Prediction Of Inflation Rates In Kenya Using Binomial Logistic Regression

ABSTRACT

Inflation is increasingly becoming an important parameter that determines financial

performance of central banks worldwide. Although many empirical studies have used time

series and vector auto regression models to analyze inflation, very few studies have relied

on the predictive ability of the logistic model which has a more intuitive interpretation

for key issues such as drivers of inflation. The primary concern for every country is

growing expectations about inflation which has resulted to a specific need for developing

probabilistic models for studying the interaction between inflation and the key drivers.

In Kenya, a similar scenario is concern is experienced. The purpose of the study was

to develop a predictive binomial logistic regression model of inflation rate with published

data from the Central Bank of Kenya (CBK), to determine the associations between CBK

rates and inflation using the logistic regression model and to analyze the accuracy level

of the using the model on predicting likelihood of inflation exceeding the set target of

5%. For this purpose, various rates were used as independent variables and inflation

(categorized as either ”on target” or ”above target”) as the dependent variable with

a binomial distribution. A binomial logistic regression was performed to ascertain the

effects of these variables on the likelihood that inflation would exceed the set target.

Results showed that increasing deposit, repo and reserve repo rates were associated with

a reduction in likelihood of inflation exceeding the target, but increasing overdraft rates,

interbank rates, USD exchange rates and interbank rates were associated with increase

in the likelihood of inflation exceeding the target. The model was statistically significant

at 2 = 32.482,p < .000 and explained 68.0% (Nagelkerke R-Squared) of the variance in

inflation by correctly classifying 95.0% of the cases.