometrics, Chicago: RandMcNaily and Company, 1966.
24. Nash, Darrel A., Deman d for Fish and Fish Products With Special Refe r-erence to New England, Division of Economics, Bureau of Commercial Fish-eries, U.S. D.I.., 1965.
25. National Association of Food Chains, Economic Effects of Pr ice Specialsfor Meat , A Report by the Food Chain Industry to the Fact Finding Com-mittee of the American National Cattlemen's Association, Washington,D. C. , January 2, 1959.
26. Nordhauser, Fred and Paul L. Farris, "An Estimate of the Short-RunPrice Elas. of Demand for Fryers," J.F.E. , 41:798-803, November 1959.
27. Scheffe, Henry, The Analysis of Variance , New York: John Wiley and
Sons Inc. , 1959.
28. Spnrr, William A. and Charles P. Bonin i, Statistical Analysis for
Business Qe
c
is ions , Homewood, Illinois: Richard D. Irwin, Inc.,
196 7.
"
29. Stanton, B . F. , "Seasonal Demand for Beef, Pork, and Broilers", Agri-
cultural Economics Research , pp . 1-14 , Vol . XII I , No . 1 , U.S .D. A.
,
January 1963 .
99
Suits, Daniel B., "Use of Dummy Variables in Regression Equations,"Journal of American Stat istical Associat ion , Vo 1 . 52, No . 280,pp. 548-551, December 1957,
Thomas , James J . , Notes on the Theory of Multiple Regression AnalysesAthens, Greece: Contos Press Co., No. 4 of Training Seminar Series o
the Center of Economic Research.
Tintner, Gerhard, Econometrics , New York: John Wiley & Sons, 1952.
Tomek, William G. , "Using Zero-One Variables with Time Series Datain Regression Equations," J.F.E . , Vol. 45, No. 4, November 1963,
pp. 814-822.
United States Department of the Interior, Food Fish Situat ion and
Outlook Report, May 14, 1968.
White, Donald J., The New England Fishing Industry^ Harvard Univer-sity Press, Cambridge , Massachusetts , 1954
.
Working, E. J., "What Do Statistical 'Demand Curves' Show?" Read-ings in Price. Theory , edited by George J. Stigler and Kenneth E.
Eoulding, Chicago: Richard D. Irwin Inc., 1952.
Yaniane, Taro, Statist ics ? An Introductory Analysis , New York: Harperand Row, 1964.