Trade liberalization under the Barcelona Initiative was inspired by standard results in trade theory which state that trade is mutually beneficial to trading partners. Computable General Equilibrium (CGE) models had been used in the late 1990s and early 2000s to assess from an ex ante perspective welfare benefits and underlying macroeconomic effects of decreases in tariff rates for the Southern Mediterranean Partner Countries (SMPC).

This study uses econometric methods to evaluate the macroeconomic effects ex post. In part I we use panel econometric methods to assess the effects of trade liberalization on the most important macroeconomic aggregates for a set of arabic SMPCs (Algeria, Egypt, Jordan, Lebanon, Morocco, Syria, Tunisia). This choice was motivated by the requirement that countries participating in the panel should be fairly homogenous, so that the cross section information can step in for the rather limited time series information on SMPCs.

Essentially, we use growth regressions, i.e. we regress the growth rates of a variable on its lagged level, a trend, a measure of aggregate external protection and further conditioning variables. In some cases we allow coefficients to be cross section specific or to differ for different groups of countries, e. g. net oil exporters versus oil importing countries.

Based on the theoretical analysis of a standard Ramsey economy we distinguish between announcement and implementation effects of trade liberalization. The announcement effect is identified as a binding commitment of a country to decrease its tariff rates over time in a well-defined stepwise fashion. We capture the announcement effect by regressing on future tariff rates. The implementation effect, by contrast, is captured by regressing on lagged tariff rates, as this would express how macro variables react to actual changes in the level of external protection.

We find significant and robust evidence for a semi-elasticity of GDP with respect to implemented tariff rate decreases of about 0.3 We also find significant evidence of a non-zero semi-elasticity for announced tariff rate decreases for some of the countries under investigation. We find that consumption, investment and imports react positively to tariff liberalization, while government expenditure shows no response. We present evidence that the impetus of trade liberalization on GDP operates through a stimulation of private consumption and supply side effects through increased capital accumulation.

In part II of the report we survey 25 studies that used a Computable General Equilibrium (CGE) model in order to assess the macroeconomic effects of trade liberalization ex ante. The studies we survey were performed in the years 1993-2007 and explored the effects in 8 Mediterranean countries: Tunisia, Morocco, Egypt, Jordan, Syria, Turkey, Israel and Malta. Each study checked a number of macroeconomic effects such as GDP, consumption, investment and imports.

The surveyed studies all employed a Computable General Equilibrium model. Although the basic framework of the CGE model is the same, each study used different economic assumptions regarding temporality (static/dynamic model), perfect or non-perfect competition, production, consumption, trade and government behaviour. In addition each study used a different set of data sources, most of the studies used the examined country's Statistical Accounting Matrix (SAM) but some used other data sources. For each study the basic structure of the model, the assumptions and the data sources used are reported.

For each study we report the results of the different macroeconomic effects assessed in it. The majority of the studies report the GDP growth (welfare change) effects assessment. A half of the studies report the Import and Export results and a quarter of them report the effects on Investment and Consumption. For each surveyed study we find the average reduction of tariffs. By dividing the macroeconomic change in the tariff reduction we compute the semi-elasticity of each macroeconomic effect. We find that the average GDP semi-elasticity in all the reviewed studies is 0.112, meaning that for each percent reduction in tariffs the GDP will grow by 0.112 percent. The average Imports semi-elasticity in the studies is 0.376 and for Exports the semi-elasticity is 0.288. The Investment and Consumption are also affected positively by tariff reduction and their average semi-elasticity is 0.116 and 0.176 respectively.

In part III we contrast the econometric (ex post) results with the ex ante CGE results. We argue that it is scientifically unwarranted to shield CGE projections from ex post critique by pointing at the ceteris paribus assumption under which these projections were derived, because this would imply that CGE projections can never be falsified. In order to subject CGE projections to data-based evaluation and critique while simultaneously taking seriously the ceteris paribus property of its methodology, we first postulate that CGE projections should be correct on average, provided unrelated shocks have an expected value of zero. We then argue that any systematic unrelated changes should (and, in general, will) be captured in the econometric setup used to find the ex post correlates of CGE projections, because any sound econometric estimate will have white noise residuals. Hence, the econometric estimates control for unrelated developments and identify, e. g. as semi-elasticity with respect to tariff rates, the isolated effect in the same spirit as CGE projections do.

We then study the distribution of deviations between ex ante and ex post measures of the macroeconomic effects of trade liberalization. Our sample consists of some 20 published CGE studies aimed at arabic SMPCs and intended to quantify the likely effects of the Barcelona Initiative. We properly distinguish between studies using a base year prior to the signing of the Association Agreement and base years at or after the point in time when the country committed to decreasing its external protection within the next twelve years or so. The latter type of data set is assumed to already contain the announcement effect, while the former is not.

We find that CGE studies, rather than being unbiased estimators of the true effects, generally underestimate the econometrically found semi-elasticities by an order of magnitude of 0.3. This is bias is very large as it is about the same size as the implementation semi-elasticity itself! We also find that confidence intervals (based on the standardized deviation from the econometric estimate) for CGE-derived semi-elasticities are so wide that CGE projections are almost meaningless.

Finally, we analyse the characteristic features of CGE models in our sample. Essentially, we perform an analysis of variance (ANOVA) by regressing the observed deviations on dummy variables which capture basic properties of the underlying CGE model, e. g. dynamic/recursive/static, full employment/labor market disequilibrium, fixed/flexible exchange rate, perfect/imperfect competition, balanced/unbalanced trade etc. We present evidence which suggests that success or failure of CGE projections seems to be largely independent of modeling choices and is possibly just a matter of good or bad luck. Essentially, the result of this exercise is limited to the finding that models which assume full employment fare better than models with unemployment and that most CGE models fail to capture the announcement effect properly.

In addition to the report there are 2 annexes. Annex I is a macro economic overview of the sampled countries. This overview reports the significant macro economic changes in the sampled countries in the last years. The main focus of the country overview is foreign trade in general and trade with the EU in particular. In Annex II we provide a special economic report on 2 main Mediterranean countries not in the sample – Israel and Turkey. The special report reviews the main characteristics of the country's economy such as foreign trade, labour market, economic development, tax policies and such.