## _wp15155

## Source details

**Canonical URL:** [_wp15155](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15155.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15155.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15155.pdf.json)

---

### Introduction and motivation
- Emerging Market Economies (EMs) increased their role in the global economy over the past five decades, representing over 50 percent of total global output at purchasing power parity (PPP).
- EMs account for over half of the global population.
- The study develops new clusters and a taxonomy to explain heterogeneity in long-term growth, its drivers, and post-2000 dynamics.

### Key high-level findings
- EMs’ share of global PPP-adjusted GDP increased from 27 percent in 1960 to around 53 percent by 2013.
- Shares in 2013: Advanced Markets (AMs) 44 percent, EMs 52.7 percent, Low-Income Countries (LICs) 3.3 percent.
- Five EMs (China, India, Russia, Brazil, and Mexico) account for about half of EMs’ total outputs as of 2013.
- Using middle-income thresholds (5 and 40 percent of U.S. GDP per capita, PPP-adjusted), 64 percent of EMs are classified as stuck in the “middle-income trap.”
- Growth clusters explain convergence better than geographic classification.
- Strong correlations documented:
  - Investment clusters highly correlated with growth clusters.
  - Total Factor Productivity (TFP) growth clusters strongly correlated with real GDP growth clusters.
- Investment-growth-unemployment nexus:
  - High investment-growth groups show low unemployment rates on average.
  - Low investment-growth groups display high unemployment rates on average.

### Cluster methodology and sample
- Sample: 25 major EMs for long-term cluster analysis (1980–2013).
- Method: Ward’s linkage hierarchical clustering.
- Real GDP growth series averaged over five-year intervals (seven intervals from 1980 to 2013); sensitivity tests with annual data largely consistent.
- Four clusters chosen after sensitivity tests.

### Clusters measured by real GDP growth (4 clusters) — membership and summary statistics
- Cluster 1 (Low Growth): Brazil, Colombia, Hungary, Mexico, Poland, and South Africa.
  - Real GDP growth: Mean 2.49, Std. dev 1.91, min -3.22, max 5.91, # of Obs 42
- Cluster 2 (Increasing Growth): Argentina, Peru, Philippines, Saudi Arabia, Turkey, Uruguay, and Venezuela.
  - Real GDP growth: Mean 3.17, Std. dev 2.49, min -2.24, max 6.83, # of Obs 49
- Cluster 3 (Declining Growth): Chile, Egypt, Hong Kong SAR, India, Indonesia, Israel, Korea, Malaysia, Pakistan, Singapore, and Thailand.
  - Real GDP growth: Mean 5.35, Std. dev 1.88, min 0.73, max 10.08, # of Obs 77
- Cluster 4 (High Growth): China.
  - Real GDP growth: Mean 9.84, Std. dev 0.96, min 8.77, max 11.40, # of Obs 7
- Interpretation notes:
  - Cluster 1: annual growth around 2.5 percent.
  - Cluster 2: started slow, accelerated later with large variation.
  - Cluster 3: began fast then declined and stabilized around 4 percent.
  - Cluster 4: exceptional high growth (China).

### Two-variable clusters: Growth plus Investment, Consumption, and TFP — key summary statistics
- Growth and Investment clusters (real GDP growth and investment share % of GDP):
  - Cluster 1 (low investment): real GDP growth Mean 3.29, Std. dev 2.15, min -3.22, max 7.28, #Obs 98
    - Investment share: Mean 18.30, Std. dev 3.65, min 10.90, max 28.67, #Obs 98
  - Cluster 2 (increasing investment): real GDP growth Mean 4.06, Std. dev 2.80, min -1.64, max 8.14, #Obs 14
    - Investment share: Mean 26.84, Std. dev 4.77, min 20.62, max 33.68, #Obs 14
  - Cluster 3 (declining investment): real GDP growth Mean 5.66, Std. dev 2.28, min 1.54, max 9.14, #Obs 21
    - Investment share: Mean 27.16, Std. dev 4.67, min 20.47, max 40.22, #Obs 21
  - Cluster 4 (high investment): real GDP growth Mean 8.25, Std. dev 2.90, min 3.37, max 13.05, #Obs 14
    - Investment share: Mean 32.61, Std. dev 5.56, min 25.13, max 44.64, #Obs 14
- Growth and Consumption clusters (real GDP growth and consumption share % of GDP):
  - Real GDP growth by consumption cluster:
    - Cluster 1 (low consumption): Mean 5.95, Std. dev 3.16, min 0.73, max 13.05, #Obs 35
    - Cluster 2 (increasing consumption): Mean 2.40, Std. dev 2.33, min -3.22, max 6.23, #Obs 42
    - Cluster 3 (declining consumption): Mean 5.72, Std. dev 1.93, min 1.61, max 10.08, #Obs 21
    - Cluster 4 (high consumption): Mean 3.92, Std. dev 1.57, min -0.18, max 6.83, #Obs 42
  - Consumption share in GDP by cluster (percent):
    - Cluster 1: Mean 60.51, Std. dev 8.20, min 45.07, max 77.52, #Obs 35
    - Cluster 2: Mean 79.22, Std. dev 5.16, min 66.66, max 88.88, #Obs 42
    - Cluster 3: Mean 75.52, Std. dev 6.23, min 65.77, max 88.94, #Obs 21
    - Cluster 4: Mean 84.35, Std. dev 7.97, min 61.28, max 107.43, #Obs 42
- Growth and TFP clusters — two reported decompositions:
  - TFP growth clusters (set 1):
    - Cluster 1 (low and sluggish TFP growth): Mean 2.48, Std. dev 1.95, min -3.22, max 6.23, #Obs 42
    - Cluster 2 (low and volatile TFP growth): Mean 2.88, Std. dev 2.77, min -2.24, max 6.83, #Obs 35
    - Cluster 3 (moderate TFP growth): Mean 4.97, Std. dev 1.97, min 0.73, max 10.08, #Obs 63
    - Cluster 4 (high TFP growth): Mean 6.72, Std. dev 2.35, min 3.16, max 11.40, #Obs 28
  - TFP growth clusters (set 2, alternative decomposition):
    - Cluster 1 (low and sluggish TFP growth): Mean -0.22, Std. dev 1.92, min -5.15, max 4.41, #Obs 42
    - Cluster 2 (low and volatile TFP growth): Mean -0.32, Std. dev 3.14, min -9.68, max 6.21, #Obs 35
    - Cluster 3 (moderate TFP growth): Mean 0.29, Std. dev 1.48, min -3.62, max 3.32, #Obs 63
    - Cluster 4 (high TFP growth): Mean 1.57, Std. dev 1.80, min -1.50, max 5.17, #Obs 28
  - Exact quantitative note: Low and Sluggish TFP Growth Group reported average -0.22% annual TFP growth in one aggregation and real GDP growth rate is 2.5%, the lowest among all groups; High TFP Growth Group reported TFP growth around 1.6 percent annual in one aggregation.

### Investment–growth–unemployment and boom–bust dynamics
- Investment and growth stylized outcomes:
  - Low investment group: investment share < 20 percent of GDP and growth underperforms (real GDP growth Mean 3.29).
  - Increasing investment group: investment share surpassed 30 percent after 2000 and growth accelerated.
  - Declining investment group: experienced boom-bust cycles (1997 Asian Crisis) with falling investment share and subsequent growth slowdown.
- External financing of investment booms:
  - Investment boom year T defined when deviation from trend of investment share exceeds 2 standard deviations of the cyclical component (HP filter smoothing parameter = 100).
  - Investment booms financed through debt rather than equity associated with more severe growth collapse and sharper increases in unemployment when booms bust.
  - Examples of equity-financed episodes: Czech Rep. 2008, Colombia 1994, Egypt 1990, India 2007, Indonesia 1997, Malaysia 1997, Pakistan 2006, Poland 1999, Singapore 1984.
  - Examples of debt-financed episodes: Chile 1981, Mexico 1981, Korea 1997, Philippines 1983, South Africa 2008, Thailand 1996, Uruguay 1981.
- Caveats:
  - Five-year averages used to smooth volatility; boom–bust cycles increase dissimilarity and can place separate episodes in same cluster.
  - Endogeneity concern: over-investment and strong growth often followed by under-investment and weak growth within boom–bust cycles; averaging does not fully resolve this.

### Convergence patterns and cluster dynamics
- Convergence by growth-cluster:
  - Low Growth Group: GDP per capita relative to U.S. almost stagnant between 25 and 30 percent.
  - Increasing Growth Group: uneven path; real GDP per capita growth ranged between 6 percent and -3 percent across episodes.
  - Declining Growth Group: average per capita income relative to U.S. improved from 15 to 45 percent over five decades.
  - High Growth Group: China only; extraordinarily high growth shortened catch-up.
- Clusters yield more homogeneity in convergence patterns than geographic groupings.

### Taxonomy (domestic and external angles) and post-2000 dynamics
- Domestic-angle taxonomy (supply and demand decomposition, median values 2000–2013):
  - Supply-side decomposition: physical capital contribution, labor contribution, TFP contribution to real output growth; countries ranked as “capital-driven”, “labor-driven”, “technology-driven”.
  - Demand-side decomposition: consumption contribution, investment contribution, net exports contribution; countries ranked as “consumption-led”, “investment-led”, “export-led”.
  - * marks countries that ran trade deficits through 2000-2012 but whose net exports had been improving and therefore contributed positively to GDP growth.
- External-angle taxonomy:
  - Grouped on financial openness, trade openness and linkages, terms of trade growth, and net commodity dependence.
  - Observations: Hong Kong SAR and Singapore have extraordinarily high levels of financial and trade openness.
  - Only 6 economies with relatively high levels of financial openness are top-tier versus 23 economies bottom-tier.
  - Only 9 EMs are large net commodity exporters; 31 economies run either a small net commodity export surplus or a net commodity export deficit.
- Composite external-exposure index:
  - Components: financial openness, trade openness, terms of trade growth, net commodity export to GDP.
  - Standardized via global min-max scaling to range 0–1 for each component.
  - Weights endogenized via principal component analysis (first principal component squared factor loadings).
  - Higher index = more exposed to world economy.

### Impact of external exposure, terms of trade, and openness (post-2000 evidence)
- Pre- vs post-GFC growth by external-exposure tertile:
  - Top tertile: post-crisis growth dropped by 6.4 percentage points from 8.4 percent to 2.0 percent.
  - Middle tertile: growth rate dropped by 3.2 percentage points from 5.8 percent to [value truncated in source excerpt].
  - Bottom tertile: growth rate dropped by 2.3 percentage points from 6.6 percent to 4.3 percent.
- Growth surprises (actual minus WEO one-year-ahead projection) during 2008-2012:
  - Bottom tertile: -0.5 percentage points.
  - Middle tertile: -1.1 percentage points.
  - Top tertile: -2.7 percentage points.
- Business cycle synchronization (correlation with world real GDP growth):
  - Bottom tertile: 0.63
  - Middle tertile: 0.66
  - Top tertile: 0.75
- Terms of trade effects:
  - Average elasticity of EM commodity exporters’ growth to terms of trade changes is about 0.14 and statistically significant.
  - Impact on non-commodity-exporters is minimal and statistically insignificant.
  - When terms of trade growth adjusted by trade openness, positive relationship on growth: openness magnifies terms-of-trade shocks for commodity exporters.
- Openness amplifies transmission of terms-of-trade shocks and external adjustment pressure on growth; cumulative current account deficits adjusted by financial openness magnify external adjustment pressure.

### Trade linkages and spillover effects
- Evolution of export destinations (averages):
  - 1990s average EM exports to Euro area 27.3 percent, to U.S. 17.3 percent, to China 2.3 percent.
  - Ten years later: exports to U.S. 28 percent, Euro area 17.5 percent, China 5.4 percent.
  - EMs' average export share to China increased by 132 percent.
- Growth performance by primary trade partner (Avg Real GDP growth %, 2000-07 and 2008-13 for top cluster countries):
  - Links with Euro Area: 5.6 (2000-07), 5.3 (2008-13)
  - Links with U.S.: 7.1 (2000-07), 1.4 (2008-13)
  - Links with China: 3.9 (2000-07), 4.4 (2008-13)
- Findings:
  - Countries trading more with China on average experienced faster growth than those trading more with Euro area or U.S., especially post-crisis.
  - Trade linkages generate spillovers: a country’s growth can benefit from or be dragged by trading partners.
- Complementary evidence:
  - Stronger trade linkage with China associated with less market volatility during 2013–2014 tapering talk (Mishra et al. 2014).

### Policy implications and conclusions
- EMs’ rise reshaped the global economic landscape; heterogeneity across EMs is substantial.
- Clusters and taxonomy provide actionable grouping beyond geography for policy design.
- Investment is necessary but not sufficient for growth; investment clusters are strongly associated with growth and unemployment outcomes.
- TFP improvements accompany higher growth clusters.
- Post-crisis recovery: investment-led countries rebounded faster and stronger; consumption-led countries rebounded slower and weaker.
- External exposure matters: degree of slowdown, growth surprise, and business-cycle synchronization positively correlated with external factor index.
- Openness amplifies transmission of external shocks (terms-of-trade, current account pressures).
- Policy implication: continuing convergence to high-income levels for EMs will be more challenging and is not guaranteed; tailored policy actions needed to account for heterogeneity beyond traditional geographic and income approaches.
- The clusters and taxonomy can be used as a reference to design both near- and long-term policies.

*Source: _wp15155 - 1. Output as Share of World GDP (PPP-adjusted) and related sections (IMF staff paper content provided).*

### 1. Output as Share of World GDP (PPP-adjusted)  ...................................................................6

### 1. Output as Share of World GDP (PPP-adjusted)

### Introduction and motivation
- Emerging Market Economies (EMs) have increased their role in the global economy over the past five decades, currently representing over 50 percent of total global output at purchasing power parity (PPP).
- EMs account for over half of the global population.
- Heterogeneity across EMs is substantial: some EMs converged toward U.S. income levels since 1960, while others remained stagnant; the Global Financial Crisis (GFC) had differential impacts across EMs.
- This study develops new clusters and a taxonomy to explain heterogeneity in long-term growth, its drivers, and post-2000 dynamics.

### Key high-level findings
- EMs’ share of global PPP-adjusted GDP increased from 27 percent in 1960 to around 53 percent by 2013.
- The shares of AMs, EMs and LICs in 2013 are 44, 52.7, and 3.3 percent, respectively.
- Five countries (China, India, Russia, Brazil, and Mexico) account for about half of EMs’ total outputs as of 2013.
- Using the middle-income thresholds (5 and 40 percent of U.S. GDP per capita, PPP-adjusted), 64 percent of EMs are classified as stuck in the “middle-income trap.”
- The paper covers 25 major EMs for long-term cluster analysis (1980–2013) using Ward’s linkage method; four clusters chosen based on sensitivity tests.

### Contributions of the paper
- Proposes clusters based on long-term development trends in growth and its driving factors over five decades.
- Shows that growth clusters have more explanatory power for convergence than geographic classification.
- Finds strong correlations:
  - Investment clusters are highly correlated with growth clusters.
  - Total Factor Productivity (TFP) growth clusters and real GDP growth clusters are strongly correlated.
- Documents an investment-growth-unemployment nexus: on average, high investment-growth groups show low unemployment rates, while low investment-growth groups display high unemployment rates.
- Provides a taxonomy from a domestic angle (revealed factor endowments) and an external angle (linkages since 2000) that helps explain post-GFC recovery, the impact of external factors on growth, amplification effects of openness, and trade spillovers.

### Role of openness and external linkages
- Degree of economic slowdown, growth surprise, and business-cycle synchronization are positively correlated with the external factor index.
- Increased openness amplifies transmission of terms-of-trade shocks and external adjustment pressure on growth when terms-of-trade changes are adjusted by commodity export and trade openness measures, and when cumulative current account deficits are adjusted by financial openness.
- Trade linkages generate significant spillover effects: one country’s growth can benefit from or be dragged by its trading partners.

---

### Long-term development and cluster analysis — Stylized facts
- Over 1960–2013, EMs grew faster than AMs and increased their share of world GDP from 27 percent to around 53 percent by 2013.
- As of 2013, the PPP-adjusted GDP sizes of China, India, Russia, Brazil, and Mexico were comparable to those of the United States, Japan, Germany, France, and the United Kingdom.
- Only a few EMs advanced to high-income status since 1960; many remain in middle-income status.

### Cluster methodology (overview)
- Sample: 25 major EMs selected based on economic size, growth rate and data availability for 1980–2013.
- Method: Ward’s linkage hierarchical clustering.
- Variable construction for one-variable clusters: real GDP growth rate averaged over five-year intervals (seven intervals from 1980 to 2013). Sensitivity analysis with annual data produces largely consistent results.
- Four clusters chosen after sensitivity tests.

### Clusters measured by real GDP growth (4 clusters)
- Cluster definitions (by country membership in Figure 3):
  - Cluster 1: Brazil, Colombia, Hungary, Mexico, Poland, and South Africa (Low Growth)
  - Cluster 2: Argentina, Peru, Philippines, Saudi Arabia, Turkey, Uruguay, and Venezuela (Increasing Growth)
  - Cluster 3: Chile, Egypt, Hong Kong SAR, India, Indonesia, Israel, Korea, Malaysia, Pakistan, Singapore, and Thailand (Declining Growth)
  - Cluster 4: China (High Growth)

- Summary statistics of cluster output (real GDP growth):
  - Cluster 1: low growth — Mean 2.49, Std. dev 1.91, min -3.22, max 5.91, # of Obs 42
  - Cluster 2: increasing growth — Mean 3.17, Std. dev 2.49, min -2.24, max 6.83, # of Obs 49
  - Cluster 3: declining growth — Mean 5.35, Std. dev 1.88, min 0.73, max 10.08, # of Obs 77
  - Cluster 4: high growth — Mean 9.84, Std. dev 0.96, min 8.77, max 11.40, # of Obs 7

- Interpretation:
  - Cluster 1: annual growth around 2.5 percent.
  - Cluster 2: started slow, accelerated later but with large variation.
  - Cluster 3: began with fast growth then declined and stabilized around 4 percent.
  - Cluster 4: China’s growth is exceptionally high across the sample period.
  - Forcing two clusters would combine clusters 1 and 2 into one group, and clusters 3 and 4 into another.

### Clusters measured by growth plus additional variables
- Approach: two-variable cluster analyses adding each of investment share (% of GDP), consumption share (% of GDP), and TFP growth separately to real GDP growth to study interactions between output and key contributors.

- Growth and Investment clusters (summary statistics from Table 1):
  - Cluster 1: low investment — real GDP growth Mean 3.29, Std. dev 2.15, min -3.22, max 7.28, #Obs 98
  - Cluster 2: increasing investment — real GDP growth Mean 4.06, Std. dev 2.80, min -1.64, max 8.14, #Obs 14
  - Cluster 3: declining investment — real GDP growth Mean 5.66, Std. dev 2.28, min 1.54, max 9.14, #Obs 21
  - Cluster 4: high investment — real GDP growth Mean 8.25, Std. dev 2.90, min 3.37, max 13.05, #Obs 14

- Investment share in GDP by cluster (percent):
  - Cluster 1: low investment — Mean 18.30, Std. dev 3.65, min 10.90, max 28.67, #Obs 98
  - Cluster 2: increasing investment — Mean 26.84, Std. dev 4.77, min 20.62, max 33.68, #Obs 14
  - Cluster 3: declining investment — Mean 27.16, Std. dev 4.67, min 20.47, max 40.22, #Obs 21
  - Cluster 4: high investment — Mean 32.61, Std. dev 5.56, min 25.13, max 44.64, #Obs 14

- Growth and Consumption clusters (summary statistics from Table 1):
  - Real GDP growth by consumption cluster:
    - Cluster 1: low consumption — Mean 5.95, Std. dev 3.16, min 0.73, max 13.05, #Obs 35
    - Cluster 2: increasing consumption — Mean 2.40, Std. dev 2.33, min -3.22, max 6.23, #Obs 42
    - Cluster 3: declining consumption — Mean 5.72, Std. dev 1.93, min 1.61, max 10.08, #Obs 21
    - Cluster 4: high consumption — Mean 3.92, Std. dev 1.57, min -0.18, max 6.83, #Obs 42
  - Consumption share in GDP by cluster (percent):
    - Cluster 1: low consumption — Mean 60.51, Std. dev 8.20, min 45.07, max 77.52, #Obs 35
    - Cluster 2: increasing consumption — Mean 79.22, Std. dev 5.16, min 66.66, max 88.88, #Obs 42
    - Cluster 3: declining consumption — Mean 75.52, Std. dev 6.23, min 65.77, max 88.94, #Obs 21
    - Cluster 4: high consumption — Mean 84.35, Std. dev 7.97, min 61.28, max 107.43, #Obs 42

- Growth and TFP clusters (summary statistics from Table 1):
  - TFP growth clusters (first set):
    - Cluster 1: low and sluggish TFP growth — Mean 2.48, Std. dev 1.95, min -3.22, max 6.23, #Obs 42
    - Cluster 2: low and volatile TFP growth — Mean 2.88, Std. dev 2.77, min -2.24, max 6.83, #Obs 35
    - Cluster 3: moderate TFP growth — Mean 4.97, Std. dev 1.97, min 0.73, max 10.08, #Obs 63
    - Cluster 4: high TFP growth — Mean 6.72, Std. dev 2.35, min 3.16, max 11.40, #Obs 28
  - TFP growth clusters (second set, possibly alternative decomposition):
    - Cluster 1: low and sluggish TFP growth — Mean -0.22, Std. dev 1.92, min -5.15, max 4.41, #Obs 42
    - Cluster 2: low and volatile TFP growth — Mean -0.32, Std. dev 3.14, min -9.68, max 6.21, #Obs 35
    - Cluster 3: moderate TFP growth — Mean 0.29, Std. dev 1.48, min -3.62, max 3.32, #Obs 63
    - Cluster 4: high TFP growth — Mean 1.57, Std. dev 1.80, min -1.50, max 5.17, #Obs 28

### Interrelationships and stylized outcomes
- Investment and growth:
  - Investment clusters are highly correlated with growth clusters.
  - Low investment group: investment share < 20 percent of GDP and growth underperforms other groups (real GDP growth Mean 3.29).
  - Increasing investment group: investment share surpassed 30 percent after 2000 and growth accelerated.
  - Declining investment group: experienced boom-bust cycles (1997 Asian Crisis) with falling investment share and subsequent growth slowdown.
- TFP and growth:
  - TFP growth clusters and real GDP growth clusters are strongly correlated.
- Investment-growth-unemployment nexus:
  - On average, countries in the high investment-growth group show low unemployment rates.
  - Countries in the low investment-growth group display high unemployment rates.

### Taxonomy, external factors, and post-2000 dynamics
- Taxonomy constructed along:
  - Domestic angle: revealed factor endowments.
  - External angle: real and financial linkages since 2000.
- Usefulness of taxonomy:
  - Domestic taxonomy explains the degree of economic recovery by distinguishing consumption-led vs. investment-led economies.
  - External taxonomy interprets pre- and post-GFC growth dynamics: degree of slowdown, growth surprises, and business-cycle synchronization are positively correlated with the external factor index.
- Amplification and spillovers:
  - Increased openness amplifies effects of terms-of-trade shocks and external adjustment pressure on growth.
  - Spillover effects from trade linkages can significantly benefit or drag a country’s growth depending on trading partners’ performance.

*Source: _wp15155 - 1. Output as Share of World GDP (PPP-adjusted) (IMF staff paper content provided).*

### 2000. In the High Investment Group, investment made up more than 30 percent of total

### _wp15155 - 2000. In the High Investment Group, investment made up more than 30 percent of total

### Clusters by GDP Growth and Investment
- Four investment-growth clusters identified: Low Investment, Increasing Investment, Declining Investment, High Investment (Figure 4).
- Key descriptive findings:
  - In the High Investment Group, investment made up more than 30 percent of total output, and its growth rate is also the highest among all groups.
  - Cluster country compositions (from Figure 4):
    - Cluster 1: Argentina, Brazil, Chile, Colombia, Hungary, Israel, Mexico, Pakistan, Peru, Philippines, Poland, South Africa, Turkey, and Uruguay.
    - Cluster 2: India and Venezuela.
    - Cluster 3: Hong Kong SAR, Singapore, and Thailand.
    - Cluster 4: China and Korea.
- Visual summary elements preserved: 5Y avg real GDP Gr (%), 5Y avg of real investment ( % of GDP), 25-75th Percentile and Median markers shown in Figure 4.

### Clusters by GDP Growth and Consumption
- Four consumption-growth clusters identified: Low Consumption Group, Increasing Consumption Group, Declining Consumption Group, High Consumption Group (Figure 5).
- Quantitative cluster characteristics:
  - Low Consumption Group: consumption share to GDP is the lowest––around 60 percent, while its growth rate is relatively high, especially before 2000.
  - Increasing Consumption Group: consumption share has risen gradually to more than 80 percent; growth rate has been the most disappointing though it accelerated moderately in recent years.
  - Declining Consumption Group: both consumption share and growth rate have gradually declined over time.
  - High Consumption Group: consumption share is around 85 percent, and growth rate remains stable around 4 percent.
- Cluster country compositions (from Figure 5):
  - Cluster 1: Chile, China, Mexico, Singapore, and Thailand.
  - Cluster 2: Hungary, Philippines, Poland, South Africa, Uruguay, and Venezuela.
  - Cluster 3: India, Hong Kong SAR, and Korea.
  - Cluster 4: Brazil, Colombia, Israel, Pakistan, Peru, and Turkey.

### Clusters by GDP Growth and TFP Growth
- Four TFP-growth clusters identified: Low and Sluggish TFP Growth Group, Low and Volatile TFP Growth Group, Moderate TFP Growth Group, High TFP Growth Group (Figure 6).
- Exact quantitative summaries:
  - Low and Sluggish TFP Growth Group: on average -0.22% annual TFP growth; average real GDP growth rate is 2.5%, the lowest among all groups.
  - Low and Volatile TFP Growth Group: similar characteristics to the previous group but shows more volatility.
  - Moderate TFP Growth Group: modest levels of TFP growth and real GDP growth.
  - High TFP Growth Group: TFP growth rate is the largest –––around 1.6 percent on an annual basis; its growth rate is also the largest among all groups.
- Note on aggregation: authors state first two groups can be combined into one "Low TFP Growth Group" and last two into "High TFP Growth Group" based on dissimilarity levels.
- Cluster country compositions (from Figure 6):
  - Cluster 1: Brazil, Hungary, Mexico, Philippines, Poland and South Africa.
  - Cluster 2: Argentina, Peru, Saudi Arabia, Uruguay, and Venezuela.
  - Cluster 3: Chile, Colombia, Egypt, Hong Kong SAR, Indonesia, Israel, Korea, Malaysia and Thailand.
  - Cluster 4: China, India, Singapore and Turkey.

### Investment–Growth and TFP–Growth Nexus: Joint Findings
- Investment-growth joint results (Figure 7a):
  - High growth countries are always associated with high investment (example: China).
  - Low growth countries are always associated with low investment.
  - Economies showing declining investment also experience declining growth.
- Growth–TFP joint results (Figure 7b):
  - Strong correlation: most low growth countries witness slow TFP growth; high growth countries are accompanied by high TFP growth.
- The cluster results are framed relative to classical growth theories (Sala-i-Martin 1997; Rostow 1959) and allow country positioning across cluster outputs.

### External Financing of Investment Booms and Boom–Bust Dynamics
- Methodology note:
  - Type of external financing classified by comparing five-year cumulative equity and debt inflows prior to an investment boom.
  - Investment boom year T defined as year when deviation from trend of investment share exceeds 2 standard deviations of the cyclical component (HP filter smoothing parameter = 100).
- Core empirical finding:
  - Investment booms financed through raising debt rather than equity can lead to capital over-accumulation and unsustainable growth.
  - Countries with more external debt financing have suffered more severe growth collapse and a sharper increase in the unemployment rate when investment booms have gone bust (Figure 8).
- Examples of episodes used:
  - External equity financing episodes include: Czech Rep. 2008, Colombia 1994, Egypt 1990, India 2007, Indonesia 1997, Malaysia 1997, Pakistan 2006, Poland 1999 and Singapore 1984.
  - External debt financing episodes include: Chile 1981, Mexico 1981, Korea 1997, Philippines 1983, South Africa 2008, Thailand 1996 and Uruguay 1981.
- Visual metrics tracked around T: Investment (% of GDP), GDP growth (%), Unemployment rate (%), for T-3 to T+3.

### Caveats on Cluster Persistence and Endogeneity
- Five-year averages used to smooth short-term volatility; boom–bust cycles increase dissimilarity and can place separate episodes within same cluster.
- Endogeneity concern: over-investment and strong growth often followed by under-investment and weak growth within boom–bust cycles; not fully resolved by averaging.

### Convergence Classification and Dynamics
- Convergence patterns by growth-cluster (Figure 9a):
  - Low Growth Group: GDP per capita relative to U.S. almost stagnant between 25 and 30 percent; only in last decade has the receded trend been altered.
  - Increasing Growth Group: path to convergence is uneven with real GDP per capita growth ranging between 6 percent and -3 percent across episodes; progress eroded by volatility.
  - Declining Growth Group: strong advancement over past five decades; average per capita income relative to the U.S. improved from 15 to 45 percent; business cycles less volatile and steady convergence achieved.
  - High Growth Group: China only; starts from very low level but extraordinarily high growth rates shortened catch-up process.
- Convergence by region (Figure 9b):
  - Convergence patterns based on one-variable cluster output show more homogeneity within clusters than geography.
  - Historical crises impacted all regions: Latin America debt crisis (mid-1980s), dissolution of the Soviet Union (early 1990s), Asian financial crisis, 2008 Global Financial Crisis.
- Cluster compositions used in Figure 9a (for clarity):
  - Cluster 1: Brazil, Colombia, Hungary, Mexico, Poland, and South Africa.
  - Cluster 2: Argentina, Peru, Philippines, Saudi Arabia, Turkey, Uruguay, and Venezuela.
  - Cluster 3: Chile, Egypt, Hong Kong SAR, India, Indonesia, Israel, Malaysia, Pakistan, Singapore, Korea, and Thailand.
  - Cluster 4: China.

### Investment–Growth–Unemployment Nexus
- Long-run relationship documented:
  - Significant negative relationship between investment level and unemployment (Figure 10).
  - Low Investment Group members tend to have high unemployment rates.
  - High Investment Group members tend to experience low unemployment rates.
- Theoretical links cited:
  - Solow Growth Theory: increases in the saving rate on investment lead to higher steady state output and higher growth.
  - Beveridge Curve: job vacancy rate and unemployment negatively correlated; robust growth creates more jobs, reducing long-term unemployment.
- Suggested extensions (left for future research): use labor market indicators and real growth rate as inputs; calculate Misery Index (unemployment + inflation) to examine living standards.

### Recent Performance and Taxonomy of Emerging Markets (post-2000 focus)
- Taxonomy methodology:
  - Short-term focus using data from 2000 onwards; sample expanded to 52 economies (43 major EMs and 9 Newly Industrialized Economies).
  - Six clusters identified per indicator using Ward’s linkage method, then reduced to three groups by judgment and sensitivity analysis; kernel density estimation used for robustness.
- Domestic-angle taxonomy (supply and demand decomposition, median values 2000–2013):
  - Supply-side decomposition: physical capital contribution, labor contribution, TFP contribution to total real output growth; countries ranked higher = more “capital-driven”, “labor-driven”, “technology-driven”.
  - Demand-side decomposition: consumption contribution, investment contribution, net exports contribution; countries ranked higher = more “consumption-led”, “investment-led”, “export-led”.
  - Figure 11 presents taxonomy output (measured by each indicator’s contribution to GDP growth as percent of GDP growth); * marks countries that ran trade deficits through 2000-2012 but whose net exports had been improving and therefore contributed positively to GDP growth.
- External-angle taxonomy:
  - Grouped countries on financial openness, trade openness and linkages, terms of trade growth, and net commodity dependence (Figure 12).
  - Observations:
    - Hong Kong SAR and Singapore have extraordinarily high levels of financial and trade openness.
    - Only 6 economies with relatively high levels of financial openness are top-tier versus 23 economies bottom-tier.
    - Trade openness distribution more uniform; about one third of economies in each category.
    - Only 9 EMs are large net commodity exporters; 31 economies run either a small net commodity export surplus or a net commodity export deficit.

### Trade Linkages and Cutoff Points
- Trade linkage cutoff points (exports to major partners):
  - Exports to the Euro area: cutoff points for high, medium, low linkages are 49 percent and 29 percent.
  - Exports to the U.S.: cutoff points are 34 percent and 14 percent.
  - Exports to China: cutoff points are 12 percent and 4 percent.
- Aggregate observation: EMs on average have stronger trade ties with the Euro area and the U.S. than with China, though China’s role is increasing.

### Recovery since the Global Financial Crisis (GFC) and Domestic Taxonomy Implications
- High-frequency dynamics (Figure 13a):
  - When the crisis began, EMs experienced negative contributions from net exports for four consecutive quarters; consumption and investment were weak.
- Regional post-crisis output changes:
  - Compared to pre-crisis growth, post-crisis growth has lowered by 4.3 percentage points for Emerging Europe and by 1.5 to 2.6 percentage points for other regions.
- Taxonomy explanatory power for recovery:
  - Countries mainly consumption-led rebounded the slowest and weakest (high consumption-led group of six countries had the deepest recession and weakest recovery).
  - Countries mainly investment-led rebounded fastest and strongest (high investment-led group of six countries had the mildest recession and most robust recovery).
  - Post-crisis growth rates of the high investment-led group have always excelled that of the medium and low investment-led groups.
- Conjecture offered:
  - Economic agents can adjust consumption more freely in response to shocks; investment is harder to adjust because of capital irreversibility and adjustment costs. Investment can stabilize the economy in the short run and promote growth in the long run.
- Robustness note: similar recovery dynamics by consumption-led and investment-led classification observed for the 1980s and 1990s.

### Composite External-Exposure Index and Impact of External Factors
- Construction of overall external index:
  - Components: financial openness, trade openness, terms of trade growth, net commodity export to GDP.
  - Standardization: global min-max scaling to range 0–1 for each component.
  - Weights endogenized via principal component analysis (first principal component squared factor loadings).
  - Countries assigned an overall index as weighted average of components; higher index = more exposed to world economy.
- Tertile comparisons pre- vs post-GFC (Figure 14a preliminary description):
  - All three tertiles experienced significant slowdowns after GFC.
  - Degree of slowdown is positively correlated with external exposure:
    - Top tertile: post-crisis growth dropped by 6.4 percentage points from 8.4 percent to 2.0 percent.
    - Middle tertile: growth rate dropped by 3.2 percentage points from 5.8 percent to [value truncated in source excerpt].

*Italic: Source: WEO, Penn World Table, IFS, Lane and Milesi-Ferretti (2007), Haver Analytics and IMF staff calculations, as presented in the supplied content.*

### 2.6 percent and the bottom tertile’s growth rate dropped by 2.3 percentage points from 6.6

### _wp15155 - 2.6 percent and the bottom tertile’s growth rate dropped by 2.3 percentage points from 6.6

### Impact of external factors on average growth and growth drops
- The bottom tertile’s growth rate dropped by 2.3 percentage points from 6.6 percent to 4.3 percent.
- The text references an observed relationship between the degree of external exposure (external factor index by tertile) and average real GDP growth across periods 2003-07 and 2008-12 (Figure 14a).

### Growth surprises and external exposure
- Growth surprises are measured as the difference between actual real GDP growth rate for country i in year t and the projected real GDP growth rate for that same country one-year ahead, as published in the IMF’s WEO of the previous year.
- During 2008-2012, growth surprise is negative for all three tertiles, likely driven by a global factor:
  - Bottom tertile (least external exposure): growth surprise is -0.5 percentage points.
  - Middle tertile: growth surprise is -1.1 percentage points.
  - Top tertile (most external exposure): growth surprise is -2.7 percentage points.
- The top tertile’s growth surprise is about twice as large as the surprise on the middle tertile and five times as large as the surprise on the bottom tertile.

### Business cycle synchronization and external exposure
- The degree of business cycle synchronization is defined as the correlation between one country’s real GDP growth rate and world real GDP growth rate.
- The degree of business cycle synchronization is positively correlated with the degree of the external factor (Figure 14c).
- Output correlations with global output by tertile:
  - Bottom tertile: 0.63
  - Middle tertile: 0.66
  - Top tertile: 0.75

### The role of openness and terms of trade on EM growth
- Favorable terms of trade change has helped commodity-exporting EMs boost growth, especially for large net commodity exporters (Figure 15).
- Average elasticity of EM commodity exporters’ growth to term of trade changes is about 0.14 and statistically significant.
- The average impact of terms of trade changes on non-commodity-exporters is minimal and statistically insignificant (Cubeddu et al. 2014).
- When growth of terms of trade is adjusted by trade openness, there is a positive relationship on growth:
  - Terms of trade shocks are magnified through the trade channel: the more open in international trade for a commodity exporter, the more likely it is for the country to enjoy the windfall on growth.
- External angle taxonomy useful to assess benefit from the commodity boom: trade openness, growth of terms of trade, and net commodity export to GDP.

* _wp15155 - 2.6 percent and the bottom tertile’s growth rate dropped by 2.3 percentage points from 6.6*

### 2007. When the crisis hit, the damage was amplified by the degree of financial openness:

### 2007. When the crisis hit, the damage was amplified by the degree of financial openness:

### E. Trade Linkages and Spillover Effects
- Evolution of EMs' export destinations:
  - In the 1990s, average EMs' exports to the Euro area, the U.S., and China as percentage of exports to the world were 27.3, 17.3, and 2.3 percent respectively.
  - Ten years later, average EMs' exports to the U.S., Euro area, and China became 28, 17.5, and 5.4 percent.
  - EMs' average export share to China increased by 132 percent.
- Growth performance by primary trade partner:
  - Results in Figure 16b indicate that countries who trade more with China on average experienced faster growth than countries who trade more with the Euro area or the U.S.
  - This pattern was more pronounced in the post-crisis period, as the Euro area’s trading partners witnessed their growth almost faltering.
- Complementary evidence:
  - Mishra et al. (2014) find that stronger trade linkage with China (measured by total trade with China to its own GDP) is associated with less market volatility during the U.S. Federal Reserve’s tapering talk in 2013–2014, ceteris paribus.
- Notes on country groupings used in analysis:
  - EMs links with China: AGO, CHL, HKG, KAZ, KOR, PER
  - EMs links with the U.S.: AGO, COL, CRI, DOM, GTM, ISR, JAM, MEX, PAN, VEN (IRQ excluded due to missing data)
  - EMs links with Euro Area: AZE, BGR, BIH, CZE, DZA, HRV, HUN, MAR, POL, ROM, SVK, SVN, TUN
  - All EMs refer to Appendix Table 3, excluding China, Swaziland and Taiwan POC.
- Statistical inference cited:
  - Before the GFC, average real GDP growth rates for EMs with strong trade links to China are statistically higher; after the GFC, average real GDP growth rates for EMs with strong trade links to Euro Area are statistically lower. Results are based on the unpaired two-sample t-test.

### Findings on openness, commodity booms, and global imbalances (Figure 15 summary)
- Empirical relationship described:
  - The impacts on growth from commodity booms and global imbalances are adjusted by openness measures.
  - Increased openness amplifies transmission of terms of trade shocks and external adjustment pressure on growth.
- Specific adjustments and variables used:
  - Growth of terms of trade adjusted by net commodity export (% of GDP) (2000-10)
  - Growth of terms of trade adjusted by trade openness (2000-12)
  - Cumulative current account deficit adjusted by financial openness (2003-07)
- Presentation notes:
  - Angola and Venezuela are omitted in Figure 15 for presentation purposes given the magnitude of their x-axis variables.
  - Hong Kong SAR and Singapore are omitted for presentation purposes given the magnitude of their financial openness.

### Key chart numeric references (Figure 16b / trade-linkage growth averages)
- Avg Real GDP growth (%), 2000-07 and 2008-13 by trade linkage (chart values shown):
  - Links with Euro Area: 5.6 (2000-07), 5.3 (2008-13)
  - Links with U.S.: 7.1 (2000-07), 1.4 (2008-13)
  - Links with China: 3.9 (2000-07), 4.4 (2008-13)
- Note: (1) Only countries in the top cluster are included in the calculation of growth by trade linkage.

### IV. Conclusions
- Emergence and heterogeneity of EMs:
  - Emerging market economies’ rise over the last five decades has greatly reshaped the global economic landscape; their contribution to world output makes them a significant economic powerhouse.
  - The study explores EM heterogeneity by identifying clusters and creating a simple taxonomy based on EMs’ fundamentals.
- Cluster and taxonomy findings:
  - The clusters point to four distinct patterns of long-term economic convergence and have more explanatory power than traditional geographic classification.
  - Investment clusters are highly correlated with growth clusters — investment is a necessary but not sufficient condition for growth.
  - Economic growth and productivity improvement tend to go hand in hand, per TFP growth and real GDP growth clusters.
  - On the investment-growth-unemployment nexus: on average, countries in the high investment-growth group tend to have low unemployment rates, while countries in the low investment-growth group are likely to have high unemployment rates.
- Post-crisis recovery and drivers:
  - EMs rebounded relatively quickly overall in the post-crisis period, but growth has slowed recently.
  - Countries that are mainly investment-led rebounded the fastest and strongest; countries mainly consumption-led rebounded the slowest and weakest.
- Role of external factors:
  - Degree of economic slowdown, growth surprise, and business cycle synchronization are all positively correlated with the degree of the external factor index.
  - Increased openness amplifies the transmission of terms of trade shocks and external adjustment pressure on growth.
- Policy implication:
  - Continuing convergence to high income level for EMs will be more challenging and is not guaranteed.
  - Tailored policy actions are needed to take into account heterogeneity in the EM universe beyond traditional geographical and income approaches.
  - The clusters and taxonomy can be used as a reference to design both near- and long-term policies.

### Appendix — Cluster Methodology (Ward’s method)
- Clustering approach:
  - Ward’s method starts with n clusters (in this case n=25) of size 1 and stops when all observations are merged into one cluster.
  - At each step, observations are combined to minimize the errors sum of squares (equivalently maximize the R-square) from the group centroid.
  - A two-stage clustering approach is conducted:
    - Stage 1: obtain cluster IDs for each observation within a given country (each country will have a series of 7 cluster IDs).
    - Stage 2: all cluster IDs of each country are utilized to determine a single cluster ID for that country.
  - The number of clusters is chosen at the discretion of the dissimilarity level; four clusters are selected as appropriate.
- Dendrogram notes:
  - The dendrogram represents dissimilarity levels among countries based on real GDP growth over 1980-2013, using Ward’s Linkage.
  - Longer horizontal lines indicate greater dissimilarity; China shows the highest level of dissimilarity among all countries.

### Appendix — Taxonomy Methodology
- Domestic indicators:
  - For each domestic indicator, period medians of each country are used after excluding outliers.
  - Outliers are defined as observations falling outside the 5th–95th percentile range of the entire cross-country sample during a 3-year crisis window (three years preceding the crisis, including banking, currency and debt crises). Crisis data from Laeven and Valencia (2012).
  - Around 1.3 to 2.8 percent of observations are dropped for the domestic indicators.
- Growth accounting (supply side):
  - Assumes standard Cobb-Douglas function: Yt = At Kt^α Lt^(1−α)
  - Decomposition in log form: Δy = α Δk + (1−α) Δl + Δa
  - Capital contribution to GDP growth equals sum of ICT and non-ICT capital contributions (from Total Economy Database).
  - Labor contribution accounts for labor quantities and labor quality.
  - TFP contribution is the residual.
- Demand side contributions:
  - Contribution of a component to GDP growth = real growth rate of component × its share in total real GDP in the previous year.
  - Consumption and investment contributions use WEO annual data in real terms.
  - Net exports contribution = export contribution minus import contribution (Lequiller and Blades 2006).
  - To preserve the sign of each factor’s contribution, absolute values of GDP growth are taken in the X Contribution / GDP growth formulation.
- External indicators:
  - For external indicators, period simple averages are used for clustering rather than medians because external indicators are relatively stable.
- Data coverage and notes:
  - Capital, labor, and TFP contribution data: Total Economy Database, 2000-2012* (2013 excluded due to >50 percent missing data).
  - Consumption, investment, net exports: WEO Database, 2000-2013.
  - Financial openness: External Wealth of Nations Database, WEO Database, 2000-2010.
  - Trade openness, terms of trade growth: WEO Database, 2000-2012 (terms of trade 2000-2012; net commodity exporters 2000-2010).
  - Export shares by destination: Direction of Trade Statistics Database, 2000-2012.

*Italic source attribution line*

*Source: WDI, WEO, IFS, External Wealth of Nations Database, Total Economy Database, Direction of Trade Statistics Database, Laeven and Valencia (2012), and IMF staff calculations as presented in the provided content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15155.pdf_