Asymmetric Impacts of Oil Price Shocks on Government Expenditures: Evidence from Saudi Arabia

Saudi Arabia, a major oil exporting country, adopted an ambitious plan known as 2030 Vision to diversify its economy through dramatic increases in domestic investment. However, with oil remaining at modest to low prices by recent historical standards, it is important to study the implications of oil price negative shocks to key macroeconomic variables. This paper investigates the effect of oil price shocks on government expenditures. By allowing for the theoretical plausibility of asymmetric effects of oil price shocks on fiscal policy, our research suggests that nothing can guarantee linearity of the impacts of oil prices positive and negative shocks to government expenditures. For this purpose, we use a quarterly dataset 1990Q-2017Q2 on government expenditures on health and education sectors, and apply a non-linear autoregressive distributed lag (NARDL) model. Our key findings show evidence of a non-linear relationship between oil prices and government expenditures in Saudi Arabia, where a negative oil price shock would have a statistically significant different impact in the long run compared to a positive shock. Finally, we build upon our empirical findings and draw some policy recommendations for the 2030 Vision.


Introduction
The oil price plummet of 2014 has revived academic and policy interests in understanding the relationship between oil prices and macroeconomic performance, especially of oil exporting countries. A large body of literature already exists focuses on how oil price volatility affects different aspects of the economy in both developed and developing countries. This research stresses on the fact that large and unanticipated changes in energy prices could have significant negative effects on oil-dependent economies. However, recent drops in oil prices, which seems to become a norm till date, stimulated researchers' appetite to study new possible equilibria at low oil price levels.
An important channel through which low oil prices are expected to affect key economic variables in oil-rich countries is government revenues. A sharp and continued drops in government revenues would certainly harm government development plans, especially in countries such as Saudi Arabia. In a country like Saudi Arabia wherein oil revenues account for more than 80% of total government revenues, it would not be surprising to see how the 2014 price crash coupled with high levels of government spending have been draining the Saudi government's budget. As a result and given the currently low oil prices, fiscal deficit in Saudi Arabia has become more than $118 billion in 2016 (about 16 % of GDP). This paper sheds light on possible implications of negative oil price shocks on government expenditures in the health and education sectors. This is an important issue for policy makers in oil dependent economies in general and in Saudi Arabia in particular 2 for several reasons. The Saudi government has recently adopted a comprehensive reform program articulated in the National Transformation Plan (NTP) -Vision 2030 -which aims at diversifying the economy away from oil within 15 years. The plan envisages the liquidation of the state-owned oil company (ARAMCO) to fund wide-ranging public investments. The privatization of Aramco through an Initial Public Offering (IPO)which produces 90% of government's revenues -is mainly to raise funds for the newly introduced reform program.
This ambitious plan focuses on three major areas. Firstly, it seeks to triple non-oil revenues by the end of the decade through levying indirect taxes and fees on public services. This includes the introduction of a value-added-tax (VAT) and developing non-oil sectors such as mining, tourism and education. Secondly, it aims to cut public spending by reducing subsidies, rationalising public investment and reducing the public wage bill by 5%. These measures are expected to inject about $53 billion into the Saudi government's budget by 2020. Thirdly, it aims to diversify both national wealth and investment portfolio abroad.
A major concern here is that low oil prices may push policy makers, in Saudi Arabia as well as in other oil exporting countries, to make difficult choices which could have serious implications on the welfare of Saudi people. Hence, our paper derives its significance from the weight of oil revenues in the kingdom of Saudi Arabia, and its important role in financing the government needs and improving the well-being of households.
More specifically, we focus on the implications of oil price negative shocks on health and education sectors given the important role investments in human capital have to play to 3 deliver on economic development.
Thus, the primary objective of the current study is to explain the interrelationship between oil prices and government expenditures on health and education sectors in the Saudi economy. In particular, we empirically estimate the relationship between oil prices and government expenditures on health and education over the last three decades.
Unlike most of the existing literature that assume linearity, we employ a non-linear autoregressive distributed lag (ARDL) model to disentangle the effects of positive and negative shocks to oil prices on government expenditures.
The remainder of this paper is organised as follows. Section 2 presents a brief overview of the literature. Section 3 provides a background on oil price changes and the evolution of government expenditures in Saudi Arabia. Section 4 presents the dataset and the econometric model. Section 5 summarises the empirical results. Finally, section 6 concludes.

Literature Review
The proportionately to the oil revenue increase.Similar results are reported in Garkaz et al. (2012), who find a statistically significant impact of oil export revenues on government expenditures in Iran.
Hamdi and Sbia (2013) study the dynamics among oil revenues, government spending, and growth in Bahrain. The authors find that oil revenues remain the principal source for growth and the main channel which government spending is financed. Dizaji (2014) examines the effects of oil shocks on government expenditures and government revenues nexus in Iran. The author finds that causality runs from oil revenues to government total expenditures. Their results also reveal that the contribution of oil revenue shocks in explaining the government expenditures is stronger than the contribution of oil price shocks. Akanbi and Sbia (2017) find empirical evidence of the effects fiscal policy on the current accounts of oil exporting countries. Medina (2016) study the impacts of commodity price shocks on fiscal policy indicators in Latin American and find that fiscal aggregates rise in response to positive shocks to commodity prices.
A major shortcoming in the previous literature, which we address in this study, is the linearity assumption of oil price shocks. To the best of our knowledge, none of the existing studies investigate the asymmetric impacts of oil price shocks on government expenditures in Saudi Arabia. In addition, the majority of the existing work tends to focus on total government expenditures while we study the impact of oil price shocks on government expenditures at a disaggregated level. More specifically, unlike the majority of the existing literature, we investigate the impacts of both negative and positive shocks to oil prices on government expenditures on the health and education sectors in Saudi 5 Arabia. The supply side is perceived, to a far extent, to be the key driver of pushing crude oil price downward by more than 50% since 2014. The low value of the price elasticity of international demand for oil results in similar movements (in direction) in export revenues, and in the Saudi economy this is very true.

Oil price changes and the Saudi economy
Saudi Arabia is an oil-based economy with strong government controls over major economic activities. Given its possession of about 16% of the world's proven petroleum reserves and its rank as the largest exporter of petroleum, Saudi Arabia plays a leading role in OPEC. The petroleum sector accounts for roughly 87% of budget revenues, 42% of GDP, and 90% of export earnings. In order to diversify its economy and create more jobs for nationals, Saudi Arabia has recently been promoting the growth of its private sector. Although there are over 6 million foreign workers, mainly concentrated in the oil and services sectors, high unemployment rates, especially among youth, remain a cause 7 of concern in Saudi Arabia. For this reason, Saudi officials are particularly focused on employing its large youth population, which generally lacks the education and technical skills that the private sector needs.
In 2015, the Kingdom incurred a budget deficit estimated at 13% of GDP, and it faces a deficit of $87 billion in 2016, which will be financed by bond sales and drawing down reserves. Although the Kingdom can finance high deficits for several years by drawing down its considerable foreign assets or by borrowing, it has announced plans to cut capital spending in 2016. Some of these plans to cut deficits include introducing a value-added tax and reducing subsidies on electricity, water, and petroleum products.
More recently, the government has approached investors about expanding the role of the private sector in the healthcare, education and tourism industries. While Saudi Arabia has emphasized their goals of diversification for some time, current low oil prices may force the government to make more drastic changes ahead of their long-run timeline.

Education Sector Performance in Saudi Arabia
Education in Saudi Arabia is segregated by sex and divided into three separately admin-

Health Sector Performance in Saudi Arabia
The Healthcare sector in Saudi Arabia is primarily managed and financed by the Government through the Ministry of Health (MOH) and number of semi-public organization who specifically operate hospitals and medical services for their employees.
In terms of main achievements incidence rate of measles marked by decreased between 2006 (3.4 to 0.6 per 100,000 population) in 2015. Also between 2006-2015 over all immunization coverage increase from 95% to more than 97%.
Life expectancy in KSA for the year 2015 (74.3 years) exceeds the regional average by 6 years and exceed the global average by 4 years. Crude death rate (per 1000 population) is (3.9) which is lower than the regional rate (6.3) and almost half of the global rate. Infant mortality rate (per 1000 live births) among Saudis for year 2014 (7.4) is 83% less than the regional rate (44) and 80% less than the global rate (37).(www.moh.gov.sa)

Data and Methodology
This study examines the hypothesis that oil price shocks would have significant impacts on government expenditures on the Health and education sectors in Saudi Arabia. It is expected that oil price changes would affect government revenues, which would ultimately affect government expenditures on different sectors. Our dataset includes quarterly data between 1990Q1 and 2017Q2 for oil price, government expenditures on education and health as percentage of total government expenditures. All data are extracted from Datastream. Table 1 presents the descriptive statistics of the dataset. To empirically investigate the relationship between oil price shocks and other variables,

12
we employ an asymmetric version of the autoregressive distributed lag ARDL model which is first introduced in Pesaran and Shin (1998) and Pesaran et al. (2001). An important feature of ARDL models is their relatively good performance in small samples as well as their flexibility with the order of integration of the regressors. This flexible modelling technique is extended by Shin et al. (2014) to include the non-linear case, wherein the impacts of positive and negative changes are allowed to be asymmetric.
More specifically, the non-linear autoregressive distributed lags (NARDL) model, which we apply in this study, uses partial sum decompositions to implement nonlinearity. Both short-run and long-run asymmetries can be encompassed in the transmission process as the NARDL modelling approach coherently models the long-term relationship as well as the dynamic adjustment pattern. The asymmetric co-integrating relationship is expressed as follows: where y t is the outcome variable, x + t and x − t are the partial sum process of positive and negative changes in a k × 1 vector of regressors x t and u t is the error term. In Eq. 1, β + and β − denote the related asymmetric long-term parameters and x t can be decomposed as: x + t and x − t are the partial sum process of positive (+) and negative (−) changes in x t , which are defined as shown below: The error correction model correlated with the same asymmetric co-integrating relationship as in Eq. 5 was developed by Shin et al. (2014) by incorporating the ARDL (p, q) extension in Eq. 1 which is referred to as NARDL model: where t is the error term, β + = −θ + /ρ and β − = −θ − /ρ are the asymmetric long-run parameters, see Greenwood-Nimmo and Shin (2013).
To test for the cointegrating relationships in the model, Shin et al. (2014) employ the conintegation bound test (F pss ) of Pesaran et al. (2001), which is based on a modified F-test. The F pss tests for the null hypothesis of no cointegration. In particular, the F pss is an F-test that tests for the joint null-hypothesis that the coefficients of the lagged level variables are jointly equal zero. That is, the null of no cointegration is stated as Pesaran et al. (2001) reports two critical bounds; the upper and the lower.
To reject the null hypothesis, and therefore concludes the existence of countegration, the test statistics (F pss ) should be greater than the upper bound. Thus, it provides evidence of a long-run equilibrium relationship. On the other hand, if the F pss statistic is smaller than the lower bound, we fail to reject the null hypothesis and a conclusion of no cointegration 14 is made. Finally, if the F pss value lies between the critical bounds, the test is inconclusive.
To test for asymmetries, the null hypothesis of no long-and short-run asymmetry is tested using the Wald test.

Empirical Results
A prominent feature of ARDL models is that individual series are allowed to be integrated of order zero and/or one. However, for the computed F-statistics to be valid it is important to ensure that there is no I(2) series included in the model. To test for the integration properties of the individual series, we employ the Augmented Dickey Fuller (ADF) test of Dickey and Fuller (1981). The ADF test results, see Table 2, show that all series are I(1) where reject the null hypothesis only when data are in the first difference form. Before we proceed to estimate our asymmetric ARDL model, we estimate an ARDL model alongside a number of diagnostic tests. We adopt a simple specification throughout the paper wherein the outcome variable is regressed on its own lagged values as well as the lagged values and first difference of oil prices expressed in their log form. While we use the AIC information criteria to determine the appropriate lag order for the ARDL model, we follow a general to specific approach for the NARDL. In both models, we apply Pesaran et al. (2001) test for cointegration, Breusch Godfrey LM test for autocorre-15 lation, Breusch Pagan Heteroscedasticity test, Ramsey RESET test for specification errors, and cumulative sum of recursive residuals for coefficient stability.
The results from the symmetric (i.e., linear) ARDL models for Education and Health as well as their diagnostic tests are reported in Table 3. The F-statistic of Pesaran et al. (2001) are relatively high (9.281 and 16.573) suggest the presence of long run relationships between oil prices and government expenditure on education and health. The adjustment coefficient is negative and significant (at the 1% level) in both ARDL models which ensures convergence towards long-run equilibrium. The speed of adjustment seems to be very slow though.  The ARDL results also show that government expenditures (on both education as well as on health ) are positively related to oil prices in the long run. However, this coefficient 16 is found to be statistically significant (at the 1% level) only in the case of education. In addition, both ARDL models seem to pass the diagnostic tests, see the lower section of Table 3 and Fig. 4. However, both models assumes linearity which does not have to be the case. In other words, since that ARDL model is symmetric, it would follow the implicit assumptions that the size of the impact of positive and negative shocks to oil prices on government expenditures to be identical, which is unlikely to be the case. Such built-in assumption if not true would lead to a misspecification problem.
Therefore, we extend the analysis to investigate the asymmetric impact of oil price shock on government expenditures as presented in section 4. Our NARDL estimation is reported in Table 4 with the selected number of lags included in the model is based on a general-to-specific approach. The table presents  The long run impact of positive and negative shocks to oil prices is statistically significant.
It is important to verify the validity of the asymmetric impact assumption the model 17   To show the adjustment process and the period of disequilibrium caused by a shock to oil prices, we use the NARDL estimates to produce the cumulative dynamic multiplier. In other words, the dynamic multiplier of the NARDL model would explain the adjustment process from the initial equilibrium to the new equilibrium point that results from a positive or negative shock. According to Shin et al. (2014), the dynamic multiplier can still depict any possible asymmetries exist in the adjustment path even if no evidence of short run asymmetry is found. This is due to the fact that the adjustment path is not determined only by long run parameters but also on the error correction coefficients and the dynamics of the model itself. The dynamics of the oil prices cumulative multiplier shows that a positive shock to oil prices would have a greater impact on government expenditures on education compared 19 to a negative shock to oil prices. However, the magnitude of the impact of a negative shock to oil prices would have a greater impact on government expenditures on health compared to a positive shock. This indicates that the negative shock moves government expenditures on health to a lower equilibrium point on the long-run. Moreover, in both models the dynamics of positive and negative shocks are straightforward and show no significant differences on the short. In addition, it is worth noting how the impact dies out gradually and the new equilibrium is reached after nearly eight years. The overall conclusion, though, is that when oil prices drop, government expenditures on education and health would not reacts immediately. However, the negative shock to the same variable would have a stronger consequences on government spending in the long-run.

Conclusion
This paper examines the asymmetric impact of oil prices shocks on government expenditures on education and health. We employ a non-linear autoregressive distributed lag NARDL model to model the relationship between oil prices and government expen-20 ditures in Saudi Arabia; a major oil exporting country. The key findings in this study point to empirical evidence of the asymmetric impact of oil price shocks on government expenditures. In particular, our results support the non-linearity proposition wherein positive shocks to oil prices would have statistically different impacts on government spending compared to negative shocks. However, our findings lend support to such asymmetries only in the long run. Moreover, our results show that negative shocks to oil prices would have relatively greater long-run impacts on government expenditures on health compared to positive shocks. In addition, the effect of positive oil prices shocks on government spending on education are found to exceed the magnitude of negative shocks on the same variable. An important policy lesson could be learnt from our results.
Although negative oil price shocks are expected to hit both variables in the long-run, government spending on the health sector may suffer relatively more compared to the education sector. Therefore, it is important to diversify funding sources on both education and health sector in the long-run to offset any negative impacts drops on oil prices would have on both sectors.