Methodology

The process of constructing the index involves data collection, imputation of missing data, normalization through panel min–max scaling, and weighting and aggregation using a two-step panel principal component analysis (PCA) procedure.



DATA COLLECTION

The first step is to collect the historical data for all economies to be used in computing the indicators. The data are updated annually to ensure the latest information is used to generate the index.



INDICATOR COMPUTATION AND IMPUTATION

To fill in data gaps, various imputation methods are implemented: linear interpolation, averaging, regression imputation, and substitution by closest available observation. A more detailed explanation of the various imputation methods is presented in Huh and Park (2018).



NORMALIZATION

As measurement units differ across indicators, normalization is required prior to aggregation to achieve commensurability. Panel min–max normalization involves normalizing indicators based on all sample years and all economies at hand. It is suitable for comparing progress across different economies through the years.



PANEL PRINCIPAL COMPONENT ANALYSIS (PCA)

A two-stage panel PCA estimation is carried out to obtain time-consistent weights for aggregation.

In the first stage, panel PCA is applied to each of the dimensions independently, and the implied weights assigned to the indicators are used for constructing a set of dimensional composite indexes.

In the second stage, panel PCA is applied again to weigh the dimensional indexes to produce an overall index, which is the ARCII.

A noticeable difference between the enhanced versus the baseline framework is the normalization of indicators, which is determined by the nature of comparison or analysis being made on the ARCII results.

In the enhanced framework, the indicators are normalized based on all regions together (i.e., worldwide normalization) prior to performing the PCA.

Weighting and aggregation of data are intended to improve comparability by utilizing all information at hand. The most important source of incomparability is said to be the systemic statistical properties of the index components that affect weighting (UNCTAD 2007). PCA remains the main statistical tool to obtain the weights utilized for compiling the ARCII.

The two-stage PCA is implemented separately for each region to account for the different features and characteristics of the region in the computation of the index.

Learn more in the ARCII Report >

Economies covered

The 48 Asian economies covered in this study are classified by subregion, according to the Asian Development Bank categorization (number of economies in parentheses):

Central Asia (8): Armenia, Azerbaijan, Georgia, Kazakhstan, the Kyrgyz Republic, Tajikistan, Turkmenistan, Uzbekistan
East Asia (6): People’s Republic of China; Hong Kong, China; Japan; the Republic of Korea; Mongolia; Taipei,China
Southeast Asia (10): Brunei Darussalam, Cambodia, Indonesia, the Lao People’s Democratic Republic (Lao PDR), Malaysia, Myanmar, the Philippines, Singapore, Thailand, Viet Nam
South Asia (8): Afghanistan, Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, Sri Lanka
The Pacific (14): Cook Islands, Fiji, Kiribati, Marshall Islands, Micronesia, Nauru, Palau, Papua New Guinea, Samoa, Solomon Islands, Timor-Leste, Tonga, Tuvalu, Vanuatu
Oceania (2): Australia, New Zealand

To allow for comparison with other regions, integration indexes were also built for other regions of the world, utilizing all available information for all the economies and years in the sample period (i.e., 173 economies for 2006–2018) in the data imputation, normalization, and weighting and aggregation. Following ADB classification, the countries covered in other regions are (number of countries in parentheses):