LOGISTIC REGRESSION ANALYSIS OF HOUSEHOLD INCOME AND " EXPENDITURE PATTERNS IN GHANA

ABSTRACT

Studies into income, expenditure and standard of living of individuals

and households are well documented in literature. Many of these studies cover

different forms of poverty indicators and variables. The methods used in this

analysis were mainly econometric models.

The study is concerned with determining the indicators and variables

that possibly influence income, expenditure and living standard of households.

The objective here is to model the determinants of living standards or poverty

status of households. To this end, secondary data were obtained from the

Ghana Statistical Service.Some sections of GLSS 3 and GLSS 4 were used to

develop the model.

Nine variables were recorded and about 4500 respondents were used

for each of the surveys, that is GLSS 3 and GLSS 4. Exploratory data analysis

was employed in preliminary stages of the analysis. Logistic regression was

used because this modeling technique is considered as the appropriate

technique for modeling income and expenditure, since the outcome or

response variable is dichotomous.

It was found out that poverty was more prevalent and severe in the

rural areas than the urban areas. In determining the poverty status of a

household head, that is, whether he or she is poor or not poor; household size,

locality, age of head, education and occupation played a significant role. The

poverty status model developed was made up of the variables mentioned

above. A household status could be determined by substituting the values of

the variables in the model. The outcome (response) may be close to one or

zero.