^{1}

^{*}

^{1}

^{2}

This paper investigates the factors that affect Egypt’s bilateral export flows to its main trading partners. Based on the panel data, the gravity model approach has been used to estimate Egypt’s exports through annual data covering the period 2000 to 2013 for 42 main trading partners. The gravity model in its fixed effects panel data explained 84 percent of the fluctuations in Egypt’s exports. The results show that Egypt’s GDP, importer’s GDP, importer’s population, regional trade agreements (RTA) and the border between Egypt and its trading partner are the main factors affecting Egypt’s exports to its main trading partners. All these factors affect Egypt’s exports positively. Transportation costs (distance variable) are found to have negative but insignificant effect on Egypt’s exports. All these results can help the government and policy makers to undertake appropriate measures to improve the performance of the Egyptian foreign trade sector.

Exports of goods and services represent one of the most important sources of foreign exchange income that ease the pressure on the balance of payments and create employment opportunities [

Egypt is one of the emerging economies in Africa and Middle East region. Egypt’s foreign trade sector constitutes an important part of its economy. The trade openness ratio increased from 39 percent in 2000 to 71.7 percent in 2007 (World Bank). Although this sector has a great importance in the Egyptian economy, it has been suffering from a chronic deficit over the last years. Egypt’s export sector accounts for only slightly more than 10 percent of the overall economy, and this relatively modest contribution cannot by itself end high unemployment or generate enough tax revenue to fix the government’s shattered finances.

Egypt is in the early stages of an export boom, suggesting its economy could begin to recover in the next few months if a minimum level of political stability is restored. Egypt’s total exports has increased from US$ 3.5 billion in 1994 to US$ 28.78 billion in 2013 and exports in gross domestic product (GDP) also increased from 5 to 10.6 percent during the same period, thus reflecting modest importance of exports sector in the economy. However, the Egyptian exports are not widely diversified, but dominated by few products such as chemicals and agricultural products that lead to higher concentration in export products, and in turn to highly concentrated export destinations.

Given this importance and the role foreign trade sector plays in the Egyptian economy, it is important to find out the economic factors influencing Egyptian exports in order to help the government and policy makers to undertake appropriate measures to improve the performance of the foreign trade sector.

Gravity model through the panel data estimation technique is a useful methodological tool in finding out the economic factors that determine trade or export of a country. This model originates from the Newtonian law of gravity through the attraction of two countries’ masses, weakened by distance between them. The gravity model of trade basically states that trade flows between two countries are determined positively by their income and negatively by the distance between them. This model was originally employed by [

Although few literatures are found with regard to analyzing trade relations and trade policies of Egypt using the gravity model, there is no particular in depth study found out the determinants of Egypt’s total export trade. This study, therefore, is an attempt to fill up this research gap.

To find out the most relevant factors affecting Egypt’s exports in the global market during the period 2000- 2013, the paper used panel data estimation technique and gravity model. The main contribution of this study is that it applies for the first time panel data approach in a gravity model framework to identify the factors influencing Egypt’s total exports.

The rest part of the paper is organized as follows. Section 2 provides an overview of Egypt’s exports and its main trading partner. Section 3 discusses the gravity model methodology and data. Section 4 presents the results of the empirical analysis, and finally conclusion and policy recommendations are briefed at the end.

As Egypt is one of the emerging economies in Africa and Middle East region, it is important to investigate the performance of Egypt’s total exports. During the period 1980-2004, Egypt’s exports increased substantially and made a major contribution to Egypt’s balance of trade account. Total exports in 1980 were around US$ 3 billion and more than doubled in value to US$ 7.7 billion by 2004, growing at an average annual rate of approximately 6 percent. Since then Egyptian exports have increased enormously and reached US$ 27.88 billion in 2013. This represents a 26.33 percent average growth rate between 2004 and 2013 (see

In terms of directions of Egypt’s exports, most have been highly concentrated among a few major trading partners. During 2000-2013, about one-half of total exports went to the top five trading partners, and sixty percent to the top 10 markets (

2000 | 2005 | 2008 | 2010 | 2011 | 2012 | 2013 | |
---|---|---|---|---|---|---|---|

EU-27 | 1919 | 3627 | 9211 | 7990 | 9680 | 7871 | 8095 |

India | 156 | 528 | 1659 | 1228 | 2265 | 2039 | 2135 |

Saudi Arabia | 140 | 381 | 1239 | 1514 | 1964 | 1816 | 1966 |

Turkey | 84 | 334 | 770 | 985 | 1526 | 1563 | 1744 |

Libya | 62 | 150 | 774 | 1220 | 557 | 1439 | 1277 |

USA | 400 | 957 | 1255 | 1547 | 1819 | 2015 | 1182 |

Jordan | 19 | 196 | 713 | 669 | 864 | 695 | 852 |

UAE | 59 | 297 | 517 | 501 | 755 | 686 | 750 |

Iraq | 63 | 56 | 335 | 383 | 450 | 570 | 736 |

Rep. of Korea | 54 | 138 | 631 | 531 | 434 | 279 | 710 |

Other | 1737 | 3967 | 8848 | 8730 | 9553 | 9126 | 8430 |

Total | 4693 | 10632 | 25952 | 25298 | 29867 | 28098 | 27878 |

Source: World Integrated Trade Solution.

Regarding the economic literature, there are a lot of empirical approaches which have used the gravity model to explain the determinants of bilateral trade. Using a gravity model applied to panel data, researcher [

The gravity model has been used extensively in the international trade literature over the past 50 years [

where X_{ij} is the exports flow between country i and j, α_{k} is a constant, β, γ, ξ, ε, μ are coefficients, weighted geometric averages, Y_{i} and Y_{j} are gross domestic product (GDP) in country i and j, respectively, N_{i} and N_{j} are population in country i and j, and D_{ij} is the distance between the countries’ capitals. The U_{ij} is a lognormal distributed error term with E(lnU_{ij}) = 0. Frequently dummy variables are also included in the model to take into account preferential trade factors between i and j.

The linear form of the previous gravity equation can be written as follows:

The standard gravity model includes distance and income as independent variables. Most models also include population and different dummy variables [

where all variables (excluding dummies) are expressed in natural logarithms. The dependent variable

Each of the independent variables included in Equation (3) has a defined impact on the level of trade flows between Egypt and its trading partners. A high level of income in the exporting country indicates a high level of production, which increases the availability of goods for exports [_{1} will be positive. By the same logic, the expected sign of the coefficient of β_{2} is also expected to be positive where a high

Variable | Description | Expected sign |
---|---|---|

GDP_{i} | Egypt’s gross domestic product (constant 2005 US$) | + |

GDP_{j} | Trading partner’s gross domestic product (constant 2005 US$) | + |

POP_{j} | Trading partner’s population | +/− |

OPN_{j} | Trading partner’s openness rate (total trade/real GDP) | + |

DIS_{ij} | Distance between Cairo and trading partner’s capital | − |

LANG_{ij} | Dummy variable takes the value 1 when Egypt and its trading partner speak Arabic language, and the value 0 otherwise | + |

RTA_{ij} | Dummy variable takes the value 1 when Egypt and its trading partner signed regional or free trade agreement, and the value 0 otherwise | + |

BORDER_{ij} | Dummy variable takes the value 1 when Egypt and its trading partner share common border, and the value 0 otherwise | + |

Variable | Obs. | Mean | Std. Dev. | Min. | Max. |
---|---|---|---|---|---|

ln(X_{ij}) | 588 | 18.67142 | 1.530177 | 13.28538 | 21.72368 |

ln(GDP_{i}) | 588 | 25.32067 | 0.187223 | 25.04612 | 25.57948 |

ln(GDP_{j}) | 588 | 26.12570 | 1.809263 | 22.45182 | 30.30507 |

ln(POP_{j}) | 588 | 17.06198 | 1.759187 | 12.85151 | 21.02882 |

ln(OPN_{j}) | 588 | 4.280725 | 0.547893 | 2.755570 | 6.086002 |

ln(DIS_{ij}) | 588 | 7.504116 | 0.808463 | 5.723585 | 8.723069 |

LAN_{ij} | 588 | 0.333333 | 0.471806 | 0.000000 | 1.000000 |

RTA_{ij} | 588 | 0.598639 | 0.490591 | 0.000000 | 1.000000 |

BORDER_{ij} | 588 | 0.023810 | 0.152585 | 0.000000 | 1.000000 |

level of income in the importing country leads to higher imports. The coefficient estimate for population of the importer β_{3}, may be positive or negative signed, depending on whether there is more opportunity for trade in a large variety of goods or not. As concern to openness, the more open the country is, the more would be the trade [_{4} will be positive. The expected sign of β_{5} is expected to be negative, where a long distance between trading partners implies higher costs and lower profit margins for importers [_{6}, β_{7} and β_{8} are expected to have a positive sign (see the summary of expected signs in

The research covers a total of 42 importing markets. These markets are chosen based on the importance of trading partnership with Egypt and the availability of required data. Eleven markets from EU (Belgium, Cyprus, Germany, Spain, France, United Kingdom, Greece, Italy, Malta, Netherlands and Portugal), fourteen Arab markets (Algeria, Iraq, Jordan, Kuwait, Lebanon, Libya, Morocco, Oman, Qatar, Saudi Arabia, Syria, Tunisia, United Arab Emirates and Yemen), eight markets from non-Asian Arab countries (China, Indonesia, India, Japan, Korea Rep., Malaysia, Pakistan and Singapore) and other nine markets (Brazil, Canada, Switzerland, Kenya, Nigeria, Russian Federation, Turkey, USA and South Africa) are included in our sample for the analysis of Egyptian exports during the period 2000-2013 (14 years). On the whole, the sample markets absorb at least 85% of Egypt’s total exports. The estimation covers 42 countries over the years 2000-2013 with one dependant variable and 8 explanatory variables (a total of observations = 588, N = 42, and T = 14), and all variables are expressed in natural logarithm. The data on Egypt’s total exports to its trading partners (country j) are obtained from the World Integrated Trade Solution (WITS) [

Many early empirical studies used classical gravity models (cross-section data) to estimate trade effects and trade relationships for a particular time such as models used by [

Panel data contains many models that can be estimated. These models are pooled, fixed effects and random effects. The main problem of the pooled model is that it does not allow for heterogeneity of countries. It does not estimate country specific effects and assumes that all countries are homogenous [^{2}) is significant, less than 0.05, then fixed effects model will be used. If P-value is greater than 0.05, then random effects model will be the most efficient model. The fixed effects model has a problem in estimating variables that do not change over time because inherent transformation combs out such variables, and as such the dummy and distance variables need to be dropped. This problem can be solved by estimating these variables in a second step through running a second stage regression with considering the individual effects as the dependent variable whereas the dummy and distance as independent variables. The estimated equation of the second stage regression will be as follows:

where IE_{ij} represents the individual effects and the other explanatory variables are as explained before.

Before estimating Equation (3), this paper analyzes the univariate characteristics of the data that entails panel unit root tests. Panel unit root test determines a potentially cointegrated relationship between the variables. If all variables have no unit root i.e. are stationary, then the traditional estimation methods can be used to estimate the relationship between the variables. If the variables have unit root test i.e. are non-stationary, a test for cointegration will be performed [

Variable | Hadri test | LLC test |
---|---|---|

Null: No unit root | Null: Unit root | |

X_{ij} | 14.7744 (0.0000)^{***} | 0.49518 (0.6898) |

GDP_{i} | 17.1184 (0.0000)^{***} | −4.59292 (0.0000)^{***} |

GDP_{j} | 17.0063 (0.0000)^{***} | 1.24253 (0.8930) |

POP_{j} | 17.9277 (0.0000)^{***} | −47.63495 (0.0000)^{***} |

OPN_{j} | 11.4271 (0.0000)^{***} | −43.54018 (0.0002)^{***} |

Note: ^{*/**/***}Denote rejection of the null hypothesis at 10%/5%/1% significance level.

Variables | Pooled model | Fixed effects model | Random effects model |
---|---|---|---|

Constant | −97.44348 (−19.58680)^{***} | −113.7947 (−29.15961)^{***} | −100.6475 (−25.81471)^{***} |

ln(GDP_{i}) | 4.073998 (20.38980)^{***} | 4.012909 (18.45325)^{***} | 4.277310 (26.11997)^{***} |

ln(GDP_{j}) | 0.498568 (15.82329)^{***} | 0.845589 (3.524467)^{***} | 0.562705 (6.849754)^{***} |

ln(POP_{j}) | 0.284574 (7.945792)^{***} | 0.528146 (1.738172)^{*} | 0.193165 (2.303473)^{**} |

ln(OPN_{j}) | 0.456490 (5.375225)^{***} | −0.057525 (−0.394106) | 0.097861 (0.767610) |

ln(DIS_{ij}) | −0.962617 (−15.53296)^{***} | - | −1.017451 (−6.048798)^{***} |

LANG_{ij} | 0.215828 (2.072738)^{**} | - | 0.392612 (1.416076) |

RTA_{ij} | 0.412817 (4.172674)^{***} | - | 0.126173 (1.232739) |

BORDER_{ij} | 1.180999 (4.989782)^{***} | - | 1.155609 (1.700600)^{*} |

No. of observations | 588 | 588 | 588 |

Adjusted R^{2} | 0.69 | 0.84 | 0.70 |

F-test | - | 340.05^{***} | - |

Hausman test | - | 14.38^{***} | - |

Notes: ^{*/**/***/}Significant at 10%/5%/1% level. The t-statistics are in parentheses.

not allow for heterogeneity of cross sections and cross section specific effects are not estimated. In addition to these problems, the pooled model assumes that all cross sections are homogenous and it assumes a single intercept and same parameters over time and cross countries. The results of pooled panel data model are in the second column of

The results for random effect model are presents in the last column of

In order to test the null hypothesis that the regressors and individual effects are not correlated, the Hausman test has been performed. This test can distinguish between fixed and random effects. Failure to reject the null hypothesis implies that the random effect model is efficient than the fixed effect model. The fixed effect model will be more efficient than the random effect model, only if the null hypothesis has been rejected.

^{2} values of 84 percent in case of fixed effects model. Almost all estimates of the coefficients are highly significant with expected signs. This indicates that the gravity model is suitable for explaining the flows of Egyptian exports to its major trading partners. As can be seen from ^{2} indicates that approximate 84 percent of the variability in the total exports between Egypt and its trading partners can be explained by the fixed effects model. The value of F-test indicates that the overall significance of the model is highly significant at 1 percent level. The explanatory variables product of GDPs are highly significant at 1 percent level, the explanatory variable importer’s population is weak significant at 10 percent level. The variable of importer’s openness is not significant factor in explaining Egypt’s exports. The results of fixed effects model in

The estimated coefficient of Egypt’s GDP is 4.01 which means that, holding constant for other variables, a 1 percent point increases in Egypt’s GDP will result in a roughly 4 percent point increase in Egypt’s total exports. On the other hand, the estimated coefficient of importer’s GDP is 0.84 which means that, holding constant for other variables, a 1 percent point increases in importer’s GDP will result in a roughly 0.84 percent point increase in Egypt’s total exports. Moreover, the estimated coefficient of importer’s population is 0.53 which means that, holding constant for other variables, a 1 percent point increases in importer’s population will result in a roughly 0.53 percent point increase in Egypt’s total exports. Importer’s openness does not show significant coefficient, and thus is not explanatory variables in the flow of Egyptian exports.

The results of second stage regression are presented in

Given the importance and the role foreign trade sector plays in the Egyptian economy, it is important to find out the economic factors influencing Egyptian exports in order to help the government and policy makers to undertake appropriate measures to improve the performance of the foreign trade sector.

The main purpose of this research was to identify and evaluate the main factors that affect Egyptian exports. The analytical procedure includes the gravity model as one of the best theoretical framework to estimate export equations using panel data for the period 2000 to 2013. A sample of 42 was chosen for the empirical analysis.

Regression analysis was performed in three techniques which include the pooled model (common intercept model), the fixed effects model, and the random effects model. In order to choose among methods, the F-test and Hausman test were performed.

Independent variable | Coefficient (t-statistics) |
---|---|

Constant | 19.34994 (25.84944)^{***} |

Distance | −0.142029 (−1.534736) |

Common language | −0.793246 (−4.884456)^{***} |

RTA | 1.057587 (7.064632)^{***} |

Common border | 0.780418 (1.937841)^{**} |

Adjusted R^{2} | 0.100 |

Note: ^{*/**/***/}Significant at 10%/5%/1% level.

The F-test rejected the null hypothesis of common intercept. When choosing between fixed and random effects, the Hausman test rejected the null hypothesis. The research demonstrated that the fixed effects model generated the most reliable outcomes and then showed the analysis using this method. The results of the regression indicated that most estimated parameters had the expected signs and were statistically significant.

Egypt’s GDP, GDP of the importing countries, population of the importing countries had a positive influence on Egypt’s exports. The results suggested a positive influence of the increase in Egypt’s GDP, since a greater income stimulated exports. The higher the income, the greater the volume of Egyptian exports which took place. The results also suggested that Egyptian exports was positively influenced by an increase in GDP in the importing countries; the higher the income in importing country, the greater the volume of Egyptian exports. The openness rate of the importing countries did not encourage Egypt’s exports, having a negative influence.

The distance variable between Egypt and the importing countries had its expected sign but it was not significant. The results also indicated that Egypt exports less products to countries where Arabic is the official language, and this is not consistent with theoretical expectation. The regional trading agreements (RTA) variable is highly significant and this implies that the importer country that has RTA with Egypt can affect Egypt’s exports. The common border between Egypt and its neighbors encouraged the exports, where the results indicated the significance and positive effect of this variable on the flows of Egyptian exports.

All these results can help the government and policy makers to undertake appropriate measures to improve the performance of the Egyptian foreign trade sector. However all these results are valuable, more researches and more data on Egypt’s trade will add more and more to the foreign trade sector.