## wp1852

## Source details

**Canonical URL:** [wp1852](https://www.imf.org/-/media/files/publications/wp/2018/wp1852.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp1852.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp1852.pdf.json)

---

### Major findings on EMDE contribution to global growth and consumption
- During 2000–08, EMDEs on average accounted for 70 percent of global growth in output and consumption in purchasing power parity terms.
- After the global financial crisis, EMDEs’ contribution rose to about 80 percent of output growth and 85 percent of consumption growth (PPP terms).
- In market exchange rate terms, EMDEs accounted for close to 70 percent of global output growth and just over 70 percent of global consumption growth during 2010–15.

### Relative income and catch-up potential
- In 90 percent of EMDEs current real income per capita (converted at purchasing power parity exchange rates) is less than half what it is in the United States.
- In 85 percent of EMDEs real income per worker is less than half that in the United States.
- Ratios are calculated based on average real income per capita over a five-year window, 2010–14, to smooth business cycle and commodity price fluctuations.
- Caveat: Some EMDEs use the single-deflation method to calculate real GDP, which may affect the accuracy of relative income calculations.
- Observation: Narrowing of EMDEs’ relative income gap since 2000 did not reflect “convergence from above”; U.S. real GDP per capita did not decline in absolute terms except during the global financial crisis.

### Role of external conditions in medium-term growth
- External conditions analyzed: external demand conditions, external financial conditions, and terms of trade.
- Medium-term growth defined as the average over five-year horizons.
- The three country-specific external conditions help explain medium-term growth outcomes and the occurrence of growth accelerations and reversals.
- Variation at the country level in external conditions, as well as global common factors, matter for EMDEs’ medium-term growth.

### Constructing measures of external conditions (country-specific)
- External demand conditions:
  - Measured by the export-weighted growth rate of domestic absorption of trading partners (trading partners Θj account for at least 50 percent of total exports).
  - One unit corresponds to a 1 percentage point growth in domestic absorption of trading partners.
- External financial conditions:
  - Proxied by gross inflows to the region (excluding the country) as a share of aggregate GDP of other economies in the same region.
  - One unit corresponds to a 1 percentage point of GDP in capital flows to regional economies.
- Commodity terms of trade (CTOT):
  - Trade-weighted average of world prices of imported and exported commodities; annual change computed as trade-weighted sum of ∆log Pj,t with weights τi,j,t = (xi,j,t−1 − mi,j,t−1) / GDPi,t−1.
  - One unit corresponds to a 1 percent increase in the commodity terms of trade index (akin to a windfall income gain of 1 percent of GDP).

### Correlations among external condition measures (pairwise)
- External Demand vs External Financial Conditions: 0.1288
- External Demand vs Commodity Terms of Trade: 0.0737
- External Financial Conditions vs Commodity Terms of Trade: -0.0016
- Interpretation: Low correlations imply each dimension can exert a separate influence; country-specific measures often deviate from global proxies.

### Empirical approach and data
- Sample: Broad unbalanced sample of more than 80 economies over 1970–2014 (five-year nonoverlapping windows).
- Estimation framework: Fixed-effects panel growth regression git = γZit + βXit + αi + μt + εit; estimated with difference GMM (Arellano and Bond 1991; Arellano and Bover 1995).
- Controls Xit include initial log GDP per capita (previous five years), average inflation rate, human capital, de jure trade and financial openness measures, combined Polity IV index.
- Primary data sources: IMF WEO database, Penn World Tables 9.0, World Bank WDI; additional sources for external conditions and policy variables.

### Key empirical findings on the effect of external conditions
- All three external conditions have economically and statistically significant effects on EMDE medium-term growth and remain significant when included jointly and with time fixed effects.
- Magnitude examples (from column (4) results in text):
  - A 1 percentage point increase in trading partners’ domestic absorption growth associated with a 0.4 percentage point increase in medium-term growth.
  - A 1 percentage point increase in commodity terms of trade increases medium-term growth by almost ½ percentage point.
  - An increase in the ratio of capital flows to regional GDP of 1 percentage point raises medium-term growth (coefficients reported in table).

### Estimation results (Table 2 summary)
- Dependent variable: annual growth of GDP per capita (PPP), averaged over five-year windows.
- Estimation methods: GMM and OLS; time fixed effects: Yes; country fixed effects: Yes.
- Sample period: 1970–2014.
- Number of observations by specification: 505, 517, 509, 497, 587, 601, 592, 578.
- Number of economies by specification: 81, 84, 83, 80, 82, 84, 83, 81.
- R2 reported: 0.411, 0.422, 0.417, 0.432 (across baseline specifications).
- Specification tests (p-values): Second-Order Correlation Test: 0.863, 0.913, 0.567, 0.507. Hansen Test: 0.149, 0.173, 0.197, 0.201.
- Key coefficient estimates (selected):
  - External Demand Conditions: 0.524**, 0.421**, 0.331, 0.243.
  - External Financial Conditions: 0.266***, 0.186**, 0.339***, 0.289***.
  - Commodity Terms of Trade: 0.453*, 0.481*, 0.539**, 0.538**.
- Note: Interquartile range for average annual change in commodity terms of trade index is –0.4 percent to 0.3 percent.

### Contribution of country-specific external conditions to per capita income growth
- Collective contribution of the three external condition variables: "almost 2 percentage points to income per capita growth over 1975–2014."
- Joint contribution increased from about 1.7 percentage points over 1975–94 to about 2⅓ percentage points during 1995–2014.
- External financial conditions contribution increased by about ½ percentage point between 1995–2004 and 2005–14 (one-third of the increase in average income per capita growth between those subperiods).

### Definitions and identification of growth episodes
- Trend growth rate g_t,t+h estimated over rolling six-year windows (baseline h = five years).
- Growth acceleration episode requires:
  - g_t,t+h ≥ 3.5 percent a year;
  - g_t,t+h − g_t,t−h ≥ 2 percentage points;
  - y_t,t+h ≥ max{y_i}, ∀ i ≤ t.
- Persistent acceleration: acceleration not associated with a subsequent reversal or banking crisis within three years before or after end of episode.
- Reversal episode requires:
  - g_t,t−h − g_t,t+h ≥ 2 percentage points;
  - ȳ_t,t+h ≤ ȳ_t−h,t.
- Episodes identified up to year 2010 using PWT 9.0 through 2014 and extended to 2015 using WEO growth rates.

### Descriptive statistics of episodes (1970–2014)
- Total acceleration episodes: 127.
  - Persistent accelerations: 95.
  - Nonpersistent accelerations: 32 (12 associated with subsequent reversals, 18 with banking crises, 2 with both).
- Reversal episodes: 125.
- Temporal patterns:
  - Accelerations picked up in the 2000s; rare in other decades.
  - Balance shifted increasingly toward persistent accelerations in recent decades.
  - Reversals were numerous in the 1970s and 1980s; declined thereafter.
- Regional patterns:
  - Accelerations relatively steady in Asia; occur in all regions.
  - Reversals concentrated in MENAP, LAC, and SSA; fewer in Asia and Europe.

### Magnitude and persistence of episode effects on real income per capita
- Persistent accelerations:
  - Associated increases in real income per capita typically range from 15 to 40 percent above the starting level during the episode.
  - Median level of income per capita during the two decades after onset increases nearly twice as much as for economies that do not experience accelerations.
  - Persistent accelerations appear associated with permanent increases in income levels.
- Reversals:
  - Real income per capita typically declines 5–30 percent relative to the initial starting level during reversals; declines as large as 50 percent observed in some cases.
  - Reversals have persistent negative effects: level does not return to start-level until around 15 years after the start of the episode.
- Persistent vs nonpersistent accelerations:
  - Similar increases during first five years; after five years nonpersistent accelerations slow, leading to lower levels eight years after onset compared with persistent accelerations.
- Association with long-term growth:
  - Larger cumulative gains during persistent accelerations correlate with faster long-term average growth.
  - Larger cumulative losses during reversals correlate with lower long-term average growth.

### Event analysis: characteristics of episodes (1970–2015)
- Persistent accelerations:
  - Median annual growth rate: 5.5 percent.
  - Median comparator growth: 1.7 percent.
  - Median trading partner growth: "just above half a percentage point" higher than comparator economies.
  - External financing (gross capital flow into the region): about 1.5 percentage points higher than comparators.
  - Median change in terms of trade (full sample): –0.2 percent (episode) vs –0.1 percent (comparator).
  - For commodity exporters: median change in terms of trade = 0.9 percent (accelerations) vs 0.1 percent (comparators).
- Nonpersistent accelerations:
  - Median change in terms of trade is positive and significantly higher for these episodes.
- Reversals:
  - Median annual growth rate: –3 percent.
  - Median comparator growth during reversals' period: 2.6 percent.
  - Trading partner growth: almost 0.7 percentage point lower than nonepisodes.
  - Capital flows to the region: roughly 0.7 percentage point lower than nonepisodes.
  - Median change in terms of trade (full sample): –0.10 percent (episode) vs –0.08 percent (nonepisode).
  - For commodity exporters in reversals: decline of about 0.75 percentage point in terms of trade vs increase of about 0.3 percentage point for nonreversal commodity exporters.

### Logit regression results and marginal effects
- Logit specifications estimate odds ratios for occurrence of persistent accelerations and reversals with various fixed-effects and controls.
- Persistent accelerations: external demand and external financial conditions show robust positive associations with odds ratios across specifications; change in terms of trade not significant on full sample.
  - Example odds ratios (country and time fixed effects): External Demand = 1.158** (0.085); External Financial = 1.098** (0.044).
- Reversals: external demand and external financial conditions generally reduce odds of reversal; terms-of-trade effects weaker once time-fixed effects included.
  - Example odds ratios (country and time fixed effects): External Demand = 0.793*** (0.061); External Financial = 0.977 (0.055); Change in Terms of Trade = 0.973 (0.028).
- Marginal effects evaluated at sample means:
  - Persistent accelerations:
    - A 1 percentage point increase in trading partner demand raises probability of acceleration by 3.9 percentage points (increasing probability to 9.7 percent).
    - A 1 percentage point of GDP increase in regional capital flows raises probability of persistent acceleration by 2.6 percentage points.
    - Terms of trade: not significantly associated with persistent accelerations for full sample; significant for commodity exporters and for nonpersistent acceleration subset.
  - Reversals:
    - A 1 percentage point increase in external demand lowers probability of a reversal by 4 percentage points (about 50 percent of the unconditional probability).
    - A 1 percentage point of GDP increase in regional capital flows is associated with a 2.4 percentage point decrease in reversal probability.
    - Change in terms of trade associated with a 0.6 percentage point reduction in likelihood of reversals (statistically significant in some specifications).
- Robustness: extending episode horizon to seven years leaves marginal effects statistically robust; point estimates change slightly.

### Conclusions and key takeaways
- EMDEs now account for more than three-fourths of global growth in output and consumption.
- Country-specific external conditions (external demand, external financial conditions, terms-of-trade shocks) are significant determinants of EMDE medium-term growth and have become increasingly influential with global integration.
- External financial conditions increased their contribution to EMDEs’ medium-term growth over time; post-2005 contribution rose by about ½ percentage point relative to 1995–2004.
- External conditions influence growth by affecting the probability of persistent accelerations and reversals:
  - Favorable external demand and financial conditions increase likelihood of persistent accelerations and reduce likelihood of reversals.
  - Terms-of-trade windfalls matter particularly for commodity exporters; for the broader EMDE sample terms of trade are more closely associated with nonpersistent accelerations and reversals in specific subsamples.
- Long-run divergence: up to two decades after onset, real income per capita still diverges from benchmark paths for economies experiencing acceleration or reversal episodes.

### Data, sample, and annex highlights
- Primary data: IMF WEO, Penn World Tables 9.0, World Bank WDI.
- Annex Table 1 lists all indicators and sources; Annex Table 2 lists the EMDE sample (examples: Albania, Algeria*, Angola*, Argentina, ... Vietnam, Yemen*, Zambia*, Zimbabwe).
- Commodity exporters flagged with * (commodity exports exceed 65 percent of total exports and net commodity exports ≥ 6 percent of GDP over sample).
- Country-year episode lists included in Annex Tables 3 and 4 (persistent accelerations and reversal episodes with specific years).

*Source: wp1852 - References (IMF working paper content provided).*

### References .............................................................................................................

### wp1852 - References

### Major findings on EMDE contribution to global growth and consumption
- During 2000–08, emerging market and developing economies (EMDEs) on average accounted for 70 percent of global growth in output and consumption in purchasing power parity terms, nearly double their contribution during the 1980s.
- After the global financial crisis, EMDEs’ contribution to global growth rose to about 80 percent of output growth and 85 percent of consumption growth (PPP terms).
- In market exchange rate terms, EMDEs accounted for close to 70 percent of global output growth and just over 70 percent of global consumption growth during 2010–15.

### Relative income and catch-up potential
- In 90 percent of EMDEs current real income per capita (converted at purchasing power parity exchange rates) is less than half what it is in the United States.
- In 85 percent of EMDEs real income per worker is less than half that in the United States.
- The paper notes that some EMDEs may be close to their own steady-state levels and unlikely to experience further catch-up growth; conversely, large gaps in output per worker suggest significant room for catch-up in many countries.
- Caveats: Ratios are calculated based on average real income per capita over a five-year window, 2010–14, to smooth business cycle and commodity price fluctuations. Some emerging market economies use the single-deflation method to calculate real GDP, which may affect the accuracy of relative income calculations.

### Role of external conditions in medium-term growth
- The paper investigates how three country-specific external conditions—external demand conditions, external financial conditions, and terms of trade—affect medium-term growth in EMDEs, the likelihood of experiencing growth accelerations or reversals, and the pace of income convergence vis-à-vis advanced economies.
- “Medium-term” growth is defined as the average over five-year horizons to smooth the influence of business cycle fluctuations.
- The analysis distinguishes between persistent and nonpersistent accelerations, where nonpersistent accelerations are associated with a subsequent growth reversal or a banking crisis.
- Key result: Variation at the country level in external conditions, as well as global factors affecting all economies in a common manner during certain intervals, matter for medium-term growth outcomes of individual EMDEs.
- The three country-specific external conditions also help explain the occurrence of growth accelerations and reversals.

### Change in importance of external financial conditions over time
- The importance of country-specific external conditions for EMDEs’ medium-term growth has increased over time, particularly in the case of external financial conditions.
- Their contribution to medium-term growth has increased by about ½ percentage point—or one-third of the increase in average income per capita growth—between 1995–2004 and 2005–14.

### Data, sample, and empirical scope
- Primary data sources: IMF World Economic Outlook (WEO) database, Penn World Tables (version 9.0), and World Bank World Development Indicators database.
- Additional databases are used to construct the external conditions variables and policy and other domestic attribute variables used in the empirical analyses.
- Annex Table 1 lists all indicators used in the paper and their sources; Annex Table 2 lists the sample of all EMDEs used in the various analytical exercises.
- Group definition: The EMDE group comprises all economies currently classified as such by the IMF’s World Economic Outlook as well as those reclassified as “advanced” since 1996 (list provided in the source). Economies with populations in 2010 below 1 million according to the Penn World Tables (PWT) 9.0 vintage are excluded from the sample.
- Data on cross-country real income, factors of production (physical and human capital, labor input), and population are from the PWT 9.0 vintage. Deaton and Aten 2016 and Inklaar and Rao 2016 discuss methodology of the 2011 International Comparison of Prices that underpins PPP real income calculations in PWT 9.0.

### Stylized facts on growth episodes and distributional patterns
- Growth across EMDEs exhibits episodes of accelerations and reversals rather than steady, automatic catch-up.
- Historical patterns: In the 1970s the median income gap remained broadly unchanged; during the 1980s and 1990s income gaps widened as EMDEs suffered a “lost decade” and financial crises.
- Distributional observation: For the bottom three quintiles of the income distribution in the 1970s, the best performers (top decile of that group) saw relative income rise from about 11 percent in the 1970s to about 21 percent in recent years, while the median relative income for that group declined over four decades. By contrast, the median relative income increased for the top two quintiles of EMDEs in the 1970s.

*Source: wp1852 - References (IMF working paper content provided).*

### 2. EMDE Relative Income per Worker

### 2. EMDE Relative Income per Worker

### Patterns in relative income per worker
- Real income per worker is reported relative to the U.S., using the 2010–14 average (percent).
- The narrowing of EMDEs’ relative income gap with the United States since 2000 did not reflect “convergence from above”: except during the global financial crisis, real GDP per capita in the United States did not decline in absolute terms during the 2000s and 2010s.
- Narrowing occurred in part because of exceptional tailwinds that supported synchronized accelerations (IMF 2014). Growth reversals in EMDEs have played an important role when gaps widened. The episodic nature of EMDE growth—including accelerations and reversals—contributes to variation in the pace at which relative income gaps narrow and widen.

### Constructing country-specific and exogenous measures of external conditions
- The paper focuses on three sets of external conditions: external demand conditions, external financial conditions, and terms of trade.
- External demand conditions:
  - Measured by the export-weighted growth rate of domestic absorption of trading partners (following Arora and Vamvakidis (2005) and IMF (2014)).
  - For emerging market economy j in year t, the growth rate of external demand is the export-share-weighted average of trading partners’ annual growth rate of real domestic absorption. Trading partners are those in the set Θj that collectively account for at least 50 percent of total exports (DOTS data). Domestic absorption uses Penn World Tables 9.0.
- External financial conditions:
  - Proxied by a quantity-based measure of capital flows to peer economies (other EMDEs within the same region) as a share of their aggregate GDP (constructed to be exogenous to each country, following Blanchard, Adler, and de Carvalho Filho 2015).
  - For economy j in year t, measured by the ratio of gross inflows to the region (excluding inflows to j) to the aggregate GDP of other economies in the same region. The numerator aggregates K_inflow_i,t; the denominator aggregates GDP_i,t−1 for i ∈ Θ\j.
  - A quantity-based metric is used to capture fluctuations in availability of diverse financial flows ranging from direct investment to cross-border bank lending (footnote 6).
- Terms of trade (CTOT):
  - Country-specific changes based on international commodity prices as in Gruss (2014) and IMF (2015), constructed as a trade-weighted average of world prices of imported and exported commodities.
  - The annual change in economy i’s CTOT index in year t is given by the trade-weighted sum of changes in log commodity prices ∆log Pj,t, with weights τi,j,t = (xi,j,t−1 − mi,j,t−1) / GDPi,t−1, where xi,j,t−1 and mi,j,t−1 are the average export and import values of commodity j by economy i between t − 1 and t − 3 (U.N. Comtrade), and GDPi,t−1 is the average GDP of economy i between t − 1 and t − 3 (U.S. dollars).
  - Commodity price series start in 1960 for 41 individual commodities covering energy, metals, food and beverages, and agricultural raw materials (detailed list provided in the text). Primary sources include IMF’s International Financial Statistics database and World Bank’s Global Economic Monitor for extensions.

### Correlations and idiosyncratic variation across external condition measures
- Pairwise correlations between country-specific external condition measures (Table 1):
  - Correlation between External Demand Conditions and External Financial Conditions: 0.1288
  - Correlation between External Demand Conditions and Commodity Terms of Trade: 0.0737
  - Correlation between External Financial Conditions and Commodity Terms of Trade: -0.0016
- Interpretation:
  - Low correlations suggest each dimension potentially exerts a separate influence; the external environment EMDEs face comprises a complex mix of factors that do not always move in the same direction.
  - Rolling correlations (Figure 5) show substantial deviations of country-specific measures from global proxies:
    - Country-specific external demand conditions often deviate significantly from aggregate world output growth.
    - External financial conditions show a common regional element but country-specific measures still exhibit substantial variability when restricted to same-region peers.
    - Country-specific CTOT indices show high cross-country variability in their correlation with oil prices or aggregate commodity prices.

### External conditions and medium-term growth: empirical approach
- Sample and period:
  - Broad unbalanced sample of more than 80 economies over 1970–2014 (five-year nonoverlapping windows to smooth business cycles).
- Estimation framework:
  - Fixed-effects panel growth regression: git = γZit + βXit + αi + μt + εit, where git is average annual growth rate of real GDP per capita (PPP) in country i over period t (five-year window).
  - Zit includes the three country-specific external conditions.
  - Xit controls include: initial level of income per capita (average log GDP per capita over previous five years); average rate of inflation; level of human capital; de jure measures of trade and financial openness (average import tariffs and an index of restrictions to the capital account); combined Polity IV index of governance characteristics.
  - αi captures time-invariant country fixed effects; μt is a time fixed effect controlling for common global factors.
  - Model estimated with difference generalized method of moments (Arellano and Bond 1991; Arellano and Bover 1995) to mitigate endogeneity concerns.

### Key empirical findings on the effect of external conditions
- All three external conditions have economically and statistically significant effects on EMDEs’ medium-term growth, remaining significant when included jointly and when controlling for time fixed effects.
- Magnitude examples (from column (4) results):
  - A 1 percentage point increase in the growth rate of domestic absorption in trading partners is associated with a 0.4 percentage point increase in medium-term growth—equivalent to around one-fifth of the average annual growth rate of GDP per capita in the sample.
  - An increase in the ratio of capital flows to GDP of EMDEs within the region of 1 percentage point raises medium-term growth (exact coefficient for this effect is reported in the table referenced in the text).

*Source: wp1852 - 2. EMDE Relative Income per Worker*

### 0.2 percentage point. A 1 percentage point increase in commodity terms of trade increases

### wp1852 - 0.2 percentage point. A 1 percentage point increase in commodity terms of trade increases

### Estimation results (Table 2: Linear Panel Growth Regression)
- Dependent variable: GDP per Capita Growth Rate (annual growth of GDP per capita in purchasing power parity terms, averaged over nonoverlapping five-year windows).  
- Estimation methods reported: GMM and OLS.  
- Time fixed effects: Yes.  
- Country fixed effects: Yes.  
- Other control variables: Yes.  
- Sample period: 1970–2014.  
- Number of observations by specification: 505, 517, 509, 497, 587, 601, 592, 578.  
- Number of economies by specification: 81, 84, 83, 80, 82, 84, 83, 81.  
- R2 reported: 0.411, 0.422, 0.417, 0.432 (across baseline specifications).  
- Specification tests (p-values): Second-Order Correlation Test: 0.863, 0.913, 0.567, 0.507. Hansen Test: 0.149, 0.173, 0.197, 0.201.  
- Notes: "One unit of external demand conditions corresponds to a 1 percentage point growth in domestic absorption of trading partners; one unit of external financial conditions corresponds to a 1 percentage point of GDP in capital flows to regional economies; one unit of the commodity terms of trade corresponds to a 1 percent increase in the commodity terms of trade index (akin to a windfall income gain of 1 percent of GDP). Robust standard errors are reported in parentheses. GMM = generalized method of moments; OLS = ordinary least squares. ***,**, and * denote significance at the 1, 5, and 10 percent level, respectively."

### Key coefficient findings (preserve reported point estimates and significance)
- External Demand Conditions:
  - Coefficients shown: 0.524**, 0.421**, 0.331, 0.243 (standard errors in parentheses as reported in table).  
- External Financial Conditions:
  - Coefficients shown: 0.266***, 0.186**, 0.339***, 0.289*** (standard errors in parentheses as reported in table).  
- Commodity Terms of Trade:
  - Coefficients shown: 0.453*, 0.481*, 0.539**, 0.538** (standard errors in parentheses as reported in table).  
- Interpretation provided in text:
  - "0.2 percentage point. A 1 percentage point increase in commodity terms of trade increases medium-term growth by almost ½ percentage point." (textual statement in source).  
  - Footnote: "It should be noted that a 1 percentage point change in the commodity terms of trade index is akin to a windfall income gain of 1 percent of GDP—a relatively large amount. The interquartile range for the average annual change in the commodity terms of trade index across all countries and periods is –0.4 percent to 0.3 percent."

### Contribution of country-specific external conditions to per capita income growth
- Collective contribution of the three external condition variables: "almost 2 percentage points to income per capita growth over 1975–2014."  
- Change over subperiods:
  - Joint contribution increased from about 1.7 percentage points over 1975–94 to about 2⅓ percentage points during 1995–2014.  
- External financial conditions contribution:
  - Increased by about ½ percentage point between 1995–2004 and 2005–14.  
  - This increase represents "one-third of the increase in average income per capita growth" between those subperiods, and "about half of the contribution from external factors since 2005, up from about one-third during 1995-2004."

### Role of common factors (time fixed effects)
- Time fixed effects capture "other common factors" including external conditions common across economies.  
- The contribution of other common factors:
  - Relatively stable during 1975–1999.  
  - Increased sharply since the early 2000s.  
- Implication: The estimates of country-specific external conditions' contribution to EMDEs’ medium-term growth may be a lower bound on the impact of external conditions because common-factor effects (captured by time fixed effects) also include common external influences.  
- Comparison with global variables:
  - The overall contribution of external conditions—and in particular external financial conditions—to medium-term growth over the past 15 years may have been larger than what is captured by country-specific external conditions variables (as suggested by comparison with global activity and financial variables).

### Growth accelerations and reversals: definitions and identification
- Trend growth rate (g_t,t+h) defined as least squares growth rate of real GDP per capita estimated over rolling windows of six years with horizon h (baseline h = five years).  
- Growth acceleration episode (baseline h = five years) requires:
  - Trend growth rate ≥ 3.5 percent a year (g_t,t+h ≥ 3.5);  
  - Trend growth rate during episode exceeds trend growth rate during preceding equal-length interval by at least 2 percentage points (g_t,t+h − g_t,t−h ≥ 2);  
  - Level of real GDP per capita at end of episode is at least as large as maximum level recorded prior to onset (y_t,t+h ≥ max{y_i}, ∀ i ≤ t).  
- Persistent acceleration:
  - An acceleration not associated with a subsequent reversal or a banking crisis (Laeven and Valencia 2013) within three years before or after the end of the acceleration episode.  
- Reversal episode definition (interval [t,t+h]):
  - Trend growth rate during the reversal is at least 2 percentage points lower than during preceding interval (g_t,t−h − g_t,t+h ≥ 2);  
  - Average level of real GDP per capita during the episode [t,t+h] is lower than the average level during [t−h,t] (ȳ_t,t+h ≤ ȳ_t−h,t).  
- Data sources and coverage for episodes:
  - Episodes identified up to year 2010 using real income per capita from PWT 9.0 through 2014 and extended to 2015 using the growth rate of real income per capita from the WEO database.  

### Descriptive statistics of episodes (1970–2014)
- Total growth acceleration episodes identified: 127.  
  - Persistent accelerations: 95.  
  - Nonpersistent accelerations: 32.  
    - Of these 32: 12 associated with subsequent reversals, 18 with banking crises, and 2 with both.  
- Reversal episodes identified: 125.  
- Temporal patterns:
  - Accelerations picked up in the 2000s; were relatively rare during other decades.  
  - Balance of accelerations shifted increasingly toward the persistent kind in more recent decades.  
  - Large number of reversals in the 1970s and 1980s; reversals have declined in number since then.  
- Regional patterns:
  - Accelerations: relatively steady in Asia; more variable elsewhere; occur in all regions.  
  - Reversals: concentrated in MENAP (Middle East, North Africa, Afghanistan, and Pakistan), LAC (Latin American and the Caribbean), and SSA (sub-Saharan Africa); Asia and Europe have seen fewer reversals.

### Magnitude and persistence of episode effects on real income per capita
- Persistent accelerations:
  - Associated increases in real income per capita typically range from 15 to 40 percent above the starting level during the episode.  
  - Median level of income per capita during the two decades after onset of a persistent acceleration increases nearly twice as much as the median level for economies that do not experience accelerations.  
  - Persistent accelerations appear associated with permanent increases in income levels.  
- Reversals:
  - Real income per capita typically declines 5–30 percent relative to the initial starting level during reversals; income drops as large as 50 percent observed in some cases (example in text: Sierra Leone in the mid-1990s).  
  - Reversals appear to have persistent negative effects: the level does not return to that attained at the start of the episode until around 15 years after the start of the episode.  
- Persistent vs. nonpersistent accelerations:
  - Levels of real GDP per capita increase similarly during the first five years of both sets of episodes.  
  - After five years, nonpersistent accelerations increase at a slower rate, leading to a lower level eight years after onset compared with persistent accelerations.  
- Association with long-term growth:
  - Economies with larger cumulative income gains during persistent accelerations tend to grow faster on average over the long term.  
  - Economies with larger cumulative income losses during reversals tend to witness lower long-term average growth rates.

*Source: Authors' calculations (from the provided content of the IMF working paper excerpt).*

### 2. Economies Experiencing Persistent or Nonpersistent

### 2. Economies Experiencing Persistent or Nonpersistent Accelerations

### Event Analysis: Persistent Accelerations and Reversals (1970–2015)
- Median annual growth rate during persistent acceleration episodes: 5.5 percent.
- Median annual growth rate for comparator economies (not in an episode over same period): 1.7 percent.
- Median annual growth rate during reversals: –3 percent.
- Median annual growth rate for comparators during reversals' period: 2.6 percent.
- For persistent acceleration episodes:
  - Median trading partner growth is "just above half a percentage point" higher than for comparator economies (difference statistically significant).
  - External financing (gross capital flow into the region) is about 1.5 percentage points higher than for comparator economies.
  - Median change in commodity terms of trade (full sample): –0.2 percent for persistent accelerations versus about –0.1 percent for comparator countries (very close to zero; marginal difference).
  - For commodity exporters only (persistent accelerations): median change in terms of trade = 0.9 percent versus 0.1 percent for comparator commodity exporters (difference statistically significant).
- For nonpersistent accelerations:
  - Median change in terms of trade is positive and significantly higher for those that experienced nonpersistent accelerations (Figure 15, panel 3).
- For reversal episodes:
  - Trading partner growth is almost 0.7 percentage point lower than for nonepisodes spanning the same interval.
  - Capital flows to the region are roughly 0.7 percentage point lower compared with nonepisode countries over the same period.
  - Median change in terms of trade (full sample) for reversals: –0.10 percent (episode) versus –0.08 percent (nonepisode) — very close to zero and not statistically significant.
  - For commodity exporters in reversal episodes: decline of about 0.75 percentage point in terms of trade versus an increase of about 0.3 percentage point for commodity exporters that did not experience a reversal.

### Logit Regression: Specification and Purpose
- Binary dependent variable: dummy = 1 for economy-years identified as persistent acceleration episodes (and zero otherwise); separate dummy = 1 for reversal episodes (and zero otherwise).
- Dummy variables also take value 1 in first lead (t + 1) and lag (t – 1) around each identified episode.
- Dependent dummy assumes value 1 for periods t, t − 1, and t + 1 of the episode.
- Independent variables: moving average of each of three external condition variables between periods t and t + 5.
- Benchmark specification (Equation 4) includes:
  - log(Pr(episode_it=1)/(1−Pr(episode_it=1))) = γZ_it + βX_it + α_i + ε_it
  - X_it are controls using moving averages between t − 3 and t − 1 (de jure integration, credibility of policy frameworks, etc.).
  - α_i captures time-invariant country fixed effects.
- Robustness: pattern of significance is robust to inclusion of time-fixed effects and additional controls; linear probability model tested as robustness check.

### Logit Estimates: Odds Ratios (Tables 3 and 4)
- Persistent accelerations (Table 3): coefficients report changes in odds ratio (value greater than 1 = increase in odds ratio).
  - External Demand:
    - No Country or Time Fixed Effects: 1.248*** (0.087)
    - Country Fixed Effects and Other Controls: 1.607*** (0.151)
    - Time Fixed Effects Only: 1.095 (0.097)
    - Country and Time Fixed Effects: 1.158** (0.085)
    - Random Effects: 1.330*** (0.119)
    - Probit Random Effects: 1.165*** (0.052)
    - Baseline Country Fixed Effects: 1.384*** (0.088)
  - External Financial:
    - No Country or Time Fixed Effects: 1.209*** (0.045)
    - Country Fixed Effects and Other Controls: 1.227*** (0.050)
    - Time Fixed Effects Only: 1.103** (0.050)
    - Country and Time Fixed Effects: 1.098** (0.044)
    - Random Effects: 1.243*** (0.049)
    - Probit Random Effects: 1.123*** (0.021)
    - Baseline Country Fixed Effects: 1.240*** (0.034)
  - Change in Terms of Trade:
    - Coefficients span: 0.970 (0.047), 1.042 (0.091), 0.935 (0.046), 1.040 (0.076), 1.007 (0.063), 1.009 (0.030), 1.052 (0.066) — none significant on full sample.
  - Model Chi-Squared Test reported values include 43.4***, 98.2***, 31,482.8***, 245.5***, 45.8***, 51.8***, 103.6*** across specifications.
  - Number of Economies and Observations vary by specification (e.g., Number of Observations: 4,176; 1,325; 4,176; 2,279; 4,176; 4,322; 2,279).
- Reversals (Table 4): coefficients report changes in odds ratio (value smaller than 1 = decrease in odds ratio).
  - External Demand:
    - No Country or Time Fixed Effects: 0.818*** (0.047)
    - Country Fixed Effects and Other Controls: 0.738*** (0.067)
    - Time Fixed Effects Only: 0.841*** (0.046)
    - Country and Time Fixed Effects: 0.793*** (0.061)
    - Random Effects: 0.736*** (0.055)
    - Probit Random Effects: 0.851*** (0.033)
    - Baseline Country Fixed Effects: 0.655*** (0.038)
  - External Financial:
    - No Country or Time Fixed Effects: 0.822*** (0.037)
    - Country Fixed Effects and Other Controls: 0.710*** (0.043)
    - Time Fixed Effects Only: 1.014 (0.061)
    - Country and Time Fixed Effects: 0.977 (0.055)
    - Random Effects: 0.788*** (0.041)
    - Probit Random Effects: 0.876*** (0.023)
    - Baseline Country Fixed Effects: 0.774*** (0.028)
  - Change in Terms of Trade:
    - No Country or Time Fixed Effects: 0.933* (0.039)
    - Country Fixed Effects and Other Controls: 0.851* (0.074)
    - Time Fixed Effects Only: 0.976 (0.041)
    - Country and Time Fixed Effects: 0.973 (0.028)
    - Random Effects: 0.935** (0.031)
    - Probit Random Effects: 0.963** (0.017)
    - Baseline Country Fixed Effects: 0.941** (0.027)
  - Model Chi-Squared Test reported values include 36.9***, 70.1***, 59,017.7***, 231.0***, 44.6***, 48.2***, 124.9*** across specifications.
  - Number of Observations examples: 4,176; 1,184; 4,176; 2,835; 4,135; 4,322; 2,835.
- Key interpretation:
  - Persistent accelerations: robust positive association between odds ratio and external demand and external financial conditions across specifications; change in terms of trade not significant on full-sample specifications.
  - Reversals: external financial conditions lose statistical significance when time-fixed effects are included (likely due to common factors in capital flows); effect of terms of trade on reversals becomes statistically insignificant when time-fixed effects are included.

### Marginal Effects from Logit Estimates (Figure 17 and discussion)
- Marginal effects evaluated at sample means of external conditions; coefficients represent impact of a one-unit increase in external condition on likelihood of episodes.
- Persistent accelerations (all external conditions at mean):
  - A 1 percentage point increase in trading partner demand raises probability of acceleration by 3.9 percentage points.
  - This represents an increase of the probability of acceleration to 9.7 percent (near-doubling relative to the unconditional probability).
  - A 1 percentage point of GDP increase in regional capital flows raises probability of persistent acceleration by 2.6 percentage points.
  - Improvement in terms of trade: not significantly associated with persistent accelerations for the entire EMDE sample.
    - Exceptions:
      - Commodity exporters: increase in terms of trade is significantly associated with increased likelihood of persistent accelerations.
      - Subset of 32 nonpersistent accelerations: increase in terms of trade significantly associated with occurrence of such episodes (terms-of-trade windfalls may trigger accelerations that are not sustained).
- Reversals:
  - A 1 percentage point increase in external demand lowers probability of a reversal by 4 percentage points (about 50 percent of the unconditional probability).
  - A 1 percentage point of GDP increase in capital flows to the region is associated with a 2.4 percentage point decrease in the probability of a reversal.
  - Change in terms of trade is associated with a statistically-significant reduction in likelihood of reversals of 0.6 percentage point.
- Robustness to alternative episode horizon:
  - Extending episode identification to seven years: marginal effects of external conditions remain robust in statistical significance; point estimates change slightly but pattern of significance is unchanged.

### Conclusions and Key Takeaways
- EMDEs now account for more than three-fourths of global growth in output and consumption.
- Country-specific external conditions (external demand, external financial conditions, terms-of-trade shocks) are significant determinants of EMDE medium-term growth and have become increasingly influential as EMDEs integrate with the global economy.
- External financial conditions have increased their contribution to EMDEs’ medium-term growth over time; comparing post-2005 with 1995–2004, their contribution increased by about ½ percentage point (about one-third of the increase in average income per capita growth for the group over this time).
- External conditions influence growth through their effects on the probability of persistent growth accelerations and reversals:
  - Favorable external demand and financial conditions increase likelihood of persistent accelerations and reduce likelihood of reversals across EMDEs.
  - Terms-of-trade windfalls matter especially for commodity exporters; for the broader sample of EMDEs, terms of trade are less influential on persistent accelerations but affect nonpersistent accelerations and reversals in specific subsamples.
- Long-run divergence: up to two decades after onset, real income per capita still appears to diverge from benchmark paths for economies experiencing acceleration or reversal episodes.

*Source: IMF Working Paper, "2. Economies Experiencing Persistent or Nonpersistent Accelerations" (contained in wp1852 PDF).*

### REFERENCES

### wp1852 - REFERENCES

### References (selected topics and authors)
- Studies on global productivity and trade slowdowns: Adler et al. 2017; Ahn and Duval (Forthcoming); Hong et al. 2016.
- Methods and econometrics: Ai and Norton 2003; Arellano and Bond 1991; Arellano and Bover 1995; Bai and Perron 1998, 2003; Chakravarti, Laha, and Roy 1967.
- Growth, growth spells, and middle-income trap: Barro and Sala-i-Martin 2004; Ben-David and Papell 1998; Eichengreen, Park, and Shin 2013; Jones 2016; Jones and Olken 2008; Klenow and Rodríguez-Clare 1997, 2005; Pritchett 2000; Pritchett and Summers 2014.
- Capital flows, financial globalization, and crises: Aizenman, Chinn, and Ito 2010; Alfaro, Kalemli-Ozcan, and Volosovych 2014; Bekaert, Harvey, and Lundblad 2005; Ghosh, Ostry, and Qureshi 2016; Gourinchas and Jeanne 2013; Kose et al. 2006; Laeven and Valencia 2013; Lane and Milesi-Ferretti 2007; Obstfeld 1998; Prasad, Rajan, and Subramanian 2007; Rajan 2006; Shambaugh 2004.
- Commodity prices, commodity exporters, and regional studies: Aslam et al. 2016; Collier and Goderis 2012; Gruss 2014; Tsangarides 2012.
- Policy, institutions, and development: Berg, Ostry, and Zettelmeyer 2012; Mauro et al. 2013; Rajkumar and Swaroop 2008; Reddy and Minoiu 2010; Rodrik 1999; Sahay et al. 2015; Henry 2000; Markusen and Venables 1999.
- IMF internal notes and working papers referenced: IMF Staff Discussion Note 17/04; IMF Staff Discussion Note 17/02; IMF Working Paper 16/27; IMF Working Paper 14/154; IMF staff discussion and policy papers cited through 2016.

### Annex — Data Sources (Indicator mapped to Source)
- Banking Crisis Indicator: Laeven and Valencia 2013
- Bilateral Cross-Border Bank Claims: Bank for International Settlements
- Capital Account Openness: Quinn 1997; Aizenman, Chinn, and Ito 2010
- Capital Inflows: IMF, Financial Flows Analytics database
- Capital Stock: Penn World Tables 9.0
- Commodity Terms of Trade: Gruss 2014
- Commodity Export Weights: United Nations Commodity Trade Statistics (Comtrade) Database; IMF, World Economic Outlook database
- Credit Boom Episodes: Dell’Ariccia and others 2016
- Current Account Balance: IMF, World Economic Outlook database
- Deposit Money Banks' Assets Ratio to GDP (Percent): World Bank, World Development Indicators database
- Employment: Penn World Tables 9.0
- Exchange Rate Stability Index: Aizenman, Chinn, and Ito 2010
- Export Value of Goods (Bilateral): IMF, Direction of Trade Statistics database
- External Debt Liabilities as a Share of GDP: Lane and Milesi-Ferretti 2007
- Free Trade Agreements by Year of Signature of Agreement: DESTA, Free Trade Area database; October 2016 World Economic Outlook
- Free Trade Agreements Coverage: WTO Regional Trade Agreements Database; October 2016 World Economic Outlook
- Human Capital: Penn World Tables 9.0
- Legal System and Property Rights Quality Index: Gwartney, Hall, and Lawson 2016
- Nominal GDP: IMF, World Economic Outlook database
- Nominal Interest Rate: IMF, World Economic Outlook database
- Oil Price in U.S. Dollars: IMF, Global Assumptions database
- Polity Score (Combined): Polity IV/Transparency International
- Population: Penn World Tables 9.0; United Nations Population database
- Public Debt as a Share of GDP: Mauro and others 2013; IMF, World Economic Outlook database
- Real GDP at Constant National Prices: IMF, World Economic Outlook database; Penn World Tables 9.0
- Real GDP in Purchasing Power Parity Terms: Penn World Tables 9.0
- Real Domestic Absorption: Penn World Tables 9.0
- Regulation Quality Index: Gwartney, Hall, and Lawson 2016
- Sound Monetary Framework: Gwartney, Hall, and Lawson 2016
- Tariffs: UNCTAD, Trade Analysis Information System; WTO Tariff Download Facility; IMF, Structural Reforms database; October 2016 World Economic Outlook

*Annex Table 1. Data Sources — Source: Authors' compilation. Note: DESTA = Design of Trade Agreements database; UNCTAD = United Nations Conference on Trade and Development; WTO = World Trade Organization.*

### Annex Table 2. Sample of Emerging Market and Developing Economies Included in the Analyses
- Albania, Algeria*, Angola*, Argentina, Armenia, Azerbaijan*, Bahrain*, Bangladesh, Belarus, Benin, Bolivia*, Bosnia and Herzegovina, Botswana, Brazil, Bulgaria, Burkina Faso, Burundi, Cambodia, Cameroon*, Central African Republic, Chad, Chile*, China, Colombia, Democratic Republic of the Congo, Republic of Congo*, Costa Rica, Côte d’Ivoire*, Croatia, Czech Republic, Dominican Republic, Ecuador*, Egypt, El Salvador, Estonia, Ethiopia, Gabon*, The Gambia, Georgia, Ghana, Guatemala, Guinea*, Guinea-Bissau, Haiti, Honduras, Hong Kong SAR, Hungary, India, Indonesia*, Islamic Republic of Iran*, Iraq*, Israel, Jamaica, Jordan, Kazakhstan*, Kenya, Korea, Kuwait*, Kyrgyz Republic, Lao P.D.R., Latvia, Lebanon, Lesotho, Liberia, Lithuania, FYR Macedonia, Madagascar, Malawi, Malaysia, Mali, Mauritania*, Mauritius, Mexico, Moldova, Mongolia*, Morocco, Mozambique, Myanmar, Namibia, Nepal, Nicaragua, Niger, Nigeria*, Oman*, Pakistan, Panama, Paraguay, Peru*, Philippines, Poland, Qatar*, Romania, Russia*, Rwanda, Saudi Arabia*, Senegal, Serbia, Sierra Leone, Singapore, Slovak Republic, Slovenia, South Africa, Sri Lanka, Sudan, Swaziland, Syria, Taiwan Province of China, Tajikistan*, Tanzania, Thailand, Togo, Trinidad and Tobago*, Tunisia, Turkey, Turkmenistan*, Uganda, Ukraine, United Arab Emirates, Uruguay, Uzbekistan, Venezuela*, Vietnam, Yemen*, Zambia*, Zimbabwe

*Note: The classification of emerging market and developing economies includes economies considered emerging markets before 1996. * denotes commodity exporters, which are economies for which commodity exports constitute the main source of export earnings during the sample period (commodity exports exceed 65 percent of total exports of goods, and net commodity exports account for at least 6 percent of GDP).*

### COUNTRY-YEAR GROWTH EPISODES – Persistent Accelerations (Annex Table 3)
- Albania 1995
- Algeria 2000
- Argentina 2003
- Armenia 2000
- Azerbaijan 2003
- Belarus 1999, 2002
- Benin 1977
- Bosnia 1995
- Botswana 1970, 1986, 1994, 2003
- Bulgaria 2003
- Burkina Faso 1994
- Cambodia 2003
- Cameroon 1970, 1976
- Chad 2000
- Chile 2002
- China 1980, 2000
- Colombia 2004
- Costa Rica 2003
- Czech Republic 2003
- Dominican Republic 1994, 2004
- Ecuador 1970
- Egypt 2004
- Estonia 2002, 2010
- Ethiopia 2003
- Ghana 2008
- Honduras 2003
- Hong Kong SAR 1976, 2003
- Hungary 1997
- India 1993, 2002
- Indonesia 1988, 2002
- Jordan 1975, 2001
- Korea 1982
- Lao P.D.R. 1979
- Lesotho 1987, 2005
- Lithuania 2002
- FYR Macedonia 2003
- Malawi 2005
- Malaysia 2002
- Mali 1974
- Mauritius 1973, 1985
- Mozambique 1994
- Myanmar 1993, 1998
- Namibia 2002
- Nigeria 2000
- Oman 1975
- Pakistan 2002
- Panama 2003
- Paraguay 2000, 2009
- Peru 2003
- Philippines 2003
- Poland 1995, 2003
- Rwanda 1975, 2003
- Sierra Leone 2009
- Singapore 1977, 1986, 2003
- Slovak Republic 2003
- Slovenia 1995
- Sri Lanka 1976, 1990, 2003
- Sudan 1997
- Swaziland 1985
- Syria 1972, 1993
- Taiwan Province of China 1984
- Tanzania 2000
- Thailand 1986, 2002
- Trinidad and Tobago 1996, 2001
- Tunisia 1995
- Turkey 2002
- Turkmenistan 2004
- Uzbekistan 2003
- Vietnam 1975, 1981

*Annex Table 3. Persistent Acceleration Episodes — Source: Authors' calculations.*

### COUNTRY-YEAR GROWTH EPISODES – Reversal Episodes (Annex Table 4)
- Albania 1988
- Algeria 1985
- Angola 1976, 1989
- Argentina 1980, 1999
- Bahrain 1981, 2006
- Bangladesh 1971
- Bolivia 1981
- Brazil 1989
- Bulgaria 1989
- Burkina Faso 1981
- Burundi 1992
- Cameroon 1985
- Central African Republic 1970, 1978, 2000, 2010
- Chad 1977, 1991
- Chile 1971
- Democratic Republic of the Congo 1974, 1989
- Republic of Congo 1986
- Costa Rica 1980
- Croatia 2009
- Côte d'Ivoire 1979, 1989, 1999
- El Salvador 1978
- Ethiopia 1973, 1982, 1988
- Gabon 1978, 1983, 1997
- The Gambia 1984
- Ghana 1973, 1979
- Guatemala 1982
- Guinea 1989
- Guinea-Bissau 1978, 1997
- Haiti 1981, 1990, 2000
- Honduras 1981
- Hungary 1988
- Iran 1976, 1984
- Iraq 1980, 1987
- Jamaica 1975, 1996, 2007
- Jordan 1986
- Kenya 1990
- Kuwait 1979, 1986, 1998, 2007
- Lebanon 1987
- Lesotho 1980
- Liberia 1979, 1989, 2003
- Madagascar 1973, 1979, 1990, 2009
- Malawi 1980, 1999
- Mauritania 1979
- Mexico 1983
- Mongolia 1989
- Mozambique 1981
- Myanmar 1985
- Namibia 1981
- Nicaragua 1976, 1985
- Niger 1971, 1982
- Nigeria 1979
- Oman 2010
- Panama 1985
- Paraguay 1983, 1996
- Peru 1980, 1987
- Philippines 1981
- Poland 1979, 1988
- Qatar 1979
- Rwanda 1985, 1990
- Saudi Arabia 1980, 1994
- Senegal 1976, 1989
- Sierra Leone 1994
- Slovenia 2009
- South Africa 1982
- Sudan 1978
- Syria 1985, 2010
- Tanzania 1979
- Togo 1972, 1979, 1989, 1998
- Trinidad and Tobago 1982
- Uganda 1976
- United Arab Emirates 1984, 2005
- Uruguay 1981, 1999
- Venezuela 1979, 1998
- Zambia 1970, 1976, 1990
- Zimbabwe 1974, 1983, 2001

*Annex Table 4. Reversal Episodes — Source: Authors' calculations.*

*Source: wp1852 - REFERENCES (IMF working paper annex content).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1852.pdf_