ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3424
www.rsisinternational.org
,
DOI:
https://dx.doi.org/10.51244/IJRSI.2025.1210000297
This study examines the impact of renewable energy consumption on Kenya’s agricultural sector from 1987 to
2023 using an ARDL model. Results reveal that renewable energy exerts initially negative but significant positive
effects on agricultural output with lags of three to six years, reflecting the time required for irrigation,
mechanization and storage systems to enhance productivity. Non-renewable energy shows negative short-term
but positive lagged impacts, highlighting cost and efficiency trade-offs. Labour contributes positively in the short
run but displays delayed negative effects, while gross capital formation displays mixed outcomes, indicating
timing and efficiency challenges. The ECM model indicates a rapid adjustment toward long-run equilibrium by
correcting 71.7% of disequilibrium annually. The F-Bounds testing confirms a stable long-run association and
the diagnostic checks, normality, heteroskedasticity, serial correlation and CUSUM tests support the model’s
robustness and reliability. The results of this study emphasize that the adoption of renewable energy delivers
substantial delayed gains in agricultural growth and productivity, thereby emphasizing the need for sustained
investment, proper implementation and maintenance of renewable energy infrastructure to drive and accelerate
sustainable sectoral growth in Kenya while ensuring environmental preservation at the same time. This study
therefore recommends aligning of Kenya’s agricultural energy transition with the Sustainable Development and
Ecological modernization frameworks by integrating decentralized renewable energy systems into Vision 2030
and rural electrification interventions so as to boost productivity, resilience and environmental sustainability.
Renewable energy consumption, agriculture sector, sustainable agricultural development
Globally, the agricultural sector is essential in making sure that there is adequate food for the population and raw
materials for the industries. The sector also provides job opportunities and contributes about 4.3% to the global
GDP (Food and Agriculture Organization of the United Nations, 2020). The document asserts that in the case of
the developed world, agriculture is highly mechanized and technology-driven, thereby boosting its growth.
According to (FAO,2020) this important sector is a backbone of Kenya’s economy as it contributes substantially
to employment, food security and the GDP. Despite its significance, the agriculture sector faces significant
energy-related challenges such as unreliable electricity supply, the high non-renewable energy costs and limited
access to modern energy technologies in rural areas. The traditional dependence on non-renewable energies
serves to increase production costs, reduces efficiency and exposes farmers to environmental and price
instabilities. Renewable energy sources such as solar, wind, geothermal and biomass offer a viable roadmap to
enhancing energy security and sustainability. While the uptake of renewable energy is increasing in Kenya, the
direct contribution of renewable energy consumption to agricultural output remains underexplored. The sector’s
energy needs, such as post-harvest processing and cold storage, as well as agro-processing, suggest that
renewable energy could play a significant role in boosting growth and productivity. On the other hand,
infrastructural readiness, technology adoption, capital deployment and human resource absorption may moderate
or delay the benefits of clean green energy adoption. Empirical evidence from other jurisdictions such as the
sub-Saharan Africa and Asia, shows positive relationships between renewable energy access and agricultural
growth and productivity. This needed to be done in the Kenyan context.
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3425
www.rsisinternational.org
According to Kenya National Bureau of Statistics (2023the agriculture sector in Kenya continues to contend
with the ever-rising power costs, intermittent power supply, power blackouts, load shedding and low integration
of renewable energy systems. Though there’s increased uptake of renewable energy, its direct impact on
agricultural output remains underexplored. The understanding of this relationship is important in guiding
investment decisions, policy formulation and in spearheading technological adoption and achieving sustainable
and climate-resilient agricultural growth. Therefore, this study sought to establish the short-run and long-run
effects of the use of renewable energy on agricultural output while at the same time exploring how labour and
gross capital formation moderate this important relationship.
The results of this study will go a long way in helping policymakers, investors and other agricultural stakeholders
with an evidence-based and data-driven picture into the role of renewable energy consumption in enhancing
sustainable agricultural development while ensuring environmental preservation and conservation. By
establishing both contemporaneous and delayed effects of renewable energy consumption, the study highlights
the timing and magnitude of renewable energy interventions that are necessary in optimizing growth and
productivity. This is meant to improve efficiency and strengthen resilience against environmental and economic
shocks. This study also sought to fill a key gap in the literature where a few studies have rigorously investigated
the dynamic and sector-specific impacts of renewable energy consumption on Kenya’s agricultural output using
advanced econometric methodologies such as the ARDL and ECM, while accounting for the moderating roles
of the non-renewable energy forms, labour and capital inputs.
Some empirical studies, such as Obange (2019), while exploring data for the period 1970-2010 using the VECM
in Kenya, found that renewable energy consumption has a substantial and positive impact on the performance of
the agricultural sector, with this effect being one-directional. The study found a one-way relationship where
energy consumption drives agricultural growth. However, it did not consider other factors that may influence the
level of agricultural growth apart from energy consumption. This study exploited other variables in the model,
as energy consumption cannot be the only factor influencing the growth of the sectors. This study also adopted
capital, labour and non-renewable energy constructs as the study’s control variables. Other factors affecting
sectoral growth, beyond the listed factors, were accounted for by the error term in this study.
Sartbayeva et al. (2023) examined the link between renewable energy consumption, economic growth and the
agro-industrial sector in Kazakhstan, utilizing data from 1991 to 2021. Their findings revealed a one-way
relationship where agricultural production influences renewable energy consumption. The study employed
hierarchical regression analysis for its assessment. However, despite being conducted in a more developed
economy, the research did not quantify the strength of the relationship. The study, just like many studies in this
area, suffers from the problem of aggregation of the impacts, hence making policymaking a nullity. This study
addressed this shortcoming by disaggregating energy consumption into the various amounts of renewable energy
consumed. This study is disaggregated to mitigate this.
Liang et al. (2020), in Kazakhstan, while analyzing the consumption patterns of livestock products in Kazakhstan
over the periods 1992–2000 and 2000–2013 using both qualitative and quantitative methods, made use of
descriptive statistics, trend analysis and assessment of socio-economic and ecological factors to analyze the
characteristics and determinants of milk, meat and egg consumption across different regions. The study
established the presence of a feedback cause-effect relationship between the consumption of renewable energy
and the performance of the agricultural sector. This is because the animal feeds needed energy to prepare and in
turn, the increased renewable energy consumption meant more output from the important sector. The study,
however, is silent on the magnitudes of the associations. The ARDL and ECM methodologies employed by this
study are critical in determining not only the magnitudes and direction but also the speeds of adjustment too.
Smagulova et al. (2023), also in Kazakhstan aimed to investigate the relationship existing between the two
variables by coming up with an econometric model intended to determine the connection existing between
electricity generation and usage and digital farms on increased output from the agricultural industry. The study
made use annual data from 2017 to 2021 and employed multiple regression analysis to examine the influence of
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3426
www.rsisinternational.org
factors such as electricity production and the total number of digital agricultural farms on gross agricultural
output in Kazakhstan. The methodology is robust but may be limited by the relatively short five-year period.
Smagulova et al. (2023) concluded that these variables possessed a significant influence on agricultural
production. The study, however, is from a more developed economy. This was needed to conduct this study in
Kenya as Kazakhstan may not reflect the Kenyan situation politically, economically, climate-wise and also in
terms of energy resources endowments. The time period under consideration was also too short for a sound
econometric study. This study addressed this shortcoming for exploiting a sufficient dataset of the period
19802023.
Zou (2022) aimed to investigate the connection between agricultural productivity and energy consumption in
China, utilizing the Toda Yamagoto tests along with data from 1953 to 2020. The findings revealed a reciprocal
relationship between energy use and agricultural output. However, the study considered energy consumption as
a single aggregate variable, highlighting the need for further analysis by separating renewable and non-renewable
energy sources to understand their individual effects. Additionally, given that China is at a different level of
economic development than Kenya, conducting a similar study in Kenya was necessary to account for regional
differences.
Ahmad et al. (2019), in its quest to establish the linkage between renewable energy uptake, economic growth
and environmental dynamics in China, while exploiting the 1981-2996 dataset, determined that there was no
Granger-causality existing between the usage of energy and the performance of the agricultural sector and
suggested that the stringent energy-saving mechanisms employed in the agricultural industry will not ultimately
adversely affect agricultural activities. However, a Granger-causality bidirectional relationship was found to
exist between the variables across eastern and western China in the period 1990-2008, but no causal relationship
in central China. The differences were attributed to economic and social differences amongst the regions (Hu et
al., 2011. The study, however, harbours the inadequacies of aggregation, which is one of the critical shortcomings
of existing literature that this study sought to mitigate.
Similarly, Tiwari
et al. (2021), in a bid to establish the linkage between electricity consumption and overall
growth of the economic sectors, drawing on data covering the period 1960-2015, established a unidirectional
long-run heterogeneous panel causality flowing from electricity usage to agricultural productivity in India. The
inquiry also established a causality flowing from industrial growth to electricity usage. The study, besides being
silent on the intensities of the association, cannot be replicated in Kenya due to the social and economic
differences between Kenya and India, hence the need for this study.
According to Martinho (2016), a study done in 12 EU member states’ farms using data for the period
1989-2009
and for the years 2004–2012 to establish the efficiency and growth implications of renewable energy use in the
agricultural sector across Europe, found that energy consumption negatively affected agricultural growth. It was
therefore concluded that though the relationship was significant, it was negative. The data under consideration
was analyzed using the different econometric techniques; the GMM and frameworks drawing from the Kaldor
developments. The study, however, did not solely focus on renewable energy but instead dealt with energy
consumption as a variable. aggregation of the impacts is one of the gaps of past knowledge that this study sought
to address.
In order to determine the influence of renewable energy consumption on the agricultural sector in Kenya, the
Solow Swan growth model was adopted. The basic Cobb-Douglas production function was linearized by the
introduction of lags. The outputs were therefore interpreted as percentages. The lagged effects were also
incorporated as results of economic decisions do not occur instantaneously but take time.
𝐿𝑛(𝐴𝐺𝑅
𝑡
) = 𝐿𝑛𝐴 + 𝛼
1
𝐿𝑛(𝑅𝐸𝐶
𝑡
) + 𝛼
2
𝐿𝑛(𝑁𝑅𝐸𝐶
𝑡
) + 𝛼
3
𝐿𝑛(𝐾
𝑡
) + 𝛼
4
𝐿𝑛(𝐿
𝑡
) + 𝛽
1
𝐿𝑛(𝑅𝐸𝐶
𝑡−
1
) + 𝛽
2
𝐿𝑛(𝑁𝑅𝐸𝐶
𝑡−
1
)
+ 𝛽
3
𝐿𝑛(𝐾
𝑡−
1
) + 𝛽
4
𝐿𝑛(𝐿
𝑡−
1
) + 𝜀
𝑡
…………………….... (3.1)
Where 𝐿𝑛(𝐴𝐺𝑅
𝑡
) is the logged agricultural output at time t, 𝐿𝑛𝐴 is the total factor productivity, 𝛼
1
𝐿𝑛(𝑅𝐸𝐶
𝑡
) is
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3427
www.rsisinternational.org
logged renewable energy consumption, 𝛼
2
𝐿𝑛(𝑁𝑅𝐸𝐶
𝑡
) is logged non-renewable energy consumption, 𝛼
3
𝐿𝑛(𝐾
𝑡
)
is logged gross capital formation, 𝛼
4
𝐿𝑛(𝐿
𝑡
) is logged labour, while the values with (t-1) are the lagged values of
the variables in the model, while 𝜀
𝑡
is the error term.
AGR (”000000”)
REC
NREC (“000000”)
L (“{000000”)
GCF(“OF_GDP”)
Mean
1102373.
19791.70
0.137666
13.54270
20.09076
Median
1077750.
16115.00
0.117500
12.73076
19.73131
Maximum
1783299.
46600.00
0.242000
23.18485
25.44904
Minimum
593460.3
6000.000
0.072000
5.341202
15.00382
Std. Dev.
348262.3
10732.72
0.057575
5.556273
2.956138
Skewness
0.368935
0.900081
0.577631(
0.205444
0.087421
Kurtosis
1.892409
2.769333
1.841721
1.682008
2.157324
Jarque-Bera
3.247219
6.038611
4.906443
3.494208
1.357899
Probability
0.197186
0.048835
0.086016
0.174278
0.507149
Sum
48504415
870835.0
6.057312
595.8787
883.9934
Sum Sq. Dev.
5.22E+12
4.95E+09
0.142540
1327.503
375.7664
Observations
44
44
44
44
44
(Source: Author,2025)
Table 4.1 provides a statistical highlight of Kenya’s agricultural sector and related variables; renewable and non-
renewable energy consumption, gross capital formation and labour. The findings show that agricultural output
averaged 1.10 trillion shillings, emphasizing the critical economic role it plays despite seasonal and climatic
fluctuations. Renewable energy consumption shows a rising pattern, averaging 19791.70 kilowatt-hours. This
reflects Kenya’s increasing efforts in adopting renewable energy uptake, while non-renewable consumption
remained dominant, averaging 137666.2 kilowatt-hours due to persistent industrial and transport dependence on
non-renewable energies. Labour recorded a steady increase, with a mean of 13.5 million. This indicates
consistent population and workforce growth supporting agricultural sector growth and productivity. Gross capital
formation averaged 20.09% of GDP, thus emphasizing Kenya’s sustained investment efforts in infrastructure and
industrial development.
Variable has a unit root
: Automatic based on AIC, maximum lags of 10
ADF
Level
First Difference
CONCLUSION
Variable
Trend & Intercept
Trend & Intercept
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3428
www.rsisinternational.org
AGR
-2.001507 ( 0.5840)
-4.964347 (0.0014)
I (1)
REC
-2.148084 (0.5053)
-5.148163 (0.0008)
I (1)
NREC
-2.591905 (0.2858)
-6.793060 (0.0000)
I (1)
L
-3.558792 (0.0494)
-2.925557 ( 0.1665)
I (0)
GCF____OF_GDP
-2.754643 (0.2212)
-5.703648 (0.0002)
I (1)
(Source: Author, 2025)
Table 4.2 presents the ADF stationarity test results. The results confirm that agricultural output, renewable energy
consumption, non-renewable energy consumption and gross capital formation are non-stationary at level but
become stationary after first differencing, hence integrated of order one. Labour, however, is stationary at level,
making it I(0). The mix of I(0) and I(1) variables fulfills the conditions for applying the ARDL bounds testing
approach.
The selection of the optimum lag length was determined using the AIC criterion in the ARDL framework. This
ensured the model captured relevant dynamics with optimal simplicity for both long-run and short-run
estimations.
Dependent Variable: AGR
Method: ARDL
Date: 07/08/25 Time: 18:24
Sample (adjusted): 1987 2023
Included observations: 37 after adjustments
Maximum dependent lags: 1 (Automatic selection)
Model selection method: Akaike info criterion (AIC)
Dynamic regressors (7 lags, automatic): REC NREC L GCF
Fixed regressors: C
Number of models evalulated: 4096
Selected Model: ARDL(1, 7, 3, 7, 7)
Variable
Coefficient
Std. Error
t-Statistic
Prob.*
AGR(-1)
0.282817
0.112999
2.502826
0.0408
REC
-0.067696
0.048539
-1.394688
0.2058
REC(-1)
-0.052828
0.036293
-1.455606
0.1888
REC(-2)
-0.012091
0.042688
-0.283229
0.7852
REC(-3)
0.112919
0.041338
2.731598
0.0293
REC(-4)
0.025640
0.038634
0.663665
0.5281
REC(-5)
-0.110642
0.032288
-3.426682
0.0110
REC(-6)
-0.088877
0.031220
-2.846809
0.0248
REC(-7)
-0.207625
0.041923
-4.952563
0.0017
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3429
www.rsisinternational.org
NREC
-0.271538
0.116333
-2.334132
0.0523
NREC(-1)
0.338585
0.127249
2.660807
0.0324
NREC(-2)
0.019098
0.108987
0.175234
0.8659
NREC(-3)
0.417027
0.129969
3.208657
0.0149
L
1.716309
1.081074
1.587595
0.1564
L(-1)
-0.301677
1.701800
-0.177270
0.8643
L(-2)
-2.268661
1.728076
-1.312825
0.2306
L(-3)
2.139133
1.886039
1.134193
0.2941
L(-4)
-3.181687
2.097392
-1.516973
0.1731
L(-5)
2.793991
2.452747
1.139127
0.2921
L(-6)
-4.855929
2.783281
-1.744678
0.1246
L(-7)
4.337322
1.543880
2.809364
0.0262
GCF
-0.032973
0.057902
-0.569456
0.5868
GCF(-1)
0.086335
0.046738
1.847214
0.1072
GCF(-2)
-0.232436
0.064752
-3.589666
0.0089
GCF(-3)
0.198371
0.063642
3.117010
0.0169
GCF(-4)
-0.106129
0.048863
-2.171972
0.0664
GCF(-5)
0.199645
0.064921
3.075217
0.0179
GCF(-6)
-0.078539
0.057053
-1.376591
0.2111
GCF(-7)
0.067598
0.063365
1.066808
0.3215
C
11.50460
2.226687
5.166690
0.0013
R-squared
0.999328
Mean dependent var
27.76565
Adjusted R-squared
0.996547
S.D. dependent var
0.269067
S.E. of regression
0.015812
Akaike info criterion
-5.499480
Sum squared resid
0.001750
Schwarz criterion
-4.193330
Log likelihood
131.7404
Hannan-Quinn criter.
-5.039001
F-statistic
359.2182
Durbin-Watson stat
3.128133
Prob(F-statistic)
0.000000
*Note: p-values and any subsequent tests do not account for model
selection.
(Source: Author, 2025)
The ARDL model in Table 4:3 for the agriculture sector was estimated using 37 adjusted observations over the
period from 1987 to 2023. The optimal lag structure was selected based on the AIC criterion and specified as
ARDL (1, 7, 3, 7, 7), reflecting one lag of the dependent variable and multiple lags for the independent variables.
The dependent variable in this model is agricultural sector output, while the independent variables include
renewable energy consumption, non-renewable energy consumption, labour and gross capital formation. All data
were log-transformed, making the interpretation elasticity-based.
The coefficient of the lagged dependent variable, AGR(-1), is 0.282817, with a standard error of 0.112999, a
tstatistic of 2.502826 and a probability value of 0.0408. This coefficient is statistically significant at the 5% level.
It indicates that a 1% increase in agricultural output in the previous period leads to a 0.282817% increase in the
current period’s output. This positive carryover effect is consistent with the notion of inertia in agricultural
production where past harvests, investment cycles and seasonal factors influence current outcomes. It
demonstrates the partial but meaningful path dependence in the performance of the agriculture sector.
The contemporaneous value of renewable energy consumption has a coefficient of -0.067696, a standard error
of 0.048539, a t-statistic of -1.394688 and a probability value of 0.2058. This result is statistically insignificant,
suggesting that a 1% increase in renewable energy consumption in the current year has no meaningful effect on
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3430
www.rsisinternational.org
agriculture output. This may reflect the sector’s limited short-term sensitivity to energy changes, particularly if
energy-intensive applications such as irrigation, drying and refrigeration are not widespread or are delayed in
their impact.
The lagged value of renewable energy consumption at lag one, REC(-1), has a coefficient of -0.052828, a
standard error of 0.036293, a t-statistic of -1.455606 and a probability value of 0.1888. This too is statistically
insignificant, indicating no notable effect one year after consumption changes. The second lag, REC(-2), has a
coefficient of -0.012091, a standard error of 0.042688, a t-statistic of -0.283229 and a probability of 0.7852,
remaining statistically insignificant. However, at lag three, REC(-3) shows a significant coefficient of 0.112919,
with a standard error of 0.041338, a t-statistic of 2.731598 and a probability of 0.0293. This implies that a 1%
increase in renewable energy consumption three years earlier is associated with a 0.112919 % increase in current
agricultural output. This lagged impact is likely driven by the gradual integration of renewable energy
technologies in farm-level activities, such as solar-powered irrigation or biogas-based processing, which require
time to implement and become productive.
Interestingly, REC(-5) has a coefficient of -0.110642, standard error of 0.032288, a t-statistic of -3.426682 and
a probability value of 0.0110, which is statistically significant at the 5% level. This result suggests that a 1%
increase in renewable energy consumption five years earlier reduces agricultural output today by 0.110642%.
Similarly, REC(-6) has a coefficient of -0.088877, a standard error of 0.031220, a t-statistic of -2.846809 and a
probability value of 0.0248, also indicating a significant negative effect. The seventh lag, REC(-7), is even more
impactful, with a coefficient of -0.207625, a standard error of 0.041923, a t-statistic of -4.952563 and a
probability of 0.0017. This implies that a 1% increase in renewable energy consumption seven years ago reduces
current agricultural output by 0.207625%. These delayed negative effects may be attributed to the initial
inefficiencies, poor targeting or technical failures in past renewable energy investments, especially in off-grid
rural areas where equipment maintenance and energy reliability are known challenges.
Regarding non-renewable energy consumption, the contemporaneous value of NREC has a coefficient of
0.271538, a standard error of 0.116333, a t-statistic of -2.334132 and a probability value of 0.0523. This
coefficient is marginally significant at the 10% level and suggests that a 1% increase in non-renewable energy
consumption in the current period leads to a 0.271538% reduction in agriculture output. This negative outcome
may stem from the rising costs of fossil fuels or the unsuitability of non-renewable energy infrastructure for the
unique demands of the agricultural sector. On the other hand, the first lag of non-renewable energy consumption,
NREC(-1), has a coefficient of 0.338585, a standard error of 0.127249, a t-statistic of 2.660807 and a probability
value of 0.0324. This coefficient is statistically significant at the 5% level, indicating that a 1% increase in NREC
one year ago raises agricultural output today by 0.338585%. This may reflect time-lagged benefits of energy
used for land preparation, storage or input production. NREC(-3) is also significant, with a coefficient of
0.417027, a standard error of 0.129969, a t-statistic of 3.208657 and a probability of 0.0149. This implies that a
1% increase in non-renewable energy consumption three years earlier results in a 0.417027% rise in output,
possibly driven by delayed input-output responses in agricultural mechanization. Other lags of non-renewable
energy consumption are statistically insignificant.
Labour enters the model through multiple lags, with only the seventh lag, L(-7), showing statistical significance.
Its coefficient is 4.337322, with a standard error of 1.543880, a t-statistic of 2.809364 and a probability value of
0.0262. This result suggests that a 1% increase in labour input seven years ago increases current agricultural
output by 4.337322%. This substantial and highly delayed effect may capture the impact of long-term
agricultural education, skill development or rural employment programs that take years to manifest through
productivity.
Gross capital formation also exhibits mixed results across lags. The second lag, GCF(-2), has a coefficient of
0.232436, a standard error of 0.064752, a t-statistic of -3.589666 and a probability of 0.0089, which is
statistically significant. This implies that a 1% increase in capital investment two years ago is associated with a
0.232436% reduction in output, possibly reflecting sunk costs or misallocated funds. On the other hand, the third
lag, GCF(-3), shows a positive and significant coefficient of 0.198371, a standard error of 0.063642, a t-statistic
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3431
www.rsisinternational.org
of 3.117010 and probability of 0.0169. This means that a 1% increase in capital investment three years earlier
leads to a 0.198371% increase in agricultural output, indicating delayed but productive investments. Similarly,
GCF(-5) has a positive and significant coefficient of 0.199645, with a standard error of 0.064921, a t-statistic of
3.075217 and a probability value of 0.0179. These findings suggest that capital investment can have either
negative or positive outcomes depending on timing, project type and implementation efficiency.
The constant term in the model is 11.50460, with a standard error of 2.226687, a t-statistic of 5.166690 and a
probability of 0.0013, which is highly significant. This value captures the baseline level of output not explained
by the regressors, including institutional, policy or environmental factors.
The ARDL results show that renewable energy consumption exerts both positive and negative effects on
agricultural output, depending on the lag period, with insignificant short-term impacts but significant effects
appearing between the third and seventh year. This reflects the long gestation period of renewable energy
investments in agriculture, such as irrigation and storage systems, which take time before boosting productivity.
Non-renewable energy provides positive medium-term support but has a negative immediate impact, suggesting
that while fossil fuels can temporarily enhance farm operations, their costs and inefficiencies undermine shortrun
output. Labour influences are highly delayed, consistent with the seasonal and structural dynamics of agricultural
employment, while capital investment displays alternating productivity effects, indicating inefficiencies or
misallocation in some periods. The findings emphasize that the agriculture–energy relationship in Kenya is
strongly dependent on timing, implementation and sector-specific dynamics.
The model has a very high explanatory power, with an R-squared value of 0.999328 and an adjusted R-squared
of 0.996547. This indicates that over 99.9% of the variation in agricultural output is explained by the included
variables and their lags. The F-statistic is 359.2182 with a probability value of 0.000000, confirming that the
regressors are jointly significant. The D-W statistic is 3.128133, which is higher than the ideal value of 2,
suggesting potential negative serial correlation and warranting further diagnostic testing.
ARDL Error Correction Regression
Dependent Variable: D(AGR)
Selected Model: ARDL(1, 7, 3, 7, 7)
Case 3: Unrestricted Constant and No Trend
Date: 07/08/25 Time: 18:24
Sample: 1980 2023
Included observations: 37
ECM Regression
Case 3: Unrestricted Constant and No Trend
Variable
Coefficient
Std. Error
t-Statistic
Prob.
C
11.50460
1.233369
9.327783
0.0000
D(REC)
-0.067696
0.026746
-2.531086
0.0392
D(REC(-1))
0.280675
0.037713
7.442427
0.0001
D(REC(-2))
0.268585
0.038212
7.028788
0.0002
D(REC(-3))
0.381504
0.039371
9.690048
0.0000
D(REC(-4))
0.407144
0.040916
9.950698
0.0000
D(REC(-5))
0.296502
0.038731
7.655370
0.0001
D(REC(-6))
0.207625
0.025145
8.256955
0.0001
D(NREC)
-0.271538
0.054894
-4.946549
0.0017
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3432
www.rsisinternational.org
D(NREC(-1))
-0.436126
0.079017
-5.519387
0.0009
D(NREC(-2))
-0.417027
0.088469
-4.713803
0.0022
D(L)
1.716308
0.648811
2.645314
0.0332
D(L(-1))
1.035831
0.806532
1.284302
0.2399
D(L(-2))
-1.232830
0.749316
-1.645274
0.1439
D(L(-3))
0.906302
0.803013
1.128627
0.2962
D(L(-4))
-2.275385
0.812007
-2.802173
0.0264
D(L(-5))
0.518607
1.187860
0.436589
0.6756
D(L(-6))
-4.337323
1.093900
-3.965009
0.0054
D(GCF)
-0.032973
0.029410
-1.121154
0.2992
D(GCF(-1))
-0.048510
0.032548
-1.490427
0.1797
D(GCF(-2))
-0.280947
0.036205
-7.759982
0.0001
D(GCF(-3))
-0.082575
0.030396
-2.716640
0.0299
D(GCF(-4))
-0.188704
0.027976
-6.745274
0.0003
D(GCF(-5))
0.010941
0.029298
0.373445
0.7199
D(GCF(-6))
-0.067598
0.027921
-2.421069
0.0460
CointEq(-1)*
-0.717183
0.077013
-9.312503
0.0000
R-squared
0.965224
Mean dependent var
0.024356
Adjusted R-squared
0.886186
S.D. dependent var
0.037389
S.E. of regression
0.012614
Akaike info criterion
-5.715696
Sum squared resid
0.001750
Schwarz criterion
-4.583700
Log likelihood
131.7404
Hannan-Quinn criter.
-5.316615
F-statistic
12.21223
Durbin-Watson stat
3.128133
Prob(F-statistic)
0.000061
* p-value incompatible with t-Bounds distribution.
(Source: Author, 2025)
The ECM model in Table 4:4 presents the short-run effects of renewable and non-renewable energy consumption,
labour and gross capital formation on agricultural sector output in Kenya, together with the adjustment speed
toward long-run equilibrium. The model is estimated using 37 observations based on the ARDL(1,7,3,7,7)
specification.
The constant term has a coefficient of 11.50460 with a standard error of 1.233369, a t-statistic of 9.327783 and
a probability value of 0.0000. This is highly significant and captures the autonomous short-run growth in
agriculture when all differenced regressors are neutral.
Renewable energy consumption shows an immediate negative impact. The coefficient of the contemporaneous
differenced value is -0.067696 with a t-statistic of -2.531086 and a probability value of 0.0392. This result
indicates that a 1% increase in renewable energy consumption in the current year reduces agricultural output by
0.0677%, which may reflect short-term inefficiencies, installation disruptions or adjustment costs in adopting
renewable systems. However, the lagged values reveal significant and positive effects. The first lag has a
coefficient of 0.280675 with a probability value of 0.0001, showing that renewable energy consumption in the
previous year raises current agricultural output by 0.281%. The second lag is 0.268585 with a probability value
of 0.0002, while the third lag is 0.381504 with a probability value of 0.0000. The fourth lag rises further to
0.407144 with a probability value of 0.0000, representing the strongest positive effect. The fifth lag is 0.296502
with a probability value of 0.0001 and the sixth lag is 0.207625 with a probability value of 0.0001. These findings
demonstrate that renewable energy consumption has clear delayed benefits, with substantial productivity gains
becoming visible between one and six years later. This suggests that renewable energy projects in irrigation,
mechanization and storage require time for full integration and adaptation before their contributions to output
materialize.
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3433
www.rsisinternational.org
Non-renewable energy consumption consistently shows negative and significant effects. The contemporaneous
coefficient is -0.271538 with a t-statistic of -4.946549 and a probability value of 0.0017, indicating that a 1%
increase in non-renewable energy consumption reduces agricultural output by 0.272% in the same year. The first
lag is -0.436126 with a probability value of 0.0009 and the second lag is -0.417027 with a probability value of
0.0022. These consistent negative impacts confirm that reliance on fossil fuels undermines agricultural
productivity, likely due to high operating costs, volatile pricing and inefficiencies linked to pollution and
environmental degradation.
Labour input reveals both positive and negative influences. The contemporaneous coefficient is 1.716308 with
a probability value of 0.0332, meaning that a 1%increase in labour raises agricultural output by 1.72% in the
same period. However, the fourth lag is negative at -2.275385 with a probability value of 0.0264 and the sixth
lag is -4.337323 with a probability value of 0.0054. These delayed contractionary effects suggest that while
labour can immediately contribute to productivity, structural inefficiencies, underemployment or low absorption
of unskilled workers eventually weigh down agricultural performance over time.
Gross capital formation exerts predominantly negative short-run effects. The contemporaneous coefficient is
0.032973 with a probability value of 0.2992, which is statistically insignificant. The second lag is strongly
negative at -0.280947 with a probability value of 0.0001, while the third lag is -0.082575 with a probability value
of 0.0299. The fourth lag is -0.188704 with a probability value of 0.0003 and the sixth lag is -0.067598 with a
probability value of 0.0460. These results indicate that gross capital formation often suppresses agricultural
output in the short run, possibly due to delayed project implementation, misallocation of resources or long
gestation periods before capital investments can be productively utilized.
The ECM term has a coefficient of -0.717183 with a standard error of 0.077013, a t-statistic of -9.312503 and a
probability value of 0.0000. This coefficient is highly significant and negative as expected, confirming
cointegration. It implies that about 71.7% of disequilibrium from the previous year is corrected in the current
period, showing a relatively fast adjustment speed where most shocks are absorbed within one to two years.
The diagnostic statistics indicate strong model performance. The R-squared is 0.965224, showing that 96.5% of
the variation in agricultural output is explained by the short-run regressors. The adjusted R-squared is 0.886186.
The F-statistic is 12.21223 with a probability value of 0.000061, confirming that the variables are jointly
significant. The D-W statistic is 3.128133, suggesting possible negative autocorrelation in the residuals, which
may call for robustness checks.
The ECM results demonstrate that renewable energy consumption has an initially negative but strongly positive
delayed effect on agricultural output, reflecting the time required for renewable systems to be integrated and
absorbed into farm operations. Non-renewable energy consumption consistently reduces output, highlighting its
cost inefficiencies and environmental burden. Labour exerts both immediate positive and delayed negative
effects, underlining productivity challenges in rural employment structures. Gross capital formation tends to
contract output in the short run, likely due to inefficiencies in project timing and sectoral alignment. The
relatively fast adjustment speed of over 71% correction annually confirms that agriculture in Kenya is highly
responsive to restoring equilibrium, even amid short-run volatility.
F-Bounds Test
Null Hypothesis: No levels relationship
Test Statistic
Value
Signif.
I(0)
I(1)
F-statistic
11.03744
10%
2.45
3.52
K
4
5%
2.86
4.01
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3434
www.rsisinternational.org
2.5%
3.25
4.49
1%
3.74
5.06
(Source: Author, 2025)
The F-statistic of 11.03744 as shown in Table 4:46, is well above the 1% upper bound critical value of 5.06.
Thus, the test confirms the presence of a long-run relationship among the variables, supporting the validity of
the ARDL model.
(Source: Author, 2025)
Residuals in Figure 4:1 assumed normality based on smooth curve and a probability value of 0.673282 which is
above the conventional 0.05.
4
Breusch-Godfrey Serial Correlation LM Test:
F-statistic
3.226306
Prob. F(2,5)
0.1259
Obs*R-squared
20.84648
Prob. Chi-Square(2)
0.0000
(Source: Author, 2025)
The Breusch-Godfrey LM test in Table 4:6 yields an F-statistic of 3.226306 and a p-value of 0.1259, which is
not significant, confirming no serial correlation. However, the observed R-squared test gives a p-value of 0.0000,
which could raise some concern, though the high D-W of 3.13 supports the LM conclusion.
Heteroskedasticity Test: Breusch-Pagan-Godfrey
F-statistic
1.606881
Prob. F(29,7)
0.2668
Obs*R-squared
32.16787
Prob. Chi-Square(29)
0.3126
0
1
2
3
4
5
6
7
-0.015
-0.010
-0.005
0.000
0.005
0.010
0.015
S
e
r
i
e
s
:
R
e
s
i
d
u
a
l
s
S
a
m
p
l
e
1
9
8
7
2
0
2
3
O
b
s
e
r
v
a
t
i
o
n
s
3
7
Mean
6.91e-15
Median
-0.000207
Maximum
0.015402
Minimum
-0.015656
Std. Dev.
0.006972
Skewness
0.332586
Kurtosis
2.734020
Jarque-Bera
0.791183
Probability
0.673282
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3435
www.rsisinternational.org
Scaled explained SS
0.998249
Prob. Chi-Square(29)
1.0000
(Source: Author, 2025)
The Breusch-Pagan-Godfrey test in Table 4:7 reports an F-statistic of 1.606881 and a p-value of 0.2668 and an
observed R-squared with a value of 0.3126, both of which are insignificant. Therefore, the residuals are
homoskedastic.
CUSUM 5% Significance
(Source: Author, 2025)
The CUSUM test in Figure 4:2 confirms that the model is stable over the sample period, as the cumulative sum
of residuals remains within the 5% confidence interval bands.
Agriculture presents a distinct pattern. Renewable energy consumption reduces output contemporaneously,
possibly due to installation disruptions and early inefficiencies, but shows strong and consistent positive effects
across all subsequent lags. This delayed yet powerful influence indicates that renewable energy supports
irrigation, mechanization and storage after sufficient adaptation and absorption. Non-renewable energy
consumption, on the other hand, is consistently negative, reflecting the high costs and inefficiencies of fossil fuel
use in farming systems. Labour exerts both immediate positive and delayed negative effects, with the latter
suggesting that structural inefficiencies in rural employment eventually outweigh short-term gains. Gross capital
formation predominantly suppresses output in the short run, further confirming the sector’s misalignment
between investment timing and productive needs. Importantly, agriculture displays a high speed of adjustment,
with about 72% of disequilibrium corrected annually, signaling resilience and flexibility in restoring equilibrium
despite short-run volatility.
This study therefore emphasizes the need to align Kenya’s agricultural energy transition with the Sustainable
Development and Ecological Modernization frameworks, emphasizing renewable energy as both a driver of
productivity and environmental stewardship. Policymakers should integrate the study’s findings into national
strategies such as Vision 2030 and the Rural Electrification Programme by prioritizing decentralized renewable
systems, ensuring maintenance support and incentivizing green investments to enhance agricultural resilience
and energy efficiency. Comparative insights from other African and global experiences further affirm that
structured, inclusive renewable energy policies yield sustained agricultural and rural growth.
Therefore, the consumption of renewable energy is an important accelerator of agricultural sector growth and
ISSN No. 2321-2705 | DOI: 10.51244/IJRSI |Volume XII Issue IX September 2025
Page 3436
www.rsisinternational.org
necessary interventions should be taken to increase their uptake. Such measures could include introduction of
carbon tax to discourage fossil energy consumption and removing taxes on the consumption of renewable energy
forms.
This body of knowledge was limited by data constraints. This is due to the fact that sector-specific renewable
energy consumption figures were derived from national aggregates. This could potentially reduce the accuracy
in capturing true sectoral effects that could inform sector-specific interventions.
Though this study is an invaluable tool for designing energy policies in Kenya, this is a macro-level study. Hence,
more efforts should be made to replicate this study at the county and regional levels.
1. Ahmad, M., Zhao, Z. Y., & Li, H. (2019). Revealing stylized empirical interactions among construction
sector, urbanization, energy consumption, economic growth and CO
2
emissions in China. Science of the
Total Environment, 657, 1085-1098.
Ahmed, Z., & Afzal, M. (2023). Renewable energy installations and
economic growth in Asia: A panel data analysis. Asian Journal of Economic Studies, 19(4), 301-318.
2. Food and Agriculture Organization of the United Nations. (2020). The State of Food and Agriculture 2020.
Overcoming water challenges in agriculture. Roma.
3. Hu, L., Niu, S., Ma, L., & Zhang, X. (2011). Relationship between China's agricultural production energy
consumption and agricultural economic growth based on panel data model. Chinese Journal of Agricultural
Engineering, 27(6), 1–6. https://doi.org/10.3969/j.issn.1002-6819.2011.06.001
4. Kenya National Bureau of Statistics (KNBS). (2023). Economic Survey 2023: Sectoral Performance and
Energy Constraints. KNBS.
5. Liang, Y., Zhen, L., Zhang, C., & Hu, Y. (2020). Consumption of products of livestock resources in
Kazakhstan: Characteristics and influencing factors. Environmental Development, 34, 100492.
https://doi.org/10.1016/j.envdev.2020.100492
6. Martinho, V. J. P. D. (2016). Energy consumption across European Union farms: Efficiency in terms of
farming output and utilized agricultural area. Energy, 103, 543-556.Menegaki, A. N. (2011). Growth and
renewable energy in Europe: A random effect model with evidence for neutrality hypothesis. Energy
Economics, 33(2), 257-263.
7. Obange, N. (2019). The nexus between energy consumption and agricultural performance in Kenya: A
disaggregated approach. Journal of Economics and Sustainable Development, ISSN 2222-1700 (Paper)
ISSN 2222-2855 (Online)
8. Sartbayeva, G., Balapanova, E., Mamytkanov, D. K., Talimova, L., Lukhmanova, G., & Myrzabekkyzy, K.
(2023). The relationship between energy consumption (renewable energy), economic growth, and
agroindustrial complex in Kazakhstan. International Journal of Energy Economics and Policy, 13(6), 227–
233.
https://doi.org/10.32479/ijeep.14866
9. Smagulova, S., Yermukhanbetova, A., Nurgaliyeva, K., Sariya, B., Baimukasheva, Z., Manap, A.,
Koyshinova, G., & Akimbekova, C. (2023). The impact of energy production on the introduction of ICT and
the growth of AIC in Kazakhstan. International Journal of Energy Economics and Policy, 13(1), 477– 488.
10. Tiwari, A. K., Eapen, L. M., & Nair, S. R. (2021). Electricity consumption and economic growth at the state
and sectoral level in India: Evidence using heterogeneous panel data methods. Energy Economics, 94,
105064.
11. Zou, G. (2022). Relationships between agricultural growth and energy consumption in China: Toda–
Yamamoto tests. Energy Reports, 8, 756-763.
