Advanced Economic Growth: Lecture 1, Introduction

Advanced Economic Growth: Lecture 1, Introduction Daron Acemoglu MIT September 5, 2007. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 1 / 48

0.00005.0001.00015.0002.00025 Density of coutries Cross-Country Income Di erences (1) There are very large di erences in income per capita and output per worker across countries today. 1960 1980 2000 0 10000 20000 30000 40000 50000 gdp per capita Figure: Estimates of the distribution of countries according to PPP-adjusted GDP per capita in 1960, 1980 and 2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 2 / 48

Cross-Country Income Di erences (2) Part of the spreading out of the distribution in the gure is because of the increase in average incomes. It is more natural to look at the log of income per capita when growth is approximately proportional I I when x (t) grows at a proportional rate, log x (t) grows linearly, that is, if x 1 (t) and x 2 (t) both grow by 10% over a certain period of time, x 1 (t) x 2 (t) will also grow, while log x 1 (t) log x 2 (t) will remain constant. The next gure shows a similar pattern, but now the spreading-out is more limited. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 3 / 48

0.1.2.3.4 Density of coutries Cross-Country Income Di erences (3) 1960 1980 2000 6 7 8 9 10 11 log gdp per capita Figure: Estimates of the distribution of countries according to log GDP per capita (PPP-adjusted) in 1960, 1980 and 2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 4 / 48

0 1.000e+09 2.000e+09 3.000e+09 Density of coutries weighted by population Cross-Country Income Di erences (4) Inequality among nations, or inequality among individuals? 1980 2000 1960 6 7 8 9 10 11 log gdp per capita Figure: Estimates of the population-weighted distribution of countries according to log GDP per capita (PPP-adjusted) in 1960, 1980 and 2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 5 / 48

0.1.2.3.4 Density of coutries Cross-Country Income Di erences (5) Theory is easier to map to data when we look at output (GDP) per worker. Moreover, key sources of di erence in economic performance across countries are national policies and institutions. 1960 1980 2000 6 8 10 12 log gdp per worker Figure: Estimates of the distribution of countries according to log GDP per worker (PPP-adjusted) in 1960, 1980 and 2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 6 / 48

Income, Growth and Welfare Overall, two important facts: 1 there is a large amount of inequality in income per capita and income per worker across countries; 2 there is a slight but noticeable increase in inequality across nations (though not necessarily across individuals in the world economy). Why care about income di erences? 1 Welfare. 2 Understanding the structure and e ciency of production and market mechanisms. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 7 / 48

5 6 7 8 9 10 log consumption per capita 2000 Income and Welfare (1) GBRISL AUS AUTDNK HKG NOR BRB FRA GER FIN NLD J PN IRL CHE CAN ESP NZLITABEL MGRC US SWE ISR TTO CZEPRT SVN MAC ARG KOR URY SYC HUN KNA GAB MEXCHL POL SVK EST ATG TUN LTU LCA BLZKAZ ZAF LVAHRV TUR RUS BRA MKD LBN BLR EGY ROM SLV BGR GEO COL IRN PAN GRD VENDMA GTMPRYSWZ DOMCRI THA VCT MAR PER ALBCPV ARM GIN GNQ DZA MYS IDN UKR LKA PHLJ SYR OR BOL M DA AZE ECU KGZ J NIC ZWE CHN AM CMR SEN CIV PAK HND COM BGD IND BEN GMB GHA M OZ COG KENNPL TJK STP LSO UGA MRWA DG MWI TGO NER ZMB TCD BFA MLI GNB BDI ETH YEM TZA USA LUX NGA 6 7 8 9 10 11 log gdp per capita 2000 Figure: The association between income per capita and consumption per capita in 2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 8 / 48

40 50 60 70 80 90 life expectancy 2000 Income and Welfare (2) J PN CHE HKG AUS CAN CRI GRCISR ESP ITAISL SWE FRA MAC NZLGBR AUT BEL FIN NLDNOR CHL IRL USA PRT DNK CZE SVN KOR BRB LKA PAN HRV MEX URY ALB ARG ECU MKD BLZ J AM ARM GEO COL BGR LBN LCA POL SVK SYR VEN TUN MYS LTU HUN TTO CHN EST M US J OROM SLV PRY VCT CPV DZA LVA BRA NIC PHL THA MAR EGY PER IRNTUR BLR HND M DA AZE UKRDOM KGZ RUS IDN GTM TJK KAZ INDBOL COM PAK NPLBGD ZAF YEM GAB GHA TGO SEN M DG BEN GMB GIN KEN COG TZA LSO MLI CIV CMR ETH BFA GNQ GNB NGA TCD M OZ SWZ ZWE BDI NER UGA MWI LUX ZMB RWA 6 7 8 9 10 11 log gdp per capita 2000 Figure: The association between income per capita and life expectancy at birth in 2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 9 / 48

Caveats Understanding how some countries can be so rich while some others are so poor a major question for economics and social sciences in general. But, income per capita not a su cient statistic for the welfare of the average citizen. I I Economic growth is generally good for welfare but it often creates winners and losers. (Joseph Schumpeter s creative destruction; Simon Kuznet s structural transformations; political economy analyses of economic growth). A stark illustration: South Africa under Apartheid. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 10 / 48

0 5 10 15 20 Density of coutries Postwar Growth Patterns 1960 1980 2000.1.05 0.05.1 average growth rates Figure: Estimates of the distribution of countries according to the growth rate of GDP per worker (PPP-adjusted) in 1960, 1980 and 2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 11 / 48

6 7 8 9 10 log gdp per capita Economic Growth in Selected Countries USA UK Spain Brazil South Korea Singapore Guatemala Botswana India Nigeria 1960 1970 1980 1990 2000 year Figure: The evolution of income per capita in selected countries, 1960-2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 12 / 48

Economic Growth and Income Di erences Why is the United States richer in 1960 than other nations and able to grow at a steady pace thereafter? How did Singapore, South Korea and Botswana manage to grow at a relatively rapid pace for 40 years? Why did Spain grow relatively rapidly for about 20 years, but then slow down? Why did Brazil and Guatemala stagnate during the 1980s? What is responsible for the disastrous growth performance of Nigeria? Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 13 / 48

.6.7.8.9 1 1.1 log GDP per worker relative to the US in 2000 Origins of Income Di erences and World Growth (1) Growth responsible for current cross-country income di erences. But postwar growth by itself is not. LUX AUT BEL NOR USA SGP IRL FRA NLD HKG ITAISR CHE DNK AUS CAN FIN ISL GBR J PN ESP SWE NZL PRT M US GRC TTO KOR CHL BRB M YS ARG GAB URY CRI PAN MZAF EX IRN VEN GNQ DZA DOMBRA COL EGY CPV PRY TUR J OR THA ROM ECU GTM SLV PER LKA M AR PHL J AM NIC IDN ZWE CHN PAK BOL IND SYR CMR HND CIV GIN LSO SEN COG GHA NPL BEN NGA COM KEN MLI BFA M OZ UGA GMB TCD ETH MWI RWA TGO ZMB GNB NER M DG TZA BDI.6.7.8.9 1 log GDP per w ork er relativ e to the U S in 1960 Figure: Log GDP per worker in 2000 versus log GDP per worker in 1960. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 14 / 48

6 7 8 9 10 log gdp per capita Origins of Income Di erences and World Growth (2) Western Offshoots Western Europe Asia Latin America Africa 1800 1850 1900 1950 2000 year Figure: The evolution of average GDP per capita in Western O shoots, Western Europe, Latin America, Asia and Africa, 1820-2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 15 / 48

6 7 8 9 10 log gdp per capita Origins of Income Di erences and World Growth (3) Western Offshoots Western Europe Asia Latin America Africa 1000 1200 1400 1600 1800 2000 year Figure: The evolution of average GDP per capita in Western O shoots, Western Europe, Latin America, Asia and Africa, 1000-2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 16 / 48

6 7 8 9 10 log gdp per capita Long-Run Growth in Selected Countries USA Spain Britain China Brazil India Ghana 1800 1850 1900 1950 2000 year Figure: The evolution of income per capita in the United States, Britain, Spain, Brazil, China, India and Ghana, 1820-2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 17 / 48

.02 0.02.04.06 annual growth rate 1960 2000 No Unconditional Convergence Consistent with the post-war patterns presented so far. KOR HKG TZA MWI GNB BDI COG CHN UGA BFA ETH LSO NPL ROM IDN PAK IND GMB KENGHA BEN TGO RWA THA CPV LKA BGD ZWE CIV CMR SEN COM M M DG OZTCD ZMB MLI NER NGA SYC BRB PRT ESP LUX GRC AUT ITA SYR TURGAB FIN EGYDOM ISR BEL BRA FRA M AR NOR PAN IRN CHL ISL TTO GBRDNK USA SWE NLD AUS CAN ECU GTM J OR M EX PHL GIN HND J PN MMYSUS J AM BOL PRY COL NIC SLV CRI PER IRL ZAF URY ARG VEN CHE NZL 6 7 8 9 10 log gdp per worker 1960 Figure: Annual growth rate of GDP per worker between 1960 and 2000 versus log GDP per worker in 1960 for the entire world. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 18 / 48

.01.02.03.04 annual growth rate 1960 2000 But a Di erent Picture Among Relatively Similar Countries Convergence among (original) OECD countries. J PN IRL PRT GRC ESP AUT ITA FIN LUX FRA BEL NOR ISL GBR DNK USA NLD SWE AUS CAN CHE 9 9.5 10 10.5 log gdp per worker 1960 NZL Figure: Annual growth rate of GDP per worker between 1960 and 2000 versus log GDP per worker in 1960 for core OECD countries. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 19 / 48

Conditional Convergence Barro (1991): focus on conditional convergence : the income gap between countries that share the same characteristics. Consider a typical Barro growth regression : g t,t 1 = β ln y t 1 + X 0 t 1α + ε t (1) Useful for describing the data, but not for estimating causal e ects of the variables in X t 1. These variables typically correlated with growth, but not necessarily exogenous (e.g., investment, human capital, life expectancy,...). Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 20 / 48

0.02.04.06.08 Average growth rate of GDP per capita 1960 2000 Correlates of Economic Growth (1) TWN CHN KOR URY ZWE NGA INDISL BRA GHA TTO TUR COL PHL VENKEN LKA LUX ISRNORPAN ITA AUT DOM FIN MUS BEL EGY GRC PAK FRA MAR USA NLD CAN DNK AUS SWE CHL GBR ETH ARG BEN MWI NZL ZAF GIN.04.02 0.02.04 Average growth of investment ratio 1960 2000 J AM UGA ZMB THA CRI GTM NIC ECU PRY BFA J OR BOL ESP CHE MYS PRT MEX PER HND IRL SLV IRN J PN Figure: The relationship between average growth of GDP per capita and average growth of investments to GDP ratio, 1960-2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 21 / 48

.02 0.02.04.06 average growth gdp per capita 1960 2000 Correlates of Economic Growth (2) KOR HKG PAK IRN NPL DZA MWI UGAGTM BGD KEN GHA GMB CMR BEN MLI BDI RWA TGO SEN MOZ NER THA CHN J PN IRL BRB PRT MUS MYS COG IDN ESP GRC ITA AUT FIN IND BRA DOM ISL BEL ISR NOR SYR EGY FRA TUR TTO PANCHL NLD CAN LKA DNK AUS LSO GBR SWE ZWE COL MEX PRY ECU CHE J OR CRI PHL URY ZAF PER ARG SLV J AM HND BOL VEN ZMB NIC USA NZL 0 2 4 6 8 10 average schooling 1960 2000 Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 22 / 48

From Correlates to Fundamental Causes (1) The correlates of economic growth, such as physical capital, human capital and technology, are the rst topic of study in economic growth analyses. But these are only proximate causes of economic growth and economic success. Why do certain societies fail to improve their technologies, invest more in physical capital, and accumulate more human capital? I I I how did South Korea and Singapore manage to grow, while Nigeria failed to take advantage of the growth opportunities? If physical capital accumulation is so important, why did Nigeria not invest more in physical capital? If education is so important, why are education levels in Nigeria still so low and why is existing human capital not being used more e ectively? The answer to these questions is related to the fundamental causes of economic growth. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 23 / 48

From Correlates to Fundamental Causes (2) We can think of the following list of potential causes: 1 luck (or multiple equilibria) 2 geographic di erences 3 institutional di erences (political economy) 4 cultural di erences Important to link di erent approaches to the process of economic growth to possible fundamental causes of long-run development. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 24 / 48

Cross-Country Income Di erences: Regressions (1) For a better understanding of proximate causes, let us consider an extended Solow model with human capital (see Chapter 3 of the book). Assume that country j = 1,..., N has the aggregate production function Y j (t) = K j (t) β H j (t) α (A j (t) L j (t)) 1 α β. Notice the somewhat unusual form of the production function (human capital as a separate factor of production; is this reasonable?) Countries di er in terms of their investment (saving) rates in physical and human capital, s k,j and s h,j, population growth rates, n j, and technology growth rates Ȧ j (t) /A j (t) = g j. As usual, de ne k j K j /A j L j and h j H j /A j L j. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 25 / 48

Cross-Country Income Di erences: Regressions (1) Unique steady state for each country, with physical and human capital to e ective labor ratios k j = h j = s k,j n j + g j + δ k s k,j n j + g j + δ k 1 β α s h,j n j + g j + δ h s h,j n j + g j + δ h α! 1 1 α β 1 β! 1 1 α β. Therefore, the balanced growth path of income for country j = 1,..., N can be expressed as: ln yj β (t) = ln Ā j + gt + 1 α β ln s k,j n j + g + δ k α + 1 α β ln s h,j. n j + g + δ h (2) Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 26 / 48

Cross-Country Income Di erences: Regressions (2) How to take this equation to data? Mankiw, Romer and Weil (1992) assume: Orthogonal technology assumption Ā j = ε j A, with ε j orthogonal to all other variables. And they take: I δ k = δ h = δ and δ + g = 0.05. I s k,j =average investment rates (investments/gdp). I s h,j =fraction of the school-age population that is enrolled in secondary school. MRW rst estimate equation (2) without the human capital term ln y j = constant + β 1 β ln (s k,j ) β 1 β ln (n j + g + δ k ) + ε j. Then, the full version with human capital. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 27 / 48

Cross-Country Income Di erences: Regressions (3) Estimates of the Basic Solow Model MRW Updated data 1985 1985 2000 ln(s k ) 1.42 1.01 1.22 (.14) (.11) (.13) ln(n + g + δ) -1.97-1.12-1.31 (.56) (.55) (.36) Adj R 2.59.49.49 Implied β.59.50.55 No. of observations 98 98 107 Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 28 / 48

Cross-Country Income Di erences: Regressions (4) Their estimates for β/ (1 β), implies that β must be around 2/3, but should be around 1/3. The most natural reason for the high implied values of β is that ε j is correlated with ln (s k,j ), I I either because the orthogonal technology assumption is not a good approximation to reality or because there are also human capital di erences correlated with ln s k,j so that there is an omitted variable bias. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 29 / 48

Implied β.30.31.36 Implied α.28.22.26 No. of observations 98 98 107 Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 30 / 48 Cross-Country Income Di erences: Regressions (5) Estimates of the Augmented Solow Model MRW Updated data 1985 1985 2000 ln(s k ).69.65.96 (.13) (.11) (.13) ln(n + g + δ) -1.73-1.02-1.06 (.41) (.45) (.33) ln(s h ).66.47.70 (.07) (.07) (.13) Adj R 2.78.65.60

Cross-Country Income Di erences: Regressions If these regression results are reliable, they give a big boost to the augmented Solow model. The Adjusted R 2 suggests that over (or close to) three quarters of income per capita di erences across countries can be explained by di erences in their physical and human capital investment behavior. The immediate implication is that technology (TFP) di erences have a somewhat limited role, con ned to at most accounting for about a quarter of the cross-country income per capita di erences. But... Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 31 / 48

Challenges to the Regression Analyses I 1 Technology di erences across countries are not orthogonal to all other variables. Ā j is correlated with measures of s h j and s k j for two reasons. 1 omitted variable bias problem: societies with high levels of Ā j will be those that have invested more in technology for various reasons; it is then natural to expect the same reasons to induce greater investment in physical and human capital as well. 2 reverse causality problem; complementarity between technology and physical or human capital imply that countries with high Ā j will nd it more bene cial to increase their stock of human and physical capital. In terms of the regression above, this implies that the key right-hand side variables are correlated with the error term, ε j. Consequently, OLS estimates of α and β and R 2 are biased upwards. 2 α is too large relative to what we should expect on the basis of microeconometric evidence. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 32 / 48

Challenges to the Regression Analyses II The working age population enrolled in school ranges from 0.4% to over 12% in the sample of countries. The predicted log di erence in incomes between these two countries is α (ln 12 ln (0.4)) = 0.66 (ln 12 ln (0.4)) 2.24. 1 α β Thus a country with schooling investment of over 12 should be about exp (2.24) 1 8.5 times richer than a country with a level of schooling investment of around 0.4. This gap is too large in view of the micro evidence on the returns to human capital. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 33 / 48

Returns to Human Capital and Cross-Country Evidence I How to go from micro returns to human capital to cross country evidence? Take Mincer regressions of the form: ln w i = X 0 i γ + φs i, (3) Micro evidence suggests φ is between 0.06 and 0.10 In practice, the di erence in average years of schooling between any two countries in the Mankiw-Romer-Weil sample is less than 12. Can we deduce from this information how much richer a country with 12 more years of average schooling should be? The answer is yes, but we need to assume: 1 That the micro-level relationship as captured by (3) applies identically to all countries. 2 That there are no human capital externalities. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 34 / 48

Returns to Human Capital and Cross-Country Evidence II Suppose that each rm f in country j has access to the production function y fj = Kf 1 α (A j H f ) α, Suppose also that rms in this country face a cost of capital equal to R j. With perfectly competitive factor markets, α Kf R j = (1 α). (4) A j H f This implies that all rms ought to function at the same physical to human capital ratio, and consequently, all workers, irrespective of their level of schooling, ought to work at the same physical to human capital ratio. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 35 / 48

Returns to Human Capital and Cross-Country Evidence III Another direct implication of competitive labor markets is that in country j, wages per unit of human capital will be equal to w j = α (1 α) (1 α)/α A j R (1 α)/α j. Consequently, a worker with human capital h i will receive a wage income of w j h i. Substituting for capital from (4), total income in country j is (1 α)/α Y j = (1 α) (1 α)/α R j A j H j, where H j is the total e ciency units of labor in country j. In view of this, a country with 12 more years of average schooling should have a stock of human capital somewhere between exp (0.10 12) ' 3.3 and exp (0.06 12) ' 2.05 times greater and thus have income per capita about twice or three times greater. Much lower than the over eightfold di erences implied by the regression analysis. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 36 / 48

Calibrating Productivity Di erences I Use the same Mincer approach for calibration. Suppose that each country has access to the Cobb-Douglas aggregate production function: Y j = Kj 1 α (A j H j ) α. (5) Notice that this is a more conventional production function, with e ciency units of labor as a factor of production. Suppose that each worker in country j has S j years of schooling. Then using the Mincer equation (3) ignoring the other covariates and taking exponents, H j can be estimated as H j = exp (φs j ) L j, Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 37 / 48

Calibrating Productivity Di erences II This approach however does not take into account di erences in other human capital factors, such as experience, in the quality of schooling and the amount of post-schooling human capital, and in the rate of return to schooling. Let the rate of return to acquiring the Sth year of schooling be φ (S). A somewhat better estimate of the stock of human capital can be constructed as H j = exp fφ (S) Sg L j (S) S where L j (S) now refers to the total employment of workers with S years of schooling in country j. A series for K j can be constructed from Summers-Heston dataset using investment data and the perpetual invented method. K j (t + 1) = (1 δ) K j (t) + I j (t), Let us assume, following Hall and Jones that δ = 0.06. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 38 / 48

Calibrating Productivity Di erences III Finally, with the same arguments as before, we choose a value of 2/3 for α. Given series for H j and K j and a value for α, we can construct predicted incomes at a point in time using the following equation Ŷ j = K 1/3 j (A US H j ) 2/3 for each country j, where A US is computed so that this equation ts the United States perfectly, i.e., Y US = K 1/3 US (A US H US ) 2/3. Once a series for Ŷ j has been constructed, it can be compared to the actual output series. The gap between the two series represents the contribution of technology. Alternatively, we could explicitly back out country-speci c technology terms (relative to the United States) as A 3/2 1/2 j Yj KUS HUS =. A US Y US K j H j Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 39 / 48

Predicted log gdp per worker 1980 7 8 9 10 11 Predicted log gdp per worker 1990 7 8 9 10 11 Predicted log gdp per worker 2000 7 8 9 10 11 Calibrating Productivity Di erences IV 45 45 45 NZL CHE NOR DNK AUS I SR SWE FIN HUN JPN AUT CANLD USA BEL CYP G G BR IRCSL FRA G UY KOR BRB I ARG RL ESP I TA PER HKG JAM ECU FJI PAN VEN ZMB SGP THA ZWE PHL CHL URY ZAF COG M US TTO M CRI BRA BOL I YS PRT M EX RN M WI CHN LKA G HA DOM TUR COL KEN NICTUN I ND SYR PRY HND SLV PAK BWA BGD JOR SEN LSO I DN PNG G TM NER EGY NPL TGO CMR BEN M LI CAF NOR NZL DNK CHE JPN GAUS FIN CAN ER SWEUSA HUN ARG I SR G RCI AUT SLNLD KOR BEL CYP G BR FRA BRB G UY PER HKG ISGP RL ESPI TA FJI PAN PHL THA CHL VEN URY M EX JAM CRI M YS TTO ZMB I RN BRA PRT CHN COG ZWE JOR ZAF LKANIC DOM LSO BOL SYR PRY M US BWA TUR COL TUN M WI I ND HND I DN KEN PNG SLV CMR EGY BGDPAK NPL G TM M RT BEN SEN TGO M LI CAF BDI GMB UGAMOZ SLE NOR JPNCHE KOR NZLSWE GCAN FIN AUS ER DNK USA AUT HUN G I SR RC I SL FRA NLD HKG BEL G BR PAN ESP ARG I TA PERTHA M YS ECU CHL I RL JAMPHL M EX LSO VEN JOR BRB PRT ZAF URY ZMB CRI CHN DZA BRA TTO ZWE TUR NIC LKA I DN PRY I RN HND BOL COL COG SYR DOM TUN M US I ND KEN SLV TGO G HA NPL CMR PAK EGY M WI BGD G TM BEN G SEN MB M LI NER RWA UGA MOZ UGA MOZ G MB RWA SLE 7 8 9 10 11 log gdp per worker 1980 7 8 9 10 11 log gdp per worker 1990 7 8 9 10 11 12 log gdp per worker 2000 Figure: Calibrated technology levels relative to the US technology (from the Solow growth model with human capital) versus log GDP per worker, 1980, 1990 and 2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 40 / 48

Predicted relative technology 0.5 1 1.5 Predicted relative technology 0.5 1 1.5 Predicted relative technology 0.5 1 1.5 Calibrating Productivity Di erences V I RL JOR I TA I TA G TM M EX ESP M US I TA FRA PRT BEL USA I SLNLD SGP ARG SLV CHE PRY BRAZAF URY I RLG RCAN COL AUT TUN TTO G BRSWE AUS DNK EGY HKG FIN SYR BRB CRI VEN NICI RN NOR NZL DOM CHL J PN TUR PAN SLE M YS I SR HND BWA FJI M US CMR ECU PER PNG CYP GMB BOL CAF KOR RWA HUN PHL MOZ M LI I PAK DN TGOBGD LKA ZWE J AM BEN LSO G UY NPL SEN THA UGA KEN I ND G HA NER CHN COGZMB MWI ESP BEL PRT USA FRA ISGP RL NLD G TM AUT CHE G IBR SL ZAF HKG CAN SWE AUS FIN G ER DNK EGY TUNM US M EX SLV COL BRB G RC J PN PRY CYP TUR BRA TTO I SR NZL URY JOR VEN KOR NOR M YS SYR CHL SLE BWA I RN ARG DOMCRI HUN PAK HND FJI PAN I DN PNG BOL BGD NIC MOZ GMB CMR J AMTHA PER LKA CAF PHL I ND ZWE UGA SEN TGO MNPL RT BEN KEN M BDI LI LSO G UY CHN COG ZMB MWI BEL USA PRT HKG BRB ESP NLD FRA AUT G BR I SL AUS DNK FIN G TM CAN TUN G ER EGY SWE TTO CHE SLV GI SR RC DOM M EX I BRA RNZAF URY NZL KOR J PN NOR CHL MARG YS TUR SYR HUN JOR BGD PAK COLVEN PRY CRI DZA PAN I ND LKA I DN THA MOZ HND BOL UGA CMR PHL ECU PER GMB SEN CHN M LI BEN NPL NICJAM NER RWA ZWE G HA KEN MWI TGOCOG LSO ZMB 7 8 9 10 11 log gdp per worker 1980 7 8 9 10 11 log gdp per worker 1990 7 8 9 10 11 12 log gdp per worker 2000 Figure: Calibrated technology levels relative to the US technology (from the Solow growth model with human capital) versus log GDP per worker, 1980, 1990 and 2000. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 41 / 48

Calibrating Productivity Di erences VI The following features are noteworthy: 1 Di erences in physical and human capital still matter a lot. 2 However, di erently from the regression analysis, this exercise also shows that there are signi cant technology (productivity) di erences. 3 The same pattern is visible in the next three gures, which plot, the estimates of the technology di erences, A j /A US, against log GDP per capita in the corresponding year. 4 Also interesting is the pattern that the empirical t of the neoclassical growth model seems to deteriorate over time. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 42 / 48

Challenges to Callibration I In addition to the standard assumptions of competitive factor markets, we had to assume no human capital externalities, a Cobb-Douglas production function, and also make a range of approximations to measure cross-country di erences in the stocks of physical and human capital. The calibration approach is in fact a close cousin of the growth-accounting exercise (it is sometimes referred to as levels accounting ) and can be done in a more general way as in growth-accounting exercises. Imagine that the production function that applies to all countries in the world is given by F (K j, H j, A j ), and countries di er according to their physical and human capital as well as technology but not according to F. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 43 / 48

Challenges to Callibration II Let us a rank countries in descending order according to their physical capital to human capital ratios, K j /H j Then we can write ˆx j,j+1 = g j,j+1 ᾱ K,j,j+1 g K,j,j+1 ᾱ Lj,j+1 g H,j,j+1, (6) where g j,j+1 is the proportional di erence in output between countries j and j + 1, g K,j,j+1 is the proportional di erence in capital stock between these countries and g H,j,j+1 is the proportional di erence in human capital stocks. In addition, ᾱ K,j,j+1 and ᾱ Lj,j+1 are the average capital and labor shares between the two countries. The estimate ˆx j,j+1 is then the proportional TFP di erence between the two countries. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 44 / 48

Challenges to Callibration This levels-accounting exercise faces two challenges. One is data-related and the other one theoretical: 1 Data on capital and labor shares across countries are not widely available. Almost all calibration or levels-accounting exercises that estimate technology (productivity) di erences use the Cobb-Douglas approach with a constant value of α K equal to 1/3. 2 The di erences in factor proportions, e.g., di erences in K j /H j, across countries are large. An equation like (6) is a good approximation when we consider small (in nitesimal) changes. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 45 / 48

Summary Regression and calibration analyses suggest that TFP di erences are important in accounting for cross-country income di erences. These are not necessarily pure technology di erences, but di erences in technology broadly construed, including di erences in the e ciency of production (e.g., due to market failures). Other evidence, for example, estimating productivity di erences using trade data, consistent with this conclusion (see Chapter 3). We will therefore pay special attention to models generating technology/tfp di erences across countries. I Endogenous technology. I Market failures and di erences in the e ciency of production. Also look for fundamental causes that can lead to technology di erences. But, it is useful to bear in mind that human and physical capital di erences also important and we should look for approaches that can account for these di erences as well. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 46 / 48

The Rest of the Course (1) Develop a range of models and di erent theoretical approaches useful for answering the set of questions raised here. Three areas of emphasis: 1 Understanding technology di erences. 2 Investigation of the process of economic development and structural transformation. 3 Linking proximate causes to fundamental causes (e.g., political economy). Also, additional important topics related to growth: I I Growth and the environment. The role of policy. Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 47 / 48

The Rest of the Course (2) The next two lectures: review of basic models of endogenous technology. Recitation this week: review of optimal control. The rest of the course divided between me and Philippe Aghion focusing on various topics within this agenda (see syllabus). Daron Acemoglu (MIT) Advanced Growth Lecture 1 September 5, 2007. 48 / 48