645 agro-processing industries; the sample for agroprocessing industries was selected using stratified
random sampling, which is the preferred sampling
when the whole population is considered (as in this the
study). Three levels of stratification were used in this
survey: type of agriculture production, establishment
size, and region (rural/urban). For the Uganda’s Enterprise Survey, size stratification was defined using
number employees into small enterprise (5 to 19),
medium enterprise (20 to 99), and large enterprise
(>99). In Uganda, many agro-based industries are
recognized and grouped under different classifications such as livestock, horticulture, cereals, grains,
vegetables, and cash crops agro-industries.
3.2 Statistical analysis
The study examined how electricity access influences
industrial output in Uganda; electricity access, labour,
and capital were the independent variables, and industrial output was the outcome variable. The data on
the study variables were extracted and managed in
excel; they were later exported to Eviews 7.0 and
Stata 15.0 (StataCorp, College Station, TX, USA)
for analysis. In order to avoid spurious results, the
study variables were tested for normality using the
Jarque-Bera statistic. The descriptive statistics were
summarized by mean, standard deviations (SD), and
Table 1. Descriptive statistics and testing for normality.
Jarque- PVARIABLES N Mean Max
Min
SD
Kurt Bera
value
LN
44 8.1
8.576 7.435 0.367 2.039 3.975 0.137
(INDOUT)
LN
44 19.6 19.996 19.258 0.202 2.062 1.841 0.398
(ELECCSP)
LN
44 14.1 14.263 13.825 0.118 2.174 1.965 0.374
(EDUC)
LN
44 7.2
7.316 7.018 0.092 1.667 4.276 0.118
(LBR)
Table 2. Testing for stationarity of the variables (DFGLS unit root test).
DFGLS test at
DFGLS Test at level
first difference
Trend and intercept
Trend and intercept
ADF
pADF
pVariables
Stat.
value
Decision
Status
Stat.
value
Decision
Status
LN
−1.79
0.081
Accept
Not
−6.62
<0.001
Reject
Stationary
(INDOUT)
stationary
LN
−3.54
0.001
Reject
Stationary
(ELECCSP)
LN
−3.04
0.004
Reject
Stationary
(EDUC)
Ln
−1.53
0.134
Accept
Not
−5.18
<0.001
Reject
Stationary
(LBR)
stationary
Ho: The null hypothesis is that the series have a unit root (Not stationary)
Ha: The alternative hypothesis is that the series have no unit root (Stationary)
Reject the null if P-value is less than 0.05
minimum and maximum values. Due to the time series
nature of our data, we tested for stationarity using the
Augmented Dickey Fuller (ADF) test for unit root.
This study tested for cointegration to find out whether
the study variables have a long-term relationship. The
null hypothesis was there is no cointegration among
the variables, while the alternative hypothesis is that
the variables are cointegrated. We used a cointegrated
vector autoregression (CVAR) as an estimation technique. According to Gujarati and Porter (2009), the
term autoregressive is due to the appearance of the
lagged value of the dependent variable on the righthand side, and the term vector is since one is dealing
with a vector of two or more variables. The VAR model
is one of the most successful and easy to use for the
analysis of multivariate time series. Johansen’s (1995)
cointegration technique based on VAR is employed to
determine the long-term relationship between industrial output and its explanatory variables using the
3SLS regression technique.
4 RESULTS AND FINDINGS
This study examined the indirect role of electricity
consumption, labour, and capital on industrial output
in Uganda. The model assumed that the productivity of capital, labour, and education is augmented by
electricity consumption, which provides the energy
for the modern production technology. The direction
of causality in the nexus was traced along the type
of electricity used, and the effect of labour, education, and capital were estimated using Stata version
15 and E-views. The results in Table 1 show that variables are normally distributed given that the P-value
is greater than 0.05 (Benjamin et al. 2018; Ioannidis
2018; Kennedy-Shaffer 2019). The results show that
electricity consumption and education were stationary at the first level, while industrial output and labour
were stationary at the first difference. Detailed analysis
is presented in Table 2.
88
random sampling, which is the preferred sampling
when the whole population is considered (as in this the
study). Three levels of stratification were used in this
survey: type of agriculture production, establishment
size, and region (rural/urban). For the Uganda’s Enterprise Survey, size stratification was defined using
number employees into small enterprise (5 to 19),
medium enterprise (20 to 99), and large enterprise
(>99). In Uganda, many agro-based industries are
recognized and grouped under different classifications such as livestock, horticulture, cereals, grains,
vegetables, and cash crops agro-industries.
3.2 Statistical analysis
The study examined how electricity access influences
industrial output in Uganda; electricity access, labour,
and capital were the independent variables, and industrial output was the outcome variable. The data on
the study variables were extracted and managed in
excel; they were later exported to Eviews 7.0 and
Stata 15.0 (StataCorp, College Station, TX, USA)
for analysis. In order to avoid spurious results, the
study variables were tested for normality using the
Jarque-Bera statistic. The descriptive statistics were
summarized by mean, standard deviations (SD), and
Table 1. Descriptive statistics and testing for normality.
Jarque- PVARIABLES N Mean Max
Min
SD
Kurt Bera
value
LN
44 8.1
8.576 7.435 0.367 2.039 3.975 0.137
(INDOUT)
LN
44 19.6 19.996 19.258 0.202 2.062 1.841 0.398
(ELECCSP)
LN
44 14.1 14.263 13.825 0.118 2.174 1.965 0.374
(EDUC)
LN
44 7.2
7.316 7.018 0.092 1.667 4.276 0.118
(LBR)
Table 2. Testing for stationarity of the variables (DFGLS unit root test).
DFGLS test at
DFGLS Test at level
first difference
Trend and intercept
Trend and intercept
ADF
pADF
pVariables
Stat.
value
Decision
Status
Stat.
value
Decision
Status
LN
−1.79
0.081
Accept
Not
−6.62
<0.001
Reject
Stationary
(INDOUT)
stationary
LN
−3.54
0.001
Reject
Stationary
(ELECCSP)
LN
−3.04
0.004
Reject
Stationary
(EDUC)
Ln
−1.53
0.134
Accept
Not
−5.18
<0.001
Reject
Stationary
(LBR)
stationary
Ho: The null hypothesis is that the series have a unit root (Not stationary)
Ha: The alternative hypothesis is that the series have no unit root (Stationary)
Reject the null if P-value is less than 0.05
minimum and maximum values. Due to the time series
nature of our data, we tested for stationarity using the
Augmented Dickey Fuller (ADF) test for unit root.
This study tested for cointegration to find out whether
the study variables have a long-term relationship. The
null hypothesis was there is no cointegration among
the variables, while the alternative hypothesis is that
the variables are cointegrated. We used a cointegrated
vector autoregression (CVAR) as an estimation technique. According to Gujarati and Porter (2009), the
term autoregressive is due to the appearance of the
lagged value of the dependent variable on the righthand side, and the term vector is since one is dealing
with a vector of two or more variables. The VAR model
is one of the most successful and easy to use for the
analysis of multivariate time series. Johansen’s (1995)
cointegration technique based on VAR is employed to
determine the long-term relationship between industrial output and its explanatory variables using the
3SLS regression technique.
4 RESULTS AND FINDINGS
This study examined the indirect role of electricity
consumption, labour, and capital on industrial output
in Uganda. The model assumed that the productivity of capital, labour, and education is augmented by
electricity consumption, which provides the energy
for the modern production technology. The direction
of causality in the nexus was traced along the type
of electricity used, and the effect of labour, education, and capital were estimated using Stata version
15 and E-views. The results in Table 1 show that variables are normally distributed given that the P-value
is greater than 0.05 (Benjamin et al. 2018; Ioannidis
2018; Kennedy-Shaffer 2019). The results show that
electricity consumption and education were stationary at the first level, while industrial output and labour
were stationary at the first difference. Detailed analysis
is presented in Table 2.
88