Investigating Okun’s Law in SAARC Countries: An ARDL Approach

Purpose: The aim of this paper is to test the applicability of Okun’s law in SAARC (South Asian Association for Regional Cooperation) counties. It also intends to nd the long-run association between unemployment rate and growth rate and investigate the impact of growth rate on unemployment in the South Asian Region. Design/methodology/approach: The study uses annual time series data for eight SAARC countries, from 1991 to 2015. To meet the objectives of the research, the graphical illustration of trend with descriptive statistics are followed by econometric analysis. Based on the stationarity of the variables, an autoregressive distributive lag (ARDL) model has been estimated to test the long-run relationship between unemployment and growth. Findings: The results indicate that per capita GDP negatively inuences the unemployment rate in the long run only in three member countries of SAARC, namely- Afghanistan, India and Sri Lanka and in the South Asian Region in aggregate. This paper also nds a negative relationship between the growth rate and unemployment rate in Bangladesh, but this association is not statistically signicant. The study doesn’t nd any negative relationship between the two variables in Bhutan, Maldives, Nepal and Pakistan. The study discovers the validity of Okun’s law, but the attained Okun coecient is less than that of the actual Okun coecient, documented by Arthur Okun. Another substantial evidence is that the signicance of the connection between the growth rate and unemployment rate varies among the SAARC countries. Research limitations: The main limitation of this paper is the unavailability of data for Afghanistan compared to other SAARC countries. Originality/value: This paper is unique as it tests the validity of Okun’s law in every member country of SAARC and as a region of South Asia. To date, no such study like this has been found in the body of literature which nds long-run relationship in all SAARC countries.


Introduction
GDP is a commonly used parameter to measure the economic growth of a country. Some other factors like in ation, unemployment and literacy rate affect GDP growth rate (Anyanwu, 2014). The South Asian Association for Regional Cooperation (SAARC) is the regional geopolitical organization and intergovernmental union of the countries in South Asia. Its member countries include Afghanistan, Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan and Sri Lanka. SAARC countries cover 3% of the total world's area, while the population of the SAARC countries is 21% of the world's total population. SAARC was established on December 8, 1985 in Dhaka, the capital of Bangladesh. Political and economic problems are similar in SAARC countries (Chauhan, 2008). The population densities of these countries are high, which leads to serious threats of creating job opportunities. As a result, it creates an Page 3/22 unemployment problem. Although the growth rates of the countries are positive on average in recent decades, the effect of such growth on the level of employment requires scienti c investigation.
Unemployment is a situation in which peoples who are actively seeking jobs, but there are no vacancies.
In labour economics, unemployment includes frictional unemployment, cyclical unemployment, structural unemployment and classical unemployment (Junankar & Price, 2016). Because of the globalization and advancement of technology (Mincer & Danninger, 2000), the SAARC countries can't cope with those changing circumstances on account of social and economic problems. Such a condition of the economy obstructs the economic growth of the SAARC countries, causing a high unemployment rate (Upreti, 2015).
Economic policymakers consider unemployment as a vital factor in both developing and developed countries. As the full employment level maximizes output, it is one of the signi cant macroeconomic goals for every government (Arnold, 2011). Nevertheless, South Asian countries are facing structural changes for the last few decades and the consequences of these structural changes are reducing the employment ratio of the agriculture sector in GDP (Sawhney, 2010). The inter-sectoral output gap ensures a couple of things: rapid economic growth and unemployment changes (McMillan & Rodrik, 2011).
In macroeconomics, Okun's law is a well-known concept that describes the negative relationship between economic growth and unemployment (Martin, 1993). In 1962, Arthur Okun, an American economist, states that if the real GDP increases rapidly, the rate of unemployment will decline. Conversely, unemployment will rise if the growth will lower or negative (Higgins, 2011). A reduction in the unemployment rate in any economy is supposed as a positive development (Arnold, 2011). It is considered as a favourable sign since most of the people in a society require employment for their livelihood. Therefore, the Okun's law is the best technique to investigate and explain the negative association between economic growth and unemployment.
Most of the previous studies (Ciprian-Ionel, 2007; Villaverde & Maza, 2009;Watts & Mitchell, 1991) investigated the relationship between unemployment rate and growth for a single economy or state within the country. Some papers (Ahmed, Aziz, & Zaman, 2013; Moosa, 1997;Sawhney, 2010) tried to nd the existence of Okun's law in some selected SAARC countries (not more than ve countries). The majority of the papers (Christopoulos, 2004 The objective of the paper is to examine the effectiveness of Okun's law in SAARC countries in the longrun. This paper also tries to nd out whether any inverse relationship between growth rate and unemployment rate exist or not in South Asia. The study uses annual time series data for eight SAARC countries from 1991 to 2015. To meet the objectives of the research, the graphical illustration of trend with descriptive statistics are followed by econometric analysis. Based on the stationarity of the variables, an autoregressive distributive lag (ARDL) model has been estimated to test the long-run relationship between unemployment and growth.  28, 192828, -March 23, 1980) who rst described it in the early 1960s. Okun originally described that in the long-run, the economy experiences a 1 percentage increase in the unemployment rate for every 3 percentage decrease in real GDP. Similarly, a 3% increase in GDP from its long-run level decreases a 1% percentage in unemployment (Knotek II, 2007).
According to his empirical result, a slowdown of economic growth causes the unemployment rate to rise. This connection is more numerical rather than organizational, economic framework. But during the stormy 1970s, when stag ation (condition of both stagnation and in ation) troubled the country, the rule no longer held.
Furthermore, no economic theory can explain this association between unemployment rate and economic growth. Therefore, the Okun's theory becomes a benchmark to measure the connection between these two variables. It is a widely used technique to nd any negative association between growth rate and unemployment rate for its simplicity.
Many versions of the Okun's law are available: gap version, difference version, dynamic version and production function approach. Based on assumptions formulated by the researchers, every method has its own characteristics for calculation. Among the different approaches, only the difference version carries decently simple statistical calculations. By this method, without making any assumptions, values can be straight calculated from the accessible empirical data. Different techniques are applied in the dynamic version, gap version and the production function version to regress GDP growth over uctuations in the unemployment rate to the data with different interpretations and assumptions. However, the difference version merely regresses GDP growth rate on the unemployment rate.
To acquire a positive growth, more production of goods and services is a pre-requisite. With the help of inputs (capital, labor), the output can be produced if more workers are employed at the given capital stock by utilizing economic resources at their optimum level, then we can obtain real GDP growth. From this theoretical background, it is clear that there is a positive relationship between real GDP growth and reduction in unemployment level, as high employment level leads to a high price level in the economy according to Phillips (1962).
That is, having a higher employment level, the economy must face in ation. Thus we observe a higher price level is associated with a low unemployment rate. If we consider the unemployment rate and real GDP growth, the two economic variables from the above discussions, at that time we can conclude the effect of the unemployment rate over GDP growth. As employment and GDP growth are positively related to each other, unemployment rate and GDP growth must be negatively associated with each other i.e., a positive economic growth weakens the unemployment rate in the economy. It is needed to study to nd the validity of Okun's law in all SAARC countries and also in the South Asia region. There are few studies (Chauhan, 2008 In the case of the US economy, Levin, Lin, and Chu (2002) analyzed the unemployment and economic growth nexus and found a signi cance inverse relationship that indicates the validity of Okun's law. Walterskirchen (1999) examined and found a negative relationship between unemployment and economic growth by time series annual data from 1968 to 1998 for EU countries. Tatoğlu (2011) nds the validity of Okun's law for European countries and the result shows that the nexus between growth rate and unemployment rate varies among the countries. Marinkov and Geldenhuys (2007) studied in the case of South Africa using annual data from 1970-2005. Also, they found that unemployment and GDP growth move in the opposite direction. Still, changes in the unemployment rate causes less in ation in the GDP growth rate as compared to Okun's coe cient. Among the EU countries, depending on Swedish data, Aranki, Friberg, and Sjödin (2010b) investigated the relationship between unemployment and economic growth. In the case of Australia, Watts and Mitchell (1991) also supported Okun's law. Their study found the long-run relationship between unemployment and economic growth is unstable. Villaverde  Okun's law and found an inverse and two-way relationship between GDP growth variable and unemployment rate for Romania during 1993-2004. Knotek II (2007) also studied the usefulness of Okun's law in which he calculated the effects of current output, past output level, past unemployment rate on unemployment rate by using its difference, gap and dynamic versions and analysed that increasing unemployment rate slowdown the economic growth in the short-run and long-run. Christopoulos (2004) also estimated Okun's law by applying unit root tests and co-integration approach on panel data at regional level in Greece and found that results are valid in six out of thirteen regions. Moosa (1997) tested Okun's law for G7 countries namely Canada, Italy, UK, US, France, Germany, Japan .The empirical studies showed that Okun's coe cient is higher in North America whereas lower in Japan.
Moreover, Al-Habees and Rumman (2012) directed a similar study for Arab countries, highlighting their economic strategies. The study found that less economic growth rate hinders new job creation, which increases the rate of unemployment. Empirical results identify the validity of Okun's law and indicate that programmes should be taken in such a way so that economic growth will rise, reducing the unemployment rate.
Similarly, I. Lal, S. Muhammad, M. A. Jalil, and A. Hussain (2010a) checked the validity of Okun's law in some Asian countries, speci cally Bangladesh, China, India, Pakistan and Sri Lanka. For this persistence, they took annual time series data during the period 1980 to 2006. Using Engle-Granger (1987) cointegration technique, they examined the long-run relationship between GDP growth and unemployment. Their results found that there is no existence of Okun's law in some Asian countries. By using time series data for Pakistan from 1972 to 2006, Tahir Hussain, MW Siddiqi, and Asım Iqbal (2010) investigated the fundamental relationship between economic growth and unemployment rate. Considering some explanatory variables like capital, labor and human capital; their results also indicate a short-run and signi cant long-run relation between growth and unemployment. Now the question may arise how this study is unique from others. This study tries to nd the validity of Okun's law in all SAARC countries in the long-run and the impact of growth on unemployment in the South Asia Region. ARDL model is used to test the validity of Okun's law. This study has tried to ful ll the above mentioned research gap. Moreover, this study also includes an update data from 1991 to 2017.

Data and Variables
All data for this study are collected from various sources. The primary source is the World Development Indicators (World Bank, 2018) and Bangladesh Bank. The data consists of the yearly time series over the period from 1991 to 2015. The time frame of data depended mainly on the available data. The dependent variable is GDP per capita growth and the independent variable is the unemployment rate as per ILO estimate. GDP per capita growth variable is measured in constant 2010 US dollars and, as well as descriptive data statistics; the unemployment rate is measured in percentage form. More data descriptions, as well as descriptive data statistics, are presented below.
A detailed description of the raw data set is provided in the form of descriptive statistics in Table 01 and  Table 02. The time-series data sample include 1991 to 2017 for all the countries except for Afghanistan due to the unavailability of data.  The above table is consists of observation, mean value, standard deviation, maximum value, minimum value and corresponding probability value of Jarque-Bera along with the value Jarque-Bera of all variables. All variables are normally distributed without the growth rate of Bhutan at the 5% level of signi cance.

Modeling & Methodology
When a study is being conducted with time-series data, it needs to check whether data stationary or not.
Because there is a possibility to be the spurious regression using time-series data is used. A time series Y is said to be stationary if its mean and variance are constant over time and the value of covariance between two time periods depends only on the distance between the two time periods and not on the actual time at which the variance is computed (Gujarati, 2009 When a study is being conducted with time-series data, it needs to check whether data stationary or not. Because there is a great possibility to be a spurious result of regression. A time series Y is said to be stationary if its mean and variance are constant over time and the value of covariance between two time periods depends only on the distance between the two time periods and not on the actual time at which the variance is computed (Gujarati, 2009). There are many ways to check data stationery. The standard approach to check the stationary of a time series is unit root test. Among the various tests, the Augmented Dickey-Fuller (ADF) test is used widely (Adenutsi, 2010), (Islam & Nasrin, 2015). Augmented Dickey-Fuller (ADF) unit root test has been used to check whether the variables are stationary or not.   Growth -4.065264*** -4.520553*** -5.236510*** -5.182008*** I(0) Note: (*), (**) and (***) denotes 1%, 5% and 10% respectively.
Here, Null hypothesis: H 0 = the variable has a unit root and Alternative hypothesis: H A = the variable has no unit root The decision rule of this test is that null hypothesis will be rejected means an alternative hypothesis will be accepted if the p-value of the ADF test is less than 5%. Otherwise, the null hypothesis cannot be rejected. If the t-statistic value obtained from the ADF unit root test is greater than 5% critical value, the null hypothesis will be rejected. Otherwise, it cannot be rejected at the same critical value. Rejection of the null hypothesis would mean that the variable has no unit root. So the data are stationary. If data are found stationary at level, the result of a regressed by these data will not be spurious (Gujarati, 2009). The results of the ADF unit root test for the variables of SAARC countries and South Asia in their levels and rst differences are presented in Table-03 and Table-

The Results of ARDL Estimation
The results of the ARDL model are the long-run output because all variables used by the ARDL model are stationary at I(0) and I(1). Note: (*), (**) and (***) denotes 1%, 5% and 10% respectively.
The Table-05 and Table-06 show the long-run relationship between GDP growth and unemployment rate.
The results indicate that per capita GDP negatively in uences the unemployment rate in the long run in some SAARC countries, namely India, Sri Lanka and Afghanistan and the coe cient of growth rate is also found statistically signi cant at 1%, 5%, 10% signi cant level respectively. Also, the negative relation between the variables in South Asia has been found by this paper. Note: (*), (**) and (***) denotes 1%, 5% and 10% respectively.
One percent point increase in per capita GDP will lead to a 0.109 percent point decrease in the unemployment rate in Afghanistan, 0.044 percent point in India, 0.122 percent point in Sri Lanka and 0.07 percent point in South Asia in the long run. This paper also nds a negative relation between the growth rate and unemployment rate in Bangladesh, but this association is not statistically signi cant. The study doesn't nd any negative relationship between the two variables in Bhutan, Maldives, Nepal and Pakistan.
This implies that GDP growth rate does not have any positive in uence in creating employment in those countries in the long-run. Several studies (Ahmed et al., 2013) also support these ndings. This model is the best t as the adjusted R 2 value is high as well as the explanatory variables can be explained dependent variable by near or above 90% for most of the countries. F-statistic value is also high and the corresponding probability value is much low. It is found that F-statistic value is statistically signi cant. Durbin-Watson value con rms that the model is not spurious. Durbin-Watson value indicates the serial correlation of residuals. The results of this regression will be spurious if Durbin-Watson value is not close to 2. The results of various test which are related with this model will be disclosed in the following section.

Diagnostic tests for ARDL Model
Serial correlation or autocorrelation refers to the correlation of a time series with its previous and future values. Autocorrelation is also sometimes called "lagged correlation," which refers to the correlation between members of a series of numbers arranged in time. Here, this study uses the Breusch-Godfrey Serial Correlation Lagrange Multiplier (LM) test for serial correlation by employing with econometric software EViews 9.
Null hypothesis, H 0 : There is on serial correlation.
Alternative hypothesis, H A : There is serial correlation.
Results for the serial correlation test have been presented by using EViews 9 in Table 07 and Table-08 below. The decision rule of the autocorrelation test is that the null hypothesis will be rejected if the p-value is less than 5% (5% level of signi cance). Otherwise, the null hypothesis cannot be rejected.
Here, the null hypothesis cannot be rejected because of the p-value being greater than 5%. This implies that there is no serial correlation in the ARDL Model. It indicates that the results obtained from ARDL with stationary data is not spurious or non-sense.
Heteroskedasticity test is inserted in following Table-09 and Table-10 Heteroskedasticity can be de ned as unequal variances of the residuals. In the presence of Heteroskedasticity, the coe cient of variables are found statistically insigni cant (Gujarati, 2009 Result for Heteroskedasticity has been presented by using EViews 9 in Table-09 and Table-10 below. The decision rule here is that if the corresponding p-value of F-statistic is greater than 5%, the null hypothesis cannot be rejected. Otherwise, the null hypothesis will be rejected.
Here, the corresponding p-value of F-statistic is greater than 5%. So the null hypothesis cannot be rejected and it can be concluded that the ARDL model is free from the Heteroskedasticity problem at the 5% level of signi cance. The normality test for the ARDL model is shown in the next Table-11 and Table-12.
Normality tests are checked to determine whether a data set follows a normal distribution or not. If the residual distribution is not normal, then obtained t-ratios are looking to be so good and the decision may not be valid. The null hypothesis of this test is that the residual series is normally distributed and the alternative hypothesis is that the residual series is not normally distributed. The results of the normality test for the residuals of the ARDL model are presented below.
Here, the appropriate test statistic of normality test is Jarque-Bera (JB) test statistic  Results for Ramsey RESET test is presented by using EViews 9 in Table 13 and Table 14 below.  Making a decision procedure is that if the corresponding p-value of F-statistic is less than 5%, the null hypothesis will be rejected. Otherwise, the null hypothesis cannot be rejected. From the above result, the null hypothesis cannot be rejected indicates that the ARDL model is correctly speci ed.

Findings And Discussion
The results indicate that per capita GDP negatively in uences the unemployment rate in the long run in some SAARC countries, namely India, Sri Lanka and Afghanistan and the coe cient of growth rate is also found statistically signi cant at 1%, 5%, 10% signi cant level respectively. Also, the negative relation between the variables in South Asia has been found in this paper. One percent point increase in per capita GDP will lead to a 0.109 percent point decrease in the unemployment rate in Afghanistan, 0.044 percent point in India, 0.122 percent point in Sri Lanka and 0.07 percent point in South Asia in the long run. This paper also nds a negative relation between the growth rate and unemployment rate in Bangladesh, but this association is not statistically signi cant. The study doesn't nd any negative relationship between the two variables in Bhutan, Maldives, Nepal and Pakistan. This implies that the GDP growth rate does not have any positive in uence in creating employment in those countries in the long-run. These ndings are also supported by previous studies (Ahmed et al., 2013). This model is the best t as the adjusted R 2 value is high as well as the explanatory variables can be explained dependent variable by near or above 90% for most of the countries. F-statistic value is also high and the corresponding probability value is much low. It is found that the F-statistic value is statistically signi cant. Durbin-Watson value con rms that the model is not spurious. Durbin-Watson value indicates the serial correlation of residuals. The results of this regression will be spurious if the Durbin-Watson value is not close to 2. The results of various test which are related with this model will be disclosed in the following section.
Findings-01: Only in three member countries of SAARC, namely Afghanistan, India and Sri Lanka among eight and also in the South Asia Region; this study nds the validity of Okun's law. But the attained Okun coe cient is less than that of the actual Okun coe cient, which was documented by Arthur Okun (Okun, 1962).
Findings-02: This study suggests that there is a long-run negative relationship between growth and unemployment rate in Afghanistan, India and Sri Lanka. No long-run association between the variables in Bhutan, Maldives, Nepal and Pakistan. For Bangladesh, this paper observed expected negative sign, but the result is insigni cant.

Conclusion
In this study, we examined the validity of Okun's law the long run term relationships between unemployment rate and economic growth in SAARC countries and the South Asia Region from 1991 to 2018 with the exception for Afghanistan. Estimation results support the unemployment jeopardize in most of SAARC countries.
The need for improvement in governance improvement is essential for all SAARC countries that will strengthen institutions that will help in promoting economic growth and stability. The government should help in having sound political, nancial and social reforms to achieve structural changes smoothly because SAARC countries have the potential to grow and that will help in increasing the productivity of the economy and economic welfare. (Sawhney, 2010). SAARC countries should develop their human resources and physical infrastructure development by spreading regional cooperation to each other by reducing restrictions on intra-regional investment and technology ows.
Working with time series data over the time period from 1976 to 2015, this study nds that economic growth have a negative relation to unemployment rate in Afghanistan, India and Sri Lanka in long run.
These results is supported by a few studies (Lal et al. (2010a). The policymakers of Bangladesh should take an export-attracting policy to be improvement of economic growth. Most of the remittances are used to buy land or house construction (Ahmed, 2010). Some Asian developing countries are prominent examples in the world who successfully removed the unemployment problem e.g., Korea, Malaysia, Singapore and China are the most recent of them. They are growing fastly because there is no political instability and corruption. Bangladesh Government should take pragmatic steps to reduce the unemployment problem as our growth increasing year by year. Now, the question is that, "Is growth rate negatively in uence the unemployment rate?" If so, then why the number of unemployed young are increasing day by day? (BBS, 2018). This study concludes with the recommendation that Okun's law are not valid in most of the developing countries because of asymmetric problems. This study again recommends that it is needed further research on the relationship between economic growth and the unemployment rate in short-run by different models with updated quarterly data. The implications of Okun's law for economic policy is that economists need to anticipate the further development of the unemployment rate for a given projected growth level, which is additionally essential to forecast unemployment costs. But our results do not support the implications of Okun's Law in some developing countries. It can be said that Okun's law interpretation may not be applicable in developing countries. Pakistani, Bangladeshi, Sri Lankan, and Indian governments and political leaders should adopt this role model and follow from those Asian countries.

Declarations
Availability of data and material: