ANALYSIS OF MARKET FACTORS ASSOCIATED WITH CUT
NATURAL CHRISTMAS TREE RETAIL SALES IN URBAN
AREAS

Dissertation for the Degree of Ph. D.
MICHIGAN STATE UNIVERSITY
LAWRENCE DAVID GARRETT
1976

I". L -. LIBRARY

IIIIIIIIIIIIIIIIIIIIIII IIIIII IIIIW

3 .129310560 9980

                                 

Mir" m
University

 

.'

This is to certify that the
thesis entitled

Analysis of Market Factors Associated with Cut
Natural Christmas Tree Retail sales in Urban Areas

presented by

Lawrence David Garrett

has been accepted towards fulfillment
of the requirements for

 

Ph. 0 D o deg-cc in ForeS try

 

 

 

 

L___
[:1

Majcrjrofessor IIJ

Date January 23L 1976

 

G7 639

 

 

 

 

‘ 121%

m

..,., , /
3:: 2,0,5

APP. I7 1.9975

 

ABSTRACT

ANALYSIS OF MARKET FACTORS ASSOCIATED
WITH CUT NATURAL CHRISTMAS TREE
RETAIL SALES IN URBAN AREAS
BV

Lawrence David Garrett

Annually, forty million natural Christmas trees
move from nlantations and wild lands throughout the United
States and Canada, through various market intermediaries,
and finally to retail sales lots. tThe market effort cul-
.inates with the sale of millions of natural Christmas trees
to consumers in three short weeks in December. In many
areas the retail marketing period is even shorter. The
market mechanisms are sophisticated and the marketing func—
tions quite complex. Yet, each year hundreds of thousands
of Christmas trees are left unsold on Christmas day in
metropolitan areas, while in other metropolitan areas insuf-
ficient numbers of trees are available for the consumers.
The market system reacts bv fewer retailers coming into the
market the following year in one area and more in other
areas. The significant problem faced by these retailers is

that insufficient market knowledge is availableI An in

Lawrence David Garrett

depth research effort was made to explain those factors
which associate greatest with natural Christmas tree sales
and to statistically model these factors to predict indi-
vidual retail lot tree sales potential.

The study identified various Christmas tree mar-
keting attributes of the Winston-Salem market. The area
absorbed an increasing number of natural Christmas trees
over the 3 years of study; in 1967, 9,227 trees were sold,
whereas in 1969, 11,941 trees were sold. A variety of trees
were available, including Balsam fir and Scotch pine from
Canada; plantation Fraser fir and wild cedar from North
Carolina; and Douglas fir from the West Coast. Average
price increased over the 3 years of study, and the percen-
tage of unsold trees vacillated from 14 to 17 percent.
Retailer mobility was high, with 31 retail lots moving in
and out of the market over the 3 years of study.

The study related associations between retail lot
market characteristics, consumers' demographic character-
istics, and lot locations. Analysis of marketing factors
revealed that successful retail lots had better trees,
located in areas of high retail sales activity} on heavily
traveled roads, and had ample parking facilities. These
lots sold more trees, obtained better prices, and had fewer
unsold trees at the end of the marketing period.

The study also disclosed that the consumer travels
considerable distances to purchase natural Christmas trees.

The research evaluates the various market associations and

Lawrence David Garrett

presents interpretations as to their combined effects on the
spatial location and performance of natural Christmas tree
retail outlets in Winston-Salem.

Using principal factor analysis and regression
analysis, the market associations observed in the Winston-
Salem natural tree market were statistically modeled. The
research identified independent spheres of influence or
factors, which were used in subjective evaluations of market
variable associations. Associating variables were further
utilized in developing predicting equations to estimate tree
sales for individual retail outlets in Winston—Salem and
Denver, Colorado.

Each predicting equation developed was tested for
its accuracy in predicting sales with market data different
from which it was developed. Interpretations were made as

to each model's efficiency and utility.

ANALYSIS OF MARKET FACTORS ASSOCIATED
WITH CUT NATURAL CHRISTMAS TREE
RETAIL SALES IN URBAN AREAS

By

Lawrence David Garrett

A DISSERTATION

Submitted to
Michigan State University
in partial fullfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Forestry

1976

ACKNOWLEDGEMENTS

I am indebted to many individuals for the comple-
tion of this research effort. It represents the culmination
of a relatively small dream in this hemisphere of high ideas
and worldly accomplishments. However, the academic accom-
plishment which this paper transmits to the author tran-
scends this. It began with a high school principal who
instilled an appreciation for knowledge, through professors
who were much more humanitarian than given credit, and ended
with a wife who understood the frustration, yet urged me to
carry the project to completion.

Having made the necessary hurdles in the academic
world, many of my concepts of that world have changed, how-
ever, one has not. This system, in spite of its failings,
still instills the desire to learn and understand. For that
fact, I thank the learning institutions I have attended, and
the many people with whom I was associated. To my major
professor I must express my gratitude for a conviction which
will no doubt follow me the remainder of my professional
life: "Because things are not completely to our satis-
faction, it does not necessarily mean they are of no value,

and not deserving of our consideration."

ii

TABLE OF CONTENTS

INTRODUCTION . . . .
PURPOSE OF THE STUDY .

A Structural Model .
Natural Christmas Tree Supply.

RESEARCH PROCEDURE .

Selecting Study Area, Variables,
and Data Collection Techniques

Study Area . .
Selecting the variable Set

Demographic variables.
Economic variables

Data Collection.

Retail lot owner survey.
Consumer survey.

Reduction of Original Variable Set: Factor
Analysis . . . . . . . . .
Deriving the Predicting Equations. Regression
Analysis . .

Testing the Model.

ANALYSIS AND DISCUSSION OF RESULTS

The Retail Market Structure for Natural Christmas
Trees.

Tree Sales

Prices .

Number and Type of Retail Lots
Retail Lot Size. . . . .
Lot Location .

Type of Street

Parking Facilities
Merchandising.

iii

l7

l8

18
21

21
21

23
23
24
25

28
30

32

Product Differentiation.
Ease of Market Entry .

Associating Retail Lot, Tree Sales, and Consumers.

Associating Market Factors with Tree Sales
Validity of the Crossover Assumption .
Principal Factor Analysis.

The Variable Set

Creation of Factors.

Subjective Assessment of Factors
Screening the Variable Set .

General variables deleted.
Specific variables deleted .

The Predicting Model

The Variable Set . . .
Specification of the Model
Observation Error. .
Empirical Models

Evaluation of the Winston-Salem Predicting Models.

The Denver Market.
The Denver Model . . .
Evaluating the Predicting Equations.

Variables Represented.
Analysis of Residuals.
Residual Plots .

Accuracy of Estimators

CONCLUSIONS AND RECOMMENDATIONS.
Summary.

Market Associations.
Analysis and Inference
Prediction Models.

Conclusions.
Recommendations.

APPENDICES
APPENDIX A .
APPENDIX B . . .
LIST OF REFERENCES

iv

101

102
105

108
117
174

Table

Table

Table

Table

Table

Table

Table

Table
Table

Table

Table

Table

Table

Table

H

L»)

\l

8.

LIST OF TABLES

analysis

1967-1969.

.--Variab1es selected for initial factor
.--Christmas trees sold in Winston-Salem, NC,
.--Average retail price of Christmas trees in

Winston-Salem, NC, 1967-69

.--Percent frequency of natural Christmas tree

retail lot types: Winston-Salem, NC, 1967—

1969

Winston-Salem, NC.

purchases out of census tract.

.--Natura1 Christmas tree market relationships:

.--Association between income class and tree

.--Association between income class and miles
traveled to purchase a natural tree

--Highest 14 factor scores on 19 factors

9A. --Identification of top 14 variable loadings

9B.

9C.

9D.

9E.

9F.

on 19 principal factors

--Identification of top 14
on 19 principal factors

--Identification of top 14

on 19 principal factors-.

--Identification of top 14
on 19 principal factors

--Identification of top 14
on 19 principal factors

--Identification of top 14
on 19 principal factors

V

variable

variable

variable

variable

variable

loadings
loadings
loadings
loadings

loadings

Page

117

34

36

37

54

62

62
124

126

127

128

129

130

131

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

9G.

9H.

91

9J

9K

9L.

9M

9N

90.

9F

9Q

9R

98

10.

11

12.

--Identification of top 14
on 19 principal factors .

--Identification of top 14
on 19 principal factors

.--Identification of top 14
on 19 principal factors

.--Identification of top 14
on 19 principal factors

.--Identification of top 14
on 19 principal factors

--Identification of top 14
on 19 principal factors

.--Identification of top 14
on 19 principal factors

.--Identification of top 14
on 19 principal factors

--Identification of top 14
on 19 principal factors

.--Identification of top 14
on 19 principal factors

.--Identification of tOp 14
on 19 principal factors

.—-Identification of top 14
on 19 principal factors

.-—Identification of top 14
on 19 principal factors

variable

variable

variable

variable

variable

variable

variable

variable

variable

variable

variable

variable

variable

loadings
loadings
loadings
loadings
loadings
loadings
loadings
loadings
loadings
loadings
loadings
loadings

loadings

--Forty-nine highest scoring variables

selected from 14 x 19 factor matrix .

.--List of 41 variables entered in stepwise

regression models -- Winston-Salem.models;

1967,

1968, 1969.

--Variables appearing in predicting models

I, II, and III; Winston-Salem, NC .

vi

Page

132

133

134

135

136

137

138

139

140

141

142

143

144

145

148

150

Table

Table

Table

Table

Table

Table

Table

13

14.

15

16

17

18

19

.—-List of variables entered in final stepwise
regression models -- Winston-Salem models;
1967, 1968, 1969, and Denver model, 1965.

--Comparative statistics on models I, II, and
III .

.--Description of model I: predicting model
for natural Christmas tree sales; Winston-
Salem, NC, 1967 .

.--Description of model II: predicting model
for natural Christmas tree sales; Winston-
Salem, NC, 1968 .

.--Description of model III: predicting
model for natural Christmas tree sales;
Winston-Salem, NC, 1969

.--Description of the Denver, CO model:
predicting model for natural Christmas
tree sales, 1965.

.--"F" tests of residuals from 4 predicting

models tested on 4 different sets of market
data.

vii

Page

152

86

154

156

158

160

95

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

Figure

LIST OF FIGURES

Page

1.--The natural Christmas tree supply function

during the retail market period . 10
2.--The transitory natural Christmas tree

supply function . 12
3.--A general study model to analyze natural

Christmas tree sales in urban areas 19
4.--Concentrations of Winston-Salem natural

Christmas tree retailers. 4O

5.--Percentage of Christmas tree lots in Winston-
Salem that advertised by various media,
1967-1969

6.--Winston-Sa1em major traffic arteries and
natural tree market held by major and minor
retail areas.

7.--Natura1 Christmas tree purchase patterns by
census tract: Winston-Salem, NC, 1969.

8.—-Squared residual plots of estimated Denver
sales with model I.

9.--Squared residual plots of estimated Denver
sales with model II

10.--Residua1 plots of estimated Denver retail
sales with model III

11.--Squared residual plots of estimated 1967
Winston-Salem retail sales with the Denver
model.

12.--Squared residual plots of estimated 1967
Winston-Salem retail sales with model II

13.--Squared residual plots of estimated 1967
Winston-Salem retail sales with model III

viii

43

55

58

162

163

164

165

166

167

Figure

Figure

Figure

Figure

Figure

Figure

l4.--Squared residual plots of estimated 1968
Winston-Salem retail sales with Denver
model.

15 --Squared residual plots of estimated 1968
Winston-Salem retail sales with model I.

16.--Squared residual plots of estimated 1968
Winston-Salem retail sales with model III

17.--Squared residual plots of estimated 1969
Winston-Salem retail sales with the Denver
model.

18.--Squared residual plots of estimated 1969
Winston-Salem retail sales with model I.

l9.--Squared residual plots of estimated 1969
Winston-Salem retail sales with model II

ix

Page

168

169

170

171

172

173

LIST OF SCHEDULES

Page

SCHEDULE I.--CHRISTMAS TREE RETAIL MARKETING
QUESTIONNAIRE . . . . . . . . . . . . . 108

ISCHEDULE II.--CHRISTMAS TREE CONSUMER QUESTIONNAIRE. 115

INTRODUCTION

Christmas tree marketing in the United States

n ”“4“”;

1'4 "
L/F‘éf’u
It“
I94 V '

annually involves the sale of approx1mately 40 million

natural trees valued at nearly $140 million (Sowder 1965).

Future sales are likely to be even larger By 1976, ex-

0"

Danding annual consumption should reach 45 million trees

(U.S. Department of Commerce 1965). This will represent a

market with a total retail value of at least $160 million

annually. In addition, some 20 million pounds of decorative

material --wreaths, roping, and yule logs-- are sold each

Year for more than $600,000 (Sowder 1965).

\WA market of this magnitude for a product as

Perishable as a natural Christmas tree requires an effective

I‘nlarket information system. However, there are definite

indications that the performance of complex marketing

me thods and channels now being used in the industry is less

One basic problem appears to be the in-

than desirable.
1/

duS try' 3 inability to accurately predict consumer demand-

This is illustrated by:

l\
./
Ellefson, P.V., and T.H. Pendleton. 1969.
Sis of Christmas tree marketing in the eastern U.S
Unpublished report. U.S. Dep. Agrie. For. Serv., For.
Prod. Mktg. Lab., Princeton, w.VA.

 

Problem analy-

1. Excess and deficit supplies of trees at the

retail level.

2. The year-to-year turnover of retailers.

Reliable estimates of the magnitude of excess and

deficit supplies of trees are varied. While certain areas

appear to have an abundance of trees on retail lots, others
face dire shortages of quality trees.: For example, studies
111 the St. Paul - Minneapolis metropolitan area have shown

tilat approximately 15 percent of the total trees available

fk>r sale at the retail level remain unsoldl/. Certain re-

taiilers in that market have been known to discard two trees

ft>r every three sold. Other studies have found similarly

11i.gh figures (U.S. Department of Agriculture 1961), (Troxell

19 70), (Drysdale and Nausedas 1970) .
0n the other hand, certain areas face critical

Estlortages of quality trees and trees of certain species- ,

(Eiullivan 1963). It is not uncommon for a consumer to be

IEt>rced to purchase a particular species which was not his

0 O 3 O 0
original cho1ce-/. Nor 13 1t uncommon for a consumer to

1.
‘l/Iillefson, P.V. 1965. The 1964 Twin Cities retail Christ—
Unpublished report. Minnesota Univer—

Unas tree markets.
sity, School of Forestry, St. Paul, MN.

2/

I IEuller, K.B. 1964. A study of consumer preference for
(3hristmas trees in a mature market area. Unpublished
biasters Thesis, Pennsylvania State University, School of

IEorestry, University Park, PA.

 

3/
‘ Iillefson, P.V., and T.H. Pendleton. 1969. Problem analy-

ssis of Christmas tree marketing in the eastern U.S.
Ianublished report. U.S. Dept. Agric. For..Serv., For.

Prod. Mktg. Lab., Princeton, W.VA.

purchase an artificial tree, or perhaps none at all, because

there were no quality natural trees available (Black 1962),

(Drysdale and Nausedas 1970).
It is obvious that losses to the retailer will be

substantial if he is unable to sell his supply of trees by

Christmas day. But equally significant, he is also finan-

cially affected if he cannot supply the quantity and type

(If trees desired by consumers. In this case, potential

sailes may be permanently lost and profits reduced.

J‘

{In a like manner, both the wholesaler and grower

feice possible losses from failing to satisfactorily esti-

In£1te the size of their markets. A wholesaler who incor-

treectly estimates sales either in total or by species will

find himself with a supply of unsaleable trees or else

tlrlable to fill an order. If the grower fails to accurately

:EtJrecast consumer sales either in total or by species, he
3153 faced with supply-demand imbalances which he must live
With for several years}

Year-to—year inconsistency in retailers is the
SeCond chief indication that inadequate market knowledge is
one of the major problems facing the Christmas tree indus-

It133?. In some markets, only six out of ten retailers return

It<3 sell trees the following year. And, of the new total,

all‘ound one-third are inexperienced retailers new to the

trade (Skok and Miles 1963), (Troxell 1970).

While these figures may not be representative of

allareas, they are sufficiently typical to indicate the

rather volatile and inconsistent nature of the natural

Christmas tree retail market (Conklin 1962), (Nelson 1961).

This erratic market entry and exit may well be due to a

lack of market knowledge of consumer demand. Instead of

attempting to estimate his potential sales, and then decid-
ing whether to enter the trade, the retailer all too fre-

quently makes his decision based upon the number of

retailers in the market during the preceding year and the

seemingly high profits they earned. The result is too

nurny retailers in certain years and too few in others.
Sllch inconsistency results in two major problems for the
irldustry:
Growers and wholesalers are often faced with
inefficient and unreliable market outlets.
Retailers frequently suffer economic losses
because they lack marketing experience.

One solution to excess and deficit supplies of

t:Ifees, retailer inconsistency, and increased consumer satis-

faCtion is to increase knowledge of those market factors

associated with natural Christmas tree sales. Accurate

estimates of market demand and subsequent coordination of

SuI>p1y and retail sales can significantly reduce losses now

0Q curring in this industry.

 

PURPOSE OF STUDY

The major objective of this study is to identify

and analyze market factors associated with cut natural

Christmas tree sales in urban areas. Further, to develop

forecasting models which will assist Christmas tree
rwetailers in estimating their potential sales in a given

nuirket.area. The research will attempt to relate retail

taree sales for a single retailer to those variables which

a1:e:fbund to be the most important determinants of his

Sailes.
In order to achieve this objective, two subobjec-

tives must be met:
1. Predicting models developed in this report

attach considerable weight to the natural tree

being purchased in close proximity to place

of residencel{g{ (Brundage, Nicewander, and

Kohly 1955), (Brundage 1958), (Roth and Brum-

mel 1971). Should the assumption be false,

1\
‘1’
IP‘uller, K.B. 1964. A study of consumer preference for
(3hristmas trees in a mature market area. Unpublished
asters Thesis, Pennsylvania State University, School of

Forestry, University Park, PA

2/

I $chweitzer, D.L. 1963. The retail Christmas tree market
3Ln.the Twin Cities. Unpublished report. Minnesota Uni-
‘rersity, School of Forestry, St. Paul, MN.

5

I
3’“;

‘.I\‘

.~VE

 

that is, people buy trees distant from their
place of residence, the model would be
improperly specified (Thompson 1964). There-
fore, a secondary objective is to determine
the validity of this assumption.

2. Because there are so many variables which can
conceivably influence tree sales, it is neces-
sary to screen a large set of economic and
demographic variables to obtain a smaller sub-
set having maximum explanatory power (Green
and Tull 1970). Once this subset is identi-
fied, it can be entered into a predicting
model as a set of independent variables.

The model then identifies the relative signif-
icance of those variables having association
with Christmas tree sales.

Accomplishment of these objectives will result in
wOirkable information that can be used by the retailer in
effective decision-making activities. That is, identifica-
tiOnof those economic and demographic variables which asso-
ciate with or influence tree sales, greatly increases the
retailer's market knowledge. Not only is he better equipped
It‘D 'make decisions on lot location, merchandising, and adver-
tising, but given a certain mix of these variables, he can

determine his probable sales performance at a given location.

n. - 1'

and u
c

pad-1

0....

u u!
”my

!
0.9-1'0

'Ovi-

n

v...,\
‘1

"—5-.

.
‘-- .—

---~ ..

u,

' V‘
a _'
p'..v

.fi“
“1".

|..
u “‘

-..

 

 

A;Structural Model

To fully understand the market relationships
between buyers and sellers of Christmas trees requires
analysis of such market factors as product characteristics
and price, consumer demographic characteristics and prefer-

ences, market characteristics, attributes of related

products, etc. That is, the model should incorporate all

factors related to the product, the buyer and seller, and

the market place .
It is normally impossible to include each and
evmery conceivable variable affecting a product in a market

ennxrironment. Yet, in characterizing a structural model,

iutzis important to identify through observation, empirical
aIlEilysiS, or a priori evidence those factors felt to be
inngaortant. A consumer demand or sales model for cut

néitzural Christmas trees would, therefore, reflect the

fC>llowing relationship:
Demand = some function of production and market
factors.

The above model must be partitioned to be mean-

illasfnl and, depending upon the objectives sought and con-

st31Eaints imposed, the list of individual variables to

represent these factors may vary widely. Generally, how-

e‘7€éry the analyst attempts to include the most influential

vSil‘iables even with the most abstract model. Such a parti-

tioning would likely include the following list:

Sales for cut Christmas trees = a function of

 

 

tree price, quality, type; substitute product price, quality,

type; complementary products; supply of natural tree, com—
plements and substitutes; consumer characteristics and

preference; number, size, quality, and type of retail out-

lets.
The sales for natural Christmas trees is predi-

cated on available supplies of the product. It was beyond

the scope of this research to study supply relationships;

therefore, the apparent interactions between sellers,

growers, wholesalers, brokers, etc., are not analyzed. How-

ever, to obtain a clear perspective of the demand relation-
Ships presented, it is important to have an intuitive under-

Standing of the natural Christmas tree supply function and

its a priori relationship to demand.

lflgtural Christmas Tree Supply

Natural Christmas trees are a perishable, non-

d1—‘ll‘able, consumer good. They are perishable in a biolog-

ical sense because they physically deteriorate. In an

eConomic sense they are also perishable. That is, cut

rlatural Christmas trees have zero market value on December

26 regardless of their biological condition, and conse-

9

quently, a demand ceases to exist. In contrast, the arti—

fiQial Christmas tree is nonperiShable and is a durable
good. In fact, post December 25 sales of artificial trees

at reduced prices have become quite popular in metropolitan

areas.

The natural Christmas tree market period, or the
period of direct interaction between supply and demand, is
generally initiated the last week of November or the first
week in December, assuming no artificial market constraints.
This is not to be confused with the production period
required for the product, which averages 8 - 10 years.

The market supply function depends a great deal
on the time interval separating markets and supply points.
For metropolitan areas, the time interval is generally
too short to negotiate significant adjustments in supply
during the market period. Producers and wholesalers are

inware of this problem and, therefore, harvest and handle
fixed product volumes based on orders placed 6 to 12 months
irl advance. Opportunity for competitive bidding of addi-
tional supply is severely limited.

Because of these factors, the supply of cut

Ilaltural trees is virtually inelastic during the market
Ipeexiod. The degree of inelasticity exhibited within a
given market place depends on many factors. Examples of
these factors are as follows.

Size of market: Large metropolitan markets are
generally supplied by distant growers, wholesalers, or
brokers who deal in large volumes. As indicated in Figure
1: :i;f the market structure is not responsive (i.e., no
trading within the market or between like outlets in nearby
mazrl‘:£ets), the retailer supply function is extremely inelas-
tic

By contrast, smaller markets are supplied by more

~|

n‘fini

 

 

.¥A~

~

 

10

 

Wholesale Price

 

O X
No. of Trees Purchased by Retailer

3Fiagure 1.—-The natural Christmas tree supply function during
the retail market period.

113<2a1 grown or wild trees from smaller growers or producers.

'Tllee supply is normally more expensive, but can be more elas-

tic (8181).

Proximityof competing demand points: Several
market areas (cities) in close proximity can have elastic
Sllrxply functions. For example, chain store retailers have
3m£ll;tiple sales outlets in competing markets. Since market
i1'1-‘1?”<2>21:‘mation is readily exchanged, supply can be shifted
from where demand and/or prices are low to a market where

demand is strong and/or prices are high.

Market structure: The market structure of the

 

 

 

.gbqsy‘

4....o-

(I)

...

.\. .

’V-u
‘

o. ‘c.

u.‘

u
00"

 

A"

.l

11

natural Christmas tree industry can affect the supply func-
tion. If several market areas in close proximity are ser-
viced primarily by different suppliers, the supply function
to each is more inelastic than if one supplier is servicing
all areas. Primarily this is due to shorter time for mar-
ket feedback and reduced transfer and handling costs. These
transfers are usually arranged to minimize losses to the
retailer relinquishing ownership of a tree shipment.

The above relationships regarding supply refer to
the retail market period or the very short run. Regressing
backward from December, the greater the time interval, the
Inore elastic the function. For example, if we were to
exPand our treatment of supply to the 9-month period,

‘Airril - December, we would observe a transitory supply func-
tion. In Figure 2, the period April - July is represented
1337 $1; the period August - October 82; and November -
December, S3,
The greatest grower volume commitments are made
Iiti period one at a lower average wholesale price, p1. As
the procurement period shortens, the grower or wholesaler
demands a higher price to cover both real carrying costs
a11c1 those imputed costs associated with risk and uncer-
tainty. Faced with a much shorter interval for disposing
‘11? tineir product, the growers' supply function becomes more
inelastic.
In the last period (within or immediately pre-

Seeding the market period) the grower becomes virtually

12

 

Wholesale Price

 

 

No. of Trees

Irigure 2 --The transitory natural Christmas tree supply
function.
irliflexible. Regardless of the price offered by the dealer,
bee can only release a fixed amount to the market. If he
Lhaici used normal marketing procedures, he would have little,
ii? any, product left for that market period. To make
realuatively small quantities of additional product available,
the grower would have to incur high unit costs. Even if
he Could charge exorbitant prices, a small sale would yield
insufficient profit to create the incentive. If the poten-
tial sale was large, he would be faced with a more serious
consideration. He would 'have to deplete his inventory,

W o
Inch has planned allocation to established future

n
.uyvA'

-..~-

.551.

mum

tau-l.
.

I‘-.q
\
a u- d o

1
I..

u
'Qv‘.»

It

:. ,e

‘ H'I

I

3 ‘

..,,. a
~ A

- Q
..~

13

markets -- markets that could be lost if not supplied.

Both of the above supply functions are important.

First, and most important, within the retail market period,
supply is virtually fixed. Second, although the transitory
supply functions directly affect the price which the
retailer must pay for his trees, this price does not com-
pletely determine the retail price of trees in the market
place. Retail market price is determined from the direct
interaction of market supply and consumer demand.

Retail sales of cut natural trees are constrained
at least at the upper limits by the fixed supply schedule.
In our study, we assumed the existence of a fairly rigid
illelastic supply schedule during the retail market period.
Demand and price relationships are, therefore, not affected

1337 a changing supply schedule. Supply is fixed, as well as
‘tlie incurred cost for its offering. It remains only to
tzxreat the demand relationships which guide transfer of
1:11is fixed supply from retailer to consumer. Consumer
cieemand is influenced by these factors:

1. Christmas tree prices.

Consumer taste and preference.
Market size.

Consumer income.

Ul-I-‘UJN

Price and market characteristics of related
products.

6. The range of goods available to the consumer.

14

Market prices for Christmas trees are, no doubt,
initially dependent on wholesale costs, which are determined
via the transitory supply schedules previously discussed.
Since the retailer is faced with a fixed supply, to maxi-
uuze profits his market strategy should be to shift the
Cbmand function upward at all price levels. However,
he is thwarted in these efforts as the market period wanes.

Consumers' preference and taste are demand
shifters. That is, if a consumer suddenly prefers a dif-
ferent species of trees, the demand for that tree shifts
Upwards at all price levels, and downward shifts occur for

Species being abandoned. It is possible to attract some
Sales back if the retailer can ascertain the reasons for
tlle change in preference, and successfully implement
corrective market strategies.

Consumer preferences and tastes are shaped by

Dnaany variables; religion, cultural background, ethnic
ggurouping, peer grouping, social status, education, etc.
Iicjwever, they are not inflexible, and do change over time.

Market size strictly represents the potential

Inairket. That is, ostensibly, the larger the population,
the larger the sales of the product in question. This is,
(31? «course, not necessarily true. For shoes, raw population
col-lnts may be correct, but one does not wear a Christmas
t17€3€e. Market size for Christmas trees is best represented
b}? itiousing units. And, if number of housing units increases,

0th Qr things being equal, more trees will be supplied at all

 

 

-
~~~i
o

.
h «-

boi-

 

Q‘-

'Q-

‘1.

n
'u‘

HA

..,:

ft)
( )

'A

.H.
i

15

price levels, or demand is shifted upward.

Income is a gross measure of the economic ability
to absorb goods and services. However, the consumer's
income is constrained by necessary purchases of government
goods and services (taxes) and private goods and services
(i.e., necessities; food, housing, clothing, transporta-
tion). Remaining are those monies which can be expended
for such items as Christmas trees. Increases or decreases
can affect the total number of trees purchased.

Price and market characteristics of related
products, whether they are substitutes or complements, are
limportant to sales of cut natural Christmas trees. Balled

IIrees, cut wild trees, cut natural trees, and artificial
tzsees have certain levels of substitutability in the market

place. Wreaths, lights, tinsel, tree stands, etc., are,
2111 most cases, complementary goods to the above products.
(311anging the price of any one type of tree may well change
1:11e position of the demand function for all substitutes.
Cit: would also shift demand for complements. That is, if
I3630ple save money buying their trees, they may buy more
tinsel at existing prices.

The range of goods available to the consumer
affects demand for the product. Intuitively, the wider the
3r‘31113e, the greater the individuals' needs are satisfied
arzcjf, therefore, the greater the sales. It is infeasible

it: ‘tllhe supply sector to create a wide range of trees, and

yet - some adjustments must be made. Quality, appearance»

I
‘..l.

".v.

-o; ‘

Luv A

ux-A

“by!

 

v0.

-
0a...

‘Oab.

 

16

species, price, packaging, etc., are all varied to conform
to different consumer preferences and needs.

An intensive analysis of two natural tree markets
and related specification of Christmas tree sales predicting
umdels required treatment of most of these factors. Some
were of lesser importance than others, and some were more
difficult to incorporate due to the type of analysis util-
ized. These elements of demand are treated herein as
factors relating to retail sales and, therefore, will be

referred to as tree sales.

.'... H
:. w

.
.nnn fi“

- I
,....»b‘

u. ‘1
ewv‘ .'
.
.0 31‘;

.n
T V
.P.
r I
‘5'
. ‘
'I A“
\n —:
~¢~b
.
.‘Ifl..L
“ ‘vuhh
a
" N‘UA
I l C
9 . I
v
‘.
‘HAH
'lsnt‘q
‘ I l
.
‘ Tu. b
"n..§‘
V 0
‘.A
-‘V‘
"‘h“
u
.
V‘e.
_V
..~‘~
I
I .
- "h‘l
a.
M‘I _‘
' I
s‘ ‘u
\_,r“

 

RESEARCH PROCEDURE

The basic hypothesis of this market analysis is
that tree sales for a given retail lot are a function of
economic variables associated with the product, the retail
lot, its market location, and demographic characteristics
of consumers in the census tract where the lot is located.
Further, the hypothesis to be tested is that equations can
be developed to accurately predict individual tree lots

Sales via estimation of these variables.
Testing the equations for accuracy requires a
IIY’pothesis which will test the difference between sales in

3- ggiven market area (Y), and the predicted sales (Y) ob-

tained from the developed equations. Stated specifically:

an: 2 (i - Y) = 0
If the equations are found to be statistically

Stiggnificant, further analysis will be completed to determine
tlleair practical utility as a management tool. The primary

CITi;terion for the acceptance will be the capability of the

ecllléitions to estimate actual tree sales within the average

HTEEZEEgt spoilage‘level, or 15 percent.
Development of this research progressed through

SQF’GBZtral stages, each having separate procedural elements,

17

 

 

 

y.-

.yqq

o...

(I)

L l

‘5

L)

18

but dependent on other stages at some point of the research
process. The research procedure used is best represented
by the following four stages, which will be discussed in
order of their formulation (Figure 3).

Stage A: Study area, variables, scope of study,
and methods of data collection and
analysis.

Stage B: Screening variable set for redundancy.

Stage C: Final reduction of variable set, deter-
mination of coefficients for socio-
economic variables, and deriving the
predicting equations.

Stage D: Testing the models.

.§S{;ecting Study Area, Variables,
EELQ_Data CoIlection Techniques

Study Area

 

Winston-Salem, North Carolina was chosen as the
Stiudy area. Since approximately two-thirds of all natural
ttrees are sold in metropolitan areasy, it was felt that a
s'tudy of a metropolitan area would result in greater bene-
fits to the industry.

Development of the forecasting model was dependent
I11.3on the Bureau of Census information system, i.e., census

JEEEfigct data. Winston-Salem is a Standard Metropolitan

\

l/
1This estimate is based upon the fact that over 70 percent

of the U.S. population resides in metropolitan areas.

‘

19

 

.mmmtm cmnt: :_ mo_mm mocu mmEum_t:u _mcsumc o~>_mcm Ou _opoe >n3um _mcocom <uu.m ot:m_m

_EOcoom mo
ufiom— Mum
>o>cam

 

 

     

  

o_;dmcmoeo
mamcou mo

 

 

 

_

 

 

 

 

 

 

 

.eoo
:m: m>mL®Q ummnam V LUmemmm
1|. u co_umet0¢ 1. o_nm_tm> mpooz mums ummm
mmoc< otuoz N c. .mcmfiho_nm_tm> mo co_uoo_mm oc_Ecouoa
mo_mm nouo_poca w
_mzuo< ummeucOo a H

 

 

 

 

           

 

m LUQOLQQ< mm—Qmmo—m> xmi—UNZ mumuom ”mum—3°C! ”~0th

 

      

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  
  

 

 

 

 

 

 

 

 

 

uconcmaopc_ u x_tumz >mo_0po;uoz _mu_u tood : :o_umo_m_u uoxtmz
amok :m: c0uumm o>_coo :>_mc< ao_o>o revue. 50—30ca uo>tomno
memo> >o>tam
m eotm m_onoz H a
ummcucoQ x_tumz
muco_o_ woOu .LLOu nouspom .Uum .pc_ mock
< zumotaa< mco_um_wcL0uiL w .moou .ttoQ_+ .mot< >esum .o>_u mmEum_tc
o_QE_m o>_ton o_aE_m o>_too noomno oc_mo _mtaumz
.Aco_mmotm .Am_m>_mc<
tom om_3aoumv Logomm _ma_o . .LULmomot
co_umauo mc_uo_e ac_tav uom o_nm mo tom: new .mv05uoe >paum .mo_nm
._onoe ecu m:_umoh Iota m:_>_too :_cm> mc_o:pom :_tm> .mo>_uoomno .mEe_nota oc_moo

 

 

 

o mo<km u mu<hm m mw<hm < mo<hm

.».v I

" flu

less.

I
"Q.

“.5

' I
e...-

Q-‘ul

AI
in“:
'5‘.“

3?

‘5

 

 

20

Statistical Area (SMSA) for which detailed demographic

information is available by census tract and city block.

Also, all Christmas tree retailers in Winston-

Salem are required by city ordinance to purchase a retail

license. As a result, the universe of retailers was

readily available. Only 6 of 35 other SMSA's surveyed

in the East and Southeast had such a city ordinance.
Winston-Salem SMSA had an estimated 1968 popula-

tion of 150,000 people. The population is divided into

white (60 percent) and negro (33 percent) racial groups,
with segregated areas of residence significantly reflected

in census tract statistics. Average income varies by

census tracts beginning at $3,000 and exceeding $20,000

at: the upper level.
Winston-Salem is recognized as a major consumer

market area in the North Carolina Piedmont Plateau. This

is evidenced by the fact that the Winston-Salem - High
Point - Greensboro area is one of the most rapidly expand—
ing market areas in the Southeast (Sales Management 1966).

The suitability of Winston-Salem as a test area
of consumer response to market goods is supported by the

fact that four major market testing firms have consumer

panels in the cityll.

\

7‘ /Source: Margaret Ernstes Market Interviewing Service,
Winston-Salem, NC. Market Testing Firms: Home Testing
Instit. , Family Opinion Poll, Speedata Inc. , and Marke-

test, Inc.

 

21

Selecting the Variable Set
Two general classes of variables are assumed
to be associated with sales of natural Christmas trees:

1. Demographic variables characterizing the

population buying trees, and

Economic variables characterizing retail
outlets from which trees are purchased.
Certain associations, such as changes in taste, were not

evaluated due to limitations of the cross section analysis.

IDemographic variables
The 1960 United States Census of Population and

I'1<:>using Report PH(l)-l74 for Winston-Salem, North Carolina,
SMSA was used to develop census tract data for the study.
Census tract data provided a complete set of demographic
cllaracteristics of consumers, and these data are readily
available at minimum cost.

Variables 48 through 104 represent the demographic

vatriables used in the study (Table 1, Appendix B). The

56 variables can be categorized into 10 general types or

E§roups as illustrated. Several measures of the same vari-

aE>1e were selected, and the most important of these were

iSolated using factor analysis (Harmon 1967).

womic variables
Marketing research has firmly established that

c3‘31‘tain marketing practices are instrumental in effecting

re tail sales of consumer goods. These practices, when

 

22

quantified, can be studied with the intent of identifying

those having greatest influence on sales. Because there

is definite interaction between economic variables and the
demographic variables mentioned above, their combined
effect on sales was studied.
Variables 1 through 47 and variable 105 (Table 1,
Appendix B) are the economic variables used in the study.
As with the demographic variables, an attempt was made to
identify specific variables which would best represent the
general group.
Economic variables were obtained from several
Isources. Variables 7 through 47 were obtained from obser-
‘Vations recorded at the retail lot and personal interviews
with retail lot owners (Schedule 1, Appendix A). Variables
1 through 6 were created by plotting lot locations on a
winston-Salemblock map and noting street distances among
t1lese various retail lots. These distances served as a
I’lflaxy variable for determining competition among retail
jlots.
Data for variable 105 was obtained from 1967-69
annual average daily traffic volume (AADV) mapsy. A
IInfiljor traffic artery carried an AADV of traffic exceeding
JLC) .000 vehicle units. It was usually a 4-1ane expressway,
‘33:“ limited access road. Two-lane roads with an AADV

\

../’
Diomth Carolina State Road Commission, Traffic Division.
1967, 1968, and 1969. Annual average daily 24-hour traf-

fic volume maps (AADV). Raleigh, NC.

 

 

 

P:-

23

greater than 10,000 were improved (widened) 2-lane roads

with asphalt berms.
The AADV variable acts as a direct measure of

potential customer traffic. The traffic volume for a

given retailer was determined by the AADV for the road
adjacent to the lot.

Data Collection

As noted previously, all data on demographic

variables were extracted from the 1960 Bureau of the Census

ptflflication PHC(1)-174. Data on economic variables were

taken from four surveys in Winston-Salem during the Christ-

Inas holiday periods of 1967, 1968, and 1969. Two schedules

Were used in these surveys, a retail lot owner interview

S<ihedule (Schedule I, Appendix A) and a consumer survey

SChedule (Schedule II, Appendix A).

5335311 lot owner survey
The two-part retail lot owner schedule was used

‘:C> obtain information on economic variables which affect

5‘ tretail Christmas tree lot's sales performance. Part I was

us ed to record observations by the researcher on the retail
lot, such as conditions of trees, lot appearance, parking

Marketing data were taken on Part II via a

Space, etc.
The

Personal interview with the lot owner or operator.

Slarvey was conducted for a three-year period; 1967, 1968,

and 1969.
Retail lots were located through the Winston-Salem

' .
4 B
‘ i

 

24

retail licensing bureau. All observations and interviews
were conducted by forest economists from the Forest Products
Marketing Laboratory at Princeton, West Virginia. Retail
lot observations (Part I) were made on December 17 - 21 for
each of the three study years. Interviews with retailers
(Part II) were conducted in January and February with the
majority of the interviews completed in January.

The total number of observations (retail lots)
varied from 53 in 1967 and 1969 to 46 in 1968. This was due
to entry and exit of retailers and not to nonrespondents.
.AJl lot owners within the city limits were contacted and 100
jpercent response was recorded on Parts I and II of Schedule

II for each of the three survey years.

Consumer survey

As indicated previously, past research has re-
‘Iealed that consumers prefer to travel only short distances
tn) purchase natural Christmas trees. However, this hypothe-
siis needs further evaluation as it poses a critical assump-
tZion to the model being used in this study.

To determine if Christmas trees are purchased near
tille consumer's residencell, and also to obtain information
(311 artificial tree purchases, and consumers who did not
IDIJrchase any trees, a survey of Christmas tree consumption

\

l/

“ TNear the consumer's residence" as defined for this study,
ls an area encompassing the census tract in which the con-
ESumer resides, plus directly adjoining census tracts.

 

25

patterns was conducted in Winston-Salem.

A private research organization conducted the

survey via house-to-house personal interviews, using Sched-

ule II, Appendix A. In total, 1,017 responses were recorded.

A stratified sample of Winston-Salem consumers was

constructed, using 1960 census block listings. The samples

within each census tract were weighed by the 1960 population

of the census tract.
No callbacks were taken on nonrespondents, how-

ever, replacement interviews were taken. In a nonresponse,

the interviewer would take the next highest house number as

a replacement. If this respondent was already in the

ssample, the next highest house number was used and so forth.

Questions 1 and 2 of the schedule were used to

1Tesolve proximity of purchase of trees to the consumer's

Imouse. Questions 3, 4, and 5 were used to establish the

nmarket segment held by artificial trees. Questions 6 and 7

Vwere used to determine the market segment which presently

dOes not purchase any type Christmas tree. Questions 8 — ll

‘Vsere used in correlating survey data to census block data.

lsSeduction of Original Variable
332st: Factor Analysis

Two general analytical models were used for data

Eitialysis. Factor analysis (Harman 1967) was used to reduce
title original variable set (Table 1, Appendix B) to a smaller
s"--IT:>set. Stepwise regression analysis was used for further

da ta screening, derivation of partial regression

‘

[IW‘ '1 m
' an.

 

26

coefficients, and development of predicting equations.

To current day statisticians and mathematicians,
the term "factor analysis" has varied meaning. The term
represents a group of techniques that are used to analyze
the intercorrelations within a set of variables. Its
primary purpose is to find a way of summarizing the infor-
mation contained in a number of original variables into a
smaller set of new variables or factors, with a minimum loss
of information -- that is, to remove the redundancy in the
original measurements.

Generally speaking, the method can incorporate
ifiour different analytical methodologies. These are briefly
described by Massy (1963) as:

1. Separation and analysis of distinct dimen-
sions that are latent in a larger set of
variables.

2. Separation and analysis of differing groups
which exist in a larger population.

3. Identification of certain likely variables
for subsequent regression or discriminant
analysis from among a much larger set of
potential independent variables.

4. Summarization of the common parts of a set of
explanatory variables into a smaller number
of new variables which can be used in regres-

sion or discriminant analysis.

27

Approaches l and 2 use the smaller number of con-

ceptual variables for factors as ends in themselves; that

is, the listing of factory scores on each independent

factor is the ultimate goal and the scores are used only to

subjectively identify the factors. Method 3 also uses

factor scores, but uses the scores for identifying variables
Method 4 requires that
These

to use in subsequent analysis.

quantified estimates of the factors be obtained.

 

estimates are then used in subsequent regression or discrim-

inant analysis.
Method 3 was used in this study to screen the

Original 105 variable set to the smaller variable subset

1lsed in the final regression analysis.

This use of factor analysis as strictly a data
reduction tool departs from the classical use of factor
analysis (1 above) as proposed by Thurstone (1931).

Much of the criticism of the use of factor analy-

813 in marketing research centers around its use in the
classical approach (Ehrenberg 1962), (Rothman 1963)- It is
purported that little is gained with factor analysis because

0
f the subjective manner in which factors are identified.

lat is, the final goal sought is a list of subjectively

i
tieratified orthogonal factors (zero correlations).

1/\
The grouping of a linerally independent set of original

Variables defines a factor. The extent to which each vari-
le is correlated to the factor is represented by its

factor score. "

 

‘v

..'I

28

Using factor analysis for strictly variable

reduction is an attempt to make the method more useful in

marketing researchy. That is, by identifying and elim-

inating redundancy and linear dependence in data sets, we

gain greater efficiency in subsequent regression or dis-

crimant analysis (Harman 1967), (Green and Tull 1970).

Deriving the Predicting
Equations: Regression Analysis

Factor analysis is employed not only to reduce the
original variable set, but to do it in such a manner to
delete those variables having the greatest interdependence,

which would cause serious multicolinearity in subsequent

regression analysis.
The use of regression analysis on a variable
SUbset which has been "screened" by factor analysis affords

certain advantages. With the initial factor analysis, the

manifold relationships between all variables can be studied
at One time without running into the problems of multi-

c . o o o
ollnear1ty that are so often encountered 1n regre331on

a . .
1161137313, Once a relatlvely "close" subset of independent

V -
ariables is identified, the greater predictive power and

h
3rpotmesis testing ability of multiple regression can be

b
bought into play.

.1/\_—_
501' a thorough treatment of the analytical model used here,
81:1e author suggests Harman's text on Modern Factor Analy-
ls. For applications in marketing see: Green anchull,

\
R\Qsearch for Marketing Decisions.

an-m;_w~
I
i -. _ .

29

Multiple regression is more widely known than

factor analysis and its use in marketing and economic

research requires little elaboration (Gatty 1966), (Draper

and Smith 1966), (Salzman 1968). Basically, multiple

regression is a method for describing a relationship between

a dependent variable and a set of independent variables.

The "regression" itself is an equation developed to "explain"

this relationship .
Multiple regression in itself is not designed to

assist in additional screening of variables which do not

make a minimum contribution to the explanation of the

regression. However, a refinement of multiple regression,

Stepwise regression, is designed to accomplish this objec-

tive.
To test the significance of a variable to deter-

mine if it should remain in the regression equation, a ratio
of Quantified added influence of the last introduced vari-
able to the residual unexplained part of the regression is

developed. This is compared with a ratio of minimal accept-

a1ice chosen by the analyst. For the models presented in

t - . . .
his study, the m1n1mal acceptance rat1o was 0.01 and the

The specific multi-
1/

In ‘ - .
lnlmal ratio for removal was 0.005.
v.
aIrriate computer program used in the analysis was BMDOZR
l\

\ /
Bb’lIDOZR. 1966. Stepwise regression. Version of May 2.
IFlealth Science Computing Facility, UCLA, CA.

 

 

il ‘rf'r—{u

30

The program also permits evaluation of dummy
variables; i.e. , those variables considered important but
tarot easily quantified. For example, extreme difficulty was
(erucountered in trying to quantify "level of merchandising"
<3r1 a retail lot. Instead, this variable was entered as a
(inunmy, or classification variable, which simply indicates
whether or not the owner merchandised his product-y. The
program gives all measures of significance on dummy vari-

aih>1es as are recorded for continuous variables.

Eating the Model

It is difficult to define absolute criteria for
use in assessing the usefulness of a predicting model. That
143. how does one assess the utility of a model, or its
ability to predict tree sales accurately? The attitude
taken here is that two criteria should be satisfied.

1. Comparative Statistics of Models: Once a
model was specified in one market from one
year of cross section data, additional sets
of data from the same market were obtained
and models developed for the next two consecu-
tive years. The standard errors of the three

models were contrasted as well as the coeffi-

cients of multiple determination.

 

%

'

Ibflerchandising Activities" relate not only "what product
(ambinations are presented," i.e., grade, species, sizes,
t:c., but also "how they are presented," i.e., in stands,
Inder lights, etc. It does not relate to product price.

 

31

2. Comparing Estimated and Actual Sales: After

deriving the model in one market area, it
was used to predict sales of retailers in

another market area. Actual sales in the

new market area were contrasted with esti-

mated sales.

As three consecutive years of data (1967-69) were

collected in Winston-Salem, it was possible to develop three

predicting models. These three models were then contrasted

as described above. The three models developed in Winston-

Salem were used to estimate cut tree sales in Denver, Color-

ado. These estimates were then compared to actual sales

for the Denver area. Also, a model developed from Denver

market data was used to estimate sales from the four years

Of market data in Winston-Salem.

1m 5:
0'
‘—

ANALYSIS AND DISCUSSION OF RESULTS

This chapter will be devoted to assessing the
Winston-Salem natural Christmas tree market in several ways.
First, the market situation will be described as it existed,

showing both the retailer and the consumer factions and

f' woww

their resultant interactions. The market will then be
mathematically modeled via principal factor analysis, to
isolate independent spheres of influence. Once this has
been accomplished, a statistical model of the market inter-

actions will be deve10ped, from which market performance

can be predicted.

The Retail Market Structure
&Natural Christmas Trees

The retail market structure refers to the organi-
zagional characteristics of a retail market that determine
the relationship among sellers, among buyers, of sellers to
b“'lbrers, and of established sellers to potential sellers
(Greenwold, Douglas, and Associates 1965). In other words,
retail market structure refers to those characteristics of

I‘etail market that determine or influence the nature of
<2()ttljpetition within the market. The natural Christmas tree
6111 market structure can best be descr1bed by.

1. Characteristics of retailers in the market.

32

h

33
2. The extent to which the outputs of the

retailers in the market are viewed as non-

identical by buyers in the market.

3. The relative ease or difficulty with which

new retailers may enter the market.

Tree Sales

Christmas tree retailers in Winston-Salem sold

 

9 , 227 natural trees in 1967, 10,152 in 1968, and 11,941

trees in 1969 (Table 2). Balsam fir was by far the biggest

Seller each year, accounting for over 40 percent of all
In 1967, the Eastern Red cedar and White pine
From 1967 to

trees sold.
Were second and third in sales, respectively.
1969, sales of cedar trees fell almost 60 percent (767 trees)

and had slipped to fifth in popularity. White pine became

t:he second most popular tree in 1969, followed by Fraser

fir and Scotch pine. The large decline in sales of cedars

was attributed to a change in taste influenced by the

SuPply of better quality pine and fir. This is supported

by the fact that sales of Fraser fir increased over 100
peI‘eent (706 trees) between 1967 and 1969.
There were more unsold trees in 1969 than in either

of the two earlier years. In 1969, 17.5 percent of the
Christmas trees in stock were not sold compared with 14.7
thcent in 1967 and only 9.7 percent in 1968. Eastern Red
QQ'Clar led the list of unsold trees every year, with lot

0
DQrators discarding about one-fifth of their cedars.

‘

   

 

. 7:... p I\ C...

.92

e .:r.~— .z. .9.

i .snuerpssk

.5

e

‘IU ~ 58“.

u.nvrvu~ s

.1 PI...» .7 s L as,» I I

U

W.

7"

-|y

34

 

 

 

 

 

.lIIatlII

.n_0m no: mo_uoam\m
.ooscam _mcsumc commoo0canco_oo <\m

m.m_ m.m m.q_ u_0mc= atom o_nm

n__m>< _muOP to N

m.~m m.om m.mm _mHOP to N

_am.__ Nm_.o_ m-.m _muoh
m.mw mom m.mm mmo m.mm _mm \m:0mco>_mz
u- in it I: m.om o: mmota>o mco~_c<
:.mm mm In I: u: nu museum o:_m
I I I .. 92 mm 83:: 323
~._m mm u: m o.oo_ m. museum >m3coz
m.wm oom :._m mmo._ m.mn owo._ cmuoo cox ccoummm
Em Sm; Nam mom; ~.mm SR 23 £308
mum flora. :m Rt. 7% 3m; 2E 323
o.mm mm I: u: :: \muu c_t mm_m30o
m.mw wwm._ :.om mam :.~m m_m c_e commcm
m._m mn_.m «.mm mmm.: o.mm Nom.m . c_t sow—mm

o_nm__m>< o_nm__m>< o_nm_mm><
mouth to N consaz mouth to & consaz mooch to N conszz oa>h
mom. mom. nom—

 

aeeée .92 £23.85; 5 28 8a: “mettle...m £3

35

Generally, fewer Scotch pine Christmas trees had to be

discarded than any other species.

Prices
The retail price of Christmas trees varied accord-
111g to species (Table 3), but Fraser fir brought the highest

price of all the major species each year, and it also
Prices of

'l

enjoyed the greatest annual increase in sales.

all species increased between 1967 and 1968. However,

urinal”— _N

between 1968 and 1969, prices of the pines fell even below
their 1967 level. This change in the price structure is

partially accounted for by the influx of Canadian imported

Pines, mainly Scotch. These trees, marketed primarily by

Chain grocers, were purchased at slightly lower wholesale

Prices than local grown pines. Also, the large number of

I‘etail dropouts between 1968 and 1969 coupled with the high

Percentage of unsold trees in 1968 seemed to cause retailers

to depress market prices in 1969.
The average retail markup of the five major species

declined notably over the three years. The average markup

was 118 percent in 1967, 98 percent in 1968, and only 91

pe:t‘czent in 1969. With the exception of Fraser fir, the

a\7erage markup for each species fell between 1967 and 1969.
The overall increase in wholesale tree prices coupled with
e

e1ler resistance to raise prices in 1969 was responsible

SE
01‘ the decline in markups.

Chain grocery stores and discount establishments

a 3;)

A.

(II)

,_.-
,_1

(I)
( 3

Table 3.--Average retail price-
Winston-Salem, NC, 1967-69.

36

1/

of Christmas trees in

 

 

Species 1967 1968 1969
Balsam fir $2.84 $3.10 $3.34
Fraser fir 4.81 6.73 6.90
Douglas fir g/ a_1/ 7.00
ite pine 4.44 4.69 4.19
Scotch pine 4.55 4.76 4.11
astern Red cedar 2.08 2.32 2.86
Norway spruce 3.25 g/ g/
ite spruce 2.00 g/ a/
Blue spruce 51/ g/ 4.00
Arizona cypress 1.10 g/ g/
alvorson 2.00 2.32 2.08
Average 3.21 3.75 3.87

O 'n r -'

 

\

1‘ lWeighted by number trees sold.

a
‘ / Species not sold.

S<>1d Christmas trees at below average markups. However,

this pricing policy was consistent with the normal pricing
pOlicies of chain stores. The practice of pricing Christ-
111613 trees as "loss-leaders" was not observed at any chain

grocer's or discount stores.

Number and Type of Retail Lots
There were 53 natural Christmas tree retailers
in Winston-Salem in both 1967 and 1969. Only 46 retailers
SQ1d Christmas trees in 1968.
A wide variety of retail outlets sold Christmas

1:tt‘ees in Winston-Salem between 1967 and 1969 (Table 4).

u

37

Table 4.--Percent frequency of natural Christmas tree retail
lot types: Winston-Salem, NC, 1967-69.

 

Type of Retail

 

Establishment 1967 1968 1969
Chain grocery stores 32.1 37.0 36.0
Independent lots 28.3 19.5 18.8 '
Discount stores 13.2 13.0 13.2 1
Independent grocery stores 11.3 10.9 11.3 I:
Nurseries and florists 7.5 10.9 9.4 la
Service stations 1.9 6.5 7.5
Other 5.7 2.2 3.8

Total 100.0 100.0 100.0

However, one-third of all Christmas tree retailers in each
of the three years were chain grocery stores. Independent
lots (including church and civic groups) accounted for about
28 percent of all retailers in 1967 and almost 20 percent
in each of 1968 and'l969. Discount stores, independent
grocers, nurseries and florists, service stations, and other
dealers made up the remaining retail outlets.

Of the seven distinct types of natural Christmas
tree retail lots in Winston-Salem, independent lots had
the best overall market performance. A three-year average
(315 44 percent of total sales was held by this group, which

QQ‘nstituted an average of only 22 percent of the total

u

 

38

retail lot types. Discount stores ranked second in perform-

ance with 15 percent of total sales while representing only

13 percent of the population of retailers. The poorest

performing group was independent grocers. In this group

only 3 percent of the market was held by 11 percent of the

retail outlet population.

The three largest groups of retailers, chain

grocery stores, independent lots, and discount stores, com-

Prised 70 percent of the total retail population and held

85 percent of the market sales.

Retail Lot Size

Relative size in the market place is normally

associated with power. With bigness comes the ability

to dictate market conditions through actions. Large Christ-

mas tree retailers did exist in the Winston-Salem market.

In each of the three study years, 20 percent of the retailer

population (10 retailers) held 60 percent of total market

SEiles. The five largest retailers sold about 40 percent

of all trees. As an indicator of market impact per retail

lot; the ten largest retailers averaged 722 sales per lot,
whereas the remaining lots averaged only 110 sales.
It is also significant to note that all ten large

r O O O O
etailers were either a chain grocery store, independent

1 .
Qt, or a discount store.

‘1'. I'mu-m
_ . I.

39

Lot Location

Christmas tree retailers in Winston-Salem

located where consumers went to spend money. In 1969,

81 percent of all tree retailers were located in business

areas; only 19 percent were found in residential areas.

Of greater significance, 66 percent of the retailers were __
lOcated in the four major and two minor shopping areas of l
the city (Figure 4). I "

Christmas tree retailers who located in retail I:

business areas sold more trees and discarded fewer each
year than did those retailers in residential areas. In
1969, the average retailer in a business area sold 241

1:I‘ees, but the average retailer on a residential street

c>t‘11y sold 156 trees. The retailer in the business area also

had fewer trees left over.
In addition to selling more trees, retailers in

business areas received a higher price for their trees. In

1969, the average price of trees sold in business areas was
$4.04; the price in residential areas was $3.76. The aver-
age price of trees in business areas in 1968 was over $1.00

IncDre than the average price in residential areas.

Type of Street
About half of all Christmas tree retailers in

1/

Winston-Salem were located on major traffic arteries- .

b

laverage daily traffic volume of over 10,000 vehicles.

¥

40

 

 

 

 

 

 

Singular Retail Locations Mile Scale
City Limits 0 k 1
e‘hsus Tract Boundaries

 

 

Multiple Retail Locationse
(Major 6: Minor Retail Areas)

17'
ignre 4.--Concentrations of Winston-Salem natural Christmas
tree retailers.

 

41

And these retailers sold more trees and received higher
prices than retailers located on minor streets.

In 1969, each retailer in Winston-Salem sold an
average of 225 trees. But the retailer located on a major
S treet sold an average of 340 trees. The retailer located
on a major street in 1968 sold an average of 351 trees,
While the retailer on a minor traffic artery sold only 111
trees. In all three study years, retailers on major streets
also discarded fewer trees.

Prices were also higher at tree lots located on
major streets. In 1969, the average price of trees sold
by retailers on major streets was $.87 higher than tree
Prices at lots on minor streets. The average price differ-
eI‘Lces in 1967 and 1968 were $.51 and $1.12, respectively.

Over the three-year period, there was a gradual
increase in the proportion of Christmas tree retailers
located on major traffic arteries. This trend will
undoubtedly continue as more and more retailers realize
the importance of selling trees in areas of consumer pur-

cI'lasing activity.

Parking Facilities
Most Christmas tree retailers in Winston-Salem
realized the importance of large and easily accessible
parking facilities to their business. Each year, about 70
Deilz'cent of the retail outlets had adequate parking facili-

ies. The others had either small and cramped parking

¥

   

_ .. \n .514“; “nu-IN

.‘J‘L‘-

42

lots -- 20 percent of all retailers -- or none at all.

Buyers of natural Christmas trees obviously

prefer the convenience of large and easily accessible

parking lots. Retailers providing adequate parking sold

Over twice as many trees as did other retailers, and they

had fewer trees left over Christmas day. In 1969, for

example, retailers with inadequate parking facilities had L

to discard two trees for every three sold.

I.= 10.7....

Tree prices were also higher at locations with

‘t
::

large parking areas. This illustrates the fact that

buyers are willing to pay extra to shop where convenient

Parking facilities are available.

Merchandising
A product may be merchandised or promoted in

Several ways. Advertising and product display were the

most widely used methods to merchandise trees at retail

in Winston-Salem.
About one-half of all retailers advertised their

Christmas trees each year. This advertising included

t1ewspaper, television, radio ads, handbills, trading stamp
£3Pecials, special displays, and direct mail campaigns
(Figure 5). Newspaper advertising was the most common.
3:11 1967, about 40 percent of the retailers advertised in
thspapers, and in each of 1968 and 1969, the number was
{3*3Lmost 50 percent. Newspaper advertising was heavily used

by chain grocery stores and discount establishments.

‘

   

43

 

 

 

 

Percent
50 Q 1967
0..
;.; [J 1968
O
O
O
- E3 1969
40 .
GOV. .
to: -
Q .
v. °
v. °
>0 .
Q .
w. -
>0
. 4 O
0! .
us
30 50 '
>0 0
. .
>0 .
. 4
>0 .
. 4
h! .
>0
. 0 0
>0
. 0
> 0 ’
Q 1
>0 -
.0 1
v. '
P . o
2 O 8.: .
>0
fix ’
O i 0
>0
0« .
50.
N3 :.
m .
P . .0
w - -
£0 '
10 av " ..
Q 4 . o
P.” o o _‘
O
9" ' o 0
D O 0 o
'z.: . . 0..
' . . ,.Q .0.
D’.‘ 9.‘ >'.‘ ..° . . v
0" 0.0 ’d 9. o o ’.s . a
’0‘ .‘. 90‘ 90 -‘- 90‘ ' ’0
04 . 00 0 . 04 a 0 V
P." . . ..‘ . 0 o ’0" o b. p o
2% a wk :9 3. an ' m’ ':
Newspaper Television Radio Handbl s 0t er

Advertising Medium

 

I‘wi—gure 5.--Percentage of Christmas tree lots in Winston-Salem

that advertised by various media, 1967 - 1969.

‘P..‘L!'. .

. ‘9‘

I“
.12.

44

Christmas tree retailers made very little use of
radio and television to reach the public. In 1969, only
six retailers used either of these two media. Three of
these retailers were independent lots and all were large
outlets. In 1968, four of the five users of radio and
television advertising were large independent lots.

Retailers who advertised sold more trees and had
fewer unsold trees than those who did not advertise. In
1968, each retailer who advertised sold an average of 337
trees and had only 6.6 percent of his original supply of
trees left over on Christmas day. The typical retailer

Who did not advertise, sold only 105 trees and had to
dispose of almost one-fifth of his trees. The retailer who
advertised in 1969 sold an average of 312 trees compared to
an average of 225 trees for all retailers.

Christmas tree prices were generally higher on
retail lots that advertised. In 1969, the average price
received by retailers who advertised was $.19 above the
average price received by those retailers who did not
advertise. The average price differences in 1967 and 1968
Were $.35 and $.64, respectively.

Like advertising, an attractive display of Christ-
mas trees is a way to lure customers to retail lots and
influence them to buy. However, Christmas tree retailers

in the Winston-Salem market did not take full advantage of

t1“-215 merchandising technique. Many of them -- 41 percent

_._...._. _ _
:1“ .v 1 ____,__, .'—; W
5
‘ i

\_ _

45

in 1969 -- simply leaned their trees against a building,

stand, or rope. And over one-third of the retailers in
1969 piled most of their trees on the ground. Only one-
fourth of the retailers displayed their trees standing
upright and separately for ease of selection. In both 1967
and 1968,

there were even fewer attractive and convenient
displays.

The manner in which Christmas trees were dis-

p layed on the lot in the Winston-Salem area affected both
tree sales and prices. A retailer whose trees stood upright
and separately, sold over twice as many trees as a retailer

who leaned his trees against a building or stacked them
in piles on the ground.

Retailers with good displays also
received higher prices for their trees.

Product Differentiation

In the three years of study, no strong evidence
of product differentiation existed in retailing natural

Czli‘l—thistmas trees in Winston-Salem.

Although several species and qualities of trees

were sold in the market place, no attempt was made to dif-
I'entiate among them. In fact, for the majority of trees

0 fiered, any attempt to differentiate would have been false
aqurtising at best.

Many trees supplied to this market
i Q:re nonplantation Canadian Balsam fir and Scotch pine of

19v quality. These trees represented a fairly homogenous
Dli‘oduct to the consumer.

 

in“

46

Brand differentiation was not used when, in fact,

it could probably have been used effectively. For example,

many consumers were aware of the poor quality associated
wi th Canadian imported trees and would have reacted favor-

ably to advertisements branding trees as "North Carolina

fir" or "native pine." It would even have been feasible

for long established independent lots to brand their trees

with their own names. Two such lots were local nursery

men and Christmas tree growers, who produced their own trees.
Tree quality was not explicitly used to differen-
Not one lot in the three years of study

tiate trees sold.

S old any trees on grade. However, retailers did use the

pricing mechanism to imply quality differences.
Slight quality differences were apparent both
among the trees on a particular lot and among lots.
PIi‘cnbably only two lots had significantly better trees, where
adVertising quality differences would have been effective.
These lots carried predominately local-grown plantation
I.‘3:‘eser fir and Scotch pine. However, the consumer would
ha-Ve had to purchase at the two lots in previous years,
ern informed by others, or visited several lots including
1:‘ln-ese to have known of the quality differences. That is to

S 337, that even the lots with the best trees in the city did

LQ1: capitalize on their product differences.

L‘ -.-' . v .n‘meo. f5
.1 .

 

47

Ease of Market Entry
Entry relates to the ease or difficulty a new
(Hiristmas tree retailer encounters in becoming a member of‘
a1 ggroup of competing natural Christmas tree retailers.
Barriers to entry in the natural Christmas tree
'Ireertail trade in Winston-Salem did not act as a constraint
t:<3> potential retailers. Ostensively, new retail lot owners L
in Winston-Salem overcame three factors of market entry that _
would appear critical to one entering this retail trade.
These factors could be clasified as:
1. Information concerning short run profits and
risks.
2. Access to a supply of natural Christmas trees.
3. Ability to overcome certain other natural or
artificial barriers to entry (e.g., supply
of trees, licensing, market outlet).
Probably the most influencing factor is informa-
.t:jL-<>n regarding short run profits. As there are no annual
[It‘siisborts released locally on the profitability of the enter-
‘F>:t:’jise, information must be dispersed by word-of—mouth.
A“km-12:31, since existing retailers are the major information
SEQ:P‘Iatrce, they can partially control future attitude toward
Gaytrlftry. This, no doubt, accounted in part for the fluctu-
EiTt2ning numbers of retail lots over the three-year period.
Tll‘nl good years, retailers reflect optimism, influencing
(>"tlhers to enter the market the following year. A depressed

[Iléirket the following year, due to too many retailers, causes

48

retailers to impart pessimistic views in discussions with

potential future retailers. This, in effect, characterizes

the market information system associated with the Winston-

S a lem market.
Once a potential retailer has made the decision

to enter the retail market, he is confronted by few con-

M,"

S traints. The element of risk is minimized due to the low

investment required. Essentially, all that is required

is a tree inventory, a leased lot, and a small labor input.

12"“‘11—33—

For a small operator this means an initial outlay of $200

to $400. Because nearly all market control factors are

internalized (i.e. , the lot owner has complete freedom in
the market arena), the risk is minimized even more.

During the three-year study period, availability

of trees for sale did not act as a constraint to market

el'ltry. As tastes changed, and the Eastern Red cedar lost

E>QIzularity, Canadian imported pines and fir were available

these trees could be pur-

This not only

£01: retailers. As with cedar,
Q1‘lased at a low unit price ($1.00 - $1.75).
af forded the new retailer an opportunity to maximize income
(high markup per unit), but it also affords minimum loss

( low investment risk).

Entry could be impeded by an insufficient supply

613 high quality natural trees in the future. The Winston-

Se.lem market is going through a transformation in which the
QQnsumer has been given the opportunity to select between

low price wild trees and high priced plantation trees, and

49

he has chosen the latter. Presently, there are insufficient

numbers of plantation-grown trees to supply the Winston-

Salem market. Because of the limited number of plantation
growers in the area, and concentrated demand from other
nearby cities such as Raleigh, High Point, and Charlotte,
several retailers are already buying trees quite distant
from the Winston—Salem marketarea. For example, the
increasing demand for Fraser fir has caused retailers to

procure trees from plantations in the mountainous regions

 

of western North Carolina. This has also resulted in a
high wholesale cost for this species, and a low markup rela-
tive to other species. Therefore, not only is the short
run profit potential depressed, but risk is increased. This
should become more important in future seasons as a natural
c-‘-C3'rlstraint to entry for new retailers.

Other natural constraints on market entry are those
costs associated with obtaining a retail license, renting
a lot, and actual operation of the lot. Renting a lot
ei ther involved a direct rental agreement for a definite
period of time, or profit sharing arrangements between
117% tailer and land owner. Retail licenses were awarded quite
freely for a $50 fee. Most retail outlets already involved
in another retail trade, had a retailer's license to con-
zEQrm to their normal business requirements. However, the
IT‘Qjority of retailers that entered the market over the

1:‘ll‘iree-year period did purchase a license as a prerequisite

he entry. The former approach is common in residential

50

areas and the latter dominates business areas.
Land rent costs in Winston-Salem were not exorbi-

tant, generally running less than 5 percent of gross sales.
Labor costs, although cheap in absolute terms ($1.00/hour),
are relatively expensive when considering cost per tree
sold. Labor costs rank second to the costs of purchasing
the trees. However, for small, independent lots, labor 1
requirements are satisfied through the lot owner, posing

little resistance to entry.

1““477r .—.— “ .I _r

An artificial constraint to entry is a city ordi-
nance restricting street vendors to within the city limits,
and restrictions on the time trees can be displayed for
S ale. For example, lots could not open in 1969 until
December 14. This automatically limits the retail season
to 11 days.

Another set of constraints are normally imposed
by existing retailers in the general market area. However,
e\72‘Ldence of strong competitiveness between retailers was
1'19 t: present in the Winston-Salem market. The normal mania
O ‘13 roadside vendors was not observed, as is generally true
0 3‘3 mellon and fruit stands in the midwest and pecan stands
5‘11 the south. Retailers made little attempt to differen-
t iate on price, selection, or quality. Strong price compe-
1:jltion, which is a way of life with roadside vendors,

appeared very weak in retailing Christmas trees.

To substantiate this lack of barriers in Winston-

Selem, one needs only to observe the number of retailers

 

51

entering the market each year. In each of the 1967 and
.1969 seasons, 53 retailers sold trees. Between 1967 and
.15968, 13 retailers left and 6 entered. And, between 1968

£11311 1969, 11 retailers entered and 4 left. Of the 65 differ-
ent retailers who sold trees between 1967 and 1969, 23 --
c>::r 35 percent -- lasted just one year. Only 37 retailers -— .
55'7' percent -- were in the tree business all three years. L!
3E1C1 total, 34 retailers either left or entered the business !
over the three-year period.

If we were to establish a cause for the number of
"ijailures"l/ in the natural tree retail trade, we would have
12><> call it poor market information. All retailers entering
and exiting the market were small, and most left the same
39'€2£ar they entered. As we have shown, these small retailers
eazgifiibited extremely poor market knowledge.

€%§§§L§§ociating Retail Lot,
-£EElsge Sales, and Consumers

One short statement could relate the market asso-

czdjL-Eations in Winston-Salem during the study years; "high

I:la-‘l:ural Christmas tree sales coincide with areas of high

I: . . .
Q tail act1v1ty."

Areas of high retail activity in Winston-Salem

LiStd several characteristics in common.

‘\\\\‘»

l

7‘ Jf'fhe term ”failure" may be somewhat misleading. A small
segment of retailers either planned to spend only one
season in the market, or did not return the following
year for personal or business reasons other than exces-

sive financial losses.

52

1. Major retail areas (Wingate and Corbin 1956)
were located on major traffic arteries (AADV
10,000). Minor retail areas were located on
traffic arteries with an average daily traffic
volume of 5,000 to 10,000l/.

2. Both major and minor retail areas were within
one mile of major medium and high population
concentrations.

3. Four of the seven retail areas (the four major
retail areas) were incorporated with a shopping
center complexg/. One of the minor shopping
areas was located near a small shopping center.

4. High retail sales areas were located nearer
high income residential areas than lower income
residential areas.

Thirty-four of the 53 retailers in 1969 were

:1“:><:ated in the major and minor retail areas of the city.
E:‘V-i’en more significant, 29, or 55 percent, of the 53 outlets
‘flr‘shnre located at the four major shopping areas (Table 5).

Winston-Salem is not unlike most American cities

:i“tl. its adoption of planned suburban shopping center complexes.
The "suburban sprawl" and "shopping center mania" of the
:i:-_

\

The distinction here between major and minor retail areas
is predicated primarily on the existence of a shopping
center complex at major retail areas.

1’A shopping center complex is defined here as a planned
retailing center housing a department store, chain grocery
store, discount store, and associated speciality stores.

53

fifties and sixties has transferred the cities retail trade
from the center city to several satellite retail areas
located in the suburban fringe. The center city still main-
tains an active retail trade, however, shifts have occurred
in types of goods offered and relative volumes sold.

Many Christmas tree retailers have, through casual
Observations or trial and error, keyed on this transfer of
retail activity. For example, the largest major retail
areas were D, E, and B in order of size (Table 5). The
number of retailers located at each was 9, 8, and 5 respect-
fully, or 22 total. Further, the two largest retailers in
the city located in or near area D (Figure 6). The next
largest retailer was located in area B, the fourth largest
I‘etailer in area B, and the sixth largest in area C.

All retail operators selling over 500 trees
a*rltlually located at a major or minor retail area, establish-

ing a definite association between lot size and their loca-
tion relative to retail sales potential. That is, it is
he. 1: by chance that the largest retail concentration occurs
at: area D, and also that the two largest retailers located
there, nor is it by chance that the second largest retailer
QQ‘mcentration is at area E and also that the third largest
thailer is located there. These two areas offered the
It‘Qqcimum sales potential.

Retailer concentration is best illustrated by

tlie fact that 41 percent of the retailers were located in

three retail areas, no larger than six city blocks. But

{thrill} '1 clinic. a

.motm __muoc 003mm_a.l

 

54

 

\u
.moum __muoc L0c_z\m
.motm __muoc cOmmz\m
.mump uomcu mamcmu omm. pcm .mump ucsoo o_wmmcu pcm .LoE:mcou .__mumn mom. :0 pommn o_nmb\fl
m-0.-0m-mm
000.0 -0_-0_-0-0-0-0
0. -000.0 000.0 000.00 0. -0_-0-_-0-__ 0.0 000
0 000.0 000.0 000.0 m 00-00-00-00 0.0 \m
00-0.
00 000.0_ 00_.0 000.00 0 -0-m-00-0_-_0 0.0 \m
00-00-00-00
00 000.00 000.0 000._m m -_0-0_-m-__-0_ 0.0 \m0
0 000.0 000.0. 000.0 0 00-00-00 0.0 \m
0. 000.0. 000.0 000.0 0 00-00-m_ 0.0 \m0
0 000.0. 000.0 000.00 0 0_-0_-0-0-m 0.0 \ma
v.0: uoxnmz mE:_o> 0500c. compo—:aod mum—_muom moc< uoxtmz c. moth ommrocam moc<
mo w u_m$mch Lossmcoo so a muomch mamcoo Oh po_o>mLh uoxcmz
>__mo .o>< mmmco>< Loeacho mm__z

 

Iuoz .Em_mmucoumc_3 "ma_rmco_um_oc uoxcms motu mmEum_Lro _mcaumzuu.m o_nmh

\—

55

 

 

 

 

 

Major Arteries'---

Mile Scale
Minor Arteries ——-——— 0 3 1
City Limits ------
1 - 39, Census Tracts Segment of
Market Held

Figure 6.--Winston-Salem major traffic arteries and natural
tree market held by major and minor retail areas.

 

56

this combined six-block area of intense retail activity
supported 74 percent of the total natural Christmas tree
market for the city.

And why are the retailers at these locations?
Categorically, we would have to say that these are the most
viable Christmas tree market areas in the city. Each re-
tailer at these locations held an average of 3 percent of
total market sales; whereas, in the remaining market areas,
retailers enjoyed only an average of 1 percent of market
sales. There are definite market associations which pro-
vide explanation for these observations.

Associating Market Factors
with Tree Sales

 

 

In evaluating the retail potential of a regional
area, a metropolitan area, or segment of a city area, as
is the case here, certain variables come to the forefront;
population, purchasing power, and distance.

Population has long been cited as a critical
factor in analyzing retail potential of an area (Thompson
1964). Some market analysts classify it as the primary
factor upon which so many other factors depend. Contemporary
market analysts argue that raw population counts are becom-
ing more important as sales predictors due to greater homo-
genity in the consuming sector (Thompson 1964).

In the study area, population was critical to the
Christmas tree retailers sales. However, this does not

mean that the retailer should locate in the center of a

57

suburban housing development. Because if he does, he will,
in effect, isolate himself from the majority of potential
consumers. He should have a location that gains access

to an area population of potential retail sales. In Figure
6 we see that areas A, B, C, D, and E are located on the
fringes of the city, in areas the census bureau defines as
suburban. Each is approximately two miles distant from

the other. If we define these as primarily separate market
areas, their potential sales population becomes as indicated
in Table 5. We see how critical population can be with the
two leading sales areas D and E. Percentage of market held
coincides with population. But, what of the third largest
retail area, area B, which has a representative population
of only 4,000 in its market area. In fact, it affords an
excellent example of why demand is not predicated on one or
two variables alone, but on a combination of several inter-
acting variables.

Retail area B actually services more population
than is indicated in Table 5. Two factors, type of road
systems serving this area and the demographic characteristics
of the consumers, contributed to making area B a more viable
area for tree sales than is indicated by the raw population
count. Area B is serviced by a major traffic artery with
an AADV count only slightly less than that of area E. All
of the traffic exiting northeast from the city passes the

area .

9..

 

2627””"3‘373‘30

1020212121303:

58

(consumers residence)
'5 I. '7 il‘

Destination of Tree Sales (Census Tract)
.910I|flfl“

ll?
all

C

 

E [\AIV \E._F./\D(B,
.II 0 .III'.
II n . .Illlu
III I00IIIIu
.I- ll 0...
ll. .II 0...
ll .II .0.-
.l .l-
.I.I cool
I. o .II'I.
.l' l .I'ILI
I-‘ coolllI
u'l 00-Il'l
.ll - llll
pl. unallll
I... I. I...
I'l .Il-llll
on .0.
.0-
. up.
a. on. cool
.I 0 "II-
" I'll.
ll' . III.
I'l l-JIIII
III I. all.
In. .l- a...
no.
a no.
ll .- llll

 

0 n n u a u u u n u n u u u u u u u
Amory ommmrocza coEJmcoo ocorzv .
Auomth mamcoov mo_mm.oock wo.c_m_co

I I l 0
I. I I
1969.

Winston—Salem, NC,

3 trees.

a
C

.4-Natural Christmas tree-purchase patterns by census
tract:

Represents sales 0

/

 

é
Figure 7

u
3'
3 7
30‘

 

 

 

ar
d5
P0
in
Sa.

at

59

A second factor causing high sales at area B is
the demographic characteristics of the population being
served by the retail area. Buyers of natural trees in this
area were medium middle income to upper middle income fami-
lies. One of their basic market attributes is that they
"shop" in the market place, rather than just "buy." Because
they shop, they travel farther, as indicated in Table 5.

Because middle income families travel farther to
purchase a tree, area B with a 2.5 mile average purchase
radius was obtaining sales from census tracts 4, 25, and 12.
These were not included in the population count because
this market area was more closely controlled by retail
areas C and D.

A counter-argument could then be presented for
Area A. It has a very large population in its retail area
(23,600), and is serviced by a major traffic artery
(15,000 AADV). Yet, area A held only 8 percent of the
market with 7 retailers. The prime contributing factor
here is income. Low income areas are not viable natural
Christmas tree market areas (Troxell 1970) and (Drysdale
and Nausedas 1970). This is true even though a high inci-
dence of children occur in these families, which is a
positive factor for sales when considering upper middle
income families. This is further supported by retail
sales in the diffuse area (G). Here 17 retailers locating
at 6 - 7 blocks distance obtained on an individual basis

only 0.7 percent of total sales. In this area two factors

 

 

 

the
ser
are
lie

"Sh

pur

rad

 

59

A second factor causing high sales at area B is
the demographic characteristics of the population being
served by the retail area. Buyers of natural trees in this
area were medium middle income to upper middle income fami-
lies. One of their basic market attributes is that they
"shop" in the market place, rather than just "buy." Because
they shop, they travel farther, as indicated in Table 5.

Because middle income families travel farther to
purchase a tree, area B with a 2.5 mile average purchase
radius was obtaining sales from census tracts 4, 25, and 12.
These were not included in the population count because
this market area was more closely controlled by retail
areas C and D.

A counter-argument could then be presented for
Area A. It has a very large population in its retail area
(23,600), and is serviced by a major traffic artery
(15,000 AADV). Yet, area A held only 8 percent of the
market with 7 retailers. The prime contributing factor
here is income. Low income areas are not viable natural
Christmas tree market areas (Troxell 1970) and (Drysdale
and Nausedas 1970). This is true even though a high inci-
dence of children occur in these families, which is a
positive factor for sales when considering upper middle
income families. This is further supported by retail
sales in the diffuse area (G). Here 17 retailers locating
at 6 - 7 blocks distance obtained on an individual basis

only 0.7 percent of total sales. In this area two factors

Y.‘ ~C-

60

were critical; traffic arteries were minor, 2,000 - 4,000
AADV and income level was the lowest for the city. Popula-
tion count was high but ineffective.

We can observe from the preceding analysis that
road systems (AADV) and average income are both associated
with distances consumers travel to purchase a natural tree
and other consumer goods. Distance traveled to purchase a
natural tree is obviously a factor affecting the sales per-
formance of all retailers. Because it also affects cross-
over sales from one retailer (census tract) to another, it
was given additional attention in this study.

Validity of the
Crossover Assumption

 

 

One of the basic assumptions of this study was
that the consumer buys his tree within the census tract
where he resides, and, therefore, does not cross census
tract boundaries to purchase trees. This hypothesis, that
sales are a function of the demographic characteristics of
only those persons residing in the same census tract where
the lot is located, is unfounded.

As we have demonstrated, the consumer departs
very little from his normal purchasing pattern when buying
a Christmas tree. He doesn't mind traveling to acquire his
tree. In fact, most often he travels two miles to purchase
it. Neither does he strive to minimize the effort on this
purchase. Retailers observed that consumers may shop two

or more lots before deciding on a tree. And 15 minutes in

61

deciding on a particular tree purchase was commonplace.

One factor disclosed concerning consumer behavior
was that a definite relationship exists between percent of
crossover sales and buyer income (Table 6). Persons with
higher incomes were more likely to purchase trees within the
census tract where they lived, partly because these tracts
were larger, and also because retail lots chose these tracts
for a location. For example, 75 percent of the households
with incomes less than $3,999 who bought a natural tree, pur-
chased it out of the census tract which they livedl/. Fifty
percent of the households with incomes between $4,000 and
$10,000 and 34 percent with incomes over $10,000 also pur-
chased trees out of the tracts where they resided. Obviously,
tree sales within a given census tract are not entirely
dependent on the demographic characteristics of the persons
living in that tract. Also, in this study it is likely that
a greater specification error occurs when low income tracts
are considered than when high income tracts are considered.

Analysis of individual consumer behavior also
disclosed that the consumer travels considerable distance
to purchase his tree rather than to purchase it at a
closer, more convenient location (Table 7). The study
results verify this conclusion.

The average distance traveled to purchase a natural

tree in Winston-Salem.was 2.5 miles. This distance varied

 

;/The census tract described here includes the census tract
being studied, plus the immediately adjacent census tracts.

62

Table 6.--Association between income class and tree purchases
out of census tract.

 

 

Income Percent In Percent Out
Class Tract Purchase Tract Purchase
$ 3,999 - 25 75
4,000 — 6,999 49 51
7,000 - 9,999 48 52
10,000 + 66 34

 

Table 7.--Association between income class and miles traveled
to purchase a natural tree.

 

 

Income Average Number of
Class Miles Traveled Observations
$ 3,999 - 1.8 13
4,000 - 6,999 2.5 89
7,000 - 9,999 2.8 34
10,000 + 2.0 46

 

depending upon the buyers and sellers involved. Buyers with
incomes under $4,000 traveled an average of 1.8 miles to

get their trees, and buyers with incomes above $10,000
traveled 2.0 miles to make their purchases. The population
making up the mass purchase power, those with incomes

between $4,000 - $10,000, traveled an average of 2.8 miles.

63

Also, consumer migration data was developed to
assess origin of sales at a particular location. From
these data a grid showing sales movement was developed to
illustrate high consumer mobility in selected areas of the
city (Figure 7). The figure is constructed to be read
from both axis, permitting an evaluation of where consumers
lived, who purchased trees, and to what census tract they
traveled to make their purchases.

For our original hypothesis to be correct, the
matrix would have all sales recorded on the diagonal, running
from the upper left to the lower right. A significant
number of tree purchases did occur within the census tract
as depicted by the grouped sales along the diagonal. How-
ever, considerable off-diagonal sales exist in the form of
bands. These represent sales originating at one of the
major or minor retail areas.

Persons in low income census tracts 1-9 purchased
their trees from the diffuse retail area and areas A, D,
and E. This is also verified by the fact that these persons
traveled about 1.8 miles to buy a tree, which is only
slightly over the distance across two census tracts in the
eastern side of the city. To have gotten out to market
areas B, C, and F would have meant traveling three miles
or farther.

Study of the bands of sales running horizontally
across the matrix, reveals the origin of tree sales for

each market area. For example, area E is composed of

64

pattern El and E2, located on the corners of census tracts
19, 9, and 10. Those lots sold to persons in census tracts
9-11 and 19-22, all of which bound census tracts 19 and 9
on the north, west, and south. Two major traffic arteries,
Silas and Petercreek Parkways, service the area.

Area D, located in both tracts 24 and 25, was
serviced by an interstate expressway, Stratford Road, and
Northeast Blvd., all high volume traffic roads (Figure 6).
The bulk of this area's sales went to census tracts 9-12,
which are immediately adjacent on the east and northeast;
tracts 21-26, which are directly adjacent on the south, west,
and northwest; and tracts 37-39, which surround the general
sales area on the south and west.

Further elaboration would only serve to strengthen
the argument that consumers do travel considerable distances
to select and purchase a natural Christmas tree. Also, the
percentage doing so is high, approaching 50 percent in most
income classes.

This factor was extremely important in the distri-
bution of natural trees in Winston-Salem. Its implication
to the predicting models was also significant, as will be
discussed later.

We have clarified one point. Success in retailing
natural trees is associated with locating the retail lot in
or near major retail sales areas. However, this does not
just happen. As we have shown, there are many market associ-

ations which contribute to, detract from, or have a neutral

65

effect on sales performance. However, as is evident from
preceding sections, we cannot fully analyze these associ-
ations without analytical models. To determine interactions
among only three variables; income, distance, and population
becomes an exercise in futility, without such analytical
models. Granted, we can observe certain associations, but
as we bring in more variables, we must develop analytical

models to assess the interactions.

Principal Factor Analysis

 

The intent of this research is not to provide a
definitive treatment of the principal factor model, nor to
assess the general competence of the model as an acceptable
analytical tool. Reference is made to Harman (1967) for a
detailed discussion of the principal factor model and
Green and Tull (1970) for its applicability to this analysis.

The term "factor analysis" represents a group of
methods for analyzing intercorrelations within a set of
variables. The basic intent for using the model in this
study is to summarize information contained in our original
105 variable set into a smaller set of linear independent
factors, with a minimum loss of information.

Factor analysis has had a wide array of applica-
tions, and, in fact, served effectively in all capacities.
Its utility relative to other analytical tools, such as
cross-classification, regression, discriminant analysis, etc. ,

is one of complementarity and not competitiveness.

66

Principal factor analysis is closely related to
principal component analysis. Principal component analysis
of an original correlation matrix will produce a set of
factors equal to the number of original variables analyzed.
That is, the model attempts to explain all the variance
associated with each variable.

Principal factor analysis does not ascribe entirely
to the above approach. Here the attempt is to maximize the
variance explained for each variable while minimizing the
number of factors created. That is, although each variable
is assumed to have a sample variance of one, we will only
attempt to explain a major portion of this variance. This
portion is determined by a method described in Harman and
is dependent on the correlations of the respective variables.
These percentages of total variance, which will be explained
for each variable, are denoted as the communalities for

each variable and will be represented by ”hz," where

0<h2<1.

The Variable Set

 

In Table 1, Appendix B, two lists of variables
are presented; a list of specific variables for which data
were recorded, and a list of general variables represented
by one or more specific variables.

In developing the survey schedules and deriving
the variable list, we purposely selected more than one

specific variable to represent general variables such as

67

race, families, households, tree prices, competition, etc.
These general variables are not to be confused with the
factors sought. Although we will make association between
the final factors created and these general variables, they
were not developed for that purpose. They simply serve as

a general reference or index to the specific variables.

Factor analysis of the independent variable set

(105 variables) is accomplished independent of the dependent
variable tree sales (YS). Therefore, no association or
inference is made regarding the structuring of the factors

and their influence on tree sales.

Creation of Factors

Nineteen principal factors were created from the
original correlation matrix (105 x 105). The original
communalities created to replace unities in the correla-
tion matrix ranged between .90 and .95. That is, the model
attempted to explain on the average somewhat less than 95
percent of the communality (variance) of each variable.

For 105 variables, the total communality was 97.81, giving
an average of 93 percent variance explained for each vari-
able.

Creation of only 19 factors resulted from selec-
tion of an eigenvalue of 1.00 as a cutoff point. A partial
factor matrix (14 x 19) with the resulting factor scores
appears in Table 8, Appendix B. The model assumes that the

19 factors or dimensions are independent of one another

68

(zero correlation) and that each in turn provides a descrip-
tion of a different linerally independent variance source.
Through inference of the highest loadings on each factor,

we were able to subjectively identify these 19 factors.

Subjective Assessment of Factors

 

Most of the beginning work in factor analysis had
as its end product the listing of variables and their respec-
tive scores on each linerally independent factor. Through
interpretation of the variable factor scores appearing on
each factor, researchers formulated a subjective description
of the factor, i.e., what the factor purportedly represented.
This introduction of subjectivity as a final result of the
analytical model caused criticism of factor analysis. Yet,
analysis of the output in this fashion appears extremely
useful, as it forces the scientist to make associations
between the analyzed data and the original field observa-
tions.

Using the factor model for screening, we selected
the highest scoring variable on each factor, because theo-
retically these variables explain the largest percentage
of total variance associated with each factor. Through
careful study of the remaining variables we can see how
and why they are linked to the factor and to the high
scoring variables selected. To gain this inference requires
some knowledge of the factors.

The principal factor method results in the

69

creation of two types of factors; General and Unique
Factors. The General Factor, normally created first, has
relatively high loadings for all variables, therefore, the
interpretive power on the factor is usually low. Unique
Factors generally have fewer variables with high loadings,
and consequently, are more readily interpreted. However,
some may occur with either all low loadings, or only one
high loading (Rummel 1970). These factors are difficult

to interpret, and will be referred to here as "Unique Low
Load Factors." It is feasible that additional matrix rota-
tion or a different rotation method might have been helpful
in their interpretation.

To properly describe either a general or a unique
factor we are entirely dependent on the factor scores.
These scores simply relate the correlation between the
variable and the factor, and can vary between -l.OO and
1.00. Scores of a magnitude less than 0.30 have been
ignored by most researchers. A negative loading implies
a negative reaction of the variable to its respective factor,
and singular high loadings, whether positive or negative,
are normally considered insignificant.

Based on model concepts, data presented, and
field observations, the following descriptions are presented
for the 19 factors. The variables and factor loadings are

presented in Tables 9A - 98, Appendix B.

70

Factor #1; Table 9A, App. B; General Factor:

 

As is the case of many factor solutions, a general factor

is somewhat difficult to interpret and relates little infor-
mation. Over 50 percent of the variables loaded on this
factor with a score exceeding 0.80.

Factor #2; Table 9B, App. B; Unique Factor:

 

Retail Area Contrast: This factor contrasts two types of

 

market areas in the city. Positive variables 60, 5, 100,
and 1 describe the suburban areas serviced by shopping
centers. Average population per household, and number of
rooms per housing unit were higher. Competition was either
between shopping centers or within them. Negative variables
describe the less viable eastern and northeastern census
tracts. The variables and their respective loadings des-
cribe a more transient, lower income, high rental area.
Factor #3; Table 9C, App. B; Unique Factor:

White Racial Groups: Both negative and positive loadings

 

identify this factor. Variables 49, 82, 83, and 97 identify
the white population as precisely as variables 50, 95, 80,
98, 73, and 16 identify the negro population. As indicated
by the high negative loadings, severe social and economic
differences did exist between white and black racial

groups. Whites are more wealthy and own their own homes.
Blacks are less wealthy, typically laborers, rent instead

of own, and are less educated.

71

Factor #4; Table 9D, App. B; Unique Factor:

Inferior Sales or Market Performance: Variables 35, 39,

 

31, 13, 26, and 29 associate with poor sales performance,
whereas variables 30, 47, 46, and 42 associate with good
sales performance. That is, poor tree sales are definitely
associated with poor lot quality (29), poor tree quality
(26), no advertisement (31), and time in business (35).

Factor #5; Table 9E, App. B; Uniqpe Factor:
The Independent Retail Lot: There is definite correlation
between the high loading descriptive variables on this
factor and the independent lots located in Winston-Salem,
especially the two largest independent retail lots. The
largest independent lot was a multiple ownership (40)
using handbills as the major means of advertisement (43).
Lots were generally located 1/4 mile distance from each
other (3 - 4 blocks).

Factor #6; Table 9F, App. B; Unique Factor:

Good Tree Quality: The variable scores are more difficult

 

to interpret here, but of the top 9 variables, 8 are defi-
nitely associated with good tree quality. That is, good
trees received average to high price (10, 8). Lots selling
good trees were higher quality, advertised, located in
business areas, and the best quality trees in the city were
at nurserys or florists (26, 42, 44, 18, 14).

Factor #7; Table 9G, App. B; Unique Factor:
Major or Minor Retail Areas: Five of the top 6 variable

scores characterize this factor as a high traffic retail

72

area. Traffic count is critical (105), and lots are
close together (2). Shopping areas were in business areas
(22), and the lots had average tree quality, not poor (25,
26). Because these areas were located around the city's
fringe, many other seemingly conflicting variables fall in.
For example, retail area A was in an older part of town and
was the only retail area with little housing construction
since 1940 (102). Area B was located near a college and
consequently near the only high group quarters population
in the city (59).

Factor #8;gTable 9H, App. B; Unique Factor:
Upper Income Areas: At the outset this factor appears
to be a low load factor, however, if we support the top 4
variable loadings with the next 4 we see possible defini-
tion. That is, the first 4 (84, 103, 77, and 101) tell us
income, school years completed, and value of home are impor-
tant. The next 3 highest loadings are more definitive,
i.e., high incomes (83), better educated (76), located in
an area where lots are in close proximity (2). An appropri-
ate area designation could be higher income suburban fringe.

Factor #9; Table 91, App. B; Unique Low Load
Factor: This factor is poorly defined by the variable
scores. The first 7 variables describe several lot types
in the city. For example, civic lots advertised with signs,
and located in business areas. However, they did not have

poor lot quality.

73

Factor #10; Table 9J, App. B; Unique Factor:
Lot Types, Civic Organization: This factor describes vari-
able 15, church or civic groups. Most such groups located
in median income areas, where obviously the largest percen-
tage of kindergarten enrollment occurs (69). And, because
they obtain free radio time, these groups made wider use of
radio than any other group.

Factor #11; Table 9K, App. B; Unique Low Load
Factor: Certain factors created have no definite pattern
in the relationship of factor scores. However, variables
which have interpretive value on other factors do appear
in these low load factors, such as the income variable (85).

Factor #12; Table 9L, App. B; Unique Factor:

Lot Type, Discount Store: This factor, as will be the case

 

of the remaining factors, is defined by only 2 or 3 vari-
ables. Variables 13 and 21 definitely relate to discount
stores and their locations. Variable 69 lends partial
support, however, one discount store was located near a
census tract with high incidence of children in kinder-
garten.

Factor #13; Table 9M, App. B; Unique Factor:
Group and Apartment Housing: This factor provides a descrip-
tion of a group housing area with variables 72 and 59.
Families with high incomes correlate negatively (83), per-
sons being out of the city prior to 1955 correlate posi-

tively (79), school enrollment for children and adults is
high (68), and variables 89 and 67 also support the factor,

74

as well as multifamily housing.

Factor #14; Table 9N, App. B; Unique Low Load
Factor: This factor has one element common to the preceding
two low load factors; the appearance of variables on past
and present tree purchases and sales.

Factor #15; Table 994 App. B.; Unique Factor:

Effective Market Strategies: This factor draws together

 

several variables having high scores on other factors. From
the top 5 variables an effective Christmas tree retail lot
market strategy can be defined. For example, effective
retail marketing requires good trees, attractive lot, high
consumer traffic, etc. Note these variables are similar to
the top negative loadings on factor 4.

Factors 16-19 were all difficult to interpret.
High loading variables were nonexistent. Further, there
were no clusters of variables showing definite patterns.

As such, no interpretations are presented.

Screening the Variable Set

 

A second use of the factor matrix is to reduce the
original variable set. The variables listed in Table 10,
Appendix B, represent the reduced variable set taken from
the factor matrix. Deriving the set from the original
matrix (105 x 19), required selecting the 2 or 3 highest
ranking variables on each factor. By doing this we follow
the thesis of principal factor analysis to its conclusion.

That is, the model presumes that the 19 factors or dimensions

are independent (zero correlation) and they each in turn

75

provide a description of a different independent variance
source. The variable loadings on each factor determine the
degree of association between the variables and the factor,
i.e., highest loadings reflect highest association. There-
fore, by picking the highest loading for each factor we
have effectively screened the data for intercorrelation and
retained those variables explaining maximum variance.

One segment of the approach is to be questioned.
The successive factors created by the principal factor
method decrease in their ability to explain total variance.
For the problem under study, the first factor explained 35
percent of the total eigenvalue (variance). The additional
percent of total variance explained by factor 2 was 10 per-
cent. Factor 20 was not created because it did not have an
eigenvalue greater than 1.00. Should we then consider
factor 1's highest loadings (variables) more important than
the high loadings on factor 2 or 8?

From a purely factor theory point of view, factor
1 has greatest explanatory power, i.e., it accounts for the
major portion of total variance. However, the goal here is
to isolate the several major independent spheres of influ-
ence (Factors) and screen these to determine the variables
of greatest association in these spheres. Of the total vari-
able set, these are the most suited for regression analysis.

Twenty-two of the original 24 general variables
are represented in the 49 specific variables taken from the

final factor matrix (Table 10, Appendix B). Nearly all

76

of the specific economic variables had loadings above .50
on the principal factors. Only a small segment of the

demographic variables had loadings above .50.

General variables deleted.--Two general variables

 

deleted in the initial screening were knowledge of busi-
ness (specific variable; number of years retailed Christ-
mas trees) and population (specific variable; total popula-
tion). The variable "number of years sold Christmas trees"
(47) had a factor score of .45 on the 19th factor created.
Its partial r2 with tree sales (Y's) was low (—.150).

Only 3 independent variables had a partial r2 with X47 that
exceeded .50. These were X12, X42, and X21, the two former
of which also appeared on factor 19 but with higher scores
than X47. Variable X12 was "type of establishment, chain
grocery store." These stores advertised mainly in news-
papers (X42) and they located at shopping centers (X21),
the two other variables correlating with X47. Therefore,
X47 could be excluded in favor of the more explanatory
variable X12.

Total population (X48) provided little explana-
tory power. Its r2 with tree sales (Y's) was .002. Its
loading in the factor matrix occurred on factor 1, at the
.96 level. However, three household variables loaded
higher on factor 1 than did X48. The household variables
are preferred due to direct association between housing

unit and place of display for the tree. Also, high loadings

of white and negro population variables on other factors

77

indicate that total population must be partitioned to be

meaningful.

Specific variables deleted.--Although nearly all
of the general variables survived the screening process,
many specific variables were deleted. A closer look at a
cross-section of the deleted variables supports the vali-
dity of the screening process.

The general economic variable "competition" was
represented by six specific variables of which only three
were retained. These three, competition between one and
two blocks, two and three blocks, and four blocks to a mile
are the location patterns identifiable in the city
(Figure 5).

The general economic variable ”retail price" is
represented by average price, explaining why the major vari-
ance is associated with "average" and "average low price."

All three "tree quality" variables were retained,
however, "good tree quality" accounted for greater variance
than did "average" or ”poor" tree quality. The same is
true for the advertising variable. The variable "adver-
tised yes" accounted for greater variance than did "adver—
tising no." All three variables "good tree quality,”

"good lot quality,’ and advertised yes" were associated
with larger, more progressive lots.
Similar relationships are also observed for the

demographic variables, which were screened much more

78

heavily than the economic variables.

For example, three specific variables were used
to represent effect of race in the study; white popula-
tion, negro population, and total foreign stock. Only
negro population was retained. Racial groups other than
caucasians and negro represented less than 1 percent of the
total population, and were grouped into total foreign
stock. Negro population was not only large with respect
to total population (33 percent of total), but was concen-
trated in the eastern section of the city. Predominantly
negro tracts had lower incomes, greater number of persons
per household, and fewer number of rooms per housing unit.
As such, more variance is explained by this variable than
the foreign stock variable.

Specific household variables were considered
important, and at the outset it appeared the model selected
poor variables; "population in group quarters" and "popu-
lation per household." However, Winston-Salem college
dominates a census tract in a prime retail area and
depresses retail sales in the census tract significantly.
It was felt that total number of households would be a more

realistic variable in the model, but population per
household" was retained. It provides a direct measure of
the variance associated with the number of children in a
household. It is felt that different purchase patterns are

evident for young couples, young families, and older couples

with no family at home.

79

The Predicting Model

 

In pulling together the previous sections of this
paper, it is obvious that the natural tree buyer is pre-
dictable in the market place. His purchase patterns are
associated with market characteristics exhibited by the
retail lot and also his own demographic characteristics and
preferences. This section will be devoted to character-

izing these associations in the form of predicting equations.

The Variable Set

The slightly altered variable set listed in
Table 11, Appendix B represents the 41 final variables
used in the stepwise multiple regression model. Variables
1-24 represent the economic variable set, 25-37 the
demographic variable set, and variables 38-41 represent
information on consumer preferences and behavior. These
latter variables were not included in the principal factor
model. However, due to the market held by artificial
trees and definite buying behavior and preference asso-
ciated with natural and artificial trees, it was necessary
to include them in the regression model.

Most of the adjustments in variables between
Tables 10 and 11 occurred in the economic variable set.
Although some variables were deleted, most of the changes
were due to combining variables and/or representing the
original variables with "dummy" variables.

The model was easily adaptive to representing

80

some variables as dummy dichotomies or multichotomies.

The new dummy variables are binary coded as zero, or one
(Suits 1957), (Johnson 1963), (Tomeck 1963). Values

for these variables are reflected in the computed inter-
cept for the regression line, i.e., the direct effect of
dummy variables is to shift the intercept value up or down.
The mean value of the intercept takes on the value B5 = BO
+ Bi-n when it is made to represent the constant term

(Bo) plus effects of the dummy variables (Bi-n)'

The competition variables for 1-3 blocks were
combined to form one variable "competition, 0-4 blocks.”
The area location variable ”Business and shopping center"
was dropped because of its high correlation with "Business
area." The variable "Residential area and vacant lot”
was combined with residential area.

A different set of variables were used to repre-
sent trees for sale and trees sold, because of the limited
interpretive power of the original variables in Table 10,
Appendix B.

Variables depicting multiownership of lots were
deleted from the sets because of their high correlation with
the variable "Enterprise types, chain grocery store."
Seven of the nine multiple ownerships were accounted for
by chain grocery stores.

Because of the weak data supporting some of the

variables in "Type of advertisement, these variables were

dropped and only one pseudo variable on advertisement was

81

retained.

Two demographic variables were removed and four
were added. "Median family and unrelated individual income"
was removed due to high correlation with "Median family
income." The variable "Divorced persons over 14" was
removed because of high correlation with the variables
"Laborers" and "Negro population."

Total population was entered because of its tradi-
tional use in demand models. The variable "Children under
18 years of age of head of household" was added because of
its selection in many studies as a determinant of natural
tree salesll. "Median number of persons per housing unit"
was added as a proxy variable for the sales variance
caused by group quarters and multiple dwelling housing
units. ”Number of households” was added to test the impor-

tance of the single dwelling housing unit.

Specification of the Model
The model purports that retail lot tree sales
are a function of market characteristics of the retail lot,
plus demographic characteristics of consumers within the

census tract where the lot is located. This hypothesis can

 

l/E11efson, P.V., and T.H. Pendleton. 1969. Problem analy-
sis of Christmas tree marketing in the eastern U.S.
Unpublished report. U.S. Dept. Agric. For. Serv., For.
Prod. Mktg. Lab., Princeton, W.VA.

82

also be stated as:

Y(cr) = Bo + Br._ - XrI_n + BrJ-n. er-n +

Bck ° Xck_ + u
Where:

Y(Cr) = Number of natural trees sold (Y) by
retailer (r) in census tract (c)

Bo = Intercept or base value

Xri_n = Dummy economic variables, i.e., merchan-
dising, advertising, etc., for retail
lot (r). Variables running i-n.

XrJ._n = Continuous economic variables associated
with retail lot (r) in census tract (c).
Variables running j-n.

Xck__n = Continuous demographic variables asso-
ciated with persons in census tract (c).
Variables running k-n.

u = Random component with expectation zero and

constant variance from observation to
observation.
The model, although developed to a workable
stage, is improperly specified. The specification error
involves improper characterization of the demographic data

M H

since customers from census tracts other than c are not
considered.
To correct the specification error would involve

tabulating all sales by census tract origin, and weighing

83

the demographic data from each census tract by the rela-
tive percentage of total sales attributed to them. An
attempt was made to do this using data from the consumer
survey, but it proved futile.

Previous discussions have illustrated that
buyers of natural Christmas trees in Winston-Salem pur-
chased trees out of census tracts in which they resided.
In fact, many purchases were made three miles distant from
place of residence, or three census tracts away. Conse-
quently, unexplained variance did occur in the predicting

models as a result of the specification error.

Observation Error

 

Another error involves the use of 1960 census
data to represent demographic characteristics of consumers
residing in Winston-Salem for the study period 1967 - 1969.
In effect, it could be said a different population occurred
in the city for each of the three study years. However,
at the time the study was conducted, only 1960 census
data were available. As previously discussed, any change
in demographic characteristics between 1960 and 1967 and
the successive study years introduces errors of observation
in the data. The extent of these errors were not evalu—

ated in the study.

Empirical Models

 

Three empirical models were developed from the

Winston-Salem data. To avoid confusion, the models will be

84

referred to as models I, II, and III, corresponding
respectively to the years 1967, 1968, and 1969.

In developing the three models, many variables
were dropped and new ones created to make the equations
more efficient. However, most variables which appeared
in one model also appeared in the remaining two models
(Table 12, Appendix B). The final list of variables re-
tained for use in all three Winston-Salem models is pre-
sented in Table 13, Appendix B. Initial computer runs
clearly identified 13 variables which would not enter the
models at the specified "F" levels. Referring to Table 11,

Appendix B, these variables are: X

X16’ X18’ X19-21’ 23'

X and X

Xza' 30' 38-41'

Two of the variables deleted, X16 (tree quality)
and X18 (advertised), were felt to be important to the
model. However, as previously mentioned, both variables
could have been represented by better data sets. Variable
X16 should have been a scalar variable, and X18 might have
been represented better by "dollars expended for advere
tising."

Variables X38-40 represent the percentage of
Winston-Salem consumers who purchased either a natural
tree, artificial tree, or no tree. These data were well
developed, but failed to have any association with the
models. This is surprising indeed when one considers that
over 35 percent of the consumers did not purchase a tree

and 37 percent used artificial trees.

85

Variables X19_21 and X23_24 were found to be
closely correlated with X22 (Y variable) and, consequently,
were not included in any models.

Interaction variables X10_12 represented effects
of location on independent lots and chain stores. The
effect of an independent lot in a business area or shopping
center is represented by X10. The effect of an inde-
pendent lot in a residential area is -X10. Effect of busi-
ness location on chain stores was X11, and shopping center
location X12. Effect of residential location was computed
from -X11 -X12.

Effects of interaction variable "Merchandising,"
interacting with three retail outlet types; Independent lot,
Discount store, and Nursery, were determined as follows:

Independent Lot Discount Store Nursery

 

 

Merchandised X13 X14 -X13 -X14
Did not
Merchandise -X13 -X14 X13 +X14

Likewise, "Effect of merchandising interacting with location"
was determined similarly:

Business Shopping Center Residential

 

 

Merchandised X X -X -X

15 16 15 16

Did not

Merchandise -X -X X +X

15 16 15 16

86

Evaluation of the
Winston-Salem Predicting Models

 

 

As indicated from the descriptive data in Table
14, all models were statistically significant at the 5
percent level. However, the models did differ as regards
R2, standard error, and computed "F" values.

A11 3 Winston-Salem models (Tables 15-17, Appen-
dix B) were developed on a fairly consistent sample size;
53, 46, 53, for models I, II, and III, respectively. Also,
the 34 variables used in deve10ping the 3 models were iden-
tical (Table 12, Appendix B). Yet, the statistics describ-
ing the 3 models and the variables entering each model
differed.

Models I and II were most effective in represent-
ing the data from which they were created. The coefficients
of determination for the 2 models were 0.80 and 0.79, respec-
tively. Model III, although significant at the .05 level,
offered explanation of only 51 percent of the variance asso-
ciated with tree sales (Y). Further, models I and II had

lower standard errors and higher "F" ratios than model III.

Table 14.--Comparative statistics on models I, II, and III.

 

 

Model N K R Std. Error F F.05, k, n-k-l
I 53 20 0.8046 111.9521 6.586 1.91
II 46 18 0.7876 175.3309 5.563 2.01

III 53 15 0.5148 239.3306 2.617 1.95

 

87

The foregoing statistical inference would inci-
cate that the 1967 equation has effectively modeled that
year's market relationships. It does in fact provide good
estimates of 1967 retailers' market performance in Winston-
Salem. However, that was not our basic objective. Our
objective was to develop a predicting model which would be
useful in estimating sales from market information differ-
ent from that which it was created.

Our empirical selection of variables, and success-
ful model development from a given set of market informa-
tion does not necessarily imply that the model will be
successful in estimating sales, given different market infor-
mation. In fact, another set of data might well reverse
the order of importance of the several variables.

To provide a more explicit measure of effective-
ness for the three models, they were used in predicting
retail Christmas tree sales for retail lot owners located

in Denver, Colorado, during the 1965 Christmas season.

The Denver Market

 

A 1965 survey of 186 natural tree retailers
located in Denver, Colorado, yielded extensive economic
sales data on the area's natural Christmas Tree market
(Troxell 1970). These data were used to develop vari-
ables paralleling those used in the Winston-Salem models.
Further, the data were used to test the predicting effi—

ciency of models I, II, and III, and to create an independent

88

estimating equation for Denver.

Denver and Winston-Salem have had growth patterns
similar to most metro areas. However, the two cities did
differ in many respects. Following are examples of apparent
differences in Winston-Salem and Denver demographic data:

'Denver had 185,000 households in 1960 as opposed
to 32,213 in Winston-Salem.

'Total 1960 population for Denver was 458,117 as
compared to 111,135 for Winston-Salem.

'Denver non-white racial groups accounted for only
9 percent of total population as compared to 33
percent for Winston-Salem.

'Although income distribution was similar for the
two cities, the mean income for Denver was
greater than for Winston-Salem.

Other demographic variables used in the models also differed,
characterizing the two retail areas as having quite differ-
ent demographic makeup.

Differences and similarities were noted for the
product retailing methods and retail outlets. Approximately
115,000 trees were sold in Denver, compared to 10,000 in
Winston-Salem. An estimated 92 percent of the Denver house-
holds had natural trees. Also, an estimated 61 percent
of the trees marketed in Denver were plantation grown, as
opposed to a much smaller percentage for Winston-Salem.

Species composition was largely Douglas fir (38 percent)

and Scotch pine (35 percent) in the Denver market, whereas

89

only a few Douglas fir were sold in Winston-Salem.

Retail outlets were similar for the two cities,
with over half the outlets being either chain grocery
stores or independent lots. Also, these outlets accounted
for over 60 percent of total sales in each city. Average
tree price of $4.77 for Denver was higher than Winston-
Salem (1967 - $3.21).

Consumer purchase patterns were quite similar for
the two cities, even though purchases per capita were
higher for Denver. Denver consumers traveled up to three
miles to purchase natural trees, and shopped more than one

lot before selecting their tree.

The Denver Model

 

A predicting model was developed from the Denver
data, conforming to the 33 independent variables used in
creating the three Winston-Salem models.

The Denver model proved to be significant at
the .05 level with an R2 = .4679 and a large standard
error (Table:18,Appendix B). However, the "F" test on the
Denver model is computed on a sample size (n = 152), three
times that used to create model I (n = 53).

The Denver model did incorporate many variables
found in two or more of the Winston-Salem models, however,
one variable (X23) did not occur in any of the three pre-
vious models. In addition, variables X13, X15-l6’ X23_24,

and X29 occurred in only one of the Winston-Salem models

(Table 12, Appendix B).

90

Evaluating the
Predicting Equations

 

 

The foregoing predicting equations have close
association to the data from which they were created; yet,
they may have little utility when market conditions change.
Although we have attempted to account for all variables
affecting retail sales in Winston-Salem and Denver, the
models presented are only abstractions. Further, they are
static, characterizing market conditions for only those
years for which data were collected.

The success of a predicting equation in estimating
sales for different markets relates primarily to differences
in market parameters. If market factors and their inter-
relationships remain precisely the same from one market
to the next, a predicting equation will be highly successful.
A similar conclusion is reached when the same market is con-

sidered, but at two different points in time.

Variables Represented

 

Thirty-three independent variables were used in
developing the three Winston-Salem models. Yet, the models
are not structured around identical variable sets. Al-
though the models share a similar number of variables; K=20,
18, and 15 for models I, II, and III, respectively, the
variables included in each differ slightly. Only 11 vari-
ables appear in all three Winston-Salem models. These repre-
sent type of retail establishment, area location, merchan-

dising, and lot competition (Table 12, Appendix B). Seven

91

variables appeared in only two models and are represented
by location, competition, tree price, and population.
Eleven variables appeared in only one model, either Model
I or II. These included household demographics, income,
education, and retail traffic count.

Ten of the 11 variables that appear in all
Winston—Salem models also appear in the Denver model
(Table 18, Appendix B). However, the Denver model also
included variables X15, l6, and 23, which did not occur
in any other model. Their inclusion results from differ-
ing marketing conditions and population demographics in
Denver. For example, merchandising of trees in business
areas was more prominent in Denver. In addition, families
were younger in Denver, the population per household was
larger (X33) and kindergarten enrollees (X29) were higher.
Also, these young families had purchasing power equivalent
to older families in Winston-Salem.

It is of little meaning to cite 10 variables as
being included in all models without characterizing the
level of association these variables held to the dependent
variable tree sales.

The 10 variables accounted for 33, 36, 27, and
27 percent of the variance in Models I, II, and III, and
the Denver model, Respectively. Other variables which
proved important to the models were prices of trees sold

(X20), competition 0-4 blocks (X17), and merchandising (X9).

92

Seven variables did not enter the Winston-Salem
models; X15, 16, 21, 23’ 26, 27, and 33. Merchandising in
business areas (X15) was practically nonexistent in
Winston-Salem. Negro population was concentrated in census
tracts in the eastern sector of Winston-Salem. These tracts
recorded the minimum sales incidence for the city, afford—
ing little explanation of variance in the models. Popula-
tion per household (X23) was better represented by persons
per housing unit (X27). Also, the general demographic
population (X33) was better represented by X22, 27, 28, and
30, and variables X26 and X29 are best represented by vari-
able X27.

The variable coefficients revealed a similar low
level of continuity across the models. The zero inter-
cept, or B0 for each model did differ, but the regression
coefficients did produce certain similarities. For example,
of 11 variables occurring in each of the 3 models, only 3
entered with a negative value.

A much more in depth comparative analysis could
be made of each equation; but it would not characterize
the predictability of the models. The true test of the
equations is their effectiveness in predicting sales for

differing market conditions.

Analysis of Residuals

 

Our hypothesis states that predicting equations

can be used to predict a retailer's sales in any given

93

market. Dr, stated differently, within a given level of
confidence there should be no statistical difference
between the actual sales and predicted sales. To test the
effectiveness of the predicting equations, each model was
used to predict sales for retail lots different than those
used to create the model. An analysis of residuals was
used to evaluate each equation.

The stated null hypothesis is; for a given retail
sales predicting model, Y = BC + B1 X1 + B2 X2 + u, its
§i's for a new market condition will equal the Yi's for
that market. Specifically stated, using the 1967 Winston-

Salem model (I) on the Denver market data;

HN:2(Y(I)DS - YDS) = 0 Where Y(I)DS = estimates
of Denver retail lot sales
with model I and, YDS =

Actual retail lot sales
in Denver.

A similar set of hypothesis are developed for each
model on each remaining set of market data. The level for
rejecting the null hypothesis (Type I error) is .05.

If the models are assumed to be correct, the unex-
plained variances could be envisioned as the observed errors
in the model. However, as previously noted, observation and
specification errors exist in the models, restricting the

identification of residuals to unexplained variances.

The question of model error prompted a visual, as

well as statistical examination of the residuals. The

94

visual analysis of residual plots against estimated Yi's
for each model did reveal model error. The statistical
analysis of residuals caused the null hypothesis for the
1969 Winston—Salem model (III) to be rejected at both the
99 and 95 percent confidence level (Table 19).

The 1969 estimating equation for Winston-Salem had
a low R2 and a high standard error. A computed "F" of 0.58
and 0.75 resulted from retail sales estimates of 1967 and
1968 sales data, respectively. However, estimates of
Denver market sales yielded a larger computed "F," which
resulted in the hypothesis not being rejected.

An overall comparison of computed "F's" for all
models in all markets identifies model I as the most effec-
tive estimator (Table 19). Yet, statistical indicators,
when taken alone, permit little inference. A direct compari-
son of computed "F" values assumes the residuals of each
‘model are, in fact, its true errors. Yet, each of the
models could have both observation and specification errors,
and the level of these errors could differ. An examina-
tion of residual plots provided greater insight into these

problems.

Residual Plots

 

Although several plots of the residuals were

accomplished, only plots of (Y1 - Yi)2

/ Y1 were selected
for discussion. Figures 8-19, Appendix B, present the

observed patterns of residual plots for each model on 3

different sets of market data.

95

Table 19.--”F" tests of residuals from 4 predicting models
tested on 4 different sets of market data.

 

 

1967 1968 1969 1965
Model Winston- Winston- Winston- Denver

Salem Salem Salem Data

Data Data Data

C/

II ++2.2753’27 ++5'5646,27 +1.8353’27 +1“12.35152,27
III 0.5853,37 0'7546,37 ‘r+2.6l753,37 +9.65152,37
Denver ++2'2253,l30++1°7146,130 ++3'2453,130 ++5.44152’130

 

++Significant at .05 and .01 level.

i Significant at .05 level.

S/Degrees of freedom.

The plots do support the statistical analysis of
the residuals. The pattern of residuals for model III fully
characterize a large variance, which is not constant. In
fact, the assumption of constant variance appears to be
violated at both high and low values of Y. Further, it is
more severe in model III than any other model.

A fact not indicated by the "computed F's" is
that estimates of Denver sales yield a more constant vari-
ance over lot size than did the other models. In fact,
residual variance created from each model‘s estimates of
Denver retail sales have the greatest uniformity of all

residuals analyzed. Conversely, residuals created from use

96

of the Denver model to predict sales on 1967-69 Winston-
Salem retail data reveal greater variances as larger retail
lots are encountered.

Sales estimates with model I exhibit the minimal
variation in residuals, although there is evidence of a
need for weighted least squares or variable transformation.
Yet, the probabilities of improvement are minimal compared
to models II and III.

The Denver model reveals the greatest potential
for improvement. It is a very efficient estimator on lot
sizes up to 1000 trees. At that point, the variance in-
creases radically, making the model completely ineffectual
for sales estimation on large lots.

Model II has similar plots to model III, revealing
considerable change in residual variance from small to large

lots. Its range of effectiveness is limited.

Accuracy of Estimators

 

0f the four models tested, only model III failed
to produce statistically acceptable estimates of Christmas
tree retail sales. Its estimates of 1967 and 1968 Winston-
Salem retail sales resulted in more variance than was accept-
able at either .01 or .05 level of confidence. In addition
to the variance being too large, it was also heterscedastic.
Further, this unequal or heteroscedastic variance occurred

to a limited extent in all models.

 

0.

1

97

A question raised by the foregoing discussion is,
”What practical limitations does the unequal variance impose
on the models?" The answer could be resolved by establish-
ing confidence intervals such that the retailer could be 90
percent confident that the estimated sales (Y1) are within
15 percent of actual sales (Yi)' As approximately 15 percent
of the trees at retail are left unsold, the models should
have this level of accuracy to stand alone as a management
tool.

An evaluation of each model would demonstrate that
none has the level of accuracy to be effective as a complete
decision tool for retailing Christmas trees. Model I, the
most accurate estimator overall, sustained an average error
of over 50 percent in its estimates of 1967 Winston-Salem
sales for only 7 of the 53 retailers. Therefore, the models
do have practical limitations. Further, although we can
accept our original hypothesis of 2(Yi-Yi)=0, it should not
be concluded that the models represent a complete informa-

tion and decision system.

CONCLUSIONS AND RECOMMENDATIONS

Summary

Natural Christmas tree retailers in Winston-Salem
sold 9,227, 10,152, and 11,941 trees in 1967, 1968, and
1969, respectively. The species mix was dominated each
year by Balsam fir, accounting for 40 percent of all trees
sold. Taste preferences for local plantation-grown trees
caused a rapid decline in the sales of wild Eastern Red
cedars, which dropped 60 percent in total sales over the
three years. This void was filled by local plantation-grown
firs, Scotch pine, and White pine.

Prices for trees sold varied significantly, with
Fraser fir bringing the highest prices and Eastern Red cedar
the lowest.

The number of retail lot outlets varied from 53 in
1967, to 46 in 1968, and 53 in 1969. Wholesale to retail
mark-ups on Christmas trees declined over the 3-year period
from 118 percent in 1967 to 91 percent in 1969. Trees left
unsold were 14.7, 9.7, and 17.5 percent for 1967 through
1969, respectively.

Retail outlet types were dominated by chain gro-
cery stores, which represented over one-third of the total

lots each year. The second most predominant lot type was

98

99

independents. Together, the two types accounted for over 50
percent of all lot types. Tree sales volume was concen-
trated in a relatively small number of lots also. In each
of the 3 years of study, 20 percent of the retailer popula-
tion held 60 percent of market sales, and even more signi-

ficant, the 5 largest retailers sold 40 percent of all trees.

Market Associations

 

Several retail market relationships were estab-
lished from evaluations of Winston-Salem tree market data.

The best sales performance for Christmas tree
retail lots occurred in areas of high retail activity for
other consumer durable and nondurable goods. Retail lots
located at shopping centers in four different areas of the
city demonstrated higher retail sales activity relative to
other retail lots. Retail lots located at these major re-
tail market areas outsold their competitors in residential
and downtown business areas three to one.

Factors which were demonstrative in increasing
tree sales were also associated with shopping centers.
Shopping centers were located on major traffic arteries,
gaining them direct access to population moving to and from
the city area. Also, shopping centers had larger and better
parking facilities and provided extensive advertising, such
as flyers, handbills, and radio ads. By associating them-
selves with these areas of high retail activity and prac-

ticing marketing strategies that have been demonstrated

100

effective in other retail goods, Christmas tree retail lot
(awners enjoyed greater tree sales, higher prices for their
trees, and fewer trees left unsold at the end of the Christ-
mas marketing season.

Distances traveled is associated with income, and
most consumers are willing to travel over two miles to pur-
chase a natural Christmas tree. In fact, many consumers
in Winston-Salem traveled over three miles to select and
purchase a natural Christmas tree. Bkhsed on these findings,
the research would strongly suggest that retail owners
locating in suburban areas not adjacent to roads bearing
high traffic counts, and using little advertising, will

severely limit their sales of natural Christmas trees.

Analysis and Inference

 

Many variables were revealed to have possible
association with performance of natural Christmas tree re-
tail outlets in Winston-Salem. Principal factor analysis
was utilized in reducing a large number of variables which
were felt to have some association with natural Christmas
tree sales. The method proved useful in reducing the origi-
nal number of variables to a more suitable subset for use
in regression analysis. In effect, the model screened the
data for intercorrelation among variables and identified
independent sources of variance from several variables. By
clustering these variables in the form of factors, and sub-

jectively identifying these associations, definite spheres

101

of variance were identified. Among those defined were;
xflhite racial groups, inferior market performance, good tree

or product quality, and major retail areas.

Predicting Models

A subset of variables defined through factor analy-
sis was used as input to stepwise regression analysis. Four
predicting models were created, one from each of the three
years of data from Winston-Salem, and an additional model
from 1965 tree sales data on Denver, Colorado. The models
were created to predict expected sales for natural Christ-
mas tree retail lot owners in metro areas.

Evaluation of regression coefficients and standard
errors for the four models revealed model I, developed from
1967 Winston-Salem data, best described the market data
from which it was created. The equation explained 89 per-
cent of the variance associated with the 1967 data.

Each model was then used to estimate sales from
market data different from.which it was created. A statis-
tical analysis of the residuals indicated the 1967 Winston-
Salem model to be the most effective estimator, and the 1969
Winston-Salem model the least effective. However, analysis
of residual plots revealed all models were limited due to
unequal variance. Further, the Denver model proved to have
the greatest accuracy up to lots handling 1000 trees. The
1967 Winston-Salem model had greatest accuracy on lots

handling greater than 1000 trees.

102

Several conclusions can be drawn from this
research. Based on these conclusions, recommendations can
be made which will not only provide greater meaning to
these investigations, but should also be valuable to

future research.

Conclusions

 

Competition did exist in the Winston-Salem retail
market for natural Christmas trees. Retailer market entry
and exit was not impeded either by natural or artificial
constraints as witnessed by high retailer turnover during
the three study years. A fairly homogeneous product was
marketed, with little effort expended by retailers to dif-
ferentiate the product. Although there was obvious concen-
tration of seller activity, there were no explicit controls
imposed on market activity by this concentration. Perfect
competition, however, could hardly be expected with the
limited information system available to retailers.

Consumer mobility in the purchase of cut natural
Christmas trees was high. Low income buyers residing in
areas where single lot locations averaged less than a mile
apart, traveled an average of 2.5 miles to purchase their
trees. Median income families traveled the greatest dis-
tance (2.8 miles), while residing in areas where concen-
trations of retailers averaged 1.2 miles from the consumer.
In all sectors of the market, consumers traveled an average

distance exceeding the average distance to the nearest

103

retail lot or lots.

Location of the greatest retail sales activity
occurred in areas of concentrated retail sales for other
products. About half of the cut natural tree retailers
located there and held approximately 70 percent of total
market sales. Shopping centers grew out of the concept of
concentrating buyer traffic and, therefore, high buyer
activity. Market products and services offered at these
centers are as important to the sales of natural trees as
they are to other products. They afford convenience to the
consumer, with large parking areas, sufficient area for
product display, and access to other stores for normal
shopping activity. Also, market activities such as adver-
tising and merchandising are greatly facilitated at these
locations. As such, these centers represent prime locations
for retailing Christmas trees in metropolitan areas.

Although many retailers appeared astute in locat-
ing at or adjacent to shopping centers, they did not effec-
tively utilize marketing strategies to better their inde-
pendent position relative to others, or to afford overall
expansion of natural tree sales in Winston-Salem. Price
competition was implicit in the market (a range of prices
existed), but retailers did not use the mechanism (promo-
tion of price) to lure consumers from other lots. This
was even true in the final weeks of the market period when
retailers should have been making deliberate moves to mini-

mize losses.

104

Retailers used minimal advertising and they did
not grade their trees. This was true in 1969 after a deci-
sive consumer taste change had occurred in favor of higher
quality trees. Retailers reacted slow to the taste change,
and, in fact, made a grave error in 1969 in oversupplying
the market with low quality trees. Also, they mistakenly
increased price. As a result, 23, 18, and 22 percent spoil-
age levels were recorded in 1967, 1968, and 1969 for low
quality trees. During the same periods, high quality plan-
tation fir and pine were enjoying good sales growth, and at
significantly higher prices. But instead of concentrating
on quality plantation trees in 1969, the market was supplied
with an increased supply of wild balsam fir and pine from.
Canada. Even though the pines entered the market at a price
below the 1967 average for pines, sales stagnated at the
1968 level.

Factor analysis was effective in deliniating
independent spheres of influence from certain variables
used in the analysis. However, the underlying concepts of
factor and regression analysis, as is true of other analy-
tical tools, are inherently dependent upon good empirical
data. Creation of the predicting equations was hampered
by lack of current and reliable census data and inability
to correctly quantify certain variables. Yet, the author
feels the models did effectively characterize the data
from which they were created and the market relationships

observed.

105

The use of predicting equations as developed
herein will have moderate effectiveness within the market
where they are developed, but only over the short—run. The
models are structured around marketing factors that are
exceedingly dynamic. They are ever-changing, and often in
differing proportions. Further, variance was not concen-
trated in one variable, but many, making it difficult to
monitor the changes.

The estimating equations, although developed to
a reasonable stage, are ineffectual as estimators at the
retail lot level. The demonstrate statistical significance
and are important in identifying market associations. Yet,
their accuracy at the retail level exceeds the allowable
tolerance for a management tool. Many variables in the
model have large variances. In combination they can produce
a significant departure from the true lot situation; or the
model could be representing the true capability of a lot
which is not being realized due to poor management by the
retailer. Speculation as to the cause of the unexplained
variance does not remove it. What is critical is that it

is too large for the models to be acceptable.

Recommendations

 

The most critical recommendation regarding this
work is that the reader fully understands its limitation.
Briefly, caution should be used in utilizing the models for

predictive purposes at the retail lot level.

106

This research has demonstrated that a myriad of
marketing and economic factors affect the sales performance
of a given retail lot. Although some are more important
than others, several must be included to have an acceptable
estimator. From one lot to another, within the same market,
these will change in level of importance to the response
variable (Y). More importantly, changes in market condi- 1
tions with time and/or geography will likely cause even {

greater variability.

 

This sensitivity at the retail lot level limits
the effectiveness of the models. The potential for predic-
ting error is too great to depend entirely on the model
for decision-making. Instead, the model could and should
be used for market analysis. If the retailer studies a
cross-section of potential lot locations, he can rank the
areas by estimated sales performance. Patterns similar to
those in Denver and Winston-Salem should become apparent.
These data could then be used for selecting the best lot
location.

Future research should concentrate on defining
inter-relationships among variables studied. This research
was directed at obtaining effective estimators, with limited
attention to assessment of individual variables. Extensive
descriptive-type studies have been accomplished, focusing
attention on many factors as being important to retail sales.
More definitive work is necessary to determine the exact

impact these variables do have on sales and each other.

107

Once analytical relationships have been estab-
lished for important variables, attempts should be made at
developing regional models. Systematic aggregation of
retail outlets to larger, more homogenous groups, either
via a priori or empirical evidence would be a natural exten-
sion of the work presented herein. It should incorporate

time as a variable to get at taste changes and competition

from other products.

ITW‘Z—cfiw

APPENDICES

I'.-.- 9"‘li‘3g tn 9
A t
A

 

APPENDIX A

ml;

pd

SCHEDULE I
CHRISTMAS TREE RETAIL MARKETING QUESTIONNAIRE

To be read or explained to retailer bLinterviewer

The questions that are about to be asked of you
are intended to provide us with the necessary information
to develop a method of forecasting retail sales of natural
Christmas trees. You, the Christmas tree retailer, will
be the chief beneficiary of the results of this study and

they will be made available to you as soon as possible. The

information that you give us will be confidential. No one

else will have access to this quationnaire or to the infor-

ma tion contained therein.

insxtructions to the interviewer

This questionnaire or data sheet is divided into
tWO sections. The first consists of those questions that
are to be asked of, and answered by the Christmas tree
retailer. The second section consists of certain character—
is tics of each retailer to be observed and recorded. This

laSt section does not include any questions to be asked.

108

 

109

Bureau of Budget No. 403 69007
Expiration Date: March 31, 1969

CHRISTMAS TREE RETAIL MARKETING QUESTIONNAIRE

 

Name of Lot Owner or Firm

 

Address of Owner or Firm

Pflnone (Office) (Home)

 

 

Lot Location

 

Identification Number

 

IX. Retail Lot Characteristics (Visual Observations)

 

1.

Number of other Christmas tree retailers within
designated distances of this retailer. Identify by
their identification number.

Distance Number Identification

 

1 block & under

 

Over 1 block - 2 blocks

 

 

 

Over 2 blocks - 3 blocks
Over 3 blocks - 4 blocks
Over 4 blocks — 1 mile

 

Over 1 mile — 2 miles

 

Type of retail establishment.
Independent lot

Chain grocery store

Discount store

Nursery or florist

Church or civic group

I\ l] l] I] I] I]

Independent grocery store

110

JD

Service station

Other (specify) /

 

Type of area in which retail lot is located.
check more than one.)

Business area

Residential area

Shopping center

Vacant lot

Is lot located on a principal traffic artery?
Yes

No

Parking facilities adjacent to lot.

Large parking area (>50 cars)

Small parking area (<50 cars)

Street parking only

HUI] DI] MEI] 3

None available

 

Quality of trees being sold at lot (ACTGA Grades).
Good

Average

Poor

Nature of tree display at lot.

Good (Standing upright and well-lighted)

Average (Leaning, few on stands, few lights)

I] l] I] I] l] I]

Poor (On ground, little or no lighting)

 

B.

111

Retail Lot Sales Data

1.

How many trees of each of the species listed below
did you have available to sell in the 1968 Christmas
season? How many did you sell? (Do not include

artificial trees.)

# of Trees
Available to Sell # of Trees Sold
Species Total 0-3' 3'-5’ Over Tetal 0-3' 3'-5' Over

Pine
White

Scotch

 

 

Red

 

 

Virginia

 

Total

 

 

Fir

Balsam

 

 

Fraser

 

 

 

 

Douglas

White

 

 

Total

 

 

Spruce

Norway

 

 

White

 

Black

 

 

Total

 

 

Eastern Red Cedar

 

 

 

Arizona Cypress

 

Other (specify)

 

 

 

TOTAL

 

 

2.

112

What were the high, low, and median prices that you

received for each of the species sold? (Do not

 

include prices of artificial trees.)

Species

Pine
White
Scotch
Red
Virginia
Fir
Balsam
Fraser
Douglas
White
Spruce
Norway
White
Black
Eastern Red Cedar
Arizona Cypress

Other (specify)

 

 

 

High Price Low Price Median Price
dollars dollars dollars

113

1 (-

What was the average price you paid ror each of
these species, and by what arrangements were the

purchases made?

Purchase Arrangements

Picked On
Up At From Con-
Deliv- Distrib- Cut 8 On Other sign-
ered utor Piled Stump Retailer ment
Species -average'pFice per tree-

 

Pine

White

Scotch

 

 

Red

 

 

 

Virginia

 

 

 

Fir

Balsam

 

 

 

Fraser

 

 

Douglas

 

 

White

 

 

Spruce

Norway

 

White

 

 

Black

 

 

Eastern Red Cedar

 

 

Arizona Cypress

 

 

Other (specify)

 

 

 

0‘0

114

Did you advertise your trees this year?

Yes

No

1‘]

 

If answer to Question 4 is yes, where did you adver-

tise and how frequently?

Other

 

 

Newspapers (number of issues) L:::7
Magazines (number of issues) L___

Handbills or cards (number of bills) [:::7’
Radio (number of spots) L:::7
Television (number of spots) L:::7
Display signs (number) L:::7

/
have you been selling trees? ____ years

How many years
How many trees
for Christmas,

How many trees

1967?

did you purchase for sale at this lot

1967? number purchased

 

did you sell at this lot for Christmas,

number sold

 

Do you operate
than this one?
Yes
No

If "Yes,"

any Christmas tree retail lots other

T a

where it is located?

 

SCHEDULE I I

CHRISTMAS TREE CONSUMER QUESTIONNAIRE

Irrterview BOB No.: 40-869102
Location Code:

 

 

/
C} T. - Int. #' Approval Expires: 2/28/70

Hello, I represent the Margaret Ernstes Interview-
.irrg,Service. I am conducting a survey for the Forest Ser-
XIimze of the U.S.D.A. Could you please assist us in com-
I>143ting the following questionnaire on natural and artificial
Christmas trees ?

1. Do (did) you have a Christmas tree this year?
Yes 1:::7’ No £:::7
2. How many trees do (did) you have? (COMPLETE A

SCHEDULE FOR EACH TREE.)

 

3. Is (was) your tree natural or artificial? (IF
NATURAL TREE, GO TO QUESTION 5.)
Natural / 7 Artificial / 7

4. In what year did you purchase your artificial tree?

 

5. What species or type of tree is (was) it? (IF
ARTIFICIAL TREE GO TO QUESTION 8.)

Natural (species)

 

Artificial (type)

 

115

 

10.

11.

116

How was your natural tree obtained?

Purchased 1:::7 Other 1:::7

Where was it purchased? (GET NAME OF RETAIL ESTAB-
LISHMENT OR OPERATOR AND STREET LOCATION.)

 

 

How old is the family member who selected the tree?
Under 10 1:7 10-14 1:7 15-19 1:7 20-24 1:7 Over 24 1:7

How many people live in your household?

 

Is your total family income:

Under $3,000 / 7 $ 9,000 - $11,999 / 7
$3,000 - $5,999 / 7 $12,000 - $14,999 / 7
$6,000 - $8,999 / 7 $15,000 and over / 7

 

How many rooms do you have in your living quarters?

(EXCLUDE BATHROOMS, UTILITY ROOMS, ENCLOSED PORCHES.)

 

APPENDIX B

117

Table l.--Variables selected for initial factor analysis.

 

REDUCTION ROUTINE

 

 

. General Description of specific
Variable number variables .
variables used
measured
Economic Variables
. . Proximity of competing retail
1 Competition lots - under 1 block
2 H Proximity of competing retail
lots - 1 to 2 blocks
3 H Proximity of competing retail
lots - 2 to 3 blocks
4 " Proximity of competing retail
lots — 3 to 4 blocks
5 " Proximity of competing retail
lots — 4 blocks to 1 mile
6 H Proximity of competing retail
lots — 1 to 2 miles
7 Trees available Trees for sale - sum
at lot
Retail prices
8 at which trees Average high price
offered
9 " " Average low price
10 " " Average price
11 Type retail Type of establishment -
establishment independent lot
12 n " Type of establishment - chain
grocery store
13 n " Type of establishment -
discount store
14 n " Type of establishment

nursery or florist

Table l.--(Cont'd.)

 

Variable number

REDUCTION ROUTINE

General
variables
measured

Description of specific
variables used

 

15

16

17

18

19

20

21

22

23

2a

25

26

27

28

29

3O

31

Type retail
establishment

H H

Location of lot

Tree quality

H H

Merchandising
at lot

Advertising
for lot

H

Economic Variables

 

Type of establishment -
church or civic group

Type of establishment —
independent grocery store

Type of establishment — other
Area location - business area
Area location - shopping center

Area location - residential
area

Area location - business area
and shopping center

Area location - business area
and vacant lot

Area location — residential
area and vacant lot

Tree quality - good
Tree quality - average

Tree quality — poor
Lot quality - good

Lot quality - average

Lot quality - poor
Advertisement - yes

Advertisement - no

119

 

REDUCTION ROUTINE

 

General Descri tion of s ecific
Variable number variables p. p
variables used
measured
Economic Variables
Sales Relationship of trees purchased
32 .
performance this year to last year - more
33 " Relationship of trees purchased
this year to last year - less
3” H Relationship of trees purchased
this year to last year - same
Relationship of trees purchased
35 " this year to last year - N.A.
(purchased no trees last year)
36 " Relationship of trees sold
this year to last year - more
37 " Relationship of trees sold
this year to last year - less
38 ” Relationship of trees sold
this year to last year - same
Relationship of trees sold
39 " this year to last year - N.A.
(sold no trees last year)
40 Size of Operates other retail lots — yes
operation
41 " " Operates other retail lots - no
Type of ,
42 . . Advertisement - newspaper
advertiSing
43 " " Advertisement - handbills
44 " " Advertisement - radio
45 " " Advertisement — TV
46 " " Advertisement - signs

 

1

(—

0

 

Variable number

REDUCTION ROUTINE

General
variables
measured

Description of specific
variables us ed

 

47

105

48

1+9

50

51

52

53

54

55

56

57

58

59

60

104

Knowledge of
business

Potential cus-

tomer traffic
volume

Population

Race

Households
(families)

H

H

Economic Variables

 

Number of years retailed
Christmas trees

Traffic count — AADV

Demographic Variables

 

Total population
White population
Negro population

Total foreign stock
Population in households

Primary families
Primary individuals
Wives of head of household

Children under 18 years of age
of head of household

Other relatives of head of
household

Nonrelatives of head of
household

Population in group quarters
(nearest hundredth of a person)

Population per household

Households

 

REDUCTION ROUTINE

 

 

. General Description of specific
Variable number variables .
variables used
measured
Demographic Variables
61 Families Married couples
62 " Married couples with own
household
63 " Married couples with own
children under 6 years of age
64 " Married couples with own
children under 18 years of age
65 " Married couples with husband
under 45 years of age
Married couples with husband
66 " under 45 years of age and with
own children under 18 years of
age
78 H Persons living in same house
in 1960 as in 1955
Young adults,
67 age, and group Unrelated individuals
quarters
72 " " Enrolled in college
79 n ” Persons living outside
Winston-Salem SMSA in 1955
88 " " Persons over 18 years of age
. Total school enrollment
C ld ’
68 hl ren, age 4 to 34 years of age
69 n H Enrolled in kindergarten
(private and public)
70 u H Enrolled in elementary

(private and public)

122

 

Variable number

REDUCTION ROUTINE

General
variables
measured

Description of specific
variables used

 

71

86

87

73

74

75

76

77

80

81

82

83

84

85

Children, age

H

Education

I!

H

H

H

Demographic Variables

 

Enrolled in high school
(private and public)

Persons under 6 years of age

Persons between 6 and 18
years of age

Persons 25 years old and over
with no school years completed

Persons 25 years old and over
with some elementary education

Persons 25 years old and over
with some high school education

Persons 25 years old and over
with some college education

Median school years completed
(to nearest tenth of year)

Families with income under
$3,000

Families with income between
$3,000 and $5,000

Families with income between
$5,000 and $10,000

Families with income over
$10,000

Median family income

Median family and unrelated
individual income

123

 

REDUCTION ROUTINE

 

 

General . . . .
Variable number variables Description Of SpelelC
measured variables used
Demographic Variables
89 Marital status Single persons over 14 years
of age
90 n " Married persons over 14 years
of age
91 n " Widowed persons over 14 years
of age
92 n " Divorced persons over 14 years
of age
93 Occupation Labor force (male and female)

Professional, skilled, and
94 " semi-skilled workers

(male and female)
95 " Laborers (male and female)

Type housing

96 unit (group Housing units
housing)
97 " " Owner occupied housing units
98 " ” Renter occupied units
99 " " Sound housing units

Median number of rooms per

loo H H . .
houSing unit (to nearest tenth)
101 n " Median number of persons per
housing unit (to nearest tenth)
102 " " Housing units built since 1940
103 n " Median value of owner occupied

housing unit

Table 8.——Highest

124

l
14 factor scores on 19 factors:-

 

 

 

Factors
1 2 3 4 5 8 7 8 9 10
Factor Score
Variable No.
—.98 .88 —.83 .89 .91 .93 .89 .91 —.79 .85
90 80 50 35 40 10 105 84 37 15
-.98 .84 —.76 .89 —.91 .85 .50 .88 .78 .83
53 5 18 39 41 9 2 103 38 44
-.97 —.77 —.75 -.70 .81 .82 .47 .84 .42 —.34
52 34 95 30 43 14 23 77 48 12
-.97 —.77 —.65 .70 .43 .79 .47 .88 .35 .30
81 87 3 31 4 8 22 101 29 89
-.97 -.75 -.80 .40 .41 .58 .48 .80 .28 .19
55 58 80 13 11 42 25 83 3 2
-.97 —.72 .58 -.38 .38 .35 -.45 .51 .25 .19
82 79 49 47 23 27 28 2 7 100
-.98 -.71 -.44 -.33 -.30 .32 -.40 .39 .25 .18
48 51 98 48 47 44 102 44 21 23
—.95 -.67 .41 .33 .30 -.28 .38 .38 .23 .18
84 8 82 28 7 28 59 78 89 101
-.95 .84 -.39 -.31 .23 .28 .35 .28 —.23 .17
78 100 73 42 27 18 100 51 28 80
-.94 -.83 -.33 .30 .23 .27 —.31 -.25 .18 .17
75 4 88 29 24 25 101 48 4 58
—.94 —.59 .33 —.28 .20 -.24 .31 —.23 —.18 .17
71 98 83 38 79 28 1 80 28 37
-.94 —.54 -.32 —.27 —.19 .18 —.23 .23 —.16 -.18
87 72 70 i2 42 5 18 85 22 77
-.94 .46 -.32 .24 -.18 -.17 .23 —.23 .13 .18
88 1 57 100 89 48 87 42 15 29
-.93 -.46 .31 -.24 —.18 .17 .21 .22 .13 -.16
83 92 97 25 12 32 72 31 17 11

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l/ Economic variables, 1967 retail lot surveys and traffic counts; 1960
Winston-Salem Census tract data. Analysis of 105 X 105 variable
matrix. Analysis type—factor R-Mode, using principal factor solu—
tion. Number of observations = 53; Eigen value out off at 1.00.

125

Table 8.--(Cont'd.)

 

Factors
ll 12 13 14 15 16 17 l8 l9

 

 

84
38

75
85

52
34

—.27
36

25
102

-.24
45

22
18

21
37

.19
33

.19
32

18
17
l2

16
23

.16
103

 

 

 

 

 

 

 

Factor Score

 

Variable No.

 

 

 

 

 

 

 

 

 

.70 .88 .89
13 72 33
.47 .58 —.73
21 59 32
.41 -.39 -.22
89 3 34
.38 —.38 .21
34 28 47
.38 —.35 -.21

7 83 45
.25 .32 -.19
23 29 8
.23 .31 .19
47 34 44
.22 .29 .18
32 79 28
.22 -.26 .14
28 58 18
.21 .24 _115_
27 88 59
.20 .23 -.14
11 87 80
.19 .23 -.13
18 89 102
.19 .21 -.12
85 31 48
.14 —.21 .12
78 30 72

.79

24

.68

27

.63

7

.51

22

.38

8

-.34
21

—.34
28

-.33
25

-.29
29

-.28
26

—.26
34

-.26
12

—.25
42

25
43

-.84

2O

-.33
25

30
26

-.28
102

24
21

-.23

21
34

19
45

19

—.18

32

—.17
28

17
29

-.17

.16
16

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.87 —.84 —.65
19 17 12
.27 —.43 .63
92 18 ll
.27 .34 -.57
38 21 45
-.27 -.24 —.54
26 26 42
.27 .23 -.45
25 46 47
—.25 .23 —.40
102 23 18
-.24 .22 .37
60 12 22
.24 .18 .36
28 l 23
-.24 .15 .27
29 13 59
.24 .14 26
59 102 3
.23 .14 —.26
4 25 83
.22 -.13 -.25
100 20 l
—.21 .13 -.23
46 9 76
-.21 —.13 .22
18 2 73

 

126

Table 9A.-—Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #1
Variable Score Description
90 -.98 Married persons over 14 years of age
53 —.98 Primary families
52 -.97 Population in household
55 -.97 Wives of head of household
62 -.97 Married couples with own household
48 -.96 Total population
64 -.95 Married couples with own children under 18 years
of age
78 -.95 Persons living in same house in 1960 as in 1955
75 -.94 Persons 25 years old and over with some high
school education
71 -.94 Enrolled in high school (private and public)
87 -.94 Persons between 6 and 18 years of age
88 —.94 Persons over 18 years of age
63 -.93 Married couples with own children under 6 years

of age

127

Table 9B.--Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #2
Variable Score Description
60 .86 Population per household
5 .84 Proximity of competing retail lots - 4 blocks
to 1 mile
34 -.77 Relationship of trees purchased this year to
last year
67 -.77 Unrelated individuals
58 -.75 Nonrelatives of head of household
79 -.72 Persons living outside Winston-Salem SMSA in 1955
51 -.71 Total foreign stock
6 -.67 Proximity of competing retail lots — 1 mile
and over
100 .64 Median number of rooms per housing unit (to
nearest tenth)
4 —.63 Proximity of competing retail lots - 3 to 4
blocks
98 -.59 Renter occupied units
72 -.54 Enrolled in college
1 .46 Proximity of competing retail lots — under
1 block

92 -.46 Divorced persons over 14 years of age

128

Table 9C.—-Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #3
Variable Score Description
50 -.83 Negro population
16 -.76 Type of establishment — independent grocery store
95 —.75 Laborers (male and female)
3 -.65 Proximity of competing retail lots - 2 to 3
blocks
80 -.60 Families with income under $3,000
49 .56 White population
98 -.44 Renter occupied units
82 .41 Families with income between $5,000 and $10,000
73 —.39 Persons 25 years old and over with no school
years completed
86 -.33 Persons under 6 years of age
83 .33 Families with income over $10,000
70 -.32 Enrolled in elementary (private and public)
57 -.32 Other relatives of head of household
97 .31 Owner occupied housing units

 

129

Table 9D.--Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #4
Variable Score Description
35 .89 Relationship of trees purchased this year to last
year (no trees)
39 .89 Relationship of trees sold this year to last year
(less)
30 —.70 Advertisement - yes
31 .70 Advertisement — no
13 .40 Type of establishment - discount store
47 -.38 Number of years retailed Christmas trees
46 —.33 Advertisement — signs
26 .33 Tree quality - poor
42 -.31 Advertisement - newspaper
29 .30 Lot quality - poor
36 -.28 Relationship of trees sold this year to last year
(more)
12 -.27 Type of establishment — chain grocery store
100 .24 Median number of rooms per housing unit (to
nearest tenth)
25 -.24 Tree quality - average

130

Table 9E.-~Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #5
Variable Score Description
40 .91 Operates other retail lots - yes
41 -.91 Operates other retail lots - no
43 .81 Advertisement — handbills
4 .43 Proximity of competing retail lots 3 to 4 blocks
11 .41 Type of establishment — independent lot
23 .36 Area location — residential area and vacant lot
47 -.30 Number of years retailed Christmas trees
7 .30 Trees for sale - sum
27 .23 Lot quality - good
24 .23 Tree quality - good
79 .20 Persons living outside Winston—Salem SMSA in 1955
42 —.19 Advertisement — newspaper
69 -.18 Enrolled in kindergarten (private and public)
l2 -.18 Type of establishment - chain grocery store

131

Table 9F.--Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #6
Variable Score Description
10 .93 Average price
9 .85 Average low price
14 .82 Type of establishment - nursery or florist
8 .79 Average high price
42 .56 Advertisement - newspaper
27 .35 Lot quality - good
44 .32 Advertisement - radio
26 -.28 Tree quality - poor
18 .28 Area location — business area
25 .27 Tree quality - average
28 -.24 Lot quality - average
5 .18 Proximity of competing retail lots - 4 blocks
to 1 mile
46 -.17 Advertisement — signs
32 .17 Relationship of trees purchased this year to last

year (more)

 

132

'Téat>le 9G.——Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #7
\faaxniable Score Description
.105 .89 Traffic count - AADV
2 .50 Proximity of competing retail lots - l to 2
blocks
23 .47 Area location - residential area and vacant lot
22 .47 Area location - business area and vacant lot
25 .46 Tree quality — average
26 -.45 Tree quality - poor
102 —.40 Housing units built since 1940
59 .36 Population in group quarters (nearest hundredth
of a person)
100 .35 Median number of rooms per housing unit (to
nearest tenth)
101 -.31 Median number of persons per housing unit (to
nearest tenth)
l .31 Proximity of competing retail lots - under
1 block
18 —.23 Area location - business area
67 .23 Unrelated individuals
72 .21 Enrolled in college

133

Table 9H.--Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #8
Variable Score Description
84 .91 Median family income
103 .88 Median value of owner occupied housing unit
77 .84 Median school years completed (to nearest tenth
of year)
101 .68 Median number of persons per housing unit (to
nearest tenth)
83 .60 Families with income over $10,000
2 .51 Proximity of competing retail lots - l to 2
blocks
44 .39 Advertisement — radio
76 .36 Persons 25 years old and over with some college
education
51 .26 Total foreign stock
46 -.25 Advertisement - signs
60 -.23 Population per household
85 .23 Median family and unrelated individual income
42 -.23 Advertisement - newspaper
31 .22 Advertisement - no

 

134

Table 9I.--Identification of tOp l4 variable loadings on 19 principal

 

 

factors.
Factor #9
Variable Score Description

37 -.79 Relationship of trees sold this year to last year
(less)

36 .76 Relationship of trees sold this year to last year
(more)

46 .42 Advertisement — signs

29 .35 Lot quality - poor

3 .26 Proximity of competing retail lots - 2 to 3
blocks

7 .25 Trees for sale - sum

21 .25 Area location - business area and shopping center

69 .23 Enrolled in kindergarten (private and public)

28 -.23 Lot quality — average

4 .18 Proximity of competing retail lots - 3 to 4
blocks

26 -.16 Tree quality - poor

22 —.16 Area location - business area and vacant lot

15 .13 Type of establishment - church or civic group

17 .13 Type of establishment - other

 

135

Table 9J.—-Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #10
Variable Score Description
15 .85 Type of establishment — church or civic group
44 .63 Advertisement - radio
12 -.34 Type of establishment - chain grocery store
69 .30 Enrolled in kindergarten (private and public)
2 .19 Proximity of competing retail lots - l to 2
blocks
100 .19 Median number of rooms per housing unit (to
nearest tenth)
23 .18 Area location - residential area and vacant lot
101 .18 Median number of persons per housing unit (to
nearest tenth)
60 .17 Population per household
58 .17 Nonrelatives of head of household
37 .17 Relationship of trees sold this year to last year
77 —.16 Median school years completed (to nearest tenth
of year)
29 .16 Lot quality - poor
11 -.16 Type of establishment - independent lot

 

136

Table 9K.--Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #11
Variable Score Description
38 .84 Relationship of trees sold this year to last year
(same)
85 .75 Median family and unrelated individual income
34 .52 Relationship of trees purchased this year to last
year (same)
36 -.27 Relationship of trees sold this year to last year
(More)
102 .25 Housing units built since 1940
45 —.24 Advertisement — TV
18 .22 Area location — business area
37 .21 Relationship of trees sold this year to last year
(less)
33 .19 Relationship of trees purchased this year to last
year (less)
32 .19 Relationship of trees purchased this year to last
year (more)
5 .l8 Proximity of competing retail lots - 4 blocks to
1 mile
12 .17 Type of establishment - chain grocery store
23 .16 Area location - residential area and vacant lot

103 .16 Median value of owner occupied housing unit

137

Table 9L.--Identification of tOp l4 variable loadings on 19 principal

 

 

factors.
Factor #12
Variable Score Description
13 .70 Type of establishment - discount store
21 .47 Area location - business area and shopping center
69 -.41 Enrolled in kindergarten (private and public)
34 .38 Relationship of trees purchased this year to last
year (same)
7 .36 Trees for sale - sum
23 —.25 Area location - residential area and vacant lot
47 -.23 Number of years retailed Christmas trees
32 -.22 Relationship of trees purchased this year to last
year (more)
28 .22 Lot quality - average
27 —.21 Lot quality - good
11 —.20 Type of establishment - independent lot
18 —.19 Area location - business area
85 —.19 Median family and unrelated individual income
76 -.14 Persons 25 years old and over with some college

education

 

138

Table 9M.——Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #13
Variable Score Description

72 .66 Enrolled in college

59 .58 Population in group quarters (nearest hundredth
of a person)

3 —.39 Proximity of competing retail lots - 2 to 3
blocks

28 —.38 Lot quality — average

83 —.35 Families with income over $10,000

29 .32 Lot quality - poor

34 .31 Relationship of trees purchased this year to last
year (same)

79 .29 Persons living outside Winston-Salem SMSA in 1955

58 -.26 Nonrelatives of head of household

68 .24 Total school enrollment, 4 to 34 years of age

67 .23 Unrelated individuals

89 .23 Single persons over 14 years of age

31 .21 Advertisement - no

30 -.21 Advertisement - yes

 

139

Table 9N.--Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #14
Variable Score Description
33 .89 Relationship of trees purchased this year to last
year (less)
32 -.73 Relationship of trees purchased this year to last
year (more)
34 -.22 Relationship of trees purchased this year to last
year (same)
47 .21 Number of years retailed Christmas trees
45 —.21 Advertisement - TV
8 -.19 Average high price
44 .19 Advertisement - radio
28 .18 Lot quality - average
18 .14 Area location — business area
59 .14 Population in group quarters (nearest hundredth
of a person)
6O -.14 Population per household
102 —.13 Housing units built since 1940
46 -.12 Advertisement - signs
72 .12 Enrolled in college

 

140

Table 90.-—Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #15
Variable Score Description

24 .79 Tree quality - good
27 .68 Lot quality — good

7 .63 Trees for sale - sum
22 .51 Area location - business area and vacant lot

8 .38 Average high price

21 —.34 Area location - business area and shopping center
28 -.34 Lot quality - average
25 -.33 Tree quality - average
29 —.29 Lot quality — poor
26 —.28 Tree quality - poor

34 —.26 Relationship of trees purchased this year to last

year (same)

12 -.26 Type of establishment — chain grocery store
42 -.25 Advertisement - newspaper
43 .25 Advertisement - handbills

 

141

Table 9P.--Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #16
Variable Score Description
20 —.84 Area location — residential area
25 -.33 Tree quality — average
26 .30 Tree quality - poor
102 —.28 Housing units built since 1940
21 .24 Area location — business area and shOpping center
2 -.23 Proximity of competing retail lots - l to 2
blocks
34 .21 Relationship of trees purchased this year to last
year (same)
45 .19 Advertisement - TV
1 .19 Proximity of competing retail lots - under 1
block
32 —.18 Relationship of trees purchased this year to last
year (more)
28 -.17 Lot quality - average
29 .17 Lot quality — poor
7 -.l7 Trees for sale - sum

16 .16 Type of establishment — independent grocery store

 

142

Table 9Q.--Identification of tOp l4 variable loadings on 19 principal

 

 

factors.
Factor #17
Variable Score Description
19 .87 Area location - shopping center
92 .27 Divorced persons over 14 years of age
38 .27 Relationship of trees sold this year to last
year (same)
26 -.27 Tree quality - poor
25 .27 Tree quality — average
102 -.25 Housing units built since 1940
60 -.24 Population per household
28 .24 Lot quality - average
29 -.24 Lot quality — poor
59 .24 Population in group quarters (nearest hundredth
of a person)
4 .23 Proximity of competing retail lots - 3 to 4
blocks
100 .22 Median number of rooms per housing unit (to
nearest tenth)
46 -.21 Advertisement - signs

18 -.21 Area location - business area

 

143

Table 9R.--Identification of top 14 variable loadings on 19 principal

 

 

factors.
Factor #18
Variable Score Description
17 -.84 Type of establishment - other
18 -.43 Area location — business area
21 .34 Area location - business area and shopping center
26 -.24 Tree quality - poor
46 .23 Advertisement - signs
23 .23 Area location - residential area and vacant lot
12 .22 Type of establishment - chain grocery store
1 .18 Proximity of competing retail lots - under 1
block
13 .15 Type of establishment - discount store
102 .14 Housing units built since 1940
25 .14 Tree quality - average
20 -.13 Area location - residential area
9 .13 Average low price
2 -.13 Proximity of competing retail lots - l to 2

blocks

 

144-

Table 9S.--Identification of tOp l4 variable loadings on 19 principal

 

 

factors.
Factor #19
Variable Score Description

12 -.65 Type of establishment - chain grocery store

11 .63 Type of establishment - independent lot

45 -.57 Advertisement - TV

42 —.54 Advertisement - newspaper

47 —.45 Number of years retailed Christmas trees

18 -.40 Area location — business area

22 .37 Area location — business area and vacant lot

23 .36 Area location — residential area and vacant lot

59 .27 Population in group quarters (nearest hundredth
of a person)

3 .26 Proximity of competing retail lots - 2 to 3

blocks

83 -.26 Families with income over $10,000

1 —.25 Proximity of competing retail lots - under 1
block

76 -.23 Persons 25 years old and over with some college
education

73 .22 Persons 25 years old and over with no school

years completed

 

145

Table 10.-—Forty—nine highest scoring variables selected from 14 X 19

factor matrix.

 

 

Variable Score Description
number

2 .50 Competition - 1—2 blocks

3 -.39 " - 2-3 blocks

5 .84 " - 4 blocks-l mile

7 .63 Total trees for sale

9 .85 Average low price

10 .93 Average price
11 .63 Type enterprise — independent lot

12 -.65 " " - chain grocery
13 .70 " " — discount store
14 .82 " " — nursery or florist
15 .85 " " - church or civic group
16 —.76 " " - independent grocery
17 -.84 " " - other
18 -.43 Area location - business area
19 .87 " " - shopping center
20 —.84 " " - residential area
21 .47 " " - business 6 shopping center
23 .47 " " - residential area 8
vacant lot

24 .79 Tree quality - good

25 —.33 " " — average

26 .30 " " - poor

27 .68 Merchandising _ good

on lot

Table lO.--(Cont'd.)

146

 

Variable

 

number Score Description
30 —.70 Advertisement - yes
32 -.73 Year to year tree purchases - more
33 .89 " " " " " - less
34 —.77 " " " " " — same
35 .89 " " " " " — NA
36 .76 Year to year tree sales — more
37 -.79 " " " " " - less
38 .84 " " " " " - same
39 .89 " " " " " - NA
40 .91 Number of lots operated — l
41 —.91 " " " " - l
43 .81 Where advertised — handbills
4” '63 " " - radio
45 -.57 " " - television
46 .42 " " - signs
50 —.83 Negro population
59 .58 Population in group quarters
6O .86 Population per household
69 -.41 Enrolled in kindergarten
72 .66 Enrolled in college
77 .84 Median school years completed
84 .91 Median family income

147

 

 

Variable Score Descri tion
number p
85 .75 Median family and unrelated individual
income
92 .27 Divorced persons over 14
95 -.75 Laborers
103 .88 Median value of owner occupied housing
units

105 .89 Traffic count AADV

 

148

Table ll.—-List of 41 variables entered in stepwise regression models--

Winston-Salem models; 1967, 1968, 1969.

 

 

 

Variable Description
number
Economic Variableiq
1 Competition - 0-4 blocks
2 Competition - 4 blocks-l mile
3 Traffic count (AADV)
4 Average low price
5 Average price trees sold
6 Type of enterprise - independent lot (dummy)
7 " " " - chain grocery store (dummy)
8 " " " - discount store (dummy)
9 " " " - nursery or florist (dummy)
10 " " " - church or civic group (dummy)
ll " " " - independent grocery store (dummy)
l2 " " " — other (dummy)
13 Area location - business area (dummy)
l4 " " — shopping center (dummy)
15 " " — residential area (dummy)
16 Tree quality (dummy)
(1) have quality (0) don't have quality
17 Merchandising (dummy)
(l)have merchandising (0) don't have merchandising
18 Advertised (dummy)
(1) did (0) did not
19 Trees for sale this year (P)

20 Trees for sale last year (P-l)

149

Table ll.—-(Cont'd.)

 

 

 

 

Variable Description
number
21 Ratio - trees for sale (P31)
22 Trees sold this year (P)
23 Trees sold last year (P-l)
24 Ratio trees sold (P/P-l)
DemOgraphic Variables
25 Negro pOpulation
26 Population in group quarters
27 Population per household
28 Enrolled in kindergarten
29 Enrolled in college
30 Median school years completed
31 Median family income
32 Laborers
33 Median value of owner occupied housing units
34 Total population
35 Children under 18 years of age of head of household
36 Median number of persons per housing unit
37 Number of households
Variables From Consumer Survey
38 Do not have a Christmas tree
39 Have an artificial Christmas tree
40 Have a purchased natural tree
41 Have a natural tree, not purchased

E

.pmpmomam mammHsm> noflnz aw Hopoz.l

 

 

\H
osowm whooosm
pcopcmaopcfl mpoa oa>e ox
oEoocw whoppmsv 1 msopw
>Hflemm cmwpoz hHHvrmx\m apogm cw coflpmasaom mox 0H>Ho no nopszo mpoa oa>e mx
A>Q<<v ad on pmwpoam :
pcsoo owmmmae AHV x UHom mommy mo ooflgm x no %pompsc mpoa oa>e x
90H unopawmmpcw
0 5H: 38:3
Nu lawnosoe moanmfihm> mafia a I
cofipomgmch AHVmHX\m mxooan : mcoflpflpwaeoo max oQOpm pcsoowwp mpoa omzs mx
hopcmo mcflmaozm
ma poa shown mQOpm panoomwp
cwmno moapmfipm> L4H: wcwmflpcmnosms
coflpomgmch AHVNHX moahmflpm> cowpomsmch :Hx zpmoopm camso mpoa maze mx
mmpm mmocflmSQ
omwpcmsopme . CH poa pampcomopcfi
pap mmcwmwpamsohmz AHVmX\w moanmflpm> cowpompoch oax wcmpcoampcfi mpoa came ax
Hopoe oco maopoe 03p macros Ham
cowpmwsomoa maao CH coflpaflsomoa >Hco aw CoHpQHQome CH
pcsom mason venom

 

.oz .anmmnCOpmcwz mHHH new .HH .H maopoe wcflpoflpoha cw mapmmadm moanmwsm>nl.ma mH@m&

151

.Umpmmadm manmflpm> Lowgz Ca Hopoz.l

 

 

3
mafia:
mcflmson mo opam> AHHmex momaaoacm mmmaaoo omx
pmvoameoo
mpmom Hoonom cmflpoz AHHVHmX\w mxooan :Io ”cowpwuoaeoo Z”x
mmpm mmonmdn
CH #OH oQOpm cwmgo
mpaocmmsoz mo popesz AHvamx mmapmwsm> GOMHOQOch Hax
was: popcoo
mammzoz pom mcomsom AHvbmx mafiaaogm mcowpmooa poq mx
mopm
whoponmq Avamx mmocflmsn mcowpmooa poq N.x
pros mco macros 034 maopoe Ham
coflpawpomoa maco ca cowpafipomom maco CH cowpawsowmn CH
pczom canon canon

 

A.®.pcoovll.ma mHAmB

152

Table 13.--List of variables entered in final stepwise regression models--

Winston—Salem models; 1967, 1968, 1969, and Denver model, 1965.

 

 

 

Vifiigiie Description
Economic Variables

Xl Type lot - independent

X2 " " - chain grocery store

X3 " " - discount store

X” " " - nursery or florist

X5 " " - church or civic group

X6 " " — independent grocery store

X7 Lot location — business area

X8 " " - shopping center

Xg Merchandising - did merchandise

Xlo Interaction variable - indapendent lots located in
buSiness areas

Xll " " - chain store lots located in
buSiness areas

Xl2 " " - chain store lot located in
shopping center

X13 " " — merchandising with independent
lot

X14 " " - merchandising with discount
store

X15 " " - merchandising in business area

X16 " " — merchandising in shopping
center

Xl7 Competition - 0-4 blocks

X " — 4 blocks—l mile

Table l3.--(Cont'd.)

153

 

 

 

Variable Description

number

X19: Traffic count (AADV)

X2O Prices of trees sold

Demographic Variables

X21 Negro population

X22 Population in group quarters
X23 POpulation per household

X2” Median family income

X25 Laborers

X26 Children under 18

X27 Persons per housing unit

X28 Number of households

X2g Kindergarten enrollees

X30 College enrollees

X31 Median school years completed
X32 Value of housing units

X33 Population

X (Y) Trees sold

34

 

154.

 

mmaacmaopde can . mcwmfiacmnohmz momsm.ama- was. memo. x
popcmo wcfladosm u coHpmooa you mssmm.momu boo. mmoo. mx
mopm mwonmnn u cowpmooa poq mmzmm.omH who. ooao. bx
shown m
>pmoosm ucopcmaoch : mazp poq ooomm.mman 000.: mmjo. x
macaw m
oa>ao no garage - wasp Log mammm.©:m mam. 4830. x
pmflsoam no sadmnsc - wasp hog mmomm.:sm- mao.- mmao. ax
mnowm pcsoomflv I wash hog mamas.oo: mma. same. mx
oLODm mpmoopw cflmno I mazu you :moso.:m n baa.u ommo. mx
pcmucmdmacw I wasp you mm::m.om 8mm. mmaa. ax
paom
mommy moapmfigm> pcopcmmom va :mx
pampmcoo ms:H:.mws- ox
pcoaoa moo cowpmamppoo 0mm ow popes:
cowpawsomoo COmewwmom mo coapznapwcoo magmamm>
. pcmmowmmooo . . .

 

A.mmm.o u m .mam:.sm

u nosnm .apm .maom.o H mm .mmmm.o u m .om u x .mm H z "mofipmmndpomhmgo Hmoapmapmpmv .soaa

HZL.Eonmnc0pmcfl3 mmoamm soap mmEpmwhso amazpmc pom HopoE mCflwowpmsm "H prOE mo cowuawsowooun.ma magme

 

155

om

 

mooHHOQCm owwaaoo wmmmm.o: : mad. mmao. x
was: wcamzog pom mcompmm bmzso.m mac. mmmo. smx
whosopmq Haaom.am moo. mmao. max
paom momsp mo oowhm mmomm.moa mam. maaa. omx
A>o<<v pcsoo oammmse mmaam.m ham. mmao. max
mafia a or madcap : - coapflpdasoo Hmomm.am use. Hmoo. max
madcap 3-0 - coapapddeou mmamo.so Ham. ammo. sax
poa pcopcoaopcfl nwwz wcfimflpcmno
name I manmwpm> coflpompovcH ommmm.sma spa. ammo. max
popcoo mcwaaonm aw woa oQOpm
cflmno I manmwsm> cowpompoch :smmm.ana moa.| ommo. max
mopm mmoaHmSQ cw uoa shown Ha
cfimno I wanmwpm> cofipompmch mmbam.mman ooa.u mzmo. x
mosm mmocwmsn cw poa unopcma
-dacH - magmanm> ceapompmch mmmsm.©s - ama. 8440. oax
pcoaoa woo coHpmaospoo o o 90 ans
nowadagomoo . .mww mo mm H A
cowmmms om ucmHOHmwmoo cowvsnwppcoo mHAMw9m>

 

A.8.pdooo--.84 84889

156

mono mmocflmsp cw mpoa

pcopcwa

 

 

toonn - onnonno> nonnomnonnn moos:.ssm om:. Oman. 04x
nonooo mnnddonm - nonnoooa non o:o:m.mmn- oon. Hooo. ox
mono moonnmon . nonnmooa no; masms.oaa ska. mafia. ax
oQOPm
snooonm nnoooodoonn - nod odsn maoao.maas 044.- ammo. ox
adopw m
on>no no nonono - non damn ommam.asm sma. momo. x
nonnoam no snomnoo - non dose :momo.m: - mno.- mmao. ax
ononm ncooomno - noa odsn sommm.moa- soo. maoo. mx
ononm snooonw anono - non odsn mmmmo.ooa- NHH.- maao. ax
nooonodooon - non wasp moamo.mmm mam. :ss.o ax
oaom
moonp I waannm> vampcomoo va :mx
noonmnoo mmooo.m:a ox
pcoaoa moo coaymamnnoo 0mm ow popes:
cofipaflnomom COmwwwwmom mo COHPSAHpvcoo magmanm>
. noononnmooo . . .
A.mom.m u n .oomm.msa
u nonnm .onm .osms.o n ma .msmm.o m .ma u g .o: "monnmnnonomnoao Hmonnmnnonmv .mooa

oz.Ewamm1COpmcHz mmmamm pony mmEumezu Annapm: pow Hopoe

wapoHpmnm

“HH 48808 no nonndnnomoo--.oa wagon

157

 

mm

 

mnnoo mnnmoog no ooam> 0m:m0.mm - 004. 0400. x
popoaaaoo mnmom Hoonom swarm: owmos.amsn mow. some. me
momaaoncm mwoaaoo mamom.:m - mod. mzao. omx
mpaozomson wo nonesz mmomm.ma mmo.u mmao. max
oeooon saneon nonoo: m00m0.m:a 0mm. 0s40. sax
unopnmsa msonm CH cowpmazaom bmmmm.smb mmo. momm. mmx
oxooan 0-0 - connnnoQEoo 040m:.mza :43. 80:0. sax
onoum pcsoomwp 54H: mcflmwpcmzo :H
1905 - manmwnm> powwomnoch :mmom.©man omo.- memo. x
mmnm mmocflmsp cw poa whoum H
chLo - manmwnm> coHvomnoch bmzmm.azau smo. Hmao. H x
c0404 moo coHpmaonLoo 0 0p no a:
cofipafinomoa pcommwwwmo Mo Goa ”map co m Mfinm
. m paoflowmmmoo .p p. u u an . >

 

A.p.pcoovuu.ma manmh

158

monm mmmamsn Ca #04 phonm

 

cflmno I mapmfinm> coflpomnmch m40m0.004I 000. 4:40. 44x
noncoo mn4aaonm - no4nmoo4 004 00:00.40 I 404. 0000. 0x
mwnm mmochsn I coflpmooa poq 0004m.m0 :00. 0040. 0x
anonm
mnooonw pcopcoampcfl I #04 ma>H 30004.004I m00.I 3000. ex
asonm m
o4>4o no conono - no4 mash 04044.00m 004. :m40. x
pmwnoam no znomnzc I #04 md>e 00043.000I 040.I maao. :x
ononm ncsoom4o - 404 mass a:omm.mm I 000. N000. 0x
mQOpm m
>nooosw Cwmzo I #04 came 00044.mm I 3:0.I 0:40. x
nnoocoaoan4 - 404 mass 00000.000 0mm. 4004. 4x
040m
moon» I manmflnm> pampcoaom va :mx
puppmcou 00:00.404I 0x
#:0404 moo coflpmaonsoo o m or so an
coflpdwnomoo cemmwwwmom mo c04pmn4npaoo 04AMWQm>
. Pcowoflmwooo . . .

 

u nonnm .ppm .m:40.0

u m .ms4s.0 u m .04 u g .00

m

4.440.m u n .0000.0mm

“mowpmwnopomnmno HmOflpmwympmv

H:A.EmammICOpmcfl3 mmoamm pony mmEpmfihzo awthm: pom 40005 mafipoflpoha

"HHH Hmpoe mo cowpmwmomeII.hH oanme

159

 

00

 

 

00044onco omo44oo 44000.00 0:4. 0:40. x
whopnmsv @5090 C4 c04404daom 04000.040I 000.I 0000. mmx
04om moonp mo 000440 00000.00 44m. 0000. 00X
0446 4 04 0x004p : I CO4w4panoo 00000.00 404. 0000. 04x
mxoo40 4-0 - no404nodeo0 04000.00 000. 00:0. 04x
onOPm vasoom40 L443 mC4m40cm£o :4
InmE I o4nm4nm> c04pomnoch :0000.004I 004. 0000. x
4:0404 moo c04pM4mnsoo o m 04 gonads
CO4pa4nomoo cowmwwwwmm Mo c04pwn4npcou o4am4pm>
. 4:0404mmmoo . . .
11f
A p.4coovII.04 04405

160

 

 

0040000on00 040 - 004040000on0: 00000.000 - 000. 0000. x
n0000o 00400o00 - 0o4nmoo4 noq 00000.0:0 400. 0000. 0x
00n0 00004000 - 0o4nmoo4 noq 400:0.000 - 004. 0000. 0x
wQOPm m
>n0oonm 40000000004 0004 404 :0000.044 000. 0000. x
QDOQW m
04540 no 0ono0o 0000 noa 00000.00 000. 0000. x
n04no40 no 0n00n00 0000 noa 000:0.000 040. 0000. 0x
0n04m 40300040 00>4 404 0000:.:0 I 004. 0000. 0x
0non0 0n0oon0 04000 0000 noq 00044.0044- 004.- 0000. 0x
00000000004 0000 noq 00040.000 004. 0000. 4x
00094 I 04004nm> 4C00C0a00 400 x
000000o0 00::0.:004 0x
00404 000 c044040pnoo o o 90 s:
COHPQHQUmmD #Commwwwwm MO C04 MMHPPCO m Mfihm
. m 000404000o0 .0 0. n 0 40 . >
4.000.0

.0000.000 0 nonnm .000 .0000. 0

.0004 .00400 0004 mm€pm4n£0 404540: 000 40008 mC440400na

m 04:00.0 u M .0m n x hN04 u z "0044049040090no 40044044040V

"4000E 00 .n0>c00 0:4 00 00440490000II.04 04400

 

161

mm

 

000440400 000400400040 00400.00 000.I 0000. x

000004 044000 004002 40000.000 000. 0000. 00x

040000000 000 0040040000 04400.000 I 004.- 0040. 00x

04044030 @5040 C4 C04404saom 00000.000 I 404.I 0040. max
404C00 wC4amonm C4 wC4040sz0 04

I405 I 040044m> C04400404C4 00000.4004I 004.I 0040. x
0040 000C4094 C4 mC4040C0£0 04

I405 I 044044m> C04400404C4 00400.00: 040.I 0000. x
04040 4C30004© £443 mC404UCwn0 :4

I405 I 0440440> C04400404CH 00440.40: 400.I 0040. x
404 4C0UC0Q00C4 £442 MC4040C0£0 04

I405 I 044044m> C04400404C4 00000.000 I 000.I 0040. x
404C00 mC4aaonm C4 404 04040 04

C4030 I 044044m> C04400404CH 00040.0004I 004. 0000. x
0040 000C4034 C4 404 04040 44

C4040 I 044044m> C04400404CH 40400.000 040.I 0:40. x
0040 000C4054 C4 404 4C00C0a 04

I00C4 I 040044m> C04400404CH 00000.004 000. 0000. x

C0404 000 C044m404400 o 04 40 3C
C0440440000 4C0mmwwwm0 mo C04 wm44 Co 0 ””40
. m 4C0404mm000 .4 4. 4 0 44 . >

 

4.0.00000II.04 04000

2(10”)

(YD-YI)

29253
2550}
2175.

1800;

1u25

300 .

'75

“M50

1050

675 f

 

"-9od

  
       

-606 4300 o

Maw-Ware . :4“ .. ‘wIwfim‘Mm M?

162

300
Y

I

600

900

an]. ¢ umwmit m e-Huyz‘emmmm*m‘

1200 1500 1800

Figure 8.--Squared residual plots of estimated Denver retail
I.

sales with model

163

1840

lSHO

Luuo

1240

)

 

: lOMO
<3
r4
&“
H 800
H
<>*
>3:
\/ 640
”HO
2H0
40
O
-600 -300 () 300 600 900 1200 1500 1800 2100
V Y \
II

Figure 9.--Squared residual plots of estimated Denver sales
with model II.

164

' 2100 .
1850

1600

350 o o o

100 on o 1.. . g

-150

-H00 -200 O 200 #00 600 800 1000 1200 1MOO

A

YIII

Figure lO.--Residual plots of estimated Denver retail sales
with model III.

165

1720

1520 o

1320

1120 '

2 4
D) (10 )
co
k)
<3

720

(Y -Y
ws67

520
320 o

120 o o

0 0000M. G
-80

—500 0 500 1000 1500 2000 2500 3000 3500 4000

YD

Figure ll.--Squared residual plots of estimated 1967 Winston-
Salem retail sales with the Denver model.

166

138

123'

108

‘i‘ 63
[\ O
(O
(f)
f; 48 .
33
O . .
18
O O O
. o. o o
O O O O. O O
3 o co 1&1.
. Co .0.- .. o

-400 -2OO 0 200 400 600 800 1000 1200 1400
II

Figure 12.--Squared residual plots of estimated 1967 Winston-
Salem retail sales with model II.

314.5.

   
  
  

167

 

.
.
C
.
.
.
. O
. C
. .
. C .
O
O . C
C .0
C . O. .
. O . O O.
C C -. . . . .
' .i'rewfixirnmwzmfimfm‘lflm«:
-30 -15 0 150 300 450 600

YIII

750

  

"* W~%W . 4%,: . «91..» ‘r‘ysfimriw-h?» mewqu: 9.4-1%. 7 mm. mm»

900

 

1050

Figure 13.--Squared residual plots of estimated 1967 Winston-

Salem retail sales with model III.

168

1720

1520

1320

720 o
m o
co

g o
31 520
o
320
o
120 o.
I». o
nib ID 0
0 ocean‘s—lb... up
-80

 

 

 

 

W __.,__ 1.121-: :1: _ , , . ‘,‘ f, , ._ . .. __ .... _._ ,._ {_fi— .._ _, p}
-500 O 500 1000 1500 2000 2500 3000 3500 4000

YD

Figure l4.--Squared residual plots of estimated 1968 Winston-
Salem retail sales with Denver model.

169

920

820

720

620

520

2 3
I) (10 )

420

(Y ~Y
ws68

320

220

 

-100 O 100 200 300 400 500 600 700 800

Figure 15.--Squared residual plots of estimated 1968 Winston-
Salem retail sales with model I.

170

43
38
33
28
1A
O
H
c;« 23 .
AH
H
H

-300 -150 0 150 300 050 600 750 900 1050
YII

Figure l6.--Squared residual plots of estimated 1968 Winston-
Salem retail sales with model III.

171

1720.

1520.

1320.

320'

120,

+80

4
My; w . , mmmmwwaw wawwuwbmw WWW-m asw‘é'mm finfimm':mwm.

-500 0 500 1000 1500 2000 2500 3000 3500 4000

YD

Figure l7.--Squared residual plots of estimated 1969 Winston-

Salem retail sales with the Denver model.

172

139

123

108

(>14 6 o
m
o
>43
V [vi-3‘
o
3 o.
o
18 . . '
o
o.
o O 00 no
0. mo. 0
O 0.00-"... o O O
-u‘o —300 -150 0 150 30c o .00 750 900
YI

Figure 18 --Squared residual plots of estimated 1969 Winston-
Salem retail sales with model I.

 

173

86

76

665

56

U6“

  
 

  
      

-._..'-»‘.t‘<ehr"'<m'w¢.w”mam:“mmzumsewzw»*. - . -. 1.,

—3OO —150 0 150 300 450 600 750 90

YII

 

Figure l9.--Squared residual plots of estimated 1969 Winston-
Salem retail sales with model II.

LIST OF REFERENCES

 

LIST OF REFERENCES

Black, R.
1962. Are growers the artificial trees best boosters?
American Christmas Tree Growers Journal 6(2):6.

Brundage, R. C.
1958. Consumer preference for Christmas trees and
natural decorative materials. Agr. Exp. Sta. Cir. 669.
Purdue, Univ., Lafayette, IN.

Brundage, R. C., W. Nicewander, and R. L. Kohly.
1955. Christmas trees purchased by families in a medium-
sized city. Agr. Exp. Sta. Cir. 416. Purdue Univ.,
Lafayette, IN.

Conklin, M. J.
1962. The market for pacific northwest Christmas trees.
American Christmas Tree Growers Journal 6(1):8-9, 58-66.

Draper, N. R., and H. Smith.
1966. Applied regression analysis. John Wiley & Sons,
Inc., NY.

Drysdale, D. P., and I. A. Nausedas.
1970. The Christmas tree market of metropolitan Toronto.
American Christmas Tree Growers Journal l4(4):15-21.

Ehrenberg, A. S. C.
1962. Some questions about factor analysis. The Statis-
tician 12(3):19l-208.

Gatty, R.
1966. Multivariate analysis for marketing research.
Applied Statistics 15:151-172.

Green, P. E., and D. S. Tull.
1970. Research for marketing decisions. 2nd Edit.,
p. 424-426. Prentice Hall, Inc., NY.

Greenwold, Douglas, and Associates.
1965. The McGraw-Hill dictionary of modern economics.
McGraw-Hill Book Co., NY.

Harman, H. H.
1967. Modern factor analysis. 3rd Edit. Univ. of
Chicago Press, IL.

174

175

Johnson, J.
1963. Econometric methods. p. 221. McGraw-Hill Book
Co., Inc., NY.

Kish, L.
1965. Survey sampling. John Wiley & Sons, Inc., NY.

Massy, W. F.
1963. Applying factor analysis to a specific marketing
problem. Proceedings of the Winter Conference of the
American Marketing Association. (December 27-28.)

Nelson, R.
1961. Retailing yule trees. American Christmas Tree
Growers Journal 5(3):10-11, 40-41.

Roth, P. L., and J. Brummel.
1971. Effect of community size on homeowners acceptance

of Christmas trees. American Christmas Tree Growers
Journal 15(1):3-5.

Rothman, J.
1968. Some considerations affecting the use of factor
analysis in market research. Journal of the Market
Research Society 10(3).

Rummel, R. J.
1970. Applied factor analysis. Northwestern Univ. Press,
Evanston, IL.

Sales Management, The Marketing Magazine.
1966. Sect. C-83, 102(12).

Salzman, L.
1968. Computerized economic analysis. McGraw-Hill Book
Co., NY.

Skok, R. A., and W. Miles.
1963. Retail Christmas tree sales in the Twin Cities
area. Minn. For. Note No. 141 Minn. Univ., School of
For., St. Paul, MN.

Sowder, A. M.
1965. Statistics for the U. S. Christmas tree industry.
Journal of Forestry 63(11):876.

Sowder, A. M.
1965. The 1964 Christmas tree data. Journal of Forestry
63(10):776-778.

Suits, D. B.
1957. Use of dummy variables in regression equations.
Journal of the American Statistical Assoc., Vol. 52,
p. 548-551.

176

Sullivan, R. J.
1963. The marketing column. American Christmas Tree
Growers Journal 7(2):9-10.

Thompson, D.
1964. Analysis of retailing potential in metropolitan
area. Univ. of Calif., Instit. of Business and Econ.
Res., Berkeley, CA.

Thurstone, L. L.

1931. Multiple factor analysis. Psychology Review 38,
p. 406—427.

Tomeck, W. G.
1963. Using zero-one variables with time series data in

regression equations. Journal of Farm Economics 45(4):
814-822.

Troxell, H. E.
1970. A Christmas tree study in the Denver metropolitan
area. American Christmas Tree Growers Journal 5(1):8-15.

U.S. Department of Agriculture, Forest Service.
1961. Production and marketing in the northwest. Ameri-
can Christmas Tree Growers Journal 5(1):9-12. Pacific
NW Region.

U.S. Department of Commerce, Bureau of the Census.
1965. Stat. abstract of the U. S. 86th Edit., 1047 p.,
illus. U.S.G.P.0.

U.S. Department of Commerce, Bureau of the Census.
1960. U. S. census of housing: city blocks. Series HC
(3)-304. U.S.G.P.O.

Wingate, J. N., and A. Corbin.
1956. Changing patterns in retailing: readings on cur-
rent trends. Richard D. Irwin, Inc., IL.

"I11'?11111111111111“