CHAPTER THREE

RESEARCH METHODOLOGY

This chapter presents the methodology used to achieve the aim and objectives of this study. The
dataset used is a secondary dataset obtained from Word Bank data ware house and the Central
bank of Nigeria statistical bulletin and spans between 1981-2022. The study makes use of
descriptive statistics and Auto-regressive distributed lag model. The independent variables are
external debt and domestic debt while dependent variable is inflation rate. All analysis was
performed using Eviews 10.

3.1 Theoretical framework of the study

As earlier stated, this study employs the robustness of the Autoregressive Distributed Lag
(ARDL) model. The ARDL model is a valuable econometric tool in time-series analysis,
specifically designed for estimating the long-run relationships between variables.

Initially proposed by Pesaran et al. in 2001, this model offers an alternative to the conventional
Engle-Granger co-integration approach, which is prone to issues like integration order sensitivity
and pretesting bias. The underlying framework of the ARDL model rests on the concepts of co-
integration and error correction. Co-integration, a statistical property, implies the presence of a
sustained equilibrium relationship among non-stationary variables. It helps prevent spurious
regressions that can emerge when variables seem connected in the short term but not in the long
term.

The error correction mechanism (ECM) within the ARDL model describes how variables adjust
back to their long-run equilibrium relationship following a shock. It posits that deviations from
this equilibrium, represented by the error term, can illuminate the short-term dynamics of the
variables. The Engle-Granger two-step approach, commonly used to estimate the ECM, involves
firstly testing for co-integration between variables and then calculating the ECM itself.

What sets the ARDL model apart is its amalgamation of co-integration and error correction
techniques in a more versatile and efficient manner compared to the Engle-Granger approach.
The ARDL model is equipped to handle both stationary and non-stationary variables, and its
estimation occurs in a single step, eliminating the need for multiple stages. Moreover, this model
excels in offering more robust tests for co-integration and causality than traditional
methodologies.

The following equation underpins the ARDL model:

𝑌𝑡 = 𝛼 + 𝛽1𝑌𝑡−1 + ⋯ + 𝛽𝑝𝑌𝑡−𝑝 + 𝛽𝑝+1𝑋𝑡 + ⋯ + 𝛽𝑚𝑌𝑡−𝑚 + 𝑡 1

Where Yt and Xt are the dependent and independent variables, respectively, and εt is the error

term.

The Autoregressive Distributed Lag (ARDL) model has several advantages over other time
series models, which have been extensively discussed in the literature. Some of these advantages
are:

1. Flexibility: The ARDL model can be applied to both stationary and non-stationary
time series data, making it a more flexible alternative to other models that are only
applicable to stationary or non-stationary data. This allows for a wider range of empirical
applications and improves the model's predictive power.

2. Efficiency: The ARDL model can be estimated using a single-step procedure,

which is more efficient and less prone to estimation errors than other models that require
a twostep procedure.

3. Robustness: The ARDL model is robust to the choice of lag length, unlike other
models that are sensitive to the choice of lag length and can produce inconsistent results
if the wrong lag length is chosen.

4. Model selection: The ARDL model allows for the selection of the optimal lag
length based on information criteria such as the Akaike Information Criterion (AIC) or
the Schwarz Bayesian Criterion (SBC), which is a more rigorous approach to model
selection than other methods such as the Schwartz Information Criterion (SIC)

5. Power: The ARDL model has more power than other models in detecting long-
run relationships between variables, especially in the presence of structural breaks or
regime shifts.
 6. Interpretability: The ARDL model provides clear and interpretable coefficients
that can be used to estimate the short-run and long-run effects of the independent
variables on the dependent variable

These advantages make the ARDL model a popular and reliable tool for modeling time-series
data, especially in empirical studies that involve macroeconomic variables, finance, and social
sciences.

While the Autoregressive Distributed Lag (ARDL) model has several advantages over other
time-series models, it also has some limitations that should be considered. Some of these
limitations are:

1. Sensitivity to lag length: Although the ARDL model allows for the selection of
the optimal lag length based on information criteria, the choice of the maximum lag
length can still affect the model's performance. In particular, choosing a high maximum
lag length may lead to overfitting and reduce the model's predictive power (Pesaran
et.al., 2001).

2. Sensitivity to the choice of dependent variable: The ARDL model assumes that
the dependent variable is stationary, and using a non-stationary variable as the dependent
variable may lead to spurious regression results. This is because the model may capture
the spurious correlation between non-stationary variables rather than the true relationship
between the variables.

3. Limited ability to capture nonlinear relationships: The ARDL model is a linear

model and may not capture nonlinear relationships between the variables. This limitation
can be addressed by using nonlinear extensions of the ARDL model, such as threshold
models or smooth transition autoregressive models.

4. Limited ability to capture structural breaks: Although the ARDL model can
capture the long-run relationships between variables in the presence of structural breaks,
it may not capture the short-run dynamics associated with these breaks. This can be
addressed by using more sophisticated models, such as Markov switching models or
autoregressive conditional heteroscedasticity models (Pesaran et. al. 2001).
3.2 Specification of the Model

This study explains the intricate relationship between Inflation rate and public debt (External
debt and Domestic debt). As specified earlier, the dependent variable is Inflation rate (INF) while
External debt (EXTD) and Domestic debt (DOMD) are independent variables.

The structural model is stated as:

INF=f ( EXTD , DOMD ) 2

To investigate and provide estimates of the short-run dynamics relationships of Equation 1

above, this study adopted the ARDL model approach, according to Pesaran et. al. (2001), the
model is specified thus:

k−1 k−1 k−1

∆ log INF t −1=α 0 + ∑ a1 i ∆ log INF t−i + ∑ a 2i ∆ log EXTD t−i + ∑ a 3i ∆ log DOMD t−i +α 4 ∆ log INF t−i +α 5 ∆ log EXT
g=1 h=1 i=1

∝1 , ∝2 and ∝3 investigates the short run dynamic relationship while ∝4 , ∝5 and ∝6 investigate the
long run relationship between dependent variable (Inflation rate) and independent variables
(External debt and Domestic debt).

The lag length of the variables was selected by Akaike Information Criterion (AIC) which is
often preferred because it gives the heaviest penalties for loss of degree of freedom.

The Error Correction Term (ECT) of the model is specified as:

k−1 k−1 k−1

∆ log INF t −1=α 0 + ∑ a1 i ∆ log INF t−i + ∑ a 2i ∆ log EXTD t−i ∑ a3 i ∆ log DOMDt −i + ECT +ε t
g=1 h=1 i=1

Where:

logINF = log of Inflation Rate

logEXTD = log of External debt

logDOMD = log of Domestic debt

∆ = Difference Operator

ECT = Error Correction Term

ε t = Error Term

t = Trend component

α 0 = Constant

∑ = Summation

3.3 Expected Results

The anticipated results from the specified methodology are expected to yield meaningful result
into the intricate relationship between inflation rate and public debt (external and domestic) over
the period from 1981 to 2022. By employing the robust Autoregressive Distributed Lag (ARDL)
model, the result will uncover both short-run dynamics and long-run relationships between
inflation rate and external/domestic debt.

3.4 Model Estimation Procedure

The model estimation procedure involves several steps to analyze the specified model
comprehensively. Initially, the short-run dynamics between the variables were analyzed and the
best model was selected through the AIC criterion. Subsequently, a co-integration test is
conducted using the F-bounds test to evaluate whether a long-run relationship exists among the
variables. Once the long-run modeling is completed, a model diagnostic is performed to identify
the presence of autocorrelation and heteroscedasticity in the fitted model. This process ensures a
thorough evaluation of the model and allows for appropriate adjustments to be made based on
the diagnostic results.

3.5 Model evaluation criteria: Akaike Information Criteria (AIC)

For the purpose of this study, the Akaike Information Criteria (AIC) was used as the model
evaluation criteria.

The Akaike Information Criterion (AIC) is a widely recognized model selection criterion
employed in econometric and statistical analysis. The AIC, developed by Hirotsugu Akaike, is a
measure that strikes a balance between model fit and complexity
(https://www.wallstreetmojo.com/akaike-information-criterion). It essentially quantifies the
trade-off between goodness of fit and the number of parameters in the model. By considering
both factors, the AIC aids in the identification of the model that best represents the underlying
data generating process while avoiding overfitting.

3.6 Data Description and Sources

Table 3.1: Preview of Dataset

External Debt Domestic Debt

Inflation rate
Year Outstanding (₦' Outstanding (₦'
(%)
Billion) Billion)
1981 2.33 11.19 219.002844
1982 8.82 15.01 14.80255192
1983 10.58 22.22 19.56894989
1984 14.81 25.67 5.653664095
1985 17.30 27.95 6.927769156
1986 41.45 28.44 5.415452584
1987 100.79 36.79 19.6694756
1988 133.96 47.03 20.17712612
1989 240.39 47.05 28.96967339
1990 298.61 84.09 6.66894187
. . . .
. . . .
. . . .
. . . .
Central Bank of Central Bank of
Nigeria and Debt Nigeria and Debt
Source Management Office Management Office World Bank
The table above shows the preview of the dataset used for the analysis in this study. It contains
External debt measured in Billion Naira, Domestic debt also measured in Billion Naira and
Inflation rate (%). The dataset spans from 1981-2022.
References

Pesaran, M.H., Shin, Y. and Smith, R. (2001) Bounds Testing Approaches to the Analysis of
Level Relationships. Journal of Applied Econometrics, 16, 289-326.
https://doi.org/10.1002/jae.616

https://www.wallstreetmojo.com/akaike-information-criterion