Understanding Challenges to Food Security in Dry Arab Micro-States: Evidence from Qatari Micro-Data

At the time of this research was undertaken, David Raboy was seconded as Chief Economist to the Qatar National Food Security Programme.

Streeten(1993) considered a country to be “small” if its population were less than 10 million people. Save for the UAE, the Arab States we consider have populations much smaller, in the 2–3 million range and therefore we refer to them as micro-states.

The collection of papers in Robinson (1960) remains a classic contribution to the theoretical analysis of micro-states. The theory of micro-states stresses the link between small size and suboptimality in terms of the minimum efficient scale of output, competition and efficiency. For recent empirical analyses on the economic performance of micro-states, see, e.g., Armstrong and Read (1995) and Easterly and Kraay (2000).

For example, the global food crisis of 2007–2008 triggered a ban on rice exports by India, which traditionally supplies over one-half of the rice consumed in the GCC countries (Shah 2010). As documented by Martin and Anderson (2010), export restrictions may have contributed as much as 35% to world rice prices and 25% to wheat prices during the 2007–2008 global food crisis.

The Economic Agreement Between GCC States, The Cooperation Council for the Arab States of the Gulf (GCC) Secretariat General, GCC Supreme Council, 22nd Session, 31 December, 2001, Muscat, Oman.

This was confirmed in a 2010 meeting with Ahmad Ali Mohammad Al-Muhannadi, chairman of the Customs & Ports General Authority, Qatar.

Ibid.

In particular, we employed two commonly used methods of detecting outliers in multivariate data. These are the (i) minimum covariance determinant estimator of location and scatter introduced by (Rousseeuw 1985) and updated by Rousseeuw and Driessen (1999) and (ii) the blocked adaptive computationally efficient outlier nominators (BACON) algorithm proposed by Billor, Hadi, and Velleman et al. (2000). Both the estimators perform relatively fast and allow one to quickly identify outliers, even on large datasets of tens of thousands of observations. Both estimators are available to implement as STATA commands.

Most of the information was obtained from World Shipping Register (www.e-ships.net), a leading company providing ships and shipping companies’ data with an outstanding number of ships, and shipping companies and detailed per-ship characteristics. Information on all other ships not covered by the World Shipping Register was obtained from personal visits to Customs & Ports General Authority, Doha, Qatar.

It is worth mentioning here that the incentive to under- or over-invoicing imports to reduce taxes or other duties is not relevant to the present context since most food items are exempted from import duties or are subject to low tariffs (maximum 5%) on certain products (meat, dairy) from specific locations (non-Middle Eastern countries such as Brazil or the USA). Furthermore, hitherto no national general sales tax exists in Qatar.

The country specific 12-month food price pass-through coefficients are Bahrain (0.349%), Kuwait (0.279%), Qatar (0.355%) and the UAE (0.413%).

Such imports are subject to strict Halal requirements in addition to SPS rules.

These include meat, dairy, fruits, vegetables and cereal commodities.

We thank the anonymous referee to bring this interpretation to our attention.

The n-firm concentration ratio, , is a concentration indicator alternative to HHI, with n typically equal to 4 and 8. However, unlike the HHI, the is not sensitive to the distribution of market shares among firms (Scherer and Ross 1990).

All unreported summary statistics are available from the corresponding author on request.

An alternative method for coding categorical variable is “orthogonal coding”. Pedhazur (1997, 340–409) demonstrates that the end results of multiple regression analyses of the same data coded by simple dummy coding, effect coding and orthogonal coding are identical. However, orthogonal coding permits the researchers more flexibility to examine more than simple omnibus tests, such as the complex coding type of contrast coding when a researcher wants to explore specific theory driven hypotheses about group differences among categories of a predictor variable (see Davis 2010). We thank the anonymous referee to bring this issue to our attention.