 

142
838
THS

 

A

Inca.
LIBRARIES

MICHIGAN STATE UNIVERSITY
EAST LANSING, MI 48824-1048

l\)
\j

This is to certify that the
thesis entitled

CONSUMER PREFERENCE FOR IMPORTED RICE IN NIGERIA—
PERCEIVED QUALITY DIFFERENCES OR HABIT PERSISTENCE?

presented by

HENRY OGADINMA AKAEZE

has been accepted towards fulﬁllment
of the requirements for the

MS. degree in AGRICULTURAL, FOOD, AND
RESOURCE ECONOMICS

 

 

ﬂl ”Aw

Major Progessor’s Signature

12/17/2010

 

Date

MSU is an Afﬁrmative Action/Equal Opportunity Employer

-- __u-|-l.-l-l-I-g-n-n-n-A-n-n-L-n-n-A-n-A-I—A-n-I-n-I-n-n-l-n-|--

-l—.---r---n-.--.-n-.A -A-o-n-nn.

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 KzlProj/Acc8PresIClRC/Dale0ue.indd

CONSUMER PREFERENCE FOR IMPORTED RICE IN NIGERIA—PERCEIVED
QUALITY DIFFERENCES OR HABIT PERSISTENCE?

By

HENRY OGADINMA AKAEZE

A THESIS
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
MASTER OF SCIENCE

Agricultural, Food and Resource Economics

2010

ABSTRACT

CONSUMER PREFERENCE FOR IMPORTED RICE IN NIGERIA—PERCENED
QUALITY DIFFERENCES OR HABIT PERSISTENCE?

By

Henry Ogadinma Akaeze

Previous studies on rice in Nigeria have inter-alia focused on the constraints to
local production and consumption, and the reasons for the growth in imported rice
consumption. However, these studies have largely neglected the role of habit persistence
and perceived quality differentials in explaining consumer preference for imported rice in
Nigeria. Therefore, the goal of this research is to develop a theoretically consistent
procedure for identifying and estimating the relative contributions of habit persistence
and perceived quality differentials in explaining consumer preference for particular
product varieties, and applies this procedure to demand for imported and domestically
produced rice in Nigeria. Using annual time series data from 1961 to 2006, results
indicate that habit persistence and perceived quality differences both play an important
role in explaining consumer preference for imported rice in Nigeria. An important
implication is that policies designed to encourage production of relatively high quality
local rice, thereby replacing imported rice in consumption baskets, will face considerable
inertia due to the persistence of consumer habits and mindset regarding purchase and
consumption of imported rice, even if the locally produced rice is of comparable quality.
Companion policies designed to shift consumer-buying habits and alter already
established cultural mindset via advertising and promotion programs may be required to

overcome this consumption inertia.

Cepyright by
HENRY OGADINMA AKAEZE
2010

To my mother and ‘father’— Mrs. Victoria J. A. Akaeze—who gave up life’s comfort

for my education.

iv

ACKNOWLEDGEMENTS

First and foremost, I would like to acknowledge my Creator and my God, the
Owner of my life for the gift of His Son Jesus Christ, who gave His life for me. Also, I
am highly indebted to my major professor, Dr. Robert Myers for the professorial prowess
he poured into this work and his guidance and support throughout my Masters degree
program. Dr. Scott Loveridge and Debbie Conway deserve my profound appreciation for
their invaluable advice and inputs starting from the ﬁrst time I arrived the department. I
am also grateful to Dr. Soji Adelaja, a goal-oriented, result-driven, wise and trusted
preceptor for his interest, concerns and assistance in my career.

Dr. Duncan Boughton was very helpful in helping me conceptualize the study at
the early stage. My profound gratitude also goes to Drs. Eric Crawford and Songqing J in
for serving as my committee members. Other faculty members and friends that
inﬂuenced my life in no small way include Drs. Scott Swinton, Mywish Maredia, Zengfei
Guan and Yohannes Haliu. I also wish to specially acknowledge the University of
Nigeria Nsukka, Enugu State, Nigeria for providing me with some funding for my
Masters program and the former Vice Chancellor, Prof. Ositadinma Chinedu Nebo, an
intrepid administrator, for his quest for and recognition of academic excellence leading to
this scholarship.

Words are not enough to express my gratitude and appreciation to my mother,
Mrs. Victoria J. A. Akaeze. She is my mother and father. Thank you, the she-he that
withstood all oppositions to greatness. Thank you, the rejected stone that ﬁnally became

the head of the corner. Thank you, the mother of mothers and the mother of all (N ne

madunchal). Thank you! I am grateful you did not give up hope but single-handedly dug
into the heart of the earth in search of water and nutrients for your abandoned children.
Today, your hard works speak for you. Nem, Chukwu goziei (My mother, God bless you)!
As long as I live, I will show you how grateful I am for this immeasurable sacriﬁce.

Many thanks to my Christian mentor, Pastor Nonso Echiemunor, and my
siblings—Queen Akaeze, Kidochukwu Akaeze, Emeka Akaeze and Osy Akaeze—for
their ﬁnancial support, prayers and encouragement. I would also want to thank my wife
and the queen of my life—Onyinye Hope Akaeze—for her “statistical ﬁngers,” unending
companionship and the insightful and constructive criticisms and comments she poured
into this work.

My stay in Michigan State University has been very pleasurable due to the
friendships with, unwavering support and inspirations I received from Annes Schneller of
the Ofﬁce of International Studies and Programs, Mary M. Gebbia-Portice, Coordinator
International Sponsored Student Programs, and other graduate students such as Nicole
Mason and Ramziath Adjao. Finally, I appreciate you, the reader, for your interest in this

research.

vi

TABLE OF CONTENTS

LIST OF TABLES ...................................................................................... ix
LIST OF FIGURES. . .......................................................................................................... x
CHAPTER 1: INTRODUCTION ................................................................... 1
1.1 Background information ................................................................ 1
CHAPTER 2: LITERATURE REVIEW ............................................................ 5
CHAPTER 3: METHODOLOGY AND DATA .................................................. 8
3.1 Methodology ............................................................................ 8
3.1.1 Effects of perceived quality differences on consumer preferences....8
3.1.2 Incorporating habit persistence ............................................. 9
3.2 Functional forms ....................................................................... 11
3.2.1 The simple linear speciﬁcation (SLS) .................................... 11
3.2.2 The linear expenditure system (LES) .................................... 12
3.2.3 The almost ideal demand model (AI) .................................... 12
3.2.4 The log-linear speciﬁcation (LSS) ....................................... 13
3.3 Estimation .............................................................................. 14
3.3.1 Endogeneity and instrumental variable choice...........................14
3.4 Data ...................................................................................... 16
CHAPTER 4: RESULTS. ................................................................................................ 18
4.1 Results of endogeneity tests of local rice consumption and price of imported
rice ........................................................................................ 18
4.2 Major results of the study ............................................................ 21
4.2.1 Conventional determinants of demand ..... - ............................. 21
4.2.2 Effects of habit persistence and perceived quality differentials on
imported rice consumption in Nigeria ................................... 23
4.2.3 Elasticity estimates in the conventional and generalized demand
models ........................................................................ 25
4. 3 Tracing the effects of habit persistence and perceived quality differentials on
imported rice consumption pattern over time. .. . ...28
4. 4 Implications of the roles of perceived quality differentials and habit
persistence on consumer preference for imported rice in Nigeria... .........30
4.5 Sensitivity of accounting for endogeneity and of habit persistence and
perceived quality differentials to alternative model speciﬁcations ............ 31
4.5.1 Sensitivity of accounting for endogeneity of local rice consumption
to alternative model speciﬁcations ....................................... 31

vii

 

4.5.2 Sensitivity of habit persistence and perceived quality differentials to

alternative model speciﬁcations ......................................... 32

CHAPTER 5. CONCLUSION AND POLICY IMPLICATIONS ........................... 36

APPENDIX ............................................................................................. 4O
Elasticity formulas for alternative model speciﬁcations in the conventional and

generalized demand models. . . . . . . .. ....................................................................... 40

1. Conventional model speciﬁcations ........................................... 40

II. Generalized model speciﬁcations .............................................. 44

REFERENCES ........................................................................................ 48

viii

 

LIST OF TABLES

Table 1. Rice production, consumption and imports in Nigeria (‘000 MT), 2000-2008 ..... 1
Table 2. Summary statistics for the dependent and explanatory variables .................. 17

Table 3. Second stage result of the robust test for endogeneity of price of imported
rice ..................................................................................................... 19

Table 4. Second stage result of the robust test for endogeneity of local rice
consumption .......................................................................................... 19

Table 5. Results of the robust test for the validity of the instrumental variable
choice...... ......................................................................................................................... 21

Table 6. Conventional speciﬁcation of demand in the log-linear demand model... ........22

Table 7. Results of non-linear least square estimation of the endogeneity-adjusted log-
linear model...... ............................................................................................................... 24

Table 8. Marshallian elasticity estimates from the endogeneity-adjusted conventional and
generalized demand models ........................................................................ 28

Table 9. Effects of endogeneity on parameter estimates ....................................... 34

Table 10. Sensitivity of habit persistence and perceived quality differentials to alternative
model speciﬁcations ................................................................................ 35

ix

‘t.f AT‘

LIST OF FIGURES

Figure l: Milled rice imports (‘000 MT), 2000-2008 ............................................ 3
Figure 2: Trajectory effects of habit persistence and quality differential on imported rice

consumption, 1961-2006 ........................................................................... 29
Figure 3: Actual per capita consumption of imported rice ...................................... 29

 

CHAPTER 1: INTRODUCTION

1.1 Background information

Rice is one of the world’s most important food crops and serves as a staple food
for a large percentage of the world’s population, especially in India, China, other parts of
Asia, and Africa. In Nigeria, rice is a vital food consumption staple but has also become
an important cash crop where it provides employment for more than 80% of the
population in the major producing areas (Okoruwa and Ogundele, 2006). Ayinde et al.
(2009), drawing on WARDA (1996), note that Nigeria is both the largest producer and
consumer of rice in the West African sub-region. Moreover, Nigeria consumes
considerably more rice than it produces (Business Day, 2009), leading to signiﬁcant

imports in recent years (Table 1).

Table 1: Rice production, consumption and imports in Nigeria (‘000 MT), 2000-
2008
Rice production Rice consumption Rice Imports

 

Ye” (Milled rice) (Mined rice) (Milled rice) RI“ “pm“
2000 40 55 15 0
2001 50 70 20 0
2002 54 . 75 21 0
2003 52 152 100 0
2004 37 148 136 0
2005 51 256 205 0
2006 51 186 135 0
2007 46 206 160 0
2008 51 241 190 0

 

Source: Index Mundi (2009), adapted from the United States Department of Agriculture
http://www.indexmundi.com/agriculture/7country=ng&commodity=milled-

rice&graph=ty-exports.

‘- 1' .-

 

Over the years, several government programs have attempted to stimulate
domestic rice production with the goal of addressing the increasing demand-supply gap
and making Nigeria more self sufﬁcient in rice. Some of these policies and programs
include the Federal Rice Research Station (F RRS), established in 1970; National
Accelerated Food Production Project (N AF PP), established in 1972; the National Cereals
Research Institute (N CR1), launched in 1974; World Bank-Assisted Development
Programs, set up in 1975; Operation Feed the Nation (OFN), started in 1976; the River
Basin Development Authorities (RBDs), established in 1977; and Abakaliki Rice Project
(ARP), established in 1978. Others include the Back to Land Program (BLP) and the
Directorate for Food, Roads and Rural Inﬁastructures (DFRRI), both introduced in 1988;
and the National Land Development Authority (NALDA), dating from 1995 (Imolehin
and Wada, 2000; Emodi and Madukwe, 2008). Two of the most recent programs are the
Presidential Initiative on Rice (PIR), established in 1999 and the National Program for
Food Security (NPF S), the ﬁrst phase of which was launched in 2001.

Despite these policies, projects and programs, domestic rice consumption has
continued to outpace domestic production leading to an ever-increasing role for rice
imports. As can be seen from Figure 1, rice imports have been growing steadily in
Nigeria and this growth is expected to continue due to increasing demand resulting from
growth in incomes, urbanization, and the associated expansion of fast food restaurants

(Daramola, 2005).

 

 

 

 

 

 

 

 

 

Figure 1: Milled rice imports (000' MT), 2000-2008
w 250
E 200
D.
a /\
'3 150 /
.2
.3: 100
O
.42“ 50
E M
g 0
9’ 2000 2001 2002 2003 2004 2005 2006 2007 2008
Year

 

 

 

Considerable previous research has focused on identifying the constraints to
increasing domestic rice production in Nigeria (Imolehin and Wada, 2000; Ekeleme et al.
2009; IRRI, 2008). Other studies have focused on explaining the growth of rice
consumption. Erenstein et al. (2003) note that the persistent increase in per capita rice
consumption in Nigeria means that rice has become a structural component in the
Nigerian diet and that changes in relative commodity prices have slowed but not reversed
the increase in rice consumption. In addition to rapid population growth, Erenstein et al.
(2003) conclude that consumption increases were due to increased consumer incomes and
changes in tastes and preferences.

This thesis centers on a neglected aspect of rice consumption in Nigeria that may
help explain the importance of rice imports in the Nigerian diet. Consumers in Nigeria
appear to display a distinct preference for imported rice over domestic rice, which has led

to a market price premium for imported rice. There are two major possible explanations

3

__J

for this preference. One is that imported rice is viewed as a better quality than locally
produced rice, and that therefore domestic and imported rice are not perfect substitutes.
The second explanation is that the long history of consuming imported rice in Nigeria has
led to habit persistence and consumption inertia, which makes it more difﬁcult for locally
produced rice to compete.

The relative contributions of these two competing explanations is important
because if perceived quality differentials explain most of the preference for imported rice,
this suggests policies will need to focus not only on increasing domestic rice production
but also on improving the quality of domestic rice if desired reductions in imports are to
be achieved. If, on the other hand, domestic and imported rice are perceived as being of
comparable quality, and the preference for imported rice is a result of habit persistence,
this suggests policies that overcome inertia and change consumer-buying habits are
needed to increase the demand for domestically produced rice. Both policy strategies will
be needed if both explanations are supported by empirical evidence from this study.

Speciﬁcally, the goal of this research is to develop a procedure for identifying and
estimating the relative contribution of perceived quality differentials and habit persistence
in explaining consumer preference for particular product varieties, and to apply this
procedure to the demand for imported and domestically produced rice in Nigeria

The organization of the remaining parts of the thesis is as follows. Chapter 2
brieﬂy reviews previous related literature and motivates the contribution of the research
reported here. Next, Chapter 3 presents the methodology, data and estimation strategies
used in the analysis. This is followed by discussion of results in Chapter 4. The study

ends with conclusions and policy implications in Chapter 5.

CHAPTER 2: LITERATURE REVIEW

Over the years, rice in Nigeria has received enormous research attention. For
instance, Akpokodje et al. (2003) assess the major supply and demand trends in rice and
ﬁnd that lowland rain-fed rice systems have a higher proﬁtability than upland rice.
Erenstein et al. (2003) argue that locally produced rice has the potential to meet food
(especially rice) demand of consumers in Nigeria if efficient production practices are
employed. Erenstein et al. (2003) also ﬁnd that the price of imported rice, which is a
ftmction of world market price, import duties, transport costs, and a quality premium,
puts a cap on the price of local produce.

Lancon et al. (2003) conduct a survey of imported rice consumer’s preferences
and suggest that imported rice cleanliness is the overwhelming technical feature
explaining the expansion of imported rice consumption in Nigeria at the cost of local rice
market development. Next to cleanliness are swelling capacity (mostly preferred by
restaurants and fast food joints), taste, availability and grain shape. Other rice studies in
Nigeria include research on rice processing (Lancon et al. 2003), improved technologies
(Imolehin and Wada, 2000), market supply response and demand for local rice (Rahji and
Adewumi, 2008), and rice irrigation (Kebbeh et al. 2003).

Moreover, several studies using stochastic frontier production functions, meta-
frontier models, and other production function models have also been used to test for
technical efﬁciency in rice production in Nigeria (See Tijani, 2006; Okoruwa and

Ogundele, 2006; Ayinde et a1. 2009; Shehu et al. 2007; Moses and Adebayo, 2007).

However, none of these studies has focused on consumer preference and the possible
combined effects of habit persistence and perceived quality differences.

In the general demand literature, there are multiple studies that address either
habit persistence or preference for quality differences in isolation. For example, Dynan
(2000) tests the time separability of preferences (presence of habit formation in consumer
preferences) with annual panel data from the US. Income Dynamics study and ﬁnds no
evidence of habit formation. On the other hand, Ferson and Constantinides (1991) ﬁnd
evidence of habit persistence using US. monthly, quarterly and annual consumption data
from the Center for Research in Security Prices of the University of Chicago and the
DRI.

Using the US household level BLS Interview Panel data to test the effects of
habit persistence, Heien and Durham (1991) ﬁnd that habit effects are highly signiﬁcant
but much smaller in cross-sectional data than in time series data. Other studies focus on
the implications of habit formation in consumption for monetary-policy models (Fuhrer,
2000), a consumption-based explanation of aggregate stock market behavior and habit
(Campbell and Cochrane, 1999), and understanding asset prices under habit formation
(Abel, 1990).

Researchers have also done some interesting work on quality variation and its role
in consumption behavior. For instance, Nelson (1991) rejects the theoretical arbitrariness
of the simple sum of physical quantities used as a measure of demand in the ‘quality’
literature, and investigates alternative measures of demand derived from restrictions on

quality variation, consumer preference, or relative prices. Also, Cox and Wohlgenant

(1986) distinguish quality effects from supply-related price variability to identify cross-
sectional demand for disaggregated food commodities.

However, none of these studies on habit and quality has focused on Nigeria, the
highest consumer of rice in the West African sub-region. The contribution of the current
study, therefore, is that it provides a theoretically consistent way of combining habit
persistence and preferences for quality differentiated goods in a single consumption
model, and applies this framework to investigate preferences for imported rice over
locally-produced rice in Nigeria. The resulting empirical model can identify the extent to
which consumer demand for imported rice in Nigeria is due to perceived quality
differentials versus habit persistence. The generic model developed here can be extended
to study the relative importance of habit persistence and perceived quality differences in
an entire system of demand equations for both agricultural and non-agricultural quality-

differentiated commodities.

 

CHAPTER 3: METHODOLOGY AND DATA

3.1 Methodology
3.1.1 Effects of perceived quality differences on consumer preferences

Let q] and q2 be the amount consumed of different varieties of the commodity of
interest and x be a vector of other goods consumed. In our case q] and q2 are imported
rice and domestically produced rice, respectively. Extension to more than two varieties is
straightforward. Assume that utility takes the form U(q1 + aqz, x) where a is a measure
of the perceived quality differentials between ql and q2. If a = 1, the varieties are viewed
as perfect substitutes, and therefore of equal quality. If a = 0 , then q2 is viewed to be of
such poor quality that its consumption does not generate utility. If a 6 (0,1) any given
amount of (12 generates less utility than the same amount of q]. Hence,a can be viewed
as an index of the perceived quality differential between q] and q2 .‘

The consumer’s problem is to:

(1) max {UICII +042,x)=P1(11 + P2q2 +WX S y}
qI’qZ’x

where p,- is the price of qi, W is a vector of prices for x, and y is consumer income.
Assuming both q] and q2 are consumed then necessary conditions for solving (1)

include:

(2) Uq: -/ip1 =0 and ath —/ip2 =0

 

' Any two varieties can obviously be reordered if necessary to ensure that a 6 [0,1].

8

—'u-'!llll'_‘V“

 

where Uq. is the derivative of U with respect to q* = q1 +aq2, and/i is the Lagrange

multiplier (marginal utility of income). Combining and rearranging (2) gives:

(3) P2 = 0P1

which shows that in order for both varieties to be consumed, their prices must be in ﬁxed
proportion given by the quality differential index. Using (3) and the deﬁnition of q* the

consumer’s problem can therefore be re-written as:

(4) maxi/(gag: plqat +wx S y}
(I ,X

which shows that q' can be viewed as a composite commodity whose price is p1 and

whose demand function satisﬁes all of the characteristic properties of conventional

consumer demands.

3.1.2 Incorporating habit persistence
So far the model incorporates quality differences between varieties but not habit

persistence. To model habit persistence, a time index is added and so (4) is re-speciﬁed

as:2

(5) max {U(q; — .3411]th )3 P114: + tht 5 Yr}

4: ,x,
where ,6 is a measure of habit persistence in the consumption of q; . If ,6 = 0 , there is no

habit persistence but as ,6 —->1 habit persistence plays an increasingly dominant role in

consumption behavior. For simplicity, habit persistence has only been modeled in the

 

2 This speciﬁcation restricts the degree of habit persistence to be identical across q] and q2 , which seems
like a reasonable assumption when looking at different varieties of the same product

9

consumption of q: but extending the model to allow habit persistence in x as well is
straightforward. This basic approach to modeling habit persistence has been used
previously in the literature (See Dynan, 2000; Naik and Moore, 1996; Fuhrer, 2000).

The budget constraint in (5) can be re-written as:

t t t

(6) P114: — ﬂpltqt—I + tht 5 Yr — ﬂPIt‘It-I-
Then deﬁning 5, = q: — ,Bqu and 32', = y, -,6p1,q:_1, the consumer’s problem (5) can be
re-written as:

(7) max{U(§’t, Xt)3P1t§t + tht 5 3’1}

‘1: ,x,
which shows that a, can be viewed as a composite commodity whose demand function

satisﬁes all of the characteristic properties of conventional consumer demands given

prices (p1, ,wt) and income it . This implies that any theoretically appropriate ﬁinctional
form can be used to specify and estimate the demand for a}. That is, for any theoretically

consistent demand function 5, = d(p1,,wt,'jr’,) an appropriate model for q], would be

(using the deﬁnitions of 5', and q: ):3

(3) ‘11: = 5(411—1 + 0“121—1)-(Z€12: + ‘1 IP11,“ ,yt - ﬂplz (411—1 + 0“121—1 )1

Note that if a = ,6 = 0 , (no habit persistence but perceived quality dominance of
ql) then consumption of qz would be zero and (8) reduces to a standard demand

equation in q]. If ,6 = 0 and a =1 there is no habit persistence and no perceived quality

 

3 Here I focus on demand for the commodity of interest q but the approach is clearly applicable to the entire
system of demand equation for both q and x .

10

differentials, so the demand for both varieties q] and q2 can be aggregated. If ﬂ = 0 and
a 6 (0,1) there is a perceived quality differential but no habit persistence. If both
a 6 (0,1) and ,8 6 (0,1) then there are both habit persistence and perceived quality

differentials with the relative magnitudes of these two parameters indicating the relative

importance of the two effects.

Given a speciﬁcation for d 0, equation (8) can be estimated to provide inferences
on a and ,6. Of course, estimation of (8) may be complicated by nonlinearity and the
likely endogeneity of q2, . However, estimation strategies will be discussed in more detail

below.

3.2 Functional forms
In this section I provide details of the model for four possible choices of d () The

ﬁrst example is based on a simple linear speciﬁcation, followed by a linear expenditure
system (LES), an almost ideal demand model speciﬁcation (AI), and a log-linear model

speciﬁcation (LL).

3. 2.1 The simple linear specification

A simple linear speciﬁcation follows the form:
N n N
qt = 7+5qplt + Zlgjwjt +5)?
I:
where n is the dimension of x (and w ). The corresponding version of (8) is then:

It
(9) 611: = 7 + ﬂIqlz—l + aq2t—l )- 05612: + 5qP11 + Z1 51"” jt + é‘I)’: '- ﬂP1:(qlt-l + 0421-1)]
J:

11

which can be estimated using procedures discussed below. Note that if a = ,6 = 0 then
(9) reduces to a standard linear demand equation for q] , and if a = 1 and ,6 = 0 then (9) is

a single equation representing aggregate demand for q] and q2 (which are perceived to

be of comparable quality).

3.2.2 The linear expenditure system (LES)

The second example is based on the linear expenditure system (LES) which has
been used extensively to investigate habit persistence in consumption choice (see Pollak

and Wales, 1992; Welsch, 1989). The LES model for 5', is:

' n 1
yr ‘7’qplt — .2 l’jot
51:78, +6 1.1
Pit

 

 

l. ..

The corresponding version of (8) is then:

(10) cm = r, +/3(qu_1 +aq2._1)—aq2.

5 n
+ — J’t ' 76PM (mt-1 + aq2t—1)- 7th‘ 2.17jo:
t 1::-

which can be estimated using procedures discussed below. Note again that if a = ,8 = 0

then (10) reduces to a standard LES equation for q].

3. 2.3 The almost ideal demand model (AI)

The third example is based on the almost ideal demand model (Al). The AI model
for 27, is

5: = (%t )[7 + 6}, 111(Plt)+ :16] ln(wj, )+ 9: 111(3)} 13)]

12

where P, is the conventional AI price index deﬁned over (p1,,wt ). The corresponding

version of (8) is then:

M - 16PM (th—I + aq2t-1)]

(11: =7 +.3(qll-l +0612t—1l—aqzt +[ m
t

(11)

 

x {7 + 5‘] 1n(Plt)+ I: 61- ln(wj, )+ «f In[yt - ﬂplt(qlé-l + 04121-1 )]}

which can be estimated using procedures discussed below. Again, if a = ,B = 0 then (11)

reduces to a standard AI model for q] .

3. 2.4 The log-linear specification (LLS)

The log-linear model for 5, is
n
J:
. . . . ~ III III N it :1:
Again. usmg the deﬁmtlons of q: = q: - .Bqt-l . yl = y: -ﬂpuqt—1. and q = 611 +042 the
corresponding version of (8) is
(12) (11: =ﬂ(qll-1+a12t—1)-aq21

n
+exp[7+6q lnplt + 2151- 11'1th +6114}: ‘ﬂPlt(QIt—l mam—1”]
J:

If 0: =1 and ,6 = O ‘, and if the natural logarithm of (12) is taken, then (12) reduces to

 

’ Notice that if a = ,6 = O , then (12) reduces to a standard log-linear demand for imported rice. That is,

n
Inqu=7+6q lnpl, + :15]- lnwj,+§lr(y,)
J:

13

n
(13) Inq: =7+6qlnpu + :1ch Inwj, +§lr(y,)
J:

where q: = q], +q2,. Equation (13) is the standard log-linear demand equation for q: ,

which is the total demand for rice aggregated across imported and locally produced rice.

3.3 Estimation

The quantity of imported rice consumed in Nigeria was used as the dependent
variable while the price of imported rice, prices of other competing commodities, proxied
by the Consumer Price Index (CPI), one-period lags of the quantities of domestically
produced and imported rice, quantity of domestically produced rice, and income are the
explanatory variables. Equations (9), (10), (11) and (12) were estimated with nonlinear

least squares using instrumental variables for the endogeneous q2, .5

3.3.1 Endogeneity and instrumental variable choice

In this study, local rice consumption qz, is hypothesized to be endogenous
because quantities of locally produced and imported rice are likely to be jointly
determined. In demand studies using aggregate data, as are used here, price variables
(here price of imported rice, 171,) may also be endogenous because there may be feedback
from aggregate consumption choices to equilibrium prices. Other variables in the model

(income and lagged quantities) can more reasonably be assumed exogenous. If

endogeneity is found then the nonlinear least squares estimators in equations (9) to (12)

 

5 Even (9) is nonlinear in parameters because of nonlinear parametric restrictions.

14

will be biased and inconsistent and an instrumental variables (IV) approach will be
required to correct for endogeneity.

To illustrate the instrumental variable approach, consider a structural equation:
(14) yl =60 +§1y2 +§221+u1
Assume z] is uncorrelated with u] and so exogenous but it is suspected that y2 is
correlated with u]. Wooldridge (2009) recommends the use of other exogenous
variable(s) uncorrelated with u] that do not explicitly appear in the structural equation, as

well as possibly 2] , as instruments. Deﬁne a vector of these external variables asz . The

basic assumptions here are that z] and z are both uncorrelated with u] , and that u] has a
mean of zero. That is,
E(u1) = 0, Cov(zl,u1) = 0 and Cov(z,u1) = 0 .

Instrumental variables need to satisfy the exogeneity condition and the relevance
condition. The exogeneity condition requires that the instrumental variable must be
uncorrelated with the error term in the demand model, while the relevance condition
requires that the instrumental variable must exibit some partial correlation with the

endogenous explanatory variable. The latter condition is testable by doing an F test on 1
in the following equation and testing p = 0 (Wooldridge 2009),

y2 = $0 +¢121+ZP+V2-
If p ¢ 0, then the chosen instrumental variables (IVs) are correlated with local rice
consumption and thus meet the relevance condition. Results of this test including the test

for endogeneity of p1, are presented in section 4.1.

15

3.4 Data

Aggregate annual time series data from 1961-2006 were collected on the quantity
of domestically produced rice consumed in Nigeria, imported rice consumed in Nigeria,
price of imported rice, prices of substitute and complementary goods as measured by the
consumer price index (CPI), Gross Domestic Product (GDP) as a measure of consumer
income, and the Nigerian population. Data on locally produced rice consumption,
imported rice consumption, and the CIF price of imported rice were obtained from the
International Rice Research Institute (IRRI), while the GDP and population data were
obtained from the World Development Indicators, 2007 (World Bank, 2010). Nigerian
CPI data were obtained from NationMaster.com.

In order to control the effects of population on aggregate demand, consumption of
local and imported rice as well as GDP were put in per capita terms by dividing by
population. Summary statistics of all the variables used in this analysis, including various
transformations are presented in Table 2. Consumption of rice imports has a low
minimum value of 0.005 because it was put in per capita terms. The AI price index
variable was constructed as the price of imported rice and prices of other competing

staples weighted by their shares of expenditure in income per capita.

16

Table 2: Summary statistics for dependent and explanatory variables

 

 

n = 46
Variable Mean Std. Dev. Min Max
ImP°“"d “cc per “31"“ 3602.699 3369.716 4.280 12541.45
(kg/person)
Consumer price index 28.759 52.445 0.240 207.398
GDP per capita (10,000 USD) 34.493 20.845 10.161 90.343
Log of CPI 1.186 2.285 -1.427 5.335
Log of consumers’ income
(10,000USD) 3.362 0.617 2.319 4.504
Price of imported rice
(1000$/ton) 0.326 0.139 0.109 0.708
Log of price of imported rice _
(millions USD) -1.195 0.419 2.228 0.345
L°g 0f quantity 0f “”0”“ -0 187 2 596 -5 356 2 627
rice per capita (tons/person) ' ' ' '
Price index for AI model 8.637 17.351 0.234 78.701

 

l7

CHAPTER 4: RESULTS

The ﬁrst part of this chapter presents results of the endogeneity test for local rice
consumption and price of imported rice, and for the test for the validity of the
instrumental variable choice. This is then followed by presentation of the major results of

the study.

4.1 Results of endogeneity tests on local rice consumption and price of imported rice
Following Wooldridge (2009), the test for endogeneity proceeded in two stages.
First, the hypothesized endogenous explanatory variable was regressed against all the

exogenous variables in the demand model, including the instrumental variables, to obtain

the IV residual, 13,. Second, this residual was then added to the original model including

the endogenous explanatory variable and a heteroskedasticity-robust t test of the
coefﬁcient on 13 was used to determine the endogeneity of the variable in question.
Results presented in Table 3 show no evidence of endogeneity of price of
imported rice. However results in Table 4 show that the null hypothesis of no
endogeneity in the quantity of locally produced rice is strongly rejected, implying that
local rice consumption is endogenous and NLS estimators of equations (9), (10), (l l) and

(12) would be biased and inconsistent, requiring an instrumental variable for q2,.

18

Table 3: Second stage results of the robust test for endogeneity of price of imported

 

 

 

rice
n = 46

Variables Coef. P-valuc
Constant 7.506 0.007
One-period lag of imported rice consumption, q1,_1 0.511 0.006
One-period lag of local rice consumption, q2,_] 0.023 0.502
Consumer price index, wj, 0,057 0.039
Consumers’ income per capita, y, 0.074 0.094
Price of imported rice, p], -2.966 0.518
Local rice consumption, ‘12: -O.6l6 0.037
Residual, 13 -3.495 0.754
Adj. R-squared 0.817

 

Table 4: Second stage results of the robust test for endogeneity of local rice

consumption

 

 

 

n = 46

Variables Coef. P-value
Constant -2.485 0.020
One-period lag of imported rice consumption, q1,_1 0.508 0.002
One-period lag of local rice consumption, q2,_1 -0.846 0.006
Consumer price index, W], -0010 0234
Consumers’ income per capita, y, 0.004 0.857
Price of imported rice, p1, 1.223 0.696
Local rice consumption, q2, 1.070 0.003
Residual, 1'3 -0.945 0.011
Adj. R-squared 0.793

 

The instrumental variable regression for testing endogeneity of price of imported

rice was speciﬁed as levels and squares of the exogenous variables. For the endogeneity

l9

test of local rice consumption, the instrumental variable regression was originally
speciﬁed as levels, cross products and squares of the model independent variables, but
due to model over-parameterization, there was need for excluding some of these
variables. Therefore, on the basis of empirical testing and consistency with demand
theory, squares of Consumer Price Index and the price of imported rice were the
instrumental variables used in the later test. The choice of squares of the IVs was further

supported by the nonlinear nature of the estimated models.

These IVs were hypothesized to be correlated with q2, but not with the error term
in the demand equation. This conjecture is supported by the validity test in Table 5.

Formally, an OLS regression of th on all the explanatory variables including the W3
was done, and the validity of the IVs was determined using an F test.

As Table 5 clearly shows, the coefﬁcient on squared CPI is statistically different
from zero. Though the square of the price of imported rice appear to be statistically
insigniﬁcant, results of an F test for joint signiﬁcance show that F(2, 40) = 3.232 with a
computed F value of 21. Thess results provide evidence of partial correlation between
the IV ’s and local rice consumption, indicating that the squares of CPI and price of

imported rice are valid IVs.

20

Table 5: Results of the robust test for the validity of the instrumental variable choice

 

 

Variables Coef. P-value
Constant 2.090 0.467
One-period lag of imported lice consumption, qu -0.002 0.994
One-period lag of local rice consumption, q2,_1 0.814 0.000
Consumer price index, wj, 0,063 0.056
Consumers’ income per capita, y, 0.066 0.251
Price of imported rice, p], -5.235 0.760
Squared Consumer Price Index, w 172 -0.0003 0.097
Squared price of imported rice, 151,2 1.385 0.952

 

4.2 Major results of the study

The generic model:

qlt = 3(411—1 + aq2l-1)_ “421 + dLPItaWt 9yt — .3171: (qlt-l + ant—I )]
was estimated using all the model speciﬁcations (9) to (12) as a possible choice for d().
However, results for the log-linear model are reported here because this speciﬁcation is
simple to estimate, easy to interpret, and provides results that are more consistent with
demand theory and a better ﬁt than other model alternatives.

In reporting the results particular attention is paid to the effects of habit
persistence and perceived quality differences on rice consumption in Nigeria by testing

parametric restrictions on a and ﬂ.

4.2.1 Conventional determinants of demand
The static demand models presented in Table 6 assume no effects of habit

persistence and perceived quality differentials, that is, the restrictions a =1 and ,6 = 0 are

21

imposed. Results are consistent with a priori expectations concerning the signs of the
determinants of a conventional demand model with income and price of other goods
having positive effects on the aggregate quantity of rice consumed in Nigeria while price
has a negative effect. All the determinants of demand are statistically signiﬁcant at the
1% level.

The coefﬁcients of the log-linear model have a direct interpretation as elasticities.
Thus, results indicate that a 1% increase in own-price decreases rice consumption by
0.23%. Conversely, as the CPI and consurners’ income increase by 1%, aggregate rice

consumption increases by 0.08% and 0.33%, respectively, ceteris paribus. This indicates

 

that rice is both a substitute for other consumption goods and a normal good in Nigeria.

Table 6: Conventional specification of demand in the log-linear demand model

 

 

Parameters Symbol Coef. P-value
Constant 7 l .385 0.000
Own-price 6., -0.231 0.004
Cross price 6,- 0079 0.000
Income 5 0.331 0.000
n 46

R-squared 0.821

Lung Box Q. Stat. 0.000

 

The goodness of ﬁt is reasonably high, indicating that about 82% of the variation
in the aggregate demand for rice in Nigeria is explained by the model. The Ljung Box
statistic shows evidence of autocorrelation in the residuals, but this will be discussed in

the next section.

22 ‘

Results ﬁom estimating this conventional demand model would suggest own
price, incomes and changes in the general price level are the main determinants of rice
consumption. However, this model does not account explicitly for perceived quality
differentials and habit persistence and so may misrepresent actual determinants of the

growth of imported rice consumption.

4.2.2 Eﬂ'ects of habit persistence and perceived quality ditferentials on imported rice
consumption in Nigeria

Table 7 presents results of the endogeneity-adjusted log-linear demand model
with perceived quality differentials and habit persistence. The effects of both quality
differentials and habit persistence are important and statistically signiﬁcant. With habit
persistence and perceived quality differentials, the conventional determinants of demand
still maintain their expected signs. Unlike in the conventional demand model results, the
own-price parameter estimate is not statistically signiﬁcant. This indicates that there is no
signiﬁcant responsiveness of demand for imported rice to changes in price due to habit
persistence and perceived quality differences. While the positive income effect is
statistically different from zero below the 1% level, the cross price effects are statistically
signiﬁcant at the 10% level.

It is interesting to note that own-price and cross-price parameter estimates are
approximately similar in the generalized and conventional demand results, while income
effects increased from 0.331 to 0.585 in the generalized model. This implies that the

importance of income growth as a partial explanation for the expansion of demand for

23

..
. ,

ﬂ ......ssﬂuﬂr

 

rice increases when perceived quality differentials and habit persistence are incorporated
into the demand model."

The signiﬁcance of habit persistence and quality differentials indicates that they
both play an important role in the demand for imported rice in Nigeria. For the perceived
quality differential parameter, the estimate of 0.18 indicates an extremely large effect of

consumers’

Table 7: Results of non-linear least square estimation of endogeneity-adjusted log-

Iinear model

 

 

Variable Symbol Coef. P-value

Constant 7 -1.178 0.166
gﬁfn‘t’ge’c‘m" “all” a 0.176 0.054
Effects of habit persistence ,3 0.491 0.002
Own-price 5., -0.243 0.354
Cross-price 5} 0.077 0.089
Income 5 0.585 0.001
n 46
R—squared 0.887

_Liu_ng Box Q stat. 0-778

 

perception of differences in quality on the consumption choice for imported rice in
Nigeria (i.e., it is much closer to zero than to one). This is because a coefﬁcient of 1
indicates no perceived quality differentials and thus the two goods can be considered as
perfect substitutes while zero indicates a large perceived quality dominance of imported

rice such that the consumption of domestic rice generates no utility.

 

6 For future research, it will be interesting to understand and estimate the effects of different income levels
or income distribution on demand for imported rice in a generalized demand ﬁ'amework that incorporates
perceived quality differentials and habit persistence.

24

Further, while it is important to test if a is signiﬁcantly different ﬁ'om zero
indicating that the consumption of local rice does not generate utility, it is more
interesting to test the substitutability effects of imported and domestically produced rice
by testing whether a is signiﬁcantly different from one. Though not presented, this test
shows that F(1, 40) = 86.79 with a P-value of 0.000, indicating that the hypothesis of
perfect substitutability between imported and locally produced rice is strongly rejected
below the 1% level. This means that locally produced rice is not perceived to be of
comparable quality with imported rice and therefore, one variety cannot be easily
substituted for the other.

The magnitude of parameter estimate of habit persistence has an opposite
interpretation because as ,6—> l , habit persistence plays an increasingly dominant role in
consumption behavior. Therefore, a parameter estimate of 0.50 indicates a moderate
effect of habit persistence on consumer preference for imported rice.

Moreover, the model ﬁts well with an R-squared of 89%. The Ljung Box Q-
statistic shows that there is no evidence of autocorrelation in the residuals, unlike in the
simple static speciﬁcation. Thus the generalized models are more robust in controlling

autocorrelation in the residuals.

4.2.3 Elasticity estimates in the conventional and generalized demand models

For the log-linear speciﬁcation in (12) given as:

‘1" = 5(41t—I + anI—I )— 0612:

n
+exP[7+5q 1111911 + .216) anjt +5114.” -ﬂplt(qlt—I +ant—1)] ,
j:

25

the elasticity estimates for the conventional demand models are obtained by estimating

equation (13), which is
:1: n
J:

Thus, the coefﬁcients of p], , w, and y, are own-price ;, cross-price (j, and income

77 elasticity estimates, respectively.

In the generalized demand models, however, the coefﬁcients of own-price, cross-
price and income are not elasticity estimates with respect to imported rice consumption
but with respect to 5, , which, itself, is a function of imported rice consumption, one-
period lags of imported and local rice consumption, local rice consumption, and the
coefficients a and B.

For the own-price, cross-price and income elasticity estimates, therefore, I take
the derivative of (12) with respect to p1, , w, and y, , respectively. These are converted to
elasticities by multiplying the outcome of these derivatives by the ratios of p1, , w, and y,

to q” accordingly. Thus the formulas for the own-price, C , cross-price, (j and income,

77 elasticities are:

(15) g = _1_[5q -M]W
(111 7’

16 .=i’— " a.

( ) C] qlt [1E1 1]

(17) n = —y-'—[i]w
(In 7’

26

n
where W = 6Xp[}’+§qlnp1t+ Z djlnwj, +§ln(7r) ], 7: = y,—,Bp1,¢ and
i=1

(6 = q1,_1 + aq2,_1 . For the elasticity formulas of other endogeneity-adjusted alternative

model speciﬁcations, see Appendix.

Table 8 presents endogeneity-adjusted elasticity estimates7 for the generalized
and also the static log-linear demand models. These estimates were computed at the mean
values of data. As discussed earlier, in the conventional model all parameter estimates
have the expected signs and are statistically signiﬁcant at the 1% level. When the effects
of habit persistence and perceived quality differentials are incorporated, the own price
and cross-price elasticity estimates are not statistically different from zero, though they
maintain their expected signs. However, income elasticity is statistically signiﬁcant at the

10% level but the magnitude of this estimate increased from 0.331 to 0.560.

 

7 The result presented here and in the Tables of all the major results reported in this study met the
homogeneity restriction imposed by demand functions—Marshallian demand functions are homogeneous
of degree zero in all prices and income:

‘1—[54 ‘—§ﬂ:p“]w + Ill-I E 511* L[£]W =0,

qlt €11: j=I <11: 7’

which would imply that the sum of all prices and income elasticities of demand must equal zero. This
equation was evaluated at the mean values as a non-linear combination of parameter estimates in the
tmconventional (dynamic) demand framework. Result shows that F(1, 40) = 1.69 and Prob. > F = 0.2010,
thus we fail to reject the null hypothesis of homogeneity of degree zero in all prices and income at any
reasonable signiﬁcance level. Therefore, the log-linear demand model is homogeneous of degree zero in all
prices and income.

27

Table 8: Marshallian elasticity estimates from the endogeneity-adjusted
conventional and generalized demand models

 

 

 

Conventional Demand Generalized Demand
Parameters Model (a = 1: .5 = 0) Model
Symbol Coef. P-value Coef. P-value
Own-price 9' -0. 231 0.000 -0. 242 0.408
Cross-price g j 0.079 0.004 0.072 0.144
Income 77 0.331 0.000 0.560 0.063

 

4.3 Tracing the effects of habit persistence and perceived quality differentials on
imported rice consumption pattern over time

Figure 2 presents a graphical representation of the effects of habit persistence and
perceived quality differentials on imported rice consumption in Nigeria using simulated
data from 1961 to 2006, while Figure 3 presents the consumption pattern of imported rice
using actual data for the same time period. The simulated imported rice consumption data
were computed recursively. First, one-period lag of the variables and the parameters
estimated ﬁ'om the model were used to compute the simulated quantity of imported rice
consumption at time t = 1. Next, the simulated values at time t = 2 were computed using
the computed values from the ﬁrst time period. In the third period, computed values from
the second period were used with the model’s parameter estimates and so on.

In order to trace the joint effects of habit persistence and perceived quality
differentials on consumption, a and 6 were set to their estimated values as shown by the
upper line in Figure 2. The lower line in Figure 2 shows the effects of perceived quality
differentials on imported rice consumption. This was obtained by setting ,6 = 0 while a
was set to its estimated value. It is shown that between 1961 and 1969 the simulated per

capita consumption was close to zero while a gradual increase in the importation of

28

foreign rice due to increase in demand was observed in the 19708. In all, the quantity
consumed resulting from the effects of habit persistence and perceived quality differences

grew higher than that without habit effects.

 

Figure 2: Trajectory effects of habit persistence and quality
differential on imported rice consumption, 1961-2006

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

O

12

s n . -
g. 10 +Quantlty
. a consumed
Z. 3 WITH habit and
8 quality

5 6

5- --Quantity

'_ 4 consumed
8. {5 ill WITHOUT
‘8 2 I habit

1.: VWVJ‘V

5

.E

m

96] 962 197i 1979 19:93 1.99] 1992 202?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

-2
Year
Figure 3: Actual per capita consumption of imported rice

16

14 r

2
.5 J ctual
:- 8 “’— quantity
3 6 __.._ consumed
ﬂ
3 4 ""“ 1%»:
.é’ 2 l
a W
8 0 ,
ts SESSQERQQSSSRSSS
9.. O\O\O\O\O\O\O\O\O\O\O\O\O\OOO

'— v—n — v— v— .— '— u—: v— v-r —- v-o — N N (\I
Year

 

 

29

What is interesting enough to be highlighted is that the graph of the joint effects
of habit persistence and perceived quality differentials in Figure 2 and the graph of the
actual per capita consumption of imported rice in Figure 3 follow a similar pattern. This
implies that the joint effects of habit persistence and perceived quality differentials truly
represent the consumption pattern of imported rice in Nigeria, ceteris paribus. Therefore,
any omission of such important determining factors that explain preference for imported
rice as habit persistence and perceived quality differentials would misrepresent actual

determinants of growth of imported rice consumption in Nigeria.

4.4 Implications of the role of perceived“ quality differentials and habit persistence
on consumer preference for imported rice in Nigeria

The parameter estimate of perceived quality differentials has an interesting
consumer utility implication. For example, the 0.176 estimate implies that any given
amount of locally produced rice is viewed to be of such poor quality that its consumption
only generates about one-ﬁfth as much utility as the same amount of imported rice.
Secondly, it suggests a high degree of perceived quality differentials between foreign rice
and domestically produced rice and thus one variety cannot be easily substituted.

However, given the important effects of perceived quality differentials on
consumption preferences, another important implication can be traced to the role habit

persistence plays in explaining the growth of imported rice consumption in Nigeria. In

 

8 It is important to emphasize that the empirical notion of quality in this thesis is essentially that of
perceived quality differences. This raises the question of what forms perception about quality. To the extent
to which such perception could be explained by real quality differentials between imported and locally
produced rice, or by other factors that impact on perception such as mindset, this measure may reﬂect
something more complex.

30

other words, while it has been established that rice consumers place high importance to
quality, this study shows that consumer habit is also an important contributing factor in
explaining consumer preference for foreign rice in Nigeria.

An important implication of habit persistence is that even if locally produced rice
is of comparable quality and thus “competitive”, its consumption will face considerable
inertia due to persistence of consumer habits for imported rice. Another implication of the
role of habit persistence in the consumers’ preference for imported rice is its effects on
the local rice market. In summary, though not equally as important as the effects of
quality perception, habit persistence plays a signiﬁcant role in explaining consumer

preference for imported rice in Nigeria.

4.5 Sensitivity of accounting for endogeneity and of habit persistence and perceived

quality differentials to alternative model specifications

4. 5.] Sensitivity of controlling for the endogeneity of local rice consumption to
alternative model speciﬁcations

The purpose of this section is to discuss the effects of controlling for endogeneity
on estimation results and its sensitivity to alternative model speciﬁcations. To do this,
estimations with and without controlling for endogeneity were done. As clearly shown in
the upper section of Table 9, though habit and perceived quality differential effects are
statistically signiﬁcant, quality differential has a negative effect on imported rice
consumption in all the model speciﬁcations. Indeed, this makes no economic sense
because it contradicts a priori expectations. This is because it violates equation (3) which

would imply that local rice consumption generates negative utility.

31

Another counter-intuitive implication of a negative perceived quality effect is that
the demand for imported rice will be positively related to the demand for local rice. In
other words, an increase in the consumption of local rice will also mean an increase in the
consumption of imported rice, ceteris paribus.

The lower section of Table 9 presents the results of well—behaved generalized
demand models. What quickly stands out from the Table is the switch in sign of the
quality parameter estimate in all the demand models. Notice that except in the AI model
own-price, cross price and income effects now have the correct signs for a normal good
in addition to positive quality effects consistent with demand theory. All parameter
estimates are consistent with a priori expectation. In the log-linear model, for example,

the magnitude of quality effects (a) changed from -0.231 to 0.176, own-price ( 6") from
1.726 to -0.243, and cross-price (61-) from -0.313 to 0.077.

Moreover, estimation results were sensitive to alternative instrumental variable
choice. For example, use of levels, squares and cross products of the model exogenous
variables as instruments produced results that were inconsistent with demand theory due
to over parameterization and extremely high explanatory power of the ﬁrst stage

instrumental variable regression.

4.5.2 Sensitivity of habit persistence and perceived quality differentials to alternative
model speciﬁcations
This section presents the sensitivity of perceived quality differentials (a) and

habit persistence (,6) to alternative model speciﬁcations. As can be seen from Table 10,

the estimate of perceived quality differentials is roughly 0.10 except in the SL8 model

32

where it is 0.43 and is marginally signiﬁcant below the 5% level in the log-linear
speciﬁcation, and at the 5% level in the SLS model.

Though the perceived quality differential effect is not statistically different from
zero in the LES and the AI models, a separate test showed that it is statistically different
from one. This is presented in the lower section of Table 10, where it is shown that
perceived quality differential is statistically different from one below 1% level in all the
model speciﬁcations. This is more interesting because, as discussed earlier, it tests the
substitutability effects between imported and locally produced rice rather than just testing
perceived quality dominance of imported rice. The parameter estimates of habit
persistence are statistically different from zero below the 5% level in all the alternative

model speciﬁcations.

33

0:30.3nt oomomgqu---

 

 

 

 

 

 

 

hood ohod wood oRd dossomi ..8/w
. . . . . . . . 03:50me
Vmo o 02 o woo o ooo o of o ooo o omo o omv o ..o 3:96 o0>m00o0m
mood oovd oood nmod oood mmnd Sod m d md 0. 020000608 “50:
Sod momd oood vood --- --- omod obod M 08005
oood Rod omnd mood --- oood mood oood we 00:93.80
med mvmd- Kod oood oood wood- Sod moom- 0% Suaéso
020>i .o0o0 020>i .0000 03—min .0000 03.0fm .0000 Spam
a: a: so do
not—«0500.5 80.0% 9352— 5009mm 53005025 23082.5
E05400..— .003 .0032 Sam—ogamm— .305..— .3054 033%
ow u = $030.: nonnaoufémoﬂmeuem
oood mood Sud mood Bogota ..8<
. . - . . - . . - . . - mﬁuaou0§o
ooo o 1mm o ooo o mm o moo o mom o _mo o 2m o B 6.29:0 @0208?“
oood mnod oood and oood 30d oood mvod m. ooooummmuom “Bum
omod moo.“ mood oﬂod --- --- mid oood M 08005
mood 2nd- oood Sod --- oood oood vood No 025-380
ombd emu; _ 24d Sod- oood mood- oovd mmmm 0% 089850
020>¢ .0000 020>i .0000 020>¢ .0000 030>d .0000 325m
ANS G a 8: 5
55005025 805$ ...—080: 880% nemuaoﬁoam 0.08.2.5
0005A.w0..— .003 .0032 0.5505an 30:5 .3054 0385

 

3. u = 303:: uuxmameeafbmuaumewzm

 

0.0025000 00.0822— :0 .E05W0ou0 no 300.3— "o 03:.

34

 

 

 

805800

 

 

 

 

 

 

oood 05d oood oood oood oood oood 09d 0 5.00.0 0050000“.
mood .ovd oood hood oood mmhd ...od m . md 5. 000000.800 5.0...
00.0.5. ...000 00.0.5. .0000 00.0078. ...000 00.0.5. ...000 .0085
no no so .0
005005025 8000.5 000809 80005 0030050005 80008000..
0000.4001— .00... .0082 085000000 .800... 0000... 0.085
S u 0. .~ 1 so
. . . . . . . . 0.0008050
go d on. o 0mm o ooo o oo. o ooo o omo o ooo o 0 5.00.0 00>.0000n.
mood .ovd oood hood oood mud 0 .od m . md 5. 00005800 0.00.4.
00.88 .0000 00.0.5. ...000 00.050. ...000 00.0.5. ...000 .0085
no a .0 so 5
003005025 8000.5 .8080: 8000.5 00.805025 80008000..
0000500.. .00... 0008.4 0005000000 0000...— 0000... 0.085
S u 04" 00
00030050000

.0008 050000020 000000-53 0... 00 3030005.. 3:00.. 00500.02. .80 00003.82. .50.. .0 b.5585 "o. 0.00...

35

CHAPTER 5: CONCLUSION AND POLICY INIPLICATIONS

Previous studies have focused on the constraints to local rice production and
consumption and the reasons for growth in aggregate rice consumption in Nigeria.
However, these studies have ignored an important aspect of rice consumption preference
in Nigeria—the role of habit persistence and perceived quality differentials in explaining
the increase in imported rice consumption relative to domestically produced rice.

Therefore, this research extends previous studies by developing a procedure for
understanding, identifying, and estimating the relative contribution of habit persistence
and perceived quality differentials in explaining consumer preference for imported rice in
Nigeria. The generic model developed in this study can be extended to study the relative
importance of habit persistence and perception of quality differences in an entire system
of demand equations for both agricultural and non-agricultural quality-differentiated
commodities.

Results show that habit persistence and consumers’ perception of quality
differences are both signiﬁcant and important in explaining consumer preference for
imported rice in Nigeria. The magnitudes of habit persistence and perceived quality
differentials were estimated to be 0.491 and 0.176, respectively. Thus, any given amount
of locally consumed rice only generates about one-ﬁfth as much utility as the same
amount of imported rice. The resulting strong preference for imported rice may have
resulted from a cultural mindset that foreign products are of higher quality than domestic

products, rather than actual quality differences between the products. This is because the

36

model estimated here only measures demand response to perceived quality differences,

not to actual differences that can be measured in some objective way.

Because habit persistence also plays an important role in explaining the relative
preference for imported rice in Nigeria, it is concluded that policies designed to
encourage production of relatively high quality local rice that will help replace imported
rice in consumption baskets will face considerable inertia due to the persistence of
consumer habits regarding purchase and consumption of imported rice, even if the locally
produced rice is of comparable quality. This will require companion policies that will
shift consumer-buying away ﬁ'om established, habitual patterns.

In general, the major ﬁndings of this study are:

1. Habit persistence and perceived quality differentials both play an important role in
explaining consumer preference for imported rice in Nigeria;

2. The strong preference for imported rice may be due to a cultural mindset that
foreign products are generally of better quality than domestic products rather than
the actual quality differences between them;

3. Consumption of any given amount of local rice generates about one-ﬁfth as much
utility as the same amount of imported rice;

4. Imported rice in Nigeria is a normal food commodity and is income inelastic;

5. The importance of income growth as a partial explanation for the growth of demand
for imported rice in Nigeria increases when the effects of perceived quality
differentials and habit persistence are included; and

6. Imported rice is a substitute for other competing consumption goods but not a

perfect substitute for locally produced rice.

37

Finally, it is concluded that while it is important to pursue policies that will
encourage production of relatively high quality local rice, such policies will face
considerable inertia if such issues as effects of perceived quality differences and
persistence of habit in purchase and consumption of imported rice are not tackled.
Therefore, companion policies designed to shift consumer-buying habits and alter
cultural mindset via advertising and promotion programs to benchmark any aggressive

quality improvement may be required to overcome this consumption inertia.

38

APPENDIX

39

APPENDIX

Elasticity formulas for alternative model speciﬁcations in the conventional and
generalized demand models

The own-price, cross-price and income elasticity formulas for the generalized

models were derived by taking the ﬁrst derivative of q], in equations (9) to (12) with
respect to p], , w, and y, , respectively, and converted to elasticity by multiplying these
by the rations of p1, , w, and y, to q“. A similar approach is taken for the conventional

models as discussed below. An exception is that in the conventional log-linear model

speciﬁcation coefficients are elasticity estimates.

1. Conventional model speciﬁcations
The elasticity formulas for testing no habit persistence but perceived quality

dominance of q] are computed by imposing a = ﬂ = 0. However, in the conventional

demand models, a =1 and ,6 = 0. This restriction tests no habit persistence and no

perceived quality differentials. Thus, q: (see deﬁnition below) is replaced with q".

[-1 : Simple linear speciﬁcation (SLS)

The SLS model is generally given as:
n
41: = 7 + 5(qu-1 + ant—l )— 0612: + 5qplt + Z] 5144’}: + 9‘ b1: - ﬂPu (411—1 + 0421-1)]
1:
For the conventional SLS where the restrictions or =1 and ,6 = O are imposed, we have

t n :3
9t =7+5qplt+ Zlajwjt +500), where 4t =qlt+q21
J:

40

(i) Own-price elasticity

t
04: ﬂ=§=ﬂ5
t a: 9
6p], Qt Qt

(ii) Cross-price elasticity

a "' 1 *2
—"—’1"':’=4,-=—.26-wﬂ
6w], Qt Qt i=1

(iii) Income elasticity

awn- _&
nix—77“ :-
atht Qt

1-2: Linear expenditure system (LES)

The LES model is given as:

5 n
Q1: 2 7g +£(Qlt—l +aQ21—1)‘aQ2t+;[)’t ‘ﬂPlt(Qlt—1 +aQ21—1)‘7qplt - ZIYjot]
t 1:

Again, the conventional LES model is:

1: 6 n
q: =7q+— yz-rqpu- 23 71W}:
P1: j=l

 

Thus,

(i) Own-price elasticity

a " n

a_q,__p%=;=_ .[M‘Zl’jwﬂj
p], Qt Ptht i=1

(ii) Cross-price elasticity

a "‘ a n
‘q—tﬂjzgj=___727jwjt
6w" Qt Ptht J'=1

41

(iii) Income elasticity

ngL=nz ”*5
Ptht

 

I-3: Almost ideal demand system (A1)

The AI is given as:

y: - [31?]: (6111—1 + “(121-1 )]
P1:

Qlt = 7+ﬂ(QIt—l +aQ21—l)-aQ2t +[

 

x {7 + aq ln(Plt)+ j}: 5]- 1n(w j, )+ 51,.[yt - ﬁmdqné—l +aq21_1 )J}

Similarly, the conventional AI model is given as:
a. n
qt = 7+[L] X {7+ 64 ln(p1,)+ 2 6]- ln(wj,)+§ln|:!P’—]}
P1: j=l t

where P, is the conventional AI price index deﬁned over (p1, ,wt ). Notice that the left-

hand side of the AI model is not shares but aggregated quantities of q] and q2.
Thus:

(i) Own-price elasticity

 

a " ”
$4.9? y’ .{5q-[7+5q1npu+ ,2 ajlnwjt+5ln(YI/Pt)]}
t (1, Ptht 1:]

(ii) Cross-price elasticity

#
6g, ﬂ=€= ijt
t *
6w], Qt Ptht

 

42

(iii) Income elasticity

at . n
ailg=n=i,{y+aq1npu+ )3 ajlnwj,+;1n(y,/P,)+g}
6y, Qt Ptht j=1

[-4: Log-linear speciﬁcation

For the log-linear speciﬁcation given as:

Qlt = ﬂ(qlt—l +aq2t—l)—anI

n
+ 6Xv[7 + 5g In P11 + Z1 5 j 1n th + 5 lnly, - ﬂplt (Qlt-l + ant—l )1]
J =
the resultant conventional demand equation becomes:

a: n
lnq, =y+6q lnp1,+ zlajmwj,+gmy,.
J:

Thus, the coefficients of p1, , w, and y, are the own-price, 5 , cross-price, 9']- and

income, 77 elasticity estimates, respectively.

(i) Own-price elasticity
*
alnq, = 4 = 5q
6 ln Plt

(ii) Cross-price elasticity

61an

 

=gj=§j

(iii) Income elasticity

61nq;

alny, =U=§

43

II. Generalized model specifications
”-1 .' Simple linear speciﬁcation (SLS)

The SLS model is given as:

n
(11: = 7 + “(111—1 + “(121—1)‘ “612: + 5gp]: + Z] 5N jt + 5b: - ﬂpu(qlt-1 + 0421-1)]
J:

(i) Own-price elasticity

ﬁﬂ=g=ﬂbq —§,6¢], where ¢ = Q1:—1+QQ21—1-
5P]: Q]: ‘11!

(ii) Cross-price elasticity

(iii) Income elasticity

5% 21 z ,, = n
5y: Q11 Q lt
Notice that the elasticity formulas for cross-price and income are the same with the
conventional SLS. This is because in the generalized SLS, cross-price and income
elasticities are not directly affected by the effects of habit and perceived quality

differentials. However, the parameter estimates could be different because they are

indirectly affected through the dependent variable.

11-2: Linear expenditure system (LES)

The LES model is given as:

6 n
(11: = 7., +ﬂ(Q1t—1 +0142t-1)—a€121 +‘p—l— J’: ‘ﬁpu(qlz—1 +a'ta-1)‘7qpu — th’jwjz]
t 1:

44

(i) Own-price elasticity

a n
-—"'—’ﬂ=c=— ‘5 yt—ZrJ-wﬂ
aPlt Qlt Pthlt j=l

 

(ii) Cross-price elasticity

6611: W1: 5 "
__=g.=_ Zyﬂwt
5W1: Qlt J Pthlt j=1 J J

 

(iii) Income elasticity

 

6cm i=7]: y: 5
5y: Qlt Pthlt

11-3: Almost ideal demand system (A1)

The AI is given as:

 

y: - ﬂpu (6111-1 + (Wm-1)]

Q]: = 7+ﬂ(Qlt—l +aq21—l)_ant +[ P1
t

 

x {7 + 5g 1n(P1:)+ :16]- ln(wj,)+ €1n[J’: — .Bpuqué-l + 61421—1)“

(i) Own-price elasticity

ﬂltﬂ= =—-1——5 — a — 5m +"5-1 -+Ir{£)]
apthlt g Pth1t{qyt ﬂplt¢( qu) Yt|:7+ q Plt 1311””): 4: Pt

where 77 = J’t —ﬂplt¢a ¢ = Qlt—l+aQ21—l

45

(ii) Cross-price elasticity

aQ1: W1: 7’ "
____5. = Z 5.
6W1: (11: j P1011: j=l J

 

(iii) Income elasticity

 

:7]:

6Q1: ﬁ y,
6y, Qlt Pthlt

7+6q lnPlt+le5 lnw j%,+§ln[ :]+ 4‘}

j=l

11-4: Log-linear speciﬁcation
This is given as:

Q1: = .3(Q1:—1 + 042:4 )" aqz:

+eXp[7+6qlnp1,+j215j1nWj-t'l’51nl)’: .Bplt(Q1:— l+aQ2:-1)]

aer’u)

P1:

 

Using the differentiation rule:

= f'(p")ef (pm)
(i) Own-price elasticity

4 = ;[,q _ 01],,

Q1: 7’

(ii) Cross-price elasticity

6-
4j= Q1t|:jz=l ]

(iii) Income elasticity

77 =y—|::]t/I, where w = exp[7+6qlnp1,+ 21-6 Hlnw,+§ln[7r]], ¢ and It as
Q1:

deﬁned earlier.

46

REFERENCES

47

REFERENCES

Abel A. B. (1990), Asset Prices under Habit Formation and Catching up with the
Joneses. The American Economic Review, Vol. 80, No. 2, pp. 38-42.

Akpokodje G., Lancon F., and Erenstein O. (2003), Nigeria ’s Rice Policy and
Development. In Erenstein O. and Lancon O. (2003) eds. The Nigerian Rice
Economy in a Competitive World: Constraints, Opportunities and Strategic
Choices. West Aﬁ'ica Rice Development Association (WARDA).

Ayinde 0., Adewumi M. and Ojehomon V. (2009), Determinants of Technical Eﬂiciency
and Varietal-Gap of Rice Production in Nigeria: A Meta-Frontier Model
Approach. International Association of Agricultural Economics 2009 Conference
Paper.

Business Day (2009), Making Nigeria a Rice Exporter.
http://wwwbusinessdayonline.com/indexphp?option=com content&view=article
&id=5040:making-nigeria-a-rice-exporter&catid=98:editorial&ltemid=35 1

Campbell J. Y. and Cochrane J. H. (1999), By Force of Habit: A Consumption-Based
Explanation of Aggregate Stock Market Behavior. Journal of Political Economy,
Vol. 107, No. 2.

Cox T. L. and Wohlgenant M. K. (1986), Prices and Quality Eﬁects in Cross-Sectional
Demand Analysis. American Journal of Agricultural Economics, Vol. 68, No. 4,
pp. 909-919.

Daramola B. (2005), Government Policies and Competitiveness of Nigerian Rice
Economy, paper presented at the ‘Workshop on Rice Policy & Food Security in
Sub-Saharan Aﬁica’ organized by WARDA, Cotonou, Republic of Benin.

Dynan K. E. (2000), Habit Formation in Consumer Preferences: Evidence ﬁom Panel
Data. The American Economic Review, Vol. 90, No. 3, pp. 391-406.

Ekeleme F., Kamara A.Y., Oikeh S. 0., Omoigui L. 0., Amaza P., Abdoulaye T. and
Chikoye D. (2009), Response of Upland Rice Cultivars to Weed Competition in
the Savannas of West Aﬁ'ica. Crop Protection; 28:90-96.

Emodi A. I. and Madukwe M. C. (2008), A Review of Policies, Acts and Initiatives in
Rice Innovation System in Nigeria. Journal of Agricultural Extension, Vol. 12 (2).

Erenstein 0., Lancon F ., Akande S. 0., Titilola S. 0., Akpokodje G., and Ogundele, O.
O. (2003), Rice production in Nigeria: A survey project report — The Nigerian

48

Rice Economy in a competitive world. Constraints, opportunities and strategies
choices. Abidjan WARDA.

Ferson W. E. and Constantinides G. M. (1991), Habit Persistence and Durability in
Aggregate Consumption, Journal of Financial Economics, Vol. 29, pp. 199-240.

Fuhrer J. C. (2000), Habit Formation in Consumption and Its Implications for Monetary-
Policy Models, The American Economic Review, Vol. 90, No. 3, pp. 367-390.

Heien D. and Durham C. (1991), A Test of the Habit Formation Hypothesis Using
Household Data, The Review of Economics and Statistics, Vol. 73, No. 2, pp.
189-199.

Imolehin E. and Wada, A. C. (2000), Meeting the Rice Production and Consumption
Demand of Nigeria with Improved Technology, International Rice Commission
Newsletter. Vol. 49.
http://foorg/docrep/x7l 64t/x7l 641 01 .htm

Index Mundi (2009), Nigeria Yearly Rice, Milled Domestic Consumption.

http: //www. indexmundi. com/agriculture/?country=ng&commodity=milled-
rice&graph=domestic-consumption.

International Rice Research Institute, [RR] (2008), World Rice Statistics.

Kebbeh, M., Haefele, S. and Fagade, S. O. (2003), Challenges and Opportunities for
Improving Irrigated Rice Productivity in Nigeria. Abidjan: WARDA.

Lancon F., Erenstein 0., Akande S. 0., Titilola S. 0., Akpokodje G. and Ogundele O. O.
(2003), Imported Rice Retailing and Purchasing in Nigeria: A Survey. In Erensten
O. and Lancon O. (2003) eds. The Nigerian Rice Economy in a Competitive
World: Constraints, Opportunities and Strategic Choices. West Aﬁica Rice
Development Association (WARDA).

Moses J. and Adebayo E. (2007), Eﬂiciency of Factors Determining Rainfed Rice
Production in Gayne Local Government Area, Adamawa State, Journal of
Sustainable Development in Agriculture and Environment. Vol. 3.

Naik N. Y. and Moore M. J. (1996), Habit Formation and Intertemporal Substitution in

Individual Food Consumption, The Review of Economics and Statistics, Vol. 78,
No. 2, pp. 321-328.

49

Nelson J. A. (1991), Quality Variation and Quantity Aggregation in Consumer Demand
for Food, The American Journal of Agricultural Economics, Vol. 73, No. 4, pp.
1204-1212.

Okoruwa V. and Ogundele O. (2006), Technical Efficiency Dijferentials in Rice
Production Technologies in Nigeria, Aﬁican Economic Research Consortium.
RP_154.
http://foorg/docrep/x7l64t/x7164101.htm

Pollak R. and Wales T. (1992), Demand System Speciﬁcation and Estimation. Oxford
University Press, New York, pp. 140-141.

Rahji M. and Adewumi M. (2008), Market Supply Response and Demand for Local Rice
in Nigeria: Implications for Self-Sufficiency Policy, Journal of Central European
Agriculture, Vol. 9, No. 3 pp. 567-574.

Shehu, J. F ., Mshelia, S. I. and Tashikalma A. M. (2007), Analysis of Technical
Eﬂiciency of Small-scale Rain-fed Upland Rice Farmers in North-west

Agricultural Zone of Adamawa State, Nigeria, Journal of Agriculture and Social
Sciences, 1813-2235.

Tijani, A. A. (2006), Analysis of the Technical Eﬁ‘iciency of Rice Farms in Ijesha Land of
0sun State Nigeria, Agrekon, Vol. 45, No 2.

Welsch H. (1989), Habit Persistence and the Structure of International Commodity
Trade. Empirical Economics Vol. 14, No. I, pp. 21-42.

West Aﬁica Rice Development Association, WARDA ( 1996), Rice Trends in Sub-
Saharan A ﬁica, Second Edition, Bouaké.

Wooldridge J. M. (2009), Introductory Econometrics, A modern approach. South-
Westem Cengage Learning, OH, US. 4'’1 Edition, pp. 517-522.

World Bank (2010), World Development Indicators (WDI) and Global Development
Finance (GDF).
http://databank.worldbank.org/ddp/home.do?Step=1&id=4

Web Materials

NationMaster.com
http://www.nationmaster.com/timepllp?stat=eco_con_pritind-economy-
consumer-price-index&countrv=ni-nigeria. Accessed April 18, 2010.

 

50

MICHIGAN STATE UNIVERSITY

l HI I III II lllllltlllllllll'es
0 96

3 1293 3163 71