## wp17231 - 2015.  The downside is that it only reports trade in goods.

## Source details

**Canonical URL:** [wp17231 - 2015.  The downside is that it only reports trade in goods.](https://www.imf.org/-/media/files/publications/wp/2017/wp17231.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17231.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17231.pdf.json)

---

### Data sources and measures
- Population, land area, and real income per capita: World Bank’s World Development Indicators (WDI) database.
- Bilateral distance, distance from the equator, dummy variables for common border and being landlocked: CEPII’s GeoDist database; bilateral distance is the population-weighted distance between major cities in the two countries.
- Top-to-bottom income per capita ratio: World Panel Income Distribution database (Lakner and Milanovic, 2015); vintages allocated to closest five-year intervals between 1988 and 2008; used to obtain cross-section estimates for 1993, 1998, 2003, 2008.
- Gini coefficients: Standardized World Income Inequality Database (SWIID, Solt, 2016).
- Rule of Law index: Worldwide Governance Indicators (WGI).
- Average temperature (1961-1999): World Bank Climate Change Knowledge Portal.
- Settler mortality rates: Acemoglu, Johnson and Robinson (2001) data from Daron Acemoglu’s website.
- For openness in goods and services (when considered using goods-only instrument based on DOTS data): balance of payments data from the World Economic Outlook database.

### Construction and quality of the geography-based instrument
- Bilateral trade regressions estimated with PPML; key coefficients reported for cross sections Year = 1995, 2005, 2015.
- Distance: Ln distance coefficients of -0.96, -0.97, -1.01 (standard errors (0.05), (0.03), (0.04)).
- Population and area effects:
  - Ln population (country i): -0.18, -0.11, -0.12 (standard errors (0.04), (0.03), (0.03)).
  - Ln area (country i): -0.01, -0.02, -0.03 (standard errors (0.03), (0.03), (0.02)).
  - Ln population (country j): 0.84, 0.78, 0.84 (standard errors (0.04), (0.03), (0.03)).
  - Ln area (country j): -0.23, -0.13, -0.13 (standard errors (0.05), (0.04), (0.03)).
- Landlocked: -0.70, -0.82, -0.76 (standard errors (0.08), (0.06), (0.06)).
- Common border: 1.26, 3.79, 3.99 (standard errors (3.47), (2.03), (1.61)).
- Interaction terms (examples):
  - x Ln distance: -0.09, 0.39, 0.40 (standard errors (0.53), (0.31), (0.24)).
  - x Landlocked: 0.59, 0.63, 0.59 (standard errors (0.21), (0.12), (0.13)).
- Sample sizes: 26,565; 28,224; 28,056.
- R2 values: 0.14, 0.29, 0.34.
- Instrument relevance:
  - Geography-induced openness (aggregated across partners) has a point estimate between 0.5 and 0.8 in the first-stage regression.
  - Rule-of-thumb F-statistic condition (F > 10) is satisfied for every cross section (with one instrument t2 = F).

### Baseline estimates — Trade’s effect on real income per capita
- IV estimates: a one percentage point increase in trade openness raises real income per capita by between 2 and 5 percent (point estimates).
- Time variation:
  - Coefficients hovered between 3-5 percent since early 1990s, fell to about 2 percent after the global financial crisis.
- Statistical significance: estimates are overwhelmingly significant for all time periods reported.
- Note on openness measure: trade openness is entered in decimal form (e.g., 20% openness is 0.2).

### Baseline estimates — Trade’s effect on inequality
- Top-to-bottom income ratio:
  - Point estimates suggest one-percentage-point higher openness causes the income of the top decile to decrease by about 5 percent relative to the bottom decile.
  - Effect is statistically significant for earlier vintages; insignificant for the latest available cross-section.
- Market and net Gini coefficients:
  - A one percentage point higher openness is associated with a 0.2-0.6 points lower net Gini coefficient.
  - Many estimates, especially for the net Gini, are statistically significant for a number of years.
- Overall pattern:
  - Almost all point estimates suggest an inequality-reducing effect of trade; some insignificant and a few positive estimates appear in more recent periods coinciding with a dramatic drop in sample size.
  - Inequality indicators tend to move slowly over time, so estimated impacts are relatively large in that context.

### OLS versus IV comparisons and interpretive remarks
- IV yields markedly stronger effects of trade than OLS for both income and inequality regressions.
- For income:
  - OLS coefficient for trade is much smaller than IV coefficient.
  - The authors find this persistent gap across over 20 vintages and argue sampling error is unlikely to explain the difference; IV estimates are taken as meaningful.
- For inequality:
  - OLS coefficients are almost always insignificant and close to zero.
  - IV using geography-induced trade shows a negative effect on aggregate inequality with many significant coefficients.
- Cautionary notes:
  - Gini data (SWIID) rely extensively on imputation across and within countries, particularly prevalent for less developed regions — interpret results with care.
  - Differences in sample composition over time affect estimates; changes in estimated coefficients are not always statistically significant.

### Robustness checks — Including services trade
- World services imports accounted for "20.4% of total world imports of goods and services in the year 2000" and "22.7% by 2015."
- First-stage diagnostics when including services:
  - "The point estimate of the first-stage regression coefficient ranges between about 0.3 and 0.8."
  - "With the exception of the cross-sections for 1994-1995, the instrument remains remarkably strong."
- Estimated effects on real income per capita including services:
  - Estimates are "well within the confidence intervals of our previous estimates" and "very similar in magnitude" except where the instrument is weak (1994 and 1995).
  - The positive effect of a one-percent increase in openness on real income per capita falls "from about 4% in the years before the crisis, to about 2.4% since."
- Figures reproduced in the source:
  - Figure 5: first-stage coefficient and t-statistic (goods-only instrument when including services).
  - Figure 6a: IV estimates of goods-and-services openness on log real GDP per capita; only estimates significant at the 20% level are shown (confidence bands at 95%).

### Robustness checks — Direct and indirect geography channels and identification
- Conceptual diagnostics:
  - The FR trade instrument (predicted openness) is "very weakly correlated with distance from the equator and temperature."
  - Distance from the equator, temperature and settler mortality each explain roughly six times better the cross-country variation of institutional quality than the trade instrument (R-square of 30% vs 5%).
- Controls and instruments used to address omitted geography channels:
  - Direct geography controlled via annual average temperature.
  - Indirect geography through institutions controlled via the Rule of Law index.
  - Institution endogenous treatments instrumented with log settler mortality in some specifications.
- Weak-instrument and identification diagnostics reported:
  - Kleibergen and Paap rk statistic (underidentification).
  - Cragg and Donald (1993) Wald statistic with Stock and Yogo (2005) critical values for weak-instrument size distortions.

### Empirical robustness — Selected table highlights (preserve reported estimates exactly)
- Table 2 (Trade’s impact on income: Robustness to climate and institutions) — example IV openness coefficients and standard errors:
  - Openness = 3.955 ∗∗∗ (0.967)
  - Openness = 4.059 ∗∗ (1.385)
  - Openness = 2.228 ∗∗∗ (0.541)
  - Temperature example: Temperature = -0.0434 ∗∗ (0.0151)
  - Rule of Law example: 0.815 ∗∗∗ (0.0913)
  - Weak ID test stat.: 26.20, 10.52, 38.72 (selected values).
  - Stock–Yogo critical values shown as "Ho: t-test size > 10pct" = 16.38 and "Ho: t-test size > 25pct" = 5.530 in many columns.
  - When institutions are instrumented with log settler mortality, "the estimated positive effect of trade on income falls dramatically and becomes statistically indistinguishable from zero" in that smaller sample.
- Table 3 (Robustness of income regression to institutions: Effect of sample selection):
  - Baseline openness repeated: 3.955 ∗∗∗ (0.967), 4.059 ∗∗ (1.385), 2.228 ∗∗∗ (0.541).
  - In reduced sample required to instrument institutions with LOGEM4, openness estimates become negative or insignificant (example: Openness = -0.734 (1.105)).
  - Using DISTEQ as instrument for Rule of Law, openness is significant on the full sample (example: Openness = 3.029 ∗∗ (0.998)) but insignificant in the reduced sample.
  - Conclusion: "the disappearing effect of trade may have to do more with the specific sample rather than with the overall primacy of institutions."
- Table 4 (Trade’s impact on top-to-bottom income ratio: Robustness to climate and institutions):
  - Openness examples: Openness = -6.036 ∗ (2.949); Openness = -4.349 ∗∗ (1.536).
  - Weak instrument diagnostics often poor; instrumenting institutions with settler mortality leads to loss of significance and instrument strength collapse in the smaller sample.
  - Interpretation: "there is no evidence that more trade openness causes a greater gap between the incomes of the top and bottom deciles."
- Tables 5 and 6 (Market Gini and Net Gini robustness) — openness estimates examples:
  - Market Gini: Openness = -34.79 ∗ (14.32), Openness = -10.99 ∗ (4.988).
  - Net Gini: Openness = -69.54 ∗∗ (26.40), Openness = -33.24 ∗∗∗ (9.421), Openness = -29.94 ∗∗ (11.28).
  - Temperature example for net Gini: Temperature = 0.400 ∗∗∗ (0.113).
  - Across inequality specifications, "the point estimates on trade openness are always negative" and "in many cases they seem significant, especially for the net Gini coefficient," but weak-instrument test statistics counsel caution.

### Additional robustness and heterogeneity results
- Non-European sample (Figure 9):
  - IV estimates plotted for non-European countries after including a Europe dummy and interactions; "plotted coefficients refer to the non-European countries."
  - Estimates for European countries are "generally highly insignificant."
- Diagnostics and correlations (Figure notes based on 2013 data):
  - Low correlation between predicted openness and distance from equator / average temperature; example R-squared values reported in notes: 0.06 for POPEN and 0.76 for TEMP in one figure note.
  - Other R-squared values reported across panels include 0.19, 0.05, 0.01, 0.33, 0.00, 0.27, 0.02, and 0.31 for different pairings of variables.

### Summary conclusions from robustness checks
- Controlling for direct geography (temperature) and the Rule of Law typically leaves baseline IV estimates of trade’s positive impact on income "virtually unchanged" in the full samples.
- Instrumenting institutions with settler mortality substantially alters point estimates, but this coincides with a marked reduction in sample size; evidence suggests sample selection rather than a definitive dominance of institutions over trade.
- Across specifications and inequality measures:
  - There is no robust evidence that greater trade openness increases overall income inequality.
  - Point estimates are generally negative for inequality, but weak instrument diagnostics limit the strength of inference.
- Policy implication reiterated by the authors: "well-designed policies can leverage trade integration to support higher and inclusive growth," subject to caveats about the difference between geographic and policy-induced trade barriers, long-run interpretation of cross-country estimates, and the lack of reduced-form identification of channels.

### Key numerical findings and statistics (preserved exactly)
- Bilateral Ln distance elasticity: -0.96, -0.97, -1.01.
- Ln population (country j) elasticities: 0.84, 0.78, 0.84.
- Landlocked coefficients: -0.70, -0.82, -0.76.
- Common border coefficients: 1.26, 3.79, 3.99.
- Interaction x Landlocked: 0.59, 0.63, 0.59.
- Sample sizes: 26,565; 28,224; 28,056.
- R2: 0.14, 0.29, 0.34.
- First-stage geography-induced openness point estimate range: between 0.5 and 0.8.
- Estimated effect of a one percentage point increase in openness on real income per capita: between 2 and 5 percent.
- Estimated effect of a one percentage point increase in openness on net Gini: 0.2-0.6 points lower.
- Estimated impact on top-to-bottom income ratio: about 5 percent decrease in top decile income relative to bottom decile for a one-percentage-point higher openness.

*Source — wp17231 (2015), excerpt provided.*

### 2015.  The downside is that it only reports trade in goods.

### wp17231 - 2015.  The downside is that it only reports trade in goods.

### Data sources and measures
- Population, land area, and real income per capita: World Bank’s World Development Indicators (WDI) database.
- Bilateral distance, distance from the equator, dummy variables for common border and being landlocked: CEPII’s GeoDist database; bilateral distance is the population-weighted distance between major cities in the two countries.
- Top-to-bottom income per capita ratio: World Panel Income Distribution database (Lakner and Milanovic, 2015); vintages allocated to closest five-year intervals between 1988 and 2008; used to obtain cross-section estimates for 1993, 1998, 2003, 2008.
- Gini coefficients: Standardized World Income Inequality Database (SWIID, Solt, 2016).
- Rule of Law index: Worldwide Governance Indicators (WGI).
- Average temperature (1961-1999): World Bank Climate Change Knowledge Portal.
- Settler mortality rates: Acemoglu, Johnson and Robinson (2001) data from Daron Acemoglu’s website.
- For openness in goods and services (when considered using goods-only instrument based on DOTS data): balance of payments data from the World Economic Outlook database.

### Construction and quality of the geography-based instrument
- Bilateral trade regressions estimated with PPML; key coefficients reported for cross sections Year = 1995, 2005, 2015.
- Distance: Ln distance coefficients of -0.96, -0.97, -1.01 (standard errors (0.05), (0.03), (0.04)).
- Population and area effects:
  - Ln population (country i): -0.18, -0.11, -0.12 (standard errors (0.04), (0.03), (0.03)).
  - Ln area (country i): -0.01, -0.02, -0.03 (standard errors (0.03), (0.03), (0.02)).
  - Ln population (country j): 0.84, 0.78, 0.84 (standard errors (0.04), (0.03), (0.03)).
  - Ln area (country j): -0.23, -0.13, -0.13 (standard errors (0.05), (0.04), (0.03)).
- Landlocked: -0.70, -0.82, -0.76 (standard errors (0.08), (0.06), (0.06)).
- Common border: 1.26, 3.79, 3.99 (standard errors (3.47), (2.03), (1.61)).
- Interaction terms (examples):
  - x Ln distance: -0.09, 0.39, 0.40 (standard errors (0.53), (0.31), (0.24)).
  - x Landlocked: 0.59, 0.63, 0.59 (standard errors (0.21), (0.12), (0.13)).
- Sample sizes: 26,565; 28,224; 28,056.
- R2 values: 0.14, 0.29, 0.34.
- Instrument relevance:
  - Geography-induced openness (aggregated across partners) has a point estimate between 0.5 and 0.8 in the first-stage regression.
  - Rule-of-thumb F-statistic condition (F > 10) is satisfied for every cross section (with one instrument t2 = F).

### Baseline estimates — Trade’s effect on real income per capita
- IV estimates: a one percentage point increase in trade openness raises real income per capita by between 2 and 5 percent (point estimates).
- Time variation: coefficients hovered between 3-5 percent since early 1990s, fell to about 2 percent after the global financial crisis.
- Statistical significance: estimates are overwhelmingly significant for all time periods reported.
- Note on openness measure: trade openness is entered in decimal form (e.g., 20% openness is 0.2).

### Baseline estimates — Trade’s effect on inequality
- Top-to-bottom income ratio:
  - Point estimates suggest one-percentage-point higher openness causes the income of the top decile to decrease by about 5 percent relative to the bottom decile.
  - Effect is statistically significant for earlier vintages; insignificant for the latest available cross-section.
- Market and net Gini coefficients:
  - A one percentage point higher openness is associated with a 0.2-0.6 points lower net Gini coefficient.
  - Many estimates, especially for the net Gini, are statistically significant for a number of years.
- Overall pattern: almost all point estimates suggest an inequality-reducing effect of trade; some insignificant and a few positive estimates appear in more recent periods coinciding with a dramatic drop in sample size.
- Magnitude note: inequality indicators tend to move slowly over time, so estimated impacts are relatively large in that context.

### OLS versus IV comparisons and interpretive remarks
- IV yields markedly stronger effects of trade than OLS for both income and inequality regressions.
- For income: OLS coefficient for trade is much smaller than IV coefficient. The authors find this persistent gap across over 20 vintages, arguing that sampling error is unlikely to explain the difference and that IV estimates are meaningful.
- For inequality: OLS coefficients are almost always insignificant and close to zero; IV using geography-induced trade shows a negative effect on aggregate inequality with many significant coefficients.
- Cautionary notes:
  - Gini data (SWIID) rely extensively on imputation across and within countries, particularly prevalent for less developed regions — interpret results with care.
  - Differences in sample composition over time affect estimates; changes in estimated coefficients are not always statistically significant.

### Key numerical findings and statistics (preserved exactly)
- Bilateral Ln distance elasticity: -0.96, -0.97, -1.01.
- Ln population (country j) elasticities: 0.84, 0.78, 0.84.
- Landlocked coefficients: -0.70, -0.82, -0.76.
- Common border coefficients: 1.26, 3.79, 3.99.
- Interaction x Landlocked: 0.59, 0.63, 0.59.
- Sample sizes: 26,565; 28,224; 28,056.
- R2: 0.14, 0.29, 0.34.
- First-stage geography-induced openness point estimate range: between 0.5 and 0.8.
- Estimated effect of a one percentage point increase in openness on real income per capita: between 2 and 5 percent.
- Estimated effect of a one percentage point increase in openness on net Gini: 0.2-0.6 points lower.
- Estimated impact on top-to-bottom income ratio: about 5 percent decrease in top decile income relative to bottom decile for a one-percentage-point higher openness.

*Italic: Source — wp17231 (2015), excerpt provided.*

### 3.3    Robustness of the Baseline Results

### 3.3    Robustness of the Baseline Results

### Including services trade
- World services imports accounted for "20.4% of total world imports of goods and services in the year 2000" and "22.7% by 2015."
- First-stage diagnostics when including services:
  - "The point estimate of the first-stage regression coefficient ranges between about 0.3 and 0.8."
  - "With the exception of the cross-sections for 1994-1995, the instrument remains remarkably strong."
- Estimated effects on real income per capita including services:
  - Estimates are "well within the confidence intervals of our previous estimates" and "very similar in magnitude" except where the instrument is weak (1994 and 1995).
  - The positive effect of a one-percent increase in openness on real income per capita falls "from about 4% in the years before the crisis, to about 2.4% since."
- Figures reproduced:
  - Figure 5: first-stage coefficient and t-statistic (goods-only instrument when including services).
  - Figure 6a: IV estimates of goods-and-services openness on log real GDP per capita; only estimates significant at the 20% level are shown (confidence bands at 95%).

### Direct and indirect effects of geography through omitted variables
- Conceptual distinctions and correlations:
  - The FR trade instrument is a weighted average of bilateral distances to trading partners (distance weighted by partner size) and is "very weakly correlated with distance from the equator and temperature."
  - Instrument vs institutions:
    - "Only our geography-based instrument (predicted openness) explains a meaningful share of cross-country variation in trade integration."
    - "Distance from the equator, temperature and settler mortality each explain roughly six times better the cross-country variation of institutional quality than our instrument for trade (R-square of 30% vs 5%)."
- Identification strategy to address omitted geography channels:
  - Control for direct geography effect using "annual average temperature."
  - Control for indirect geography effect through institutions using the "Rule of Law" index.
  - Diagnostic tests reported: Kleibergen and Paap rk statistic (underidentification) and Cragg and Donald (1993) Wald statistic with Stock and Yogo (2005) critical values for weak-instrument size distortions (nulls that the actual 5% t-test size is greater than 10 or 25 percent).

### Empirical robustness: income regressions and instrumentation (summary of key table results)
- Table 2 (Trade’s impact on income: Robustness to climate and institutions)
  - Columns (1)-(3) (Year 2003, 2008, 2013 baseline IV openness coefficients and standard errors):
    - Openness = 3.955 ∗∗∗ (0.967)
    - Openness = 4.059 ∗∗ (1.385)
    - Openness = 2.228 ∗∗∗ (0.541)
  - Temperature enters negative where reported: e.g., Temperature = -0.0434 ∗∗ (0.0151) in one specification.
  - Rule of law coefficients when included: e.g., 0.815 ∗∗∗ (0.0913).
  - Weak identification / instrument diagnostics (selected values):
    - Weak ID test stat.: 26.20, 10.52, 38.72 (for selected columns).
    - Reported Stock–Yogo critical values shown as "Ho: t-test size > 10pct" = 16.38 and "Ho: t-test size > 25pct" = 5.530 in many columns (table notes).
  - When institutions are instrumented with log settler mortality (columns with two endogenous variables), "the estimated positive effect of trade on income falls dramatically and becomes statistically indistinguishable from zero" in that smaller sample.
- Table 3 (Robustness of income regression to institutions: Effect of sample selection)
  - Reproduces baseline openness coefficients and shows:
    - Baseline openness again: 3.955 ∗∗∗ (0.967), 4.059 ∗∗ (1.385), 2.228 ∗∗∗ (0.541).
    - In the reduced sample required to instrument institutions with LOGEM4, openness estimates become negative or insignificant (e.g., Openness = -0.734 (1.105) in a reduced-sample column).
    - Using DISTEQ as instrument for Rule of Law, openness is significant on the full sample (e.g., Openness = 3.029 ∗∗ (0.998)) but insignificant in the reduced sample.
  - Conclusion drawn: "the disappearing effect of trade may have to do more with the specific sample rather than with the overall primacy of institutions."
- Table 4 (Trade’s impact on top-to-bottom income ratio: Robustness to climate and institutions)
  - Openness coefficients across columns are negative in all reported specifications; examples:
    - Openness = -6.036 ∗ (2.949) in one column; Openness = -4.349 ∗∗ (1.536) in another.
  - Weak instrument diagnostics often poor: many weak ID test stats are low and in many cases "we cannot reject the null that the actual size of the 5% t-test is above 25%" and "we can never reject that it is above 10%."
  - Instrumenting institutions with settler mortality leads to loss of significance and instrument strength collapse in the smaller sample.
  - Interpretation: "there is no evidence that more trade openness causes a greater gap between the incomes of the top and bottom deciles."
- Tables 5 and 6 (Market Gini and Net Gini robustness)
  - Openness coefficients on market Gini often negative; examples:
    - Openness = -34.79 ∗ (14.32), Openness = -10.99 ∗ (4.988).
  - Openness coefficients on net Gini show larger negative magnitudes in several columns; examples:
    - Openness = -69.54 ∗∗ (26.40), Openness = -33.24 ∗∗∗ (9.421), Openness = -29.94 ∗∗ (11.28).
  - Temperature sometimes positive and significant for net Gini (e.g., Temperature = 0.400 ∗∗∗ (0.113) in one column).
  - Across inequality specifications, "the point estimates on trade openness are always negative" and "in many cases they seem significant, especially for the net Gini coefficient," but low weak-instrument test statistics counsel caution about inference.
  - Overall statement: "there is certainly no indication in any of these robustness checks that trade could have a detrimental impact on overall income inequality."

### Additional robustness checks and heterogeneity
- Non-European sample (Figure 9):
  - IV estimates plotted for non-European countries after including a Europe dummy and interactions; "plotted coefficients refer to the non-European countries." Estimates for European countries are "generally highly insignificant."
- Figure 7 and Figure 8 diagnostics (based on 2013 data):
  - Figure 7: low correlation between predicted openness and distance from equator / average temperature (R-squared reported as 0.06 for POPEN and 0.76 for TEMP in the figure note).
  - Figure 8 panels show R-squared values reported in the notes: e.g., "R-squared= 0.19" for trade openness vs geography (one panel) and "R-squared= 0.05" for rule of law vs geography (another panel); other panels list R-squared values including 0.01, 0.33, 0.00, 0.27, 0.02, and 0.31 for different pairings of variables shown.

### Summary conclusion from robustness checks
- Controlling for direct geography (temperature) and the Rule of Law typically leaves baseline IV estimates of trade’s positive impact on income "virtually unchanged" in the full samples.
- Instrumenting institutions with settler mortality substantially alters point estimates, but this coincides with a marked reduction in sample size; evidence suggests sample selection rather than a definitive dominance of institutions over trade.
- Across specifications and inequality measures, there is no robust evidence that greater trade openness increases overall income inequality; point estimates are generally negative for inequality, but weak instrument diagnostics limit the strength of inference.
- Policy implication reiterated: "well-designed policies can leverage trade integration to support higher and inclusive growth," subject to caveats about the difference between geographic and policy-induced trade barriers, long-run interpretation of cross-country estimates, and the lack of reduced-form identification of channels.

*Source: IMF Working Paper — section 3.3, "Robustness of the Baseline Results."*

### References

### References

### Econometrics and Identification
- Cragg, John G., and Stephen G. Donald.1993. “Testing Identifiability and Specification in Instrumental Variable Models.”Econometric Theory, 9(2): 222240.
- Kleibergen, Frank, and Richard Paap.2006. “Generalized reduced rank tests using the singular value decomposition.”Journal of Econometrics, 133(1): 97–126.
- Santos-Silva, J. M. C., and Silvana Tenreyro.2006. “The Log of Gravity.”The Review of Economics and Statistics, 88(4): 641–658.
- Stock, James, and Motohiro Yogo.2005. “Testing for Weak Instruments in Linear IV Regression.”Identification and Inference for Econometric Models, , ed. Donald W.K. Andrews, 80–108. New York:Cambridge University Press.

### Trade, Gravity, and Trade-Related Empirics
- Autor, David H., David Dorn, and Gordon H. Hanson.2013. “The China Syndrome: Local Labor Market Effects of Import Competition in the United States.”American Economic Review, 103(6): 2121–68.
- Disdier, Anne-C ́elia, and Keith Head.2008. “The Puzzling Persistence of the Distance Effect on Bilateral Trade.”The Review of Economics and Statistics, 90(1): 37–48.
- Feyrer, James.2009a. “Distance, Trade, and Income - The 1967 to 1975 Closing of the Suez Canal as a Natural Experiment.” National Bureau of Economic Research, Inc NBER Working Papers 15557.
- Feyrer, James.2009b. “Trade and Income – Exploiting Time Series in Geography.” National Bureau of Economic Research, Inc NBER Working Papers 14910.
- Frankel, Jeffrey A., and David Romer.1999. “Does Trade Cause Growth?”American Economic Review, 89(3): 379–399.
- Head, Keith, and Thierry Mayer.2014. “Gravity Equations: Workhorse, Toolkit, and Cookbook.” InThe Handbook of International Economics. , ed. Gita Gopinath, Elhanan Helpman and Kenneth Rogoff, Chapter 3, 131–195. Elsevier.
- Noguer, Marta, and Marc Siscart.2005. “Trade raises income: a precise and robust result.”Journal of International Economics, 65(2): 447 – 460.
- Santos-Silva, J. M. C., and Silvana Tenreyro.2006. “The Log of Gravity.”The Review of Economics and Statistics, 88(4): 641–658.

### Trade, Inequality, and Distributional Effects
- Fajgelbaum, Pablo D., and Amit K. Khandelwal.2016. “Measuring the Unequal Gains from Trade.”The Quarterly Journal of Economics, 131(3): 1113.
- Goldberg, Pinelopi K.2015.Trade and Inequality.Edward Elgar.
- Helpman, Elhanan.2016. “Globalization and Wage Inequality.” National Bureau of Economic Research Working Paper 22944.
- Autor, David H., David Dorn, and Gordon H. Hanson.2013. “The China Syndrome: Local Labor Market Effects of Import Competition in the United States.”American Economic Review, 103(6): 2121–68.

### Institutions, Geography, and Development
- Acemoglu, Daron, Simon Johnson, and James A. Robinson.2001. “The Colonial Origins of Comparative Development: An Empirical Investigation.”American Economic Review, 91(5): 1369–1401.
- Dollar, David, and Aart Kraay.2003. “Institutions, trade, and growth.”Journal of Monetary Economics, 50(1): 133 – 162.
- Engerman, Stanley L, and Kenneth L. Sokoloff.2002. “Factor Endowments, Inequality, and Paths of Development Among New World Economics.” National Bureau of Economic Research Working Paper 9259.
- Hall, Robert E., and Charles I. Jones.1999. “Why do Some Countries Produce So Much More Output Per Worker than Others?”The Quarterly Journal of Economics, 114(1): 83–116.
- Hallak, Juan Carlos, and James Levinsohn.2004. “Fooling Ourselves: Evaluating the Globalization and Growth Debate.” National Bureau of Economic Research, Inc NBER Working Papers 10244.
- Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi.2004. “Institutions Rule: The Primacy of Institutions Over Geography and Integration in Economic Development.”Journal of Economic Growth, 9(2): 131–165.
- Rodriguez, Francisco, and Dani Rodrik.2001. “Trade Policy and Economic Growth: A Skeptic’s Guide to the Cross-National Evidence.” InNBER Macroeconomics Annual 2000, Volume 15. , ed. Ben S. Bernanke and Kenneth Rogoff, 261–338. MIT Press.

### Income Distribution, Inequality, and Databases
- Jenkins, S.2015. “World income inequality databases: an assessment of WIID and SWIID.”The Journal of Economic Inequality, 13(4): 629–671.
- Lakner, Christoph, and Branko Milanovic.2015. “Global Income Distribution: From the Fall of the Berlin Wall to the Great Recession.”The World Bank Economic Review, 1–30.
- Solt, Frederick.2016. “The Standardized World Income Inequality Database.”Social Science Quarterly, 97(5): 1267–1281.

### Climate, Scale, and Regional Analysis
- Masters, William A., and Margaret S. McMillan.2001. “Climate and Scale in Economic Growth.”Journal of Economic Growth, 6(3): 167–186.
- Bernanke, Ben S.2009. “Asia and the Global Financial Crisis.”Speech given at the Federal Reserve Bank of San Francisco’s Conference on Asia and the Global Financial Crisis, Santa Barbara, California.
- IMF.2017. “Cluster Report: Trade Integration in Latin America and the Caribbean.”International Monetary Fund IMF Staff Country Reports 17/66.

*References (as listed in the source PDF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17231.pdf_