Openness Inequality and Poverty Endowments Matter Julien Gourdon

  • Slides: 31
Download presentation
Openness, Inequality, and Poverty: Endowments Matter Julien Gourdon (World Bank - CERDI) Nicolas Maystre

Openness, Inequality, and Poverty: Endowments Matter Julien Gourdon (World Bank - CERDI) Nicolas Maystre (University of Geneva) Jaime de Melo (CEPR, University of Geneva)

Introduction - motivation • We focus on the impact of trade liberalization on inequality

Introduction - motivation • We focus on the impact of trade liberalization on inequality within countries • Why study inequality? • Controversial subject (see next 2 slides) 1) Mixed empirical evidence (see summary of evidence on inequality and income) 2) Very little support for HOS (see summary of studies)

View 1: Growth is good for the poor (Dollar and Kraay (2001)) Figure =

View 1: Growth is good for the poor (Dollar and Kraay (2001)) Figure = scatter from 129 growth episodes of 6. 8 years average: 31 from East Asia; 50 from LAC; 15 from MENA; 15 from South Asia; 15 from SSA

View 2: One must look beyond averages = Scope for policy to affect poverty

View 2: One must look beyond averages = Scope for policy to affect poverty (Bourguignon, Ravallion) Fit a log-normal distribution to the data (allows to decompose change in poverty into two components: (i) growth effect; (ii) change in distribution First slide: Mexico-type (G=0. 55) country: Assumption is gy =3% p. a H by 7 percentage points in 10 years (but much better if G back to 0. 45, the level of the mid 80 s. Bottom: low-income (G=0. 4). Same assumption: gy =3% p. a. H by 15 percentage points in 10 years! (But if G 0. 45, then reduction in H cut in half) Conclusion: Inequality matters for poverty. Does policy matter for inequality, and what about trade policy?

Econom etric Definition of openness inequality index Results: effect of openness on inequality GMM

Econom etric Definition of openness inequality index Results: effect of openness on inequality GMM Trade/GDP with both expressed in nominal dollar terms decile mean income/overall mean income (from WYD) greater inequality under 8000 $PPP, smaller inequality over 8000 $PPP OLS Sachs-Warner measure Income share of bottom quintile or Gini (from DS) Mildly pro-inequality Unbalanced panel; 5 -year intervals; 285 obs; 92 countries Fixed Effect Trade (in 1985 dollars) / GDP (in 1985 prices but measured in international dollars)* Income share of bottom quintile (mostly WIDER) Insignificant 19601995 Unbalanced panel; 5 -year intervals; 284 obs; 102 countries GMM Trade/GDP Gini from DS Ravallion (2001) 19601994 Unbalanced panel; 5 -year intervals; 159 obs. OLS Export/GDP Gini from DS Pro-inequality in poor countries Barro (2000) 19601990 Balanced panel; 10 -year intervals; 214 obs Gini from DS Pro-inequality in poor countries in level Insignificant in Fixed Effect Higgins and Williamson (1999) 19601990 Unalanced panel; 10 -year intervals; 219 obs, 85 countries Trade/GDP Gini from DS Insignificant in FE for poor & rich countries Pro-equality in level for poor & rich countries Savvides (1998) 19781994 Balanced panel; 7 -year intervals; 34 countries i) Average tariffs ii) Sachs Warner Gini from DS Authors Period Sample, Milanovic (2005) 19881998 Unbalanced panel; 5 -year intervals; 322 obs; 129 countries Lundberg and Squire (2003) 19601998 Unbalanced panel; 5 -year intervals; 119 obs; 38 countries Dollar and Kraay (2002) 19601999 Calderon and Chong (2001) OLS Trade/GDP adjusted for Fixed country size Effect OLS Fixed Effect Changes Insignificant in rich countries Pro-equality in poor countries Insignificant in rich countries Pro-Inequality in poor countries

 Unbalanced panel, 5 years intervals GMM Trade/GDP Gini from DS Pro-inequality in mining

Unbalanced panel, 5 years intervals GMM Trade/GDP Gini from DS Pro-inequality in mining rich countries Pro-equality in agricultural rich countries Pro -inequality in skilled labor rich countries Insignificant in capital rich countries Bensidoun and al. (2005) 19702000 Unbalanced panel, 4 years intervals 140 obs, 41 countries Change Net exports Gini from DS Pro-inequality in noed-labor rich countries Fisher (2001) 19651990 Unbalanced panel; 5 -year intervals; 66 countries Fixed Effect Sachs-Warner measure Gini from DS Pro-inequality in skilled-labor rich countries Pro-equality in capital rich countries OLS Own index of openness adjusted for endowments, economic size and geographic distance to eventual partners Gini from DS Pro-inequality in skilled-labor rich countries Pro-equality in capital rich countries Pro -equality in land rich countries OLS Average tariffs in manufacture, net export in agriculture, net export in mining products Gini and share of the poorest 20% of the population Pro equality concerning decrease in manufact. tariffs pro inequality for country specialized in mining prod. Perry and Olarreaga (2006) Spilimbergo, Londono and Szekely (1999) Bourguignon and Morrisson (1990) 19651992 Unbalanced panel; 320 obs; 34 countries 1970's Cross country; 35 developing countries

What’s new (besides more & better data—next slide) • Analyze the changes within countries

What’s new (besides more & better data—next slide) • Analyze the changes within countries rather than using the between countries dimension. • Analyze according to factor endowments (HOS) (finer disaggregation: 3 types of workers + split of natural resources) • Use of a policy-based variable for “openness” (i. e. tariffs = import duties / total imports) rather than outcomebased variables (Rodrik 2000)

Data • 2 different data sets for inequality: - Deininger & Squire (Gini coefficient)

Data • 2 different data sets for inequality: - Deininger & Squire (Gini coefficient) 1980 -2000 (4 5 yr. periods) 61 countries, 198 observations - World Income Distribution (decile) 1988 -1998 (3 5 yr. periods) 55 countries, 146 observations Good representation across regions and developing countries are adequately represented 2/3 of developing countries (same for observations) However, we drop transition countries from the sample.

Technique • Use of country-fixed effects • Correct the standard errors for heteroskedasticity and

Technique • Use of country-fixed effects • Correct the standard errors for heteroskedasticity and correlation within countries (panel-corrected standard errors, Beck & Katz) ≈ robust cluster • Many controls – macro-economic (inflation) – year-dummies (globalization) – sources of data • (individual/household; income/expenditure; gross/net income)

3. 1 Openness , Income and Inequality (no endowments yet) • HOS interpretation: &

3. 1 Openness , Income and Inequality (no endowments yet) • HOS interpretation: & turning point: • Specification 1: impact according to GDPpc level: usual test of HOS. • Motivation: distinction between skilled and unskilled or between labor and capital through the level of GDPpc.

Table 3 1 2 3 5 OLS OLS FE Gini GDPpc -0. 0691 (1.

Table 3 1 2 3 5 OLS OLS FE Gini GDPpc -0. 0691 (1. 48) -0. 0341 (0. 65) -0. 0712 b (2. 00) 0. 0740 (1. 22) Tariffst-5 -5. 7662 b (2. 29) -4. 9716 c (1. 86) -5. 1414 b (2. 41) 4. 1129 c (1. 89) Tariffst-5 * GDPpc 0. 6503 b (2. 25) 0. 5477 c (1. 75) 0. 6100 b (2. 45) -0. 4647 c (1. 80) Education 0. 0310 (0. 47) 0. 0385 (0. 50) -0. 0332 (0. 60) -0. 0227 (0. 33) -0. 4185 a (4. 65) -0. 3253 a (3. 63) -0. 3475 a (4. 47) 0. 0580 (0. 53) Ethnicity 0. 0229 (1. 62) 0. 0302 b (2. 19) 0. 0291 b (2. 48) Civil Liberties 0. 0450 (0. 88) 0. 0437 (0. 81) 0. 0565 (1. 38) 0. 0318 (0. 70) Inflation 0. 0783 b (2. 34) -0. 0001 (0. 00) 0. 0665 a (3. 28) 0. 0063 (0. 51) Dependent variable Mature M 2/Gdp -0. 1569 a (3. 48) Gov. expenditure 0. 1232 b (2. 55) Observations 224 172 224 215 # Countries 75 61 75 66

 1980 1995 Gini Tariffs GDP per capita

1980 1995 Gini Tariffs GDP per capita

3. 2 Trade Liberalization, Endowments Inequality • Measure of relative factor endowments:

3. 2 Trade Liberalization, Endowments Inequality • Measure of relative factor endowments:

 Improvement: finer skilled disaggregation 3. 2 Trade Liberalization, Endowments Inequality • Motivation –

Improvement: finer skilled disaggregation 3. 2 Trade Liberalization, Endowments Inequality • Motivation – Wood (1995): No-educated Workers are not in the manufactured export sector – Similar arguments can be found in Milanovic (2002), Kremer-Maskin (2003), Bensidoun & al. (2005) Method: No-educated (NO-ED): uneducated or primary uncomp. Based Educated (BS-ED): primary comp & sec. uncomp. Skilled Educated (SK-ED): secondary completed & more

Table 4 1 2 3 4 5 Gini Gini Tariffst-5 -0. 4915 (1. 14)

Table 4 1 2 3 4 5 Gini Gini Tariffst-5 -0. 4915 (1. 14) -0. 4147 (0. 96) -0. 0531 (0. 14) Tariffst-5 -0. 4834 (1. 16) -0. 5865 (1. 44) (AT/L) *(Tariffst-5) 0. 3467 (0. 76) 0. 3710 (0. 80) -0. 1141 (0. 29) (AT/L) *(Tariffst-5) 0. 3470 (0. 78) 0. 6036 (1. 28) (K/L) *(Tariffst-5) -0. 4335 b (2. 56) -0. 2809 c (1. 78) 0. 3710 c (1. 76) (K/L) *(Tariffst-5) -0. 4252 b (2. 18) -0. 4854 b (2. 48) (MF/L) *(Tariffst-5) ((NO-ED)/L) *(Tariffst-5) -1. 1000 b (2. 40) ((BS-ED)/L) *(Tariffst-5) 0. 1436 (1. 37) ((SK-ED)/L) *(Tariffst-5) -0. 1121 (0. 80) (SK-ED/BS-ED) *(Tariffst-5) -0. 6877 a (2. 71) -0. 6110 b (2. 38) ((SK-ED+BS-ED)/NO-ED) *(Tariffst-5) 0. 9021 a (3. 19) 1. 0188 a (3. 66) -0. 8094 a (3. 68) Observations 198 198 Observations 198 # Countries 61 61 61 # Countries 61 61 + ratio (Wood): (SK-ED+BS-ED)/NO-ED SK-ED/BS-ED

3. 2 Trade Liberalization, Endowments Inequality Robustness • Application with quintile on a smaller

3. 2 Trade Liberalization, Endowments Inequality Robustness • Application with quintile on a smaller database: results hold • Addition of macro variables : ok • Use other index of openness: results change sometimes (because of heterogeneity between trade indexes (e. g. Prichett 1996)) • Exclusion of outliers (7 observations): results hold

4. Openness, Inequality and Poverty: Further Results World Income Distribution (WYD) database: • New

4. Openness, Inequality and Poverty: Further Results World Income Distribution (WYD) database: • New data set (more homogeneous) (drawn almost entirely from household surveys) • Allows to study the impact of trade liberalization on each decile (does not depend of the inequality measure) Allows to check for robustness Evaluation also on poverty use of mean income rather than GDP per capita - Deaton (2005)

Table 5 lnshare 1 lnshare 2 lnshare 3 lnshare 8 lnshare 9 lnshare 10

Table 5 lnshare 1 lnshare 2 lnshare 3 lnshare 8 lnshare 9 lnshare 10 lngini Tariffs 14. 6459 a (3. 88) 6. 0793 a (4. 12) 3. 2486 a (3. 23) -0. 5300 a (3. 79) -0. 7518 a (4. 85) -0. 2154 (0. 56) -1. 0508 a (2. 75) (Capital/Labor) * Tariffs 3. 7957 a (3. 31) 1. 4825 a (3. 15) 0. 8644 a (2. 70) -0. 3332 a (3. 31) -0. 4344 a (4. 07) 0. 0992 (0. 36) -0. 2417 (0. 94) (Arable Land/Labor) * Tariffs -12. 2183 a (3. 03) -4. 6897 a (3. 79) -2. 3084 a (2. 74) 0. 3207 a (2. 73) 0. 5824 a (4. 46) 0. 4285 (1. 47) 0. 7787 b (2. 01) (Mining&Fuel/Labor) * Tariffs -0. 5832 (1. 07) -0. 2138 (0. 93) -0. 1741 (1. 04) -0. 0216 (0. 50) -0. 0996 b (2. 28) -0. 0280 (0. 34) -0. 0169 (0. 19) (Skill/Basic-Edu) * Tariffs 0. 6794 (0. 56) 0. 0873 (0. 14) 0. 0788 (0. 18) 0. 0552 (0. 78) -0. 0447 (0. 44) -0. 5956 a (3. 27) -1. 0939 a (5. 65) [(Skill + Basic-Edu) / No-Edu] * Tariffs -1. 9963 (1. 42) -0. 9401 (1. 60) -0. 5866 (1. 18) -0. 1438 a (2. 92) -0. 0278 (0. 40) 0. 4496 b (2. 28) 0. 5304 b (2. 48) Inflation -0. 1946 b (2. 38) -0. 0540 (1. 52) -0. 0176 (0. 71) 0. 0113 a (2. 99) -0. 0022 (0. 26) 0. 0210 b (2. 02) 0. 0300 a (3. 17) Data source (Income vs expenditure) -0. 2013 a (2. 80) -0. 0943 b (2. 56) -0. 0782 a (2. 63) 0. 0093 (1. 51) 0. 0008 (0. 17) 0. 0231 c (1. 73) 0. 0407 a (4. 29) Observations 146 146 # Countries 55 55

Robustness - endogeneity • Reverse causality: Since we do not have a good instrument

Robustness - endogeneity • Reverse causality: Since we do not have a good instrument for tariffs, we ran the same regression using future trade rather the past values. Results become mostly insignificant, suggesting that reverse causality should not be a problem here. • Endogeneity In absence of worthwhile instruments we take refuge behind argument that plausibly endogeneity may not be too worrisome. E. g. Goldberg and Pavcnik (2004) note that the preoccupation about the endogeneity of tariffs is lessened by the fact that many countries moved towards a reduction in protection and more uniformity in their tariff structures when they became full members of the GATT/WTO.

Quantify the results For 3 groups of developing countries: - Sub-Saharan Africa Ghana (2),

Quantify the results For 3 groups of developing countries: - Sub-Saharan Africa Ghana (2), Lesotho (2), Kenya (2), Uganda (2), and Zimbabwe (2) - Latin America Argentina (3), Bolivia (3), Brazil (3), Colombia (3), Costa Rica (3), Dominican Republic (3), Ecuador(3), Jamaica (3), Mexico (3), Nicaragua (2), Panama (3), Paraguay (2), Peru (3), Uruguay (3) and Venezuela (3) - East, South and South-East Asia Bangladesh (2), China (2), India (3), Indonesia (2), Korea (3), Malaysia (3), Pakistan (3), Philippines (3), Singapore (3), Sri Lanka (3) and Thailand (3)

Impact of a diminution of tariffs of 5 points (next slides explain each result)

Impact of a diminution of tariffs of 5 points (next slides explain each result) Africa Latin America Asia Share 1 Share 2 Share 3 Share 8 Share 9 Share 10 Gini A 1. 9% - 2. 3% 2. 9% - 3. 1% 3. 9% - 4. 0% 11. 2% - 10. 7% 15. 6% - 15. 4% 38. 0% - 39. 4% B 230 - 272 350 - 368 460 - 474 1291 - 1234 1774 - 1744 4191 - 4341 C -1. 70% -0. 50% -0. 30% 0. 40% 0. 20% -0. 40% 0. 464 – 0. 437 A 1. 3% - 1. 0% 2. 5% - 2. 3% 3. 6% - 3. 4% 11. 6% - 11. 5% 16. 6% - 17. 1% 38. 0% - 38. 1% B 348 - 280 704 - 636 1007 - 947 3306 - 3293 4763 - 4929 10994 - 11040 C 2. 1% 1. 0% 0. 6% 0. 0% -0. 3% 0. 0% A 3. 0% - 2. 2% 4. 3% - 3. 8% 5. 2% - 4. 9% 11. 6% - 11. 5% 15. 1% - 15. 3% 29. 6% - 29. 5% B 613 - 445 955 - 834 1184 - 1103 2704 - 2658 3486 - 3549 6692 - 6679 C 3. 1% 1. 3% 0. 7% 0. 2% -0. 2% 0. 0% Row A reports each decile’s share in income before and after the simulated reduction in tariff protection Row B reports the corresponding incomes in $1993 PPP Row C shows the corresponding annual real growth (over the 10 years) that would be necessary to keep each decile’s income at its initial value 0. 482 – 0. 483 0. 358 – 0. 357

Impact of a diminution of tariffs of 5 points (I) (effect on income and

Impact of a diminution of tariffs of 5 points (I) (effect on income and on shares) Africa Latin America Asia Share 1 Share 2 Share 3 Share 8 Share 9 Share 10 Gini A 1. 9% - 2. 3% 2. 9% - 3. 1% 3. 9% - 4. 0% 11. 2% - 10. 7% 15. 6% - 15. 4% 38. 0% - 39. 4% B 230 - 272 350 - 368 460 - 474 1291 - 1234 1774 - 1744 4191 - 4341 C -1. 70% -0. 50% -0. 30% 0. 40% 0. 20% -0. 40% 0. 464 – 0. 437 A 1. 3% - 1. 0% 2. 5% - 2. 3% 3. 6% - 3. 4% 11. 6% - 11. 5% 16. 6% - 17. 1% 38. 0% - 38. 1% B 348 - 280 704 - 636 1007 - 947 3306 - 3293 4763 - 4929 10994 - 11040 C 2. 1% 1. 0% 0. 6% 0. 0% -0. 3% 0. 0% A 3. 0% - 2. 2% 4. 3% - 3. 8% 5. 2% - 4. 9% 11. 6% - 11. 5% 15. 1% - 15. 3% 29. 6% - 29. 5% B 613 - 445 955 - 834 1184 - 1103 2704 - 2658 3486 - 3549 6692 - 6679 C 3. 1% 1. 3% 0. 7% 0. 2% -0. 2% 0. 0% Row A reports each decile’s share in income before and after the simulated reduction in tariff protection Row B reports the corresponding incomes in $1993 PPP Row C shows the corresponding annual real growth (over the 10 years) that would be necessary to keep each decile’s income at its initial value 0. 482 – 0. 483 0. 358 – 0. 357

Impact of a diminution of tariffs of 5 points (II) (effect on Gini varies

Impact of a diminution of tariffs of 5 points (II) (effect on Gini varies across regions) Africa Latin America Asia Share 1 Share 2 Share 3 Share 8 Share 9 Share 10 Gini A 1. 9% - 2. 3% 2. 9% - 3. 1% 3. 9% - 4. 0% 11. 2% - 10. 7% 15. 6% - 15. 4% 38. 0% - 39. 4% B 230 - 272 350 - 368 460 - 474 1291 - 1234 1774 - 1744 4191 - 4341 C -1. 70% -0. 50% -0. 30% 0. 40% 0. 20% -0. 40% 0. 464 – 0. 437 A 1. 3% - 1. 0% 2. 5% - 2. 3% 3. 6% - 3. 4% 11. 6% - 11. 5% 16. 6% - 17. 1% 38. 0% - 38. 1% B 348 - 280 704 - 636 1007 - 947 3306 - 3293 4763 - 4929 10994 - 11040 C 2. 1% 1. 0% 0. 6% 0. 0% -0. 3% 0. 0% A 3. 0% - 2. 2% 4. 3% - 3. 8% 5. 2% - 4. 9% 11. 6% - 11. 5% 15. 1% - 15. 3% 29. 6% - 29. 5% B 613 - 445 955 - 834 1184 - 1103 2704 - 2658 3486 - 3549 6692 - 6679 C 3. 1% 1. 3% 0. 7% 0. 2% -0. 2% 0. 0% Row A reports each decile’s share in income before and after the simulated reduction in tariff protection Row B reports the corresponding incomes in $1993 PPP Row C shows the corresponding annual real growth (over the 10 years) that would be necessary to keep each decile’s income at its initial value 0. 482 – 0. 483 0. 358 – 0. 357

Impact of a diminution of tariffs of 5 points (III) (Estimated growth to keep

Impact of a diminution of tariffs of 5 points (III) (Estimated growth to keep income at initial value) Africa Latin America Asia Share 1 Share 2 Share 3 Share 8 Share 9 Share 10 Gini A 1. 9% - 2. 3% 2. 9% - 3. 1% 3. 9% - 4. 0% 11. 2% - 10. 7% 15. 6% - 15. 4% 38. 0% - 39. 4% B 230 - 272 350 - 368 460 - 474 1291 - 1234 1774 - 1744 4191 - 4341 C -1. 70% -0. 50% -0. 30% 0. 40% 0. 20% -0. 40% 0. 464 – 0. 437 A 1. 3% - 1. 0% 2. 5% - 2. 3% 3. 6% - 3. 4% 11. 6% - 11. 5% 16. 6% - 17. 1% 38. 0% - 38. 1% B 348 - 280 704 - 636 1007 - 947 3306 - 3293 4763 - 4929 10994 - 11040 C 2. 1% 1. 0% 0. 6% 0. 0% -0. 3% 0. 0% A 3. 0% - 2. 2% 4. 3% - 3. 8% 5. 2% - 4. 9% 11. 6% - 11. 5% 15. 1% - 15. 3% 29. 6% - 29. 5% B 613 - 445 955 - 834 1184 - 1103 2704 - 2658 3486 - 3549 6692 - 6679 C 3. 1% 1. 3% 0. 7% 0. 2% -0. 2% 0. 0% Row A reports each decile’s share in income before and after the simulated reduction in tariff protection Row B reports the corresponding incomes in $1993 PPP Row C shows the corresponding annual real growth (over the 10 years) that would be necessary to keep each decile’s income at its initial value 0. 482 – 0. 483 0. 358 – 0. 357

Figure 2: Simulated changes in quintile mean incomes of a 5 percentage points reduction

Figure 2: Simulated changes in quintile mean incomes of a 5 percentage points reduction in tariffs Figure 2 a: bottom quintile*

Figure 2 b: top quintile* * Simulated quintile share before tariff reduction on the

Figure 2 b: top quintile* * Simulated quintile share before tariff reduction on the horizontal axis, and changes in quintile share following the tariff reduction (here, a 5 percentage points) on the vertical axis. For example, the average income share of the poorest 20% of Indonesia (IDN) is reduced from 6% of total income to 4% after the tariff reduction

Conclusions • Paper restores role for HOS and factor endowments effects in study of

Conclusions • Paper restores role for HOS and factor endowments effects in study of trade liberalization effects on inequality. Impact of trade liberalization on income distribution (Summary) Countries relatively well endowed in: Impact on inequality: • Capital / labor • Arable land / labor • Mining resources • High skilled workers • Low skilled workers • No-educated workers Mostly insignificant

Conclusions (bis) • Poor countries with no-ed (in non-tradables) could see inequality rise with

Conclusions (bis) • Poor countries with no-ed (in non-tradables) could see inequality rise with trade liberalization • Rather robust results though omitted variable bias surely still there. • Some quantification and supports for “looking beyond averages”