Abstract:
Population size and growth rate have great impact on the society and economy of every country.
Finding the determinants and growth rate of the population have become fundamental to pol-
icy makers in developing countries like Ghana, and its capital Accra in particular. Population
growth models like the Exponential and Logistic Growth Models are mostly used in population
dynamics. Most variables in this area are count variables. Count models are appropriate for
modeling count variables. However, count models are hardly used in this eld. Most count
data have the problem of overdispersion, where data exhibits more variability than expected
under an assumed model. There are possible advantages of including overdispersion in the
modeling process. This study sought to investigate the e ects of overdispersion in count data,
and compare the Negative Binomial Model which is able to account for overdispersion to the
Exponential and Logistic Growth Models which do not account for overdispersion. The results
were then applied to a real dataset of the Greater Accra Region of Ghana.
The results showed that avoiding overdispersion when it do exist has dire consequences. Partic-
ularly, standard errors of the estimates are understated and the signi cance of some covariates
are overstated. This was more evident when the data from the Greater Accra Region was
modeled. Unless overdispersion is treated, inferences on such results are misleading. The study
also revealed that the Negative Binomial Model actually performed better than the population
growth models in modeling the population growth. The study recommends that analysts con-
sider overdispersion when modeling count data and recommends the Negative Binomial Model
to be used in modeling population data as compared to the Exponential and Logistic Growth
Models.
