^{1}

^{*}

^{1}

This study is conducted to revisit the empirical relationship between exports, imports and economic growth in Bangladesh using annual time series data from 1981 to 2017. To capture the objective, the study used Johansen Co-integration test and Granger-causality test in Vector Error Correction Model (VECM) framework. Based on the results of Johansen Co-integration test, it confirms that there is statistically significant long-run equilibrium relationship between exports, imports and economic growth. The results of the VECM Granger Causality test assure that the disequilibrium in long-run GDP growth rate is corrected or adjusted by 24% in short-run following the next year. Furthermore, the study found short-run causality running from exports to economic growth and from economic growth to imports. Finally, findings of the study will help the policy makers and development partners of Bangladesh to rethink about the current policies regarding the exports, imports, inflation, gross capital formation (investment) and economic growth.

The theoretical and empirical nexus between exports, imports and economic growth has long been a subject of much interest and controversy in international trade literature because every nation wants to increase GDP and improve the quality of life for the citizen [

There are four partitions among the researchers considering the relationship among the three macro-economic variables exports, imports and economic growth [

Economic growth is considered as one of the important macroeconomic indicators of welfare, changing economic structure, uplifting GDP per capita and development of a country. Bangladesh is a developing country whose current economic growth rate is more than 7% for the last 2 years [

As the macroeconomic and development conditions of the world are not same over the time, moreover it is changing in nature that’s why the relationship among the macroeconomic variables are not one time solution but continuous. As the relationship is inconclusive in nature, again and again it demands investigation [

The objective of this research work is to econometrically revisit the nexus between exports, imports and economic growth in Bangladesh using annual time series data from 1981 to 2017. To be more specific, this paper tried to reinvestigate empirically the relationship, whether exports lead economic growth or imports lead economic growth or economic growth leads exports and imports or no relationship among the three variables. The findings and recommendations of this empirical analysis will be helpful for all development partners and policy makers of Bangladesh.

This study is different from other papers by following ways: first of all, the study used long period data sets (1981-2017). Second, this paper has greater policy implications as it used more appealing and effective econometrics models such as Johansen Co-integration model and VECM. Finally, this paper used some dominant control variables namely inflation, Gross capital formation (Investment) in its model specification to isolate the true effects of exports and imports on economic growth.

The rest of the paper is organized as follows: literature review about the similar studies is given in Section 2; data and methodology of the research are discussed in Sector 3; after then, the results and analysis, as well as conclusion of the study are presented in Section 4 and 5 respectively.

Academicians and policy makers studied and investigated the nexus between exports, imports and economic growth again and again as the issue is very important for a country. Researchers found different result regarding the nexus among these variables. Some of them are shown below:

Bakari, S. and Mabrouks, M. [

Mukhtar, W. L. (2017) [

Gulzar, A. and Zhaohua, L. [

Riyath, M. and Jahfer, A. [

Yuksel, S. and Zengin, S. [

Ridzuan, A. R. and et al. [

Ucan, O. and Akyildisz, A. [

Saaed, A. J. and Hossain, M. A. [

Sachin N. Mehta [

Saaed, A. J. and Hossain, M. A. [

Hussain, M. and Saaed, A. [

Mahmood, M. et al. [

Velnampy, T. and Achchuthan, S. [

Iqbal et al. [

Khan, T. F. and Kundu, N. [

Yuhong, L. et al. [

Based on the literature review what the paper described above, one could conclude that there is a wide scope to revisit the relationship between exports, imports and economic growth. The recent statistics of international trade in Bangladesh also suggest revisiting the impacts of exports and imports on economic growth. Thus, this study attempted to investigate the nexus among the variables under consideration.

This empirical study used annual time series data for all variables whether dependent or independent for the period of 1981-2017 (37 observations). The data of Gross Domestic Product (GDP) in annual growth rate, Exports (EP), Imports (IMP) and Gross Capital Formation (GCF) in percentage of GDP are collected from World Development Indicator [

Empirical studies suggest that including control variables in a model is an essential way to isolate the impact of independent variable on the dependent variable. This study considers GDP as a dependent variable; exports and imports as independent variables and inflation as well as gross capital formation as control variables. To apply empirical model and analysis all data are converted into natural logarithm form. Every estimations and diagnostic tests are carried out using Econometric views (E-views) version 10.0 statistical software.

To examine the empirical nexus between exports, imports and economic growth an econometric model is used to link among the variables of the study.

Economic Growth = f ( Exports,Imports,Inflation,Investment ) (1)

ln G D P t = β 0 + β 1 ln E P t + β 2 ln I M P t + β 3 ln I N F t + β 4 ln G C F t + є t ⋯ (2)

where,

β_{0} = Constant term; GDP = Gross domestic product at annual growth rate; EP = Exports (% of GDP); IMP = Imports (% of GDP); INF = Inflation rate at percentage change in consumer price index (CPI); GCF = Investment growth rate measured by using Gross Capital Formation; є t = error term assumed to be normally, identically and independently distributed, while β_{1}, β_{2}, β_{3}, β_{4} are coefficients and ln indicates the natural logarithm form of variables.

Assumptions of Gauss-Markov are used in this study for testifying the validity and strength of the Ordinary Least Square (OLS). These assumptions include, model’s linearity, unbiased estimation (β_{0}, β_{1}, β_{2}, β_{3}, β_{4}) with expected value of zero i.e. E ( є t ) = 0 and distribution with equal variance (homoscedasticity).

The long-run and short-run relationship between exports, imports and economic growth is examined using Johansen co-integration approach and Granger causality test in Vector Error Correction Model (VECM) framework.

In co-integration analysis especially in time series data, researchers have to be conscious about the stationary of the data as there may raise a problem of spurious regression. It is well discussed issue that macroeconomic time series variables follow a random walk model that means exhibit a unit root behavior. So, to address the time series issues and related unit root test, the most popularly used techniques: The Augmented Dickey-Fuller [

ADF and PP test models are as follows respectively:

Δ Z t = x + ( ρ − 1 ) Z t − 1 + γ T + δ Δ Z t − 1 + ε 1 t (3a)

Δ Z t = λ 0 + λ 1 t + δ Z t − 1 + ε 2 t (3b)

After running unit root test, if the variables are not stationary at level but stationary at first difference that is if the variable under study found stationary at same order, say in their first difference it is possible to run the regression. After running the regression, if the error terms are found stationary at level, then the linear combination of the individually non stationary variables are said to be stationary. In this case it can be said to be integrated and economically interpretable as long run relationship among the macroeconomic variables [

According to Utkulu [

Johansen’s procedure will be applied to overcome the problems of Engle-Granger two steps procedure. Similar to the Engle-Granger approach, here also stationary of data will be checked. If all variables are found to be integrated stationary at the same order then the co-integrating analysis continues without suffering from spurious regression [

Since long-run relationships are mostly explained in static equilibrium form. For this reason it is difficult to explain the dynamics of structural and institutional changes occur in the economy within the short-run. Considering this limitation, it is necessary to study the short-run relationship and short-run dynamism of the variables under the study. The VECM is the best possible way to assess the short-run dynamic structure of the model. The Granger Representation Theorem [

Δ L N G D P t = μ 11 + ∑ i = 1 k μ 12 Δ L N G D P t − i + ∑ i = 1 k μ 13 Δ L N E P t − i + ∑ i = 1 k μ 14 Δ L N I M P t − i + ∑ i = 1 k μ 15 Δ L N I N F t − i + ∑ i = 1 k μ 16 Δ L N G C F t − i + μ 17 E C T t − 1 + є t (1)

Δ L N E P t = μ 21 + ∑ i = 1 k μ 22 Δ L N G D P t − i + ∑ i = 1 k μ 23 Δ L N E P t − i + ∑ i = 1 k μ 24 Δ L N I M P t − i + ∑ i = 1 k μ 25 Δ L N I N F t − i + ∑ i = 1 k μ 26 Δ L N G C F t − i + μ 27 E C T t − 1 + ϵ t (2)

Δ L N I M P t = μ 31 + ∑ i = 1 k μ 32 Δ L N G D P t − i + ∑ i = 1 k μ 33 Δ L N E P t − i + ∑ i = 1 k μ 34 Δ L N I M P t − i + ∑ i = 1 k μ 35 Δ L N I N F t − i + ∑ i = 1 k μ 36 Δ L N G C F t − i + μ 37 E C T t − 1 + є t (3)

Δ L N I N F t = μ 41 + ∑ i = 1 k μ 42 Δ L N G D P t − i + ∑ i = 1 k μ 43 Δ L N E P t − i + ∑ i = 1 k μ 44 Δ L N I M P t − i + ∑ i = 1 k μ 45 Δ L N I N F t − i + ∑ i = 1 k μ 46 Δ L N G C F t − i + μ 47 E C T t − 1 + ϵ t (4)

Δ L N G C F t = μ 51 + ∑ i = 1 k μ 52 Δ L N G D P t − i + ∑ i = 1 k μ 53 Δ L N E P t − i + ∑ i = 1 k μ 54 Δ L N I M P t − i + ∑ i = 1 k μ 55 Δ L N I N F t − i + ∑ i = 1 k μ 56 Δ L N G C F t − i + μ 57 E C T t − 1 + ϵ t (5)

where, μ’s capture the short-run effects of the explanatory variables on the dependent variable, E C T t − 1 captures the rate at which the dependent variable (GDP) adjusts to the equilibrium state after structural or institutional shocks that occur.

The VECM is based on the classical linear regression model that residuals are normally distributed, no autocorrelation on the residuals and absence of correlation among the explanatory variables.

The descriptive statistics shown in

It is necessary to identify the order of integration in time series data. So, to check the stationary of the data we use Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) unit root tests and their results are shown in

According to

After getting stationary of all the variables at first difference that is integrated in order 1 i.e. I (1); this study used the Johansen co-integration test to examine the long-run association between exports, imports and economic growth. In

Since the Johansen co-integration test suggest the long-run relationship among the variables under the study consideration. Now it is time to observe whether there any short-run deviation exists or not from the long-run equilibrium path. This study used the VECM framework to check the disequilibrium considering

GDP | EP | IMP | INF | GCF | |
---|---|---|---|---|---|

Mean | 5.103167 | 11.49278 | 17.83503 | 7.473568 | 22.18730 |

Median | 5.121278 | 11.43115 | 16.36958 | 7.040000 | 22.72000 |

Maximum | 7.284208 | 20.16159 | 27.94933 | 14.54500 | 30.51000 |

Minimum | 2.134328 | 3.396255 | 11.69775 | 1.908000 | 15.47000 |

Std. Dev. | 1.356727 | 5.154666 | 4.944116 | 3.037789 | 4.927237 |

LNGDP | LNEP | LNIMP | LNINF | LNGCF | |
---|---|---|---|---|---|

Mean | 28.54223 | 26.26442 | 26.78250 | 1.911709 | 3.074728 |

Median | 28.53325 | 26.39267 | 26.73924 | 1.951608 | 3.123246 |

Maximum | 30.61459 | 28.71987 | 29.01846 | 2.677247 | 3.418054 |

Minimum | 26.52502 | 23.49991 | 24.56718 | 0.646056 | 2.738903 |

Std. Dev. | 1.175922 | 1.659311 | 1.414400 | 0.486256 | 0.227572 |

Correlation | LNGDP | LNEP | LNIMP | LNINF | LNGCF |
---|---|---|---|---|---|

LNGDP | 1.000000 | ||||

LNEP | 0.994050 | 1.000000 | |||

LNIMP | 0.995451 | 0.995089 | 1.000000 | ||

LNINF | −0.287756 | −0.309800 | −0.258423 | 1.000000 | |

LNGCF | 0.951741 | 0.963223 | 0.962826 | −0.311092 | 1.000000 |

Variables | ADF | PP | Remarks | ||||
---|---|---|---|---|---|---|---|

At level | First difference | Lag: SIC Max | At level | First difference | Bandwidth (NW) | ||

LNGDP | 0.42 (0.981) | −4.22*** (0.002) | 9 | 0.31 (0.976) | −4.19*** (0.002) | AS | I(1) |

LNEP | −0.73 (0.826) | −8.72*** (0.000) | 9 | −0.92 (0.770) | −9.12*** (0.000) | AS | I(1) |

LNIMP | −0.14 (0.937) | −5.75*** (0.000) | 9 | −0.13 (0.939) | −5.75*** (0.000) | AS | I(1) |

LNINF | −2.04 (0.270) | −7.99*** (0.000) | 5 | −3.26 (0.024) | −9.80*** (0.000) | AS | I(1) |

LNGCF | −0.54 (0.871) | −4.02*** (0.004) | 9 | 0.19 (0.968) | −3.94*** (0.005) | AS | I(1) |

ADF: Augmented Dickey-Fuller; PP: Phillips-Perron; NW: Newey-West; AS: Automatic Selection; ***indicates 1% level of significance: I (1): Integrated in 1^{st} order level.

the rate of convergence to the long-run equilibrium path as well as to determine the granger cause of exports and imports on economic growth and other macroeconomic variables and vice versa.

Lag | LogL | LR | FPE | AIC | SC | HQ |
---|---|---|---|---|---|---|

0 | 29.07032 | NA | 1.67e−07 | −1.415901 | −1.191436 | −1.339352 |

1 | 225.0904 | 322.8566 | 7.27e−12 | −11.47590 | −10.12912* | −11.01661 |

2 | 251.7560 | 36.07697 | 7.30e−12 | −11.57388 | −9.104767 | −10.73184 |

3 | 293.4580 | 44.15508* | 3.62e−12* | −12.55635* | −8.964915 | −11.33157* |

*indicates lag order selected by the criterion; LR: sequential modified LR test statistic (each test at 5% level)

FPE: Final prediction error; AIC: Akaike information criterion; SC: Schwarz information criterion; HQ: Hannan-Quinn information criterion.

Hypothesized No. of CE(s) | Eigenvalue | Trace | Max-Eigen | ||||
---|---|---|---|---|---|---|---|

Statistic | 0.05 Critical Value | Prob.** | Statistic | 0.05 Critical Value | Prob.** | ||

None* | 0.8877 | 161.44 | 69.81 | 0.000 | 72.16 | 33.87687 | 0.0000 |

At most 1* | 0.7745 | 89.28 | 47.85 | 0.000 | 49.15 | 27.58434 | 0.0000 |

At most 2* | 0.5237 | 40.12 | 29.79 | 0.002 | 24.47 | 21.13162 | 0.0162 |

At most 3* | 0.2827 | 15.64 | 15.49 | 0.047 | 10.96 | 14.26460 | 0.1558 |

At most 4* | 0.1321 | 4.67 | 3.84 | 0.030 | 4.67 | 3.841466 | 0.0306 |

Trace test indicates 5 co-integrating eqn(s) at the 0.05 level; Max-eigenvalue test indicates 3 cointegrating eqn(s) at the 0.05 level; *denotes rejection of the hypothesis at the 0.05 level; **MacKinnon-Haug-Michelis [

This study also conducted diagnostic tests for every equation shown in

Short-run | ∆LNGDP | ∆ LNEP | ∆ LNIMP | ∆ LNINF | ∆ LNGCF | ||
---|---|---|---|---|---|---|---|

∆LNGDP | 0.33 (1.55) | 0.92** (−2.35) | 2.02 (0.58) | 0.19** (2.03) | |||

∆ LNEP | −0.09* (−1.71) | 0.08 (0.31) | 1.11 (0.81) | 0.02 (0.85) | |||

∆ LNIMP | −0.01 (−0.08) | −0.002 (−0.005) | −1.90 (−1.29) | −0.08** (−2.46) | |||

∆ LNINF | 0.05** (2.55) | 0.19* (1.77) | 0.16* (1.89) | 0.03*** (4.00) | |||

∆ LNGCF | 0.13 (0.33) | 1.86 (0.97) | −2.24* (−1.97) | 0.99 (0.17) | |||

Long-run | ECT01(−1) | −0.24* (−1.90) | −0.31 (−0.54) | −0.38 (−0.85) | −1.43* (−1.70) | −0.25*** (−4.24) | |

Diagnostic Tests | |||||||

Jarque-Bera Normality Test | 5.69 [0.060] | 0.67 [0.715] | 0.15 [0.929] | 0.22 [0.895] | 0.51 [0.775] | ||

Breusch-Godfrey Serial Correlation LM Test | 0.23 [0.873] | 1.49 [0.264] | 0.42 [0.743] | 0.12 [0.949] | 1.11 [0.382] | ||

ARCH Heteroskedasticity Test | 0.99 [0.412] | 0.82 [0.596] | 0.88 [0.464] | 0.22 [0.878] | 0.16 [0.920] | ||

Stability CUSUM Test and CUSUM of Square Test | Within the bands | Within the bands | Within the bands | Within the bands | Within the bands | ||

LM: Lagrange Multiplier; ARCH: Autoregressive Conditional Heteroscedasticity; CUSUM: Cumulative Sum; t-statistics in () and p-value in [ ]; ***represents 1%, **represents 5% and *represents 10% level of significance respectively.

stability test considering Cumulative Sum (CUSUM) of Recursive Residuals and Cumulative Sum (CUSUM) of Squares of Recursive Residuals confirm that all the equations of VECM framework are stable in 5% level of significance.

The environment of international trade is changing day by day because of globalization effect. So, economists, researchers and policy makers need to revisit the relationship between exports, imports and economic growth again and again. Considering the importance, this study examines the long-run causal relationship among exports, imports and economic growth in Bangladesh using annual time series data over the period of 1981 to 2017. The stationary (has no unit root) of the data is checked using ADF and PP tests. To find the long-run association among the variables, the Johansen co-integration test is used, and to find the short-run dynamics and granger causality between variables, VECM framework is used. Johansen co-integration test finds that exports, imports and economic growth are co-integrated. VECM results demonstrate that there is a long-run equilibrium relationship among the variables and a unidirectional causality between the export and economic growth in the short run. Further, no strong evidence is found in short-run that import causes economic growth. Meanwhile, it has been found that economic growth granger causes import in short-run. Moreover, major implication of our findings is that export is matter for the economic growth of Bangladesh than import. Finally, findings of the study will help the policy makers and development partners of Bangladesh to rethink about the current policies regarding the exports, imports, inflation, gross capital formation (investment) and economic growth.

The authors declare no conflicts of interest regarding the publication of this paper.

Miyan, Md.S. and Biplob, Md.N.K. (2019) Revisiting Exports, Imports and Economic Growth Nexus: Empirical Evidence from Bangladesh (1981-2017). Modern Economy, 10, 523-536. https://doi.org/10.4236/me.2019.102036