Productivity Differences Within and Between Countries Melissa Dell MIT OECD Lagging Regions Meeting, June 2010 Dell (MIT) Productivity Differences OECD Lagging Regions 1 / 21
Motivation Introduction and Conceptual Framework Income differences across geographic space are both large and persistent. This talk provides a comparative perspective from the Americas, in order to: Introduce a context for guaging the generalizability of patterns that emerge in the OECD Highlight methodologies Offer additional insight into why regional inequalities arise and persist Dell (MIT) Productivity Differences OECD Lagging Regions 2 / 21
Introduction and Conceptual Framework Outline for Presentation 1 Quantify the magnitude of income differences between regions within countries, relative to the magnitude of income differences across countries. 2 Briefly discusses why differences arise and persist Geography Public infrastructure Example of detailed quantitative study on persistence Dell (MIT) Productivity Differences OECD Lagging Regions 3 / 21
Inequality Patterns in the Americas Examining Inequality in the Americas: Data Population censuses and household expenditure surveys for 18 countries in the Western Hemisphere. 11 countries with income data geo-referenced to municipalities (municipality mean population between 26,000 and 108,000) 7 additional countries with data geo-referenced to bigger regions. Results both for labor income and household expenditure. Construct comparable measures across surveys for both. Results both with regionally-deflated incomes and nationally-deflated incomes. Dell (MIT) Productivity Differences OECD Lagging Regions 4 / 21
Inequality Patterns in the Americas Figure A2: Labor incomes in Mexico and Central America Incomes in Central America Mean Labor Income (PPP $) <4,000 4,000-7,000 7,000-10,000 10,000-15,000 >15,000 Dell (MIT) Productivity Differences OECD Lagging Regions 5 / 21
Inequality Patterns in the Americas Incomes in South America Figure A3: Labor incomes in South America Mean Labor Income (PPP $) <4,000 4,000-7,000 7,000-10,000 10,000-15,000 >15,000 Dell (MIT) Productivity Differences OECD Lagging Regions 6 / 21
Inequality Patterns in the Americas Empirical Strategy Decompose inequality within the Western Hemisphere into between-country, between-municipality and within-municipality differences. Decompose labor income into predicted (from a Mincer regression) and residual components. Relevant for understanding the role of human capital vs. residual factors Residual factors similar to technology in cross-country models (physical capital mobile within national boundaries). Use additively decomposable measures of inequality (from the General Entropy class). Theil Index and Mean Log Deviation Index Dell (MIT) Productivity Differences OECD Lagging Regions 7 / 21
Results: Summary Inequality Patterns in the Americas Large within-country differences. Between-municipality differences about half to a quarter of between-country differences with the United States included. Without the United States, about twice between-country. Large between-country predicted income differences Human capital factors explain about half of the between-country and between-municipality differences. Factors related to differences in productive efficiency (i.e. local institutions, public goods) appear important for regional variation. Dell (MIT) Productivity Differences OECD Lagging Regions 8 / 21
Inequality Patterns in the Americas Results: Between vs. Within Income Inequality (Theil Index) Between Within Between Cntry. Cntry. Mun. sample (pop. weights) (1) (2) (3) Mun. (actual) 0.250 0.544 0.058 Mun. (equal) 0.285 0.622 0.088 No US (equal) 0.048 0.706 0.114 All (actual) 0.253 0.542 0.054 All (equal) 0.235 0.619 0.061 No US/CA (equal) 0.071 0.726 0.081 Dell (MIT) Productivity Differences OECD Lagging Regions 9 / 21
Inequality Patterns in the Americas Results: Predicted vs. Residual Predicted Labor Income Residual Labor Income Btwn Btwn Within Btwn Btwn Within Cntry Mun/Reg Mun/Reg Cntry Mun/Reg Mun/Reg (1) (2) (3) (4) (5) (6) Ref. to Mun. Mun. (actual) 0.170 0.015 0.131 0.033 0.040 0.389 Mun. (equal) 0.166 0.040 0.142 0.041 0.053 0.404 No US (equal) 0.031 0.053 0.157 0.040 0.057 0.421 All (actual) 0.163 0.014 0.130 0.040 0.037 0.392 All (equal) 0.158 0.026 0.140 0.045 0.043 0.433 No US/CA (equal) 0.081 0.035 0.158 0.042 0.050 0.467 Dell (MIT) Productivity Differences OECD Lagging Regions 10 / 21
26 C Climate 4 C Average Annual Temperature State Boundaries Channels Median Income (PPP $) <1,500 1,500-3,000 3,000-4,500 4,500-6,000 >6,000 State Boundaries Brazil - Temperature Brazil - Labor Income Mean Annual Temperature 28 C 14 C State Boundaries Median Income (PPP $) <4,500 4,500-5,500 5,500-6,500 6,500-8,000 >8,000 State Boundaries Dell (MIT) Productivity Differences OECD Lagging Regions 11 / 21
Climate Channels Use high-resolution data on long-run climate averages to examine the relationship between climate and economic prosperity. Dependent Variable is: Log per capita GDP (PWT) Log labor income (1) (2) (3) (4) (5) Temperature -0.085* -0.089-0.085** -0.012*** -0.019** (0.017) (0.072) (0.004) (0.004) (0.009) Precipitation 0.000 0.019-0.003** 0.000 0.002 (0.015) (0.047) (0.001) (0.001) (0.001) Elevation, slope, coast no no yes yes yes Country F.E. no no no yes yes State F.E. no no no no yes R-squared 0.23 0.21 0.61 0.82 0.88 Number of clusters 260 260 260 Number of observations 134 12 7684 7684 7684 Dell (MIT) Productivity Differences OECD Lagging Regions 12 / 21
Road Infrastructure Channels Theil Index Income Regressions Between Within Country Country Baseline Controls (1) (2) (3) (4) Brazil 1.049-0.022-0.019 (0.004) (0.003) Mexico 0.379-0.124-0.096 (0.011) (0.010) Panama 0.756-0.157-0.138 (0.026) (0.025) United States 0.795-0.080-0.076 (0.026) (0.021) Venezuela 0.747-0.017 0.010 (0.006) (0.006) All (actual) 0.439 0.815 All (equal) 0.286 0.656 No U.S. (equal) 0.249 0.655 Dell (MIT) Productivity Differences OECD Lagging Regions 13 / 21
Channels A Historical Example Now we will discuss a study examining the fundamental determinants of sub-national income differences in Andean Peru These income differences are large To understand their origins, we must briefly discuss institutions in colonial Peru Dell (MIT) Productivity Differences OECD Lagging Regions 14 / 21
The Mining Mita Channels The mining mita was instituted by the Spanish government in 1573 and abolished in 1812 It required over 200 indigenous communities in Peru and Bolivia to send one seventh of their adult male population to work in the Potosí silver and Huancavelica mercury mines The Potosí mines, discovered in 1545, provided the largest deposits of silver in the Spanish Empire The mita assigned 14,181 conscripts from southern Peru and Bolivia to the Potosí mines and 3,280 conscripts from central and southern Peru to the state-owned Huancavelica mines (Bakewell, 1984, p. 83) Dell (MIT) Productivity Differences OECD Lagging Regions 15 / 21
Channels The mita s extent Huancavelica! Potosi! Uyuni Salt Flat Study Boundary Mita Boundary 5000 m 0m Dell (MIT) Productivity Differences OECD Lagging Regions 16 / 21
Results Channels The mita s long run effects lower household consumption by around 25% in subjected districts today and increases malnutrition in children by around six percentage points. Channels Land tenure Large negative effect on the concentration of haciendas - few large landowners in mita districts Impact persisted through the 20th century Public goods Mita districts less integrated into road networks Historically lower levels of education Markets and subsistence Residents of mita districts are substantially more likely to be subsistence farmers Dell (MIT) Productivity Differences OECD Lagging Regions 17 / 21
Channels Interpretation My hypothesis: The long-term presence of large landowners in non-mita districts provided a stable land tenure system that encouraged public goods provision. Why is public goods provision higher outside the mita catchment? 1 Large landowners controlled a larger percentage of the productive factors (land and labor) 2 Property rights less secure in mita districts Incentives to protect peasant rights to land disappeared when the mita was abolished De facto communal land tenure, numerous land confiscations, widespread livestock rustling and banditry (Flores Galindo, 1987; Jacobsen, 1993; Tamayo Herrera, 1982) 3 Landowners possessed the political connections required to secure public goods Roads twist to pass through as many haciendas as possible (Stein, 1980) Dell (MIT) Productivity Differences OECD Lagging Regions 18 / 21
6 12 Results Channels Figure 2 16 S 15.5 S 15 S 14.5 S 14 S 13.5 S 13 S 70 Obs. 102 Obs. Obs. 38 Obs. <5.50 6.20 6.90 7.60 8.30 >9.00 16 S 15.5 S 15 S 14.5 S 14 S 13.5 S 13 S Obs. <0.10 5998 Obs. 0.18 0.26 11984 Obs. 0.34 0.42 17970 Obs. >0.50 74 W 73 W 72 W 71 W 74 W 73 W 72 W 71 W (a) Consumption (2001) (b) Stunting (2005) 16 S 15.5 S 15 S 14.5 S 14 S 13.5 S 13 S 0.00 5.00 10.00 15.00 20.00 >25.00 16 S 15.5 S 15 S 14.5 S 14 S 13.5 S 13 S 0.00 0.26 0.52 0.78 1.04 >1.30 74 W 73 W 72 W 71 W 74 W 73 W 72 W 71 W (c) Haciendas (1689) (d) Haciendas (1845) Dell (MIT) Productivity Differences OECD Lagging Regions 19 / 21
Channels Results 13 S 15 S 14.5 S 14 S 13.5 S 0.00 0.18 0.36 0.54 0.72 >0.90 16 S 74 W 73 W 72 W 73 W 74 W 71 W 15 S 50 Obs. 0.00 0.10 0.20 0.30 0.40 >0.50 13 S 5378 Obs. 10706 Obs. 13.5 S 16034 Obs. 14 S 16 S 16 S 71 W 14.5 S 14.5 S 14 S 13.5 S <5.00 18.00 31.00 44.00 57.00 >70.00 15 S 15.5 S 72 W (f) Education (1876) 15.5 S 13 S (e) Haciendas (1940) 74 W 73 W 72 W (g) Road Density (2006) Dell (MIT) <0.01 0.04 0.07 0.09 0.12 >0.15 16 S 15.5 S 15 S 14.5 S 14 S 13.5 S 15.5 S 13 S Figure 2 (cont.) 71 W 74 W 73 W 72 W 71 W (h) Ag. Market Participation (1994) Productivity Differences OECD Lagging Regions 20 / 21
Channels Summary Subnational income differences are large, even relative to cross-country income differences When U.S. excluded, about twice as large as cross-country differences Evidence that both geography and institutional arrangements play a role Dell (MIT) Productivity Differences OECD Lagging Regions 21 / 21