The Determinants of Intra-Oceanian Imports from 2001 to 2015: A Panel Gravity Model Approach

ABSTRACT

This article aims at identifying the main determinants of annual bilateral imports flows amongst 23 Oceanian territories from 2001 to 2015. The analysis quantifies the impact of adherence to the World Trade Organization and Pacific Islands Countries Trade Agreement on the import flows. When adherence to only one trade organization is considered, estimates establish a discrepancy between the official aims of trade agreements and their effective impact on bilateral imports flows. Adherence to only one trade agreement has a negative impact. However, when adherence to both trade agreements is jointly considered, estimates establish a positive impact from agreement adherence.

KEYWORDS:

• gravity models
• imports
• Oceania
• trade agreements

Disclosure statement

No potential conflict of interest was reported by the author.

Supplementary material

Supplemental data for this article can be accessed here.

Notes

¹ Their analysis is focused on the impact of common currency arrangements on the volume of trade.

² The Human Development Index (HDI) is calculated by the United Nations Development Programme only for United Nations members. This implies that this index is not available for several territories amongst the 23 considered here. HDI data for non-UN members have been collected. One should be very cautious when comparing these data because they are not associated with the same time period as this study (i.e., observations extend from 2005 to 2015). This implies the presence of a methodological bias because a country observed in 2015 should be more developed than it was five or more years earlier. This explains why no ranking has been associated with any data observed before 2015. See Table 1 for details on sources.

³ In 2017, the Pacific Islands Forum included 18 members: Australia, Cook Islands, Federated States of Micronesia, Fiji, French Polynesia, Kiribati, Nauru, New Caledonia, New Zealand, Niue, Palau, Papua New Guinea, Republic of the Marshall Islands, Samoa, Solomon Islands, Tonga, Tuvalu, and Vanuatu. Additionally, Tokelau has been an associate member since 2014, while Wallis and Futuna, American Samoa, and Guam – amongst others – are registered as observers.

⁴ https://www.un.org/Depts/los/general_assembly/contributions_2015/PIFS.pdf.

⁵ http://www.forumsec.org/resources/uploads/attachments/documents/PACER.pdf.

⁶ See Table A1 in the online appendix. The appendix can be found online at www.tandfonline.com/uitj.

⁷ The model considered here is, therefore, not affected by the “silver medal mistake” that arises when one considers the average of bi-directional bilateral trade before applying the logarithm operator (Baldwin and Taglioni 2006, 9–10). Such a method leads to biased estimates.

⁸ This hypothesis is also considered in Mölders and Ulrich (2011, 437).

⁹ The database used here consists of 23 countries and 15 periods which leads to 690 dummies (i.e., 23 * 15 * 2).

¹⁰ The measure of the internal distance rests on data collected from the GeoDist database published by the CEPII. This database is available at http://www.cepii.fr/distance/geo_cepii.xls. The $d_{ij}$ variable has been introduced through two measures of distance: the distance between two countries measured from their respective capital cities and the distance between the two countries measured in kilometers.

¹¹ Trade resistance could also be taken into consideration through the restrictiveness of Rule of Origin.

¹² https://www.wto.org/english/thewto_e/whatis_e/what_stand_for_e.htm.

¹³ The document is available at https://wits.worldbank.org/GPTAD/PDF/archive/picta.pdf.

¹⁴ The fixed effects approach suffers from a severe limitation because the presence of fixed effects is incompatible with any time-invariant variable in the model to be estimated. The random effects approach does not suffer the same limitations. However, a common point between these two estimation methods is the necessity to linearize the gravity model through the logarithm operator. The problem is that this method cannot be implemented with null trade flows because the logarithm operator is only defined for strictly positive values. Therefore, a classical estimation method through fixed effects or random effects is not optimal with numerous zero trade data.

¹⁵ Recall that this index corresponds to a country’s average weighted distance from its trading partners, with the partner’s countries’ shares of world GDP used as weights (Head 2003, 8).

¹⁶ Estimated coefficients for time dummies are not detailed.

¹⁷ For example, “Limports” indicates that the “Imports” variable is considered in logarithm form.

¹⁸ This reasoning rests on the existence of a positive correlation between the intensity of intra-industry trade between two countries and the similarity of their relative factor endowments (Krugman 1981).

¹⁹ The Relationship (2) given in Section 3 for the remoteness index implies that $\left(\mathrm{\partial }{\mathrm{R}}_{\mathrm{i}}/\mathrm{\partial }{\mathrm{d}}_{\mathrm{i}\mathrm{j}}\right)$ < 0. Given that $\left(\mathrm{\partial }{\mathrm{M}}_{\mathrm{i}\mathrm{j}}/\mathrm{\partial }{\mathrm{d}}_{\mathrm{i}\mathrm{j}}\right)$ < 0, it implies that $\left(\mathrm{\partial }{\mathrm{M}}_{\mathrm{i}\mathrm{j}}/\mathrm{\partial }{\mathrm{R}}_{\mathrm{i}}\right)$ > 0.

²⁰ Estimated coefficients for these annual dummies are not presented.

²¹ This result contrasts with Model (1) where the estimated coefficient is negative.

²² This interpretation rests on the rejection of the hypothesis that “D(Imports)” does not Granger-cause “Fdi_imp”.

²³ More details are given in Section 2.

²⁴ Australia and New Zealand are excluded here because both countries are not allowed to adhere to PICTA.

²⁵ Rose (2002, 11) obtains negative and insignificant coefficients for the dummy variables for one or both of the countries who are GATT/WTO members. He does not give any interpretation to these results and prefers to consider them as a mystery.

²⁶ Australia and New Zealand are ignored because they are not allowed to adhere to PICTA.

²⁷ In September 2016, French Polynesia and New Caledonia were granted full membership to PICTA. Even if they do not officially pertain to the WTO, their strong political, historical, and economical links with France imply their informal adherence to the WTO. Thus, these two territories can be considered as adherents to PICTA and the WTO as well. This statement does not contradict our previous result given the level of the average of their nominal GDP in the 2001 to 2015 period: New Caledonia has the second highest level of average nominal GDP while French Polynesia occupies the third rank. To sum up, amongst the nine most economically advanced territories in terms of average nominal GDP, eight have formally or informally adhered to PICTA and the WTO’s rules.

²⁸ The binary variables involving PICTA are equal to 1 for years where this trade agreement is in force. The beginning of the trade agreement is not the year of its official signature but the year of its entering into force. See Table A1 in the online appendix.

²⁹ When the “Picta_wto_one” dummy is corrected so as to take into account the hypothesis that French Polynesia, New Caledonia, and Wallis and Futuna are members of the WTO, the coefficient associated with the “Picta_wto_one” dummy is positive and significant in Model (12).

³⁰ Estimated coefficients for these annual dummies are not presented.

³¹ Most estimated coefficients are significant, and RESET tests indicate that the specifications are correct. These elements suffice to meet the target, i.e., to assess the impact of FTAs on bilateral import flows.