CHAPTER III

DESIGN OF THE STUDY

This chapter briefly sketches the rationale behind the choice of the study area, the

sampling framework, and the statistical and econometric techniques used for analysing the

data collected.

3.1 Choice of study area

The Second-Tier cities of Tamil Nadu state were considered as the universe of this

study. Among the five Second-Tier cities of Tamil Nadu, two corporation cities, viz.,

Coimbatore and Salem (Fig. 3.1) were selected randomly. Increased industrial growth

coupled with rapid urbanization that in turn triggered increasing inhabitations of different

income strata are the characteristic features of these two selected corporation cities. The

residents of these cities not only had the benefit of urbanization in terms of a variety of

consumer products, but also had reaped the advantages of accessing to fresh livestock

products, as the leading livestock production pockets were adjacent to these cities. Of these

two cities, Coimbatore is the second largest city of Tamil Nadu, known for its textile and

manufacturing factories, engineering firms and automobile parts manufacturers, while Salem

is the fifth largest city of the State, sheltering largely cottage industries, besides a number of

industries including Steel Authority of India Limited (SAIL) and an exclusive Electrical and

Electronics Industrial Estate.

3.2 Selection of households

A multistage sampling procedure was adopted to select the respondents of the study.

In the first stage, as stated above, two Second-Tier cities, viz., Coimbatore and Salem were

selected randomly. In the second stage, eight zones, four from each of the two selected cities,

were chosen and in the third stage, 16 wards, two from each chosen zone were selected using

simple random sampling technique. In the fourth stage, 160 household respondents, 10 from

each of the selected wards were chosen randomly. Thus, this study had the sample size of 160

household consumers comprising 80 from each of the cities (Table 3.1).

Table 3.1: Sample households in the study area

Sl. No. District Name of the Zone Name of the chosen wardNo. of

households selected

1.

Coimbatore

East Sowripalayam 10Upplipalayam 10

2. West R.S.Puram 10Thillai nagar 10

3. North Peelamedu 10Gandhipuram 10

4. South Gopalapuram 10Ashok nagar 10

Total 801.

Salem

Suramangalam Jagirammapalayam 10Seerangapalayam 10

2. Hasthampatty Swarnapuri 10Suramangalam 10

3. Ammapet Ammapet 10Kichipalayam 10

4. Kondalampatty Kondalampatty 10Linemedu 10

Total 80Grand Total (Sample size) 160

3.3 Period of study

The reference year for the study was 2007-08 and the data collection was undertaken

during the months of January, February, March and April, 2008.

3.4 Method of enquiry and collection of data

From the household consumers so selected, relevant data were collected to achieve

the objectives of the study. The data were collected through personal enquiry, by

interviewing the sample respondent households with the help of a structured and pilot-tested

interview schedule.

The interview schedule had the demographic and socio-economic details of the

consumers, followed by the expenditure on ‘food’ items (cereals, pulses, vegetables, spices,

condiments, milk, meat, egg, etc.) and ‘non-food’ items (education, recreation, medical

expenses, rent, etc.). The interview schedule also had the provision to assess the quality

perceptions, awareness and preferences for various products of milk and meat.

Secondary data relevant for the study were collected from the Directorate of

Economics and Statistics, Government of Tamil Nadu and also from the offices of

Coimbatore and Salem Corporations.

3.5 Classification of households

The selected households were classified based on the monthly income of the

household, religion and educational level of the head of the household. Following Rao

(1982), consumption units of the households were worked out as below, so as to facilitate

comparison among households.

Table 3.2: Consumption units of households

Sl. No. Categories Consumption units

1. Adult male above 14 years 1.00

2. Adult female above 14 years 0.83

3. Child between 10.1 - 14 years 0.83

4. Child between 6 - 10 years 0.73

5. Child below 6 years 0.50

3.6 Tools of analysis

The analyses of the data were carried through conventional, tabular and functional

methods.

3.6.1 Functional analysis

In order to assess the interrelationships between consumption of livestock products

and the socio-economic and psychological factors, as described by Daisy Rani (1995), three

Semi-log functions, one each for milk, meat and eggs, were fitted. The functional forms were

as below:

1nY1 = α + β1X1 + β2X2 + β3X3 + β4X4 + β5X5 + β6X6 + β7X7 + β8X8 + β9X9 + β10X10 + β11X11 + ξ

1nY2 = α + β1X1 + β2X2 + β3X3 + β4X4 + β5X5 + β6X6 + β7X7 + β8X8 + β9X9 + β10X10 + β11X11 + ξ

1nY3 = α + β1X1 + β2X2 + β3X3 + β4X4 + β6X6 + β7X7 + β8X8 + β9X9 + β10X10 + β11X11 + ξ

Where,

Y1 = Quantity of milk consumed per household per month (in litres)

Y2 = Number of eggs consumed per household per month

Y3 = Quantity of meat consumed per household per month (in kg)

X1 = Family size in consumption units

X2 = Hindu religion dummy (1-if Hindu; 0-otherwise)

X3 = Christian religion dummy (1-if Christian; 0-otherwise)

X4 = Educational level of the head of the household (0-if Illiterate; 1-if Primary; 2-if Secondary; and 3-if college)

X5 = Food habit dummy (1-if Non-vegetarian; 0-otherwise)

X6 = Region dummy (1-if Coimbatore; 0-otherwise)

X7 = Child dummy (1-if family had child-below 14 years; 0-otherwise)

X8 = Aged dummy (1-if family had aged person-above 60 years; 0-otherwise)

X9 = Low income dummy (1-if household income is less than Rs.10000; 0-otherwise)

X10 = Middle income dummy (1-if household income is Rs.10000-20000; 0-otherwise)

X11 = Physical exertion dummy (1-if family member does physical work; 0-otherwise)

ξ = Stochastic disturbance term

βi = Regression coefficients to be estimated

α = Intercept

3.6.2 Ordered Probit Model

Ordered-response models recognize the indexed nature of various response variables;

in this study, consumers’ preferences towards attributes of milk and meat quality were the

ordered responses. Underlying indexing in such models is a latent, but continuous descriptor

of the response. In an Ordered Probit Model, the random error associated with this continuous

descriptor is assumed to follow a normal distribution.

In contrast to Ordered-response models, Multinomial Logit and Probit Models neglect

the data’s ordinality, require estimation of more parameters (in the case of three or more

alternatives, thus reducing the degrees of freedom available for estimation), and are

associated with undesirable properties, such as the independence of irrelevant alternatives in

the case of a Multinomial Logit (Ben-Akiva and Lerman, 1985) or lack of a closed-form

likelihood as in the case of a Multinomial Probit (Greene, 2000).

An individual consumers’ utility function or preference ordering was hypothesized to

be represented by consumers’ importance ratings R (R=1-strongly no; R=2-no; R=3-slightly

no; R=4-slightly yes; R=5-yes; and R=6-strongly yes) on different milk and meat quality

attributes. Ratings (R’s) are determined by a 1 x 1 vector (X) consisting of socio-economic,

geographic and demographic factors of the representative household respondent (Table 3.3).

Table 3.3: Description of variables used in Ordered Probit analysis

Explanatory variables Levels Specification

Family size Continuous Consumption units in the household

Sex Male; Female 1 – if male; 0 – otherwise

Hindu Hindu;Non-Hindu 1 – if household is a Hindu; 0 – otherwise

Christian Christian: Non-Christian 1 – if household is a Christian; 0 – otherwise

Education FourEducational level of the head of the household (0 – if Illiterate; 1 – if Primary; 2 – if Secondary; and 3 – if College)

Food habit Non-vegetarian; Vegetarian 1 – if Non-Vegetarian; 0 – otherwise

Region Coimbatore; Salem 1 – if Coimbatore; 0 – otherwise

Child Two 1 – if family had child(ren)-below 14 years; 0 – otherwise

Aged Two 1 – if family had aged person(s)-above 60 years; 0 – otherwise

Low income Two 1 – if household income is less than Rs.10000; 0 – otherwise

Middle income Two 1 – if household income is Rs.10001-20000; 0 – otherwise

Physical exertion Two 1 – if respondent does physical work; 0 – otherwise

The ordered probit models of this study were estimated using STATA 9.0® software

packages. The following model specification was used here:

Where, = latent and continuous measure of preference of the respondent n in the study,

= a vector of explanatory variables describing the respondent,

β = a vector of parameters to be estimated, and

n = a random error term (assumed to follow a standard normal distribution).

The observed and coded discrete preference variable, was determined from the model as below:

=

where, the represent thresholds to be estimated (along with the parameter vector

). The probabilities associated with the coded responses of an Ordered Probit Model are as

follows:

where n is an individual, k is a response alternative, is the probability that

the individual n responds in manner k, and (.) is the standard normal cumulative

distribution function.

The interpretation of this model’s primary parameter set, β, is that positive signs

indicate higher preference as the value of the associated variables increase, while negative

signs suggest the converse.

3.6.3 Factor Analysis

In addition to assessing the determinants of consumers’ importance ratings for limited

livestock products’ attributes, Factor analysis was used to detect the interrelationships for

consumer’s ratings (in six point scale, as in Ordered Probit analysis) among the detailed

questions placed on attributes (Table 3.4). Thus, the Factor analysis used in this study

identified the latent factors underlying important ratings (R) on the set of livestock products’

attributes and explained the structure of importance ratings (R) on these selected products’

attributes using the reduced number of latent factors (Peng et al., 2005). The following

equation describes the Factor analysis model in a matrix form:

Where,

‘X’ is the matrix of variables;

‘A’ is the matrix of factor loadings (aij);

‘F’ is the matrix of dimensions;

‘aij’ is the net correlation between jth dimension ith observed variable;

‘n’ is the number of variables; and

‘m’ is the number of dimensions.

Following Dziuban and Shirkey (1974), the sampling adequacy was determined by the

Kaiser-Meyer-Olkin (KMO) measure. According to them, the KMO index in the 0.90s was

‘marvelous’, in the 0.80s ‘meritorious’, in the 0.70s ‘middling’, in the 0.60s ‘mediocre’, in the

0.50s ‘miserable’ and below 0.50 ‘unacceptable’. The Principal Component Analysis method

was followed to extract dimensions and the initial orthogonal solution was rotated by the varimax

method with Kaiser normalisation (Dave et al., 2007).

Table 3.4: List of factors used for Factor analysis

Milk quality attributes Meat quality attributesHigh price is permeable for safety Tenderness is importantHigh price is permeable for high sensory property High price is permeable for safetyHigh SNF is important High cookery property (parts and form) is importantLow fat is important High price is permeable for high sensory propertyColour is important Retail cut is importantFeeding condition is important Ageing is importantLow price is important Meat colour is importantGood odour is important Feeding condition is importantTexture is important Low price is importantGood taste is important Good odour is importantSafety is important Juiciness is importantFreshness is important Good taste is importantShopping environment is important Safety is importantGood flavour is important Leanness is goodHomogenization is important Freshness is importantPasteurization is important Shopping environment is important Good packaging is importantCow milk is preferredBuffalo milk is preferredMilking environment is important(R=1-strongly no; R=2-no; R=3-slightly no; R=4-slightly yes; R=5-yes; and R=6-strongly yes)