Why Is Tunisia’s Unemployment So High? Evidence From Policy Factors

Aymen Belgacem; Jérôme Vacher

doi:10.5089/9798400257025.001.A001

Jump to Content

Why Is Tunisia’s Unemployment So High? Evidence From Policy Factors

Author:

Aymen Belgacem

Aymen Belgacem
Search for other papers by Aymen Belgacem in
Current site
Google Scholar

and

Jérôme Vacher

Jérôme Vacher
Search for other papers by Jérôme Vacher in
Current site
Google Scholar

Publication Date:: 27 Oct 2023

eISBN:: 9798400257025

Language:: English

Keywords:: labor market deregulation; labor market dynamics; labor market demand; business regulation; labor market distortion

Download PDF (1.9 MB)

Tunisia has one of the highest unemployment rates within the Middle East and Central Asia. We look at the extent to which institutional factors explain those high unemployment levels. We also assess unemployment cyclicality, by looking at the determinants of labor market sensitivity to the output gap. We find that during the last decade the deterioration of institutional factors that affect labor demand explain not only about a quarter of the unemployment rate increase in Tunisia, but also Tunisia’s excess sensitivity of unemployment to the output gap. Our results suggest that an improved business environment and product market competition, increased labor market flexibility as well as reduced financial constraints and informality would help reduce Tunisia’s unemployment.

Abstract

I. Introduction

With unemployment reaching 16.2 percent in the first quarter of 2023, Tunisia has one of the highest unemployment rates within the Middle East and Central Asia region, with substantial disparities related to gender, age, and regional dimensions (Figures 1.1 to 1.4). The young, recent graduates and households falling into the lowest income deciles are more likely to be unemployed. For the youth between 15 and 29 years, unemployment also appears negatively correlated with education levels and is more pronounced for males than for females (World Bank, 2023). These relatively persistent disparities – each with significant social impact- are indicative of the challenges that the Tunisian economy faces, not only to absorb its young and growing labor force, but also to provide new opportunities for its stock of long-term unemployed. This calls for an urgent and adequate response to tackle Tunisia’s high and persistent unemployment.

On the labor supply side, Tunisia is suffering from structural bottlenecks. This includes the stark duality of its labor market, with informal employment accounting for more than 60 percent in some sectors (Figure 1.5). Informal employment is strongly associated with the level of education, as informality seems prevalent among non-graduates (IMF, 2022), with 80 percent of the informal workforce having only reached a primary or a secondary education level, and 10 percent being illiterate (World Bank, 2014). Self-employment of the young is also common and informal, especially in rural areas (World Bank, 2014). On the other hand, the large cohorts of university graduates tend to be unemployed rather than working in the informal sector. As the wage grid and social benefits in the public sector on average tend to be higher than for the equivalent positions in the private sector, especially for entry level positions, 78 percent of new graduates would also rather wait for a public sector job, rather than accept a job in the private sector (TLMPS, 2014). The existence of this “wage premium for public jobs” of about 40 percent would provide incentives to “queue” for public sector jobs and raise the reservation wage in the formal private sector (World Bank, 2023).

On the labor demand side, several distortions are potentially at play. These include the low productivity and competitiveness of the Tunisian private sector, which has resulted in a low potential growth during the last decade. The manufacturing and service sectors (19 and 52 percent of the labor force, respectively) continue to lead job creation and employment. However, they are far from being the most productive sectors. Conversely, sectors with potentially a higher productivity are either regulated or dominated by state owned enterprises, which does not help to develop the private sector and thus employment (figure 1.8). In addition to the low productivity, institutional factors have also been identified by the literature as potentially having a significant effect on the dynamics of the labor market. These factors include the employment protection legislation, wage bargaining system, product market regulation, access to finance and taxation.

The aim of this paper is to address to what extent those institutional factors explain the unemployment level and its evolution in Tunisia during the last decade. Furthermore, as Tunisia is a small open economy, with a high share of international trade and a sizeable informal sector, we also look at unemployment cyclicality, by identifying the determinants of the labor market sensitivity to the output gap^². Assessing the impact of institutional factors on the Tunisian labor market is important for two main reasons:

First, as Tunisian labor regulations have not changed since the 1990s, and with institutions potentially explaining differences in employment dynamics (Blanchard and Wolfers, 2000), stepping back and assessing the impact of institutional factors on employment in the current context would not only be useful to address challenges that the Tunisian labor market faces, but also novel in the case of Tunisia.
Second, our analysis seeks additional insights on the linkages between different types of regulations and policies, and thus a broader view on potential solutions in the context of high and persistent unemployment.

For Tunisia more specifically, among different structural and policy factors that could affect the labor market, the literature has focused more on the linkages between unemployment and supply related factors, including skill mismatches (Boughzala, 2017; Assad and Boughzala, 2018; OECD, 2021) and wage subsidies (e.g., Marouani, 2010; WB, 2014). The literature has also examined, to a lesser extent, the role of labor market regulation in influencing unemployment trends (Boughzala, 2017; EMNES, 2018).

To our knowledge, this is the first study that attempts to fill the gap in the literature by exploring the relationship between the Tunisian labor market and a broad set of institutional factors simultaneously. Specifically, we aim at measuring the impact of demand-side institutional and policy distortions on Tunisia’s high unemployment during the last decade. We also explore the linkages and interactions between factors, rather than studying them separately, in a more comprehensive approach.

The remainder of the paper is as follows: the second section gives an overview of the literature on the impact of institutional factors on the labor market. Section 3 assesses the relative institutional factors’ contribution to the unemployment in Tunisia during the last decade. Section 4 presents to what extent institutional factors explain unemployment’s sensitivity to business cycles in the Tunisian context.

^{Figures 1.1 to 1.8:}

Tunisia: Labor Market Indicators

Citation: IMF Working Papers 2023, 219; 10.5089/9798400257025.001.A001

II. Literature Review

Economic studies have looked at several demand-side institutional determinants of unemployment, with the main ones being labor market regulation, including wage bargaining system, product market regulation and access to finance:

▪ Employment protection legislation (EPL): Finding an equilibrium between protecting workers and the efficiency of the labor market is important for supporting growth and enhancing labor market dynamics. EPL is generally used to ensure job protection, increase stability while improving working conditions, and protect workers’ welfare (OECD, 2013). EPL protects workers from firing and gives insurance against dismissal, especially in the presence of financial market imperfections (Pissarides, 2010). It can also help managers invest in human capital and thus enhance firm productivity (Hatayama, 2022). However, when it is excessive, EPL can limit employers’ ability to efficiently adjust the workforce, thereby causing labor market distortions. This is more pronounced in industries characterized by rapid technological change. Furthermore, excessive EPL could increase the cost of formal employment, which in turn increases informality and unemployment (Boughzala, 2017).

▪ Wage bargaining system: The literature has identified a hump shaped relationship between unemployment and the degree of wage bargaining centralization (Calmfors and Driffils, 1988; Bassanini and Duval, 2006; Jaumotte, 2011). In a centralized system, unions internalize better the negative impact of excessive wage pressures on employment. On the other hand, the small bargaining power of unions in a fully decentralized negotiation system, coupled with market discipline, could reduce wage claims, thus helping preserve employment. In such circumstances, the intermediate bargaining system would tend to display the worst outcomes. First, as it does not benefit from both advantages of the aforementioned systems (Bassanini and Duval, 2006). Second, as the strong bargaining power of unions tend to increase wages above productivity, resulting in poor employment performance. Third, with negotiations taking place at the sectoral level, competition in product markets can be limited, as negotiations tend to lead to more harmonized wages at the sectoral level and the competition between firms offering close substitutes cannot be at play (European employment observatory, 2011). Hence, the intermediate bargaining system, coupled with a stringent employment protection legislation could constitute an entry barrier, hampering a more rapid development of the private sector.

▪ Product market regulation: The impact of product market regulation on the labor market is ex ante unclear. On the one hand, a lighter product market regulation has a positive effect on employment as it (i) lowers entry barriers, boosts activity level and employment demand, (ii) boosts real wages via lower prices, which stimulates labor supply and (iii) lowers market rents which reduces wage claims and therefore the gap between productivity and real wages. On the other hand, a high level of competition may reinforce the bargaining power of employees, leading to an increase in employment costs and consequently the unemployment rate (Jaumotte, 2011).

▪ Access to finance: Interactions between labor and financial markets imperfections have also an impact on employment. In a market with labor imperfections, i.e., hiring and firing costs as well as weak flexibility, financial development allows firms to finance labor adjustment costs more efficiently as (i) firms become less dependent on internal financial sources, and (ii) financial development could lead to more competition in the credit market (Gatti et al 2012, Rendon, 2001). On the other hand, when financial imperfections exist, making the labor market more flexible has a strong impact on employment, as the removal of labor adjustment costs would free firms from financing them (Rendon, 2001).

▪ Taxation: The impact of taxation on unemployment rates is ex ante ambiguous: In a perfectly competitive labor market, a tax wedge increase would be fully transmitted into lower net wages, leaving the unemployment rate unchanged (Bassanini and Duval, 2006). However, in countries with powerful trade unions and intermediate wage bargaining system, coupled with a stringent labor market regulation, higher taxes would be transmitted into higher real wages, which in turn would lead to a decrease in labor demand. The literature is more conclusive about the impact of taxation on informality, as it shows that an increase in the tax wedge generates disincentives to work formally, amplifying further the dual market problem (Bassanini and Duval, 2006). IMF (2022), for instance, points out that the reduction in corporate tax rates (from 45 to 22.5 percent) as well as the automation of tax collection and declaration have partly contributed to the decrease of informality in Egypt between 2005 and 2017.

III. The Contribution of Policy and Institutional Factors in Explaining Tunisia’s Unemployment Rate

These policy and institutional determinants are likely to be important factors in the evolution of the labor market in Tunisia. Drawing on the recent literature, this section assesses to what extent those factors explain the unemployment level during the last decade.

First, we focus on a panel of 157 countries during the period 2007–2017^³ and then try to extract the impact of policy factors on the Tunisian labor market. Specifically, our first step is to estimate a panel regression including, in addition to the lagged unemployment rate (to control for unemployment persistence), the output gap (denoted “Gap”, the difference between the real GDP and the HP filter of the real GDP^⁴, as percent of the latter), a proxy of labor market regulation (“Labor”), the degree of centralization in wage bargaining (“Barg” and its square, to allow for nonlinearity as explained in section II), business regulation (“Bus”), financial market regulation (“Fin”) and taxation (“Tax”). As the small-time dimension used would lead to potential biases in estimation (especially in the presence of a lagged variable), we also use the Arellano-Bond GMM dynamic model with one lag instrumental variables.^⁵ :

\begin{matrix} U_{i, t} = α U_{i, t - 1} + β G a p_{i, t} + γ L a b o r_{i, t} + δ B u s_{i, t} + θ_{1} B \arg_{i, t} + θ_{2} B \arg_{i, t}^{2} + σ F i n_{i, t} + ϑ T a x_{i, t} + μ_{i} + τ + ε_{i, t} & (1) \end{matrix}

Where μ_i and τ represent country and time fixed effects, respectively. Details of the data used are in Annex 5. Table 1 presents several specifications including different variants of the variables below, extracted from a wide range of sources to check robustness^⁶. Some of the estimations also include sub-indicators to fine tune the results and allow for the identification of policies having a significant impact on the unemployment rate^⁷.

Estimation results show that policies and regulations related to the labor market, the product market, and financial development seem to significantly affect the unemployment rate:

A greater flexibility in the product market is associated with less unemployment. Looking more specifically at labor market policies however, we find that it is mostly firing procedures and costs that are having a significant impact on the unemployment level: specifications 2 to 5 show that the latter increases with augmented severance payments, or more stringent dismissal procedures. Product market flexibility including through competition intensity, combined with less burdensome administrative costs are associated with a lower unemployment rate.
The flexibility of credit market regulations seems also to affect the unemployment rate significantly and negatively.
However, results do not show a significant direct effect of wage bargaining on the unemployment level. As lagged unemployment – which reflects the persistence of the unemployment – is significant, wage bargaining could affect indirectly the current unemployment level. This would be by increasing the persistence of unemployment, given that insiders (currently employed) could directly influence the wage bargaining system and thus, lead to an increase in the level of the natural unemployment rate (O’Shaughnessy, 2011; Blanchard and Summers, 1987).
Estimation results also do not show any significant impact of the level of taxation on unemployment. With the latter considering employment in the informal sector, the literature shows that a burdensome tax system increases the cost of formal work, which in turn could affect informality, rather than the unemployment level^⁸ (IMF, 2022 for North Africa).

For Tunisia, while the output gap and the persistence of unemployment explain most of the increase in the unemployment rate between 2009 and 2017, changes in institutional policy factors explain about a quarter of it:

Based on specification 4 and 6 (Table 1 and figures 2.1 and 2.2), about 20 percent^⁹ of the increase in the unemployment rate in Tunisia during the last decade would be attributable to the deterioration of product markets and 5 percent is due to the credit market conditions. During the same period, the de jure employment protection legislation (ILO indicator) is unchanged, which explains its lack of impact on the unemployment change in the estimation.
The rest of the increase in unemployment would mostly be explained by the output gap, as well as the persistence of unemployment^¹⁰.
The importance of the lagged unemployment contribution in the model goes in tandem with the hypothesis of relatively long-lasting persistence in unemployment caused by economic shocks. Blanchard and Summers (1987) distinguish between two groups of workers: insiders (currently employed) and outsiders (unemployed). With unemployment increasing, the number of insiders would diminish, and the number of outsiders increase. However, the bargaining system would play an important role in maintaining the gap between both groups, and hence an increase in the natural unemployment rate.
the impact of institutional factors on the increase of unemployment in Tunisia between 2009 and 2017 is more pronounced for males than for females and youths (Figures 2.6 to 2.8). This suggests that female and youth employment in Tunisia are more influenced by economic cycles.

^{Table 1:}

Institutional determinants of unemployment

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

^{^1/}Sources of the data are in parentheses. All the indicators above, except the employment protection legislation indicators from the ILO, are given in an ascending scale in terms of efficiency/ flexibility: The higher the indicator is, the more flexible/efficient the market is.

^{Table 1:}

Institutional determinants of unemployment

		(1)	(2)	(3)	(4)	(5)	(6)
	VARIABLES ^1/	Unemployment rate
	Lagged Unemployment	0.5664*** (0.0376)	0.6030*** (0.0335)	0.6063*** (0.0398)	0.5817*** (0.0407)	0.5721*** (0.0450)	0.6947*** (0.0343)
	Output gap	-0.0769*** (0.0207)	-0.0765*** (0.0168)	-0.0800*** (0.0186)	-0.0767*** (0.0180)	-0.0751*** (0.0181)	-0.0526*** (0.0131)
Employment protection legislation	Employment protection legislation (ILO)	-0.0390 (0.0278)
	Procedural requirements for dismissals (ILO)		0.0370*** (0.0056)	0.0449*** (0.0045)	0.0427*** (0.0055)	0.0440*** (0.0042)
	Severance payments (ILO)		0.0101 (0.0063)	0.0112** (0.0053)	0.0090 (0.0057)	0.0105* (0.0054)
	Flexibility of hiring and firing regulations (Fraser)						0.0010 (0.0007)
Collective bargaining system	Centralized collective bargaining (Fraser)	-0.0006 (0.0043)	0.0012 (0.0045)		0.0007 (0.0047)
	Square of Centralized collective bargaining	0.0001 (0.0004)	-0.0001 (0.0004)		-0.0001 (0.0004)
	Flexibility of wage determination (WEF)			-0.0063 (0.0084)		-0.0111 (0.0082)	-0.0162** (0.0080)
	Square of flexibility of wage determination			0.0004 (0.0009)		0.0009 (0.0009)	0.0014* (0.0008)
Business/Product market regulation	Administrative requirements (Fraser)	-0.0023** (0.0011)	-0.0023** (0.0010)	-0.0021* (0.0011)
	Regulatory Burden (Fraser)	-0.0010** (0.0004)	-0.0009** (0.0004)	-0.0009** (0.0004)
	Goods market efficiency (WEF)				-0.0110*** (0.0032)	-0.0107*** (0.0033)	-0.0069*** (0.0025)
	Intensity of local competition (WEF)						0.0008 (0.0021)
Taxation	Tax burden (Heritage)	0.0001 (0.0002)	0.0001 (0.0002)	0.0001 (0.0002)	-0.0000 (0.0002)	-0.0000 (0.0002)	0.0001 (0.0001)
Financial markets	Financial market development (WEF)	-0.0036 (0.0024)				-0.0021 (0.0024)
Financial markets	Credit market regulations (Fraser)		-0.0018** (0.0008)	-0.0016* (0.0009)	-0.0015* (0.0009)		-0.0021*** (0.0007)
	Constant	0.0728** (0.0327)	0.0335 (0.0250)	0.0560* (0.0287)	0.0765*** (0.0277)	0.1054*** (0.0333)	0.1012*** (0.0267)
	Observations	745	769	750	752	754	1,376
	R-squared	0.5908	0.5990	0.6026	0.6016	0.6015	0.6592
	Number of c_id	91	93	91	91	91	138

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

^{Table 1:}

Institutional determinants of unemployment

		(1)	(2)	(3)	(4)	(5)	(6)
	VARIABLES ^1/	Unemployment rate
	Lagged Unemployment	0.5664*** (0.0376)	0.6030*** (0.0335)	0.6063*** (0.0398)	0.5817*** (0.0407)	0.5721*** (0.0450)	0.6947*** (0.0343)
	Output gap	-0.0769*** (0.0207)	-0.0765*** (0.0168)	-0.0800*** (0.0186)	-0.0767*** (0.0180)	-0.0751*** (0.0181)	-0.0526*** (0.0131)
Employment protection legislation	Employment protection legislation (ILO)	-0.0390 (0.0278)
	Procedural requirements for dismissals (ILO)		0.0370*** (0.0056)	0.0449*** (0.0045)	0.0427*** (0.0055)	0.0440*** (0.0042)
	Severance payments (ILO)		0.0101 (0.0063)	0.0112** (0.0053)	0.0090 (0.0057)	0.0105* (0.0054)
	Flexibility of hiring and firing regulations (Fraser)						0.0010 (0.0007)
Collective bargaining system	Centralized collective bargaining (Fraser)	-0.0006 (0.0043)	0.0012 (0.0045)		0.0007 (0.0047)
	Square of Centralized collective bargaining	0.0001 (0.0004)	-0.0001 (0.0004)		-0.0001 (0.0004)
	Flexibility of wage determination (WEF)			-0.0063 (0.0084)		-0.0111 (0.0082)	-0.0162** (0.0080)
	Square of flexibility of wage determination			0.0004 (0.0009)		0.0009 (0.0009)	0.0014* (0.0008)
Business/Product market regulation	Administrative requirements (Fraser)	-0.0023** (0.0011)	-0.0023** (0.0010)	-0.0021* (0.0011)
	Regulatory Burden (Fraser)	-0.0010** (0.0004)	-0.0009** (0.0004)	-0.0009** (0.0004)
	Goods market efficiency (WEF)				-0.0110*** (0.0032)	-0.0107*** (0.0033)	-0.0069*** (0.0025)
	Intensity of local competition (WEF)						0.0008 (0.0021)
Taxation	Tax burden (Heritage)	0.0001 (0.0002)	0.0001 (0.0002)	0.0001 (0.0002)	-0.0000 (0.0002)	-0.0000 (0.0002)	0.0001 (0.0001)
Financial markets	Financial market development (WEF)	-0.0036 (0.0024)				-0.0021 (0.0024)
Financial markets	Credit market regulations (Fraser)		-0.0018** (0.0008)	-0.0016* (0.0009)	-0.0015* (0.0009)		-0.0021*** (0.0007)
	Constant	0.0728** (0.0327)	0.0335 (0.0250)	0.0560* (0.0287)	0.0765*** (0.0277)	0.1054*** (0.0333)	0.1012*** (0.0267)
	Observations	745	769	750	752	754	1,376
	R-squared	0.5908	0.5990	0.6026	0.6016	0.6015	0.6592
	Number of c_id	91	93	91	91	91	138

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

^{Figures 2.1 to 2.5:}

Share of institutional factors in the unemployment increase between 2009 and 2017 Using regressions 4 (LHS figures) and 6 (RHS figures)

Citation: IMF Working Papers 2023, 219; 10.5089/9798400257025.001.A001

^{Figures 2.6 to 2.8:}

Share of institutional factors in the unemployment increase between 2009 and 2017

Citation: IMF Working Papers 2023, 219; 10.5089/9798400257025.001.A001

Source: Authors calculations from regression 4.

Among all institutional factors, the deterioration of product market regulation is likely to be the most relevant factor. During the 2009–2017 period, and despite the adoption of important laws to strengthen the business climate (e.g., the revision of the competition law^¹¹ in 2015 and a new investment law^¹² in 2016), improved transparency and a decrease in trade tariffs, several product market and business regulations’ indicators in Tunisia (based on surveys^¹³) experienced a deterioration between 2009–2017 . The most important deterioration is in competition and administrative requirements, as the market power of monopolies and oligopolies has actually been reinforced during the last decade. Moreover, administrative requirements seem to be an important obstacle for investment and business creation in Tunisia. The new investment and competition laws may not have been effective yet in improving the business climate and competition, perhaps as the publication of those laws may not have been sufficient and should be accompanied by the effective implementation of new procedures (decrees and other explanatory documents) and the elimination of inconsistent laws.

^{Figure 3:}

Change In the Product Market Efficiency and business regulation in Tunisia between 2009–2017 (In percent)

Citation: IMF Working Papers 2023, 219; 10.5089/9798400257025.001.A001

Sources: WEF and authors calculation. Solid Bars are the results of surveys, a decrease implies a detrioration of the indicator.

Part of the increase in unemployment during the last decade is attributable to labor market regulation, in the presence of an intermediate bargaining system and rigid product market regulation. Estimates from specification (4) show that an alleviation of dismissal procedures by one score point^¹⁴ leads to a decrease in the unemployment rate by 4 ½ percent. It should be noted that the score for this index is low in Tunisia (0.25, from a score between 0 and 1), reflecting mainly the low protection of temporary in comparison with permanent workers. Overall, the sub indicators of labor market policies between 2009 and 2017 show a move toward a more rigid labor market and a more centralized bargaining system, while still remaining at an intermediate level of bargaining (i.e., neither fully centralized, nor fully liberalized). The difference between de facto indicators (Fraser Institute’s indicators – based on surveys, in figure 5) and de jure indicators (the ILO labor indicators-based) suggests that, even though Tunisian labor market laws have barely changed during the last decade, employers perceive that a more centralized bargaining system, coupled with a deterioration of the business climate, has made labor market regulations more rigid and as a consequence, assess labor market policies as deteriorating. As product market deregulation also incentivizes labor market deregulation – with an overall positive effect on aggregate employment (Blanchard and Giavazzi, 2003), it is quite likely that the deterioration in product market regulations has not helped to improve labor market conditions, and thus led to an increase in the unemployment level.

^{Figure 4:}

Change in Labor Market regulation and its subcomponents between 2009 and 2017

(In percent)

Citation: IMF Working Papers 2023, 219; 10.5089/9798400257025.001.A001

Source: Fraser Institute and authors calculations

Access to credit matters for the labor market in Tunisia, especially in the presence of a rigid labor market regulation. Estimation in Table 1 shows that credit access explains about 5 percent of the unemployment increase during the period 2009 and 2017. Perceptions of banking sector conditions have deteriorated during the last decade. This has been associated with a few developments: a continuous crowding out from the public sector, shocks affecting durably bank soundness (e.g., through the weakened tourism sector and the increase in NPLs), lack of progress in credit access which impedes job creation through new innovating firms’ (Acemoglu, 2001). In addition to the role played by financial inclusion in job creation, the Tunisian labor market presents some imperfections, coupled with a strong wage bargaining system. Financial development would help firms to better internalize those imperfections and allow for lower labor adjustment costs.

^{Figure 5:}

Change in Credit Market Regulation and its subcomponents in Tunisia between 2009 and 2017

(In percent)

Citation: IMF Working Papers 2023, 219; 10.5089/9798400257025.001.A001

Sources: Fraser Institute. WEF and Authors calculation1/ The extent of government borrowing relative to private-sector borrowing 2/ between 2011 and 2017

IV. Unemployment in Tunisia and its Sensitivity to the Output Gap

The Tunisian economy has been growing at a pace below its potential during the period, and therefore short of creating enough jobs to absorb new entrants. The output gap contributed positively – to the tune of a cumulative 11 percent – to the increase in the unemployment rate in Tunisia between 2009 and 2017 (charts 2.3 and 2.4). The persistence of the output gap is a result of the different domestic and external shocks that Tunisia faced, as well as the falling investment and productivity that in turn led to a decline in growth (OECD, 2018). Another possible explanation for Tunisia’s persistent growth below potential could come from institutional factors that could, not only directly affect the unemployment level, as explained in the previous section, but also affect the speed of adjustment of employment to business cycles in Tunisia^¹⁵. Institutional factors could impede adjustments to shocks by reducing the turn-over in the labor market. They can also slow down the recovery in the labor market during economic upturns, thereby preventing output from reaching its target level (David et al. 2019).

One way to empirically address this question is to estimate the sensitivity of unemployment to economic activity (Okun’s coefficient), and then determine to what extent the latter could be affected by institutional factors. As in David et. al (2019), this section uses a broad panel of countries to explore the cross-country variation in Okun’s coefficients and draws on the results to focus on the Tunisian case. The literature (Ball et al 2017, David et al, 2019, Ahn et al. 2017) points out that the sensitivity of unemployment to business conditions depends on the level of income. In light of this, we consider various estimations using a panel of 151 countries, including Tunisia. More specifically, we use a dynamic common-correlated effects model with heterogeneous coefficients in a panel over 21 time periods (from 1996 to 2017). This methodology allows us to deal with possible cross-sectional dependency through the inclusion of common factors in the estimation. Moreover, it is shown to perform well even in the case of dynamic models with weakly exogenous regressors and with the inclusion of non-stationary variables (Kapetanios et al. 2011, Chudik and Pesaran, 2013). Kapetanios et al. (2011) show that the estimators are also robust to structural breaks¹⁶.

The regression format is as follows:

\begin{matrix} Δ u_{i, t} = λ_{i} Δ u_{i, t - 1} + β_{i} Δ y_{i, t} + ϑ_{i, t} & (2) \\ ϑ_{i, t} = α_{i} + Σ_{l = 1}^{P_{T}} δ_{i, l} f_{l, t} + ε_{i, t} \end{matrix}

Where Δu_i,t = u_i,t – u_i,t-1 is the (absolute) change in unemployment rate, Δy_i,t = y_i,t – y_i,t-1 is the change in log of output (real GDP)^¹⁷, α_i a unit-specific fixed-effect; ε_i,t is a cross section unit specific error term, which is assumed to be independent and identically distributed and f_m,tare unobserved common factors, with (δ_i) is the heterogenous, country-specific loading factor. β_i represents the Okun’s coefficient and is country specific. Following Pesaran (2006) and Chudik and Pesaran (2013), we include two lags of the exogenous variables as well as the cross-sectional averages and estimated the model in two steps. This in turn allows us to account for unobserved dependance between cross sectional units, which leads to unbiased ordinary least squares regression results.

Estimation results of equation (2) are presented in table 2, using annual data for a (unbalanced) panel over the period 1996 to 2017. Specification (1) includes all the 151 countries in our sample and specification (2) includes Tunisia, among middle income countries. This is to test whether unemployment changes exhibit a different reaction to output variations in middle income countries. The results are robust to strong cross sectional dependance (Pesaran, 2015). A potential challenge for the specifications below is that output change could affect unemployment change and thus, lead to an endogeneity problem within the model. To tackle this issue, we also estimated separately an augmented version of the model, by adding instrumental variables (the first four lagged variables of the output change). Results show no significant difference between the two versions of the model^¹⁸.

Unemployment’s responsiveness to variations in output depends on country income levels. The estimation of Okun’s coefficient for the full sample (specification 1) is about -0.09 and statistically significant at the 1 percent level^¹⁹. For the sample of advanced economies, the coefficient seems to be lower than the one estimated by David et al (2019) and Ball et al (2017), suggesting that the level of income explains only in part the level of unemployment responsiveness to output changes. To confirm this hypothesis, we also conduct the estimation for several panels of countries, by income level (specifications 2 and 3). Results from the different specifications (table 2) show substantial differences across groups especially between low – middle income countries and advanced countries. Advanced economies show overall a substantially higher responsiveness of unemployment to business conditions.

^{Table 2:}

Estimation of Okun’s coefficient (Equation 2)

^{Table 2:}

Estimation of Okun’s coefficient (Equation 2)

	(1) All sample	(2) Middle income countries	(3) high income countries
	Δ Unemployment	Δ Unemployment	Δ Unemployment
Δ GDP_t	-0.094*** (0.016)	-0.073*** (0.014)	-0.244*** (0.035)
Δ GDP_t-1	-0.026*** (0.011)	-0.02 (0.015)	-0.057* (0.032)
Δ Unemployment_t-1	0.118*** (0.034)	0.142*** (0.047)	0.175*** (0.071)
Constant	0.001 (0.001)	0.002 (0.002)	-0.001 (0.001)
Number of countries	151	41	48

^{Table 2:}

Estimation of Okun’s coefficient (Equation 2)

	(1) All sample	(2) Middle income countries	(3) high income countries
	Δ Unemployment	Δ Unemployment	Δ Unemployment
Δ GDP_t	-0.094*** (0.016)	-0.073*** (0.014)	-0.244*** (0.035)
Δ GDP_t-1	-0.026*** (0.011)	-0.02 (0.015)	-0.057* (0.032)
Δ Unemployment_t-1	0.118*** (0.034)	0.142*** (0.047)	0.175*** (0.071)
Constant	0.001 (0.001)	0.002 (0.002)	-0.001 (0.001)
Number of countries	151	41	48

For Tunisia, the responsiveness of unemployment to the business cycle is substantially higher than for the average of middle-income countries and the MENA region and is comparable to high income countries (Figure 6). This result is in line with IMF (2022) and Ball et al. (2019), showing that Tunisia has a relatively high Okun’s coefficient, and that unemployment variations in Tunisia appear to be much more driven by changes in the business cycle than in countries with peer countries.

^{Figure 6:}

Okun’s coefficient

(in units)

Citation: IMF Working Papers 2023, 219; 10.5089/9798400257025.001.A001

Authors calculation

To better assess the reasons behind such a high sensitivity to the business cycle in Tunisia, we conduct a cross sectional estimation of the determinants of the Okun’s coefficient. We include in the regression explanatory variables that have been shown in the literature to have an impact on unemployment’s responsiveness to the business cycle (David et al. 2019, Furceri et al. 2020, Ball et al. 2019). We consider the following variables:

The unemployment rate: Furceri et al. (2020) shows that the response is more pronounced for countries with a high unemployment level. Institutional factors affecting unemployment could also explain this positive relationship: a high degree of product and/or labor market flexibility allows firms to easily adjust employment when output changes, raising the Okun’s coefficient (Ball et al, 2017).
Labor market and business regulations: employment and product protection legislation are important institutional factors affecting the labor market as well as its responsiveness to business conditions (Jaumotte, 2011, David et al. 2019). In previous sections we show that flexibility in both areas can play a significant role in employment conditions, while the relationship tends to be neither positive, nor linear (see detailed discussion in section 2). Data on labor and business regulation are taken from the Fraser Institute and World Economic Forum (WEF). The advantage of this data is that (i) index’s components are standardized from different sources, and this allows to calculate subcomponents by taking two or three ones, in order to capture the interaction between them, and (ii) subcomponents are available and calculated which allows for a more granular and precise estimation.
The level of informality: the informal sector can have a stabilizing effect during the business cycle, especially at the time of recessions and downturns, as it may absorb workers who lose their formal jobs (IMF, 2022). Informality could then be at play as a buffer and reduce the sensitivity of employment to business cycles (Ball et al. 2019). Consequently, it is expected that the higher the level of informality, the lower is the response of unemployment to cyclical variations. The Medina-Schneider index of informality is used as a proxy of the informal sector (more details of the index in Medina and Schneider, 2018).
Global economic integration (degree of trade openness): the response of the unemployment rate to the business cycle can be a function of the countries’ dependance on global economic conditions. The share of exports and imports to GDP from the World Bank is used for that purpose.

The size of the informal sector is negatively correlated with trade openness and flexibility of business as well as labor market regulations (table 5). This is in line with the literature which documents that informality tends to decrease as product markets are less rigid and the business climate is improved. As this significant correlation could affect the responsiveness of unemployment to business conditions, we include, in addition to the specification with all the variables described above, other separate estimations of individual variables, as well as correlated ones. We estimate the following specification:

\begin{matrix} {\hat{β}}_{i} = α_{i} + \emptyset^{'} X_{i} + ξ_{i} & (3) \end{matrix}

Where ${\hat{β}}_{i}$ is the Okun’s coefficient estimated by equation (2)^²⁰, X_i is the vector of control variables explained above. Results of the OLS estimation with robust standard deviations and using the crosscountry data are presented in table 3^²¹.

Unemployment is less responsive to cyclical conditions in countries with higher levels of informality^²² and/or lower levels of unemployment. This confirms that informality acts as a “buffer” to labor market disruptions in the formal economy (David et al. 2019; IMF, 2022). As the informal sector may absorb workers who lose their formal jobs during economic downturns, the adjustment to business cycles in economies with high informality is likely to occur more through wages and/or working hours in the informal sector, rather than through a reduction in the number of employed. Moreover, we observe that the higher the level of unemployment, the stronger (in absolute terms) the Okun’s coefficient is.

Unemployment is more responsive to cyclical conditions in countries with more flexibility in their business regulations. While the individual specifications show no response of the Okun’s coefficient to labor market regulation nor to trade openness, specification (7) from Table (3) that includes all explanatory variables show a significant effect of both. One plausible explanation is that the combination of the informal sector, business and labor market regulations have simultaneously an impact on the Okun’s coefficient, while labor market regulation taken individually would not significantly affect the response of unemployment to business activity. This is in line with Fiori et al. (2007) and Cournede et al. (2016) who argue that labor and business regulations are mutually reinforcing.

To assess the impact of policy factors and test the hypothesis of their complementarity, we regress the Okun’s coefficient on several subcomponents of business and labor market regulations. Results reported in Table 4 confirm the interactions between the informal sector, the business and labor market regulations, and their combined effect on the Okun’s coefficient:

With respect to product market regulation, the regulatory burden and business barriers are the most significant drivers of the responsiveness of unemployment. The lower the regulatory burden and entry barriers, the higher (in negative terms) the Okun’s coefficient.
When it is significant, the coefficient associated with labor market regulation or its subcomponents is positive, implying that unemployment in countries with a higher share of informal sector, higher labor market flexibility and weaker business flexibility respond less to changes in the business cycle and vice versa.
This confirms the adjustment role of the informal sector as well as the complementarity between business and labor regulations. This result is also relevant in Tunisia’s context, as it highlights that a high level of product market flexibility, if not matched with labor flexibility could reinforce the bargaining power and worsening employment conditions (Jaumotte, 2011). As Tunisia is facing a high and persistent unemployment rate, further flexibilization of the product market without taking into consideration labor market regulations, would lead to a higher sensitivity of unemployment to business cycles, already high in the Tunisian context. This in turn would increase vulnerability of the Tunisian labor market during downturns.
Estimations suggest also that the collective bargaining system does not have a significant impact on the responsiveness of unemployment to business activity, but its interaction with other labor regulations -including the hiring and firing rules- affects the Okun’s coefficient.

Tunisia’s Okun’s coefficient is 69 percent higher than the average of middle-income countries. Figure 7 compares the unemployment level as well as policies described above (informality, labor market and business regulations) with middle income countries. Regression results of equation (3) show that this significant difference between Tunisia and the average of the middle-income countries is mainly attributable to the unemployment level, which is 69 percent higher in Tunisia than the average of middle-income countries and explains about 48 percent of the difference in the Okun’s coefficient between Tunisia and middle-income countries. 22 percent of the difference between Tunisia and middle-income countries is attributable to the higher flexibility in business regulation in Tunisia and 3 percent to the lower informality in Tunisia. Conversely, the slightly higher flexibility in the Tunisian labor market regulation and the higher share of trade openness in Tunisia have contributed to lower the difference between Tunisia and middle-income countries.

^{Figure 7:}

Factors’ contributions to the difference in Okun’s coefficient between Tunisia and middleincome countries

(In percent)

Citation: IMF Working Papers 2023, 219; 10.5089/9798400257025.001.A001

Sources: Authors calculation using regression (7)

The results of this section highlight the complementarity between institutional factors and their impact of the labor dynamics. As Tunisia’s Okun’s coefficient is relatively high, it is worth noting that fixing structural impediments should go in tandem. This, in order not to increase further the sensitivity of the labor market to economic cycles. An improvement in the flexibility of business regulation alone, without further flexibility in the labor market and/or reduction in informality, would lead to a greater sensitivity of unemployment, which would weigh on the labor market, especially during economic shocks, which tend to be frequent in Tunisia.

^{Table 3:}