## wp1875

## Source details

**Canonical URL:** [wp1875](https://www.imf.org/-/media/files/publications/wp/2018/wp1875.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp1875.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp1875.pdf.json)

---

### Global context and the "new normal"
- "The global recovery is continuing, but it is still incomplete" (World Economic Outlook, October, 2017).
- World economy growth rate projected to be 3.9 percent in 2018.
- Five year average of real global GDP growth rate: 3.5 percent (¾ percent smaller than average of upswing periods since 1980; more than 1.5 percent lower compared to 2003-2007).
- "New Normal": lower output growth expected as the new norm, especially in advanced economies, due to prolonged deleveraging and re-orientation toward a new growth paradigm.

### Research question, approach, and data
- Core questions:
  - Do more globally integrated countries have higher short-term output growth on average?
  - Do more integrated countries suffer larger short-term output contractions when a financial crisis occurs?
- Empirical strategy:
  - Globalization used as a transition variable inducing regime shifts in an autoregressive growth model.
  - Dynamic panel with an endogenously and recursively estimated globalization threshold via GLM following Hansen (1996); 15 percent trimming value; robustness checks include GMM and alternative trimming values.
  - KOF Globalization Index (and subindices) as transition variable.
  - Model: short-term output growth depends on lags and crisis dummies (banking, currency, stock market), with regime-dependent constants, autoregressive coefficients, and crisis impacts.
- Data:
  - Real GDP growth from IMF WEO up to 2016.
  - Crisis dummy database expanded to 102 countries, 1970–2016 (core from Reinhart and Rogoff, supplemented with HBS, Laeven and Valencia (2012), and calculated dummies).
  - KOF Globalization Index (economic, social, political subindices).
  - Balanced panel: 92 countries, yearly, 1970–2014.

### Key high-level empirical findings (summary)
- In higher-globalization regimes (KOF or subcomponents above estimated threshold):
  - Short-term output growth is higher on average.
  - Output growth contracts more when a financial crisis occurs (banking and stock market crises particularly).
  - The autoregressive growth coefficient is lower in the higher-globalization regime, implying slower recovery and producing a persistent (not permanent) "new normal" of lower output growth following crises.
- Heterogeneity by income:
  - High-income countries: stock market crisis negative regardless of globalization; effect stronger above threshold; banking crises negative above threshold.
  - Low- and middle-income countries: currency and stock market crises negative in higher-globalization regimes.
  - Low- and middle-income countries reach benefits and vulnerabilities earlier in globalization.

---

### Estimation tests and approach (preliminary)
- Block-bootstrap-based Log Likelihood Ratio test (Hansen 1996) used to test linearity; if rejected, model estimated via GLM with KOF indices sequentially as transition variables.
- Threshold maximizes Log Likelihood Ratio; confidence bounds via block bootstrap with 10,000 replications.

### Overall KOF Globalization Index results
- Linearity test p-value: <1%.
- Full-sample (selected coefficients, Table 1):
  - Intercept: 2.806***, se. 0.689
  - g_{i,t−1}: 0.226***, se. 0.017
  - Banking_{i,t−1}: -0.432, se. 0.295
  - Currency_{i,t−1}: -0.630***, se. 0.229
  - Stock_{i,t−1}: -1.900***, se. 0.330
- Estimated threshold (average): 64.
- Threshold confidence interval: [57 : 95].
- Sample (γ >64 [57 : 95]) estimates:
  - Intercept: 0.480, se. 0.335
  - g_{i,t−1}: -0.162***, se. 0.039
  - Banking_{i,t−1}: -1.303***, se. 0.509
  - Currency_{i,t−1}: 0.443, se. 0.571
  - Stock_{i,t−1}: -1.676***, se. 0.531
- Timing and cross-country pattern:
  - Most countries cross the globalization threshold in the 1990s or early 2000s.
  - Countries above threshold tend to be high income; some middle- or low-income cross in early 2000s; others later or never.
- Economic interpretation and magnitudes:
  - In high-globalization regime (overall KOF >64) countries have about 0.48 percent higher output growth (term marginally significant; t-stat = 1.45).
  - Banking crisis in higher-globalization regime → decline in growth by 1.3 percent (Banking_{i,t−1} = -1.303***).
  - Stock market crisis in higher-globalization regime → decline in growth by 1.7 percent (Stock_{i,t−1} = -1.676***).
  - Recovery slower in higher-globalization regime because combined AR coefficient = +0.064 (0.226 − 0.162).
  - Currency crises do not negatively impact growth in a high-globalization regime.
- Synthesis: globalization raises growth potential but increases financial exposure; about half the countries pass the estimated threshold by 2014.

### KOF subcomponents as transition variables (selected results)
- Full-sample coefficients (selected from Table 3):
  - Intercepts (full-sample): 2.725*** (KOF Economics), 2.747*** (KOF Actual Flows), 2.804*** (KOF Social), 2.602*** (KOF Political) with respective standard errors.
  - g_{i,t−1}: 0.239*** (Economics), 0.234*** (Actual Flows), 0.228*** (Social), 0.262*** (Political).
  - Banking_{i,t−1}: -0.526*** (Economics), -0.599*** (Actual Flows), -0.444 (Social), -0.518 (Political).
  - Currency_{i,t−1}: -0.478*** (Economics), -0.482*** (Actual Flows), -0.605*** (Social), -0.912*** (Political).
  - Stock_{i,t−1}: -2.132*** (Economics), -2.155*** (Actual Flows), -2.166*** (Social), -1.236*** (Political).
- Estimated thresholds and confidence intervals:
  - KOF Economics: γ >73 [54−76]
  - KOF Actual Flows: γ >76 [60−77]
  - KOF Social: γ >65 [49−69]
  - KOF Political: γ >56 [43−56]
- Higher-globalization regime patterns (economic / actual flows):
  - Average growth in higher regime higher by about 1¾ percent.
  - Banking and stock market crises induce an extra ~2 percent decline.
  - Only about one third of countries pass economic and actual flows thresholds by 2014; mainly high income countries passing in mid-1990s or earlier.
- Interpretation: trade/globalization aspects contribute positively to higher growth but also increase vulnerability to financial crises.

### Results by income-per-capita groups
- Clustering into two groups: high-income countries vs. middle- and low-income countries.

- High-income countries (selected, Table 4):
  - Full-sample (KOF Overall):
    - Intercept: 2.410***, se. 0.738
    - g_{i,t−1}: 0.375***, se. 0.038
    - Banking_{i,t−1}: 0.184, se. 0.642
    - Currency_{i,t−1}: -0.271, se. 0.483
    - Stock_{i,t−1}: -2.025***, se. 0.539
  - Estimated thresholds vary by index (examples): KOF Overall γ >64; KOF Economics γ >39; KOF Actual Flows γ >73; KOF Social γ >65; KOF Political γ >76.
  - In higher-globalization regime (e.g., γ >64), banking crises negatively affect growth: Banking_{i,t−1} = -1.883***, se. 0.750 (KOF Overall sample).
  - Stock market crises significantly negative in higher regimes (e.g., Stock_{i,t−1} = -1.226**, se. 0.667).

- Middle- and low-income countries (selected, Table 5):
  - Full-sample (KOF Overall):
    - Intercept: 0.983, se. 0.834
    - g_{i,t−1}: 0.166***, se. 0.039
    - Banking_{i,t−1}: -0.251, se. 1.067
    - Currency_{i,t−1}: 0.394, se. 0.650
    - Stock_{i,t−1}: 5.346, se. 3.412
  - Estimated thresholds (examples): KOF Overall γ >27; KOF Economics γ >24; KOF Actual Flows γ >29; KOF Social γ >16; KOF Political γ >39.
  - In higher-globalization regimes for middle- and low-income countries:
    - Intercept example: 1.997***, se. 0.439 (KOF Overall sample).
    - g_{i,t−1}: 0.025, se. 0.045 (KOF Overall higher-regime); 0.128***, se. 0.049 (KOF Economics higher-regime).
    - Currency crises often negative and significant (e.g., Currency_{i,t} = -1.148**, se. 0.710 for KOF Overall sample).
    - Stock market crises in higher regimes strongly negative (e.g., Stock_{i,t−1} = -7.692***, se. 3.429 for KOF Overall sample).
  - Interpretation:
    - Globalization thresholds much lower for middle- and low-income countries (~27) than for high-income (~64).
    - Low-income countries gain about 1−2 percent higher GDP growth once they reach their threshold.
    - In higher-globalization regimes currency and stock market crises affect growth more negatively; banking systems may be insulated but markets are shallow and illiquid.
    - These countries benefit sooner from opening up but also become more vulnerable.

### Robustness checks (selected)
- GMM estimation with contemporaneous regressors (Table 6):
  - GMM results consistent with GLM results; GMM thresholds lie within block-bootstrapped confidence bounds.
  - Full-sample GMM examples:
    - KoF Overall Intercept: 2.209***, se. 0.857
    - g_{i,t−1}: 0.402***, se. 0.026
    - Banking_{i,t}: -0.170, se. 0.241
    - Stock_{i,t}: -1.967***, se. 0.259
  - GMM estimated thresholds examples: KoF Overall γ >64; KoF Economics γ >78; KoF Actual Flows γ >79; KoF Social γ >51; KoF Political γ >69.
  - In high-globalization regimes under GMM: positive constants, negative and significant autoregressive coefficients, and significant negative coefficients for banking and stock market crisis dummies for overall and economic subindices.

- Alternative transition variable: price of brent (Table 7):
  - Linearity test p-value: <1%.
  - Full-sample (brent transition) coefficients:
    - Intercept: 3.241***, se. 0.692
    - g_{i,t−1}: 0.110***, se. 0.020
    - Banking_{i,t}: -1.001***, se. 0.333
    - Currency_{i,t}: -0.875***, se. 0.299
    - Stock_{i,t}: -1.861***, se. 0.382
  - Estimated brent threshold: γ >38 (US dollars per barrel).
  - Sample (brent >38) coefficients:
    - Intercept: -1.206***, se. 0.206
    - g_{−1}: 0.208***, se. 0.031
    - Banking_{i,t}: 0.397, se. 0.467 (not significant)
    - Currency_{i,t}: 0.820***, se. 0.389 (positive and significant)
    - Stock_{i,t}: -1.172***, se. 0.498
  - Differences vs. globalization-driven regimes:
    - Banking crisis dummy not significant when brent >38 US dollars per barrel, whereas banking crises are significant under globalization transitions.
    - Constant highly negative in high-oil-price regime (consistent with oil-importing country effects).
    - In a high-oil-price regime currency crisis dummy positive, implying currency crises can foster growth in oil-exporting contexts due to realignment effects.

---

### Stylized facts and takeaways (summary and magnitudes)
- Main stylized result: in high-globalization regimes banking and stock market crises have larger negative growth effects; recovery is slower.
- Stylized correlations (Table 8: banking crisis vs. growth):
  - Periods: 1970-1994, 1995-2016
  - Sample: -0.18, -0.28
  - KOF<64: -0.18, -0.17
  - KOF>64: -0.18, -0.40
  - Interpretation: for KOF>64 correlation much more negative in second part of sample (−0.4 vs. −0.18).

- Evolution by period and average KOF (Table 9):
  - Periods: 1970-1990, 1991-2000, 2001-2010, 2011-2016
  - All countries: -0.14, -0.28, -0.43, -0.40
  - Average KOF∈(25,40]: -0.16, 0.01, N/A, N/A
  - Average KOF∈(40,64]: -0.25, -0.27, -0.27, N/A
  - Average KOF∈(64,80]: -0.12, -0.63, -0.38, -0.76
  - Average KOF>80: N/A, -0.38, -0.54, -0.37
  - Interpretation: largest negative correlation for overall sample during 2001-2010 (−0.43); most globalized countries show lowest average growth during 2001-2010 but absence of a recent banking crisis implies no automatic "new normal" for all highly globalized countries.

- Average real GDP growth by period and KOF (Table 10):
  - Columns: sample (1970-2016), 1970-1990, 1991-2000, 2001-2010, 2011-2016
  - All countries: 3.47, 6.32, 1.76, 2.65, 3.64
  - Average KOF∈(25,40]: 3.55, 5.94, 2.44, 5.48, 2.72
  - Average KOF∈(40,64]: 3.80, 6.88, 3.44, 2.62, 4.32
  - Average KOF∈(64,80]: 3.06, 5.85, -2.37, 1.01, 4.75
  - Average KOF>80: 2.24, N/A, 1.36, 2.13, 1.29
  - Interpretation: for 2011-2016 average real output growth for all groups similar to or higher than in the 1990s; KOF tends to increase over time; globalization increases growth on average but above a threshold increases vulnerability to larger short-term negative impacts from financial crises.

- Conceptual takeaway: "tale of two narratives"
  - Globalization (especially economic) raises output growth significantly.
  - In high-globalization regimes high-income countries: banking crises negative and stock market crises stronger negative.
  - Low-income countries: currency and stock market crises negative; benefits and vulnerabilities accrue earlier.
  - Paper documents a "new normal": persistently (but not permanently) lower real GDP growth regime following a financial crisis for countries in a high-globalization regime; eventual return to higher growth in higher-globalization equilibrium.
  - Results robust to specifications and not driven solely by 2007−2008 global financial crisis.

### Policy implications and recommendations (selected, precise)
- Broad prescriptions:
  - Sound macroeconomic and financial policies essential to obtain globalization’s benefits and mitigate vulnerabilities.
  - Appropriate regulation and supervision of banks and capital markets.
  - Good international financial policy coordination.
  - Avoid synchronizing financial deregulation with increases in globalization.
- Specific measure discussed:
  - Consider saving benefits of higher growth (e.g., international reserves or fiscal surpluses) into a financial stabilization fund to be used to recapitalize banks or financial institutions facing a systemic crisis (practical implementation to be analyzed in future research).
- Consistency with other assessments:
  - Findings align with 2017 BIS Annual Report and literature (Cecchetti (2012), Shin (2017), Broner and Ventura (2016)) that globalization is beneficial up to a point unless mitigating policies are in place.
  - A globalization reversal would not be optimal given remaining gains from deeper integration for many countries.

*Source: wp1875 — IMF Working Paper (sections 1, 4, and 7).*

### 1. Introduction

### 1. Introduction

### Global context and the "new normal"
- "The global recovery is continuing, but it is still incomplete" (World Economic Outlook, October, 2017).
- The world economy’s growth rate is projected to be 3.9 percent in 2018, more than 1 percent lower than it was during the upswing prior to the global financial crisis.
- The 5 year average of the real global GDP growth rate, at 3.5 percent, is ¾ percent smaller than the average of all similar upswing periods since 1980, and more than 1.5 percent lower compared to the average global growth rate during 2003-2007.
- "New Normal" (term coined by McKinsey and PIMCO in 2009) denotes a notion that lower output growth is expected to be the new norm, especially in advanced economies, due to prolonged deleveraging and a re-orientation towards a new growth paradigm.

### Literature, theory, and empirical gaps
- Prior empirical findings:
  - Cerra and Saxena (2008): output losses following financial and some political crises are highly persistent.
  - Candelon, Carare, and Miao (2016): allowing common factors raises short-term output growth and the negative impact of crises; labeled a common factor "globalization".
- Theoretical channels by which globalization affects growth:
  - Trade openness: (ia) rising market size of production, (ib) increased competition among firms.
  - Financial openness: (iia) more efficient allocation of capital, (iib) technology and know-how transfer, (iic) greater opportunities for returns and risk diversification.
- Empirical literature finds positive contributions from trade openness; gains from financial openness are mixed due to greater vulnerability to crises (Ashenfelter and others, 2017).
- Identified gap: scarce comprehensive empirical evidence reconciling the dual narratives that globalization raises average growth while increasing crisis vulnerability.

### Research question and empirical strategy
- Core questions:
  - Do countries that open up and integrate into the world have, on average, higher short-term output growth?
  - Do more integrated countries suffer larger short-term output contractions when a financial crisis occurs?
- Approach:
  - Allow globalization to affect growth simultaneously in both directions by using globalization as a transition variable that induces regime shifts in an autoregressive growth model.
  - Use a dynamic panel with an endogenously and recursively estimated globalization threshold via a Generalized Linear Model (GLM) following Hansen (1996).
  - Use the KOF Globalization Index (and subindices) as the transition variable to capture economic, social, and political dimensions together.
  - Model specification highlights:
    - Short-term output growth depends on its lags and crisis dummies (following Cerra and Saxena (2008)), with fixed effects and regime-dependent constants, autoregressive coefficients, and crisis impacts.
    - The threshold γ is estimated; significant positive constant in high-globalization regime implies higher average growth above γ; significant negative crisis coefficients above γ imply greater crisis sensitivity in high-globalization regimes.
    - Estimation: recursive GLM with a 15 percent trimming value; robustness checks include GMM and alternative trimming values.

### Data
- Real GDP growth: IMF World Economic Outlook (WEO) database, data up to 2016.
- Crisis dummy database:
  - Core from Reinhart and Rogoff (banking, currency, stock market crashes) covering 70 countries from 1970 until 2010.
  - Supplemented with Harvard Business School (HBS) database and calculated dummies using Reinhart and Rogoff definitions, WEO exchange rate data, stock indices, and Laeven and Valencia (2012) for banking crises.
  - Resulting crisis dummy coverage: 102 countries, from 1970 up to 2016, yearly frequency (Appendix 1 contains definitions and counts).
- Globalization index:
  - KOF Globalization Index (Dreher and others, 2006) used as composite indicator (economic, social, political globalization).
  - The economic subindex includes trade of goods, financial flows, and trade and capital account restrictions; social subindex includes personal contact, information flows, cultural proximity; political subindex includes country representation in international structures and treaties.
- Sample used for estimations:
  - Balanced panel for 92 countries, yearly frequency, from 1970 to 2014. These 92 countries match the longest span available and the expanded crisis dummy database.
- Stylized data patterns:
  - Globalization increases steadily over time; after a brief slowdown in the mid-eighties, average globalization speeds up until 2007 with no sudden stops or strong accelerations.
  - Economic globalization increases faster since the mid-nineties, especially for the least open country; social globalization increases slightly less, and political globalization rises faster in the early nineties.

### Key empirical findings (summary)
- In a higher-globalization regime (KOF Globalization Index or subcomponents above an estimated threshold):
  - Short-term output growth is higher on average.
  - Output growth contracts more when a financial crisis occurs.
  - The autoregressive growth coefficient is lower in the higher-globalization regime, implying slower recovery and creating a persistent (yet not permanent) "new normal" of lower output growth following crises.
- Heterogeneous findings by income group:
  - High-income countries: a stock market crisis has a negative impact on short-term growth regardless of globalization level; once globalization reaches a threshold, this effect is stronger, and banking crises also have a negative impact.
  - Low- and middle-income countries: a currency crisis and a stock market crisis have a negative effect on growth in a higher-globalization regime.
  - Benefits and vulnerabilities accrue earlier in the globalization process for low- and middle-income countries, consistent with Broner and Ventura (2016).
- Robustness:
  - Results are robust to various KOF Globalization subindices and alternative empirical specifications, including ones allowing for endogeneity (GMM reported in Robustness Checks).
  - Findings are not driven by the tail end of the sample, nor solely by the 2007-08 crisis.

### Contributions and normative implications
- Contributions:
  - Thoroughly document the existence of a "new normal".
  - Expand crisis dummy databases to over one hundred countries up to 2016.
  - Reconcile two previously opposing narratives: globalization raises average growth but increases crisis sensitivity, by introducing threshold effects of globalization on growth.
- Policy implications:
  - Systematic evidence of a new normal informs policy debate and theoretical modeling.
  - Recommended policy responses: implement stabilization policies accompanied by strong structural reforms to mitigate long-run effects of negative short-run developments; ensure adequate financial regulation and supervision to reap the rewards of globalization.
  - Theoretical models should allow for slower convergence to a unique steady state to reflect persistent post-crisis lower growth dynamics.

*Source: wp1875 - 1. Introduction (wp1875 - 1. Introduction).*

### 4. Results

### 4. Results

### Preliminary tests and estimation approach
- A block-bootstrap-based Log Likelihood Ratio test like in Hansen (1996) is performed for the dynamic panel model to check whether the null of linearity is rejected or not.
- If linearity is rejected, the dynamic panel model (3) is estimated via GLM, using the overall KOF Globalization Index and the four subindices (economic, actual flows, social, political) sequentially as transition variables.
- The threshold estimate corresponds to the value that maximizes the Log Likelihood Ratio.
- Confidence bounds are obtained via block bootstrap with 10.000 replications.

### Overall KOF Globalization Index as transition variable
- Linearity test p-value: <1%.
- Full-sample estimates (Table 1):
  - Intercept: 2.806***, se. 0.689
  - g_{i,t−1}: 0.226***, se. 0.017
  - Banking_{i,t−1}: -0.432, se. 0.295
  - Currency_{i,t−1}: -0.630***, se. 0.229
  - Stock_{i,t−1}: -1.900***, se. 0.330
- Estimated threshold (average) for overall KOF Globalization Index: 64.
- Threshold confidence interval: [57 : 95].
- Sample (γ >64 [57 : 95]) estimates:
  - Intercept: 0.480, se. 0.335
  - g_{i,t−1}: -0.162***, se. 0.039
  - Banking_{i,t−1}: -1.303***, se. 0.509
  - Currency_{i,t−1}: 0.443, se. 0.571
  - Stock_{i,t−1}: -1.676***, se. 0.531
- Timing and cross-country pattern:
  - Most countries cross the globalization threshold in the 1990s or early 2000s.
  - Countries above the threshold over the sample, or from an earlier stage, are high income countries.
  - Some middle- or low-income countries cross the threshold in the early 2000s; some pass later or never.
  - Figure 3 reports cumulative counts of countries crossing the estimated threshold every year since 1970.
- Economic interpretation:
  - In a high-globalization regime (overall KOF >64), countries have about 0.48 percent higher output growth rates (term marginally significant; t-stat equals 1.45).
  - Higher globalization increases exposure to crises, in particular banking and stock market crises.
  - In the higher-globalization regime, occurrence of a banking crisis leads to a decline in growth by 1.3 percent.
  - In the higher-globalization regime, occurrence of a stock market crisis leads to a decline in growth by 1.7 percent.
  - Recovery is slower in the higher-globalization regime because the autoregressive coefficient is lower: combined coefficient = +0.064 (0.226 - 0.162).
  - Currency crises do not negatively impact growth in a high-globalization regime.
- Synthesis: globalization raises growth potential (trade gains) but increases financial exposure; about half the countries pass the estimated threshold by 2014, the others can still gain short-term growth from opening up.

### KOF subcomponents as transition variables
- Overall result: estimations with KOF subindices produce similar results to overall KOF index (Table 3).
- Full-sample coefficients (selected):
  - Intercepts: 2.725*** (KOF Economics), 2.747*** (KOF Actual Flows), 2.804*** (KOF Social), 2.602*** (KOF Political) with respective standard errors.
  - g_{i,t−1}: 0.239*** (Economics), 0.234*** (Actual Flows), 0.228*** (Social), 0.262*** (Political).
  - Banking_{i,t−1}: -0.526*** (Economics), -0.599*** (Actual Flows), -0.444 (Social), -0.518 (Political).
  - Currency_{i,t−1}: -0.478*** (Economics), -0.482*** (Actual Flows), -0.605*** (Social), -0.912*** (Political).
  - Stock_{i,t−1}: -2.132*** (Economics), -2.155*** (Actual Flows), -2.166*** (Social), -1.236*** (Political).
- Estimated thresholds and confidence intervals:
  - KOF Economics: γ >73 [54−76]
  - KOF Actual Flows: γ >76 [60−77]
  - KOF Social: γ >65 [49−69]
  - KOF Political: γ >56 [43−56]
- Sample (higher-globalization regimes) patterns:
  - For economic and actual flows subindices, average growth in the higher regime is higher by about 1¾ percent.
  - Banking and stock market crises have a much stronger negative impact in those subindices: an extra 2 percent decline.
  - Only about one third of countries pass the economic and actual flows thresholds by 2014; these are mainly high income countries that tend to pass in the mid-1990s or earlier.
- Interpretation: trade/globalization aspects (economic, actual flows) contribute positively and significantly to higher growth but also increase vulnerability to financial crises.

### Results for income-per-capita groups of countries
- Clustering:
  - Two groups estimated due to data limits: high-income countries (OECD and non-OECD) vs. middle- and low-income countries.
- High-income countries (Table 4):
  - Full-sample (KOF Overall) coefficients:
    - Intercept: 2.410***, se. 0.738
    - g_{i,t−1}: 0.375***, se. 0.038
    - Banking_{i,t−1}: 0.184, se. 0.642
    - Currency_{i,t−1}: -0.271, se. 0.483
    - Stock_{i,t−1}: -2.025***, se. 0.539
  - Estimated thresholds across indices: e.g., KOF Overall γ >64; KOF Economics γ >39; KOF Actual Flows γ >73; KOF Social γ >65; KOF Political γ >76.
  - In the higher-globalization regime (e.g., γ >64), banking crises negatively affect growth: Banking_{i,t−1} = -1.883***, se. 0.750 (KOF Overall sample).
  - Stock market crises are significantly negative in higher-globalization regimes (e.g., Stock_{i,t−1} = -1.226**, se. 0.667 for KOF Overall sample).
  - Linearity tests: P−value <1% across indices.
- Middle- and low-income countries (Table 5):
  - Full-sample (KOF Overall) coefficients:
    - Intercept: 0.983, se. 0.834
    - g_{i,t−1}: 0.166***, se. 0.039
    - Banking_{i,t−1}: -0.251, se. 1.067
    - Currency_{i,t−1}: 0.394, se. 0.650
    - Stock_{i,t−1}: 5.346, se. 3.412
  - Estimated thresholds: e.g., KOF Overall γ >27; KOF Economics γ >24; KOF Actual Flows γ >29; KOF Social γ >16; KOF Political γ >39.
  - In higher-globalization regimes for middle- and low-income countries:
    - Intercept examples: 1.997***, se. 0.439 (KOF Overall sample).
    - g_{i,t−1}: 0.025, se. 0.045 (KOF Overall sample); 0.128***, se. 0.049 (KOF Economics higher-regime).
    - Currency crises in higher-globalization regimes often negative and significant (e.g., Currency_{i,t} = -1.148**, se. 0.710 for KOF Overall sample).
    - Stock market crises in higher regimes strongly negative (e.g., Stock_{i,t−1} = -7.692***, se. 3.429 for KOF Overall sample).
  - Interpretation:
    - Globalization threshold estimates are much lower for middle- and low-income countries (around 27) than for high-income countries (around 64).
    - Low-income countries on average gain about 1−2 percent higher GDP growth once they reach their threshold.
    - In higher-globalization regimes for these countries, currency and stock market crises affect growth more negatively; banking systems may be insulated but markets are shallow and illiquid.
    - These countries benefit sooner from opening up but also become more vulnerable to financial crises.

### Robustness checks
- First robustness: GMM estimation with contemporaneous regressors (Table 6).
  - GMM results are consistent with GLM results (Tables 1 and 3).
  - GMM estimated thresholds lie within the block-bootstrapped confidence bounds previously estimated.
  - Full-sample GMM examples:
    - KoF Overall full-sample Intercept: 2.209***, se. 0.857
    - g_{i,t−1}: 0.402***, se. 0.026
    - Banking_{i,t}: -0.170, se. 0.241
    - Stock_{i,t}: -1.967***, se. 0.259
  - Estimated GMM thresholds: e.g., KoF Overall γ >64; KoF Economics γ >78; KoF Actual Flows γ >79; KoF Social γ >51; KoF Political γ >69.
  - In high-globalization regimes under GMM: positive constants, negative and significant autoregressive coefficients, and significant negative coefficients for banking and stock market crisis dummies for overall and economic subindices.
- Second robustness: alternative transition variable — price of brent (Table 7).
  - Model (3) estimated with brent price as transition variable; trimming Value 15%.
  - Linearity test p-value: <1% (rejects single-regime null).
  - Full-sample (brent transition) coefficients:
    - Intercept: 3.241***, se. 0.692
    - g_{i,t−1}: 0.110***, se. 0.020
    - Banking_{i,t}: -1.001***, se. 0.333
    - Currency_{i,t}: -0.875***, se. 0.299
    - Stock_{i,t}: -1.861***, se. 0.382
  - Estimated brent threshold: γ >38 (US dollars per barrel).
  - Sample (brent >38) coefficients:
    - Intercept: -1.206***, se. 0.206
    - g_{−1}: 0.208***, se. 0.031
    - Banking_{i,t}: 0.397, se. 0.467 (not significant)
    - Currency_{i,t}: 0.820***, se. 0.389 (positive and significant)
    - Stock_{i,t}: -1.172***, se. 0.498
  - Differences vs. globalization-driven regimes:
    - Banking crisis dummy not significant when brent >38 US dollars per barrel, whereas banking crises are significant under globalization transitions.
    - The constant is highly negative in the high-oil-price regime (consistent with oil-importing country effects).
    - In a high-oil-price regime, the currency crisis dummy coefficient is positive, implying currency crises can foster growth in oil-exporting contexts due to realignment effects.

*Italic: Source: wp1875 - 4. Results (IMF Working Paper PDF wp1875 - 4. Results).*

### 7. Results, Stylized Facts, and Takeaways

### 7. Results, Stylized Facts, and Takeaways

### Key empirical findings on globalization and crisis effects
- In a high-globalization regime a banking and stock market crisis have larger negative growth effects.
- Mechanisms and intuition:
  - During a banking crisis a credit crunch ensues, caused by capital losses on non-performing loans or a tightening of the regulatory environment; the more severe the banking crisis, the larger the credit crunch.
  - The effects are most likely longer lasting the more globally interconnected banks and economies are.
  - Financial crises that produce capital flight impair funding of financial institutions, which in turn cannot support international trade and growth.
  - Broken bank linkages reduce trade-finance channels (e.g., long-term interbank lending on export flows, letters of credit, documentary collections), leading to declines in net exports and output; if banks remain unconnected, only a domestic credit crunch contributes to output decline.

### Stylized correlations between crises and growth
- Table 8: Average correlation between the real GDP growth and banking crisis dummy by period
  - Periods: 1970-1994, 1995-2016
  - Sample: -0.18, -0.28
  - KOF<64: -0.18, -0.17
  - KOF>64: -0.18, -0.40
- Interpretation:
  - For overall KOF Globalization Index below the estimated threshold the correlation is almost unchanged over time.
  - For overall KOF Globalization Index above 64 the correlation is much more negative in the second part of the sample (−0.4 versus −0.18), consistent with more globalized countries experiencing larger growth losses from banking crises.

- Table 9: Average correlation between real GDP growth and banking crisis dummy by period and average overall KOF Index
  - Periods: 1970-1990, 1991-2000, 2001-2010, 2011-2016
  - All countries: -0.14, -0.28, -0.43, -0.40
  - Average KOF∈(25,40]: -0.16, 0.01, N/A, N/A
  - Average KOF∈(40,64]: -0.25, -0.27, -0.27, N/A
  - Average KOF∈(64,80]: -0.12, -0.63, -0.38, -0.76
  - Average KOF>80: N/A, -0.38, -0.54, -0.37
- Interpretation:
  - The highest negative correlation for the overall sample occurs during the decade including the global financial crisis (−0.43).
  - Lower-globalization countries (KOF between 25 and 40) do not experience banking crises after 2000; their lower average growth is recorded in earlier decades (e.g., the 1990s).
  - Moderately globalized countries (KOF between 40 and 64) show similar correlations across the first three periods and do not have banking crises during 2011-2014.
  - The most globalized countries show the lowest average growth during 2001-2010, but absence of a recent banking crisis implies no automatic “new normal” for all highly globalized countries.

### Average real GDP growth by period and globalization
- Table 10: Average real GDP growth by period and average overall KOF Globalization Index
  - Columns: sample (1970-2016), 1970-1990, 1991-2000, 2001-2010, 2011-2016
  - All countries: 3.47, 6.32, 1.76, 2.65, 3.64
  - Average KOF∈(25,40]: 3.55, 5.94, 2.44, 5.48, 2.72
  - Average KOF∈(40,64]: 3.80, 6.88, 3.44, 2.62, 4.32
  - Average KOF∈(64,80]: 3.06, 5.85, -2.37, 1.01, 4.75
  - Average KOF>80: 2.24, N/A, 1.36, 2.13, 1.29
- Interpretation:
  - For 2011-2016, average real output growth for all groups is similar to, or higher than, in the 1990s.
  - Because the KOF Globalization Index tends to increase over time for all countries, the data support the conclusion that countries are, overall, better off being more open and integrated.
  - Globalization increases growth on average, but above a threshold it also increases vulnerability to larger negative short-term growth impacts from financial crises.

### Takeaways and broader conclusions
- The effect of globalization on growth is a “tale of two narratives”:
  - Globalization (especially economic globalization) raises the growth rate of output significantly over the sample for all country groups.
  - In high-globalization regimes, high-income countries experience negative growth effects from banking crises and stronger negative effects from stock market crises.
  - In low-income countries, currency and stock market crises have negative effects on growth; benefits and vulnerabilities can accrue earlier in the globalization process.
- The paper documents a “new normal”: a persistently (but not permanently) lower real GDP growth regime following a financial crisis for countries in a high-globalization regime. Eventually countries return to higher growth in the higher-globalization equilibrium.
- Results are robust to various specifications and are not driven solely by the tail end of the sample (e.g., the global financial crisis of 2007−2008).

### Policy implications and recommendations
- Sound macroeconomic and financial policies are essential to reap globalization’s benefits while mitigating vulnerabilities:
  - Appropriate regulation and supervision of banks and capital markets.
  - Good international financial policy coordination.
  - Avoid synchronizing financial deregulation with increases in globalization, as this would undermine globalization’s positive growth impact.
- Specific measures discussed:
  - Consider saving benefits of higher growth (e.g., international reserves or fiscal surpluses) into a financial stabilization fund that could be used to recapitalize banks or financial institutions facing a systemic crisis (practical implementation to be analyzed in future research).
- Consistency with other assessments:
  - Findings align with the 2017 BIS Annual Report and literature (Cecchetti (2012), Shin (2017), Broner and Ventura (2016)) that globalization is beneficial up to a point unless mitigating policies are in place.
  - A globalization reversal would not be optimal given remaining gains from deeper integration for many countries.

*Source: 7. Results, Stylized Facts, and Takeaways — wp1875*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp1875.pdf_