Full article: Untapped regional integration potential: A global frontier analysis

ABSTRACT
Formulae display:?Mathematical formulae have been encoded as MathML and are displayed in this HTML version using MathJax in order to improve their display. Uncheck the box to turn MathJax off. This feature requires Javascript. Click on a formula to zoom.

ABSTRACT

This paper proposes a novel approach to estimate untapped regional integration potential across geographical subregions of the world. We first construct an empirical production possibility frontier for regional integration outcomes based on two composite indices capturing enabling factors and achieved levels of regional integration outcomes across various domains. We then use non-parametric frontier analysis to rate the performance of subregions in terms of integration relative to their estimated potential. The obtained efficiency scores allow us to quantify and compare the empirical magnitudes of untapped integration potential across individual subregions. Our results suggest that, globally, subregional integration levels are currently at 60 percent of the estimated potential on average. Furthermore, there is large variation in integration outcomes, and subregions with large untapped integration potential are spread across all parts of the world. We also demonstrate how, from a policy perspective, the proposed method can be used for assessing achievements in targeting subregions with certain needs and characteristics, and guiding decisions about types of interventions that future programs aimed at fostering higher integration should prioritize in each subregion.

KEYWORDS:

Regional integration frontier composite index data envelopment analysis

JEL CLASSIFICATIONS:

F02 F13 F15 F60

1. Introduction

Regional integration has long been viewed as a key strategy for reducing conflict and fostering economic growth. ¹ Despite historically varying degrees of success and the recent resurgence of nationalistic rhetoric and protectionism in some parts of the world, policymakers continually argue that deeper regional integration constitutes a crucial factor in creating an environment conducive to sustainable economic development, peace, and security (UNESCAP 2016; ACET 2017; EU 2018; UNECA 2018).

Despite the prominence of efforts aimed at increasing regional integration on the current agenda of many international development organizations, there is very little empirical evidence that allows policymakers to quantify and compare integration levels across different regions and subregions, and map achieved progress against stated goals. ² While the economic literature comprises many studies providing detailed insights on single dimension of regional integration, e.g. studies focusing exclusively on trade or migration, policymakers are often in need of more comprehensive, yet compact, measures of regional integration. The recent wave of policy-driven studies constructing composite measures of regional integration outcomes is evidence of this need (De Lombaerde et al. 2008; AfDB et al. 2016; ADB 2017; Claveria and Park 2018).

This paper addresses this gap by proposing a new approach to estimate untapped regional integration potential and demonstrating how the method can be used to evaluate achievements in targeting subregions with certain needs and characteristics, and guiding decisions about future policy interventions aimed at fostering higher regional integration. ³ In measuring regional integration, we follow the recent notion in the literature and construct a composite index that aggregates information from various empirical indicators covering five dimensions of regional integration: trade integration, financial integration, regional investment and production networks, movement of people, and peace and security. The obtained composite regional integration (CRI) index is then used as the output variable in a non-parametric frontier analysis to estimate the empirical magnitudes of untapped regional integration potential across geographical subregions of the world. This is done by setting the CRI index in relation to a set of proxy measures of the enabling environment for regional integration (capturing enabling factors related to trade openness, cross-border infrastructure, and business regulation environment), and applying data envelopment analysis (DEA). ⁴ In particular, this method calculates the empirical production possibility frontier for regional integration outcomes, where the frontier specifies the potential level of integration (quantified by the CRI index) as a function of currently available resources and enabling factors needed to foster economic integration in each subregion.

Finally, we use the estimated frontier to rate achieved integration levels relative to subregions’ empirical potential to obtain an estimate of their untapped integration potential. This allows us to identify those subregions that are currently achieving relatively high levels of integration in comparison to other subregions facing a similar enabling environment, and those subregions that are apparently falling short of their potential.

There are several ways in which the method we propose for estimating regional integration potential contributes to the literature. First, using a non-parametric frontier approach has the benefit that it involves fewer assumptions about the structure of the data than parametric methods such as standard regression analysis. ⁵ As a consequence, the results generated from the DEA-based method tend to be much easier to interpret, and are not subject to the same econometric challenges that studies based on regression analysis typically face. Second, our analysis provides results that go beyond the insights obtained from a simple comparison of absolute levels of regional integration outcomes, as done in previous studies, because such a comparison does not take into account the individual conditions faced by each subregion. Augmenting our composite index of regional integration with a proxy measure of the enabling environment in each subregion when applying DEA thus generates results that provide an additional perspective on regional integration outcomes which complements the insights of existing studies in this context. Third, our methodology and selection of variables are designed to compare subregions across all parts of the world rather than focusing on a single geographical region, as existing studies in this context do, e.g. Africa in the case of AfDB et al. 2016, and Asia in the case of ADB 2017; Claveria and Park 2018; and Naeher 2015.

Our theoretical insights and empirical results are in several ways useful from a policy perspective. First, the obtained results can help development organizations with a mandate to foster regional integration to assess their achievements in targeting subregions with certain needs and characteristics, e.g. subregions that are performing relatively well despite low absolute levels of integration or subregions that are still far away from reaching their estimated integration potential. Second, our analysis provides insights about which enabling factors, e.g. trade-related political institutions, physical infrastructure, or private sector regulations, are particularly strong or weak in certain subregions. For example, these insights can be used to (i) evaluate the relevance and appropriateness of current interventions in a given subregion, (ii) review the ‘frontier’ regions and subregions with the highest untapped potential, and (iii) decide on the types of interventions that future programs should prioritize in each subregion. ⁶

The remainder of the paper is structured as follows. Section 2 explains the methodology underlying the construction of the CRI index and estimation of untapped integration potential via data envelopment analysis. Section 3 describes the involved variables and data sources. Section 4 presents the results of a global comparison of regional integration levels and estimated magnitudes of untapped integration potential. Section 5 discusses the robustness of the results. Section 6 concludes.

2. Methodology

2.1. Composite regional integration index

When it comes to the construction of composite indices, no unique standard procedure has been established in the literature. Rather, the applied methods have to be adapted to the specific context and purpose at hand (De Lombaerde et al. 2008). This is reflected in our analysis, which is based both on methods that are specifically designed to capture regional integration outcomes as well as on standard normalization and aggregation methods that are also used in the construction of other well-known composite indices (e.g. the Doing Business Index or the Human Development Index). The following provides a detailed description of the methodology underlying the construction of the CRI index, which builds upon the approach first outlined in Naeher (2015) and is in line with the guidelines laid out by the OECD (2008). ⁷ The robustness of the resulting estimates to alternative specifications is discussed as part of the robustness tests in Section 5.

Figure 1 shows a schematic illustration of the construction of the CRI index. The final composite index captures five dimensions of regional integration: (I.) trade integration, (II.) financial integration, (III.) regional investment and production networks, (IV.) movement of people, and (V.) peace and security. Each dimension comprises two or three individual indicators, e.g. the trade dimension is composed of intraregional exports and intraregional imports. These indicators are obtained from 11 empirically observed variables measuring different aspects of regional integration.

Figure 1. Composite regional integration index.

Display full size

In contrast to other studies in this context, we distinguish between ultimate regional economic integration outcomes (e.g. actual flows of goods and people across borders) and intermediate outcomes that can be perceived as means for achieving higher ultimate outcomes (such as signed FTAs, available infrastructure, and prevailing business regulations). ⁸ While the latter will be used as proxies of the enabling environment for regional integration, the CRI index is designed to capture subregions’ performance in terms of ultimate integration outcomes. For this reason, we also include peace and security as one of the dimensions of the CRI index, capturing important noneconomic aspects of regional integration outcomes such as reduction in conflict and increased stability within subregions.

Different to composite indices in other contexts, the construction of a multidimensional index of regional integration outcomes requires measures based on bilateral (dyadic) data rather than national data. There are several possible ways to construct such measures. ⁹ In order to guarantee comparability across the different dimensions of the composite index, we focus on variables that can be measured as intraregional shares of bilateral data (the only exception is the peace and security dimension for which no bilateral data could be identified). ¹⁰

Based on a country-by-country matrix containing information about flows $F_{i j}$ between countries $i$ and $j$ , the intraregional share is defined as the fraction of flows between the countries in subregion $R$ (denoted $F_{R R}$ ) and total flows between the countries in $R$ and all countries in the world ( $F_{R W}$ ). Formally, intraregional shares are calculated as (1) $\frac{F_{R R}}{F_{R W}} = \frac{\sum_{i \in R} \sum_{j \in R, j \neq i} F_{i j}}{\sum_{i \in R} \sum_{j \in W, j \neq i} F_{i j}}$ (1) To facilitate aggregation into an overall index, the data must be normalized such that higher values indicate higher degrees of regional integration and all variables feature a comparable range of values. For those subindicators representing intraregional shares, the former is already achieved. For the peace and security dimension, the data is inverted by subtracting the respective scores from the highest possible value (i.e. 10). Regarding the range of values, several possible methods exist for rescaling, each featuring its own set of advantages. We apply standard min–max rescaling, which ensures that all normalized variables feature the same range from 0 to 1. ¹¹ For each subregion $i$ in the overall sample $N$ , indicator $I$ is normalized using the formula (2) $I_{i}^{*} = \frac{I_{i} - mi n_{i \in N} (I_{i})}{ma x_{i \in N} (I_{i}) - mi n_{i \in N} (I_{i})}$ (2)

As shown in Figure 1, the construction of the CRI index involves two steps of aggregation. First, the overall composite index incorporates information along five dimensions of regional integration. Second, each of these dimensions is composed of multiple individual indicators – mostly intraregional shares of bilateral variables. At both levels of aggregation, an equal weighting scheme is applied to combine the respective subindicators. This facilitates the interpretation of the results and is in line with many other studies that construct composite indices. ¹²

Overall aggregation follows the scheme illustrated in Figure 1, using the variables listed in Table 1 and the normalization and weighting schemes described above. Since each dimension enters the index with equal weight, the resulting score of the CRI index can be interpreted as the average performance of a given subregion along the considered dimensions of regional integration. The same applies to each of the five dimensions individually, across the respective subindicators. Alternative weighting schemes (including principal component analysis) are explored as part of the robustness tests in Section 5.

Table 1. Variable descriptions and data sources.

CSV Display Table

2.2. Data envelopment analysis

Data envelopment analysis (DEA) is a nonparametric method for estimating production possibility frontiers using linear programing (see Chames, Cooper, and Rhodes 1978; Coelli et al. 2005; Sickles and Zelenyuk 2019). It can be used to measure rates of relative efficiency within a set of comparable units of observation. ¹³ In its standard form applied here, DEA assumes the existence of a convex production possibility set. The frontier of this set is empirically estimated as the maximal attainable level of output for a given level of input. The distance between an observed input–output combination and the estimated production possibility frontier is then used as a measure of the unit’s relative efficiency. For example, unit A is considered to be relatively inefficient if another unit B uses less or an equal amount of inputs to generate more amount of output (i.e. the distance between unit A and the frontier is larger than the distance between unit B and the frontier). The obtained efficiency scores are normalized to range between 0 and 1, where units located on the frontier are assigned the maximum value of 1. ¹⁴

In the context of regional integration, the key assumption underlying DEA is that subregions that feature similar levels of enabling factors for integration (such as, e.g. quality of cross-border infrastructure or institutional arrangements that facilitate trade and multinational private sector activities) should in principle be able to achieve similar levels of regional integration outcomes. Estimated efficiency scores below one can thus be interpreted as untapped potential in regional integration. In the analysis, untapped regional integration potential is therefore defined as the distance between the currently achieved level of regional integration (measured by the CRI index) and the theoretically possible level corresponding to the estimated frontier.

DEA requires two types of variables, input variables and output variables. In estimating untapped regional integration potential, we use the CRI index as the DEA output variable and set it in relation to a proxy measure of the enabling environment (DEA input variable). In the context of this study, the enabling environment consists of factors that facilitate higher levels of ultimate regional integration outcomes as captured by the CRI index. While there are many potential factors that may affect regional integration outcomes, including geographical features (e.g. distance and natural characteristics) and cultural background (e.g. common language), we focus on factors that are more directly in control of governments and international policy makers. In particular, we consider factors related to trade openness, cross-border infrastructure, and business regulation environment. These are chosen, because they represent key enabling factors behind the three processes that are often seen as driving regional integration. ¹⁵

2.2.1. Market-led processes

A common view is that regional integration arises naturally as a result of economic forces when the benefits of agglomeration (e.g. economies of scope, scale, and speed) outweigh the costs of agglomeration, such as congestion (McKay, Armengol, and Pineau 2005; Marinov 2015). Such market-led processes are often driven by reductions in barriers to trade and investment, as well as by the development of regional transportation and communication infrastructure (Francois and Manchin 2013). To account for market-led processes in the analysis, we include regional average scores of the World Bank’s Logistic Performance Index (LPI) as one of the proxies for the enabling environment. The LPI captures a wide range of relevant aspects in this context, including the quality of trade and transport-related infrastructure, efficiency of customs clearance processes, and various other export and import-related conditions such as ease of arranging shipments, quality of logistics services, and ability to track consignments.

2.2.2. Institution-led processes

Another driving force of regional integration is based on institutional arrangements, such as regional trade agreements, customs unions, and bilateral free trade agreements between economies within the same region. In order to account for institution-led processes in the analysis, we include a measure of intraregional FTA coverage, which is the percentage of country pairs within each subregion that have signed free trade agreements. The data comes from the Design of Trade Agreements Database (Dür, Baccini, and Elsig 2014), which provides information for around 730 preferential trade agreements, covering various types of agreements that liberalize trade, including bilateral, regional, and inter-regional agreements (we include only those agreements that are listed by the WTO).

2.2.3. Private sector-led processes

A third driving force behind regional integration is due to private sector-led initiatives, such as the establishment of regional production networks through multinational corporations and development of subregional economic zones (Peng 2002; Yoshimatsu 2002). ¹⁶ To account for private sector-led processes in the analysis, we include subregional average scores of the World Bank’s Doing Business Index (DBI) as a third proxy for the enabling environment. The DBI scores capture various important factors in this context, including business entry regulations, financing constraints, and taxation issues.

To facilitate the computation of DEA, the three identified variables capturing enabling factors for regional integration are combined into a single measure, the ‘DEA input index’. This is done by applying the same aggregation methods as used in the construction of the CRI index, i.e. normalization via min–max rescaling and aggregation based on equal weights. The obtained values for each subindicator are reported in Table A1 in the appendix. Overall, the DEA input index should be thought of as a proxy composite measure of the enabling environment for regional integration faced by individual subregions.

It is important to note that the estimates obtained from the performed DEA are based on currently available resources and conditions, not on potential future developments. In particular, our analysis does not seek to forecast integration outcomes under possible scenarios of changes in political or economic conditions, nor provide an assessment of the potential tradeoff between (sub)regional integration and efforts aimed at increasing economic integration at a global level. Instead, the analysis compares levels of integration outcomes across different subregions and identifies those subregions that, relative to other subregions with a similar enabling environment, are currently achieving lower levels of integration than they should potentially be able to. Furthermore, it should be noted that the obtained estimates relate only to the dimensions of regional integration captured by the CRI index and do not provide direct implications for potential welfare or growth effects. ¹⁷

3. Data and variables

The construction of any composite index capturing regional integration levels is limited by the availability of empirical measures of regional integration outcomes. For indicators that can be represented as intraregional shares of bilateral data, the number of potential variables to be included is even more restricted. Nevertheless, several key dimensions of regional integration can be covered this way.

Table 1 provides a complete list of the variables used in the analysis and respective data sources. All data used in this study are publicly available. Dimensions I to IV capture economic integration outcomes and are broadly in line with the variables used in other studies on regional economic integration (e.g. AfDB et al. 2016; ADB 2017; Huh and Park 2018). Dimension V is based on data from the Global Conflict Risk Index (GCRI), a database designed to provide global risk assessments based on economic, social, environmental, security and political factors, including the risk of confrontation with other states (GCRI 2017; Halkia et al. 2017). The GCRI quantifies the statistical risk of violent conflict occurring in the subsequent 1–4 years, based on a total of 24 individual indicators and a final scale from 0 (no conflict) to 10 (highly violent conflict).

For most of the included variables listed in Table 1, the latest data available are for 2017, 2016, or both. In these cases, we either use the most recent year available or combine the information from both years to achieve better coverage (see Appendix B for details). For indicators III.b and IV.c, the most recent data we were able to obtain are from 2015 and 2012, respectively. In case of the latter, we use the average annual growth over the previous five years to linearly extrapolate the data to 2017.

Based on data availability, the sample consists of 193 economies, grouped into 19 subregions spanning all geographical regions of the world. In defining subregions, we follow the classifications of the UN (2017). Note that while looking at geographical subregions may not always be fully in line with political objectives of the corresponding countries, working with the UN classifications has the advantage that it provides a complete set of country groupings, i.e. each country is mapped to exactly one subregion. ¹⁸ A complete list of the included economies and subregional groupings is provided in Appendix A. In some cases, information in the original datasets is missing for some economies, such that the affected subregions are only represented by a subset of the corresponding economies. In order to minimize potential biases, some attempts were made to adjust for missing values, e.g. by imputation. A detailed description of the extent and handling of missing data is provided in Appendix B.

4. Empirical results

4.1. Global comparison of regional integration levels

Figure 2 shows the ranking of subregions in terms of achieved levels of integration as measured by the CRI index (exact values are provided in Table A1 in the appendix). The two subregions with the highest CRI scores are Western Europe (0.78) and East Asia (0.71). After a considerable gap, Southeast Asia (0.50), Northern Europe (0.47), and North America (0.45) follow next. At the lower end of the ranking are Central Asia (0.16), Northern Africa (0.11), and Middle Africa (0.06). The last two subregions feature CRI scores below one third of the average CRI index across all 19 subregions, which equals 0.34.

Figure 2. CRI index: global comparison.

Notes: Author’s calculation based on the methodology illustrated in Figure 1 and data sources described in Table 1. Exact numerical values of the CRI index for each subregion are reported in Table A1 in the appendix.

Display full size

Figure 3 depicts results for geographical regions, where the thick squares represent simple averages and the lines indicate ranges of values for subregions in a given region (bounded by the minimum and maximum value within each region). In terms of average CRI levels, Europe (0.50) achieves the highest result, Asia (0.36) and the Americas (0.33) are close to the overall sample mean across all subregions (0.34), and Africa (0.21) lags behind. The gap between Africa and Europe becomes even more apparent when looking at the depicted ranges. While all regions feature considerable heterogeneity in regional integration levels, the most integrated subregion in Africa (Southern Africa) has a lower CRI score than the least integrated subregion in Europe (Eastern Europe). Asia shows the by far largest range of CRI scores, being the only region that comprises subregions both at the very top and bottom end of the CRI index.

Figure 3. CRI index: ranges for geographical regions.

Notes: For each geographical region, the minimum (lower line), maximum (upper line), and average value (thick square) of the CRI index across the corresponding subregions are shown.

Display full size

Overall, the results in Figures 2 and 3 are in line with the common view that economies in Europe (in particular the Western European economies belonging to the European Union) currently feature the highest level of regional integration in the world (Baldwin and Wyplosz 2006; Freund and Ornelas 2010). However, while the gap between Europe and other regions may be large in terms of institutional integration (e.g. as measured by the five stages of regional integration defined by Balassa 1961), the findings in Figure 2 suggest that in terms of actual integration outcomes as measured by the intraregional shares included in the CRI index, some subregions in Asia are currently achieving outcomes that are comparable to those achieved by European subregions.

Figure 4 shows the resulting rankings of subregions for individual dimensions of regional integration. Disaggregating the results in this way reveals that Western Europe is leading the rankings for trade and financial integration as well as for regional investment and production networks. East Asia performs very well in terms of trade integration, investment and production networks, and movement of people, such that the lower CRI score compared to Western Europe can be mainly attributed to financial integration and peace and security. Middle Africa, Northern Africa, and Central Asia are among the lowest ranked subregions for almost all five dimensions, suggesting that the low values of the overall CRI index for these subregions are not driven by any particular dimension. While the two subregions comprising island states, the Caribbean and Pacific & Oceania, are ranked among the bottom half for most dimensions, they perform very well in terms of peace and security (dimension V). The opposite holds for Eastern Europe, which achieves relatively high scores for dimensions I to IV, but ranks low for dimension V.

Figure 4. CRI index: dissagregated by dimension.

Notes: Subregions with special relevance for the IEG’s evaluation study are marked in dark blue. Exact numerical values for each subregion are reported in Table A1 in Appendix C.

Display full size

As shown in Figure 4, there is a large gap in financial integration between the highest ranked subregion (Western Europe) and all other subregions. One might worry that some of the results are primarily driven by the high value of Western Europe for financial integration. However, as shown in Section 5, the overall ranking of subregions remains almost unchanged when different weighting schemes are applied, suggesting that the results are not merely driven by one particular dimension.

It should be noted that the ranking for dimension IV (movement of people) differs relatively strongly from the rankings for the other dimensions. The low scores for North America and some of the European subregions for dimension IV are likely due to the high global mobility that people in these subregions enjoy, rather than to constraints on movement within these subregions. For example, small intraregional shares for migration in richer subregions may be due to the fact that individual migration decisions in these subregions are not restricted to neighboring countries (within the same subregion), as might be the case for many people in poorer and less-developed subregions.

4.2. Frontier analysis of untapped integration potential

We now demonstrate how the constructed composite index of regional integration outcomes (CRI index) can be used in a DEA to obtain estimates of untapped regional integration potential. As described above, the CRI index enters the DEA as output variable and is set in relation to a set of proxy measures of the enabling environment for regional integration in each subregion (the DEA input index). The obtained values of the DEA input index are reported in the first column of Table 2. According to the results, the subregions with the strongest enabling environments for regional integration are North America (0.97), Western Europe (0.95), and Northern Europe (0.92). The subregions with the weakest enabling environments are Eastern Africa (0.37), the Caribbean (0.34), and Middle Africa (0.18). Looking at simple averages across geographical regions, Europe (0.76) obtains the highest score, while the Americas (0.57) and Asia (0.48) are ahead of Africa (0.42).

Table 2. Data envelopment analysis estimates of untapped integration potential.

CSV Display Table

Figure 5 plots the CRI index over the DEA input index and shows the resulting production possibility frontier for regional integration (dotted line). For our sample of 19 subregions, the frontier turns out to be defined by four subregions. At the lower end of the enabling environment, the frontier is defined by Middle Africa, which is mainly due to the fact that there are no other subregions with similarly small values of the DEA input index. In the middle of the sample, the frontier is defined by Southest Asia and East Asia, both of which outperform many other subregions with similar levels of the DEA input index. At the upper end, the frontier is defined by Western Europe, which achieves a much higher CRI score than the two subregions with similar enabling environment, North America and Northern Europe.

Figure 5. Regional integration frontier.

Notes: The dotted line represents the production possibility frontier for regional integration as measured by the CRI index. Exact numerical values of the CRI index and the DEA input index for each subregion are reported in Table A1 in Appendix C.

Display full size

Table 2 presents the resulting estimates for untapped regional integration potential (DEA scores), along with each subregion’s rank. Larger ranks correspond to smaller scores and indicate higher potential for increasing regional integration levels based on current conditions (an estimated score of one indicates the subregion is located on the frontier). ¹⁹ Most of the subregions achieve scores between 0.4 and 0.7, suggesting that these subregions are currently achieving around 40–70 percent of their potential integration levels. ²⁰ South America and the Caribbean obtain scores larger than 0.65, which suggests that these subregions are performing considerably well given their enabling environments. The three subregions with the largest untapped integration potential are Central America, Central Asia, and Northern Africa, which, according to the estimates, are currently achieving only less than one third of their potential integration levels.

According to the average scores reported at the bottom of Table 2, all geographical regions include subregions with considerable untapped integration potential. On average, the subregions in Europe achieve around 67 percent of their integration potential (with Eastern and Southern Europe featuring the lowest DEA scores). In Asia, average untapped integration potential is around 40 percent (with the lowest score obtained by Central Asia). The subregions of the Americas and Africa are found to achieve 57 percent of their integration potential on average (with the lowest scores for Central America and Northern Africa, respectively). Globally, average regional integration levels across all subregions are found to be at 60 percent of the estimated potential.

It should be noted that the quantitative results should be interpreted with caution, as data availability and quality for the used indicators are limited. Still, the obtained rankings along the CRI index as well as for individual dimensions of regional integration appear to be plausible in comparison to the findings of other studies in this context. In addition, we show below that the obtained results are generally robust to moderate changes in the aggregation methods underlying the construction of the CRI index (as well as the exclusion of the peace and security dimension from the CRI index). While the analysis itself focuses exclusively on economic integration and conflict reduction, the findings may also be used as a basis for discussions on further advances in regional integration at the institutional level, or as a starting point for investigations into the deeper reasons behind each subregion’s performance (both of which go beyond the scope of this study).

5. Robustness

The construction of the variables used in the analysis involves a number of decisions which may affect our findings. This section reports a series of tests to assess the robustness of our findings to different specifications, focusing on the selection of indicators and weights used in the aggregation of the CRI index and the DEA input index.

Table 3 reports the values and rankings of the CRI index resulting from different weighting schemes, including principal component analysis (PCA) and the exclusion of dimension V (peace and security) from the composite index. The first two columns in Table 3 replicate the baseline results obtained from the methodology described in Section 2. In the subsequent columns, four different weighting schemes are explored, where each scheme assigns double weight to one particular dimension. For example, in the column for dimension I., the indicator for trade integration is assigned a weight of 2/6, while the indicators for the remaining three dimensions are each assigned a weight of 1/6. The two columns thereafter report the resulting values and ranking when the CRI index is constructed using principal component analysis. In this case, the weights are based on the relative importance of each dimension in the overall composite index as quantified by the factor loadings obtained from PCA, while keeping all five dimensions in the index (the corresponding eigenvalues and scoring coefficients are reported in Table A2 in Appendix C). ²¹ The last two columns in Table 3 show the results when the CRI index is constructed without dimension V. In addition to the standard Pearson correlation coefficient for continuous variables, the bottom of Table 3 also shows Spearman correlation coefficients which measure the similarity between discrete rankings. The Spearman correlation coefficient equals 1 if both rankings are identical, and values smaller than 1 indicate less agreement (a value of 0 indicates that the rankings are completely independent). For all the specifications tested in Table 3, both the Pearson and Spearman correlation coefficients tend to be very close to one and are always significant at the 1% significance level. Accordingly, the rankings of subregions show only relatively small changes across the considered specifications.

Table 3. Robustness of CRI index to different weighting schemes, PCA, and exclusion of dimension V.

CSV Display Table

Table 4 reports the results of a similar exercise for the DEA input index. Again, the correlation coefficients are significant at the 1% significance level for all specifications and, except for the Spearman correlation coefficient in the fourth column (where the weight is doubled for the dimension capturing trade openness), all coefficients are larger than 0.95.

Table 4. Robustness of DEA input index to different weighting schemes and PCA.

CSV Display Table

Overall, the results in Tables 3 and 4 show that the baseline indices used in the frontier analysis in Section 4 are robust to moderate changes in the underlying weighting schemes. For most subregions, the respective ranks of the CRI index and DEA input index show only relatively small changes across the considered alternative specifications, and correlation coefficients are always significant at the 1% significance level. In particular, this suggests that it is unlikely that the main results reported in Section 4 are driven purely by the aggregation methodology.

Another concern may be missing data. This applies particularly to dimension II (financial integration), where data is only available for around 40% of economies and the subregions Western Africa and Middle Africa are not covered at all. As an additional robustness check, we therefore also compute the CRI index without dimension II and compare the associated ranking to the baseline results. The resulting ranks of the two affected subregions change only marginally (the rank of Western Africa changes from 13 to 11 and the rank of Middle Africa remains unaffected) and only one other subregion changes ranks by more than two (West Asia changes from 11 to 14). This suggests that the imputed values for dimension II are not driving the overall results (see also Appendix B).

6. Conclusion and policy implications

This paper proposes a new method to address several important questions in the context of regional cooperation and integration: (i) how integrated are specific subregions compared to other subregions in the world when looking at multiple key dimensions of regional integration; (ii) what are the levels of available resources and prevailing conditions in each subregion that are needed to achieve higher regional integration outcomes, i.e. how weak or strong is the enabling environment for regional integration in different subregions; (iii) how well are individual subregions doing relative to other subregions facing a similar enabling environment; and (iv) how large is the untapped integration potential for each subregion based on currently available resources and conditions. To answer these questions, we first construct a composite index of regional integration that allows to quantify and compare integration outcomes across geographical subregions along five key dimension: trade, financial integration, regional investment and production networks, movement of people, and peace and security. In a second step, we set the CRI index in relation to a proxy measure of the enabling environment for regional integration and use a non-parametric frontier method to quantify magnitudes of untapped integration potential in each subregion.

Overall, our proposed method and analysis give rise to the following findings and implications. While Western Europe features the highest level of regional integration along the considered dimensions of the CRI index, some subregions in Asia (especially East Asia) follow closely behind. This holds particularly in terms of economic integration levels (such as trade and regional investment and production networks), where several subregions in Asia achieve outcomes that are comparable to those achieved by European subregions. The subregions with currently the lowest levels of integration are Middle and Northern Africa, Central Asia, and Central America. These subregions are ranked low across almost all of the five dimensions covered by the CRI index, suggesting that these subregions may face systematic barriers to higher integration rather than difficulties in just a single area.

Other subregions perform relatively well in terms of overall integration levels but seem to struggle with issues in one particular area. Taking these issues into account may be critical for the success of programs aimed at fostering stronger cooperation and integration. This may be particularly relevant for subregions facing challenges related to peace and security, as challenges in this domain are likely to also affect the outcomes in other areas of integration. According to the disaggregated results for each dimension, this applies to South Asia and Eastern Europe, as well as to all African subregions except Southern Africa. In East Asia, which achieves very high levels of economic integration, special attention should be paid to peace and security as well as to financial integration.

According to our measure of the enabling environment, subregions that currently feature relatively poor conditions for regional integration include Western, Eastern, and Middle Africa, as well as South Asia and the Caribbean. In these subregions, programs aimed at facilitating higher regional integration should aim primarily at improving the conditions needed to foster higher regional integration levels, such as trade openness, cross-border infrastructure, and business regulations. Southeast Asia appears to be a benchmark and potential role model for these subregions, as, despite featuring a similarly low enabling environment, Southeast Asia is achieving considerably higher levels of ultimate regional integration outcomes according to the CRI index.

In case strategic priorities are to be given to specific subregions, our estimates of untapped integration potential suggest that these should not be restricted to a single geographical region. Rather, we find that subregions with large untapped integration potential are spread across all geographical regions of the world. In particular, subregions which currently achieve only up to half of their estimated potential include Northern and Southern Africa, Central America, Central and South Asia, and Eastern Europe. Globally, average regional integration levels across all subregions are found to be at 60 percent of the estimated potential.

Table 1. Variable descriptions and data sources.

Dimension	Indicator		Description	Data source and year
*Panel A: Ultimate regional integration outcomes (CRI index, DEA outputs)*
I. Trade integration	I.a	Intraregional goods exports	Percentage of intraregional goods exports compared to total goods exports	IMF – Direction of Trade Statistics (DOTS), 2017
I. Trade integration	I.b	Intraregional goods imports	Percentage of intraregional goods imports compared to total goods imports	IMF – DOTS, 2016/17
II. Financial integration	II.a	Intraregional equity liabilities	Percentage of intraregional cross-border equity liabilities compared to total cross-border equity liabilities	IMF – Coordinated Portfolio Investment Survey (CPIS), 2016/17
II. Financial integration	II.b	Intraregional debt liabilities	Percentage of intraregional cross-border debt liabilities compared to total cross-border debt liabilities	IMF – CPIS, 2016/17
III. Regional investment and production networks	III.a	Intraregional FDI positions	Percentage of intraregional inward FDI positions compared to total inward FDI positions	IMF – Coordinated Direct Investment Survey (CDIS), 2015/16; UNCTAD – Bilateral FDI Statistics 2014
III. Regional investment and production networks	III.b	Intraregional intermediate goods imports	Percentage of intraregional inter-mediate goods exports compared to total intraregional goods exports	World Integrated Trade Solutions (WITS), 2015; UN – Commodity Trade Statistics (COMTRADE), 2017
IV. Movement of people	IV.a	Intraregional migration	Percentage of intraregional outbound migration compared to total outbound migration	UN Population Division – Trends in International Migrant Stock, 2017
	IV.b	Intraregional remittances	Percentage of intraregional remittances inflows compared to total remittances inflows	World Bank – Migration and Remittances Data, 2017
	IV.c	Intraregional tourism	Percentage of intraregional outbound tourists compared to total outbound tourists	UNWTO – Yearbook of Tourism Statistics, 2012 (extrapolated to 2017)
V. Peace and security	V.a	Absence of conflict	Subregional mean of GCRI scores for security risk area (current conflict situation, history of conflict)	EU – Global Conflict Risk Index (GCRI), 2017
V. Peace and security	V.b	Social cohesion and security	Subregional mean of GCRI scores for social risk area (social cohesion and diversity, public security and health)	EU – GCRI (2017)
*Panel B: Intermediate outcomes/enabling environment (DEA inputs)*
Trade openness			Percentage of country pairs within subregion that have signed FTAs	Design of Trade Agreements Database (DESTA), 2017
Cross-border infrastructure			Subregional mean of Logistics Performance Index (overall score)	World Bank – Logistics Performance Index (LPI), 2016
Business regulation environment			Subregional mean of Doing Business Index (distance to frontier score)	World Bank – Doing Business Index (DBI), 2017

FDI = foreign direct investment, FTA = free trade agreement, DEA = data envelopment analysis.

Table 2. Data envelopment analysis estimates of untapped integration potential.

Subregion	DEA input index	Score	Rank
Western Africa	0.405	0.62	5
Northern Africa	0.508	0.17	16
Eastern Africa	0.367	0.64	4
Middle Africa	0.177	1.00	1
Southern Africa	0.629	0.42	12
North America	0.974	0.57	7
Central America	0.486	0.29	14
Caribbean	0.337	0.72	2
South America	0.497	0.69	3
West Asia	0.420	0.55	8
Central Asia	0.557	0.22	15
East Asia	0.536	1.00	1
South Asia	0.390	0.48	11
Southeast Asia	0.411	1.00	1
Pacific & Oceania	0.586	0.38	13
Western Europe	0.952	1.00	1
Northern Europe	0.921	0.61	6
Eastern Europe	0.624	0.51	10
Southern Europe	0.534	0.54	9
*Regional Averages:*
Africa	0.42	0.57
Americas	0.57	0.57
Asia	0.48	0.60
Europe	0.76	0.67
Global Average	0.54	0.60

Notes: The last two columns report DEA scores and corresponding ranks based on output oriented analysis under variable returns to scale (VRS), estimated via DEAP 2.1 (Coelli 1996; Coelli et al. 2005). Input variable: DEA input index; output variable: CRI index (see Table A1).

Table 3. Robustness of CRI index to different weighting schemes, PCA, and exclusion of dimension V.

			Weights doubled for individual dimensions:												Excluding Dimension V
Subregion	CRI Index Baseline^a	Rank	I.	Rank	II.	Rank	III.	Rank	IV.	Rank	V.	Rank	PCA^b	Rank	CRI	Rank
Western Africa	0.31	10	0.29	11	0.26	11	0.29	10	0.39	8	0.30	12	−1.03	13	0.36	9
Northern Africa	0.11	18	0.10	18	0.10	18	0.10	18	0.15	18	0.12	18	−2.03	18	0.04	18
Eastern Africa	0.27	13	0.27	12	0.23	14	0.25	13	0.33	10	0.27	14	−1.09	14	0.27	11
Middle Africa	0.06	19	0.05	19	0.05	19	0.05	19	0.08	19	0.06	19	−2.30	19	0.08	19
Southern Africa	0.31	9	0.31	9	0.26	10	0.31	9	0.31	11	0.34	10	−0.28	10	0.23	12
North America	0.45	5	0.48	4	0.45	5	0.43	6	0.40	7	0.47	5	1.48	4	0.37	6
Central America	0.18	16	0.16	16	0.15	16	0.19	16	0.20	16	0.20	16	−1.36	15	0.16	15
Caribbean	0.26	14	0.24	14	0.23	13	0.22	14	0.29	13	0.33	11	−0.94	12	0.19	14
South America	0.44	6	0.45	6	0.39	6	0.46	4	0.48	4	0.43	6	0.51	7	0.45	4
West Asia	0.29	11	0.29	10	0.28	9	0.27	11	0.30	12	0.29	13	−0.49	11	0.23	10
Central Asia	0.16	17	0.15	17	0.13	17	0.15	17	0.16	17	0.19	17	−1.42	16	0.08	17
East Asia	0.71	2	0.74	2	0.65	2	0.72	2	0.72	1	0.69	2	2.86	2	0.73	2
South Asia	0.22	15	0.21	15	0.19	15	0.20	15	0.29	14	0.21	15	−1.57	17	0.31	13
Southeast Asia	0.50	3	0.52	3	0.46	4	0.51	3	0.55	3	0.49	4	0.99	5	0.50	3
Pacific & Oceania	0.27	12	0.25	13	0.25	12	0.26	12	0.25	15	0.35	8	−0.27	9	0.14	16
Western Europe	0.78	1	0.82	1	0.82	1	0.80	1	0.70	2	0.78	1	4.56	1	0.77	1
Northern Europe	0.47	4	0.48	5	0.48	3	0.44	5	0.42	5	0.55	3	1.63	3	0.35	7
Eastern Europe	0.37	8	0.41	7	0.35	8	0.36	8	0.40	6	0.34	9	0.10	8	0.41	5
Southern Europe	0.38	7	0.38	8	0.38	7	0.36	7	0.36	9	0.42	7	0.66	6	0.30	8
Pearson corr.	1.00	–	1.00	–	0.99	–	1.00	–	0.98	–	0.99	–	0.97	–	0.97	–
Spearman corr.^c	–	1.00	–	0.99	–	0.99	–	0.99	–	0.96	–	0.97	–	0.97	–	0.95

Notes: ^aSimple average, i.e. equal weights assigned to each dimension (as in Table A1, column 1). ^bBased on principal component analysis. ^cThe Spearman correlation coefficient ranges inside the interval [−1,1] and takes the value 1 if the agreement between two rankings is perfect (i.e. the two rankings are identical), the value 0 if the rankings are completely independent, and the value −1 if one ranking is the reverse of the other.

Table 4. Robustness of DEA input index to different weighting schemes and PCA.

			Weights doubled for individual dimensions:
Subregion	DEA Input Index Baseline^a	Rank	Trade Openness	Rank	Cross-border Infrastructure	Rank	Business regulation	Rank	PCA^b	Rank
Western Africa	0.41	15	0.52	10	0.33	16	0.36	16	−1.09	15
Northern Africa	0.51	10	0.63	7	0.44	10	0.45	12	−0.27	10
Eastern Africa	0.37	17	0.45	14	0.32	17	0.34	18	−1.40	17
Middle Africa	0.18	19	0.27	19	0.13	19	0.13	19	−2.92	19
Southern Africa	0.63	4	0.72	4	0.56	6	0.60	6	0.70	4
North America	0.97	1	0.98	1	0.96	2	0.98	1	3.45	1
Central America	0.49	12	0.52	11	0.44	11	0.50	10	−0.46	12
Caribbean	0.34	18	0.38	18	0.27	18	0.36	17	−1.65	18
South America	0.50	11	0.57	9	0.45	9	0.48	11	−0.36	11
West Asia	0.42	13	0.40	15	0.41	14	0.45	13	−0.99	13
Central Asia	0.56	7	0.67	5	0.43	12	0.57	9	0.13	7
East Asia	0.54	8	0.40	16	0.59	4	0.62	5	−0.07	8
South Asia	0.39	16	0.45	13	0.35	15	0.37	15	−1.22	16
Southeast Asia	0.41	14	0.38	17	0.41	13	0.44	14	−1.06	14
Pacific & Oceania	0.59	6	0.66	6	0.53	7	0.57	8	0.35	6
Western Europe	0.95	2	0.96	2	0.96	1	0.93	3	3.28	2
Northern Europe	0.92	3	0.93	3	0.89	3	0.94	2	3.03	3
Eastern Europe	0.62	5	0.63	8	0.57	5	0.67	4	0.65	5
Southern Europe	0.53	9	0.50	12	0.51	8	0.59	7	−0.08	9
Pearson corr.	1.00	–	0.95	–	0.99	–	0.99	–	1.00	–
Spearman corr.^c	–	1.00	–	0.86	–	0.95	–	0.96	-	1.00

Notes: ^aSimple average, i.e. equal weights assigned to each dimension (as in Table A1, column 7). ^bBased on principal component analysis. ^cThe Spearman correlation coefficient ranges inside the interval [−1,1] and takes the value 1 if the agreement between two rankings is perfect (i.e. the two rankings are identical), the value 0 if the rankings are completely independent, and the value −1 if one ranking is the reverse of the other.

Table A1. Regional integration and enabling environment: disaggregated by dimensions.

	Ultimate Regional Integration Outcomes						Enabling Environment
Subregion	CRI index	Trade Integration	Financial Integration	Investment and Production Networks	Movement of People	Peace and Security	DEA Input Index	Trade Openness	Cross-border Infrastructure	Business Regulation Environment
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Western Africa	0.31	0.20	0.02	0.23	0 . 78	0.29	0.41	0 . 88	0.10	0.23
Northern Africa	0.11	0.05	0.01	0.01	0.31	0.19	0.51	1 . 00	0.23	0.29
Eastern Africa	0.27	0.25	0.04	0.17	0.63	0.26	0.37	0.68	0.17	0.25
Middle Africa	0.06	0.01	0.02	0.00	0.19	0.08	0.18	0.53	0.00	0.00
Southern Africa	0.31	0.33	0.03	0.33	0.35	0.49	0.63	1 . 00	0.36	0.53
North America	0.45	0.66	0.48	0.33	0.15	0.62	0 . 97	1 . 00	0 . 92	1 . 00
Central America	0.18	0.06	0.01	0.23	0.31	0.30	0.49	0.63	0.28	0.55
Caribbean	0.26	0.14	0.08	0.00	0.45	0.64	0.34	0.50	0.08	0.43
South America	0.44	0.47	0.14	0.55	0.69	0.37	0.50	0 . 78	0.29	0.42
West Asia	0.29	0.32	0.24	0.17	0.39	0.30	0.42	0.35	0.37	0.53
Central Asia	0.16	0.12	0.00	0.12	0.19	0.36	0.56	1 . 00	0.06	0.61
East Asia	0.71	0 . 93	0.39	0 . 82	0 . 79	0.61	0.54	0.00	0.73	0 . 87
South Asia	0.22	0.18	0.04	0.09	0.65	0.15	0.39	0.64	0.22	0.32
Southeast Asia	0.50	0.58	0.24	0.55	0 . 75	0.41	0.41	0.30	0.41	0.53
Pacific & Oceania	0.27	0.13	0.12	0.22	0.13	0 . 75	0.59	0.88	0.35	0.52
Western Europe	0 . 78	0 . 98	1 . 00	0 . 89	0.29	0 . 75	0 . 95	1 . 00	1 . 00	0 . 86
Northern Europe	0.47	0.50	0.50	0.27	0.17	0 . 94	0 . 92	0 . 97	0 . 80	1 . 00
Eastern Europe	0.37	0.58	0.22	0.32	0.54	0.20	0.62	0.65	0.41	0 . 81
Southern Europe	0.38	0.37	0.39	0.29	0.24	0.61	0.53	0.40	0.42	0 . 78
*Regional Averages:*
Africa	0.21	0.17	0.02	0.15	0.45	0.26	0.42	0 . 82	0.17	0.26
Americas	0.33	0.33	0.18	0.28	0.40	0.48	0.57	0.73	0.40	0.60
Asia	0.36	0.37	0.17	0.33	0.48	0.43	0.48	0.53	0.36	0.56
Europe	0.50	0.61	0.53	0.44	0.31	0.62	0 . 76	0 . 76	0.66	0 . 86
Global Average	0.34	0.36	0.21	0.29	0.42	0.44	0.54	0.69	0.38	0.55

CRI = composite regional integration, DEA = data envelopment analysis.

Notes: Values above 0.75 are marked in bold.

Table A2. Robustness tests: principal component analysis.

	I. Trade Integration	II. Financial Integration	III. Investment and Prod. Networks	IV. Movement of People	V. Peace and Security
Eigenvalues	3.042	1.392	0.369	0.138	0.059
Proportion	0.608	0.278	0.074	0.028	0.012
Cumulative Proportion	0.608	0.887	0.961	0.988	1.000
Scoring Coefficients:
Component 1	0.543	0.528	0.514	0.027	0.402
Component 2	0.193	−0.180	0.283	0.810	−0.440
Component 3	−0.192	−0.380	−0.056	0.416	0.801
Component 4	0.137	0.535	−0.746	0.370	0.043
Component 5	−0.783	0.508	0.310	0.181	−0.019

Disclosure statement

This research builds upon work which the authors conducted as part of a World Bank evaluation study.

Notes

1 Potential channels through which regional integration may improve economic outcomes include creating larger and more efficient markets (e.g. based on economies of scale and forces of competition), introducing additional investment opportunities, and driving technological change (Krugman 1991; Baldwin and Venables 1995; Fernandez and Portes 1998; Sapir 2011). In addition, regional integration is often seen as a ‘building block’ for global trade liberalization and multilateralism (Bhagwati 1993; Baldwin 2006; Calvo-Pardo, Freund, and Ornelas 2011).

2 The term `subregion’ refers to a set of (typically bordering) countries located in the same geographical region (see examples below). Although we provide average results also for geographical regions (i.e. continents), the main focus of this paper is on subregions as defined in Appendix A. This focus is in line with other recent studies on regional integration (e.g. Claveria and Park 2018 and the studies summarized in De Lombaerde et al. 2008). For simplification, we use the term regional integration even when talking about subregions.

3 The method we propose builds and expands upon the approach developed in Naeher (2015).

4 DEA is a nonparametric method for estimating production possibility frontiers based on linear programing. Some details on the methodological background are provided in Section 2 (for a more elaborated introduction to DEA, see Coelli et al. 2005).

5 In principle, DEA and regression analysis are alternative methods for performance assessments, each featuring its own set of advantages and weaknesses (Thanassoulis 1993; Sickles and Zelenyuk 2019). While regression analysis provides more tools for identifying causal relationships in observational data, it also tends to involve stronger assumptions on the structural form of the underlying data than non-parametric methods, causing it to be more sensitive to econometric challenges such as endogeneity and omitted variable bias. Since the purpose of our analysis, i.e. to quantify and compare empirical magnitudes of untapped integration potential across subregions, does not involve making any causal claims, DEA appears to be an appropriate choice.

6 It should be noted that the obtained estimates of untapped integration potential are exclusively based on currently available resources and conditions, not on potential future developments. In particular, our analysis does not seek to forecast integration outcomes under possible scenarios of changes in political or economic conditions, nor provide an assessment of the potential tradeoff between (sub)regional integration and efforts aimed at increasing economic integration at a global level.

7 We make several refinements to the methodology in Naeher (2015), add additional variables, and expand the analysis to the global sample of subregions rather than focusing on Asia. It should be noted that some authors have recently expressed concerns about the general usefulness of composite indices in evaluating economic performance, and instead argue in favor of an alternative ‘dashboard’ approach of monitoring each component separately (Stiglitz, Sen, and Fitoussi 2009; Ravallion 2010). In an attempt to accommodate this view, we also provide the results of the CRI index in disaggregated form for all individual subindicators.

8 Other composite indices pool these two sets of outcomes together, e.g. the Africa Regional Integration Index (AfDB et al. 2016) and the Asia-Pacific Regional Integration Index (Huh and Park 2018).

9 Possible alternatives to the intraregional shares used here include intraregional correlation coefficients and intensity indices.

10 Since the peace and security dimension differs from the other dimensions of the CRI index in that it represents noneconomic outcomes and in that it is not based on bilateral data, Section 5 reports the resulting values of the CRI index and associated rankings both with the peace and security dimension and without.

11 The same method is used in the construction of other well-known composite indices, such as the Africa Regional Integration Index, Doing Business Index, and the Unit Nation’s Human Development Index.

12 This includes popular indices such as the Human Development Index and the Africa Regional Integration Index, as well as indices constructed particularly for data envelopment analysis (e.g. Afonso, Schuknecht, and Tanzi 2005). In Afonso, Schuknecht, and Tanzi (2005), the use of an equal weighting scheme is justified by stating that this weighting ‘is quite straightforward and economically intuitive (even though it is still somewhat ad hoc). It avoids the problem of lack of economic justification of a more complex statistical approach.’

13 DEA has been applied to a wide range of fields, including efficiency of expenditures on health and education (Clements 2002; Herrera and Pang 2005), agricultural efficiency (Latruffe et al. 2004), and overall public sector efficiency (Afonso, Schuknecht, and Tanzi 2005, 2010; Gupta and Verhoeven 2001). Naeher (2015) uses DEA in the context of regional economic integration in Asia.

14 As noted in Section 1, DEA and regression analysis such as Ordinary Least Squares (OLS) are alternative methods for performance assessments, each featuring its own set of advantages and weaknesses. While DEA is based on a function that is determined by the most efficient units in the sample, OLS techniques are based on comparisons relative to an average unit. For a more elaborated discussion, see Thanassoulis (1993) and Sickles and Zelenyuk (2019). In our analysis in Section 4, the estimation is carried out using the software tool DEAP 2.1 (see Coelli 1996; Coelli et al. 2005).

15 In addition, some of the included indicators for cross-border infrastructure may also (at least partially) capture geographic conditions, e.g. by representing de-facto distances between economies in terms of the time and cost for transportation.

16 While in the past, the private sector was often viewed as merely reacting to institution and market-led processes, an increasing number of studies argue that in many parts of the world, private sector-led multinational initiatives have become an important driving force itself, with the role of governments and international organizations being limited to at most facilitating business initiatives by ensuring appropriate local policy condition (Peng 2002; ASEAN 2004).

17 Implications of economic integration in this direction are discussed elsewhere (e.g. Baldwin and Venables 1995; Sapir 2011).

18 An alternative approach would be to use country groupings corresponding to regional economic communities (RECs), as done in AfDB et al. (2016). However, this would lead to methodological difficulties for the subsequent frontier analysis, which would have to deal with (i) countries that are not part of any REC, (ii) countries that are part of multiple RECs, and (iii) the fact that intra-regional shares would not be consistent anymore with total global flows of the respective variables. At the same time, potential issues due to differences in the sizes of subregions would remain.

19 Note that the obtained results should be interpreted as lower bounds, since for subregions located on the frontier untapped potential is assumed to be zero by definition, even though there may still be scope for further enhancement in these subregions (there are simply no other subregions in the sample that can serve as benchmarks).

20 Note that the reported estimates are based exclusively on currently available resources and prevailing conditions, and thus do not allow for interpretations of how close subregions are to their general integration potential if economic or political conditions improve in the future.

21 Note that we abstain from dropping variables to reduce dimensionality, as sometimes done in other studies, since the numbers of dimensions of our composite indices are already quite low (which is mainly due to the limitations in data availability described in Section 2).

References

ACET . 2017. African Transformation Report 2017 . Accra : African Center for Economic Transformation. [Google Scholar]
ADB . 2017. Asian Economic Integration Report 2017 . Manila : Asian Development Bank. [Google Scholar]
Afonso, A. , L.Schuknecht, and V.Tanzi . 2005. “Public Sector Efficiency: An International Comparison.” Public Choice 123 (3–4): 321–347. doi: 10.1007/s11127-005-7165-2 [Crossref], [Web of Science ®], [Google Scholar]
Afonso, A. , L.Schuknecht, and V.Tanzi . 2010. “Public Sector Efficiency: Evidence for New EU Member States and Emerging Markets.” Applied Economics 42 (17): 2147–2164. doi: 10.1080/00036840701765460 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
African Development Bank, African Union Commission, and UNECA . 2016. Africa Regional Integration Index: Report 2016. African Development Bank, African Union, and United Nations Economic Commission for Africa. https://www.integrate-africa.org . [Crossref], [Google Scholar]
ASEAN . 2004. “AFTA and Beyond: What Drives Economic Integration in ASEAN?” ASEANOne 1, November 11. Association of Southeast Asian Nations. http://www.asean.org/uploads/archive/Lens%20-%20Myrna.pdf . [Google Scholar]
Balassa, B. 1961. The Theory of Economic Integration . Homewood : Richard Irwin. [Google Scholar]
Baldwin, R. E. 2006. “Multilateralising Regionalism: Spaghetti Bowls as Building Blocs on the Path to Global Free Trade.” World Economy 29 (11): 1451–1518. doi: 10.1111/j.1467-9701.2006.00852.x [Crossref], [Web of Science ®], [Google Scholar]
Baldwin, R. E. , and A. J.Venables . 1995. “Regional Economic Integration.” In Handbook of International Economics , edited by G. M.Grossman and K.Rogoff , 1597–1644. North-Holland : Elsevier. [Google Scholar]
Baldwin, R. E. , and C.Wyplosz . 2006. The Economics of European Integration . London : McGraw-Hill. [Google Scholar]
Bhagwati, J. 1993. “Regionalism and Multilateralism: An Overview.” In New Dimensions in Regional Integration , edited by J.de Melo and A.Panagariya , 22–51. Cambridge : Cambridge University Press. [Crossref], [Google Scholar]
Calvo-Pardo, H. , C.Freund, and E.Ornelas . 2011. “The ASEAN Free Trade Agreement: Impact on Trade Flows and External Trade Barriers.” In Costs and Benefits of Economic Integration in Asia , edited by R.Barro and J.Lee , 157–186. Oxford : Oxford University Press. [Crossref], [Google Scholar]
Chames, A. , W.Cooper, and E.Rhodes . 1978. “Measuring the Efficiency of Decision Making Units.” European Journal of Operational Research 2 (6): 429–444. doi: 10.1016/0377-2217(78)90138-8 [Crossref], [Web of Science ®], [Google Scholar]
Claveria, R. , and C. Y.Park . 2018. “Constructing the Asia-Pacific Regional Cooperation and Integration Index: A Panel Approach.” ADB Economics Working Paper, no. 544. Asian Development Bank, Manila. [Google Scholar]
Clements, B. 2002. “How Efficient is Education Spending in Europe?” European Review of Economics and Finance 1 (1): 3–26. [Google Scholar]
Coelli, T. J. 1996. “A Guide to DEAP Version 2.1: A Data Envelopment Analysis (Computer) Program.” CEPA Working Paper, No. 96/08. University of New England. [Google Scholar]
Coelli, T. J. , D. S.Rao, C. J.O’Donnell, and G. E.Battese . 2005. An Introduction to Efficiency and Productivity Analysis . New York : Springer Science and Business Media. [Google Scholar]
De Lombaerde, P. , E.Dorrucci, G.Genna, and F. P.Mongelli . 2008. “Quantitative Monitoring and Comparison of Regional Integration Processes: Steps towards Good Practice.” CRIS Working Paper No. W-2008/9. United Nations University, Tokyo. [Google Scholar]
Dür, A. , L.Baccini, and M.Elsig . 2014. “The Design of International Trade Agreements: Introducing a New Database.” Review of International Organizations 9 (3): 353–375. doi: 10.1007/s11558-013-9179-8 [Crossref], [Web of Science ®], [Google Scholar]
EU . 2018. Statement about Regional Integration on the website of the European Commission. https://ec.europa.eu/europeaid/node/680_bg . [Google Scholar]
Fernandez, R. , and J.Portes . 1998. “Returns to Regionalism: An Analysis of Nontraditional Gains from Regional Trade Agreements.” World Bank Economic Review 12 (2): 197–220. doi: 10.1093/wber/12.2.197 [Crossref], [Web of Science ®], [Google Scholar]
Francois, J. , and M.Manchin . 2013. “Institutions, Infrastructure, and Trade.” World Development 46: 165–175. doi: 10.1016/j.worlddev.2013.02.009 [Crossref], [Web of Science ®], [Google Scholar]
Freund, C. L. , and E.Ornelas . 2010. “Regional Trade Agreements.” World Bank Policy Research Working Paper, No. 5314. World Bank, Washington DC. [Google Scholar]
GCRI . 2017. Global Conflict Risk Index: Data and methodology. Joint Research Center of the European Commission. http://conflictrisk.jrc.ec.europa.eu . [Google Scholar]
Gupta, S. , and M.Verhoeven . 2001. “The Efficiency of Government Expenditure: Experiences from Africa.” Journal of Policy Modelling 23: 433–467. doi: 10.1016/S0161-8938(00)00036-3 [Crossref], [Web of Science ®], [Google Scholar]
Halkia, M. , S.Ferri, I.Joubert-Boitat, and F.Saporiti . 2017. Conflict Risk Indicators: Significance and Data Management in the GCRI . Ispra : Joint Research Centre. [Google Scholar]
Herrera, S. , and G.Pang . 2005. “Efficiency of Public Spending in Developing Countries: An Efficiency Frontier Approach.” World Bank Policy Research Working Paper, No. 3645. World Bank, Washington DC. [Google Scholar]
Huh, H. S. , and C. Y.Park . 2018. “Asia-Pacific Regional Integration Index: Construction, Interpretation, and Comparison.” Journal of Asian Economics 54: 22–38. doi: 10.1016/j.asieco.2017.12.001 [Crossref], [Web of Science ®], [Google Scholar]
Krugman, P. 1991. “The Move Toward Free Trade Zones.” Economic Review 76 (6): 5–25. [Google Scholar]
Latruffe, L. , K.Balcombe, S.Davidova, and K.Zawalinska . 2004. “Determinants of Technical Efficiency of Crop and Livestock Farms in Poland.” Applied Economics 36 (12): 1255–1263. doi: 10.1080/0003684042000176793 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
Marinov, E. 2015. “Economic Determinants of Regional Integration in Developing Counties.” International Journal of Business and Management 3 (3): 22–39. doi: 10.20472/BM.2015.3.3.003 [Crossref], [Google Scholar]
McKay, J. , M. O.Armengol, and G.Pineau . 2005. Regional Economic Integration in a Global Framework . Frankfurt : European Central Bank. [Google Scholar]
Naeher, D. 2015. “An Empirical Estimation of Asia's Untapped Regional Integration Potential Using Data Envelopment Analysis.” Asian Development Review 32 (2): 178–195. doi: 10.1162/ADEV_a_00056 [Crossref], [Google Scholar]
OECD . 2008. Handbook on Constructing Composite Indicators: Methodology and User Guide . Paris : OECD Publishing. [Crossref], [Google Scholar]
Peng, D. 2002. “Subregional Economic Zones and Integration in East Asia.” Political Science Quarterly 117 (4): 613–641. doi: 10.2307/798137 [Crossref], [Web of Science ®], [Google Scholar]
Ravallion, M. 2010. “Mashup Indices of Development.” World Bank Policy Research Working Paper, No. 54332, World Bank, Washington DC. [Crossref], [Google Scholar]
Sapir, A. 2011. “European Integration at the Crossroads: A Review Essay on the 50th Anniversary of Bela Balassa’s Theory of Economic Integration.” Journal of Economic Literature 49 (4): 1200–1229. doi: 10.1257/jel.49.4.1200 [Crossref], [Web of Science ®], [Google Scholar]
Sickles, R. C. , and V.Zelenyuk . 2019. Measurement of Productivity and Efficiency: Theory and Practice . Cambridge : Cambridge University Press. [Crossref], [Google Scholar]
Stiglitz, J. , A.Sen, and J. P.Fitoussi . 2009. Report by the Commission on the Measurement of Economic Performance and Social Progress. www.stiglitz-sen-fitoussi.fr . [Google Scholar]
Thanassoulis, E. 1993. “A Comparison of Regression Analysis and Data Envelopment Analysis as Alternative Methods for Performance Assessments.” Journal of the Operational Research Society 44 (11): 1129–1144. doi: 10.1057/jors.1993.185 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]
UN . 2017. World Population Prospects: The 2017 Revision . New York : United Nations Population Division. [Google Scholar]
UNECA . 2018. Statement about Regional Integration and Trade on the website of the United Nations Economic Commission for Africa. https://www.uneca.org/our-work/regional-integration-and-trade . [Google Scholar]
UNESCAP . 2016. Enhancing Regional Economic Cooperation and Integration in Asia and the Pacific . Note by the Secretariat: E/ESCAP/72/5. Bangkok : United Nations Economic and Social Commission for Asia and the Pacific. [Google Scholar]
Yoshimatsu, H. 2002. “Preferences, Interests, and Regional Integration: The Development of the ASEAN Industrial Cooperation Arrangement.” Review of International Political Economy 9 (1): 123–149. doi: 10.1080/09692290110101117 [Taylor & Francis Online], [Web of Science ®], [Google Scholar]

Appendices

Appendix A. List of economies and subregional groupings

Economies are grouped into geographical regions and subregions following the UN’s (2017) classification (the only exception is Azerbaijan, which we include in Central Asia to be more in line with World Bank classifications). The total sample consists of 193 economies which are grouped into the following geographical regions and subregions (number of economies in parentheses):

Europe (39):

Eastern Europe (10): Belarus, Bulgaria, Czechia, Hungary, Moldova, Poland, Romania, Russia, Slovakia, Ukraine
Northern Europe (10): Denmark, Estonia, Finland, Iceland, Ireland, Latvia, Lithuania, Norway, Sweden, United Kingdom
Southeastern Europe (12): Albania, Bosnia and Herzegovina, Croatia, Greece, Italy, Macedonia, Malta, Montenegro, Portugal, Serbia, Slovenia, Spain
Western Europe (7): Austria, Belgium, France, Germany, Luxembourg, Netherlands, Switzerland

Americas (37):

Caribbean (15): Antigua and Barbuda, Aruba, Bahamas, Barbados, Cuba, Dominica, Dominican Republic, Grenada, Haiti, Jamaica, Puerto Rico, Saint Lucia, Saint Kitts and Nevis, Saint Vincent and the Grenadines, Trinidad and Tobago
Central America (8): Belize, Costa Rica, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama
North America (2): Canada, United States
South America (12): Argentina, Bolivia, Brazil, Chile, Colombia, Ecuador, Guyana, Paraguay, Peru, Suriname, Uruguay, Venezuela

Asia (65):

Central Asia (6): Azerbaijan, Kazakhstan, Kyrgyzstan, Tajikistan, Turkmenistan, Uzbekistan
East Asia (8): China, China Hong Kong SAR, China Macau SAR, DPR Korea, Japan, Mongolia, South Korea, Taiwan
South Asia (9): Afghanistan, Bangladesh, Bhutan, India, Iran, Maldives, Nepal, Pakistan, Sri Lanka
Pacific and Oceania (15): Australia, Cook Islands, Fiji, Kiribati, Marshall Islands, Micronesia, Nauru, New Zealand, Palau, Papua New Guinea, Samoa, Solomon Islands, Tonga, Tuvalu, Vanuatu
Southeast Asia (11): Brunei Darussalam, Cambodia, Indonesia, Lao PDR, Malaysia, Myanmar, Philippines, Singapore, Thailand, Timor-Leste, Vietnam
West Asia (16): Armenia, Bahrain, Cyprus, Georgia, Iraq, Israel, Jordan, Kuwait, Lebanon, Oman, Qatar, Saudi Arabia, Syria, Turkey, United Arab Emirates, Yemen

Africa (52):

Eastern Africa (17): Burundi, Djibouti, Eritrea, Ethiopia, Kenya, Madagascar, Mauritius, Malawi, Mozambique, Rwanda, Seychelles, Somalia, South Sudan, Tanzania, Uganda, Zambia, Zimbabwe
Middle Africa (8): Angola, Cameroon, Central African Republic, Chad, Dem. Rep. of the Congo, Congo, Equatorial Guinea, Gabon
Northern Africa (6): Algeria, Egypt, Libya, Morocco, Sudan, Tunisia
Southern Africa (5): Botswana, Lesotho, Namibia, South Africa, Swaziland
Western Africa (16): Benin, Burkina Faso, Cabo Verde, Cote d'Ivoire, The Gambia, Ghana, Guinea, Guinea Bissau, Liberia, Mali, Mauritania, Niger, Nigeria, Senegal, Sierra Leone, Togo

Appendix B. Handling of missing data

The analysis is affected by two types of data unavailability: missing values for individual economy-pair observations and complete absence of economies in some of the original datasets. To address the first issue of single missing values, we augment the data with information from the previous year (i.e. missing values are imputed with the corresponding values from the previous year if available) for those variables that are most affected, i.e. indicators I.b, II.a, II.b, and III.a. Besides this procedure, no additional imputations for single missing values are performed.

Regarding absent economies, none of the variables we use provides information on all 193 economies included in the analysis. However, in most cases coverage is well above 95 percent. The only variables that cover less than 95 percent of economies are indicators II.a, II.b,, III.a, V.a and V.b of the CRI index (i.e. FDI positions and the two indicators for financial integration and peace and security, respectively), as well as the Logistics Performance Index, which is used as input variable in the DEA.

For indicator III.a, data on bilateral FDI positions from the IMF’s Coordinated Direct Investment Survey (CDIS) are only available for 115 economies (around 60 percent). To increase coverage, we use data from UNCTAD’s Bilateral FDI Statistics 2014 to add information for those countries missing in the CDIS. This leads to a total of 192 economies represented in the respective indicator for FDI. Similarly, we use data from the UN’s Commodity Trade Statistics Database (COMTRADE) to impute the intraregional share of intermediate goods exports (III.b) for Southern Africa, the only subregion that is not sufficiently covered in the WITS dataset.

For indicators II.a and II.b, data are only available for 75 and 76 economies, respectively (around 40 percent). Coverage is particularly low for the Caribbean, Pacific & Oceania, and the subregions in Africa, where in some cases data are only available for a single economy within a subregion. For Western and Middle Africa, no data at all are available. As we were unable to identify an alternative data source, we impute the missing values for these two subregions by using the respective mean values across the other three African subregions. While this approach certainly provides only a very rough approximation, we believe it likely helps to reduce the potential bias that might otherwise occur if the CRI index was computed without the dimension of financial integration for these two regions. When we compute the CRI index without the financial integration dimension at all, the resulting ranks of the two affected subregions change only marginally (the rank of Western Africa changes from 13 to 11 and the rank of Middle Africa remains unaffected) and only one other subregion changes ranks by more than two (West Asia changes from 11 to 14). This suggests that the imputed values for dimension II are not driving the overall results.

For the two indicators of dimension V., data are only available for 136 economies (70 percent), as most of the European economies are not included in the Global Conflict Risk Index. For the Logistics Performance Index, data are available for 165 economies (85 percent). As all of the subregions are covered, we abstain from additional imputations for these three indicators and simply compute the respective values based on the subsets of economies available for each subregion.

Table A1. Regional integration and enabling environment: disaggregated by dimensions.

CSV Display Table

Figure A1. Global CRI index: dimensions by subregions.

Display full size

Table A2. Robustness tests: principal component analysis.

CSV Display Table

The Journal of International Trade & Economic Development

Original Articles

Untapped regional integration potential: A global frontier analysis

Original Articles

Untapped regional integration potential: A global frontier analysis

ABSTRACTFormulae display:?Mathematical formulae have been encoded as MathML and are displayed in this HTML version using MathJax in order to improve their display. Uncheck the box to turn MathJax off. This feature requires Javascript. Click on a formula to zoom.

1. Introduction

2. Methodology

2.1. Composite regional integration index

Published online:

Published online:

Table 1. Variable descriptions and data sources.

2.2. Data envelopment analysis

2.2.1. Market-led processes

2.2.2. Institution-led processes

2.2.3. Private sector-led processes

3. Data and variables

4. Empirical results

4.1. Global comparison of regional integration levels

Published online:

Published online:

Published online:

4.2. Frontier analysis of untapped integration potential

Published online:

Table 2. Data envelopment analysis estimates of untapped integration potential.

Published online:

5. Robustness

Published online:

Table 3. Robustness of CRI index to different weighting schemes, PCA, and exclusion of dimension V.

Published online:

Table 4. Robustness of DEA input index to different weighting schemes and PCA.

6. Conclusion and policy implications

Disclosure statement

Notes

References

Appendix A. List of economies and subregional groupings

Europe (39):

Americas (37):

Asia (65):

Africa (52):

Appendix B. Handling of missing data

Published online:

Table A1. Regional integration and enabling environment: disaggregated by dimensions.

Published online:

Published online:

Table A2. Robustness tests: principal component analysis.

Related articles

Information for

Open access

Help and info

Keep up to date

ABSTRACT
Formulae display:?Mathematical formulae have been encoded as MathML and are displayed in this HTML version using MathJax in order to improve their display. Uncheck the box to turn MathJax off. This feature requires Javascript. Click on a formula to zoom.