Sectoral employment analysis for Saudi Arabia

ABSTRACT

This study aims to explore the impact of output and wage on labour demand in Saudi Arabia at sectoral level. We applied cointegration and equilibrium correction methods to the time-series data of 10 sectors over 1995–2016 using the demand side framework and considering the structural breaks in the data. We found that in the long run, the employment is positively affected by the output while the impact of the wage was negative in all sectors. In the short-run, employment growth in all sectors reacted to the wage growth except for the government sector. While only some sectors responded to the output growth, we also found that employment can adjust to the desired equilibrium level in all sectors but time horizon for the adjustment processes varies across the sectors. Differences in estimated coefficients imply that policies should be sector-specific as a ‘one-fits-all’ policy would fail to consider the sectoral specificities.

KEYWORDS:

• Employment
• wage
• output
• cointegration
• ecm

JEL CLASSIFICATION:

• J20
• J23
• J43

I. Introduction

Achieving a desired level of employment is central to macroeconomic policy. Regulators must understand employment dynamics in order to design appropriate policies and test their impact. Healthy employment levels not only benefit household income and the production factor of firms, but also help maintain sustainable economic growth and reduce poverty. Employment is a central element in the concept of inclusive growth (UN 2006; Bhalla 2007). Therefore, the dynamics of employment determinants have been the subject of considerable research to date.

As the Saudi Arabia proceeds with historic reforms, the question of what drives employment is crucial for the country’s policymakers. Saudi Vision (2030) (SV 2030), the government’s strategic roadmap for the Kingdom’s future development, and related initiatives, such as the National Transformation Program, establish numerous employment-related schemes and set targets for employment levels. Between 2016 and 2030, SV 2030 aims to reduce unemployment from 11.6% to 7%; expand women’s participation in the workforce from 22% to 30%; increase the employment rate of Saudi nationals in the oil and gas sectors from 40% to 75%; raise the private sector’s contribution to GDP from 40% to 65%; and grow the share of non-oil exports in non-oil GDP from 16% to 50% (SV 2030; NTP, 2019). To achieve these and other targets, the authorities should conduct a comprehensive examination of employment in the country; this in turn necessitates an assessment of the relationships between employment and its main determinants.

Like other emerging economies, Saudi Arabia faces constant structural change and large, persistent differences between sectors in productivity and earnings. In the context of the Kingdom, microeconomic dynamics, government support programs for certain economic activities, sectoral growth patterns, and capital-enhancing technological progress can have widely varying impacts on employment across sectors. From a policy perspective, sector-level growth and employment trends have important implications for future development strategies. Therefore, the objective of this research is to investigate the impacts of the primary determinants of employment – namely, GDP and wages – in different sectors of the Saudi Arabian economy.

We consider 10 economic sectors: agriculture and forestry (AGR); construction (CON); distribution and retail, wholesale, hotels and catering (DIS); finance, insurance and business services (FIBU); government services (GOV); non-oil manufacturing (MANNO); non-oil mining (MINOTH); other services (including community, social and personal services) (OTHS); transport, storage and communication (TRACOM); and utilities: electricity, gas and water (U).

Our analysis addresses two primary research questions:

For each sector, what are the impacts of real wages and output on employment in the long and short run?

For each sector, how quickly does employment return to its long-run equilibrium path after deviating in the short run?

Output, value added, income, production, and economic activity are interchangeably used in the literature. To be consistent throughout the paper, we use the word of output. We apply cointegration and equilibrium correction methods to time-series data for the above ten sectors in Saudi Arabia for 1995 through 2016. Demand-side modelling of employment provides a theoretical foundation for this research.

We find that there is a long-run relationship between employment and its determinants in all the 10 sectors. In other words, the employment is positively affected by the value added while the impact of the wage is negative across the sectors in the long run. It is also found that growth rates of the values added, and wage have statistically significant positive and negative impacts, respectively, on the employment growth across the sectors in the short run. Finally, we find that the short-run dynamics in the employment converges to the long-run relationship in the all sectors since the sectoral speed of adjustment (SoA) coefficients are statistically significant and negative, indicating that the short-run dynamics converge with the long-run relationships in all sectors. However, the magnitudes of the effects vary across the sectors due to their idiosyncratic features.

The results of this study offer policymakers quantitative insight into how the key determinants of employment, such as income and wages, can affect sector-level employment in Saudi Arabia, in both the long and short run. The findings can also help authorities seeking to balance employment across sectors. Furthermore, our observation that the impacts of output and wages vary across the 10 sectors highlights the importance of designing and implementing economic policies at the sectoral level and the shortcomings of a ‘one size fits all’ approach, which would fail to consider the distinct characteristics of each sector.

This research contributes to the existing literature on employment in Saudi Arabia in the following ways. First, this research relies on the theoretical foundation and investigates employment effects of both output and wages, whereas earlier studies considered only the former. Second, we evaluate the relationships between employment and these two drivers at the sector level for 10 primary sectors of the Kingdom’s economy. Third, we use up-to-date econometric techniques and tools including Autometrics, which provides parsimonious and theoretically interpretable specification and accounts for any policy and/or regime changes and the stability of the estimated coefficients.

The rest of the paper is structured as follows. Section II provides a brief literature review and section III discusses labor market structure in Saudi Arabia. Theoretical framework is provided in section IV. Section V presents the data and econometric methodology used. The estimation results and discussion of the findings are provided in section VI and VII, while section VIII concludes the study and provides policy recommendations.

II. Literature Review

In this section, we review relevant literature on the determinants of employment, especially economic output and wages, in Saudi Arabia. According to the existing literature, the most common factors of employment are labour costs (i.e. wages) and economic output (e.g. Nickell 1984; Barker and Peterson 1988; Pesaran, Pierse, and Kumar 1989). Other explanatory variables can also be included in employment analysis, such as exchange rates (Bruno, Falzoni, and Helg 2004), inflation (e.g. Loboguerrero and Panizza 2003; Skare and Caporale 2014), trade (e.g. Freeman 2004; Bruno, Falzoni, and Helg 2004), oil prices (Papapetrou 2001; Uri 1996; Davis and Haltiwanger 2001), and taxation (e.g. Nickell 2003; Bassanini and Duval 2006).

To the best of our knowledge, only few such studies exist for the Kingdom.¹ These studies include Alkhateeb et al. 2017a, b, c, d); Khodeir and AL Nuwaiser (2016), and Aljebrin (2012). Alkhateeb et al. (2017a) assess the role of financial market development in job creation in Saudi Arabia by examining annual time-series data for 1980 to 2015. The authors model employment as a function of total credit to GDP, remittances outflows, human capital index and investment. The authors also consider Saudization by measuring it with a dummy variable, which takes unity after the implementation of the policy in 1985 and zero prior to 1985. The results of their ARDL bounds testing approach indicate that in Saudi Arabia, financial market development, Saudization, and investment tend to promote employment, whereas outflow remittance and human capital index erode it (though the latter was found statistically insignificant). This research does not consider either output or wages.

Another paper by Alkhateeb et al. (2017b) investigates the role of oil price fluctuations and economic growth in employment in Saudi Arabia, again examining annual data from 1980 to 2015. The authors measure the symmetric and asymmetric impacts of oil prices on employment using linear and non-linear ARDL models. The parameter estimates show that oil prices and economic growth each have statistically significant positive effects on generating more jobs within the economy.

A third study by the same authors examines the impacts of oil revenue, government expenditures, and economic growth on employment in Saudi Arabia from 1991 to 2016 (Alkhateeb et al. 2017c). Based on the results of a Johansen cointegration test, they conclude long-run relationships between employment and considered explanatory variables. In addition, the authors find Granger-causality running from oil revenue, public spending, and economic growth to employment over the long run whereas in the short run, only oil revenues and public spending cause employment.

Alkhateeb, Sultan and Mahmood (2017d) also examine whether trade openness, government spending on education, and economic growth influence employment in Saudi Arabia, once more examining annual time-series data covering 1980–2015. The results obtained from ARDL estimations reveal that all these variables have positive long-run impacts on employment.

Khodeir and AL Nuwaiser (2016) investigate the determinants of industrial employment in Saudi Arabia. They model industrial employment as a function of foreign direct investment (FDI), inflation (measured by consumer price index), and exports as a percentage of gross domestic product (GDP). The authors utilize annual time-series data spanning from 1990 to 2014 and employ the autoregressive distributed lag (ARDL) bounds testing method developed by Pesaran, Shin and Smith (2001). Their long run estimates reveal a negative relationship between FDI and employment. On the other hand, the study concludes that higher exports and inflation have statistically significant positive impact on employment. In the short run, all variables have negative but statistically insignificant impacts on employment. The study does not consider the main theoretically predicted determinants of employment, namely, output and wages.

Aljebrin (2012) examined the determinants of labour demand in Saudi Arabia for the period 1990–2008 by using fully modified ordinary least squares (FMOLS) approach. He found that real income, investment, government expenditure and exports have significant positive, while import has negative impact on labour demand in Saudi Arabia.

None of the studies above discuss why wages, one of the theoretically articulated determinants of employment, have not been considered. The authors also do not explain why GDP growth, rather than GDP level, is considered in the analysis.

Additional research analyzes the link between economic growth and unemployment in Saudi Arabia based on Okun’s law (e.g. Al-Habees and Rumman 2012; Khrais and Al-Wadi 2016; Abou Hamia 2016). A few other studies (e.g. Alotaibi 2017) describe employment in the Kingdom but do not apply any statistical and/or econometric tools to estimate the relationships in the labour market.

Due to the lack of empirical research that models employment in Saudi Arabia by considering its theoretically articulated determinants, we extend this review to studies of other economies that explore the relationships between employment and its drivers. These studies are documented in Appendix A of the online supplementary file.

In concluding the literature review, we offer three remarks. First, to the best of our knowledge, few studies exist that investigate employment in Saudi Arabia. Second, they do not analyse wages and consider output growth instead of output level for their long-run analysis. Third, the studies of other economies reviewed in Appendix A and those in section 2 confirm that economic output and wages are viewed as the main determinants of employment in prevailing empirical analysis.

Therefore, this research contributes to the existing literature by modelling the long-run and short-run impacts of economic output and wages on employment for 10 sectors in Saudi Arabia. The results of our empirical analysis can help inform policymakers regarding the extent to which economic output and wages impact employment in the long and short run.

III. Labour Market Structure in Saudi Arabia

There are many features characterizing labour market in Saudi Arabia. First, the public sector tends to absorb the large share of nationals entering the market, while the private sector relies heavily on foreign labour. On average, citizens usually have better education level compared to foreigners and prefer working on the public sector, where they are well-paid, have job security, have more benefits, and work less hours compared to work in the private sector. Most of expats usually are employed in the private sector and concentrated in sectors such as wholesale and retail, agriculture, trade, personal services, transport, and construction. Expats working in these sectors usually are low-skilled and paid less wages; they also may work for longer hours. Second, significant wage gap in the private sector shifts labour demand towards expats as Saudis are paid higher than non-Saudis. Indeed, there is also significant wage gap between nationals and foreigners working in the public sector as Table 1 documents.

Table 1. Wage gap between nationals and foreigners, SAR

	2019Q1	2019Q2	2019Q3	2019Q4	2020Q1	2020Q2
Overall wage gap	6427	6294	6407	6487	6272	5834
Governmental wage gap	1654	1104	1707	887	1902	1125
Private wage gap	3039	3023	3027	3164	2996	2635
Source: General Authority for Statistics, Haver Analytics, and authors calculations.

Third, even though economic participation rate has been increasing for Saudi, it is still low especially for females; available statistics for 2020Q2 indicates that only 48.8% of Saudis participate compared to 40.3% in 2017Q2. Female labour force participation increased from 17.4% in 2017Q2 to 31.4% in 2020Q2. Despite this improvement, it remains low compared to male participation rate of 65.6% in 2020Q2. Finally, there is an empirical evidence suggesting the presence of skill mismatches in the labour market as citizens may not be well equipped with skills that are needed by the private sector (International Monetary Fund 2013).

In order to foster job creations for nationals, the authorities over the past years have implemented various policies promoting employment in the private sector. For example, Hafiz programme was launched to support nationals seeking for jobs with monthly allowance of SAR 2000 for a maximum period of one year. Nitaqat programme aiming to increase the number of nationals working in the private sector. This programme also imposed sanctions on firms not hiring nationals. While there is a minimum wage for nationals working the public sector about SAR 4,000 there is no wage floor for foreigners. The authorities have imposed expats levy on foreigners and their dependents since mid-2016. Labour marker reform initiative has gradually increased the share of Saudi citizen in the private sector. The Saudi citizen employment in private sector has grown by an average of 8% per year between 2005 and 2017, while the Saudi citizen employment in public sector has grown by an average of 4% during the same period. In November 2020, the authorities announced easing the restrictions on expats working in the private sector enable them not only to change their jobs, but also leave the country without their employers’ consent.

Examining the labour market dynamics at sectoral level can be quite helpful in understanding the patterns of overall labour market development in Saudi Arabia. In Table 2, we report the private sector employment share and their growth rates.

Table 2. Private sector employment, growth rate and share

		Growth rates, % per annum			% share in total employment
Sector		2005–2009	2010–2014	2015–2017	2005	2010	2015	2017
AGR	S	0.48	21.98	0.51	2.17	1.80	2.72	3.18
	NS	10.46	6.97	−4.49	97.83	98.20	97.28	96.82
MIN	S	−4.79	19.27	0.61	67.14	59.99	76.14	80.01
	NS	4.48	−3.6	−6.27	32.86	40.01	23.86	19.99
MANNO	S	3.59	26.52	−1.33	11.9	12.10	21.79	23.16
	NS	5.07	2.75	−3.74	88.1	87.90	78.21	76.84
U	S	−7.61	79.11	−1.33	45.51	37.50	80.05	81.21
	NS	2.4	−8.45	−2.18	54.49	62.50	19.95	18.79
CON	S	2.02	37.95	1.55	8.77	6.71	11.26	12.75
	NS	12.96	14.02	−2.94	91.23	93.29	88.70	87.25
DIS	S	0.58	28.12	−33.3	11.21	12.59	20.18	22.94
	NS	0.57	6.12	−33.3	88.79	87.41	79.82	77.06
FIBU	S	8.28	99.73	4.22	30.29	9.55	43.81	49.53
	NS	7.54	−6.78	−3.49	69.71	90.45	56.19	50.47
TRACOM	S	113.46	65.5	−33.3	15.63	15.30	21.00	24.29
	NS	120.96	8.6	−33.3	84.37	84.70	79.00	75.71
OTHS	S	5.64	94	7.68	14.67	31.61	21.71	23.94
	NS	4.56	173.6	2.81	85.33	68.39	78.28	76.06
Total Labour Force	S	2.33	28.5	3.06	11.63	10.37	16.38	18.60
	NS	7.78	8.79	−2.12	88.37	89.63	83.62	81.40

Note: S = Saudi Employment, NS = Non-Saudi Employment, MIN = Mining. Data Source: SAMA (2019) (2019).

With the exception of few sectors, private sector employment is dominated by expatriates. Table 2 indicates that the dependency of private sector of Saudi economy on foreign labour force has declined over the time and proportion of Saudi employed in all sectors has increased between 2005 and 2017. However, the rate of employment restructuring in all the sectors is very slow. Few sectors of the economy such as Agriculture, Manufacturing, Construction, Distribution, Transport and Communication, Other services are still dominated by foreign labour force, resulting in high unemployment among the rapidly growing Saudi labour force. Only in Mining and utility sectors witnessed greater Saudi employment than any other sectors. Overall proportion of Saudi labour force in private sector has increase from 11.6% in 2005 to 18.6% in 2017.

At comparable levels of education, private sector wages are lower for non-Saudi workers than for Saudis. This is the main reason for the high proportion of foreign workers in the private sector observed from Table 2. Easy access to low-skilled foreign workers with low wages has meant that sectors such as wholesale and retail trade, personal services, transport, and construction have been the main drivers of private sector growth² Increasing trend of proportion of Saudi labour share and drastically increasing growth rates of Saudi labour forces in all sectors is the result of labour market reforms in Saudi Arabia.

IV. Theoretical Framework

By following studies such as Lewis and MacDonald (2002), Dowrick and Wells (2004), and Hutchings and Kouparitsas (2012), we use standard profit maximization problem for derivation of labour demand function. The profit maximization problem can be written as follows:

(1) $ma{x}_{L,K}{\pi }_t=p_t{Y}_{it}-w_t\left(1+{\tau }_t\right)L_t-r_t{K}_t$(1)

Subject to a constant return to scale Cobb-Douglas type production function (Cobb and Douglas, 1928):

(2) $Y_t=A{K}_t^{\alpha }{L}_t^{1-\alpha }$(2)

where ${\pi }_t$ is profit at time t, p_t is output price, Y_t is output, w_t is nominal wages, ${\tau }_t$ is the tax rate; L_t is employment, r_t is the rental rate of capital, K_t is capital; A is Hicks-neutral technical change.

Substituting (2) in to (1):

(3) $ma{x}_{L,K}{\pi }_t=p_t\left[A{K}_t^{\alpha }{L}_{it}^{1-\alpha }\right]-w_t\left(1+{\tau }_t\right)L_t-r_t{K}_t$(3)

The first order necessary condition for profit maximization with respect to labour implies $\frac{\mathrm{\partial }{\pi }_t}{\mathrm{\partial }{L}_t}=0$, which using (3) can be expressed as below:

(4) $w_t\left(1+{\tau }_t\right)=A{p}_t\left(1-\alpha \right)K_t^{\alpha }{L}_t^{-\alpha }=A{K}_t^{\alpha }{L}_t^{1-\alpha }\left(1-\alpha \right)p_t\frac{1}{L_t}$(4)

Substituting (2) into (4):

(5) $\frac{w_t\left(1+{\tau }_t\right)}{p_t}=A\left(1-\alpha \right)\frac{Y_t}{L_t}$(5)

Taking natural logarithms of (5) and solving for labour ($L_t$) implies:

(6) $ln{L}_t=\mathrm{l}\mathrm{n}\left(A\left(1-\alpha \right)\right)+ln{Y}_t-ln\left(\frac{w_t\left(1+{\tau }_t\right)}{p_t}\right)$(6)

(7) $ln{L}_t={\alpha }_o+ln{Y}_t-ln\left(\frac{w_t\left(1+{\tau }_t\right)}{p_t}\right)$(7)

where ${\alpha }_o=\mathrm{l}\mathrm{n}\left(A\left(1-\alpha \right)\right)$

In equation (7)(7) $ln{L}_t={\alpha }_o+ln{Y}_t-ln\left(\frac{w_t\left(1+{\tau }_t\right)}{p_t}\right)$(7) , long run labour demand is a function of three terms, i.e. a constant, real output and real wage.

In case of Saudi Arabia, ${\tau }_t=0$ can be considered as there is no income tax in the country. Additionally, to capture the technological progress, we introduce trend component following Pesaran, Pierse, and Kumar (1989). Finally, we add an error term (${\mu }_t$) to equation (7)(7) $ln{L}_t={\alpha }_o+ln{Y}_t-ln\left(\frac{w_t\left(1+{\tau }_t\right)}{p_t}\right)$(7) to make it an econometric specification. Thus, our final long-run equation for labour demand gets the following form:

(8) $e{t}_t={\alpha }_o+a_{1}gv{a}_t+a_2{w}_t+a_{3}t+{\mu }_t$(8)

where ${{\alpha }_i}^{\prime }s$ are the coefficients to be estimated econometrically. Variables in small letters mean they are in the natural logarithmic forms. Labour demand rises (declines) when output is booming (contracting), i.e. ${\alpha }_1>0$. There is a negative relationship between labour demand and the cost of it, that is, wage and hence, ${\alpha }_2<0$. It is theoretically expected that ${\alpha }_3>0$.

Note that Pesaran, Pierse, and Kumar (1989) derived the same specification as equation (8)(8) $e{t}_t={\alpha }_o+a_{1}gv{a}_t+a_2{w}_t+a_{3}t+{\mu }_t$(8) using cost minimization approach. Additionally, equation (8)(8) $e{t}_t={\alpha }_o+a_{1}gv{a}_t+a_2{w}_t+a_{3}t+{\mu }_t$(8) is similar to that of Peterson (1988) and Baker et al. (2001).

Again, following the studies above, short run equation, i.e. Equilibrium Correction Model (ECM) specification in a general form can be specified by introducing Equilibrium Correction Term (ECT) that is one-period lagged residuals of the long-run equation, lagged first differences of log employment, contemporaneous and lagged first differences of log of output and log of real wage:

(9) $\begin{array}{r}\mathrm{\Delta }e{t}_t=\phi \left[e{t}_{t-1}-{\alpha }_o-{\alpha }_{1}gv{a}_{t-1}-{\alpha }_2{w}_{t-1}-{\alpha }_3\left(t-1\right)\right]+\\ +\sum _{i=1}{\gamma }_{1i}\mathrm{\Delta }e{t}_{t-i}+\sum _{i=0}{\gamma }_{2i}\mathrm{\Delta }ga{v}_{t-i}\\ +\sum _{i=0}{\gamma }_{3i}\mathrm{\Delta }{w}_{t-i}+{\epsilon }_t\end{array}$(9)

Where, $\phi <0$ represent the speed of adjustment (SoA), at which actual employment adjusts to its desired long-run equilibrium level. We will estimate equations (8)(8) $e{t}_t={\alpha }_o+a_{1}gv{a}_t+a_2{w}_t+a_{3}t+{\mu }_t$(8) and (9(9) $\begin{array}{r}\mathrm{\Delta }e{t}_t=\phi \left[e{t}_{t-1}-{\alpha }_o-{\alpha }_{1}gv{a}_{t-1}-{\alpha }_2{w}_{t-1}-{\alpha }_3\left(t-1\right)\right]+\\ +\sum _{i=1}{\gamma }_{1i}\mathrm{\Delta }e{t}_{t-i}+\sum _{i=0}{\gamma }_{2i}\mathrm{\Delta }ga{v}_{t-i}\\ +\sum _{i=0}{\gamma }_{3i}\mathrm{\Delta }{w}_{t-i}+{\epsilon }_t\end{array}$(9) ) for the above-mentioned ten sectors of Saudi Arabia in the empirical analysis.

V. Data in Brief and Econometric Methodology

Data

We use annual time-series data from 1995 to 2016 for the following variables.

$E{T}_i$ is thousands of employed people in a given sector.

$GV{A}_i$ is gross value-added (GVA) in a given sector, measured in millions of Saudi Arabian Riyal (SAR) at constant 2010 prices.

$W_i$ is real wages in a given sector, measured in millions of SAR at constant 2010 prices. To obtain the real values, a sector’s nominal wage values were divided by its GVA deflator.

Sectoral employment data are taken from Saudi Arabia General Authority for Statistics (GaStat, 2018), with the exception of government sector employment data, which come from Saudi Monetary Authority (SAMA 2019). The sectoral GVA and nominal wage data are also provided by GaStat, (2018).³ The sectoral GVA deflators are calculated using the relevant nominal and real values, again from GaStat, 2018.

Figure B1 in Appendix B of the online supplementary file illustrates the natural logarithmic levels and growth rates of employment, output and wages in each of 10 sectors. Structure of the employment by sector in terms of share in total, and growth rate, are documented in Appendix B.

Econometric Methodology

The empirical assessment strategy in this study is as follows. We first check time series properties of variables. In the case of non-stationarity of the variables, we test whether they are cointegrated. If the variables are cointegrated, then we estimate long-run and then short-run as well as SoA coefficients of the relationships. If they are not cointegrated, then we will estimate only short-run coefficients without Equilibrium Correction Term (ECT). We follow the General to Specific Modelling (GSM) approach using Autometrics, a cutting-edge econometric technique, in estimating short-run relationships (see Hendry, Johansen, and Santos 2008; Doornik and Hendry 2009; Doornik 2009; Doornik and Hendry 2018). Details of the unit root and cointegration tests as well as long- and short-run estimation methods are described in Appendix C of the online supplementary file.

VI. Estimation Results

Unit root tests results

The results of unit root tests indicate that all the variables are non-stationary at the log levels and their growth rates are stationary as Table D1 in Appendix Do f the online supplementary file presents. For d(gva), the ADF test does not reject the null hypothesis of unit root in the first difference of gva in three sectors, DIS, MINOTH and TRACOM. However, the DF-GLS test (Elliott, Rothenberg, and Stock 1996) concludes stationarity at 5% significance level for DIS and TRACOM, while the KPSS test indicates the stationarity for d(gva) in MINOTH. We note that as Juselius (2006) discusses, stationarity at second differenced form does not seem reasonable when the sample size is smaller than 30 observations. Thus, we conclude that all the variables are non-stationary at their log-levels, but their growth rates are stationary. In other words, all variables under consideration follow I(1) process. The various cointegration tests reported in Table D2 of Appendix D confirm the existence of long-run relationship in all 10 sectors.

Long-run estimation results

Table 3 shows the long-run estimation results of employment equation (8) by sector.

Table 3. Long-run estimation results

Regressor	AGR ^D	CON ^V	DIS ^D	FIBU ^V	GOV ^D	MANNO ^F	MINOTH ^D	OTHS ^F	TRACOM ^V	U ^A
gva	0.93 ^a	1.06 ^a	0.27 ^a	0.65 ^a	1.28 ^a	0.27 ^c	0.23 ^a	1.11 ^a	0.65 ^a	0.42 ^a
w	−0.78 ^a	−0.56 ^a	−0.38 ^a	−0.52 ^a	−0.26 ^a	−1.53 ^a	−1.61 ^a	−0.60 ^a	−0.85 ^a	−0.49 ^a
C	0.92	0.60	7.87 ^a	3.51 ^a	−6.09 ^a	16.97 ^b	21.92 ^a	1.47	−7.54	5.68 ^a

Notes: ^D DOLS is used, ^F FMOLS is used, ^V VECM is used, ^A ARDL is used; Significance levels: a – 1%, b – 5%, and c – 10%; C denotes Intercept ; The following pulse dummies are used for the following sectors: DP2007 for TRACOM, DP2008 for CON, DP2009 for OTHS, DP2012 for TRACOM and U, DP2013 for TRACOM, DP2014 for OTHS and U, DP2015 for CON. DPXXXX takes 1 for the given XXXX year and 0 otherwise; Time trend was statistically significant only in AGR with the coefficient of 0.03.

The signs of the estimated coefficients are positive for value-added and negative for wages, in line with the theoretical expectations discussed in Section 4. The results show that value-added and wages have statistically significant long-run impacts on employment in all sectors. The impact of value-added ranges from 0.23 for non-oil mining to 1.28 for government services. For wages, results likewise vary across sectors and are lowest (in absolute value) for government services (−0.26) and highest (in absolute value) for the non-oil mining sector (-1.61).

Short-run estimation results

The short-run estimation results (i.e. the final ECM specifications) are reported in Table 4. The estimated elasticities are all statistically significant. Additionally, all the post-estimation tests results are in favour of the validity of the selected final ECM specifications. Moreover, the estimated parameters, including SoA coefficients, have the expected signs.

Table 4. Final ECM specifications by sector

Variable	AGR	CON	DIS	FIBU	GOV	MANNO	MINOTH	OTHS	TRACOM	U
$EC{T}_{t-1}$	−0.568^a	−0.260^a	−0.966^a	−0.485^b	−0.198^b	−0.130^a	−0.134^a	−1.441^a	−0.095^a	−0.348^a
$C$	0.035^a	−0.016	0.033^a	0.003	0.007	0.021	0.034^a	0.001	0.047^a
$\mathrm{\Delta }e{t}_{t-1}$	−0.169^a					0.109 ^c
$\mathrm{\Delta }e{t}_{t-2}$		−0.232^b	0.435^a
$\mathrm{\Delta }gva$	1.984^a		0.190 ^c		0.686^a			0.759 ^c		0.550^a
$\mathrm{\Delta }gv{a}_{t-2}$			0.229^b
$\mathrm{\Delta }w$	−0.785^a	−0.699^a	−0.534^a	−0.393^a		−0.208^b	−0.189^b	−0.456^a	−0.087 ^c	−0.190^a
$\mathrm{\Delta }{w}_{t-1}$			−0.134^a			0.122
D(DP2008)		−0.152^a
DP2003		−0.271^a						0.073^b
D(DP2009)		−0.145^a						−0.074^b
DP1983						−0.399^a
DP1984						−1.032^a
DP2013			−0.091^a	−0.162^b	0.075^a
DP2012									−0.139^a
DP2008					0.061^a
DP2007									0.109^a
DB1213										−0.211^a
DP2009										0.144^b
DB1011						−0.348^a
DB0910							−0.188^a
DB1415							0.172^a			0.204^a
DP2001						0.182^b
DP2002							0.122^a
DB1314									−0.082^a
R^2	0.974	0.931	0.986	0.921	0.819	0.939	0.919	0.724	0.856	0.926
Adjusted R^2	0.967	0.897	0.976	0.901	0.776	0.921	0.894	0.618	0.801	0.898
SER	0.027	0.039	0.016	0.043	0.014	0.070	0.033	0.038	0.025	0.049
SC	0.282 (0.759)	0.849 (0.457)	1.184 (0.350)	0.908 (0.427)	2.607 (0.107)	0.572 (0.572)	0.726 (0.501)	1.247 (0.325)	0.345 (0.716)	0.391 (0.686)
HET	0.089 (0.985)	1.096 (0.418)	0.555 (0.778)	0.599 (0.625)	1.260 (0.324)	0.644 (0.734)	1.470 (0.254)	0.923 (0.497)	0.506 (0.767)	0.627 (0.706)
JB	2.907 (0.234)	0.552 (0.759)	2.864 (0.239)	0.470 (0791)	5.720 (0.057)	4.915 (0.086	0.441 (0.802)	0.274 (0.872)	0.858 (0.651)	2.114 (0.348)
RR	0.376 (0.551)	1.044 (0.329)	0.102 (0.757)	0.595 (0.454)	0.250 (0.624)	2.681 (0.114)	2.414 (0.141)	0.161 (0.695)	0.000 (0.991)	2.275 (0.157)

Notes: SER = Standard Error of Regression; SC = Breusch-Godfrey Serial Correlation LM Test; HET = White test for Heteroscedasticity; JB = Jarque–Bera normality test; RR = Ramsey’s Mis-specification test; p-values are in parentheses; d = difference operator.

VII. Discussion of the findings

Our results indicate that value-added and wages have statistically significant impacts on employment for all 10 sectors in the long run. As Table 3 presents, we found elastic and inelastic long-run elasticities for employment with respect to both value-added and wages across the sectors, while all the short-run elasticities are inelastic (with one exception of value-added for agriculture), as shown in Table 4. This corroborates the economic theory that impacts are usually expected to be higher in the long-run than in the short-run. Generally, our empirical findings can be explained through the Hicks–Marshall laws of derived labour demand due to scale effect and substitution effect as articulated in the labour demand theory (see e.g. Ehrenberg and Smith 2016, Chapter 4). We also note that employment responds differently across sectors, primarily due to two factors: the idiosyncrasies of a given sector and its stylized features in the Saudi economy.

Four sectors – agriculture, construction, government services and other services – exhibit a nearly one-to-one relationship between employment and value-added in the long run. For these sectors, a 1% increase in value-added will result in an equivalent rise in employment.

Government services and non-oil mining exhibit the highest and the lowest value-added elasticities, respectively. We offer three explanations for the former. First, the government sector is heavily labour intensive. Hasanov et al. (2020) estimated that the labour and capital elasticities of value-added are 0.48 and 0.21, respectively, in the production function of the government service sector in Saudi Arabia for the period 1996 to 2016. The labour elasticity being more than twice higher of the capital elasticity indicates that the sector’s activity is mostly driven by labour. Hence, any changes in sector’s activity or output will be associated with corresponding changes in sector’s employment. Second, as in other developing economies, the government service sector exhibits a room for efficiency improvement in Saudi Arabia. Empirical studies, such as Al-Faris (2002), Joharji and Starr (2010), Alshahrani and Alsadiq (2014), and Eid and Awad (2017) highlight the inefficiencies in various government services. Saudi Vision 2030 (SV 2030), the Kingdom’s masterplan for development, particularly the Fiscal Balance Program includes numerous initiatives to increase government efficiency (FBP 2017). Additionally, the SV 2030 Human Capital Program identifies increasing government efficiency and providing world-class government services as its key priorities. Such inefficiency means that additional government services activity requires more than the optimal number of government employees and thereby, elastic output elasticity of employment. Third, Saudis prefer working in government sector to private sector. Official statistics show that on average, Saudis comprised only 13.2% of private sector employment from 2005 to 2016 (SAMA 2019). Additionally, an official labour survey conducted in 2016 shows that the public sector accounted for 67% of total employment of Saudi Nationals (SLMR 2016). According to Ministry of Civil Service via SAMA (2019), share of nationals in the government-sector employment increased from 78.9% in 1995 to 94.6% in 2016 with the period average of 90.1%. All these numbers show that nationals are heavily dominate in GOV sector and given that Nitaqat policy encourages haring and maintaining a high level of nationals, expanding government activities will be associated with more increase in the sector’s employment.

For non-oil mining, we attribute the observed low value-added elasticity of employment and the highest wage elasticity of employment (see Table 3) with the commonly accepted fact that mining is a less labour-intensive economic activity. This implies that factors other than employment, such as capital and total factor productivity, drive the development of the sector and that its expansion will be associated more with such other variables, and also that the labour absorption capacity of the sector is limited. This general observation is also the case for Saudi non-oil mining sector. It is worth considering that the average shares of non-oil mining employment in total and non-oil employment were 1.4% and 1.5%, respectively, and even, the maximum number of the shares were 1.9% for the studied period of 1995–2016 according to the available statistics (SAMA 2019). Although we could not find any study that estimated labour elasticity of output in the non-oil mining sector in Saudi Arabia, which would provide an information about how labour is important in the sector’s activity, calculation using available data shows that labour share (employment compensation) in value added declined from 37.3% to 20.8% in 2016 with the period average of 25.5%.⁴ These statistics point out two main things. First, low labour share indicates that production factors other than employment such as capital play important roles in the activity of non-oil mining sector. Resultantly, any changes in the sector’s activity will be associated with more changes in the other production factor, say capital than in employment. Simply because this is not labour that mainly drives sector’s activity and consequently inelastic output elasticity of employment originates. Second, declining labour share, which implies growing capital share, shows that employers in the sector has substituted labour with capital over time. This is consistent with the theoretical articulation emphasizing that substitution effect, which usually happens in the long run is one of the reasons for obtaining elastic wage elasticity of employment (see e.g. Ehrenberg and Smith 2016, Chapter 4). The intuition here is that since sector’s activity/output is mostly determined by capital not employment, employers in the sector will respond to wage increases by firing more employees and tending to replace them with capital over time. Usually, the mining sector is capital-intensive, and activities are driven mainly by advanced technologies that require specialized skilled labour. As a result, wages level in the sector are usually higher than that in other sectors. This common sense understanding of the sector is also true for the Saudi Arabian case: the average wage level in the sector comes the second highest after the government sector during the studied period 1995–2016. Thus, a capital-intensive sector with higher wages, although labour share is not high, can encourage employers to substitute employees with capital in the long run when they face wage increases. This would be another possible explanation for the elastic wage elasticity of employment in the non-oil mining sector as it is theoretically articulated (see, e.g. Ehrenberg and Smith 2016, Chapter 4). Note that other studies for other countries’ mining sector also found elastic wage elasticity of employment. For example, Hamermesh (1996) reports the elasticity for the British coal mines to be in the range of −1.0 and −1.4.

Regarding the long-run impact of real wages on sectoral employment, we find it interesting that sectors, which are less driven by value-added, namely non-oil mining and non-oil manufacturing, they are the mostly sensitive to wages (see Table 3). Precisely speaking, a 1% increase in real wages decreases employment in these two sectors by around 1.5% and 1.6%, respectively. Commonly, manufacturing is labour intensive, and the sector exhibits a tight relationship between employment and value-added. However, our estimated value-added elasticity of employment is small in magnitude (0.27) and this indicates that the common sense understanding of the sector might not be the case in Saudi Arabia. Indeed, this is supported by previous research: Alkhateeb et al. (2017c) argue that the Kingdom’s industry sector heavily depends on capital-intensive technology and Hasanov et al. (2020) estimate the employment elasticity of value-added being 0.23 for the country’s non-oil manufacturing sector for the period 1995–2016. These clearly show that the role of employment in the non-oil manufacturing activity in Saudi Arabia is considerably small and consequently, wage plays more role than output in the formation of the sector’s employment. Therefore, when wage rate in the sector is raised, then the employers in the sector can decrease their demand more than what the wage increase is because this is not employment but capital and other production factors that mostly drive the sector’s activity. The substitution effect between labour and other factors of production would be another explanation for why labour demand is more responsive to changes in real wages in the non-oil manufacturing sector. The discussion above highlighting the limited role of labour in the sector’s activity increases the probability that the substitution effect prevails in the sector. Besides, inspecting the development paths of wages, employment and value added in the sector shows that increases (decreases) in the former one were associated with declines (increases) in the middle one, but these did not lead to declines (increases) in the latter one. For example, real wages declined after 1999 till 2001, while employment increased considerably in the same period whereas value added did not demonstrate any noticeable changes. Or real wage increased noticeably in 2008 and 2010 while employment declined in both years, but value added did not change its trend. These may indicate that substitution effect was prevailing while scale effect, articulating that declines in employment due to the high wages lead to declines in output, was less likely the case for the non-oil manufacturing sector. Another explanation for the elastic wage elasticity of employment in the sector is the following. Protective government policies and unions’ action lead to inelastic wage elasticity of employment as discussed in the labour market literature (see, e.g. Ehrenberg and Smith 2016, Chapter 4). In this regard, recall that the Saudi national employment policy, Nitaqat aims for encouraging and maintaining high levels of Saudi employment and indirectly discouraging hiring foreigners in the sectors of the economy.⁵ Non-oil manufacturing sector is heavily dominated by foreign labour. Precisely speaking, Table 2 above reports that 77% of the total employment in the sector is foreigners, which is a quite high number. This indicates that the flux of foreign labour in the sector is quite high. Thus, given that Nitaqat is not for foreigners and they are dominant in the non-oil manufacturing labour market, the employers in the sector can cut off foreign labour more than what the wage increase is.

Among the ten sectors, government services exhibit the least sensitivity of employment to wages, with elasticity of −0.26. This can be explained by the above-mentioned preference of Saudi nationals to work in the government. They often prefer a public sector job regardless of the wage level, probably due to the job security, high level of wage compared to other sector as well as other allowances and benefits/preferences of being a government employee. Another fact that more than 90% of the government sector’s employment are nationals and Nitaqat policies aim for encouraging and maintaining high levels of Saudi employment. Combination of these two implies that the sector would not fire employees mainly nationals at a greater extent if wages are reduced say due to economic recession or low oil price and thus low government revenues

Before concluding the discussion of long-run findings, it is worthwhile to underscore that employment levels in non-oil manufacturing and non-oil mining are found be the least value-added driven and the most wage sensitive. Together with agriculture, these sectors comprise the bulk of Saudi Arabia’s non-oil tradable production. Development of the non-oil tradable industries can help the Kingdom avoid the so-called ‘Dutch disease’ of overreliance on a single sector and boost the SV 2030 export-led growth strategy; thus, these three sectors deserve more detailed research in future.

Turning to the short-run estimations reported in Table 4, we highlight the following results. First, SoA coefficients appear statistically significant and negative, as expected, for all ten sectors. This indicates that the long-run relationships of employment to value-added and wages are stable over time and the impacts of shocks or changes to the corresponding long-run relationships are temporary and do not create permanent deviations. Additionally, the statistically significant negative SoA coefficients imply long-run causality running from value-added and wages to employment. Moreover, the SoA coefficients for all sectors, with the exception of other services, fall within the range of (0;-1). Thus, for these nine sectors, employment takes more than one year to correct from short-run disequilibrium back to long-run equilibrium, while requiring less than a year in the case of other services. (For the details of correction and overcorrection processes, see Loayza and Ranciere 2005; Enders 2015, pp. 374, 377–378; Shittu, Yemitan, and Yaya 2012; Olczyk and Kordalska 2017.)

Second, wage growth has statistically significant contemporaneous negative impact on the employment growth in all sectors, except for the government service sector. This indicates that wages play an important role in shaping employment growth outside the public sector. As for government services, we do not find any contemporaneous or lagged effects of wage growth to be statistically significant, in line with the above findings that the sector exhibits the least long-run sensitivity of employment to wages. We again believe this stems from the preference of Saudi nationals for government employment.

Third, value-added growth has statistically significant positive impact on the employment growth only in five sectors: agriculture, distribution, government services, utilities, and other services. Also, these positive effects happen contemporaneously. By combining the long-run and short-run effects of value-added on employment across the sectors, we see that the majority of these five sectors also have quite high value-added elasticity in the long-run. This implies that employment in these sectors is responsive to value-added/activities probably caused by their nature. We do not find any statistically significant positive influence of value-added growth on employment growth of the remaining five sectors. Finally, among the ten sectors, agriculture exhibits the most sensitivity in employment growth to value-added and wages, according to the short-run estimation results in Table 4.

VIII. Conclusions and Policy Recommendations

Policymakers in any economy benefit from an advanced understanding of employment dynamics as they design economic programmes and conduct relevant monitoring and evaluations. Sustainable economic growth also requires the promotion of a healthy and stable employment market. Therefore, the determinants of employment dynamics have been investigated extensively in the literature.

Saudi Arabia has set various targets for economic developments under SV 2030. These include lowering the unemployment rate from 11.6% to 7% and increasing women’s participation in the workforce from 22% to 30% by 2030 (Saudi Vision 2030). To successfully achieve these goals and to better understand employment dynamics in general, employment should be modelled as a function of its determinants. Hence, this study investigates sectoral employment in Saudi Arabia in the long and short run utilizing cointegration and ECM modelling techniques.

Our study provides policymakers with empirical insight into how employment can be impacted via its identified determinants – output and wages – in different sectors of the Kingdom’s economy in both the long and short run. The results can also help authorities balance employment growth across sectors. Because the magnitudes of the impacts of output and wages on employment are different in each sector, our findings highlight the value of designing and implementing employment policies at the sectoral level rather than ‘one size fits all’ aggregate level policy.

The government influence on sectoral employment dynamics in the private sector may be relatively limited but it can still play a role, primarily by impacting output and wages. First, the government can create additional demand for domestically produced goods and services, both directly and indirectly. For example, by building new schools or bridges, it increases demand directly for construction services and indirectly for the manufacturing and mining products and services used in the construction. Second, the authorities can promote substitution of imported goods and services with domestic ones. These measures will raise demand in targeted sectors, thereby encouraging greater production and revenue; resulting increase in economic activity in turn will boost employment.

This would be also a desirable contribution to local content, which is one of the main elements of the economic diversification in Saudi Arabia. Even if the government can do it only for limited number of sectors, this can create a multiplier effect, i.e. spread out to other related sectors in a simplistic Keynesian view. This measure will increase demand for the sectors’ goods and services and consequently their output and employment will raise. The process will continue until resources, mainly labour and capital, will rich to their full utilization.

The above measures would improve demand for goods and services for all ten sectors we examined, thereby enhancing sector-level employment in the long run. However, in the short run, the effectiveness varies across sector. Agriculture and other services display the highest short-run sensitivity to the stimulus, followed by government services and utilities. The distribution and retail sector exhibits less short-run promise. The other sectors indicate potential for favourable impact in the long run only. It appears intuitive that from the above-discussed measures (i.e. creating employment through income/economic activity) standpoint, agriculture and other services sectors seem to be more promising, public service and utilities sectors are seemed to be promising, the distribution sector is seemed to be less promising in the short-run. While, construction, FIBU, non-oil manufacturing and mining as well as transport and communication sectors does not seem favourable in the short run. However, over the long run all the mentioned unfavourable sectors also become favourable. It is quite reasonable to think that the government investment demand for goods and services produced locally can lead to the growth of the sectors in the long run, in particular through infrastructure development, efficiency and productivity increase, which all would increase employment.

Turning to the other examined determinant of employment, labour is a primary expense for the private sector and higher wages will reduce profit. Accordingly, private businesses will increase wages only if they have additional demand for their output requiring additional employment although there are different conceptual options to increase production (see, e.g., Hasanov 2020 and refereces there in). The government can further increase employment by inducing more of the working age population to participate in the labour force. This would increase the supply of labor, which may lower the market wage rate, leading to an increase in the demand for labor. Additionally, some relevant measures are outlined in Nitaqat policy to support employment of Saudi nationals. Also, increasing employment-subsidy for Saudi employees financing it by expat levy collections would increase labour force participation rate of Saudis. The experience of Kuwait and other GCC countries should be considered for implementations of such policies. Moreover, the government can impact the determinants of the real wage, such as productivity and price levels to increase employment in the sectors. Given that the energy price reform (EPR) is on the policy agenda until 2023 as highlighted in the Fiscal Balance Program, there is a limited room to play with the price tool. Instead, the government, by using revenues from EPR and other reforms, may wish to create positive externalities by further improving economic and social infrastructure, supporting innovation projects, and investing human capital, which all lead to an increase in productivity of the sectors in the long run. Note that wage-oriented measures would be the most impactful for MANNO and MINOTH sectors, the least impactful for GOV and DIS sectors and moderately impactful for the rest sectors in the long-run. Regarding the effects of the wage-oriented policy measures in the short-run, mainly AGR, CON, DIS and OTHS sectors will be affected.

Disclosure confilts of interest Statement

No potential conflict of interest was reported by the author(s).

Acknowledgments

The authors would like to thank to reviewers for their valuable comments and the participants of the 22^nd Dynamic Econometrics Conference. We are indebted to David F. Hendry and Jennifer Castle for their constructive comments and suggestions on the KAPSARC discussion paper version of this manuscript. This is a joint research project between Saudi Central Bank and KAPSARC. The views expressed in this paper are those of the authors and do not necessarily represent the views of their affiliated institutions.

Supplemental Material

Supplemental data for this article can be accessed here.

Data availability statement

The data that support the findings of this study are available from the corresponding author, Muhammad Javid, upon reasonable request.

Notes

¹ There are studies analysing the link between economic growth and unemployment in Saudi Arabia (e.g. Al-Habees and Rumman 2012; Khrais and Al-Wadi 2016; Abou Hamia 2016). Moreover, a few studies (e.g. Alotaibi 2017) also describe labour market in Saudi Arabia without doing quantitative assessments. We do not consider these studies here as they are not in line with the objective of our research.

² IMF Country Report No. 13/230

³ In some cases, to get the same span of 1995–2016 data the values of the earlier years for the variables are interpolated using backcasting techniques

⁴ For the methodologies to calculate labour share see Schneider (2011), Giandrea and Sprague (2017), inter alia.

⁵ For example, the employers, who have more foreigners than national have to pay higher amount of expatriate levy (FBP 2017).