Persistence of Relative Income for Countries and Populations David N. Weil Brown University and NBER 25th February 2014
2nd Type of Persistence: Levels of Development by Place Economists are increasingly interested in drawing the link from things that happened long ago to current development. Hibbs and Olsson (2004, 2005): date of transition to agriculture correlated with current income. Chanda and Putterman (2006, 2007): state history and timing of agricultural transition predict income and population density in 1500. Bockstette, Chanda and Putterman (2002): index of the presence of state-level political institutions from year 1 to 1950 has positive correlation with current income. Comin, Easterly, and Gong (2006) show that the state of technology in a country 500, 1500, or even 3000 years ago has predictive power for the level of output today.
Comin, Easterly, and Gong Was the Wealth of Nations Determined in 1000 BC (2010) They measure level of technology for years 1000 BC, 0, and 1500 AD Examine dierent categories of technology (e.g. agriculture, military, communications) Categories of technology and scaling of sophisticaion varies with year examined
Comin, Easterly, and Gong Was the Wealth of Nations Determined in 1000 BC (2010) They measure level of technology for years 1000 BC, 0, and 1500 AD Examine dierent categories of technology (e.g. agriculture, military, communications) Categories of technology and scaling of sophisticaion varies with year examined Data in each sector is scaled relative to max, and all sectors are then averaged. Data coded to current country borders Focus is extensive margin (max within borders) not intensive (how much used) Parts of empire get technology of metropole (e.g. Ottoman) 1500 means pre-1492 in Americas
Distribution of Technology in 1500
A Famous Case of Non-Persistence Acemoglu, Johnson, and Robinson, Reversal of Fortune: Geography and Institutions in the Making of the Modern World Income Distribution (QJE 2002) Among countries colonized by Europeans, there is a negative correlation between income per capita in 1500 and income today.
The Meaning of Reversal of Fortune Strong evidence against simple geography hypothesis, i.e. poor places are poor because they have bad geography. but changes in technology (temperate climate crops, animals, iron tools, and old-world tropical diseases make things more complicated) allow one to maintain a more sophisticated geography hypothesis. Their story is that dense pre-colonial populations led colonizers to create persistent bad institutions (more later) It is hard to ignore the observation that the people who live in the most reversed countries are not the descendents of the people who lived there in 1500.
Third Type of Persistence: Relative Development of Population Groups Putterman and Weil Post-1500 Population Flows and the Long-Run Determinants of Economic Growth and Inequality (2010) Things that happened long ago in a place are not the same as things that happened long ago to the people who now live in a place. Depending on the theory that you are examining, the looking at the history of the people will be better than looking at the place.
Putterman and Weil Contributions 1 The migration matrix: A tool for adjusting the data so that we can look at the early development of the people who live in a place rather than the place itself. 2 We show that this greatly improves the predictive power of measures of early development. 3 Use our matrix to look at the determinants of inequality, which has previously not been addressed at all in this early development literature. 4 The matrix of migration ows can also be used as an object of study in its own right.
The Migration Matrix N receiving countries N source countries (these don't need to be equal, but we do it that way) Each entry is the fraction of the current population in country i that is descended from country j Example: China India Indonesia Malaysia Philippines Malaysia.26.075.04.60.025
Caveats Sources migration includes slavery, etc. We impose modern borders on the world from 1500. Encyclopedias, ethnographic databases, national sources, CIA factbook, etc. (See voluminous appendices.) We also use genetic data in some cases.
Example of Using Genetic Data: Mexico Population groups for Mexico: 29.5% Amerindian, 60% Mestizo, 9% European, 0.7% African, and 0.8% other. How should we attribute the ancestry of mestizos? Bonilla et al's (2005): 19 estimates from ve dierent studies, covering 14 states (65% of population) review to obtain genetic admixture estimates for Mexican mestizos. We average studies within state and weight by state population to get national estimate. Result: Mexican mestizos are 35% European, 55% Amerindian, and 10% African. Dividing up the Mestizos gives: 30% of Mexico's ancestors were European, 62.5% Amerindian (Mexico), 6.7% African (Africa Mix), and 0.8% other.
Interesting Findings from Genetic Data Caribbean Amerindian populations not completely wiped out (3.6% of ancestry in Dominican Republic) Mitochondrial DNA vs. Y Chromosome. Whites are not pure European: Costa Rica 86.5% describe selves as Spanish, but population ancestry is 61% Spanish, 30% Amerindian, and 9% African. More African ancestry among mestizos than usually ascribed (for example, 10% among Mexican mestizos).
Early Development Data Statehist: extent of centralized, indigenous government over the period 1-1500 illustrative values: Ethiopia (1), China (.906), Egypt (.760), Spain (.562), Senegal (.398), US (0) Agyears: Number of millennia since a country transitioned from hunting and gathering to agriculture illustrative values: Jordan (10.5), Iraq (10), China (9), Ecuador (4), Congo (3), Haiti (1), Australia (0.4)
Ancestry Adjusting data adjusted statehist =P statehist Entries from the Matrix: China India Indonesia Malaysia Philippines Malaysia.26.075.04.60.025 Statehist data: China.906 India.688 Indonesia.550 Malaysia.594 Philippines 0 adjusted statehist =.666
Ancestry Adjusting State History
Ancestry Adjusting Agyears
Basic Horserace Regression
Robustness Basic Horserace result is robust to Neo Europe adjustment Inclusion of % of population speaking a European language Inclusion of % of population of European descent Inclusion of % of population not descended from natives Eurasia dummy, latitude, landlocked
Early Development Indicators Revisted Dep Var: ln(gdp/capita in 2000) geo conditions 0.752 (.075) ancestry-adjusted geo conditions 0.952 (.069) bio conditions.746 (.081) ancestry adjusted bio conditions.947 (.074) R 2.415.574.417.581 Observations 105 105 105 105 geo conditions = rst principal component of climate suitability, latitiude, size of landmass on which located, and east-west orientation of landmass bio conditions = FPC of the number of heavy seeded grasses and domestecatble animals that existed in a region in pre-history
Predictive Eect of Technology Revisited Dep Var: ln(gdp/capita in 2000) Technology Index 1 CE 0.0942 (.3758) ancestry-adjusted Technology 2.51 Index 1 CE (.059) Technology Index 1500 CE 1.55 (.30) ancestry adjusted Technology 3.26 Index 1500 CE (.30) R 2.000.133.183.525 Observations 125 125 114 114
Channels? Executive Expropriation Government Constraints Risk Eectiveness statehist 0.158 0.658 0.455 (0.274) (0.287) (0.271) ancestry adjusted 0.670 1.33 1.32 statehist (0.309) (0.33) (0.30) Rsq.002.033.047.134.019.123 Obs 141 141 111 111 144 144 Dependent variables normalized to have SD of 1, positive is good
Early Development and Inequality Theory Empirics Whatever is good about early development may apply to people within a country. Heterogeneity in early development may have produced nonegalitarian institutions. Regress inequality on variance of early development Compare heterogeneity of early development with other measures of heterogeneity as predictors of current inequality Look at incomes of dierent Statehist groups within a country
Within Country Variance of Early Development adjusted statehist =P statehist Entries from the Matrix: China India Indonesia Malaysia Philippines Malaysia.26.075.04.60.025 Statehist data: China.906 India.688 Indonesia.550 Malaysia.594 Philippines 0 adjusted statehist =.666 Var(statehist) =.26 (.26.666) 2 +.075 (.688.666) 2 + etc.
Heterogeneity of Early Development mean of standard deviation standard deviation of standard deviation statehist.097.089 agyears.764.718
Historical Determinants of Current Inequality Dependent Variable: Gini Coecient standard deviation.456.408 statehist (0.088) (.084) ancestry adjusted -.148 statehist (.036) standard deviation.0512.0571 of agyears (.0121) (.0108) ancestry adjusted -.0217 agyears (.0052) Rsq.140.267.108.260 Obs 135 135 140 140
Ethnic Fractionalization Dependent Variable: Gini Coecient Ethnic Fractionalization.117.0517 (0.037) (.0355) Historical Fractionalization.0134 -.0116 (.034) (.0489) standard deviation.392.435 of statehist (.085) (.124) ancestry adjusted -.130 -.148 statehist (.040) (.037) Rsq.073.101.276.267 Obs 132 135 132 135 Ethnic Fractionalization: Probability that two randomly selected people will be in dierent ethnic groups (current denitions) Historical Fractionization: Probability that two randomly selected people will have same national ancestry (as if individuals had pure ancestry)
Correlation of Within Country Rank and Statehist Are people from countries with earlier development at the top of the income distribiution? We looked at the 10 countries with highest SD of statehist plus the United States (rank #18). All are former colonies, all but 3 in Americas. We look at locally dened ethnic groups (Mestizo, Colored, etc.) We assign them average statehist using same techniques used in estimating the matrix. We compare average statehist rank to average income ranking using local sources.
Country Standard Rank Ethnic Percent Statehist Income Rank Dev Group of Pop (average) of Statehist Fiji.346 1 Other 4 0.745 High Indo-Fijian 41 0.688 Middle Fijian 55.000 Low Cape Verde.301 2 White 1.723 High Creole 71.473 Middle Black 28.142 Low United States.232 18 White not 67.4.650 Upper Middle Hispanic Asian 4.2.640 High Hispanic of 14.1.485 Lower Middle any race Black 12.8.240 Low American Indian 1.0.000 Lower Middle or Alaska Native For 9 out of 11 cases, average statehist and income rank dovetail perfectly
What Does it All Mean? Early development conferred some advantage(s) on the countries where descendents of early developers live and on the descendents themselves. Candidates include: human capital, culture, social capital, institutions, genes. Inequality results suggest it is not just good institutions
Spolaore and Wacziarg PW seem to suggest that something good for growth is carried with populations: insitutions, culture, human capital, etc. Spolaore and Wacziarg (2009, 2013) take a dierent view dierences in culture/ethnicity/language serve as barriers to transmission of technology/institutions as well as trade. These are _measured_ with genetic distance, even though the genes examined are not relevant for any outcomes. earlier work nds genetic distance between populations negatively aects bilateral trust and trade
Phylogenic Tree of Human Populations