Sustainable Employment in Developing and Emerging Countries: Testing Augmented Okun’s Law in Light of Institutional Quality

Abstract

This article is motivated by the recent debate raised on Okun’s Law regarding the divergence of the magnitude of Okun’s coefficient across countries and time horizons, its asymmetry and non-linearity, and the methodological issues associated with the estimate of this law. We tested the proposition that institutional quality—as measured by governance indicators—can explain this divergence and non-linearity. We estimated an augmented version of Okun’s Law showing the non-linear responsiveness of unemployment to fluctuations in GDP in the presence of institutional quality in a sample of 88 developing and emerging countries over the 1985–2019 period. Estimates are run using the 3sls regressor. We found evidence confirming that the responsiveness of unemployment to changes in output is all the more remarkable in countries with stronger institutions. We show that improving the institutional environment—as proxied by Government Effectiveness (GE), Control of Corruption (CC), Regulatory Quality (RQ), the Rule of Law (RL), and Voice and Accountability (VA)—is as important as increased economic growth in the strategy of reducing unemployment and reaching full employment in these countries, which is one of the 17 Sustainable Development Goals (SDGs) issued by the United Nations.

1. Introduction

Reaching full employment is one of the 17 sustainable development goals (SDGs) issued by the United Nations to face the complex global economic situation. It remains at the center of any economic policy. In ref. [1], Okun reported an empirical regularity based on quarterly data for the U.S. economy from 1948-q2 to 1960-q4: a short-run negative relationship between unemployment fluctuations and real output changes and a value absolute of 0.3 for the responsiveness of unemployment to fluctuations in output around the trend. Since Okun’s pivotal founding, a growing empirical literature has emerged intending to test the validity of Okun’s relationship and analyze this relationship for other countries and over various time horizons.

Overall, while most of these studies tend to confirm a negative relationship between the unemployment rate and the change in real GDP, they nevertheless underline that the magnitude of Okun’s coefficient remains controversial. Moreover, some studies have shown that Okun’s Law is neither constant nor linear or symmetric over long time horizons. Finally, some other studies raised methodological issues in estimating Okun’s relationship, particularly the endogeneity of the GDP growth variable and the problems of missing variables, and suggested controlling them to obtain better estimates of the Okun coefficient. Therefore, recent studies in this stream of literature have raised a new debate about the validity of Okun’s Law in developed and developing countries, the instability and non-linearity of Okun’s coefficient, and the factors that may shape such a relationship.

In this context, the present article contributes to this recent debate by testing the hypothesis that the institutional dimension is a possible factor likely to have importance in determining the Okun’s relationship. Indeed, since the founding works of the New Institutional and Development Economics in refs. [2,3,4] and [5], remarkable efforts have been made in this literature to demonstrate that institutional quality matters for economic growth, social outcomes, and development. Therefore, in this paper, we exploit the crucial findings from this branch of literature to test our conjecture that the value of Okun’s coefficient, i.e., the responsiveness of unemployment to output changes depends on the performance of countries in terms of institutional quality, as can be proxied by the worldwide governance indicators. Although this conjecture seems straightforward, no empirical work has been completed to support it. Therefore, this article aims to fill this gap. To the author’s knowledge, this article is the first to consider an ‘augmented’ version of Okun’s Law that considers the role of institutional quality.

Since the finding in ref. [1] regarding the link between unemployment and the growth of the real gross national product in the U.S., new studies have emerged over the next sixty years, testing the validity of Okun’s Law in other countries and regions and over different time horizons. By employing various estimation techniques, this body of research then offered new perspectives to the existing theory by bringing to light the following aspects: (i) the evolution or instability of the Okun coefficient value over the time horizon considered and across the sample of countries; and (ii) the asymmetry of Okun’s relationship over the past sixty years of research.

Most of the literature on the validity of Okun’s Law concerns advanced economies, some of which have focused mainly on the U.S. case [6,7,8]. The studies concluded that (i) the fixed-coefficient specification commonly used to characterize Okun’s Law could lead to inappropriate or incorrect results; (ii) there exists a threshold in the effect of fluctuations in output on fluctuations in unemployment (non-linearity); and (iii) the working hours and utilization affect the short- and long-term fluctuations of the time-varying Okun coefficients.

Alongside these studies on the American case, others have been interested in analyzing the validity of Okun’s relationship in groups of advanced countries, mainly European, and have provided a comparison of Okun’s coefficients among these countries. They came up to the conclusion of the divergence in the value of the Okun’s coefficient between the countries considered. Such divergence is explained partly by the idiosyncratic characteristics of national labor markets [9,10,11,12]. Some other researchers have enlarged the sample of advanced countries studied and found similar results, particularly the divergence of Okun’s coefficients, their asymmetry, and their non-linearity. For example, Okun’s Law is tested for the G7 countries in refs. [13,14], for the OECD countries in [15,16,17], and for selected developed countries in [18,19,20].

Alongside the part of the research focusing on advanced economies, many studies have recently focused on analyzing Okun’s Law for developing economies, either as regions or as individual countries. For instance, refs. [21,22,23,24] tested Okun’s Law for different samples, regrouping developing and developed countries. The main findings of these studies show (i) differences in Okun’s coefficient between high- and low-income countries, (ii) the additional impact of the financial crisis on employment in developed economies, (iii) the Okun coefficient is about half as large in developing countries as it is in advanced countries, (iv) the Okun coefficient is negatively associated with the poverty rate and skill mismatch, and (v) the faster real GDP growth is associated with lower overall male and female unemployment rates in developing and developed countries. Some studies have examined the validity of Okun’s Law in particular developing countries and have used similar econometric approaches based on cointegration techniques or Granger causality in the context of Vector Error Correction (VEC) models or Autoregressive Distributed Lags (ARDL). Most of these studies have found evidence in favor of the validity of Okun’s Law. However, some studies have found this coefficient low, reflecting rigidities associated with labor market protections. One can refer to ref. [25] for the Saudi case, ref. [26] for India, ref. [27] for the Tunisian case, ref. [28] for South Africa, ref. [29] for Jordan, ref. [30] for Algeria, ref. [31] for Nigeria, and ref. [32] for Mauritius, among other studies.

Bearing in mind the above, this article aims to contribute to the recently raised debate concerning the validity of Okun’s Law, its stability, and the non-linearity of its coefficient. Our contribution to this debate is multidimensional. First, we test the hypothesis that the responsiveness of employment to fluctuations in output—i.e., Okun’s coefficient—depends on the institutional quality of countries that we measure using worldwide governance indicators. Second, we study this responsiveness for a large sample of developing and emerging (from Africa, Latin America, and Asia) and non-advanced economies as observed in most previous work in this field, which would help reduce the relative lack of work completed on these countries. We consider a relatively long period, from 1985 to 2019, long enough to lead to solid and conclusive results on the stability of employment. Third, we address the problem of potential endogeneity in Okun’s relationship due to the non-exogeneity of real GDP fluctuations by specifying an augmented Okun model as a system of two equations that we estimate using the 3sls regressor. Finally, we also control for the idiosyncratic country and time-specific effects.

2. Materials and Methods

2.1. Augmented Okun’s Law

According to Okun’s Law, shifts in aggregate demand cause output to fluctuate around potential. These output fluctuations cause firms to hire and lay off, thus modifying employment. In turn, changes in employment move the unemployment rate in the opposite direction. Therefore, there would be a negative short-run relationship between fluctuations in output and unemployment around their long-term levels. As explained above, we seek to test this law in its original version, which we enrich with the institutional quality effect. More specifically, our main objective is to show how fluctuations in unemployment are affected by changes in production and how the latter interact with the quality of the institutional environment. We test the hypothesis that the responsiveness of unemployment to the output gap should be all the more remarkable in countries with stronger institutions. We perform this while accounting for the endogeneity aspect of the output gap, a feature that has been neglected in most previous studies when estimating the original Okun’s Law. Based on these considerations, we simultaneously estimate the following system of two equations:

$$\{\begin{array}{c}{U}_{it}^c={\alpha }_0+{\alpha }_1{y}_{it}^c+{\alpha }_{2}Ins{t}_{it}+{\alpha }_3\left(Ins{t}_{it}\ast y_{it}^c\right)+{\alpha }_4{n}_{it}+{\mu }_i+{\vartheta }_t+{\epsilon }_{it} \left(1\right)\\ \\ y_{it}^c={\beta }_0+{\beta }_{1}Ins{t}_{it}+{\displaystyle \sum }{\beta }_s{X}_{sit}+{\theta }_i+{\phi }_t+e_{it} \left(2\right)\end{array}$$

where the subscripts i and t refer to country and time, respectively.

The first equation reflects the augmented Okun’s relationship, where—in addition to the output gap—we use institutional quality and the population growth rate as explanatory factors for the unemployment gap. The second equation expresses the production fluctuations around the trend, depending on certain explanatory factors. In this system, the variables:

-

${\mathit{U}}^{\mathit{c}}$ refers to the cyclical component of the unemployment rate, i.e., the unemployment gap around the trend obtained using the Hodrick–Prescott high-pass filter, such that: $U^c=U-U^{*}$, where $U^{*}$ refers to the unemployment trend component. Data on the labor force unemployment rate is from the ILO’s estimates.

-

${\mathit{y}}^{\mathit{c}}$ is the cyclical component of the output, i.e., the deviation from the trend ($y^c=y-y^{*}$, where $y^{*}$ refers to the output trend component). $y$ is the log of the output expressed in the constant 2015 $U.S.

${\alpha }_1$ captures Okun’s coefficient, which should be negative.

-

$\mathit{I}\mathit{n}\mathit{s}\mathit{t}$ is the institutional quality indicator. We use the six sub-indicators given in the Worldwide Governance Indicators (WGI) database, including Governance effectiveness (GE), Control of Corruption (CC), Political Stability and Absence of Violence (PS), Regulatory Quality (RQ), Rule of Law (RL), and Voice and Accountability (VA). The indicators are measured on a scale from −2.5 to 2.5, where high values indicate effective institutions. We also use the index based on the mean of these six sub-indicators (Gov) as a proxy for the overall level of institutional quality. More details about what these indicators capture is provided in the Appendix A.

Notice that $Inst$ enters in both system equations. It enters the $U^c$ equation as an independent variable and as a variable interacting with the output gap, $y^c$, to capture the direct and indirect effects of institutional quality on unemployment shifts. It also enters the $y^c$ equation as a possible explanatory variable, similar to the other variables in ${\displaystyle \sum }X$.

-

$\mathit{n}$ is the log of the population growth rate used as a proxy of the labor-force growth rate and is introduced as an additional explanatory variable in the $U^c$equation.

-

${\displaystyle \sum }\mathit{X}$ denotes a set of variables explaining output fluctuations, some of which have been identified in the founding literature on economic growth [33,34]. We consider the following variables (in logarithm):

-

Inv/GDP: physical investment as a percentage of GDP,

-

Open/GDP: openness ratio calculated as the sum of exports and imports as a percentage of GDP,

-

FDI/GDP: the inward flow of Foreign Direct Investment as percentage of GDP,

-

DebtServ/Exp: the ratio of debt service to exports,

-

CPI: the Consumer Price Index (2010 = 100),

-

(${\mathit{u}}_{\mathit{i}}$ and ${\mathit{\theta }}_{\mathit{i}}$), (${\mathit{\vartheta }}_{\mathit{t}}$ and ${\mathit{\phi }}_{\mathit{t}}$), and (${\mathit{\epsilon }}_{\mathit{i}\mathit{t}}$ and ${\mathit{e}}_{\mathit{i}\mathit{t}}$) refer to the unobserved country-specific effects, time-specific effects, and residual terms, respectively.

In Equation (1) of this system, we expect the coefficient ${\alpha }_1$ to have a negative sign as in the original Okun’s Law. The direct impact of the institutional dimension, ${\alpha }_2$, on the unemployment gap should depend on whether the sample of countries studied has, on average, a high or low level of institutional quality. At the same time, the coefficient ${\alpha }_3$on the interaction variable should have a negative sign, which would be evidence that the responsiveness of unemployment to output shifts is higher in countries with stronger institutions. Regarding the impact of population growth, ${\alpha }_4$ we expect it to be positive because a higher population growth rate would translate to a higher labor-supply growth rate, which in turn would raise the unemployment rate.

In Equation (2), we expect the output gap to increase with more effective institutions (${\beta }_1>0$) following the results of the previous empirical literature belonging to the New Institutional and Development Economics. Moreover, the output gap is expected to increase with the Inv/GDP, Open/GDP, and FDI/GDP ratios. On the contrary, it should decrease with the ratio of debt service to exports (DebtServ/Exp) and the price index (CPI). Indeed, as the debt service burden gets higher relative to the export revenues, the country is constrained to reallocate more financial resources from productive activities to the payment of the debt service burden. Such a negative crowding-out effect is likely to occur in highly indebted countries. Furthermore, inflationist pressures raise investment costs, reduce real investment returns, reduce consumer demand, and hamper the country’s international competitiveness and exports. These are possible channels for explaining how inflation may inhibit output growth.

2.2. Data Description

Our econometric analysis relies on an unbalanced annual dataset for a sample of 88 emerging and developing countries belonging to Africa, Latin America, and Asia, covering the period from 1985 to 2019.

Table A1. A: Countries used in the sample. B: Description and sources of the used variables.

Region	Countries	Number
Africa	Algeria, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Central African Republic, Chad, Congo Dem. Rep., Congo Rep., Cote d’Ivoire, Djibouti, Egypt, Equatorial Guinea, Ethiopia, Gabon, Gambia, Ghana, Guinea, Guinea-Bissau, Kenya, Lesotho, Liberia, Libya, Madagascar, Malawi, Mali, Mauritania, Mauritius, Morocco, Mozambique, Namibia, Niger, Nigeria, Rwanda, Senegal, Sierra Leone, South Africa, Sudan, Tanzania, Togo, Tunisia, Uganda, Zambia, Zimbabwe.	44
Latin America	Bolivia, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, Venezuela.	18
Asia	Bangladesh, Bahrain, China, Hong Kong, India, Iran Islamic Rep., Indonesia, Jordan, Korea Rep., Kuwait, Lebanon, Malaysia, Myanmar, Nepal, Oman, Pakistan, Qatar, Saudi Arabia, Singapore, Sri Lanka, Turkey, Yemen, Philippines, Thailand, United Arab Emirates, Vietnam.	26
Variable	Variable description	Source
U	Total unemployment rate as % of the total labor force	International Labor Organization
y	Log of GDP measured in constant 2015 $US.	World Bank Development Indicators Database
n	Log of Population growth (annual %)
GE, CC, PS, RQ, RL, VA, Gov	We use the indicators provided in the Worldwide Governance Indicators (WGI) database as proxies for institutional quality. These are: - Government Effectiveness (GE): captures perceptions of the quality of public services, the quality of the civil service and the degree of its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government’s commitment to such policies. - Control of Corruption (CC): captures perceptions of the extent to which public power is exercised for private gain, including petty and grand forms of corruption and “capture” of the state by elites and private interests. - Political Stability and Absence of Violence/Terrorism (PS): measures perceptions of the likelihood of political instability and/or politically-motivated violence, including terrorism. - Regulatory Quality (RQ): captures perceptions of the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development. - Rule of Law (RL): captures perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular, the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence. - Voice and Accountability (VA): captures perceptions of the extent to which a country’s citizens can participate in selecting their government, as well as freedom of expression, freedom of association, and free media. - The Governance Index (Gov): is a proxy for the global institutional quality, and we calculated it as the arithmetic mean of the above six indicators. These indicators are available for over 200 countries from 1996 to 2019. They are measured on a scale from −2.5 to +2.5, where higher values indicate more effective institutions.	World Bank Database, the WGI by Kaufmann and Kraay (2020).
Ln(Inv/GDP)	Log of: Ratio of Gross fixed capital formation to GDP (%)	World Bank Development Indicators Database
Ln(FDI/GDP)	Log of: Ratio of Foreign Direct Investment inflows to GDP (%)
Ln(Open/GDP)	Log of: Openness ratio: Ratio of “Sum of imports and exports of goods and services to GDP (%)
Ln(DEBT_SERV/EXP)	Log of: Ratio of total debt service (% of exports of goods, services and primary income) (%)
Ln(CPI)	Log of: Consumer Prices Index (base 100 = 2010)

Source: The author.

and

Table A2. Panel Unit-Root Tests.

		Level
	P.U.R.T	Statistic	p-Value	Trend	I(.)
Uc	I-P-S	−8.749	0.008	--	I(0)
	Fisher-type	20.63	0.000	--	I(0)
yc	I-P-S	−7.879	0.000	--	I(0)
	Fisher-type	15.00	0.000	--	I(0)
GE = (Government_Effectiv)	I-P-S	−1.774	0.038	--	I(0)
	Fisher-type	5.469	0.000	--	I(0)
CC = (Control_Corruption)	I-P-S	−3.455	0.000	--	I(0)
	Fisher-type	7.280	0.000	--	I(0)
PS = (Political_Stability)	I-P-S	−3.073	0.001	--	I(0)
	Fisher-type	8.201	0.000	--	I(0)
RQ = (Regulatory_Quality)	I-P-S	−7.049	0.000	Yes	I(0)
	Fisher-type	3.695	0.000	--	I(0)
RL = (Rule_Law)	I-P-S	−7.202	0.000	Yes	I(0)
	Fisher-type	2.315	0.010	--	I(0)
VA = (Voice_Accountability)	I-P-S	−6.066	0.000	Yes	I(0)
	Fisher-type	6.467	0.000	--	I(0)
Gov = (Governance)	I-P-S	−7.775	0.000	Yes	I(0)
	Fisher-type	2.846	0.002	--	I(0)
N	I-P-S	−8.711	0.000	--	I(0)
	Fisher-type	4.153	0.000	--	I(0)
Ln(Inv/GDP)	I-P-S	−8.146	0.000	Yes	I(0)
	Fisher-type	4.926	0.000	--	I(0)
Ln(Open/GDP)	I-P-S	−6.731	0.000	Yes	I(0)
	Fisher-type	6.712	0.000	--	I(0)
Ln(FDI/GDP)	I-P-S	−12.40	0.000	--	I(0)
	Fisher-type	29.45	0.000	--	I(0)
Ln(DebtServ/EXP)	I-P-S	−4.143	0.000	--	I(0)
	Fisher-type	5.596	0.000	--	I(0)
Ln(CPI)	I-P-S	−9.466	0.000	yes	I(0)
	Fisher-type	38.04	0.000	--	I(0)

Source: Author’s compilations. Notes: I-P-S indicates the Im, Pesaran, and Shin (2003) test. The null hypothesis assumes all panels have a root process, while the alternative is that the fraction of stationary panels is nonzero. The reported statistic is the “Z-t-tilde-bar.” Fisher-type stands for the Fisher-type test with the Augmented Dickey-Fuller approach. The null hypothesis assumes all panels contain unit roots based on unit root tests conducted individually. The reported statistic is the “Modified inv. chi-squared”. The Phillips-Perron unit-root test with lag(1) is used in the Fisher-type test.

in the Appendix A provide the list of countries used by region and the data sources for the variables employed in the study, respectively.

Table 1. Summary statistics.

Variable	Obs	Mean	Std. Dev.	Min	Max
U	2792	7.128004	6.133986	0.11	37.97
y	3037	24.02164	1.791336	18.69897	30.31422
n	3163	2.2532311	1.437547	−6.766132	17.51221
GE	1825	−0.3481789	0.7672633	−2.078399	2.436975
CC	1825	−0.4076674	0.7430188	−1.815811	2.32558
PS	1825	−0.4848393	0.880126	−2.844653	1.61567
RQ	1825	−0.2994436	0.7659269	−2.363184	2.260543
RL	1825	−0.4170487	0.7412437	−2.346105	1.880382
VA	1825	−0.4713938	0.718458	−2.000246	1.314993
Gov	1825	−0.404762	0.6689733	−2.100317	1.639372
Inv/GDP	2716	22.20876	8.49991	0.1	93.54746
FDI/GDP	3030	3.233422	6.61351	0	161.8237
Open/GDP	2866	74.43833	54.47162	0.1746816	442.62
Debtserv/Exp	2381	17.8012	14.42645	0.1834247	156.8582
CPI	2889	81.8842	99.23047	0.1	3364.82

Source: data from 88 DCs over the 1985-2019 period.

summarizes the descriptive statistics of these variables.

The standard deviations of U in the sample reflect differences in countries’ unemployment situations and their strategies to combat unemployment. The average values of the indicators of institutional quality are negative, implying low institutional quality on average for the sample of countries and the period considered. The standard deviations of these indicators are high, so countries split between those with negative institutional quality values and those with positive values. Some countries have even reached results close to the maximum values (Singapore, Hong Kong, and South Korea in recent years). Figure 1, below, presents the simple plot of the original version of Okun’s Law, showing how cyclical unemployment fluctuations, U^c, are associated with the cyclical output fluctuations, y^c.

The scatter chart shows how the unemployment fluctuations around the trend are associated with output fluctuations in the studied countries, as suggested by Okun’s Law. The plot does not depict a clear association, which appeals to a deeper exploration of this law by assessing it econometrically. We perform such an investigation in the next section. More specifically, we will test to what extent the institutional dimension matters in affecting Okun’s coefficient. We will, therefore, test for the Augmented Okun’s relationship by the institutional dimension. We will carry out this exercise while controlling for the endogenous aspect of the output fluctuations by regressing the system above using simultaneous equations and not a unique equation, as was performed in previous empirical studies based on the original version of Okun’s Law.

We estimate our system of structural equations via three-stage least squares (3sls) with fixed effects. This regressor is appropriate for systems where the endogenous explanatory variables are dependent variables from other equations in the system, as in our model. The regressions are performed while we control for both country and time-specific effects. Nonetheless, we must first ensure the stationarity of the variables used in the model. Given the panel nature of the data, we use panel unit-root tests to investigate the order of integration of the series in the model. Two well-known panel unit-root tests are employed: the Im-Pesaran-Shin (IPS) [35] and the Fisher-type test. Their results are shown in Table A2 in the Appendix A. Both tests indicate that all the series are stationary in levels, i.e., I(0), allowing us to use the 3sls regressor.

3. Results and Discussions

3.1. Estimation Results

We present the estimation results in

Table A3. Estimation of the Augmented Okun’s Law (3sls regressions with country and time fixed-effects): Equation (1).

Dependent Variable: Uc
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
n	---	0.2962 ***	0.354 ***	0.378 ***	0.344 ***	0.353 ***	0.382 ***	0.313 ***	0.350 ***
	---	(0.0921)	(0.1114)	(0.1128)	(0.1118)	(0.1127)	(0.1116)	(0.1089)	(0.110)
yc	−2.753	−3.464 *	−13.03 ***	−15.72 ***	−8.091 *	−12.93 ***	−13.29 ***	−10.60 ***	−12.03 **
	(1.8854)	(1.9310)	(4.4338)	(5.1935)	(4.5419)	(3.7050)	(4.7967)	(3.9743)	(4.802)
GE			0.101
			(0.2118)
yc × GE			−7.785 **
			(3.6240)
CC				0.141
				(0.1993)
yc × CC				−8.834 **
				(4.5579)
PS					−0.057
					(0.0993)
yc × PS					−1.695
					(2.6801)
RQ						0.027
						(0.2397)
yc × RQ						−6.630 ***
						(2.5475)
RL							−0.208
							(0.1836)
yc × RL							−6.495 *
							(3.6887)
VA								0.176
								(0.1393)
yc × VA								−5.544 *
								(2.9486)
Gov									−0.125
									(0.2498)
yc × Gov									−6.906 *
									(3.8686)
Constant	−2.221 *** (0.3177)	−2.524 *** (0.3307)	−1.938 *** (0.3276)	−1.942 *** (0.3316)	−2.095 *** (0.3252)	−1.967 *** (0.3951)	−2.153 *** (0.3469)	−1.813 *** (0.3410)	−2.104 *** (0.3717)
Count eff	Yes	Yes	Yes	Yes	Yes	yes	yes	Yes	Yes
Time eff	Yes	Yes	Yes	Yes	Yes	yes	yes	Yes	Yes
Obs.	1597	1587	1143	1143	1143	1143	1143	1143	1143
Countries	88	88	88	88	88	88	88	88	88

Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Fixed country and time-specific effects are not reported here to save space.

and

Table A4. Estimation of the Augmented Okun’s Law (3sls regressions with country and time fixed-effects): Equation (2).

	Dependent Variable: yc
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
GE			0.0688 ***
			(0.0066)
CC				0.0448 ***
				(0.0069)
PS					0.0267 ***
					(0.0031)
RQ						0.0735 ***
						(0.0066)
RL							0.0576 ***
							(0.0067)
VA								0.0200 ***
								(0.0056)
Gov									0.0934 ***
									(0.0078)
Ln(Inv/GDP)	0.0394 *** (0.0050)	0.0387 *** (0.0049)	0.0292 *** (0.0064)	0.0310 *** (0.0065)	0.0219 *** (0.0066)	0.0341 *** (0.0063)	0.0270 *** (0.0065)	0.0308 *** (0.0066)	0.0210 *** (0.0063)
Ln(Open/GDP)	0.0012 (0.0042)	0.0017 (0.0041)	0.0100 ** (0.0051)	0.0101 ** (0.0052)	0.0141 *** (0.0051)	0.0083 * (0.0050)	0.0131 ** (0.0051)	0.0101 * (0.0053)	0.0119 ** (0.0050)
Ln(FDI/GDP)	0.0051 *** (0.0012)	0.0049 *** (0.0011)	0.0077 *** (0.0014)	0.0071 *** (0.0014)	0.0080 *** (0.0014)	0.0047 *** (0.0014)	0.0065 *** (0.0014)	0.0081 *** (0.0014)	0.0069 *** (0.0014)
Ln(DebtServ/EXP)	−0.0070 *** (0.0020)	−0.0067 *** (0.0020)	−0.0084 *** (0.0023)	−0.0069 *** (0.0023)	−0.0066 *** (0.0023)	−0.0080 *** (0.0022)	−0.0061 *** (0.0023)	−0.0060 ** (0.0024)	−0.0085 *** (0.0022)
Ln(CPI)	−0.0094 *** (0.0026)	−0.0090 *** (0.0026)	−0.0198 *** (0.0049)	−0.0198 *** (0.0050)	−0.0241 *** (0.0049)	−0.0205 *** (0.0048)	−0.0236 *** (0.0025)	−0.0211 *** (0.0051)	−0.0226 *** (0.0048)
Constant	−0.0838 **	−0.0849 ***	−0.0074	−0.0278	−0.0015	0.0225	0.0027	−0.0296	0.0580
	(0.0295)	(0.0288)	(0.0407)	(0.0417)	(0.0415)	(0.0408)	(0.0416)	(0.0427)	(0.0409)
Country effects	Yes	Yes	Yes	yes	Yes	yes	yes	yes	Yes
Time effects	Yes	Yes	Yes	yes	Yes	yes	yes	yes	Yes
Obs.	1597	1587	1143	1143	1143	1143	1143	1406	1143
Countries	88	88	88	88	88	88	88	78	88

Source: Autho’s compilations. Notes: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Country and time fixed specific effects are not reported here to save space.

in the Appendix A. Table A3 shows the regression results of Equation (1), while those of Equation (2) are presented in Table A4. In both tables, we consider several specifications of Okun’s Law. Note that country and time effects are estimated but not reported in the regressions for clarity of presentation. Let us start with the estimation results of the unemployment gap equation presented in Table A3. These results can be organized into the following points:

First, Specification (1) reflects the basic version of this law, where unemployment responsiveness is only a function of real output fluctuations. Okun’s coefficient has the correct sign but is non-significant. In specification (2), we add the population growth rate, n, as an additional regressor to control for the labor supply’s growth rate. The estimates show that the reactivity of the unemployment rate to output fluctuations is reinforced. Moreover, the population growth rate coefficient turns out to be highly significant (at 1%) and positive, as expected. Therefore, the unemployment gap tends to be higher in countries with the fastest population growth rates. Indeed, higher demographic rates must be seen as the corollary of more substantial labor supply pressures on the labor market and, therefore, on the unemployment rate.

Second, in the Specifications from (3) to (9) of Table A3, we consecutively supplement the original version of Okun’s Law with the institutional quality indicators (GE, CC, PS, RQ, RL, VA, and Gov). These indicators enter as such to capture their direct impacts on unemployment gaps and indirectly through their interactions with the output fluctuations variable, y^c. The following crucial results arise. The first one concerns the direct impact of institutional factors on unemployment gaps. In all the specifications, this impact, although with the expected negative sign, is statistically insignificant. This result leads us to conclude that the average institutional quality observed in the sample of countries considered was not good enough to affect unemployment directly. As indicated above, the average levels of the indicators of institutional quality in our countries are below the threshold of moderate institutional quality, i.e., zero.

The second crucial result concerns the interaction term, which captures the indirect impact of institutional factors. The output fluctuations in interaction with the institutional indicators have a negative and statistically significant impact on unemployment gaps. The significance level is between 1 and 10%. All institutional indicators except PS are significant when interacting with the output gap. When the Gov indicator is used as the average index of the six indicators, the results show that the interaction variable is significant at 10%. Therefore, a better governance index increases the effectiveness of output fluctuations in reducing unemployment gaps. This result is likely to confirm our hypothesis that the responsiveness of unemployment to variations in production depends on the quality of the institutions in place, in the sense that this responsiveness is all the greater when the institutions are sound. Government Effectiveness, Control of Corruption, Regulatory Quality, Rule of Law, Voice and Accountability are all institutional factors that significantly determine the extent of this responsiveness. And since these factors differ from country to country and change over time, their magnitude will vary within the countries considered in the sample.

Moving on to the estimation results of Equation (2), Specifications (1) and (2), show that y^c is significantly affected by the explanatory factors considered. Indeed, as expected, output growth is significantly higher in countries with higher Inv/GDP and FDI/GDP ratios. On the other hand, it is negatively and significantly associated with the Debt-service/exports ratio and the level of the CPI. These results are in line with what we had expected.

The impacts of these factors are highly significant (at 1%) in Specifications (1) and (2).

The next specifications, from (3) to (9), where we introduced the institutional quality indicators, show two essential observations. First, all these indicators are statistically significant (at 1%) with a positive sign. That is, output growth is higher in countries with better institutional quality, which is expected in light of the findings of the theoretical and empirical studies of the New Institutional and Development Economics. Second, the economic determinants of output fluctuations are robust to specification changes. Inv/GDP, FDI/GDP, Open/GDP, Debt-service/exports ratio, and the CPI are all statistically significant (at 1%) with the correct sign. The Open/GDP ratio has become, in these specifications, significantly positive after being insignificant in the first two specifications where we omitted the institutional dimension.

3.2. Discussion

As the last step of our test for the Augmented Okun’s Law, we provide in

Table 2. Okun’s coefficients calculations.

Impact on Uc
A: Impact without the Institutions interaction term (i.e., ${\mathsf{\alpha }}_1$): Column 1 of Table A3	B: Impact with the Institutions interaction term (i.e., ${\mathsf{\alpha }}_1+{\mathsf{\alpha }}_3\ast \mathrm{Inst}$): Column 9 of Table A3, evaluated at Gov = (.)
−2.753	−1.0737 B	−0.4047 M	0.2642 A
	−4.614 *	−9.2351 *	−13.854 *

Source: Compilations from the author. Note: Evaluated at Gov = (.)^B: One standard deviation below the mean; (.)^M: The sample mean; (.)^A: One standard deviation above the mean. * indicates significance at 10%.

below a summary of our main findings. The first part of the table (section A) reports the point estimate obtained without the institutional quality interaction variable, i.e., the result depicted in column (1) of Table A3 in the Appendix A. This result reports a negative but non-significant Okun’s coefficient (−2.753).

Part B of Table 2 shows the “augmented” Okun’s coefficients when we include the interaction variable. These coefficients capture total unemployment responsiveness to output fluctuations given by the expression (${\alpha }_1+{\alpha }_3\ast Inst$) in equation (1) of the model. To assess this impact, we use the Gov variable (the average value of the six institutional quality indicators) and the point estimates reported in column 9 of Table A3. This impact is calculated at different levels of institutional quality: the mean value of Gov (M) in the sample, one standard deviation below this mean (B), and one standard deviation above this mean (A). The calculated impacts are much higher than those reported in the upper part of the table, significantly negative and increasing in absolute size as the institutional quality improves from B to M then to A. Countries with low institutional quality (at level B) have far lower impacts than those whose institutional quality is at the M or A levels. Indeed, the effect is −4.614 for a country with quality institutions at B, compared with −9.235 and −13.854 for countries with M and A levels of institutional quality. In other words, a one-point variation in the GDP gap would induce a deviation of the opposite sign in unemployment of around 4.6%, 9.2%, or 13.8% if the institutional quality is at the level of B, M, or A, respectively.

Therefore, these results unambiguously show that Okun’s coefficient is higher in countries with better institutional quality, as measured by the governance index (Gov). Specifically, this impact amplifies with greater Government Effectiveness (GE), better Control of Corruption (CC), better Regulatory Quality (RQ), more promoted Rule of Law (RL), and a higher level of Voice and Accountability (VA). These results tend to confirm some recent studies showing the harmful effects of bad governance on growth and sustainable development, suggesting adopting policies to improve the institutional quality in developing countries [36,37,38].

Comparing our Okun’s coefficient point estimates (an average of 9.2%) with estimates from previous studies, it appears that ours are much lower than those reported for the advanced countries, as in ref. [9], for example, where the estimates are 45% for the United States, 15% for Japan, and 80% for Spain. Our estimates, therefore, go against the results found in the recent study by [22], showing that, on average, Okun’s coefficient is about half as high in developing countries as it is in advanced countries. We argued that because developing countries have worse institutions than developed countries, the responsiveness of unemployment to GDP growth must be greater in the latter group and not the other way around. Furthermore, our conclusion concerning the role of institutions in determining the magnitude of Okun’s coefficient is helpful to understand why some previous studies have observed that this coefficient is high in countries with low unemployment and vice versa, as in [16]. We conjecture that low-unemployment countries are likely those with good institutions, and for this reason, they are those with high Okun’s coefficients, and vice versa. Finally, our hypothesis that institutional quality determines unemployment responsiveness to GDP fluctuations is also consistent with the conclusions of previous studies, suggesting unemployment rate adjustments are faster in more flexible labor-market countries [10,11,13,19,20], among others. However, these studies raised the institutional dimension only from the perspective of labor market conditions, which is only a tiny aspect of the institutional environment that is likely to affect unemployment responsiveness. On the contrary, in this paper, we have successfully demonstrated that Okun’s coefficient is also sensitive to other aspects of the institutional quality, such as governance indicators, with the adjustments being significantly higher when these indicators are better.

4. Conclusions

This study aimed to contribute to the recent debate raised in the empirical literature on Okun’s law regarding the magnitude of the responsiveness of unemployment to GDP fluctuations, its asymmetry and non-linearity, and the methodological issues in estimating this law. We tested the hypothesis that institutional quality—measured by governance indicators—is a possible explanation for the points raised in this debate. To this end, we tested an augmented version of Okun’s law represented by a system of two simultaneous equations taking into account the direct and indirect impacts of institutional quality on the responsiveness of unemployment to output gaps. We used annual data for a sample of developing and emerging countries over the period 1985–2019, used the 3sls regressor with country- and time-specific fixed effects, and controlled for the endogeneity of output fluctuations across a set of explanatory economic and institutional variables. The main results are as follows. First, the responsiveness of unemployment to changes in output is greater in countries with sounder institutions. Second, the magnitude of Okun’s coefficient in the sample of countries considered is lower than that estimated in previous work in the case of advanced economies. Finally, the GDP fluctuation variable is significantly affected by the economic and institutional explanatory factors considered in the model. These results contribute greatly to the recent debate on Okun’s law, with important policy implications for the group of developing and emerging countries. In particular, improving the institutional environment—by improving government effectiveness (GE), control of corruption (CC), regulatory quality (RQ), rule of law (RL), and voice and accountability (VA)—is as important as spurring economic growth in the strategy to reduce unemployment in these countries.