Fiscal federalism and public service provision in Ethiopia: A mediating role of sub-national governments capacity

Figures & data

Figure 1. Conceptual model.

Source: Developed by the Authors, 2023.

Table 1. Variables used in the Study

Variables	Definition	Items	Source
Dependent Variable
SNG public service provision	The regional government’s (SNG) public service outcomes include health, education, and road.	It is computed: clean water, education access as % of the total population, and rural road per 1000km square.	WB, IMF
Independent Variables
Expenditure decentralization	It is SNG’s allowed spending from its income for the tasks it has been given. Since capital expenditure generates efficiency that can bring regional economic growth, we also include the expenditure composition to measure capital expenditure autonomy.	% Share of SNG compared to federal government expenditures (Exp _Dec), and SNG expenditure per capita (Expenditure per capita)	Schneider(2003); Bushashe and Bayiley (2023b, 2023a)
Revenue Decentralization	It indicates the extent of SNGs’ financial independence from reliance on federal grants to pay for their expenses.	SNG own revenues as % of total revenues (Rev_share) and revenue as % of Regional Growth Domestic Product (RGDP)	Bushashe & Bayiley (2023a, 2023b); Scheneider (2003)
Mediating Variable
SNG (Regional) Capacity	The composite of the regional government’s income level, tax capacity, and fiscal dependency (administrative autonomy)	SNG (Regional) capacity is measured using three indicators. The first indicator is computed: SNG inter-governmental transfer (IGT) share as a percentage of total SNG expenditure (IGT share). The second indicator is computed: SNG’s tax revenue per capita (Tax capacity). Finally, the third indicator is Real SNG (i.e., Regional) GDP per capita (GDP)	WB; IMF

Source: Developed by the researcher (2022).

Figure 2. The outer model of the Study.

Source: Study Results (2023).

Table 2. General guideline for measurement model evaluation

The strength of indicators is examined through the outer loading values. Besides, the variable reliability is measured using Composite and Cronbach’s alpha(Henseler et al., 2014, Hair et al. 2017).	Minimum Outer loading is 0.70 (or above 0.4 in exploratory research) (Lowry & Gaskin, 2014, Hair et al., 2017, Hair et al., 2021). To ensure construct reliability, Cronbach’s alpha and composite value must be above 0.7 (or above 0.6 in exploratory research).
Convergent validity gauges the magnitudes of constructs/variables that converge to their constructs. Discriminant validity gauges the magnitudes of variables that diverge from other variables in the model (Bushashe, 2023).	The construct with AVE ≥ 0.50 show convergence. Construct with, in a case of conceptually similar constructs, HTMT value less than 0.90 validate discriminant validity(Lowry & Gaskin, 2014, Henseler et al. 2014). Construct with, in a case of conceptually different constructs, HTMT value ≤ 0.85 confirms discriminant validity (Lowry & Gaskin, 2014).
The validity of formative constructs is an issue that needs to be checked. The strength of indicators can be examined through the significance of outer weights. Discriminant validity is evaluated through the inner VIF values.	Statistically significant Outer Weights are considered relevant and acceptable (Henseler et al. 2014, Hair et al.,2021). To attain the discriminant validity of formative constructs, the inner should be VIF ≤ 5 (Garson, 2016, Hair et al. 2021).

Source: Compiled by the researcher based on literature (2022).

Table 4. Diagnosis of constructs’ reliability and validity

Construct	Cronbach’s alpha	Composite reliability (rho_a)	Composite reliability (rho_c)	AVE
Expenditure Decentralization	0.802	0.802	0.910	0.835
Regional Capacity	0.534	0.630	0.759	0.520
Revenue Decentralization	0.636	0.852	0.832	0.715

Source: Study Results, 2023.

Table 5. Diagnosis of discriminant validity

	Expenditure Decentralization	Public Provisions	Regional Capacity	Revenue Decentralization
Fornell-Larcker criterion
Expenditure Decentralization	0.914
Public Provisions	0.314
Regional Capacity	−0.464	0.160	0.721
Revenue Decentralization	−0.239	0.127	0.830	0.846
Heterotrait-monotrait ratio (HTMT) – Matrix
Expenditure Decentralization
Regional Capacity	0.826
Revenue Decentralization	0.353		1.111

Source: Study Results, 2023.

Table 6. General guideline for SEM (inner) model evaluation

Criterion	Explanation	Rule of thumb for Decision
R^2 of the endogenous variable	The R2 helps to gauge variability in results explained by the explanatory variables (Hair et al., 2017).	The R2 Values 0.67, 0.33, and 0.19 are significant, modest, and weak, correspondingly (Hair et al., 2021).
Beta coefficient (β)	It evaluates numerous coefficients of association among the dependent and dependent constructs.	The t value equals 2.58 (p<0.01), 1.96 (p<0.05), and 1.64 (p<0.10) can be considered significant (Latan & Ghozali, 2013).
Effect size (f2)	It checks whether the measure of the prediction enhancement (Cohen, 1988).	f2 Values 0.02, 0.15, and 0.35 are considered to have feeble, mediocre, and significant effects, respectively (Cohen, 1988).
Prediction relevance (Q^2)	It examines the model’s predictive ability. It is calculated using Q^2 = (F^2 included— F^2 exclude)/(1- F^2 included) (Hair et al. 2021).	The values of Q^2 0.02, 0.15, and 0.35 are feeble, mediocre, and substantial effects, respectively (Hair et al. 2017, 2021).

Source: Compiled by the researchers based on literature (2022).

Table 7. PLS predict relevance assessment

Result of Predict Relevance	Q^2	PLS-SEM_RMSE	PLS-SEM_MAE	LM_RMSE	LM_MAE
Access to School	0.080	0.377	0.264	0.726	0.309
Clean Water	0.018	33.080	28.421	36.033	24.152
Rural road	0.068	2.280	1.839	8.260	2.507
IGT_share	0.121	0.209	0.144	1.453	0.257
Tax Capacity	0.638	0.021	0.013	0.021	0.006
rRGDP_c	0.183	0.000	0.000	0.000	0.000

Source: Study Results, 2023.

Table 8. Examining path coefficients

Hypotheses	Path Directions	β Coff.	Coff. of Determination	Effect Size	Inner VIF	t-statistics	Sig values	Report
H1	Revenue Decentralization -> Public Provisions	−0.275	0.239	0.028	3.529	0.915	0.360	Not Supported
H2	Expenditure Decentralization -> Public Provisions	0.546	0.239	0.280	1.396	4.572	0.000	Supported
H3	Revenue Decentralization -> Regional Capacity	0.763	0.764	2.328	1.060	14.360	0.000	Supported
H4	Expenditure Decentralization -> Regional Capacity	−0.281	0.764	0.316	1.060	5.245	0.000	Supported
H5	Regional capacity -> Public Provisions	0.641	0.239	0.127	4.239	2.118	0.034	Supported
H6	Revenue Decentralization -> Regional Capacity -> Public Provisions	0.489	NA	NA	NA	1.853	0.064	Supported
H7	Expenditure Decentralization -> Regional Capacity -> Public Provisions	−0.180	NA	NA	NA	1.761	0.078	Supported

Source: Study Results, 2023.

Notice: Since the study aimed to explore, the study used a 90% significance (Sig) level. Besides, it is challenging to calculate Coff Determination (R²), Effective Size (f²), and VIF for indirect effect; therefore, NA denotes Not Applicable.

Table 3. Results of construct indicators

	Expenditure Decentralization	Public Provisions	Regional Capacity	Revenue Decentralization	VIF	P values
Outer Loadings
Exp_Dec	0.915				1.811	0.000
Expenditure per capita	0.912				1.811	0.000
IGT_share			0.662		1.204	0.000
Rev to RGDP				0.942	1.278	0.000
Revenue Autonomy				0.737	1.278	0.040
Tax Capacity			0.883		1.309	0.000
rRGDP_c			0.584		1.101	0.000
Outer Weight
Access to School		0.860			1.082	0.000
Clean Water		−0.099			1.141	0.088
Rural road		0.431			1.104	0.020

Source: Study Results, 2023.

Table A1. Federal government budget in real Terms (in Billions)

No	Sector	Fiscal year 2020/21	Fiscal year 2021/22
1	General Service	12.8	15.9
2	Justice and Security	10.3	10.3
3	National Defense	16.5	18.3
4	Education	56.8	55
5	Health	19.4	17
6	Urban Development and Construction	71	69.7
7	Trade and Industry	2.5	2.7
8	Water and Energy	21.5	14.6
9	Agriculture and Rural Development	15.3	15.4
10	Public Debt	37	37.5
11	Provisions	13.4	9.4
12	Others	16.3	21.7
Total		293.7	287.6

Source: Authors’ Computation Based on Data from MOFEC and NBE (2020–2022).

Table A2. Examining the endogeneity of the model

Examining Path GC tests	Original sample (O)	Sample mean (M)	Standard deviation	t statistics	Sig values
GC (Revenue Decentralization) -> Public Provision	0.300	−0.348	0.641	0.5611	0.554
GC (Expenditure Decentralization)-> Public Provisions	−0.861	−0.431	0.761	1.211	0.233
GC (Revenue Decentralization) -> Regional Capacity	0.587	0.569	0.358	1.211	0.221
GC (Expenditure Decentralization)-> Regional Capacity	0.218	0.260	0.392	0.777	0.441
GC (Regional Capacity) -> Public Provisions	0.580	0.354	0.442	1.131	0.1821

Source: Study Results, 2023.

Schneider, A. (2003). Who gets what from whom? The impact of decentralization on tax capacity and pro-poor spending IDS Working Paper, p. 179, IDS.

 Google Scholar

Bushashe, M. A., & Bayiley, Y. (2023b). Fiscal decentralization and macroeconomics stability nexus: Evidence from the sub-national governments context of Ethiopia. Cogent Economics & Finance, 11(2). https://doi.org/10.1080/23322039.2023.2244353

 Web of Science ®Google Scholar

Bushashe, M. A., & Bayiley, Y. (2023a). The effect of fiscal decentralization on economic growth in sub-national governments of Ethiopia: A two-step system general methods of moments (GMM) approach. Public and Municipal Finance, 12(2), 32–42. https://doi.org/10.21511/pmf.12(2).2023.03

 Google Scholar

Henseler, J., Ringle, C. M., & Sarstedt, M. (2014). A new criterion for assessing discriminant validity in variance-based structural equation modeling. Journal of the Academy of Marketing Science, 43(1), 115–135. https://doi.org/10.1007/s11747-014-0403-8

 Web of Science ®Google Scholar

Hair, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2017). A primer on partial Least Squares Structural Equation Modeling (PLS-SEM). Sage.

 Google Scholar

Lowry, P. B., & Gaskin, J. (2014). Partial Least Squares (PLS) Structural Equation Modeling (SEM) for building and testing causal behavioral theory: when to choose it and how to use it. IEEE Transactions on Professional Communication, 57(2), 123–146. https://doi.org/10.1109/tpc.2014.2312452

 Web of Science ®Google Scholar

Hair, J. F., Hult, G. T., Ringle, C. M., Sarstedt, M., Danks, N. P., & Ray, S. (2021). Partial least squares structural equation modeling (PLS-SEM) using R. Classroom Companion: Business. https://doi.org/10.1007/978-3-030-80519-7

 Google Scholar

Bushashe, M. A. (2023). Determinants of private banks performance in Ethiopia: A partial least square structural equation model analysis (PLS-SEM). Cogent Business & Management, 10(1). https://doi.org/10.1080/23311975.2023.2174246

 Web of Science ®Google Scholar

Garson, G. (2016). Partial least squares regression and structural equation models. Statistical Associates.

 Google Scholar

Latan, H., & Ghozali, I. (2013). Partial least squares: Concept and application path modelling using program XLSTAT-PLS. Badan Penerbit Universitas Diponegoro.

 Google Scholar

Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Lawrence Erlbaum.

 Google Scholar