This analysis research is concerned with studying the structure and evolution of government spending in Iraq for the period (2004–2023), with the main goal of identifying how the nature of distribution between financial resources allocated to current and investment spending and measuring the differential impact of each of the two components on economic activity using the ARDL model that allows studying the correlation in the short- as well as long-term. Moreover, we aim to explore the extent to which external factors such as oil price shocks and global economic conditions can change the correlation between the two components and determine the future course of fiscal policy in Iraq. This research has several objectives. First, it seeks to better understand the dynamics of this correlation in Iraq’s economy, for example, this correlation has become particularly relevant for the recent years following the massive price shocks that hit the oil market and the significant structural disruptions that the economy underwent. Finally, the last goal of the research is encouraging policymakers to adopt more efficient and flexible fiscal policies that would help sustain the economic activity and shift away from excessive dependence on oil revenues by directing government spending toward more production sectors with a higher potential to significantly contribute to achieving the economy’s diversification and stability.

Research Methodology: To achieve the research objective and test its hypothesis, the descriptive (analytical) approach within economic theories and the quantitative (econometric) approach dependent on the econometric method using the cointegration methodology were adopted, relying on the statistical software (Eviews 12).

Descriptive Analysis: Analyzes the reality of government expenditure components, oil prices, and inflation rates, in addition to their correlation with economic growth in Iraq during the period (2004–2023).

Econometric Analysis: Analyzes the correlation between government expenditure components, oil prices, and inflation rate and their impact on economic growth. We will use an econometric approach dependent on instrumental variable models. First, we will test for stationarity, then test for cointegration between the economic variables under study, and third, find predicts for long-run and short-run factors and the error correction term, and then conduct diagnostic tests for the model.

Results of Descriptive Analysis: Analysis of the correlation between government expenditure components, oil prices, and inflation and their impact on economic growth.

Figure 1 shows how government expenditure components, oil price, and inflation rate affect the dependent variable (annual GDP change rate). GDP growth and current expenditure as a proportion of GDP were inversely related. Spending increased from 25.3% in 2004 to 37.4% in 2022, while growth fluctuated between 2.5% and 7% and declined to 10.4% in 2020 with expenditure at 40.1%. This expenditure is going toward operational consumption rather than productive investment, which defies economic theory that ties well-planned government spending to growth. Investment expenditure as a proportion of GDP has a moderate positive association, rising to 6.2% in 2012 with 8.7% expenditure and 11% in 2016 with 5.9%. The 2020 crisis caused growth to decline to (−10.4%) despite expenditure at (6.8%), showing how sensitive this association is to economic and political settings. Despite the increase in health expenditure to 3.8% in 2016, GDP grew 11% due to increased oil prices, not health sector efficiency. Health expenditure grew to 3.9% in 2020 while GDP fell to 10.4% due to the oil crisis and COVID-19 pandemic. This suggests that health expenditure did not increase real productivity or economic growth, which contradicts the economic theory that human capital investment (such as health) boosts labor force productivity and growth. Education spending as a percent of GDP is correlated. Education expenditure surged to 5.2% in 2012 and 4.8% in 2016, respectively, boosting economic growth to 6.2% and 11%. In 2022, the GDP grew 7% and education spending about 4.5%). Dependent on economic theory, increasing educational spending increases human capital efficiency and productivity. Oil prices, on the other hand, directly affect growth, rising to (97 dollars/barrel) in 2008 and (107.5 dollars/barrel) in 2011, respectively, and falling to (41.7 dollars/barrel) in 2020, causing growth to contract to (−10.4%), reflecting the high dependence on oil revenues. GDP growth and inflation are inversely related. In 2020, inflation was 0.6% with a contraction of −10.4%, in 2008, it was 12.5% with growth of 9.2% due to the oil boom, and in 2018, it was 0.3% with moderate growth of 2.3%, supporting the theory that high inflation pressures GDP growth.

Findings of the econometric analysis: Findings of measuring the correlation between government expenditure components, oil prices, inflation, and their impact on economic growth. 1.

Defining the model variables and functional specification:

Table 1 demonstrates the research variables used in the econometric model, their symbols, and their functional specification.

Accordingly, the model for the impact of disaggregating government expenditure components on economic growth considering price shocks and structural disturbances can be formulated dependent on the following functional form: Y = f ( X1 , X2 , X3 , X4 , X5 , X6 ) (1) 2.

Testing the stationarity of time series for research variables:

Regression models dependent on time series data rely in their predictions on the assumption that these series are characterized by stationarity. However, non-stationary series (i.e., those containing a unit root) often lead to spurious regression findings. The reason for this is the nature of time series data, which are often affected by a general trend resulting from common conditions that are reflected in all variables and make them move in the same direction, despite the absence of a real causal correlation between them. Therefore, it becomes necessary to test the stationarity property for all variables under study and determine the integration order of each variable before proceeding with the prediction. Table 2 demonstrates the findings of the time series stationarity test for research variables dependent on the Phillips-Perron (PP) test. It is observed that the calculated probability magnitudes (prob) were greater than (5%) at the original level I(0), which means that the time series are non-stationary and suffer from the presence of a unit root in all three cases (with constant, with constant and trend, and without them). Therefore, the null hypothesis, which states the existence of a unit root, is accepted. To address this problem, a unit root test was performed for the first difference of the time series data, and the probability magnitudes (prob) became less than (5%), which means that the time series for the research variables became stationary and integrated of the first order I(1), whether with an intercept, with an intercept and a general trend, or without them. Therefore, the alternative hypothesis, which states that the time series are free of a unit root, is accepted. 3.

Formulation of the Econometric Model and prediction Methodology.

After conducting the stationarity test for the time series of the research variables, it became appropriate to use the modern cointegration methodology dependent on the (ARDL) model and to predict the equilibrium correlation in the short and long run and analyze it using the following linear form: ∆ Y t = c + B 1 y t − 1 + B 2 X 1t − 1 + B 3 X 2t − 1 + B 4 X 3t − 1 + B 5 X 4t − 1 + B 6 X 5t − 1 + B 7 X 6t − 1 + ∑ i = 1 ɋ λ 1 ∆ y t − i + ∑ i = 1 ɋ 1 λ 2 ∆ x 1 t − i + ∑ i = 1 ɋ 2 λ 3 ∆ x 2 t − i + … . + ∑ i = 1 ɋ 6 λ 7 ∆ x 6 t − i + ε t (2)

Where:

Yt: Economic growth (annual rate of change in GDP), X1: Current expenditure as a percent of GDP, X2: Investment expenditure as a percent of GDP, X3: Education expenditure as a percent of GDP, X4: Health expenditure as a percent of GDP, X5: Oil price (dollar/barrel), X6: Inflation rate, βi: Long-run correlation factors, λi: Short-run correlation factors, ∆: First differences of variable magnitudes, q: Optimal lag length, ε_t: Random error term.

Quarterly data (80 observations) for the period (2004–2023) were used, applying the standard program equation and the linear form of the (ARDL) model for prediction to obtain stable linear correlations where the predicts are interpreted as marginal propensities. The time series data for the model variables are publicly available in the Zenodo repository ( https://doi.org/10.5281/zenodo.19666496). 4.

Results of the Econometric Model prediction.

a)

Testing the Long-Run Equilibrium Correlation between Model Variables.

To test whether there is a long-run equilibrium correlation between the independent variables under study and the dependent variable, the F-statistic magnitude is calculated. If the calculated F-statistic magnitude is greater than the upper bound of the critical F-magnitude, we reject the null hypothesis (H0) and accept the alternative hypothesis (H1), which states that there is cointegration between the variables (existence of a long-run equilibrium correlation). Conversely, if the calculated F-statistic magnitude is less than its upper bound critical magnitude, we accept the null hypothesis (H0) and reject the alternative hypothesis (H1). Table 3 demonstrates the findings of the bounds test for the (ARDL) model.

Table 3 demonstrates the findings of the Bounds Test, where the calculated F-statistic magnitude reached (6.649300), which is higher than the critical upper bound F-magnitude of (3.99) at a significance level of (1%). This means rejecting the null hypothesis and accepting the alternative hypothesis, indicating a long-term equilibrium correlation moving from the set of independent variables towards the dependent variable (annual percent change in GDP). This confirms the validity of the research hypothesis, which necessitates predicting the long-term and short-term factors and the error correction term. b)

Findings on predicting short-term and long-term factors and the error correction term for the (ARDL) model:

After confirming the existence of a long-term equilibrium correlation (cointegration) between the independent variables and the dependent variable (annual percent change in GDP), it is necessary to predict the long-term factors, short-term factors, and the error correction term. Table 4 illustrates this.

This indicates the existence of a cointegration correlation between the independent variables and the dependent variable. This is confirmed by the error correction term (Coint Eq(−1)), which is (−0.705803), negative and significant at a significance level less than (1%). This means that the two basic conditions for this coefficient are met, implying that (0.70) of the short-term errors are automatically corrected over time to reach long-term equilibrium. In other words, the short-term imbalance can be adjusted in the long term, thus returning to an equilibrium state. This indicates that the adaptation in the econometric model used was relatively slow. c)

Evaluation of the statistical and econometric quality of the predicted model (diagnostic tests of the model):

•

Evaluation of the statistical quality of the model:

Adjusted R ² : The magnitude of the adjusted coefficient of determination for the model (Adj R2 = 0.923415) indicates that the explanatory variables, which represent the determinants of income distribution inequality under study, explain (92%) of the change occurring in the dependent variable (annual rate of change in GDP), and the remaining (8%) is due to other random factors.

t-test: The prediction findings indicate the significance of the predicted model factors for the independent variables and the constant term in both the long and short terms, as demonstrated in Table 4.

- F-statistic magnitude: Which was (40.31766) with a probability magnitude (prob F = 0), indicates the significance of the predicted model as a whole in explaining changes in the dependent variable.

From the above, the model is free from potential statistical problems. •

Evaluation of the econometric quality of the model:

- Normality test of residuals: It is clear from the magnitude of (JB = 0.717886) with a significance level higher than (5%), reaching (0.698414), that the error terms follow a normal distribution. This indicates the acceptance of the null hypothesis (H0) which states that the residuals follow a normal distribution, as demonstrated in Figure 2.

Autocorrelation test: Table 5 demonstrates the findings of the autocorrelation test for the predicted model. It is observed that the calculated F-statistic magnitude reached (0.1488) with a non-significant probability magnitude of (0.7026). Therefore, we accept the null hypothesis that there is no autocorrelation problem (H0:ρ = 0) in the model and reject the alternative hypothesis. This is supported by the Durbin-Watson statistic magnitude of (D.W = 2.0925).

- Heteroskedasticity Test: It is inferred from Table 6 that the predicted model does not suffer from the problem of heteroskedasticity, because the calculated F-statistic magnitude reached (1.148882) with a non-significant probability magnitude of (0.3306). This means accepting the null hypothesis, which states the constancy of the variance of the random error term of the predicted econometric model.

- Test of the functional form’s suitability for the model: Table 7 demonstrates the findings of the test for the suitability of the functional form for the model under study. The calculated F-statistic magnitude was (0.481320) with a probability magnitude of (0.4910), while the t-statistic magnitude was (0.693772) with a probability magnitude of (0.4910). Both probability magnitudes were higher than the significance level (5%), which means accepting the null hypothesis stating the correctness of the functional form specification used in the predicted econometric model.

Structural Stability Test for Model Coefficients: Conducting a structural stability test for the short-term and long-term coefficients of the ARDL model requires ensuring that the data is free from any structural changes. This is done by using one of two tests: either the Cumulative Sum of Recursive Residuals (CUSUM) test or the Cumulative Sum of Squares of Recursive Residuals (CUSUM SQ) test, as demonstrated in the following figure:

It is clear from the figure above that the graphical line for the Cumulative Sum of Recursive Residuals (CUSUM) test falls within the critical limits (upper and lower limits) at a significance level of (5%), which means that the predicted coefficients for the Unrestricted Error Correction Model (UECM) used are structurally stable over the research period.

- It is inferred from the findings of the diagnostic tests above the reliability of the predicts and findings obtained from the model, which indicates the quality of the predicted model and its freedom from measurement and statistical problems, and thus its findings can be relied upon and interpreted in a manner consistent with the economic reality of Iraq.

The following is inferred from the findings of the (ARDL) model prediction for the impact of government expenditure components on economic growth in the short and long run, as demonstrated in Table 4 above: 1.

The current expenditure indicator as a percent of GDP suggests (X1) that the relationship between current expenditure as a percent of GDP and economic growth as annual rate of change in GDP is significantly negative in the short and long run. Therefore, an increase in current expenditure as a percent of GDP by one unit decreased the annual rate of change in GDP by (2.3055) unit in the short run and by (0.4620) unit in the long run even when other factors are held constant. This is because current expenditure in the economy represented by salaries and wages, subsidies, and transfers do not contribute to an increase in the productive capacity of the economy; thus, it is considered non-productive consumption expenditure. In the long run, current expenditures crowd out public and private investment; hence the growth rate will be very weak. It deprives nations from opportunities for realizing sustainable economic growth, particularly in the Iraqi economy, which heavily depends on oil revenues and does not have income diversification. Thus, an increase in current expenditures does not stimulate the economy in reality but rather it limits it from expanding.

This study did not require ethical approval as it relied on secondary and publicly available data.