AREC 345: Global Poverty and Economic Development Lecture 4 Professor: Pamela Jakiela Department of Agricultural and Resource Economics University of Maryland, College Park Does Absolute Latitude Explain Underdevelopment?
Does Latitude Explain Underdevelopment? Scatter plot of log GDP per capita by latitude QAT KWT LUX ARE NOR OMN HKG SAU USA CHE BHR AUS AUT BEL DEU NLD IRL DNK CANSWE JPN ESP ITA FRA GBR FINISL LBY ISR KORGRCNZL MLT PRT SVN CZE SVK HUN EST HRV KAZ POL LTU RUS GAB VEN AZE BGR BLR BRA BRB IRN ROM LVA SUR PAN URY TUR CRI THA BWA DZA IRQ DOM EGY JOR MKD TUN TKM BLZ BIH JAM NAM LKA SLVGTM UKR PRY BTN MAR ARM GEO NGA BOL NID VNM IND UZB MDA STP CIV DJI ZMB KGZ KEN PNG SEN BGD LSO TJK COM BFA MDG ZWE UGA RWA TGO ETH ERI NERMOZ BDI LBR MWI ZAR Q: What is the correlation between income and latitude? A: 0.580 Lecture4,Slide3 Does Latitude Explain Underdevelopment? Scatter plot of log GDP per capita by latitude QAT KWT LUX ARE NOR OMN HKG SAU USA CHE BHR AUS AUT BEL DEU NLD IRL DNK CANSWE JPN ESP ITA FRA GBR FINISL LBY ISR KORGRCNZL MLT PRT SVN CZE SVK HUN EST HRV KAZ POL LTU RUS GAB VEN AZE BGR BLR BRA BRB IRN ROM LVA SUR PAN URY TUR CRI THA BWA DZA IRQ DOM EGY JOR MKD TUN TKM BLZ BIH JAM NAM LKA SLVGTM UKR PRY BTN MAR ARM GEO NGA BOL NID VNM IND UZB MDA STP CIV DJI ZMB KGZ KEN PNG SEN BGD LSO TJK COM BFA MDG ZWE UGA RWA TGO ETH ERI NERMOZ BDI LBR MWI ZAR So, is that, like, um, large or small? On average, how large of a change in log GDP per capita is associated with each 1 degree increase absolute latitude? Could differences in distance from the equator (i.e. tropicality) explain most of the observed variation income per capita? Lecture4,Slide4
Does Latitude Explain Underdevelopment? Average GDP per capita 500 1,000 10,000 100,000 Bar graph of GDP per capita by absolute latitude (distance from the equator So, is that, like, um, big or small? On average, how large of a change in log GDP per capita is associated with each 1 degree increase absolute latitude? Could differences in distance from the equator (i.e. tropicality) explain most of the observed variation income per capita? Lecture4,Slide5 Does Latitude Explain Underdevelopment? Linear regression of log GDP per capita on latitude QAT KWT LUX ARE NOR OMN HKG SAU USA CHE BHR AUS AUT BEL DEU NLD IRL DNK CANSWE JPN ESP ITA FRA GBR FINISL LBY ISR KORGRCNZL MLT PRT SVN CZE SVK HUN EST HRV KAZ POL LTU RUS GAB VEN AZE BGR BLR BRA BRB IRN ROM LVA SUR PAN URY TUR CRI THA BWA DZA IRQ DOM EGY JOR MKD TUN TKM BLZ BIH JAM NAM LKA SLVGTM UKR PRY BTN MAR ARM GEO NGA BOL NID VNM IND UZB MDA STP CIV DJI ZMB KGZ KEN PNG SEN BGD LSO TJK COM BFA MDG ZWE UGA RWA TGO ETH ERI NERMOZ BDI LBR MWI ZAR So, is that, like, um, big or small? On average, how large of a change in log GDP per capita is associated with each 1 degree increase absolute latitude? Could differences in distance from the equator (i.e. tropicality) explain most of the observed variation income per capita? Lecture4,Slide6
Simple Linear Regression Linear regression: fitting a line to the data Dependent variable: log GDP per capita in 2010 (on y-axis) Independent variable: absolute latitude (on x-axis) We regress the dependent variable on the independent variable On average, how large of a change in the dependent variable is associated with each 1 unit increase in the independent variable? We are assuming that the relationship is linear: Each 1 unit increase in the independent variable is associated with the same size change in the average value of the dependent variable Lecture4,Slide7 What Is a Line? A set of points that satisfy the equation: y = a + b x a is the intercept b is the slope The intercept tells us the value of the dependent variable where the line cross the (horizontal) y-axis in other words, where the line starts The slope tells us how big of a change in the dependent variable we see (on the line) for every one unit increase in the independent variable We are usually more interested in the slope than the intercept Lecture4,Slide8
What Is a Line? An example of a line: y =1+ 1 2 x Lecture4,Slide9 Linear Regression: Fitting a Line to Data All lines take the form: y = a + b x A one unit increase in x is associate with a unit change in y? The line we are interested in: predicted GDP per capita } {{ } y, the dep. var. = a + b absolute latitude } {{ } x, the ind. var. We will use the data in the scatter plot to estimate this line By estimate this line we mean figure out what a and b should be The estimated value of b (the slope of the line) will tell us: a 1 unit increase in latitude is associated how big of a change in log GDP? Lecture4,Slide12
Linear Regression: Fitting a Line to Data Linear regression of log GDP per capita on latitude STP PRY HRV SWE It easy if all countries were like Sao Tome & Principe, Paraguay, Croatia, and Sweden... Lecture4,Slide13 Linear Regression: Fitting a Line to Data The regression line gives us a predicted relationship Linear regression of log GDP per capita on latitude absolute latitude = 15 BRB THA SLVGTM NIC HND VNM ZMB SEN GMB BFA ERI MWI NER On average, countries that have absolute latitudes of about 15 degrees have values of log GDP per capita around a + b 15 Lecture4,Slide14
Linear Regression: Fitting a Line to Data The difference between the actual value of the dependent variable and the average value predicted by the linear model is called the residual Lecture4,Slide15 Are GDP per Capita and Latitude Related? Each data point (country) is associated with its own residual We observe a value of the latitude variable for each country Gives us a regression prediction of that country s log GDP per capita Very few countries will have actual levels of log GDP per capita that line up precisely with the prediction of the regression model The line estimated through simple linear regression: Minimizes the sum of the squares of the residuals Lecture4,Slide17
Are GDP per Capita and Latitude Related? Linear regression of log GDP per capita on latitude QAT KWT LUX ARE NOR OMN HKG SAU USA CHE BHR AUS AUT BEL DEU NLD IRL DNK CANSWE JPN ESP ITA FRA GBR FINISL LBY ISR KORGRCNZL MLT PRT SVN CZE SVK HUN EST HRV KAZ POL LTU RUS GAB VEN AZE BGR BLR BRA BRB IRN ROM LVA SUR PAN URY TUR CRI THA BWA DZA IRQ DOM EGY JOR MKD TUN TKM BLZ BIH JAM NAM LKA SLVGTM UKR PRY BTN MAR ARM GEO NGA BOL NID VNM IND UZB MDA STP CIV DJI ZMB KGZ KEN PNG SEN BGD LSO TJK COM BFA MDG ZWE UGA RWA TGO ETH ERI NERMOZ BDI LBR MWI ZAR Estimated slope coefficient: 0.0424 Lecture4,Slide18 Are GDP per Capita and Latitude Related? Linear regression of log GDP per capita on latitude QAT KWT LUX ARE NOR OMN HKG SAU USA CHE BHR AUS AUT BEL DEU NLD IRL DNK CANSWE JPN ESP ITA FRA GBR FINISL LBY ISR KORGRCNZL MLT PRT SVN CZE SVK HUN EST HRV KAZ POL LTU RUS GAB VEN AZE BGR BLR BRA BRB IRN ROM LVA SUR PAN URY TUR CRI THA BWA DZA IRQ DOM EGY JOR MKD TUN TKM BLZ BIH JAM NAM LKA SLVGTM UKR PRY BTN MAR ARM GEO NGA BOL NID VNM IND UZB MDA STP CIV DJI ZMB KGZ KEN PNG SEN BGD LSO TJK COM BFA MDG ZWE UGA RWA TGO ETH ERI NERMOZ BDI LBR MWI ZAR E(log GDP) = 7.915 + 0.0423 latitude Latitude Constant Dep. Var. = Log GDP OLS (1) 0.0423 (standard error of b) 7.915 (standard error of a) Lecture4,Slide21
Are GDP per Capita and Latitude Related? Regression results: E(log GDP per capita) = 7.915 + 0.0423 latitude Interpretation: Location Latitude E(Log GDP) E(GDP) Actual GDP Equator 0 7.915 $ 2,738 Kinshasa, DRC 4.4 8.098 $ 3,287 $ 619 Tropic of Cancer 23.4 8.907 $ 7,381 Washington DC 38.9 9.564 $ 14,242 $ 48.357 Oslo, Norway 59.6 10.456 $ 34,755 $ 57,739 explains about 1 3 of the observed variation in income Lecture4,Slide22 Does Geography Explain Underdevelopment? Positive correlation between distance from the equator (absolute latitude) and development (specifically, income per capita and child mortality) Greater distance from equator Higher GDP per capita Lecture4,Slide23
ZAR GAB STP KEN UGA RWA BDI NGA CPV CIV ZMB SEN COM BFA MDG ZWE TGO ETH ERI NER MOZ LBR MWI BWA NAM LSO PNG LKA PHL THA VNM IND HKG BGD AUS BTN NZL SUR VEN PAN BRA CRI BRB DOM BLZ JAM SLVGTM BOL NIC HND PRY URY Are GDP per Capita and Latitude Related? Sub-Saharan Africa South & East Asia Latin America, etc JPN KOR 0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50 Regression results: Regression results: Regression results: intercept = 7.229 intercept = 9.092 intercept = 9.053 slope = 0.038 slope = 0.005 slope = 0.001 Lecture4,Slide25 Study Guide: Key Terms linear regression independent variable dependent variable intercept slope linear prediction residual Lecture4,Slide26