Does Remittance Fuel Labour Productivity in Bangladesh the Application of an Asymmetric Non Linear ARDL Approach

Applied Economics

ISSN: 0003-6846 (Print) 1466-4283 (Online) Journal homepage: http://www.tandfonlin http://www.tandfonline.com/loi/raec2 e.com/loi/raec20 0

Does remittance fuel labour productivity in Bangladesh? The application of an asymmetric non-linear ARDL approach Md. Al Mamun, Kazi Sohag, Nahla Samargandi & Farida Yasmeen To cite this article: Md. article:  Md. Al Mamun, Kazi Sohag, Nahla Samargandi & Farida Yasmeen (2016) Does remittance fuel labour productivity in Bangladesh? The application of an asymmetric non-linear ARDL approach, Applied Economics, 48:50, 4861-4877, DOI: 10.1080/00036846.2016.1167825 To link to this article: http://dx.doi.org/10.1080/00036846.2016.1167825

Published online: 19 May 2016.

Submit your article to this journal

Article views: 118

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com http://www.tandfonline.com/action/journalIn /action/journalInformation?journa formation?journalCode=raec20 lCode=raec20 Download by: [Charles by: [Charles University in Prague]

Date: 22 Date: 22 January 2017, At: 05:33

APPLIED ECONOMICS, 2016 VOL. 48, NO. 50, 4861 –4877 http://dx.doi.org/10.1080/00036846.2016.1167825

Does remittance fuel labour productivity in Bangladesh? The application of an asymmetric non-linear ARDL approach Md. Al Mamuna,d, Kazi Sohag

b

, Nahla Samargandi c and Farida Yasmeend

Department of Economics and Finance, La Trobe University, Bundoora, Australia; bFaculty of Social Science & Humanities, National University of Malaysia, Bangi, Malaysia; cFaculty of Economics and Administration, King Abdulaziz University, Jeddah, Saudi Arabia; dFaculty of Business and Economics, East West University, Dhaka, Bangladesh a

ABSTRACT

KEYWORDS

Bangladesh, a labour-abundant country, is finally learning to reap gains from her labour abundance. The historical under-optimized capital –labour ratio of Bangladesh is slowly moving towards its optimum as the migration of its labour force and consequent inflow of remittance strategically converts its abundant labour into capital. Though the fear that remittance as an alternative source of earning may result in Dutch disease for existing family members; annual data for 1982–2013 and through applying standard testing methodologies, our empirical findings suggest that such a fear is not valid since remittance flow significantly improves domestic labour productivity in Bangladesh in the long run.

I. Introduction

Remittance inflow into the developing world represents the changing pattern of global labour market dynamics as the globalization of the labour market becomes increasingly more visible. Since Levitt (1983) introduced the term ‘globalization of markets ’ it has not only been Nokia or Samsung that have found new market opportunities; it also has been the labour force, irrespective of their skill levels, that have found new hope and opportunities for a better life (Al Mamun et al.   2015). The size of global remittance is $436 billion with a growth rate of  5.7% and developing countries receive about 75% of all remittances and supply 80% of the global migrant workers in 2013 (World Bank   2013). Remittance receipts are the second largest source of  foreign exchange earnings and exceed private capital flows and foreign aid to most developing countries, including Bangladesh. Prior literature sheds light on the reasons and consequences of remittance for both host and home country economies. Theoretically, for host countries, mostly low and lower middle-income countries, an inflow of remittance can optimize the current suboptimum, labour–capital ratio. Optimization of the CONTACT  Md. Al Mamun   [email protected]; [email protected]

3086, Australia © 2016 Informa UK Limited, trading as Taylor & Francis Group

Bangladesh; labour productivity; remittance; non-liner ARDL  JEL CLASSIFICATION

E22; E24; E240; J880

labour–capital ratio can improve domestic productivity, including labour, at any given level of technological development. In fact, optimization of labour and capital is the building block of local productivity in Solow ’s (1957) classical framework. For countries such as Bangladesh, any mechanism to optimize the current sub-optimum labour–capital ratio should result in long-run economic growth. Remittance, therefore, can be a potential growth driver. Prior literature also suggests a different possibility. Mamun and Nath (2010) suggest that in countries such as Bangladesh, remittance can effectively lead to more leisure for existing family members of  migrant workers. The argument is, if remittance is mostly used to fulfill immediate family needs, less effort will be exerted for extra income by the family  members and relatives of the migrant workers in their home countries. Therefore, one can expect a lower domestic labour participation, high unemployment, and possibly an overall lower labour productivity, especially in countries with labour-intensive production. In addition, since remittance inflows strengthen the home currency, it negatively impacts industrial output and exports (Acosta, Lartey, and Mandelman 2009; Amuedo-Dorantes and Pozo   2004). Therefore,

Department of Economics and Finance, La Trobe University, Bundoora, Vic

4862

M d. A l MA MU N ET A L.

an appreciation of local currency will lead to low  domestic labour productivity. As Bangladesh is on the edge of economic prosperity or fragility with its increased investment in human capital and a growing share of the young population on one hand, and ongoing political uncertainties and increased social division on the other, we argue that the economic prospects of  Bangladesh will largely depend on mobilizing its huge labour force in a productive manner. Given this argument, we investigate the possible effect of  remittance receipt on domestic labour productivity  in Bangladesh. Using a sample period of 1982 –2013 and employing advanced time series estimators, we show that the influx of remittance significantly spurs domestic labour productivity in Bangladesh. In fact, compared to domestic fixed capital formation, remittance exerts greater impact in augmenting domestic labour productivity. However, the interaction between remittance and fixed capital formation yields a far greater impact on domestic labour productivity than remittance or fixed capital formulation alone. This finding is new in macroeconomic literature. We also test the possibility of Dutch disease due to the huge influx of remittance in Bangladesh. We find no evidence of such phenomenon. Finally, we show that successive governments’ efforts to depreciate local currency against the US dollar has countered the possible negative effect of  remittance inflow by augmenting the export sector and thus contributing to augmenting higher labour productivity. In   Section II, we present the conceptual framework, a brief literature review, and empirical predictions. In Section III we present a brief discussion on remittance flow in Bangladesh.  Section IV  presents our data and the methodology. We present our empirical results in  Section V, and Section VI  concludes the article and presents relevant policy implications of our article.

II. Conceptual framework, prior evidence and empirical predictions Conceptual framework 

We present a simple Cobb–Douglas production function to draw the link between remittance and domestic labour productivity. The function is

Φ  ¼  f ðK; ALÞ,

where Φ  is the production level,  K  is the capital stock (investment and remittance are taken as the proxy of capital) and AL   is effective labour. Φ

 AL

 K   AL  K  

 ¼  F 

;  AL  AL

¼  F 

 AL

;1

(1)

Here  Φ  /  AL  and  K / AL are the output per unit of  effective labour and capital, respectively. We further symbolize ϕ ¼  Y = AL; k  ¼  K = AL; and, thus it would be rewritten as Φ ¼  f ðkÞ. Moreover, production per unit of labour Φ= AL   relies on the ratio of capital and labour (K = ALÞ. The following function shows that how remittance augments the capital stock is K_ ðt Þ ¼  s Φðt Þ  δ Kt  1 þ Rt 

(2)

where Rt = Remittance inflow at time t . In addition, the labour supply function in the context of migration is defined as L_ ðt Þ ¼  L t  þ nLðt Þ  Lmt 

(3)

where L_ ðt Þ is the labour supply at time  t , and n  is the growth rate of labour. Lmt    is the migrant labour supply at time t , which reduces the local labour supply. We further assume a frictionless international financial market and zero transaction cost.  ðLmt Þ  ERt Þ Ret  ¼  f ðγ  w

(4)

where Ret  is the remittance influx at time t , γ is the fraction of earnings of migrant labour transferred to   is the mean wage rate, Lmt  is the local economy, w the number of migrant labour in a foreign land at time t   and, ERt  is the real exchange rate at time t . Furthermore, following the framework of Barai (2012), the remittances consumption function can be derived as Rct  ¼  C  þ RBt  þ Rht 

(5)

where  R ct  is the remittance consumption function,  C  is used for consumption expenditure,  R Bt  is the portion of remittance used as investment and, finally, Rht  is the remittance used for human capital development. Thus, the capital –labour ratio is as follows: k  ¼

Φsðt Þ

 δ Kt  1 þ RBt  Lt  þ nLðt Þ  Lmt 

(6)

Hence, an inflow of remittance increases marginal productivity of labour. In addition, it helps to foster

APPLIED ECONOMICS

human capital through utilizing the remittance for schooling and gaining skills. Moreover, the migration of local labour to foreign lands contributes to reducing the labour–capital ratio in the local economy. Empirics and predictions

Given the conceptual framework, we further investigate the previous literature to develop our empirical predictions. Prior literature on remittance primarily  focuses on reasons and consequences of remittance. Regarding the motivation, an early research (Lucas and Stark   1985) explains remittance in terms such as ‘tempered altruism’ and ‘enlightened self-interest ’; while, another (Poirine   1997) explains remittance within the family loan agreement framework. Regarding the economic impact of remittance, the literature sheds light on both the host and home country ’s perspectives. Prior literature links remittance with financial development (Nyamongo and Misati   2011; Aggarwal, Demirguc Kunt, and Martinez Peria   2010), with sustainable economic development like the welfare effect (Gupta, Pattillo, and Wagh   2009; Siddiqui and Kemal   2006; Adams and Page 2005; Adams 1993), with economic growth (Taylor and Wyatt   1996; Nyamongoa et al.   2012), with domestic consumption (Quartey and Blankson 2004), with human capital formation (Edwards and Ureta 2003; Hanson and Woodruff   2003) including education and schooling (Calero, Bedi, and Sparrow  2008; Adams and Cuecuecha 2010), with Dutch disease (Acosta, Lartey, and Mandelman 2009; Bourdet and Falck 2006; Lartey, Mandelman, and Acosta 2008; Vargas-Silva   2009), with nominal exchange beha viour (Amuedo-Dorantes and Pozo   2004; Chami, Fullenkamp, and Jahjah 2003), etc. Most of the literature seemingly presents mixed evidence on the economic impact of remittance, especially for recipient countries. A thorough examination of the existing literature on the consequences of remittance presents a research gap about the possible effect of remittances on domestic labour productivity for recipient countries. There are some plausible reasons for this research gap.  First , studies on remittance and financial development by Nyamongo and Misati ( 2011) and Aggarwal, Demirguc Kunt, and Martinez Peria (2010) show that remittances channelled through the

4863

formal financial sector improve growth. We argue that with growing restrictions on  Hundi, a primary  model of international fund transfer, a smooth channelling of remittance into home countries requires standardizing the local banking sectors up to international standard. Therefore, increased efficiency  and an improved regulatory framework in the domestic financial sector will help in achieving improved allocative efficiency. This improved efficiency will augment greater economic productivity  at any given level of capital and labour. Therefore, we expect a second order effect of remittance on domestic labour productivity. Second , Edwards and Ureta (2003) and Hanson and Woodruff (2003) show that remittance leads to increased human capital formation in Latin America. Using data from Ecuador, Calero, Bedi, and Sparrow (2008) show that remittance improves literacy rates especially among girls by financing schooling in rural areas and also prevents drop-out. In a similar vein, Adams and Cuecuecha (2010) show that households receiving remittances increase expenditure on education and housing. We argue that an increased investment of remittance in human capital development in the home country  should increase the potential supply of skilled labour forces in the local market. This increase, in turn, should also lead to higher labour productivity. Third , Bayangos and Jansen (2011) using the Philippines’   experience, show that remittance has a significant positive effect on the domestic labour market by increasing the competitiveness of the domestic economy. In a theoretical model, McCormick and Wahba (2001), present a complete decision process at the micro level and show that the decision to migrate is a utility maximizing decision for the household, which will reduce employment, output and demand in the home country with an offsetting benefit of increased inflow of remittances. Therefore, at the macro level, an optimal decision is struck when the lost production due to loss of labour supply and appreciation of the nominal exchange rate causing so-called Dutch disease (Acosta, Lartey, and Mandelman 2009; Vargas-Silva 2009) in the home country is compensated by increased remittance flow. Furthermore, there is a potential that remittance may bring negative consequences in the economy by an appreciation of the local currency. An appreciation of the local currency reduces

4864

M d. A l MA MU N ET A L.

the international competitiveness of domestic products (Amuedo-Dorantes and Pozo   2004; Chami, Fullenkamp, and Jahjah   2003), increases consumption of foreign goods by local consumers and also means less exports. As an immediate consequence, domestic industrial production will be replaced by  the foreign products. So, remittance can reduce local factor productivity, including labour. Furthermore, Mamun and Nath (2010) argue that the use of  remittance, increases consumption, repayment of  loans, and the building of houses. Thus, among many possible channels through which remittance may produce Dutch disease, the consideration of  remittance as the substitute for own income and increasing leisure activities within the remittanceleisure framework by the migrant’s family members, is the most prominent one. Increased leisure reduces the labour participation rate, labour productivity, and domestic economic growth. Finally , one of the most important theoretical arguments for a possible link between remittance and domestic labour productivity emanates from the idea that remittances can increase domestic employment by the reinvestment of remittanceinduced national savings; capital accumulation. Barai (2012) states that ‘there is direct, trickle down, and indirect benefits of remittances for many of the developing countries . Unfortunately, though, despite having a strong conceptual appeal, there is a real dearth of empirical studies focusing on the effect of  remittance on domestic labour productivity. We test these possibilities in the case of Bangladesh. ’ 

III. Why Bangladesh?

Bangladesh, historically, is an agrarian economy, sunk into an era of economic backwardness due to the nationalization of the industrial and financial sectors during the authoritarian one-party era after independence. It has recently emerged as an international source of cheap labour. Later governments, starting with General Ziaur Rahman particularly, have taken initiatives to open up the economy. This opening up was an unprecedented move because U.S. interests in the region at that time were almost absent, and the region primarily followed closed economic cooperation with India and China. Furthermore, the government from 1978 to 1981 recognized that the economic potential of 

Bangladesh relies on the successful transformation of its abundant labour into productive resources. Migration to and remittance from the Middle East emerged as a new policy concern in the government’s economic agenda. Over the past four decades, the population of  Bangladesh has become one of its core competitive advantages in the international marketplace. The abundant labour force has started to pay off, mainly  through positioning itself as a source of cheap labour supply in the garments manufacturing sectors and as an international source for cheap labour across the world. According to the Bureau of Manpower, Employment and Training (BMET   2013), from 1977 Bangladesh has sent over an estimated eight million migrant workers to more than 140 countries around the world. Today, Bangladesh is the seventh largest remittance recipient country in the world (World Bank   2013). Moreover, during the past two decades, remittance flow in Bangladesh has increased significantly. Economic liberalization during the time of  Khaleda Zia, in the early part of the 1990s, helped Bangladesh to achieve a 15% growth in remittance. Furthermore, the growth rates have also doubled from 9.26% (1990–2001) to 19.39% (2002–2012) during the past two decades. Both the Bangladesh Bank and World Bank statistics suggest that from 1993 when remittance inflow in Bangladesh reached US$ 1.0 billion for the first time, the figure has been continuously growing. From 1990 onward, it took  almost 10 years for Bangladeshi remittance flow to double while it took only 4 years to double from the 2001 figure of US$2.5 billion to $5.0 billion. By 2013, the size of the remittance stood at $14 billion, which is 12% of its foreign currency earnings (World Bank  2013). Remittance accounts for 95% of current transfers, equivalent to 62% of exports and could finance 48% of the value of imports in 2009 (Bangladesh Bank  2012). Raihan et al. (2009) suggests that outflow of Bangladeshi workers in the 2000s doubled from the previous decade to 410,000 workers per year. Among migrant workers, more than half are lowskilled, while only 2.26% are of the professional class. One interesting observation from the data source of  BMET (2013) is that there was a drastic fall in the professional level as well as the total migrants from 2008 until 2012. The fall in remittance and the

APPLIED ECONOMICS

migrant workers can be explained by the global economic meltdown as well as the foreign policy  failure at the political level by the government, especially in the case of Malaysia, KSA and other Middle Eastern countries. In fact, from the peak 2008 figure of 0.83 million migrant workers from Bangladesh, the number of migrants had drastically fallen to 0.43 million in 2012 (BMET 2013). Therefore, despite an increase in the remittance flow in Bangladesh over the period until today, there has been a significant fall in the number of migrants from Bangladesh. Interestingly for Bangladesh, a country with a traditional family structure, there has been a remarkable increase in the number of female migrant workers from Bangladesh over the past decade. In 2004 alone, there was a 378% increase in female migrant workers from Bangladesh and by today there are around 20,000 female migrant workers leaving Bangladesh each year. This increase is around 4.6% of the total migrant workers leaving Bangladesh (BMET 2013). A closer look at the year-wise growth rate reveals that the growth rate of remittance has been a slightly  inverted U-shape over the past two decades. However, such growth rates are very unstable over the period [see   Figure 1(a)]. In contrast, to the growth rate of remittance, the growth rate of the number of migrant workers from Bangladesh shows a U -shape pattern [see   Figure 1(b)]. These contrasting growth rates are interesting (given our theoretical argument in   Section II) and present a remarkable case for investigating the impact of  remittance on domestic labour productivity in Bangladesh.

Prior research on the effects of remittance on Bangladesh is limited both in numbers as well as diversity. These studies primarily focus on trends of remittance (Farid et al.  2009; Barai 2012), remittance and financial development (Chowdhury  2011), and remittance, financial development and economic growth (Noman and Uddin   2011; Ahmed 2010; Siddique, Selvanathan, and Selvanathan  2010; Barai   2012). Chowdhury (2011) in her most comprehensive and methodologically sound study so far, shows that remittance and financial development are causally linked. The study posits a significant positive impact of remittance on financial development in Bangladesh. However, the development of financial sectors has a neutral effect on the inflow of  remittances. Noman and Uddin (2011) document that both remittance and banking sector development influence per capita income for four major South Asian countries including Bangladesh. However, neither domestic products nor advancement in the banking sector has a significant impact on remittance flows. However, despite being a panel study in nature, they employed a time series analysis that reduces the quality of their policy prescription. Moreover, their findings are at odds with many  theoretically and empirically established studies. Unfortunately, Noman and Uddin (2011) present no economically possible justification for their result. On the growth front, A. Siddique, Selvanathan, and Selvanathan ( 2010) use time series for a 25 years period, showing that growth in remittances leads to economic growth in Bangladesh. Siddique, Selvanathan, and Selvanathan ( 2010) finding supports our empirical prediction.

        5   .         1

        3   .

        1

        2   .

        5   .

        1   .

        0

        0

        1   .    -

        5   .    -

95% CI

predicted migrant workers grow thrate

Migrant Workers Growth Rate

(a)

4865

(b)

Figure 1.  (a) Remittance growth rate in Bangladesh. (b) Migrant worker ’s growth rate in Bangladesh.

4866

M d. A l MA MU N ET A L.

As there is a scarcity of literature focusing on the impact of remittance on the domestic labour productivity in the international context, to the best of our knowledge, this is the first empirical study to examine the possible relation between remittance and labour productivity in the case of labour-abundant Bangladesh. We argue that, for a country like Bangladesh, any long-run growth strategy should focus on improving the productivity of domestic labour. Bangladesh is one of the global growth generator countries in the world (Buiter and Rahbari   2011). At this critical juncture, a proper combination of the capital–labour ratio is necessary to achieve optimum productivity in Solow ’s (1957) classical view  of growth. Remittance, a direct function of  migration, can help Bangladesh determine the optimum level of migration from Bangladesh that will reduce labour slack domestically. Remittance inflow can strategically convert domestic labour into capital, which will help labour-abundant and capital-scarce Bangladesh to optimize the capital – labour ratio. The optimization can increase the overall productivity along with the productivity of labour. However, the real worry is that, if remittance is considered as an alternative to the existing family members ’  drive for income, especially in a nuclear family structure (which is common in Bangladesh), then it may increase the leisure time of the domestic labour force and thus may become a counterproductive strategy for growth. Besides, if the inflow  of remittance increases the value of the local currency, then an export-driven growth strategy  may face a problem, which in turn can reduce domestic labour productivity. We test all these possibilities in   Section V. Therefore, overall, our article is a significant academic undertaking with immense policy implications.

IV. Data and methodology Variables and sources

In our study, we define domestic labour productivity as the real GDP contribution per economically  active population, i.e. real GDP chain per workers 1

http://pwt.econ.upenn.edu/Documentation/append61.pdf .

(RPW).1 Our definition is slightly different from the traditional definition followed by Weiss (1993) and Jeanneney and Hua (2011) but consistent with Al Mamun et al. (2015). The definition of an economically active population excludes traditional household workers. In our study, RPW is the dependent variable while remittance (REM) is the primary endogenous variable. Following Romer (2006), we include fixed capital formation (FC) while we include nominal exchange rate (RER) following Jeanneney and Hua (2011). We investigate the link between remittance and labour productivity in Bangladesh for the period 1982 –2013 using data from a variety of sources including the World Bank, BMET and International Labour Organization (ILO).

Standard ARDL bounds testing approach to cointegration

We employ a standard univariate Dickey Fuller-generalised least squares (DF-GLS) test, a modified version of the conventional ADF t -test, proposed by  Elliott, Rothenberg, and Stock (1996), to examine the time series properties of the variables of interest. The DF-GLS test performs better compared to the ADF test for a small sample, can capture the hidden trend and mean in the data (Elliott, Rothenberg, and Stock   1996; Fosu and Magnus  2006), and provide a more accurate predictive power. We use Pesaran, Shin, and Smith (2001) ARDL bounds testing approach for cointegration. This overcomes the limitations of the Ganger causality  of Engle and Granger (1987), and the cointegration test of Johansen (1988,   1991) on the order of the integration of variables and the inability to provide both short- and long-run dynamics. Furthermore, Pesaran, Shin, and Smith (2001) process allows the examination of the long-run cointegrating relations and dynamic interactions among the variables and provides critical leverage in the estimation process. This includes: (i) the cointegration relation can be estimated using the OLS method easily after choosing the lags order of the model; (ii) in contrast to Johansen and Jesulius’ (1990) procedure, this test procedure is appropriate irrespective of the order of integration, i.e. I (0) or I (1) or mutual

APPLIED ECONOMICS

cointegration and (iii) the test is competent in small and finite data size. The ARDL version of the vector error correction model (VECM) as proposed by Ang and McKibbin (2005), Khan, Qayyum, and Saeed (2005) and Fosu and Magnus (2006) is as follows:

X  γ ln RPW  X  δ  ln REM  þ X  ϕ ln FC  þ X  η ln RER þ X  ζ  lnðREM ÞðFC Þ þ

Δ ln RPW  ¼  β 0  þ

ln FC t 1 þ  λ4 ln RERt 1

þ λ5

lnðREM ÞðFC Þt 1

iΔ

t i

Δ  j¼1  j q Δ l ¼1 l  q Δ m¼1 m q Δ m¼1 m

t  j

t k

t m

t n

þ θ ECM t 1 þ εt 

(9) In additional, we run a cumulative sum of recursive residuals (CUSUM) and the cumulative sum of  squares of recursive residuals (CUSUMSQ) tests to check the stability of the estimated parameters in the spirit of Pesaran and Shin (1998).

 p

X X þ X þ X þ X þ þ

i¼1

q

Δ ln RPW  ¼  β 0  þ  λ1 ln RPW t 1  þ  λ2 ln REM t 1

þ λ3

 p

4867

 γ  ln RPW t i i¼1 i q

 δ  ln REM t  j  j¼1  j q  ϕ  ln FC t k l ¼1 l  q  η  ln RERt m m 1 m ¼

q

 ζ   lnðREM ÞðFC Þt n  þ εt  m¼1 m

Cointegration under structural break 

(7)

At first, we estimate Equation (1) under the OLS approach, and then we test the null hypothesis of no cointegration among the variables ( H 0:  β1 = β2 = β3 = β4   = 0), against the alternative hypothesis of cointegration among the variables (H a: β1 ≠ β2 ≠ β3 ≠ β4 ≠   0). The computed F - statistics are evaluated to obtain the critical  value (upper and lower bound) (Pesaran, Shin, and Smith   2001). Accordingly, the F -statistic being greater than the upper critical value indicates the existence of cointegration and  vice versa. The F - statistic within upper and lower bounds indicates an inconclusive cointegrating decision. After establishing the cointegrating relation among the variables, we estimate the long-run coefficient of the ARDL framework using Equation (2). We use the Schwarz Bayesian criterion to choose the appropriate lag length of the ARDL model for all the variables. Finally, we estimate Equation (3) for short-run or error correction coefficients.  p

X  γ  ln RPW  X  δ  ln REM   þ X  ϕ  ln FC  þ X  η  ln RER ; þ X  ζ   lnðREM ÞðFC Þ  þ ε þ

ln RPW  ¼  β 0 þ

q

i¼1 i

q

t m

 j¼1  j q m¼1 q

t m

m

m¼1 m

l ¼1

t m

l 

t m

t m

t 

(8)

Given that the remittance flow in Bangladesh has increased markedly in the early 1990s and early  2000s, we expect the existence of a break in the series. Hence, we apply a unit root test with a structural break suggested by Zivot and Andrews (2002). The test is appropriate if the series shows a potential structural break. The test takes three possible alternatives, as modelled below. Δ X t  ¼  a

þ ax t 1 þ bt  þ cDU t  k

X d   X  þ

 j Δ t  j þ

μt 

 

(10)

 

(11)

 j¼1

Δ X t  ¼  b

þ bx t 1 þ ct  þ bDT t  k

þ

X d   X 

 j Δ t  j  þ

μt 

 j¼1

Δ X t  ¼  c

þ cx t 1 þ ct  þ dDU t  þ dDT t  k

þ

X d   X 

 j Δ t  j  þ

μt 

 

(12)

 j¼1

where the dummy variables indicated by  DU t  show a mean shift at each point with a time break (TB) while DT t  is showing the TB for each variable. So, DU t  = 1 .. . if  t   > TB or 0 . . . if t   < TB. Moreover, DU t  =  t  –  TB . . . if  t  > TB or 0 . . . if  t  < TB. The null hypothesis of the unit root break date c = 0 indicates that the series is not stationary with a drift nor has information about the structural break point, while the  c  < 0 hypothesis implies that the variable is

4868

M d. A l MA MU N ET A L.

found to be stationary with one unknown time break. Zivot and Andrews (2002) unit root test considers all potential break points and estimates them successively  _ and finally picks the break when c ð¼  c  1Þ ¼  1 from the region where the end points of the sample period are excluded. More importantly, we apply Gregory  and Hansen (1996a, 1996b) framework for cointegration that considers the single endogenous structural breaks. Our labour productivity model is as follows: LP t  ¼  α þ β1 REM t  þ  β2 MERt  þ  β3 FCF t    þ εt 

(13)

where  LP t  is labour productivity,  REM t  is remittance inflow, MERt  is nominal exchange rate and FCF t  is fixed capital formation. Gregory and Hansen (1996a, 1996b) propose three different models with variant assumptions. Model: level shift with trend Y t  ¼ μ 1 þ μ2 f tk  þ  β1 t  þ α1 X t  þ εt 

 

(14)

Model: regime shift where intercept and slope coefficients change Y t  ¼  μ 1 þ μ2 f tk þ  β1 t  þ α1 X t  þ α2 X t f    tk  þ εt    (15)

Model: regime shift where intercept, slope coefficients and trend change Y t  ¼ μ 1 þ μ2 f tk  þ  β1 t  þ β2 tf tk  þ α1 X t    þ α2 X t f    tk  þ εt 

(16)

In the above equations, Y   is the dependent variable while  X  are independent variables. Moreover, k is the break date while φ is a dummy variable such that  f tk  ¼  0   if t hk and f tk  ¼  1 if t ik:

The above frameworks endogenously determine a single break and provide the predicted time of break  within the sample. The framework selects the break  date where the test statistic is the least vis-à-vis the absolute ADF test statistic is the highest. Finally, we compare the calculated value of this approach with MacKinnon’s (1991) critical value to ensure breaks.

applying Lind and Mehlum ’s (2010) test. This test is robust since this approach satisfies the necessary  condition to test the presence of  U  or inverted U shape. This approach argues that the conventional econometric model is unable to test the composite null hypothesis that the left side of the interval relationship is decreasing and at the right side of  the interval relationship is increasing or   vice versa (Sohag, Begum, and Abdullah   2015; Begum et al. 2015). Therefore, we test the presence of a U -shape relation using the following equation: LP t  ¼  α REM t  þ bREM t 2 þ Z t C þ εt 

The Sasabuchi –Lind –Mehlum test for a U-shape relation

We also investigate the nature of the relation between remittance and labour productivity by 

(17)

After estimating the equation, we conduct the  joint hypothesis test H 0 : ðα þ b2REM min    0 Þ [ðα þ b2REM max    0 Þ   against the alternative hypothesis: H 1 :  ð α þ b2REM min >  0 Þ [ðα þ b2 REM max < 0Þ. Here,  REM min  and  REM max  represent the maximum and minimum value of remittance. If  the null hypothesis is rejected, this confirms the existence of the  U -shape.  Asymmetric non-linear ARDL framework 

There is an increasing popularity for the non-linear econometric framework since the nature of relation among the variables is not always linear. We, therefore, employ Shin, Yu, and Greenwood-Nimmo’s (2011) non-linear ARDL (NARDL) framework that incorporates an error correction mechanism. This framework has some distinct features. For instance, parameters are estimated by OLS through incorporating persistent and stationary variables in a coherent pattern. The framework also assumes asymmetry in the long-run relation. 0

0

   y  ¼  β þ x þ t  þ β x t  þ ut ; Δx t  ¼  v t ;

 

(18)

where y t  represents a scalar  I (1) variable and  x t  is a  k × 1 vector of regressors defined such that  x t  ¼  x 0 þ x þ t  þ x t  , where x 0  is the initial value and where x þ t  ¼ t 

Non-linear approaches

 



t 

 j¼1

t 

t 

þ

P Δx  ¼ P  j

 j¼1

max (Δx  j; 0Þ and x  t  ¼

P Δx  ¼ P min (Δx  0Þ   are partial sum processes

 j¼1

 j

 j¼1

 j;

of positive and negative changes in  x t . Following Shin, Yu, and Greenwood-Nimmo (2011), we assume a single known threshold value of zero to ensure that the framework retains a clear economic interpretation.

APPLIED ECONOMICS

The NARDL ( p,q) in the levels framework, embedding (1) is written as follows:  p

q

þ0 þ 0  t  j  þ θ  j x t  j Þ

X ; y   þ Xðθ  x   y   ¼  j t  j

t 

þ εt 

 j

 j¼1

 j¼0

(19) where ; j  are autoregressive parameters, θ  jþ and θ  j contain the asymmetric distributed-lag parameters, and εt  is an independent and identically distributed (i.i.d.) process with zero mean and constant variance, σ 2ε . The associated error correction representation is  y t  ¼ ρ y t 1 þ

0 0  θ  jþ x þ t  j  þ θ  j x t  j  þ

 p1

X γ y 

 j t  j

 j¼1

q1

þ0 þ 0  θ   þ  j t  j  j x t  j Þ þ εt 

Xðθ  x  þ

 

(20)

 j¼0

where  p

 p

X ;   1; γ  ¼ X θ  for j ¼  1;  ;  ρ  ¼ X θ  ; θ  ¼ X θ  ; φ ¼ θ   ; φ  p  1; θ  ¼ X θ   for j ¼ 1;  :; q  1; φ ¼ θ   ; φ ¼ X θ   for j ¼ 1;  :; q  1; ¼  j

 . . . . . .

i

 j

 j¼1

i¼ jþ1

q

þ

q

þ

 j



 j¼0



 j

þ 0

þ 0

 0

 0

þ

 j

 j¼0

q

þ

 . . .

 j



 j

i¼ jþ1 q



 . . .

 j

i¼ jþ1

θ þ and  β ¼  and  β ¼ θ  = ρ  ρ are the asymmetric long-run parameters. By specifying a marginal data generating process of the form þ

q1

Δx t  ¼

P

 j¼1

Λ j Δx t  j  þ νt ,

and expressing εt   condition-

ally on νt   such that εt  ¼ ω 0 ðΔx t  

q1

P

 j¼1

Λ j Δx t  j Þ

þ et ,

where  e t  is uncorrelated with ν t  by construction, it is straightforward to derive the conditional non-linear ECM as follows:

4869

þ   þ where π þ 0 ¼  θ 0 þ ω; π 0 ¼  θ 0 þ ω; π  j

¼  φ jþ  ω0 Λ j and  π  j ¼  φ jω0 Λ j  for j ¼  1; . . . :; q  1: Generally, there are two tests that are applied to examine the existence of a stable long-run level relation, t BDM -statistic and F PSS. The t BDM-statistic proposed by Banerjee, Dolado, and Mestre (1998) tests H 0 :  ρ  ¼  0 against H 1 :  ρ <   0, while the F PSS statistic proposed by Pesaran, Shin, and Smith (2001, PSS) tests the joint null H 0 :  ρ  ¼  θ þ ¼  θ  ¼ 0:   The bounds testing framework of PSS offers a means to obtain valid inference of the existence of  both stationary and non-stationary variables, a characteristic, which Shin, Yu, and Greenwood-Nimmo (2011) view as highly desirable in the existence of  partial sum decompositions and may show complex interdependencies. At an empirical level, Shin, Yu, and Greenwood-Nimmo (2011) propose counting the regressors in  x t  before decomposition and choosing the suitable critical values from those tabulated in PSS to endorse conservatism. Equation (15) represents the general form, considering both long-run and short-run asymmetries. The Wald joint significance test can be applied to examine the null hypotheses of long-run symmetry  π þ ¼  π  . In addition, restrictions of short-run sym metry are considered in two different ways:  π þ i ¼  π i q1

for all i   = 0. . . q

–

1 or

þ

q1



P π  ¼ P π   . However, i¼0

i

i¼0

i

both forms can be evaluated applying the standard Wald test. Furthermore, we also examine the sym metry of the impact multipliers ði:e: H 0 :  π þ 0 ¼  π 0 ). Shin, Yu, and Greenwood-Nimmo (2011) reveal that the asymmetric dynamic multiplier effect on Y t is  related with unit changes in x þ t  and  x t    and is measured recursively from the parameters of the NARDL in levels representation in Equation (13).

V. Result and discussion Standard ARDL bounds test 

Δ y t 

¼  ρ y t 1 þ

0 0  θ  jþ x þ t  j  þ θ  j x t  j  þ

 p1

X γ y 

 j t  j

 j¼1

q1

þ0 þ 0  t  j  þ π  j x t  j Þ þ εt 

Xðπ  x  þ  j

 j¼0

 

(21)

Table 1  reports the unit root test result and shows that all the respective variables have a unit root problem at levels except for NER. However, the unit root problem ceases to exist at first difference. This mixed order of integration clearly confirms the

4870

M d. A l MA MU N ET A L.

Table 1.  DF-GLS unit root test. Log levels ( Z t )

Log 1st difference ( Z t )

Variable ln LP  ln REM ln FCF  ln NER

SIC lag 0 0 1 1

DFGLS stat 1.6018 −1.7951 −2.2751 −3.3769**

Variable Δln LP  Δln REM Δln FCF  Δln NER

SIC lag 1 1 1 1

DFGLS stat −5.5473*** −5.4399*** −3.725** −2.2494**

I (d) I (1) I (1) I (1) I (0)

Note: ** at 5%, ***at 1% and * at 10%. SIC: Schwarz information criterion.

Table 2.  Normalization result from bound test. Dep. var. F LP(LP| REM, NER,FCF) F REM(REM| LP, NER, FCF) F FC(FCF| LP, REM, NER) F RER(NER| LP, REM, FCF)

SIC lag 1 1 1 1

Table 3. Estimated long run coefficients using the ARDL

F statistic Probability Outcome 6.581 0.016** Cointegration 8.425 0.008*** Cointegration

3.413 0.429

0.075* 0.784

Cointegration No cointegration

Note: ** at 5%, ***at 1% and * at 10%.

appropriateness of using the ARDL framework of  Pesaran, Shin, and Smith (2001). Upon confirming the suitability of the Pesaran, Shin, and Smith (2001) approach, we use Equation (1) to run the ARDL framework. We choose the SIC maximum lag length of 1 for the conditional ARDLVECM. At first, we estimate OLS for the first difference in Equation (1) and then lagged variables are included to test the joint significance of the parameters.   Table 2   presents the calculated F -statistics along with the probability of having cointegrated relations. We also introduce the normalization method proposed by Pesaran and Pesaran (1997) by  considering each variable as a dependent variable in ARDL-OLS. The results show that F LP   (LP| REM, NER, FCF) = 6.581, for F REM   (REM| LP, NER, FCF) = 8.425 and for F FCF   (FC| LP, NEM, RER) = 3.413. All these results are higher than the upper bounds of the bounds test critical value at 1%, 5% and 10%, respectively. Thus, we conclude that there exists a long-run cointegrating relation among the variables when the regressions are normalized on LP, REM and FCF. Our result is immune from the problem of endogeneity, since Pesaran, Shin, and Smith (2001) also solve such a problem by taking enough lag orders. We estimate Equation (2) using the ARDL (1, 1, 0, 0) specifications to test the long-run impact of  each variable.   Table 3   reports that remittance imparts a significant positive effect on domestic labour productivity in Bangladesh. Specifically,

approach. ARDL(1,1,0,0) selected based on Schwarz Bayesian criteria. Dependent variable is InRPW Regressor C 

ln ln ln ln

REM NER FCF  REM*FCF 

Coefficient 6.604 0.102 −0.269 0.048 0.143

SE 1.501 0.031 0.144 0.065 0.085

T - ratio

4.422 3.290 −1.864 0.748 3.194

Probability 0.000 0.004 0.077 0.000 0.004

ceteris paribus, a 1% increase in the remittance inflow leads to approximately a 0.102% increase in domestic labour productivity. Alternatively, the effect of remittance-induced leisure is significantly  lower than remittance-induced labour productivity. In addition, the result also rejects the possibility of  remittance-induced Dutch disease in Bangladesh. These results are interesting findings. Mamun and Nath (2010) show that vocational training, education expenditure and investment used to establish small business enterprises are notable alternative uses of  the remittance, which indirectly enhances the potential for increased labour productivity. The impact of domestic fixed capital is also very  significant for increasing domestic labour productivity. The estimated coefficient (Table 3) suggests that a 1% increase in the fixed capital formation leads to approximately a 0.048% increase in domestic labour productivity. The economic reasoning for such a result is quite appealing. Fixed capital formation optimizes the suboptimum capital–labour ratio in Bangladesh, which in turn increases labour productivity within the classical Cobb –Douglas framework. Given the result, we argue that remittance further contributes to the domestic capital formation, vis-à vis overall productivity. Our result shows that the interaction between remittance and fixed capital formation leads to a greater level of domestic labour productivity. The result suggests that a joint increase in both fixed capital formation and remittance inflow 

APPLIED ECONOMICS

by 1% approximately results in a 0.143% increase in the real GDP chain per workers. Therefore, both remittance and domestic capital formation are crucial in augmenting the economic potential of the abundant labour force in Bangladesh. Finally, the impact of the nominal exchange rate on domestic labour productivity is negative and statistically significant. The result shows that  ceteris paribus   a depreciation of local currency against foreign currency (primarily US$) by 1% will result in a 0.269% decline in domestic labour productivity. This finding is inconsistent with previous literature. Jeanneney and Hua (2011) argue that there is a positive effect of real depreciation upon the labour productivity. The positive impact is based on the notion that depreciation of the nominal exchange rate may  improve the competitiveness of local enterprises in the international market. It also raises exports and creates an environment of import substitution with the growth of the industrial sector, inviting additional foreign direct investments. Industrial growth and FDI are favourable factors of efficiency and technical progress and hence labour productivity. Why is there a different story in the case of  Bangladesh? One of the potential explanations is that Bangladesh is an import-dependent country  and a reduction in nominal exchange rate increases the trade deficit in two ways.  First , it increases the import bills of raw materials and other capital goods which reduce the optimal labour –capital combination imparting a negative effect on the value of  marginal labour productivity (MLP).   Second , the imbalance in trade increases since domestic infrastructure readiness to export to Bangladesh or even to substitute the import sector in the local market by  local firms is very slim. Therefore, a depreciation of 

4871

local currency increases import bills without any  positive impact on exports, which leads to lower productivity of the local industry. In addition, a depreciation of the local currency, will theoretically  bring a positive result for the export sectors. Therefore, a shift from non-tradable to the export sector is likely. However, the export sector is always  vulnerable to international economic realities compared with non-tradable sectors. Thus, the long-run labour productivity is likely to be most unstable with a change in the nominal exchange rate. Table 4 (panel A) suggests that the sign of the shortrun dynamics is consistent with the sign of long-run relations. However, the impact of remittance on labour productivity is statistically insignificant. Thus, remittance does not have a first order short-run impact on labour productivity, rather it takes time. The estimated long-run relation in Equation (2) holds since the error correction coefficient is negative and statistically significant. It suggests that there is a high speed of adjustment towards the long-term equilibrium relation between remittance and labour productivity and remittance. Specifically, the error correction coefficient suggests that the disequilibrium that caused the previous year ’s temporal shock is adjusted by 54% per year. The overall model fit for ARDL regression Equation (1) is statistically satisfactory since the model passes most of the model check tests, including a test for serial autocorrelation, normality, the heteroscedasticity problem, and functional form as indicated in Panel B of  Table 4. Finally,   Figure 2   presents the stability test for coefficients. The value of CUSUM and CUSUMSQ plot is within the critical boundaries at the 5% significance level. Thus, the long-run coefficients and all the short-run coefficients in the error correction model are stable over the sampled periods.

Table 4.  Error correction representation for the selected ARDL approach. Panel A: ARDL (1,1,0,0) selected based on Schwarz Bayesian criterion. Dependent variable is d RPW  Regressor Coefficient SE 0.006 0.026 Δln REM 0.026 0.035 Δln FC  0.066 Δln RER −0.144 Δln FC *REM 0.086 −0.018 0.161 −0.535 ECM  ( –1) ECM  = RPW  –  0.10273REM  + 2

= 0.8294, adjusted SER  = 0.52836  AIC   = 70.1331 R

0.26986RER –  0.048779FC   + 0.143REM*FC  –  6.6408C  2 R = 0.8326 RSS = 0.0043561 SIC   = 66.3588

Panel B: ARDL-VECM model diagnostic tests Serial correlation  X 2 ð1Þ=0.664 [0.215] Functional form  X 2 ð1Þ= 3.427 [0.064]

T -ratio

0.223 0.733 −2.189 0.894 −3.320

Probability 0.825 0.472 0.040 0.324 0.003

F -stat

[4, 21] = 5.6012 [0.003] DW-statistics = 2.1061

Normality X 2 ð2Þ=0.476 [0.078] Heteroscedastic X 2 ð1Þ=1.468 [0.226]

4872

M d. A l MA MU N ET A L.

 Plot of Cumulative Sum of Squares of Recursive Residuals

 Plot of Cumulative Sum of Recursive Residuals 15

1.5

10

1.0

5 0

0.5

 –5

0.0

 –10  –15 1983

1988

1993

1998

2003

2008

2011

 –0.5 1983

1988

1993

1998

2003

2008

2011

 The straight lines represent critical bounds at 5% significance level

The straight lines represent critical bounds at 5% significance level

Figure 2.  The test for stability of the coefficients.

Remittance  –  labour productivity nexus: assumption of structural break 

Our sample year consists of the change in both the political regime and significant economic policies in Bangladesh. For example, in the early 1990s, Bangladesh moved towards parliamentary democracy and started liberalizing her financial sector. We contextualize this change by applying the Zivot–Andrews unit-root test.   Table 5   shows that our variables of interest exhibit unit-root properties at the level, but are stationary with one endogenous structural break, at the first difference. The result also provides information about the time at which the structural break took place. For instance, most notable structural break of remittance, nominal Table 5.  Zivot–Andrews structural break unit root test. Z&A test for level Variable LREM NER LLP  LFCF 

T statistic −3.271 −4.107 −2.366 −3.079

TB 2002 1992 1998 1994

Outcome Unit root Unit root Unit root Unit root

Z&A test for 1st difference T Statistic −9.036a −5.474b −7.158a −4.123c

TB 1992 2007 2006 2007

Outcome Stationary Stationary Stationary Stationary

Note: a, b and c indicate 1%, 5% and 10% significance level, respectively.

exchange rate, labour productivity and fixed capital formation that took place in 1992, 2007, 2006, and 2007, respectively. Since we find strong evidence for a structural break of all variables, we further evaluate whether these variables are cointegrated under the assumption of an endogenous structural break. We apply the Gregory –Hansen cointegration approach.  Table 6  presents the result. The result is consistent with a long-run relation under all three assumptions. For instance, ADF and  Z t  tests consistently confirm the existence of cointegration Table 6. Gregory–Hansen test for cointegration with regime

shifts: (model: change in level). Test ADF

Statistic Breakpoint Date 1% 5% 10% 6 1987 −5.47 −5.77 −5.28 −5.02 6 1987  Z t  −5.60 −5.77 −5.28 −5.02 6 1987 −27.56 −63.64 −53.58 −48.65  Z a Gregory–Hansen test for cointegration with regime shifts: model: change in regime ADF 11 1992 −5.99 −6.51 −6.00 −5.75 6 1987 −6.72 −6.51 −6.00 −5.75  Z t  6 1987  Z a −36.44 −80.15 −68.94 −63.42 Gregory–Hansen test for cointegration with regime shifts: model: change in regime and trend ADF 11 1992 −6.81 −6.89 −6.32 −6.16 6 1987 −6.42 −6.89 −6.32 −6.16  Z t  6 1987 −35.50 −90.84 −78.87 −72.75  Z a

Figure 3.  Structural break-test for labour productivity for 1988 –2008.

APPLIED ECONOMICS

between labour productivity and remittance under the three assumptions.  Table 6 also detects the year of breaks, which mainly occurred in 1987 and 1992. A closer look at Figure 3 provides additional insight consistent with the reality. For example, the figure shows that a downward spiral starts in 1987, a period marking the beginning of mass protests against the military dictator leading towards a democratic system in 1991. Afterwards, the downward spiral stopped and from 1991 until 2002 there was a relatively stable rate of labour productivity. In this period, Bangladesh moved closer towards democracy  and financial sector liberalization. From 2002, the labour productivity rate shot up significantly, and the growth phase continued until 2007. Afterwards, the growth trend took an inverted- U  turn (Figure 3). This result is partly attributed to the adverse impact of global financial crisis as well as the political stampede that has beset the country since 2007.

Remittance-labour productivity: non-linear  approach

The existence of a structural break in the cointegrating relation motivates us to check for the possible presence of non-linearity. We primarily apply Lind and Mehlum (2010) and the NARDL approach to detect non-linearity in our estimation. Sasabuchi –Lind –Mehlum tests for U-shape relation

The application of the SLM  U -test in augmenting the linkage between remittance-labour productivity in the context of Bangladesh is based on some anecdotal and theoretical grounds. In the spirit of Lucas and Stark (1985) and Poirine (1997), one can argue that the impact of remittance on labour productivity  does not follow a straight line. Furthermore, pre vious studies on the possible positive and negative consequences of remittance on gross domestic productivity reaffirm this understanding.   Table 7 reports that the lower bound slope of remittance is Table 7.  Sasabuchi–Lind–Mehlum test for  U -shape relationship. Interval Slope T -value P -value Hypothesis test Overall test presence of a  U -shape

Lower bound 6.429 −0.100 −3.352 0.0012 T -value 3.35

Upper bound 11.238 0.079 4.941 0.000 P -value 0.001

4873

negative (–0.100), while the upper bound slope of  remittance is positive (0.079) and statistically significant. This result implies the relationship between remittance and labour productivity follows a U -shape. Furthermore, the null hypothesis of no U shape is rejected against the alternative of  U -shape relationship between remittance and labour productivity. The NARDL approach

Table 8   presents the results of the unrestricted NARDL estimator under a long-run and short-run asymmetric assumption. It is worthy of mention that both the t BDM and F PSS   statistics reject the null hypothesis of no cointegration under the assumption of long-run and short-run asymmetry in the model. It may also be noteworthy that our previous estimates under the standard ARDL approach also fail to detect this asymmetric long-run relationship. However, the NARDL estimator efficiently detected both the short-run and long-run asymmetry under the unrestricted error correction mechanism. Table 8 clearly shows that the long-run coefficients on  REM + and REM – are 0.133 and –0.417, respectively. Statistically, when there is an increase in remittance by 1%, the labour productivity increases by 0.133%, while a decline in remittance by 1%, results in a decrease in labour productivity by 0.417% in the long run. The economic and policy significance of these findings are robust. Regarding the economic importance, the receipt of remittance by Bangladesh is utilized more productively and as a labour abundant country the optimization of the capital –labour ratio is immensely importance to the overall productivity  growth of Bangladesh. In recent times, the continued political violence has had an adverse impact on FDI flow. However, the flow of remittance as an alternative source of foreign capital can still grow according to the understanding underscored by Lucas and Stark (1985) and Poirine (1997). Regarding policy  significance, the government should recognize and work out their current diplomatic standoff with countries, such as the United States, United Kingdom, KSA, Malaysia and Australia, which will enable Bangladesh to reap further gain from its abundant labour. In addition, an appreciation of nominal exchange rate does not have any impact on labour

4874

M d. A l MA MU N ET A L.

Table 8.  Linkage between remittance and labour productivity ΔLabor

productivity

r 1 þ REM1  REM1 þ RER1  RER1 þ FCF 1  FCF 1 Δr 1 ΔREMþ ΔREMþ 1  Α

 

Observations R-squared

Coefficient −0.957** 0.128** 0.399* 0.178 −0.0420 0.00805 0 0.0992 0.0485 0.0116 7.027** (2.711) 28 0.847

–

SE (0.366) (0.0448) (0.180) (0.292) (0.151) (0.102) (0) (0.310) (0.0539) (0.0310)

 NARDL approach.

ΔLabor

productivity

ΔREM ΔREM 1 ΔRERþ ΔRERþ 1 ΔRER ΔRER 1 ΔFCF þ þ ΔFCF  1 ΔFCF   ΔFCF  1

Coefficient 0.100 −0.270** −0.171 −0.587 −0.0382 0.204 0.128 −0.114 0 0

SE (0.148) (0.0856) (0.226) (0.344) (0.186) (0.208) (0.146) (0.211) (0) 0

Long-run effect [+] Long-run effect [−] Coeff. F -stat P -value F -stat P -value 11.14 0.008 11.48 0.007 REM −0.417 0.276 0.611 0.044 0.071 0.794 NER 0.006 0.937 0.000 0.000 – FCF  Long-run asymmetry Short-run asymmetry F -stat P -value F -stat P -value 5.064 0.048 2.962 0.082 REM 0.255 0.624 2.686 0.132 NER 0.006 0.937 0.006 0.939 FCF  Cointegration test statistics: Model diagnostics: Heteroscedastic X 2 ð1Þ=1.307 [0.252] T BDM = -4.614 (upper bound critical value 3.99 at 5%) F PSS = 7.0636 (upper bound critical value 4.01 at 5%) Functional form X 2 ð1Þ= 0.839 [0.514] Normality X 2 ð2Þ=0.109 [0.947] R-square 0.72. Adjusted R  sq. 0.64, root MSE 0.012 Exogenous variable

Coeff. 0.133 0.186 0.008

productivity, while and depreciation of nominal exchange rate leads to an increase in labour productivity in the long run. Interestingly, this is in harmony with previous literature (see Jeanneney and Hua 2011) but in contrast to our previous findings under the standard ARDL estimator. We attribute this result of the NARDL estimate to the robust assumptions, estimation efficiency, and the improved and unrestricted assumptions of NARDL compared with standard ARDL. Finally, we also find that an increase of fixed capital formation fosters labour productivity. Moreover, the asymmetric test confirms that only remittance asymmetrically adjusts in both in the long-run and short-run context, while the rest of the regressors follow a symmetric adjustment in both the long term and short run. The diagnostic test of the NARDL approach shows that the model does not contain any heteroscedastic, functional and normality problems. The adjusted R-square value of 0.64 clearly highlights a fit of the model. In addition, the stability test (Figure 4) confirms the stable impact of remittance and nominal exchange rate as the parameter for them goes through the 95% critical boundary.

VI. Conclusion and policy implications

This article investigates the role of remittance on domestic labour productivity in one of the most densely populated countries in the world. Bangladesh has historically inherited a huge young labour force. Any growth agenda for Bangladesh, therefore, should primarily aim at increasing domestic labour productivity. Migration of the labour force from Bangladesh contributes to this end in two different ways –  first migration reduces the stock of  domestic labour supply providing much-needed relief on the labour market and remittance sent by  migrant workers working as a foreign supply of  capital –   both contribute in optimizing the current under-optimum, capital–labour ratio in Bangladesh. We employ several testing approaches including standard ARDL and NARDL for the sample period of  1982–2013. Our result suggests that remittance imparts a positive and significant impact on domestic labour productivity both in the short and long rununder linear, structural break and non-linear assumptions. These findings are robust and can contribute to the overall improvement of domestic labour productivity in

APPLIED ECONOMICS

Cumulative effect of LNREM on LNRPW

4875

Cumulative effect of LNRER on LNRPW

       2  .

       1

       0

       5  .

       2  .    –

       0

       4  .    –

       5  .    –

       6  .    –

       1    –

0

10

20

30

40

Time periods positive change asymmetry

0

10

20

30

40

Time periods negative change CI for asymmetry

positive change asymmetry

Note: 95% bootstrap CI is based on 100 replications

negative change CI for asymmetry

Note: 95% bootstrap CI is based on 100 replications

Cumulative effect of LNFC on LNRPW        4  .        2  .        0        2  .    –        4  .    –

0

10

20 Time periods

positive change asymmetry

30

40

negative change CI for asymmetry

Note: 95% bootstrap CI is based on 100 replications

Figure 4.  Stability test of coefficient in NARDL approach.

Bangladesh. Moreover, with an increased diplomatic impasse between Bangladesh and its development partners coupled with the developing partners themselves being suffocated by the ongoing financial crisis, the flow  of FDI has started to wilt for Bangladesh. At this juncture, a constant stream of remittance as alternative sources of foreign capital can provide a recipe for continued economic growth in Bangladesh. Today remittance inflow accounts for approximately  12% of GDP of Bangladesh. Given our result, we argue that proper political diplomacy, provision of necessary  skills and education for its young population (70% of  overall population) and substantial policy reform to manage the global demand for Bangladeshi workers, can only increase this inflow and continue to augment domestic labour productivity for Bangladesh. Disclosure statement No potential conflict of interest was reported by the authors.

ORCID Kazi Sohag   http://orcid.org/0000-0002-0976-2357

References Acosta, P. A., E. K. K. Lartey, and F. S. Mandelman.   2009. and the Dutch Disease.”   Journal of  “Remittances International Economics   79: 102–116. doi:10.1016/j.  jinteco.2009.06.007.

Adams, R., Jr.   1993. “The Economic and Demographic Determinants of International Migration in Rural Egypt.”  Journal of Development Studies  30: 146–167. doi:10.1080/ 00220389308422308. Adams, R. H., Jr., and A. Cuecuecha. 2010. “Remittances, Household Expenditure and Investment in Guatemala.” World Development   38 (11): 1626–1641. doi:10.1016/j. worlddev.2010.03.003. Adams, R. H., and J. Page. 2005. “Do International Migration and Remittances Reduce Poverty in Developing Countries?”   World Development    33 (10): 1645–1669. doi:10.1016/j.worlddev.2005.05.004. Aggarwal, R., A. Demirguc Kunt, and M. S. Martinez Peria. 2010. “Do Remittances Promote Financial Development? ”  Journal of Development Economics. doi:10.1016/j.  jdeveco.2010.10.005. Ahmed, M. S.   2010.   Migrant Workers Remittance and  Economic Growth: Evidence from Bangladesh. ASA University Review. Dhaka: ASA University. Vol. 4, No. 1 (January –June). Al Mamun, M., K. Sohag, G. S. Uddin, and M. Shahbaz.  2015. “Remittance and Domestic Labor Productivity: Evidence from Remittance Recipient Countries.”  Economic Modelling  47: 207–218. doi:10.1016/j.econmod.2015.02.024. Amuedo-Dorantes, C., and S. Pozo.   2004. “Workers’ Remittances and the Real Exchange Rate: A Paradox of  Gifts.” World Development   32 (8): 1407–1417. doi:10.1016/  j.worlddev.2004.02.004. Ang, J. B., and W. J. McKibbin.   2005. “Financial Liberalization, Financial Sector Development and Growth: Evidence from Malaysia.”   Brookings Discussion Papers in International Economics  168: 1–29. Banerjee, A., J. Dolado, and R. Mestre.   1998. “ErrorCorrection Mechanism Tests for Cointegration in a Single-Equation Framework.”   Journal of Time Series  Analysis  19 (3): 267–283. doi: 10.1111/jtsa.1998.19.issue-3.

4876

M d. A l MA MU N ET A L.

Bangladesh Bank.   2012.   Wage Earners Remittance Inflows (Yearly). Dhaka: The Central Bank of Bangladesh, Government of the People ’s Republic of Bangladesh. http:// www.bangladesh-bank.org/econdata/wageremitance.php# . Barai, M. K.  2012. “Development Dynamics of Remittances in Bangladesh.”   SAGE Open   2 (1): 1–13. doi:10.1177/ 2158244012439073. Bayangos, V., and K. Jansen.   2011. “Remittances and Competitiveness: The Case of the Philippines.”   World  Development    39 (10): 1834–1846. doi:10.1016/j. worlddev.2011.04.019. Begum, R. A., K. Sohag, S. M. S. Abdullah, and M. Jaafar. 2015. “CO2  Emissions, Energy Consumption, Economic and Population Growth in Malaysia.”   Renewable and  Sustainable Energy Reviews   41: 594–601. doi:10.1016/j. rser.2014.07.205. BMET.   2013.   Bureau of Manpower Employment and  Training . Dhaka: Government of the Peoples ’  Republic of  Bangladesh. http://www.bmet.org.bd. Bourdet, Y., and H. Falck.  2006. “Emigrants’   Remittances and Dutch Disease in Cape Verde.”   International  Economic Journal    20 (3): 267–284. doi:10.1080/ 10168730600879323. Buiter, W., and E. Rahbari.   2011. “Global Growth Generators: Moving beyond ‘Emerging Markets’ and ‘BRIC’.”  Global Economics View  21: 1 –9. Calero, C., A. S. Bedi, and R. Sparrow.  2008. “Remittances, Liquidity Constraints and Human Capital Investments in Ecuador.”  Working Paper no. 458 . The Hague: Institute of  Social Studies. Chami, R., C. Fullenkamp, and S. Jahjah. 2003. “Are Migrant Remittance Flows a Source of Capital for Development? ” IMF Working Paper . Washington, DC: International Monetary Fund, August. Chowdhury, M. B.   2011. “Remittances Flow and Financial Development in Bangladesh. ”  Economic Modelling   28 (6): 2600–2608. doi:10.1016/j.econmod.2011.07.013. Edwards, C., and M. Ureta.  2003. “International Migration, Remittances, and Schooling: Evidence from El Salvador. ”  Journal of Development Economics   72: 429–461. doi:10.1016/S0304-3878(03)00115-9. Elliott, G., T. Rothenberg, and J. Stock. 1996. “Efficient Tests for an Autoregressive Unit Root. ”  Econometrica  64: 813– 836. doi:10.2307/2171846. Engle, R. F., and C. W. J. Granger. 1987. “Co-Integration and Error Correction: Representation, Estimation, and Testing.”   Econometrica   55 (2): 251–276. doi:10.2307/ 1913236. Mar., 1987. Farid, K. S., L. Mozumdar, M. S. Kabir, and M. S. Hossain. 2009. “Trends in International Migration and Remittance Flows: Case of Bangladesh.”   Journal of Bangladesh  Agricultural University  7 (2): 387–394. Fosu, O.-A., and F. J. Magnus.   2006. “Bounds Testing Approach to Cointegration: An Examination of Foreign Direct Investment Trade and Growth Relationships.”  American Journal of Applied Sciences   3: 2079–2085. doi:10.3844/ajassp.2006.2079.2085.

Gregory, A. W., and B. E. Hansen.   1996b. “Tests for Cointegration in Models with Regime and Trend Shifts. ” Oxford Bulletin of Economics and Statistics   58 (3): 555– 560. doi:10.1111/j.1468-0084.1996.mp58003008.x. Gregory, A. W., and B. E. Hansen.  1996a. “Residual-Based Tests for Cointegration in Models with Regime Shifts.”  Journal of Econometrics  70 (1): 99–126. doi:10.1016/03044076(69)41685-7. Gupta, S., C. Pattillo, and S. Wagh.   2009. “Effect of  Remittances on Poverty and Financial Development in Sub-Saharan Africa.”   World Development   37 (1): 104– 115. doi:10.1016/j.worlddev.2008.05.007. Hanson, G. H., and C. Woodruff (2003).   Emigration and Educational Attainment in Mexico. Working Paper, University of California. Jeanneney, G. S., and P. Hua.   2011. “How Does Real Exchange Rate Influence Labour Productivity in China? ” China Economic Review   22: 628–645. doi:10.1016/j. chieco.2010.04.008. Johansen, S.   1988. “Statistical Analysis of Cointegration Vectors.”   Journal of Economic Dynamics and Control  12 (2–3): 231–254. 18. doi:10.1016/0165-1889(88)90041-3. Johansen, S.   1991. “Estimation and Hypothesis Testing of  Cointegration Vectors in Gaussian Vector Autoregressive Models.”   Econometrica   59 (6): 1551–1580. doi:10.2307/ 2938278. Johansen, S., and K. Juselius.   1990. “Maximum Likelihood Estimation and Inference on Cointegration –   with Applications to the Demand of Money. ”   Oxford Bulletin of Econometrics and Statistics  52 (2): 169–210. Khan, M. A., A. Qayyum, and A. S. Saeed.  2005. “Financial Development and Economic Growth: The Case of Pakistan. ” The Pakistan Development Review  44 (2): 819–837. Lartey, E. K. K., F. S. Mandelman, and P. A. Acosta.  2008. “Remittances, Exchange Rate Regimes, and the Dutch Disease: A Panel Data Analysis. ”   Working Paper 2008-12 . Atlanta: Federal Reserve Bank of Atlanta Levitt, T.   1983. “The Globalization of Markets.”   Harvard  Business Review, (Issue May –June), 92–102. Lind, J. T., and H. Mehlum. 2010. “With or without U? the Appropriate Test for a U-Shaped Relationship.”   Oxford  Bulletin of Economics and Statistics   72 (1): 109–118. doi:10.1111/obes.2009.72.issue-1. Lucas, R. E. B., and O. Stark.  1985. “Motivations to Remit: Evidence from Botswana. ” Journal of Political Economy  93: 901–918. doi:10.1086/261341. MacKinnon, J. G.   1991. “Critical Values for Cointegration Tests.” In   Long run Economic Relationships: Readings in Cointegration, edited by R. F. Engle and C. W. J. Granger, 267–276. Oxford: Oxford University Press. Mamun, K. A., and H. K. Nath.  2010. “Workers’  Migration and Remittances in Bangladesh.”   Journal of Business Strategies  27 (1): 29. McCormick, B., and J. Wahba.   2001. “Overseas Work  Experience, Savings and Entrepreneurship Amongst Return Migrants to Ldcs.”   Scottish Journal of Political  Economy   48: 164–178. doi:10.1111/sjpe.2001.48.issue-2.

APPLIED ECONOMICS

Noman, M. A., and S. G. Uddin.   2011. “Remittances and Banking Sector Development in South Asia.” The International Journal of Banking and Finance  8 (4): 47–66. Nyamongo, E. M., and R. N. Misati. 2011. “Remittances and Banking Sector Development in Sub Saharan Africa.” Paper presented at the Global Development Forum, Dubai, United Arab Emirates, November. Nyamongoa, E. M., R. N. Misatib, L. Kipyegonb, and L. Ndirangu. 2012. “Remittances, Financial Development and Economic Growth in Africa.”   Journal of Economics and  Business  64: 240–260. doi:10.1016/j.jeconbus.2012.01.001 . Pesaran, H. H., and Y. Shin. 1998. “Generalized Impulse Response Analysis in Linear Multivariate Models. ”  Economics Letters  58: 17–29. doi:10.1016/S0165-1765(97)00214-0. Pesaran, M. H., and B. Pesaran. 1997.  Working with Microfit  4.0: Interactive Econometric Analysis. Oxford: Oxford University Press. Pesaran, M. H., Y. Shin, and R. J. Smith.   2001. “Bounds Testing Approaches to the Analysis of Level Relationships.”  Journal Applications Economics   16: 289– 326. doi:10.1002/jae.616. Poirine, B.   1997. “A Theory of Remittances as an Implicit Family Loan Arrangement. ”  World Development   25: 589– 611. doi:10.1016/S0305-750X(97)00121-6. Quartey, P., and T. Blankson.   2004. “Do Migrant Remittances Reduce the Impact of Macro-Volatility on Poor Households in Ghana?”   Final report to the Global  Development Network, International Monetary Fund, Washington, DC. http://cloud2.gdnet.org/cms.php? id=research_paper_abstract&research_paper_id=9035. Raihan, S., B. H. Khondker, G. Sugiyarto, and S. Jha.  2009. “Remittances   and Household Welfare: A Case Study of  Bangladesh.”  ADB Economics Working Paper Series, 189, Asian Development Bank. Romer, D.   2006.   Advanced Macroeconomics. 4th ed. New  York: McGraw Hill Series Economics. Shin, Y., B. Yu, and M. Greenwood-Nimmo.   2011. Asymmetric Cointegration and Dynamic “Modelling

4877

Multipliers in a Nonlinear ARDL Framework. ”   Working  Paper (Version of November 2011 ). Siddique, A., E. A. Selvanathan, and S. Selvanathan.   2010. “Remittances and Economic Growth: Empirical Evidence from Bangladesh, India and Sri Lanka.”   Australia Discussion Paper 10.27 , University of Western. Siddiqui, R., and A. R. Kemal.   2006. “Remittances, Trade Liberalization, and Poverty in Pakistan: The Role of  Excluded Variables in Poverty Change Analysis.” The Pakistan Development Review  45 (3): 383–415. Sohag, K., R. A. Begum, and S. M. S. Abdullah.   2015. “Dynamic Impact of Household Consumption on Its CO2 Emissions in Malaysia. ”   Environment, Development  and Sustainability  17 (5): 1031–1043. Solow, R. M.   1957. “Technical Change and the Aggregate Production Function.”   The Review of Economics and  Statistics  39 (3): 312–320. doi:10.2307/1926047. Taylor, J. E., and T. J. Wyatt.   1996. “The Shadow Value of Migrant Remittances, Income and Inequality in a Household-Farm Economy.”   Journal of   Development Studies   32: 899–912. doi:10.1080/ 00220389608422445. Vargas-Silva, C.   2009. “The Tale of Three Amigos: Remittances, Exchange Rates, and Money Demand in Mexico.”  Review of Development Economics  13 (1): 1–14. doi:10.1111/rode.2009.13.issue-1. Weiss,   T. J. 1993. “Long-Term Changes in US Agricultural Output per Worker, 1800 to 1900.”   Economic History  Review  46: 324–341. World Bank.   2013. Migration and Remittance Flows: Recent Trends and Outlook, 2013–2016.   http://sitere sources.worldbank.org/INTPROSPECTS/Resources/ 334934-1288990760745/MigrationandDevelopment Brief21.pdf . Zivot, E., and D. W. K. Andrews. 2002. “Further Evidence on the Great Crash, the Oil-Price Shock, and the Unit-Root Hypothesis.”  Journal of Business & Economic Statistics 20 (1): 25–44. doi:10.1198/073500102753410372.