Impact of Unemployment and Governance on Poverty in Pakistan: A Fresh Insight from Non-linear ARDL Co-integration Approach

Abstract

This article aims to give a fresh insight into the non-linear relationship between unemployment, governance, and poverty in Pakistan. For the purpose, the study utilizes data from 1984 to 2016 by employing a nonlinear ARDL co-integration approach. The findings provide an insight that poverty responds asymmetrically due to positive or negative shocks in unemployment and governance. Moreover, the results suggest that applying linear models on poverty modelling may mislead the inference. The findings of the study imply that the policymakers and academicians must consider nonlinear behaviour of poverty for better policymaking.

Keywords

Poverty in Pakistan nonlinear co-integration hidden co-integration nonlinear ARDL

Introduction

Indeed, poverty is still an important economic irritant, specifically for developing nations like Pakistan. It is one of the most important challenges of the twenty-first century to reduce poverty (Alpízar & Ferraro, 2020). Worldwide, approximately 10 per cent of the world population or 700 million people are just striving for basic needs like safe drinking water, food, health and education and facing extreme poverty (United Nations, 2020).¹ About 188 million people are jobless globally (United Nations, 2020).² The recent statistics report a very painful fact that that globally almost 821 million people go to bed every night with an empty stomach (World Food Programme, 2020).³ These alarming statistics have highlighted importance of the issue and pushed international organizations and researchers worldwide to explore the underlying factors, determinants and causes to root out this social and economic disease called poverty.

The issue has grabbed the attention of not only policymakers but also academicians and researchers in the field of economics and finance. International literature has identified some major determinants of poverty including growth, development expenditures, unemployment, direct tax, gender inequality, inflation, private investment, productivity of agriculture sector, charity, literacy rate and governance (Dollar & Kraay, 2002; Grindle, 2004; Holmberg et al., 2009; Kwon & Kim, 2014; Mogale, 2005; Santos et al., 2019; Sarker & Rahman, 2007; Yousaf & Ali, 2014). In Pakistan, various researchers found various factors as a major determinant of poverty. Shirazi (1995) came forward with very interesting findings that Sadqat affect poverty significantly, Chaudhry et al. (2006) report agriculture growth rate, GDP growth rate, trade openness index and unemployment as major determinants of poverty in Pakistan, while Cheema and Sial (2012) found that household size, foreign remittances and education are the major factors affecting poverty. The present study selected unemployment and governance as major factors of poverty. On the flip side, the literature on poverty and unemployment suggest conflicting findings. For example, Gustafsson and Johansson (1999) are of the view that poverty had no relationship with unemployment. Further, Aiyedogbon and Ohwofasa (2012) found the positive and significant association of unemployment with poverty. Likewise, similar evidence of the positive relationship between unemployment, globalization, inequality and poverty was reported by Ukpere and Slabbert (2009). Moreover, Grindle (2004) claimed that good governance can significantly reduce poverty if the government plays its role honestly. It is imperative to mention here that literature reports institutional quality as a gauge of governance mechanism (Torgler & Schneider, 2009). Specifically, in developing countries, citizens, practitioners and academics consider good governance as a critical factor for bringing down the poverty figures as it is for encouraging development. Interestingly, researchers focused their studies more on finding the relationship of poverty with indicators of governance (government effectiveness, rule of law, absence of violence and political stability, voice and accountability, control of corruption, and regulatory quality) mentioned by World Bank (Kaufmann et al., 2013). However, these indicators provide mixed evidence like, corruption has an indirect influence on poverty (Chetwynd et al., 2003), while Khan (2009) claimed the positive relationship between the two. Also, lack of accountability, corrupt government, inadequate infrastructure and political instability aggravate the poverty situation (Rakodi et al., 2000).

An Overview of Poverty Level in Pakistan

Figure 1 reveals the relationship between poverty and unemployment in Pakistan. Poverty has been measured by the headcount ratio at PPP US$1.90 a day in the percentage of the total population, while unemployment is measured by the International Labor Organization (ILO) modelled estimate (total youth, % inactive population of age 15–64). From Figure 1 it is evident that the rate of poverty increased in 2001 and beyond that year, a declining trend has been observed, while the rate of unemployment remained almost stable from 1998 to 2010. Beyond 2010, the rate of unemployment in Pakistan started increasing and a gradual increase has been observed afterward. An interesting phenomenon is observed here that both the rates were equal in 2013. From Figure 1 it is found that poverty decreases with the increase in unemployment in Pakistan.

After an extensive review of the past literature on poverty modelling, it is found that poverty is one of the major concerns for developing counties like Pakistan. However, to the best of our knowledge, we observed these major studies in Pakistani context (Amjad & Kemal, 1997; Gillani et al., 2009; Jamal, 2006), have taken into account the symmetric behaviour of macro-economic factors affecting poverty. However, recent literature on the issue suggests that macroeconomic factors affect poverty in a nonlinear fashion. Ayala et al. (2011) found that unemployment affects poverty asymmetrically during periods of economic recession and expansion. Saleem et al. (2019) suggest that multi-dimensional poverty is significantly more in rural areas than in urban areas. A recent study carried out by Meo et al. (2018a, b) also claimed that various macroeconomic variables including unemployment affect poverty asymmetrically.

Empirical research provides many credible reasons for existence of asymmetric relationship between unemployment and poverty. Agénor (2002) proposed the Labor Hoarding hypothesis, which suggests that the main reason behind the asymmetric relationship between unemployment and poverty is an economic recession. According to the Labor Hoarding hypothesis, when it comes to an economic downturn, we may observe an interesting phenomenon that firms tend to downsize their labour force. However, skilled labour retains its positions while unskilled labour has to lose their jobs. This hoarding of skilled labour causes organizations to bear high turnover costs such as training and development cost, hiring and firing cost. However, this shock is deemed to be temporary regardless of its magnitude. With the passage of time, when economic activities are restored to their normal pace, firms try to fetch back forgone opportunities and wipe away economic loss during the period when the economy was in bad waters. They may rigorously attempt to reinstate high skilled labour to a state of high productivity. As, there is a greater degree of complementarity between fixed asset investment and use of skilled labour (though, elasticity between the factor is lower than substitution between the factors and unskilled labour), firms are more inclined to invest in tangibles instead of escalating demand for unskilled labour. Ultimately, wages of unskilled labour may fall proportionately more than the skilled labour. This way price of input capital and wages of unskilled workers will adjust downwards. Furthermore, firms’ inclination to hire high skilled labour in good times may be higher if the firm has experienced sunk cost associated with an investment in tangibles prior to the contraction in output. Therefore, this unique combination of low degree of substitution between skilled labour and fixed investment and high turnover costs may result in a formidable degree of relationship between poverty and unemployment, which is expected to continue for a longer period as a consequence of negative shock to output.

Literature also confirms that many of the macro-economic variables exhibit asymmetric properties specifically concerned with the business cycle (Falk, 1986; Neftci, 1984). Therefore, unemployment has been observed to have a non-linear effect on many macroeconomic variables (Cevik et al., 2013; Koutroulis et al., 2016). Furthermore, the current study also considers governance/institutional quality as a determinant of poverty. Kraay and Kaufmann (2002) suggest that institutional quality positively links with per capita income (a proxy of poverty). Moreover, it is observed that institutional quality/governance in the form of a better rule of law and corruption control directly affects people’s income, which ultimately leads to a decrease in poverty. Fosu (2017) examined that governance asymmetrically affects poverty through growth channel in Africa. Likewise, Doumbia (2019) also recently found that governance affects poverty asymmetrically.

Figure 1.

Source: World Development Indicators (WDI).

One of the major reasons behind the nonlinear effect of governance on poverty is the institutional instability of Pakistan. In the current era, nonlinearities among the macroeconomic variables got huge attention of the researchers worldwide. Previous studies on unemployment and governance are conducted in a linear framework, while many researchers argued that the linear model provides misleading results in the presence of nonlinear relationships among the variables. Kahneman and Amos (1979) highlighted the importance of nonlinearities in variables and suggested to account for the issue. According to Anoruo (2011), one of the limitations of linear modelling is that it assumes the variables to be linear; however, in reality, they exhibit nonlinearity. This implies that if we employ linear methodology for investigating factors of poverty, we may get spurious results. Bildirici and Turkmen (2015) explained that the models with nonlinear functional specifications have more explanatory power as these can incorporate the asymmetric behaviour and structural breaks in time variables. Besides, if negative and positive components of time series are co-integrated, then it may have hidden co-integration (Granger & Yoon, 2002). Furthermore, Po and Huang (2008) found that linear models also are unable to produce valid result in the presence of short-term changes and structural breaks. Therefore, considering the importance of nonlinear relationship among the proposed variables, it is the first study in Pakistani environment, which analysed the nonlinear relationship between unemployment, governance and poverty using Non-linear Autoregressive Distributed (NARDL) model or asymmetric ARDL co-integration approach. Hence, the purpose of the current study is to examine the nonlinear effect of unemployment and governance on the poverty of Pakistan.

The rest of the article is organized into four sections. A comprehensive review of the literature is discussed in the second section. The third section consists of methodology, which is followed by data analysis in the fourth section. The study ends with a conclusion and recommendations.

Literature Review

In recent literature on the determinants that influence poverty, researchers have identified a number of factors including direct tax, unemployment, earning per member of the family, development expenditures, inflation, GDP per capita, education level, social capital, gender inequality, private investment and quality of local governance (Jan et al., 2008; Mogale, 2005; Yousaf & Ali, 2014; Zhang et al., 2017). Past studies applied statistical models by assuming the linear relationship between these variables. The current study is an attempt to make a contribution to the existing literature by focusing on institutional quality/governance and unemployment to scrutinize their impact on the most important socio-economic problem, that is, poverty, by employing asymmetric Autoregressive Distributed Lag Model (ARDL).

According to International Labor Organization (ILO), global unemployment rate for 2017 is expected to be 13.1 per cent, which constitutes 71 million unemployed young people. Particularly for South Asian region, increasing trend of unemployment rate will continue in 2017 by claiming 13.9 million young people suffering from unemployment (ILO, 2016). Poverty and unemployment both have become, at present, a contemporary challenge of social existence for human beings worldwide.

Oduwole (2015) explored the relationship between unemployment and poverty through content analysis. He further identified that these issues have deep down roots in poor governance, leadership and security. Gallie et al. (2003) studied the vicious cycle of social exclusion on EU member states. They presented the strong evidence that unemployment intensifies the possibility of social isolation and poverty and in turn, poverty and social exclusion make it more problematic for people to get back to employment.

There are various studies that examined the relationship between unemployment and poverty. This relationship has been examined in two fashions. First, in current literature, authors explored nonlinear association and secondly, many authors studied relationships in a linear framework. In a recent study, Meo et al. (2018a) examined the asymmetric association between unemployment and poverty in Pakistan using NARDL approach. They found that a nonlinear relationship exists between poverty and unemployment. Aiyedogbon and Ohwofasa (2012) estimated the impact of unemployment on poverty and confirmed that unemployment positively affects poverty in Nigeria. Gillani et al. (2009), using a co-integration approach, also report that unemployment positively and significantly affects poverty. Another evidence of Ukpere and Slabbert (2009) argued a positive and significant relationship between unemployment and poverty. Furthermore, Gillani et al. (2009) confirmed, in Pakistan, that unemployment and poverty significantly correlate with each other in the long run. A study by Saunders (2002) argued that unemployment increases poverty and inequality, while Martínez et al. (2001) also examined the association between unemployment and poverty in OECD. They found that unemployment is one of the major underlying factors behind poverty.

Apart from many other determinants of poverty, researchers have also shed light on the vital role of governance on poverty eradication (Grindle, 2004; Holmberg et al., 2009; Kwon & Kim, 2014; Mogale, 2005; Sarker & Rahman, 2007). Literature has offered several definitions of governance. For example, Kaufmann et al. (2013) expressed governance as ‘Traditions and institutions by which authority in a country is exercised for the common good’. In addition, it involves the procedures through which governments are chosen and replaced; the government’s capability of effectively devising and executing policies; and veneration for the institutions that reign social and economic interactions. Researchers at World Bank have recently reported six indicators of governance quality that incorporate government effectiveness, rule of law, absence of violence and political stability, voice and accountability, control of corruption and regulatory quality (Kaufmann et al., 2013). Grindle (2004) argued that good governance can be an important factor for reducing or alleviating poverty if the role of government is considered sincerely.

Moreover, scholars have also scrutinized some governance indicators with poverty specifically. Like, Gupta et al. (2002) employed OLS and instrumental variable (IV) techniques to investigate the link of corruption with poverty and income inequality. Their results revealed a positive relationship and suggest that growing corruption proliferate poverty and income inequality. Kumar (2019a) used logit model to investigate the linkage between international remittances and poverty. This article revealed that if a household receives international remittances, the probability of that household may be reduced by 28.07 per cent. The same direction of linkage was found by Kumar (2019b).

A study conducted by Chetwynd et al. (2003) contended that corruption has an indirect impact on poverty via economic model and governance model. Governance model claims that increasing corruption firstly induces government institutions for producing high-quality services, promotes capital projects rather than investing in public needs and pulls down the conformity to health and safety regulations. These factors posit severe challenges to good governance practices and hence stimulate poverty. According to the economic model, corruption reduces investment, saddles with the competition, augment the cost of conducting businesses and upsurge income inequalities. The undermining of these economic factors exacerbates poverty. Many researchers (Kaufmann & Bellver, 2005; Kaufmann et al., 2006; Mauro, 1997; Yusuf et al., 2014) contended the indirect link between corruption and poverty and thus validated the above-mentioned direction of the economic model while few asserted the direct link between the two (Khan, 2009).

Also, Rakodi et al. (2000) studied that political instability, lack of accountability and corruption in government, and poor infrastructure played their role in uplifting poverty outlook in Mombasa. In addition, African’s poverty and deprivation were assessed by Mbaku (2014), who professed that a proper constitution and effective institutional arrangements for law enforcement can eradicate poverty.

As reported in the literature, poverty is one of the contemporary issues worldwide and thus researchers are trying hard to dig out its determining factors to develop effective policies. It is observed that time series-based research on this phenomenon in South Asian countries has not gained attention yet. Findings on unemployment relationship with poverty were contrasting in direction and governance variables remain uninvestigated as a determinant of poverty in Pakistan. These reasons make our study more promising and justified to be estimated by filling the gap in existing literature from both theoretical and methodological standpoints. Interestingly, past studies on poverty with its contributing factors assumed their linear relationship that is somehow a restrictive assumption when the relationship is non-linear. Furthermore, the functional forms of the relationships were ignored and focus remained on scrutinizing the causes or factors of poverty. Bildirici and Turkmen (2015) explained that the models with nonlinear functional specifications have more explanatory power as these can incorporate the asymmetric behaviour and structural breaks in time variables. Besides, if negative and positive components of time series are co-integrated, then it may have hidden co-integration (Granger & Yoon, 2002). Hence, for all these reasons, this study applies the asymmetric Autoregressive Distributed Lag (ARDL) co-integration model to estimate the relationship of unemployment and governance on poverty in Pakistani context to satiate the research gap on non-linear poverty modelling.

Methodology

Data Sources

This study investigates the asymmetric (non-linear) relationship between unemployment, governance and poverty using annual data from 1984 to 2016. Due to the gap in poverty (headcount ratio) data, we interpolated the data with the help of the linear interpolation method. This method searches for a straight line that passes through the endpoints A and B (Lepot et al., 2017). A study by Chen et al. (2012) argued that linear interpolation is an easy method of interpolation. Gnauck (2004) suggested a linear interpolation as one of the efficient methods that produce more reliable missing values as compared to the non-linear interpolation. Furthermore, for the robustness checking of the outcomes, we also used ‘Population below national poverty line’ as second a proxy of poverty. Information regarding the units and data sources of other variables is presented in Table 1.

Table 1.

Data Sources and Description

Variables	Variables Description	Units of Measurement	Data Sources	Expected Relationship
Pov	Poverty	Head count ratio^a	WDI
Unemp	Unemployment	Youth unemployment rate	WDI	Positive
Gov	Governance index	Index	ICRG	Negative

Source: The authors.

Note: ^a Poverty headcount ratio at US$1.90 a day is the percentage of the population.

Empirical Model

The functional form of model in linear framework is stated below.

Pov = f (Unemp and Gov)

As suggested by the literature, the relationship between two or more variables such as unemployment, governance and poverty is normally explored by a means of the standard time series techniques such as co-integration, error-correction modelling, and Granger causality. Although all these techniques can evaluate both long-run and short-run relations, they assume symmetric relations only. They are not enough to capture potential asymmetries in the relations. Recently, Shin et al. (2014) developed a nonlinear ARDL co-integration approach (NARDL) as an asymmetric extension to the well-known ARDL model of Pesaran et al. (1999) and Pesaran et al. (2001), which captures both long-run and short-run asymmetries. This study adopts this approach for capturing both the long- and short-run asymmetric effects of unemployment and governance on poverty. The specific asymmetric long-run equation of poverty is as follows:

{Pov}_{t} = \emptyset_{0} + \emptyset_{1} ({Unemp}_{t}) + \emptyset_{2} ({Gov}_{t}) + μ_{t}

(1)

In Equation (1), Pov, Unemp and Gov refer to poverty, unemployment and governance index respectively. While long-term coefficients and error terms are denoted by $\emptyset_{0}, \emptyset_{1}, \emptyset_{2} and μ_{t}$ , respectively. This study employs the Autoregressive Distributed Lag (ARDL hereafter) model of Pesaran et al. (2001) to find long-term and short-term relationships among proposed variables. The ARDL approach to co-integration has various advantages over other co-integration approaches. First, this technique works well (Romilly et al., 2001). Johansen’s (1988) process of estimating long-term association among the purposed variables required I(0) order of integration. Like other co-integration approaches, the ARDL model is equally applicable to the series, which are stationary at I(0) or I(1) or I(0) and I(1). According to Ibrahim (2015), stationary checking is not compulsory for the Autoregressive Distributed Lag (ARDL) model. However, this approach cannot be applied in the presence of any variable, which is stationary at second order I(2), because, in the presence of I(2) variables, the F-statistic value of co-integration test becomes invalid (Ibrahim, 2015).

Considering the advantages of the ARDL model, this study formulates the following general ARDL equation in a linear framework.

Δ {Pov}_{t} = \emptyset_{0} + \emptyset_{1} {Pov}_{t - i} + \emptyset_{2} {Unemp}_{t - i} + \emptyset_{3} {Gov}_{t - i} + \sum_{i = 1}^{a} \emptyset_{4} {Pov}_{t - i} + \sum_{i = 1}^{b} \emptyset_{4} {Unemp}_{t - i} + \sum_{i = 1}^{c} \emptyset_{4} {Gov}_{t - i} + μ_{t}

(2)

In the Equation (2) $\emptyset_{1}, \emptyset_{2} and \emptyset_{3}$ refers to long-term coefficients, while $\emptyset_{4}, \emptyset_{5} and \emptyset_{6}$ are short-term coefficients of linear ARDL model. However, the prime objective of the study is to investigate the long-term asymmetric (nonlinear) relationship between unemployment, governance, and poverty. The phenomena of hidden co-integration, discussed in Granger and Yoon’s (2002) study, claimed that if the positive and negative components of any variable are co-integrated with each other, it is called hidden co-integration. To consider the asymmetric relationship of unemployment and governance on poverty, we followed the following nonlinear co-integration regression.

{Pov}_{t} = β^{+} {Unemp}^{+}_{t} + β^{-} {Unemp}^{-}_{t} + β^{+} {Gov}^{+}_{t} + β^{-} {Gov}^{-}_{t} + μ_{t}

(3)

Equation (3) refers to the long-term coefficients of partial sums of positive and negative changes of unemployment and governance respectively. This article derived partial sums of positive and negative shocks of unemployment and governance as following from Equations (4) to (7).

{Unemp}_{t}^{+} = \sum_{i = 1}^{t} {ΔUnemp}_{t}^{+} = \sum_{i = 1}^{t} \max ({ΔUnemp}_{i}, 0)

(4)

{Unemp}^{-} = \sum_{i = 1}^{t} {ΔUnemp}_{t}^{-} = \sum_{i = 1}^{t} \min ({ΔUnemp}_{i}, 0)

(5)

{Gov}_{t}^{+} = \sum_{i = 1}^{t} {ΔGov}_{t}^{+} = \sum_{i = 1}^{t} \max ({ΔGov}_{i}, 0)

(6)

{Gov}_{t}^{-} = \sum_{i = 1}^{t} {ΔGov}_{t}^{-} = \sum_{i = 1}^{t} \min ({ΔGov}_{i}, 0)

(7)

Equation (2) only provides the long-term and short-term relationship of unemployment, governance and poverty in symmetric (linear framework). However, our concern is to check nonlinear relationships among purposed variables; therefore, this study followed nonlinear procedure advanced by Shin et al. (2014) by incorporating negative and positive shocks of unemployment and governance in Equation (2) to make model nonlinear (asymmetric). The following Equation (8) is a general form of a non-linear ARDL model.

\begin{array}{l} Δ {Pov}_{t} = \emptyset_{0} + \emptyset_{1} {Pov}_{t - i} + \emptyset_{2} {Unemp}^{+}_{t - i} + \emptyset_{3} {Unemp}^{-}_{t - i} + \emptyset_{4} {Gov}^{+}_{t - i} + \emptyset_{5} {Gov}^{-}_{t - i} + \\ \sum_{i = 1}^{a} \emptyset_{6} {Pov}_{t - i} + \sum_{i = 1}^{b} \emptyset_{7} {Unemp}^{+}_{t - i} + \sum_{i = 1}^{c} \emptyset_{8} {Unemp}^{-}_{t - i} + \sum_{i = 1}^{b} \emptyset_{9} {Gov}^{+}_{t - i} + \sum_{i = 1}^{c} \emptyset_{10} {Gov}^{-}_{t - i} + μ_{t} \end{array}

(8)

Now in Equation (8) ARDL approach to co-integration is a nonlinear (asymmetric) framework, which produces long-term and short-term asymmetric coefficients.

The empirical implementation of the NARDL approach incurs the following steps. First, the stationarity of all variables is examined by unit root tests through the Augmented Dickey–Fuller and Phillips–Perron test. It is explored to check the order of integration of variables, that is, whether they are integrated by order zero or one, namely, I(0) or I(1). In the second step, Equation (6) is estimated by the ordinary least squares (OLS) method. Also, to arrive at the final specification of the NARDL model by trimming insignificant lags, the information criterion SIC or general-to-specific is adopted following Katrakilidis and Trachanas (2012). In the third step, a test is performed for checking the presence of co-integration among variables using a bound testing approach of Pesaran et al. (2001) and Shin et al. (2014), which involves the Wald F test of the null hypothesis. Finally, with the presence of co-integration, examination of long-run and short-run asymmetries in the relations between unemployment and governance and poverty is made and inferences are drawn. In the third step, a test is performed for checking the presence of cointegration among variables using a bounds testing approach of Pesaran et al. (2001) and Shin et al. (2014) which involves the Wald F test of the null hypothesis, $φ_{1} = φ_{2} = φ_{3} = φ_{4} = 0$ . Finally, with the presence of cointegration, estimation of long-run and short-run asymmetries in the relations between unemployment and governance and, poverty is made and inferences are drawn. In this step, the cumulative dynamic multiplier effects of both positive and negative of explanatory variables is also derived.

Results

The summary statistics of the variables are presented in Table 2. The unit root checking is the most important pre-condition in time series analysis. Therefore, this study employed an ADF (Augmented Dickey-Fuller test) and PP (Phillips-Perron test). Table 3 demonstrates the results of ADF and PP unit root tests. It is found that all the variables have unit root except Gov in level. However, all the remaining series (poverty and unemployment) become stationary after their first difference. Hence, the order of integration of the series is a mixture of level and first differences, that is, I(0) and I(1). According to Sezgin and Yildirim (2002), for the ARDL co-integration approach, unit root checking is not compulsory because the autoregressive distributed lag model is equally applicable on the series, which are stationary at I(0) or I(1) or a mixture of I(0) and I(1). However, this approach cannot be applied in the presence of any variable, which is stationary at second order I(2), because, in the presence of I(2) variables, the F-statistic value of the co-integration test becomes invalid (Ibrahim, 2015; Ouattara, 2004). Therefore, in the ARDL approach, unit root checking is still important. The outcomes of ADF and PP confirm the absence of any variable, which is stationary at second order; therefore, this study can proceed to the ARDL model.

Table 2.

Descriptive Statistics

	Pov	Unemp	Gov
Mean	21.90833	5.589313	–5.21E-16
Median	16.20000	5.652500	–0.217757
Maximum	62.20000	8.270000	3.788403
Minimum	3.900000	3.071000	–3.589834
Std. dev.	19.42559	1.498787	1.534017
Skewness	1.297620	–0.039041	0.287135
Kurtosis	3.314152	2.297045	3.223728
Jarque–Bera	3.416981	0.666990	0.458973
Probability	0.181139	0.716415	0.794942

Source: The authors.

Table 3.

Unit Root Tests

Variables	Unit Root at Level I(0)	Unit Root at Level I(I)
Augmented Dickey–Fuller test statistic
Pov	0.616 (0.843)	–6.706 (0.001)*
Unemp	–2.092 (0.530)	–3.151 (0.044)**
Gov	–2.595 (0.012)**	–4.043 (0.001)*
Phillips–Perron test statistic
Pov	–0.282 (0.576)	–16.919 (0.000)*
Unemp	0.156 (0.725)	–6.766 (0.000)*
Gov	–2.416 (0.018)**	–4.062 (0.002)*

Source: The authors.

Note: * and ** Refer to level of significance at 1% and 5%, respectively.

There are various studies on nonlinear ARDL estimation which employed general to specific approach (Ibrahim, 2015; Katrakilidis & Trachanas, 2012; Meo et al., 2018b; Shin et al., 2014). The prime objective of this approach is to remove all the significant lags of the regression to avoid noise in dynamic multipliers (Katrakilidis & Trachanas, 2012). Furthermore, we employed a maximum of two lags for regressors and endogenous variables. Before proceeding with NARDL estimation, various diagnostic tests are being performed such as Ramsey RESET test for the specification issue, Jarque–Bera test for the normality of residuals and Breusch/Pagan heteroscedasticity test. The results of all the diagnostic tests are presented in Table 4. It is confirmed that the model is free from any of the said problem; therefore, this study can be used for NARDL estimation.

Table 4.

Dynamic Asymmetric Estimation of Poverty

Variables	Coef.	Std. Err.	t-ratio	Sig.
$Pov(- 1)$	–1.299	0.172	–7.77	0.004
$Unemp_P$	2.418	0.657	3.68	0.021
$Unemp_N$	0.270	0.125	2.19	0.043
$Gov_P$	–1.976	0.575	–3.44	0.026
$Gov_N$	–.381	0.072	–5.26	0.030
$Δ Pov (- 1)$	1.238	0.364	3.40	0.027
$Δ Pov(- 2)$	0.065	0.184	0.36	0.740
$Δ Unemp_P$	0.137	0.247	0.55	0.609
$Δ Unemp(- 1)$	–1.717	0.492	–3.49	0.025
$Δ Unemp (- 2)$	–0.604	0.409	–1.48	0.213
$Δ Unemp_N$	–0.957	0.398	–2.40	0.074
$Δ Unemp_N(- 1)$	2.030	0.727	2.79	0.049
$Δ unemp_N (- 2)$	–0.363	0.598	–0.61	0.577
$Δ Gov_P$	–6.026	1.311	–4.60	0.010
$Δ Gov_P(- 1)$	–0.243	0.567	–0.43	0.691
$Δ Gov_P(- 2)$	–2.155	0.768	–2.81	0.049
$Δ Gov_N$	0.915	0.447	2.05	0.110
$Δ Gov_N(- 1)$	–1.080	0.515	–2.10	0.104
$Δ Gov_N(- 2)$	–0.719	0.432	–1.67	0.171
C	30.584	5.717	5.35	0.006
Adj R-squared		0.930

Source: The authors.

Table 5.

Asymmetric Co-integration Using Bounds Test

Co-integration test statistics:	t_BDM = –5.770
	F_PSS = 8.745

Source: The authors.

Table 5 represents the results of nonlinear co-integration among unemployment, governance/institutional quality, and poverty. The null hypothesis of the co-integration is that there is no co-integration. The F_PSS values refer to the F-statistic value by Pesaran and t_BDM refers to the t-statistic of Banerjee et al. (1998) to test the null hypothesis of no co-integration. However, from Table 5, it is confirmed that there is a long-term relationship between unemployment, institutional quality and poverty. Therefore, this study can move forward for NARDL estimation and examine the long-term asymmetric relationship among the purposed variables.

Table 5 reveals the long-term asymmetric relationship between unemployment, governance and poverty. It is found that the positive and negative components of unemployment () are significantly associated with poverty in the long run with positive signs. The long-run asymmetric outcomes show that 1 per cent increase in Unemp _P leads to a positive change in poverty by 1.855 per cent, while a 1 per cent decrease in Unemp_N reduces poverty by 0.207 per cent. The findings are in line with the study of (Meo et al., 2018a; Ukpere & Slabbert, 2009). It implies that positive change in unemployment has a larger effect on poverty compared to a negative change in unemployment. As suggested by Taylor and Saunders (2002), there are plenty of social costs associated with unemployment. It ranges from the psychological and physical health of an individual to an adverse effect on family and community life. The current study empirically supports the argument that the increasing effect of unemployment on poverty is larger and far-reaching than the decreasing effect. This study also found that positive and negative components of Gov are significantly associated with poverty in the long-run with negative signs. A 1 per cent increase in Gov_P leads to a negative change in poverty by 1.52 per cent, while a 1 per cent decrease in Gov_N increases poverty by 0.29 per cent. It implies that positive change governance has a larger effect on poverty as compared to a negative change in governance. These interesting sets of findings indicate the most vital role of governance in alleviating poverty through a high employment rate. The results specify a larger effect of better-quality governance on poverty reduction. Strong governance results in the rule of law and better institutional quality, which leads to a long-lasting positive societal impact. The study has strong implications for policymakers that how improved governance has extended social impact. The findings are in line with Karim et al. (2013) and Meo et al. (2018a). We can easily deduce from Table 6 whether asymmetric relationship exists or not among our proposed variables on the basis of differences in negative and positive components. However, this article also employed Wald test to confirm asymmetric long-run relationships, results presented in Table 7, though asymmetric relationships among the proposed variables also can be confirmed from the dynamic multiplier graph, presented in Figure 2.

Table 6.

Long-term Asymmetric Relationship

Exog. var.	Coef.	F-stat	Sig.
	1.855	5.94	0.038
	0.207	7.69	0.015
	–1.521	6.69	0.032
	–0.293	14.75	0.008

Source: The authors.

Table 7.

Long-run and Short-run Asymmetry

Exog. var.	Long-run Asymmetry		Short-run Asymmetry
Exog. var.	F-stat	Sig.	F-stat	Sig.
	20.4	0.011	2.624	0.181
	19.59	0.011	17.58	0.014

Source: The authors.

Table 8.

Dynamic Asymmetric Estimation of Poverty and Long-run Coefficients (robustness checking with population below national poverty line as a proxy of poverty)

Variables	Coef.	Std. Err.	t-Ratio	Sig.
Pov (–1)	–0.934	0.042	–22.238	0.000
Unemp_P	0.705	0.033	21.363	0.000
Unemp_N	0.122	0.027	4.518	0.001
Gov_P	–0.782	0.098	–7.979	0.000
Gov_N	–0.097	0.019	–5.105	0.030
ΔUnemp_P	0.352	0.112	3.142	0.032
ΔUnemp_N	–0.652	0.019	–34.315	0.000
ΔGov_P	–0.863	0.923	–0.934	0.213
ΔGov_N	0.893	0.447	1.997	0.111
C	10.273	0.717	14.327	0.000
LUnemp_P	0.754*
LUnemp_N	0.173*
LGov_P	–6.409*
LGov_N	–0.124*
Adj R-squared		0.231
F-statistic		10.821
Ramsey RESET test		0.313
Jarque–Bera test		0.813
Breush/Pagan Heteroskedasticity test		0.132

Source: The authors.

Note: LUnemp_P, LUnemp_N, LGov_P and LGov_N refers to long run coefficients and * means 1% level of significance.

Table 8 shows results of dynamic estimation and long-run coefficients of poverty with a population below the national poverty line as a proxy of poverty. The findings are consistent as shown in Table 4.

This study has also checked stability of estimated long-run parameters by applying CUSUM (Cumulative Sum) and CUSUMSQ (CUSUM of Squares).⁴ Figures 3 and 4 indicate that CUSUM and CUSUMSQ are inside the critical lines, which confirms that model is free from sudden change, or structural breaks.

Figure 2.

Source: The authors.

Figure 3.

Source: The authors.

Figure 4.

Source: The authors.

Conclusion and Discussion

The prime objective of this study is to investigate asymmetric relationship between unemployment, institutional quality and poverty using annual data from 1984 to 2016. The asymmetric relationship is checked using NARDL model recently advanced by Shin et al. (2014). The findings of the study confirmed a nonlinear positive and significant relationship between unemployment and poverty. It is found that positive shocks in unemployment have almost double effect on poverty than negative shock. The core reason of this finding is that Pakistan is one of the poor countries, where unemployment is a major issue. Furthermore, this article also found that quality of governance/institutional quality also has a significant and negative effect on poverty and it is confirmed that positive change in governance has a larger effect on poverty as compared to negative change.

Footnotes

Acknowledgement

This article has not received any funds from any government and non-government organizations. However, authors of this article are grateful to the anonymous referees of the journal for their extremely useful suggestions to improve the quality of the article. The authors are also grateful to BK School of Research for its technical support.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

ORCID iDs

Bezon Kumar

Sumayya Chughtai

Notes

References

Amjad

, & Kemal

A. R.

(1997). Macroeconomic policies and their impact on poverty alleviation in Pakistan. The Pakistan Development Review , 39–68.

Agénor

P. R.

(2002). Business cycles, economic crises, and the poor. The Journal of Policy Reform , 5(3), 145–160.

Aiyedogbon

J. O.

, & Ohwofasa

B. O.

(2012). Poverty and youth unemployment in Nigeria, 1987–2011. International Journal of Business and Social Science , 3(20), 269–279.

Alpízar

, & Ferraro

P. J.

(2020). The environmental effects of poverty programs and the poverty effects of environmental programs: The missing RCTs. World Development , 127, 104783.

Anoruo

(2011). Testing for linear and nonlinear causality between crude oil price changes and stock market returns. International Journal of Economic Sciences and Applied Research , 4(3), 75–92.

Ayala

, Cantó

, & Rodríguez

J. G.

(2011). Poverty and the business cycle: The role of the intra-household distribution of unemployment (ECINEQ WP, 222).

Banerjee

, Dolado

, & Mestre

(1998). Error-correction mechanism tests for cointegration in a single-equation framework. Journal of Time Series Analysis , 19(3), 267–283.

Bildirici

M. E.

, & Turkmen

(2015). Nonlinear causality between oil and precious metals. Resources Policy , 46, 202–211.

Brown

R. L.

, Durbin

, & Evans

J. M.

(1975). Techniques for testing the constancy of regression relationships over time. Journal of the Royal Statistical Society . Series B

10.

Cevik

E. I.

, Dibooglu

, & Barişik

(2013). Asymmetry in the unemployment–output relationship over the business cycle: Evidence from transition economies. Comparative Economic Studies , 55(4), 557–581.

11.

Chaudhry

I. S.

, Malik

, & Ashraf

(2006). Rural poverty in Pakistan. Pakistan Economic and Social Review , 44(2), 259–276.

12.

Cheema

A. R.

, & Sial

M. H.

(2012). Incidence, profile and economic determinants of poverty in Pakistan: HIES 2005–06. Management Science and Engineering , 6(2), 120–129.

13.

Chen

Z. Y.

, Fan

Z. P.

, & Sun

(2012). A hierarchical multiple kernel support vector machine for customer churn prediction using longitudinal behavioral data. European Journal of Operational Research , 223(2), 461–472.

14.

Chetwynd

, Chetwynd

, & Spector

(2003). Corruption and poverty: A review of recent literature. Management Systems International , 600, 5–16.

15.

Dollar

, & Kraay

(2002). Growth is good for the poor. Journal of Economic Growth , 7(3), 195–225.

16.

Doumbia

(2019). The quest for pro-poor and inclusive growth: The role of governance. Applied Economics , 51(16), 1762–1783.

17.

Falk

(1986). Further evidence on the asymmetric behavior of economic time series over the business cycle. Journal of Political Economy , 94(5), 1096–1109.

18.

Fosu

A. K.

(2017). Governance and development in Africa: A concise review (No. 082017). GDI, The University of Manchester.

19.

Gallie

, Paugam

, & Jacobs

(2003). Unemployment, poverty and social isolation: Is there a vicious circle of social exclusion. European Societies , 5(1), 1–32.

20.

Gillani

S. Y. M.

, Rehman

H. U.

, & Gill

A. R.

(2009). Unemployment, poverty, inflation and crime nexus: Cointegration and causality analysis of Pakistan. Pakistan Economic and Social Review , 47(1), 79–98.

21.

Gnauck

(2004). Interpolation and approximation of water quality time series and process identification. Analytical and Bioanalytical Chemistry , 380(3), 484–492.

22.

Granger

C. W.

, & Yoon

(2002). Hidden cointegration (Economics Working Paper, (2002–02)). University of California. http://papers.ssrn.com/sol3/papers.cfm

23.

Grindle

M. S.

(2004). Good enough governance: Poverty reduction and reform in developing countries. Governance , 17(4), 525–548.

24.

Gupta

, Davoodi

, & Alonso-Terme

(2002). Does corruption affect income inequality and poverty. Economics of Governance , 3(1), 23–45.

25.

Gustafsson

, & Johansson

(1999). In search of smoking guns: What makes income inequality vary over time in different countries. American Sociological Review , 64(4), 585–605.

26.

Holmberg

, Rothstein

, & Nasiritousi

(2009). Quality of government: What you get. Annual Review of Political Science , 12, 135–161.

27.

Ibrahim

M. H.

(2015). Oil and food prices in Malaysia: A nonlinear ARDL analysis. Agricultural and Food Economics , 3(1), 1–14.

28.

ILO. (2016). World employment and social outlook. International Labour Organization. Methodological , 37(2), 149–192.

29.

Jamal

(2006). Does inequality matter for poverty reduction? Evidence from Pakistan's poverty trends. The Pakistan Development Review , 439–459.

30.

Jan

, Chishti

A. F.

, & Eberle

P. R.

(2008). An analysis of major determinants of poverty in agriculture sector in Pakistan (No. 382–2016–22593).

31.

Johansen

(1988). Statistical analysis of cointegration vectors. Journal of economic dynamics and control , 12(2–3), 231–254.

32.

Kahneman

, & Amos

(1979). Prospect theory: An analysis of decision under risk. Econometrica , 47, 263–292.

33.

Karim

K. M.

, Zouhaier

, & Adel

B. H.

(2013). Poverty, governance and economic growth. Journal of Governance and Regulation , 2(3), 19–24.

34.

Katrakilidis

, & Trachanas

(2012). What drives housing price dynamics in Greece: New evidence from asymmetric ARDL cointegration. Economic Modelling , 29(4), 1064–1069.

35.

Kaufmann

, & Bellver

(2005). Transparenting transparency: Initial empirics and policy applications. Available at SSRN 808664.

36.

Kaufmann

, Kraay

, & Mastruzzi

(2006). Governance matters V: Aggregate and individual governance indicators for 1996–2005 . The World Bank.

37.

Kaufmann

, Kraay

, & Mastruzzi

(2013). The Worldwide Governance Indicators (WGI) project . World Bank.

38.

Khan

(2009). Is ‘good governance’ an appropriate model for governance reforms? The relevance of East Asia for developing Muslim countries . In Development models in muslim contexts: Chinese, “Islamic” and Neo-Liberal Alternatives (pp. 195–230). Edinburgh University Press.

39.

Koutroulis

, Panagopoulos

, & Tsouma

(2016). Asymmetry in the response of unemployment to output changes in Greece: Evidence from hidden co-integration. The Journal of Economic Asymmetries , 13, 81–88.

40.

Kumar

(2019a). The impact of international remittances on poverty alleviation in Bangladesh. Remittances Review , 4(1), 67–86.

41.

Kumar

(2019b). Remittances, poverty and welfare: Evidence from Cumilla, Bangladesh. American Journal of Data Mining and Knowledge Discovery , 4(1), 46–52.

42.

Kraay

, & Kaufmann

(2002). Growth without governance . The World Bank.

43.

Kwon

H. J.

, & Kim

(2014). Poverty reduction and good governance: Examining the rationale of the millennium development goals. Development and Change , 45(2), 353–375.

44.

Lepot

, Aubin

J. B.

, & Clemens

F. H.

(2017). Interpolation in time series: An introductive overview of existing methods, their performance criteria and uncertainty assessment. Water , 9(10), 796.

45.

Martínez

, Ayala

, & Ruiz-Huerta

(2001). The impact of unemployment on inequality and poverty in OECD countries. Economics of Transition , 9(2), 417–447.

46.

Mauro

(1997). Corruption and growth. Trends in Organized Crime , 2(4), 67.

47.

Mbaku

J. M.

(2014). The rule of law and poverty eradication in Africa. African and Asian Studies , 13(4), 530–557.

48.

Meo

M. S.

, Chowdhury

M. A. F.

, Shaikh

G. M.

, Ali

, & Masood

S. S.

(2018a). Asymmetric impact of oil prices, exchange rate, and inflation on tourism demand in Pakistan: New evidence from nonlinear ARDL. Asia Pacific Journal of Tourism Research , 23(4), 408–422.

49.

Meo

M. S.

, Khan

V. J.

, Ibrahim

T. O.

, Khan

, Ali

, & Noor

(2018b). Asymmetric impact of inflation and unemployment on poverty in Pakistan: New evidence from asymmetric ARDL cointegration. Asia Pacific Journal of Social Work and Development , 28(4), 295–310.

50.

Mogale

T. M.

(2005). Local governance and poverty reduction in South Africa (I). Progress in Development Studies , 5(2), 135–143.

51.

Neftci

S. N.

(1984). Are economic time series asymmetric over the business cycle. Journal of Political Economy , 92(2), 307–328.

52.

Oduwole

T. A.

(2015). Youth unemployment and poverty in Nigeria. International Journal of Sociology and Anthropology Research , 1(2), 23–39.

53.

Ouattara

(2004). Modelling the long run determinants of private investment in Senegal (No. 04/05). Credit Research Paper.

54.

Pesaran

M. H.

, Shin

, & Smith

R. J.

(2001). Bounds testing approaches to the analysis of level relationships. Journal of Applied Econometrics , 16(3), 289–326.

55.

Pesaran

M. H.

, Shin

, & Smith

R. P.

(1999). Pooled mean group estimation of dynamic heterogeneous panels. Journal of the American Statistical Association , 94(446), 621–634.

56.

W. C.

, & Huang

B. N.

(2008). Tourism development and economic growth–A nonlinear approach. Physica A: Statistical Mechanics and its Applications , 387(22), 5535–5542.

57.

Rakodi

, Gatabaki-Kamau

, & Devas

(2000). Poverty and political conflict in Mombasa. Environment and Urbanization , 12(1), 153–170.

58.

Romilly

, Song

, & Liu

(2001). Car ownership and use in Britain: A comparison of the empirical results of alternative cointegration estimation methods and forecasts. Applied Economics , 33(14), 1803–1818.

59.

Saleem

, Shabbir

M. S.

, & Khan

(2019). Re-examining multidimensional poverty in Pakistan: A new assessment of regional variations. Global Business Review . doi:10.1177/0972150919844412.

60.

Santos

M. E.

, Dabus

, & Delbianco

(2019). Growth and poverty revisited from a multidimensional perspective. The Journal of Development Studies , 55(2), 260–277.

61.

Sarker

A. E.

, & Rahman

M. H.

(2007). The emerging perspective of governance and poverty alleviation: A case of Bangladesh. Public Organization Review , 7(2), 93–112.

62.

Saunders

(2002). The direct and indirect effects of unemployment on poverty and inequality. Australian Journal of Labour Economics , 5(4), 507.

63.

Sezgin

, & Yildirim

(2002). The demand for Turkish defence expenditure. Defence and Peace Economics , 13(2), 121–128.

64.

Shin

, Yu

, & Greenwood-Nimmo

(2014). Modelling asymmetric cointegration and dynamic multipliers in a nonlinear ARDL framework. In Festschrift in honor of Peter Schmidt (pp. 281–314). New York, NY: Springer.

65.

Shirazi

N. S.

(1995). Determinants of poverty in Pakistan. Pakistan Economic and Social Review , 33(1/2), 91–101.

66.

Taylor

, & Saunders

(Eds.). (2002). The price of prosperity: The economic and social costs of unemployment . UNSW Press.

67.

Torgler

, & Schneider

(2009). The impact of tax morale and institutional quality on the shadow economy. Journal of Economic Psychology , 30(2), 228–245.

68.

Ukpere

W. I.

, & Slabbert

A. D.

(2009). A relationship between current globalization, unemployment, inequality and poverty. International Journal of Social Economic , 36(1/2), 37–46.

69.

Yousaf

, & Ali

(2014). Determinants of poverty in Pakistan. International Journal of Economics and Empirical Research (IJEER) , 2(5), 191–202.

70.

Yusuf

, Malarvizhi

C. A.

, Mazumder

M. N. H.

, & Su

(2014). Corruption, poverty, and economic growth relationship in the Nigerian economy. The Journal of Developing Areas , 48(4), 95–107.

71.

Zhang

, Zhou

, & Lei

(2017). Social capital and its contingent value in poverty reduction: Evidence from Western China. World Development , 93, 350–361.