## _wp1076 - Appendix A. Country Groups

## Source details

**Canonical URL:** [_wp1076 - Appendix A. Country Groups](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1076.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1076.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1076.pdf.json)

---

### I. Introduction
- Purpose: revisit the relationship between inflation and growth and estimate the speed at which inflation beyond a threshold becomes harmful to growth using a Logistic Smooth Transition model (LSTR).
- Methodological advance: estimate both a threshold level for inflation and the speed of transition (γ) from low-cost to high-cost inflation regimes; validate estimates using bootstrapping and Monte-Carlo experiments.
- Core empirical summary:
  - Study estimates a threshold of about 10 percent for most country groups (except advanced economies).
  - Estimated transition speed is fairly high, implying inflation becomes harmful to growth soon after it exceeds the threshold.
  - Bootstrapping shows the range of estimates in the literature can be driven by data-process variability as well as estimation technique.
  - For developing countries, more than 90 percent of the bootstrapped sample places the estimated inflation threshold between 7 and 13 percent.
  - Inflation appears more costly for oil-exporters than for other groups when the dependent variable is non-oil GDP growth (as opposed to total GDP growth).
- Monte-Carlo result: LSTR provides additional flexibility, but distinction between LSTR and TAR is not crucial for these data-generating processes.

### II. The Backdrop (2007/08 inflation episode)
- Global context and numeric snapshots:
  - Advanced economies: headline inflation rose to 3.5 percent in May 2008 (year-on-year); core inflation remained at 1.8 percent.
  - Emerging and developing economies: headline inflation rose to 8.6 percent; core inflation rose to 4.2 percent.
- Inflation drivers identified:
  - Strong capital inflows, rapid credit growth, tightening labor markets, widening current account deficits in several countries (notably emerging Europe).
  - Rising export earnings and domestic demand in commodity-exporting countries (GCC, Russia); capacity constraints (e.g., housing, raw materials) in GCC.
  - Surging commodity (food and fuel) prices boosting inflation globally, with especially large effects in emerging markets and low-income countries.
- Policy responses observed:
  - Use of monetary, trade, and fiscal measures; price and quantity controls, administered prices, subsidies, buffer-stock building.
  - Policy choice influenced by exchange rate regime: flexible regimes relied more on monetary/fiscal mix; pegged regimes relied more on fiscal measures, capacity increases, and prudential actions.
- Outcome note: inflation remained high until housing-market effects and the subsequent crisis reduced rents and commodity/asset prices.

### III. Literature review — key findings relevant to threshold estimation
- Main empirical questions:
  - Is there a robust negative relationship between inflation and growth?
  - Is there a “kink” (nonlinearity) where low inflation can correlate positively with growth but high inflation correlates negatively?
  - Is there a minimum threshold beyond which inflation’s adverse growth effects become serious?
- Representative prior findings (numeric thresholds preserved exactly):
  - Mubarik (2005, Pakistan): inflation below 5 percent positive for growth; depresses growth if above 9 percent.
  - Rangarajan (1997, India): suggested range of 5 to 7 percent.
  - Samantaraya and Prasad (2001): estimated threshold at 6.5 percent.
  - Sarel (1996, 87 countries, 1970–90): structural break at 8 percent — positive correlation below 8 percent, negative above.
  - Ghosh and Philips (1998): positive correlation at about 2 to 3 percent; reversed at higher rates; convex relationship implies larger growth decline for increases from 10 to 20 percent than from 40 to 50 percent.
  - Khan and Senhadji (2001, 140 countries, 1960–98): threshold between 1 and 3 percent for industrial countries and 7 and 11 percent for developing countries.
  - Drukker, Gomis-Porqueras, and Hernandez-Verme (2005): threshold about 19 percent.
  - Pollin and Zhu (2006): threshold around 15-18 percent (sample divided by decades).
  - Burdekin et al (2004): for developing countries, inflation costly when higher than 3 percent; second break at 50 percent with marginal growth costs decreasing by 25 percent.
- This study’s contribution relative to prior work:
  - Extends period and number of countries.
  - Distinguishes emerging markets and oil exporters.
  - Estimates speed of transition between regimes (γ).
  - Uses bootstrap techniques to assess precision of threshold and transition-speed estimates.

### IV. Inflation and growth — Threshold level (Estimating the LSTR model)
- Data and specification:
  - Panel of 165 countries covering 1960–2007.
  - Data averaged over 5-year periods (time dimension: 9 periods of 5-year averages).
  - Controls: ratio of investment to GDP, population growth, initial GDP, rate of change in the terms of trade, variability of the terms of trade.
  - Country dummies and time dummies included; excluded periods of severe recession (losses > 5 percent of GDP per year over 5 years) for specific noted episodes.
  - Explanatory variable f(π) = ln(1+π).
  - LSTR uses weight W_high and transition speed γ to allow smooth transition between regimes.
- Interpretation of LSTR parameters (numeric examples preserved exactly):
  - W_high weight goes from 0 (low-inflation regime) to 1 (high-inflation regime); W_low = 1 − W_high.
  - Example (with threshold π* = 10 and st.dev = 0.5 shown in Figure 3):
    - If γ = 3 and inflation = 13 percent: effect captured by 0.3*β_low + 0.7*β_high.
    - If γ = 15 and inflation = 13 percent: effect captured by 0.003*β_low + 0.997*β_high.
  - γ captures speed of transition; γ → ∞ corresponds to standard TAR (instantaneous regime switch).
- Main empirical estimates and exact reported figures:
  - Full sample (165 countries): threshold inflation ≈ 9 percent.
  - When inflation is above 9 percent, a doubling of inflation decreases GDP by 0.7 percentage point per year.
  - Advanced economies: estimated threshold is 1 percent; costs of inflation almost three times smaller than for full sample.
  - Post-1990 sub-sample: threshold about 11 or 12 percent for all countries; for advanced economies non-linearity reduced and threshold estimated at 14 percent (precision hampered by smaller sample).
  - Developing countries (bootstrap): for more than 90 percent of the bootstrapped sample, the estimated inflation threshold lies between 7 and 13 percent.
  - Oil-exporting countries: LSTR point estimate of the threshold inflation is 74 percent; TAR model would give threshold = 10 percent — result for oil exporters is not robust and bootstrapping modes should be considered for judgment.
  - Non-oil GDP specification: estimates stronger than previous ones; text reports "a doubling of inflation from higher than 13 percent decreases real non-oil GDP by" (numeric completion not provided in source).
- Robustness checks and inference:
  - Standard errors built using 500 bootstrapping iterations conditioned on threshold from base estimation.
  - Bootstrapping confirms robustness for some thresholds (e.g., 1 percent threshold for whole sample estimated precisely) but also shows wide bootstrap distributions explaining differing literature estimates.
- Policy implication emphasized by γ:
  - High γ implies inflationary pressures must be tackled promptly because inflation becomes costly almost immediately after exceeding the threshold.
  - TAR models (γ → ∞) do not provide information on transition speed; LSTR fills this gap.

### V. Main findings on inflation-growth non-linearities and thresholds
- Using a panel of 165 countries covering 1960–2007, the study finds that for emerging market economies inflation above a threshold of about 10 percent quickly becomes harmful to growth.
- For advanced economies, the inflation threshold is much lower (mode of the bootstrap distribution is at 1 percent; the threshold is found to be lower than 5 (resp. 18) percent in only 43 (resp. 75) percent of the bootstrapped samples).
- For oil exporting countries, estimates are less robust: bootstrapping shows two potential modes for the threshold distribution (one TAR mode at 10 percent and a cluster of simulations with estimated threshold exceeding 100 percent).
- When using non-oil GDP growth for oil exporters, the threshold estimate is robust around 13 percent; the adverse effect of higher inflation on (non-oil) growth is stronger for oil producing countries than for other emerging markets.
- Across models, the link between growth and inflation at high levels of inflation is almost always negative; coefficients linking growth and inflation at low levels are not consistently different from zero.

### VI. Model estimates and interpretation (selected numeric parameter estimates preserved)
- Table 2 and associated tables report thresholds and parameters across groups. Selected exact figures:
  - Thresholds reported in table headers/entries: 9, 8, 1, 1, 10, 11, 74, 10, 13, 13.
  - “All countries” (column (a)): β_low = 0.002, β_high = -0.01, gamma = 6, (st.dev) = 1.08.
  - “All countries (if gamma = 15)” (column (c)): β_low = 0.002, β_high = -0.01, gamma = 15, (st.dev) = 1.08.
  - “Advanced economies” (column (d)): β_low = 0.008, β_high = -0.008, gamma = 1, (st.dev) = 0.84.
  - “Emerging and developing economies” (column (e)): β_low = 0.001, β_high = -0.01, gamma = 1, (st.dev) = 1.12.
  - “Effect on non-oil real GDP” (column (f)): β_low = 0.028, β_high = -0.082, gamma = 1, (st.dev) = 0.98.
- Model behavior:
  - LSTR models provide estimates of γ that are fairly high, implying costly effects of inflation shortly after threshold is reached.
  - Exception: LSTR estimate for oil producers suggests a very low γ and a very high threshold; this is not robust and likely driven by outliers.
  - TAR models were used as checks; TAR appears more robust than LSTR for oil exporters in this dataset.
  - Models assuming medium-term causality from inflation to growth (5-year averages) and robustness checks with lagged inflation and lagged investment yield similar threshold estimates (slightly higher for Emerging Markets).

### VII. Bootstrapping inference (panel-aware)
- Procedure: panel-aware bootstrap accounting for estimated non-linearities.
- Key bootstrap outcomes:
  - For developing countries: estimated inflation threshold lies between 7 and 13 percent for more than 90 percent of the bootstrapped sample.
  - For advanced economies: bootstrap mode at 1 percent; threshold <5 percent in 43 percent of bootstraps and <18 percent in 75 percent of bootstraps.
  - For oil exporters: bootstrap histogram shows two modes; TAR mode at 10 percent and a cluster with thresholds >100 percent; conclusion: oil-exporter group is not homogeneous enough for robust estimation for total GDP growth.
  - When using non-oil GDP growth for oil exporters: threshold robust around 13 percent.
- Standard errors for TAR threshold estimates are usually small, around 1 percentage point (as reported in the text).
- Bootstrapping explains much of the variation in threshold estimates across the literature.

### VIII. Monte-Carlo analysis: TAR vs LSTR
- Experiment designs:
  - (i) Artificial data from a low-γ LSTR model (γ = 6) estimated as TAR.
  - (ii) Artificial data from a TAR model (approximated by γ ≥ 15) estimated as LSTR.
- Findings:
  - LSTR estimation captures the value of the threshold and the negative sign of β_high, but β_low is often not statistically different from 0 (matches bootstrap results).
  - LSTR often underestimates γ; for a TAR (γ → ∞ approximated by γ ≥ 15), LSTR estimated γ ≥ 15 only 25 percent of the time.
  - For simulated LSTR DGP estimated as TAR, TAR does not bias the threshold estimate significantly and obtains a significantly negative β_high; TAR cannot evaluate the size of γ.

### IX. Quantitative effect examples (selected entries from Table 2)
- Table 2 reports β_low, β_high, γ and standard deviations, and effects on annual GDP growth for various initial and new inflation pairs. Selected reported values:
  - Thresholds: 9, 8, 1, 1, 10, 11, 74, 10, 13, 13.
  - “All countries” parameters: β_low = 0.002, β_high = -0.01, gamma = 6, (st.dev) = 1.08.
  - “Advanced economies” parameters: β_low = 0.008, β_high = -0.008, gamma = 1, (st.dev) = 0.84.
  - “Emerging and developing economies” parameters: β_low = 0.001, β_high = -0.01, gamma = 1, (st.dev) = 1.12.
  - “Effect on non-oil real GDP” parameters: β_low = 0.028, β_high = -0.082, gamma = 1, (st.dev) = 0.98.
- Example effect reported in text:
  - For full sample: when inflation is above 9 percent, a doubling of inflation decreases GDP by 0.7 percentage point per year.

### X. Policy implication and conclusion
- LSTR models confirm that inflation tends to be costly quickly, implying that inflation has to be tackled sooner rather than later.
- For emerging market economies: inflation above about 10 percent quickly becomes harmful to growth, suggesting the need for a prompt policy response at or above that threshold.
- For advanced economies: the much lower threshold (bootstrap mode at 1 percent) implies a smaller tolerance for inflation before growth is harmed.
- For oil exporters: heterogeneity warrants caution in applying uniform thresholds; non-oil growth responses indicate a threshold around 13 percent but stronger adverse effects of higher inflation relative to other emerging markets.

*Source: _wp1076 - Appendix A. Country Groups (excerpts provided in source content).*

### Appendix A. Country Groups .............................................................................................

### _wp1076 - Appendix A. Country Groups

### I. Introduction
- Purpose: revisit the relationship between inflation and growth and estimate the speed at which inflation beyond a threshold becomes harmful to growth using a Logistic Smooth Transition model (LSTR).
- Key methodological advance: estimate both a threshold level for inflation and the speed of transition (γ) from low-cost to high-cost inflation regimes; validate estimates using bootstrapping and Monte-Carlo experiments.
- Core empirical summary:
  - Study estimates a threshold of about 10 percent for most country groups (except advanced economies).
  - Estimated transition speed is fairly high, implying inflation becomes harmful to growth soon after it exceeds the threshold.
  - Bootstrapping shows the range of estimates in the literature can be driven by data-process variability as well as estimation technique.
  - For developing countries, more than 90 percent of the bootstrapped sample places the estimated inflation threshold between 7 and 13 percent.
  - Inflation appears more costly for oil-exporters than for other groups when the dependent variable is non-oil GDP growth (as opposed to total GDP growth).
- Monte-Carlo result: LSTR provides additional flexibility, but distinction between LSTR and TAR is not crucial for these data-generating processes.

### II. The Backdrop (2007/08 inflation episode)
- Global context and numeric snapshots:
  - Advanced economies: headline inflation rose to 3.5 percent in May 2008 (year-on-year); core inflation remained at 1.8 percent.
  - Emerging and developing economies: headline inflation rose to 8.6 percent; core inflation rose to 4.2 percent.
- Inflation drivers identified:
  - Strong capital inflows, rapid credit growth, tightening labor markets, widening current account deficits in several countries (notably emerging Europe).
  - Rising export earnings and domestic demand in commodity-exporting countries (GCC, Russia); capacity constraints (e.g., housing, raw materials) in GCC.
  - Surging commodity (food and fuel) prices boosting inflation globally, with especially large effects in emerging markets and low-income countries.
- Policy responses observed:
  - Use of monetary, trade, and fiscal measures; price and quantity controls, administered prices, subsidies, buffer-stock building.
  - Policy choice influenced by exchange rate regime: flexible regimes relied more on monetary/fiscal mix; pegged regimes relied more on fiscal measures, capacity increases, and prudential actions.
- Outcome note: inflation remained high until housing-market effects and the subsequent crisis reduced rents and commodity/asset prices.

### III. Literature review — key findings relevant to threshold estimation
- Main empirical questions in literature:
  - Is there a robust negative relationship between inflation and growth?
  - Is there a “kink” (nonlinearity) where low inflation can correlate positively with growth but high inflation correlates negatively?
  - Is there a minimum threshold beyond which inflation’s adverse growth effects become serious?
- Representative prior findings (preserve reported numeric thresholds and ranges exactly):
  - Mubarik (2005, Pakistan): inflation below 5 percent positive for growth; depresses growth if above 9 percent.
  - Rangarajan (1997, India): suggested range of 5 to 7 percent.
  - Samantaraya and Prasad (2001): estimated threshold at 6.5 percent.
  - Sarel (1996, 87 countries, 1970–90): structural break at 8 percent — positive correlation below 8 percent, negative above.
  - Ghosh and Philips (1998): positive correlation at about 2 to 3 percent; reversed at higher rates; convex relationship implies larger growth decline for increases from 10 to 20 percent than from 40 to 50 percent.
  - Khan and Senhadji (2001, 140 countries, 1960–98): threshold between 1 and 3 percent for industrial countries and 7 and 11 percent for developing countries.
  - Drukker, Gomis-Porqueras, and Hernandez-Verme (2005): threshold about 19 percent.
  - Pollin and Zhu (2006): threshold around 15-18 percent (sample divided by decades).
  - Burdekin et al (2004): for developing countries, inflation costly when higher than 3 percent; second break at 50 percent with marginal growth costs decreasing by 25 percent.
- Contribution of this study relative to literature:
  - Extends estimation period and number of countries.
  - Distinguishes emerging markets and oil exporters.
  - Estimates speed of transition between regimes.
  - Uses bootstrap techniques to assess precision of threshold and transition-speed estimates.

### IV. Inflation and growth — Threshold level (Estimating the LSTR model)
- Data and specification details:
  - Panel of 165 countries covering 1960–2007.
  - Data averaged over 5-year periods (time dimension: 9 periods of 5-year averages).
  - Controls: ratio of investment to GDP, population growth, initial GDP, rate of change in the terms of trade, variability of the terms of trade.
  - Country dummies and time dummies included; excluded periods of severe recession (losses > 5 percent of GDP per year over 5 years) for specific noted episodes.
  - Explanatory variable f(π) = ln(1+π) (logarithmic function with constant 1) to model multiplicative shocks in inflation.
  - Use of LSTR: allows continuous (smooth) transition between low- and high-inflation regimes via weight W_high and transition speed γ.
- Interpretation of LSTR parameters (preserve numeric examples exactly):
  - W_high weight goes from 0 (low-inflation regime) to 1 (high-inflation regime); W_low = 1 − W_high.
  - Example weights (with threshold π* = 10 and st.dev = 0.5 shown in Figure 3):
    - If γ = 3 and inflation = 13 percent: effect captured by 0.3*β_low + 0.7*β_high.
    - If γ = 15 and inflation = 13 percent: effect captured by 0.003*β_low + 0.997*β_high.
  - γ captures speed of transition; γ → ∞ corresponds to standard TAR (instantaneous regime switch).
- Main empirical estimates and exact reported figures:
  - Full sample (165 countries): threshold inflation ≈ 9 percent.
  - When inflation is above 9 percent, a doubling of inflation decreases GDP by 0.7 percentage point per year.
  - Advanced economies: estimated threshold is 1 percent; costs of inflation almost three times smaller than for full sample.
  - Post-1990 sub-sample: threshold about 11 or 12 percent for all countries; for advanced economies non-linearity reduced and threshold estimated at 14 percent (precision hampered by smaller sample).
  - Developing countries (bootstrap): for more than 90 percent of the bootstrapped sample, the estimated inflation threshold lies between 7 and 13 percent.
  - Oil-exporting countries: LSTR point estimate of the threshold inflation is 74 percent; TAR model would give threshold = 10 percent — result for oil exporters is not robust and bootstrapping modes should be considered for judgment.
  - Non-oil GDP specification: estimates stronger than previous ones; text reports "a doubling of inflation from higher than 13 percent decreases real non-oil GDP by" — (text ends here; no further numeric completion provided in source).
- Robustness checks and inference methods:
  - Standard errors built using 500 bootstrapping iterations conditioned on threshold from base estimation; more general bootstrap results discussed in text.
  - Bootstrapping confirms robustness of some threshold estimates (e.g., 1 percent threshold for whole sample estimated precisely) but also shows wide bootstrap distributions that help explain differing literature estimates.
- Policy implication emphasized by γ estimate:
  - High γ implies inflationary pressures must be tackled promptly because inflation becomes costly almost immediately after exceeding the threshold.
  - Standard TAR models (γ → ∞) do not provide information on transition speed; LSTR fills this gap.

*Source: _wp1076 - Appendix A. Country Groups (excerpts provided in source content).*

### 2.7 percent per year. Although we do not have the data to investigate the cause of this

### _wp1076 - 2.7 percent per year. Although we do not have the data to investigate the cause of this

### Main findings on inflation-growth non-linearities and thresholds
- Using a panel of 165 countries covering the period 1960–2007, the study finds that for emerging market economies inflation above a threshold of about 10 percent quickly becomes harmful to growth.
- For advanced economies, the inflation threshold is much lower (mode of the bootstrap distribution is at 1 percent; the threshold is found to be lower than 5 (resp. 18) percent in only 43 (resp. 75) percent of the bootstrapped samples).
- For oil exporting countries, estimates are less robust: bootstrapping shows two potential modes for the threshold distribution (one TAR mode at 10 percent and a cluster of simulations with estimated threshold exceeding 100 percent), indicating heterogeneity and influence of outliers.
- When using non-oil GDP growth for oil exporters, the threshold estimate is robust around 13 percent, consistent with Emerging Market group estimates; however, the adverse effect of higher inflation on (non-oil) growth is stronger for oil producing countries than for other emerging markets.
- Across models, the link between growth and inflation at high levels of inflation is almost always negative; coefficients linking growth and inflation at low levels are not consistently different from zero.

### Model estimates and interpretation
- LSTR (Logistic Smooth Transition Regression) models provide estimates of γ that are fairly high, implying that starting from a level of inflation of 10 percent, doubling inflation is more costly when the speed of transition (γ) is high.
- Exception: LSTR estimate for oil producers suggests a very low speed of transition (very low γ) and a very high threshold; this result is not robust and likely driven by outliers.
- TAR (Threshold AutoRegressive) models were estimated as a check; results are comparable and TAR appears more robust than LSTR for oil exporters in this dataset.
- The study identifies models assuming medium-term causality runs from inflation to growth (variables expressed as 5-year averages); robustness checks use lagged inflation and lagged investment to GDP ratio with 3-year averages and find threshold estimates very similar (slightly higher for Emerging Markets) to contemporaneous inflation estimates.

### Bootstrapping inference (panel-aware, accounts for non-linearities)
- Statistical inference performed using a bootstrapping technique that accounts for the panel structure and estimated non-linearities.
- For developing countries, the estimated inflation threshold lies between 7 and 13 percent for more than 90 percent of the bootstrapped sample.
- For advanced economies, bootstrap mode is at 1 percent; threshold <5 percent in 43 percent of bootstraps and <18 percent in 75 percent of bootstraps.
- For oil exporters:
  - Bootstrapping histogram shows two modes; TAR mode at 10 percent and a cluster with thresholds >100 percent.
  - Conclusion: country group of oil exporters is not homogeneous enough for robust estimation for total GDP growth.
  - When using non-oil GDP growth, threshold is robust around 13 percent.
- Standard errors for TAR threshold estimates are usually small, around 1 percentage point (as reported in the text).
- The bootstrapping exercises suggest most results are robust (see Figures 4 to 6 in the source).

### Monte-Carlo analysis: TAR vs LSTR
- Monte-Carlo experiments generate artificial data from (i) a low-γ LSTR model (γ = 6) and estimate TAR, and (ii) a TAR model (approximated by γ ≥ 15) and estimate LSTR.
- Findings:
  - LSTR estimation captures the value of the threshold and the negative sign of β_high, but β_low is often not statistically different from 0 (matches bootstrap results).
  - LSTR often underestimates γ; for a TAR (γ → ∞ approximated by γ ≥ 15), LSTR estimated γ ≥ 15 only 25 percent of the time.
  - For simulated LSTR DGP estimated as TAR, TAR does not bias the threshold estimate significantly and obtains a significantly negative β_high; TAR cannot evaluate the size of γ.

### Quantitative effect examples (from Table 2)
- Table 2 reports β_low, β_high, γ and standard deviations, and effects on annual GDP growth for various initial and new inflation pairs for different groups. Selected reported values:
  - Thresholds reported in Table 2: 9, 8, 1, 1, 10, 11, 74, 10, 13, 13 (as shown in the table header/entries).
  - For “All countries” (column (a)) reported parameters include β_low = 0.002, β_high = -0.01, gamma = 6, (st.dev) = 1.08.
  - For “All countries (if gamma = 15)” (column (c)) reported parameters include β_low = 0.002, β_high = -0.01, gamma = 15, (st.dev) = 1.08.
  - For “advanced economies” (column (d)) reported parameters include β_low = 0.008, β_high = -0.008, gamma = 1, (st.dev) = 0.84.
  - For “emerging and developing economies” (column (e)) reported parameters include β_low = 0.001, β_high = -0.01, gamma = 1, (st.dev) = 1.12.
  - For “effect on non-oil real GDP” (column (f)) reported parameters include β_low = 0.028, β_high = -0.082, gamma = 1, (st.dev) = 0.98.
- Table 1 and Table 3 present coefficients, thresholds, standard deviations, γ estimates, Hansen F-tests, R-squared, observations, and robustness checks across subsamples (All countries, Advanced economies, Emerging Markets, Oil producers, Oil producers - non oil GDP, Post 1990, lagged inflations, etc.). Selected exact figures from Table 1 and Table 3 appear in the source tables (e.g., Threshold values 9, 8, 1, 1, 10, 11, 74, 10, 13, 13; Stdev threshold 1.13, 1.32, 1.58, 1.48, 0.91; Gamma figures 61, 36, 11, 1 in table segments).

### Policy implication and conclusion
- The LSTR models confirm that inflation tends to be costly quickly, implying that inflation has to be tackled sooner rather than later.
- Motivated by the global inflation episode of 2007–08 and risk of monetary policies during the global financial crisis spurring global inflation, the paper underscores that for emerging market economies inflation above about 10 percent quickly becomes harmful to growth, suggesting the need for a prompt policy response at or above that threshold.
- For advanced economies, the much lower threshold implies that tolerance for inflation before growth is harmed is smaller.
- For oil exporters, heterogeneity warrants caution in applying uniform thresholds; non-oil growth responses indicate a threshold around 13 percent but stronger adverse effects of higher inflation relative to other emerging markets.

*Source: IMF working paper text provided in the content unit.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1076.pdf_