The African Financial Review July-August 2014 | Page 37

E(εi,t) = 0
E(εi,t εi,s )=σ2E si t=s etE(εi,t εi,s)=0 si t ≠ s
Implying E(εi,t ε’i) = σ2EIT where IT denotes the identity matrix
(Ti,Ti)
E(εi,t εj,s) = 0 whatever j ≠ s and whatever (t,s)

Findings and discussion

of hypotheses:
H0 : E(μi/Xi) = 0 ( the random effects estimators are non-biased)

H1 : E(μi/Xi) ≠ 0 (the random effects estimators are biased).

μi denotes the individual stochastic component.
The P-values associated to the x2(6) statistic within our estimations
is about 0.0000 [x2(6) = 51.96] so we conclude that the fixed effect
estimator is more consistent to our model.
Our model is therefore written as follows:

INFit = αi + β1 (DEM)it + β2 (POLSTAB)it + β3 (MONgrowth)it +
β4 (LnRES)it + β5 (Growth)it + β6 (Trade)it + εit (2)
Where αi is the unknown intercept for each entity. αi = α + μi and
α refers to the fixed component. We assume that residuals εit are
independent, identically distributed (iid) and fulfils the following
conditions, whatever i C [1, N] and t C [1, Ti]:

The estimation method that we followed is to apply OLS (Least
Square Dummy Variables) to our model, in which we have
introduced a dummy variable for each country. The objective is to
estimate our fixed effects model after correcting the (t) of Student
from heteroscedasticty using the White method. This method
provides the same values for the parameters estimated by OLS,
the difference lies in the estimated standard deviations. Correcting
errors from heteroscedasticity provides robust estimators. The
estimations results for 124 countries and for the period 19962012 are found overleaf (Table 1)

The Fisher-test [F(6,1744) = 8.34] indicates that our
explanatory variables are jointly significant at 1% level and the
R-Squared value indicates that they explain 30% of the amount
of variance of inflation. The signs in the coefficients indicate
that greater growth of money and quasi- money, and greater
trade openness is associated to more inflation. However, greater
growth of GDP, of total reserves and further efforts in insuring
democracy and political stability reduce inflation.

The positive sign of the estimated coefficient relative to
(MONgrowth) variable was expected. The increase of money
supply decreases the value of the economy’s medium of exchange,
namely money and quasi-money, and, then, produces a rise in
price levels.

The results indicate that the variable (LnRES) is not
statistically significant. But we can explain its negative correlation
with the dependent variable by the fact that reserves accumulation
produces a currency appreciation which leads to prices decrease.

As for the positive relationship between trade openness and
inflation, it can be explained by the phenomenon of imported
inflation. But this variable hasn’t either a significant effect on
inflation.

The estimated coefficient of the variable (Growth) gets also
the expected sign. As the GDP of given economy grows especially
through productivity increase, price should decrease.

The results also suggest that political stability is negatively
correlated with inflation and this institutional variable is
statistically significant at the 5% level.

A stable political environment improves the degree of
protection of property rights and social cohesion, which has
a positive effect on macroeconomic performance. Against by,
an unstable political environment improves the rent seeking
behaviour, promotes corruption and arbitrary decisions and
reduces the state control on the informal sector, which develops
the black market and produces price increases.

The result that we are mostly interested in concerns the
impact of democracy, proxied by the Kaufmann et al. (2012)
democracy index. Democracy is statistically significant at the
1% level and is negatively correlated with inflation. This result
confirms the above theoretical comments. Democracy, when it is
correctly conceptualized, by integrating participation, contestation

The African Financial Review | 37