Development Economics ECON 570
Popular in Course
Popular in Economcs
This 0 page Class Notes was uploaded by Miss Romaine Grimes on Sunday November 1, 2015. The Class Notes belongs to ECON 570 at Pennsylvania State University taught by Staff in Fall. Since its upload, it has received 7 views. For similar materials see /class/233105/econ-570-pennsylvania-state-university in Economcs at Pennsylvania State University.
Reviews for Development Economics
Report this Material
What is Karma?
Karma is the currency of StudySoup.
Date Created: 11/01/15
Notes on Investment and Investment Prices Econ 570 Spring 2005 1 Introduction A very robust nding is that the investment rate is highly correlated with income 0 Mankiw et al nd that differences in investment rates account for around half of the income disparity across countries using the Solow type model 0 Levine and Renelt nd the investment rate as the lone robust correlate with growth in income per person using the standard Barro type regression1 o DeLong and Summers nd that it is machinery investment that is key2 in quotIndia like in Argentina the savings is relatively high but equipment is expen sivelndia demonstrates not that boosting investment is unproductive but that policies that boost saving while simultaneously raising the relative price of invest ment in equipment and structures are unproductive We suspect that restrictions on imports of capital goods have ensured that the Indian government s attempts to support investment have had effects not on quantities but on prices lndia s policies have managed to enrich industrialists instead of encouraging industry quot o Richer countries invest more than poorer countries This is clear from gure 1 Notice that PPP investment rates are 2 3 times higher for US and Norway compared with poor 1A standard Barroetype regression is a crossecountry regression of pericapita income growth over some period Dyt on a host of variables Dyt Oz 51yy71 th71 711t 72 5t where h is educational attainment and the 35 s are various policy variables such as the investment ratio corruption openness in ation rate democracy rule of law etc 2De Long and Summers 1 396 stress the distinction quotbetween investment effort 7 share of national product saved plus capital in ows 7 and investment 7 buildings constructed and machines put into productive use Many of the policies that have been followed in the posteWWH period especially in the developing world seem designed to maximize 7investment effort7 while ensuring that each unit of 7investment effort7 translates into as little actual investment as possiblequot Like so many other aspect of economic policy what was merely a disease in developing countries was a pathology in the socialist wor Investment Ratios Spring 2004 countries like Mali and Kenya Thus both faster growing and richer countries invest more than slow growing and or poor countries 0 How does this t with the nding that TFP is what really matters What explains these nding 0 One set of arguments focuses on low savings rates and the F H hypothesis Classical low savings theories 0 Another set of arguments focuses on distortions to investment In particular it is pointed out that the relative price of investment is higher in poor countries This could be because of taxes or other distortions that make investment more expensive It could re ect trade policies barriers to entry prohibitions corruption and all sorts policy distortions that make investment more expensive the classic reference is to Argentina where the price of capital may have doubled in the Peron era 10 Price Distortions and the Lucas Puzzle Recall the Lucas puzzle 7 with common technologies and 04 4 India has a marginal product of capital that is 58 times that in the US So capital does not ow But there is another simple exercise that is informative lf capital is paid its marginal product then 04 FKg But this means that we can estimate the return to capital across countries by MPK a 1 where i indexes the country It is rather easy to get data on capital output ratios across country Before collecting it however we should also take into account the fact that output will be sold at H in country i and that the cost of capital goods is P1 Then we should examine the true holding return M PK de ned by P39 P Y JWPIC l4PIQ l 2 1 PI PIaKgti z Investment Ratios Spring 2004 Rather than measuring the capital output ratio at world prices we adjust each for domestic price levels and distortions This will be important as this ratio will differ dramatically across countries Suppose we do this what do we nd From 1 we can see that much of the conundrum goes away Without the domestic price adjustment we can already explain a lot Take 1ndia With a 13 the ratio of Ig 226 But using 1 we nd MPKIndm 055 compared with 037 for the US Essentially 1ndia uses less capital per unit of output than the US and this shows up in the ratio 1t is due to lower productivity of course When we do the calculation according to 2 then we have MPKIndm 032 compared with 046 for the US This shows up similarly for most countries The point being that distortions in domestic prices make the returns to capital across countries look normal3 Another way to look at this result is with regard to Argentina With the Lucas calculation its return to capital is 11 times that of the US At international prices it is about 20 higher than the US 043 versus 037 But correcting for distortions it is about 30 lower than in the US 035 versus 046 Hence correcting for the distortions we see why capital does not ow to Argentina very easily4 These price distortions suggest that at domestic prices poor countries are not investing so much less than rich ones The problem is that they get less investment for a unit of expenditure At domestic prices the correlation goes away Thus while the correlation between y and is 050 when measured at PPP the correlation between iy at domestic prices and y at PPP is only 005 1t is well known that investment rate differenHSL39502bmpe when measured at a common set of prices while very small when measured at domestic prices Evidently the domestic price of investment goods is high relative to consumption in poorer countries Why 1s this due to tariffs and taxes or is it more fundamental 3This should not be surprising since capitaleoutput ratios do not vary much across countries In the Solow model capital labor ratios differ in different steady states but capital output ratios vary less 7 one of Kaldor s stylized facts 4This may be the place to point out the importance of studying Argentina and Latin America for devele opment In 1913 Argentina s GDP pericapita was 80 of the OECD average By 1987 it was 32 This is quite dramatic and explaining why a country diverges must be as important as why it converges Investment Ramos Spnng 2mm mg quotH mm m u gure 1 Estimated Maxgnd Products of 0mm Investment Ratios Spring 2004 Hsieh and Klenow pursue this and show that investment goods are no more expensive in poor than rich prices The high relative price of capital is primarily driven by lower consumption goods prices It is not 5 that differs so much but g The argument is that in poor countries investment sectors have low productivity relative to their consumption sectors This is a corollary to the Balassa Samuelson hypothesis Tradable investment goods but not all tradable consumption goods Nontradable consumption goods in poor countries are cheaper than tradable investment goods That is what causes the relative price difference But what causes this productivity difference 2 Model The Hsieh Klenow model has two sectors and two tax rates a nontradable consumption sector a tradable investment sector a tax rate on importing investment goods 717 and a tax rate on capital income TKjO Otherwise it is a standard neoclassical growth model Each of Lj workers in inelastically supply one unit of labor per period They choose consumption to maximize A a 3 A subject to 191 1 7 Mg 113 4 chytc 139113 1111 PW 7K 1 e 63 K13 T13 5 and Rj 1 7 TKJ 73 60 7 719 P1 01 Bit WE 6 where R is the rental price of capital and T are transfers from the government and are rebated tax collections Expression 6 is just an expression for the rental cost of capital Rest is standard Investment Ratios Spring 2004 Key Assumption Consumption goods cannot be traded internationally but investment goods are tradable The pre tax world price of investment is taken as given PIW Then P17 717PIW 7 In each country rms rent capital and hire labor in competitive spot markets They sell output in competitive markets to maximize current pro ts given by H0 2 P070 7 ijCj i Rqu H1 2 PIWIj i ijIj i Pth because investment goods are tradable This can be rewritten as H1 2 PIjIj i TIjPIWIj i ijIj i RjKIj Production technologies in the two sectors are c ACnglecf I AIngL 1 8 10 11 where A17 and A07 are productivity indexes It is assumed that 04 is the same across countries and across sectors within countries The FOC s that come from this problem allow us to write K 0471 31 aPIWAI 1 7 and5 Pay A 7 1 PW ch 5Why are the capital labor ratios equal From the FOC for labor we have w 1 7 0413914016 1 7 commie c ail sci Q 153 Q f where the last equality comes from the FOC wrt to capital Hence7 k1 kg 6 1 a 6 12 Investment Ratios Spring 2004 which implies E A1739 13 P17 ch17391j Expression 12 equates the rental price of capital to the marginal product Expression 13 says that the relative price of consumption goods is inversely related to relative TEP in the two sectors and is decreasing in the tax rate on importing goods These are the two key factors The important thing is to separate out the two factors To make this tractable they assume that some basic parameters 6 039 and 6 are the same across countries Moreover sectoral TFP s grow at the constant rate gA across sectors and across countries6 The parameter values that can vary across countries are 719717141 and A0 Thus TFP grows at similar rates but they can differ across sectors or countries at a point in time Variation in the parameters 719717141 and A07 generate cross country variations in steady state levels of the investment rate at domestic prices the domestic price of investment and the domestic price of consumption They also yield different levels of PPP levels of y at a point in time across countries Notice that because of 7 and the assumption that consumption is nontradable there are no opportunities for international commodity arbitrage There are also no incentives for international capital mobility because capital is taxed where it is located After tax after depreciation real interest rates are the same in all countries and equal to 1 9 71 14 this follows from the consumption Euler equation 7 it is the Keynes Ramsey rule 7 plus the steady state assumption With no capital ows we have savings equal to investment within countries The investment rate at international prices can be written using 10 and 11 as L 471717 PfWAIj k1 739 a 1 PPPA1ltkgtaLL 3 PgWAcxkcm 7 6This latter assumption is not critical Investment Ratios Spring 2004 which can be rewritten as L I V 171717 P1 Au T7 imp 7 15 PfWAIjLL I Pg ch 1 7 because the capital labor ratios in the two sectors are equal The key implication of this equa tion is that low TFP in investment relative to that in consumption can cause the investment rate at PPP to be lower Now consider what happens if A1 falls holding AC xed From 15 it is clear that invest ment rate falls This means that the fall in aggregate TFP will be greater than just the direct effect via its size7 The reason is that capital accumulation is also affected This is impor tant Not only do poor countries have lower TFP than rich ones but they have especially low 41 s Their low sectoral TFP s contribute to their low aggregate TFP and their low 2 ratios contribute to their low physical capital intensity in PPP terms Figuring out why A1 is so low is very important 3 CrossCountry Evidence If the quotinvestment barriersquot hypothesis is true then investment prices should be cheaper in rich countries So if the prices of machinery and equipment are converted to dollar prices at o icial exchange rates or using black market premia then there should be a negative relationship with PPP income But this does not appear in the data see gure for o icial exchange rates Suppose that one regresses the price of investment goods machinery or xed investment on the log of ya either using o icial exchange rates or black market rates to convert in vestment prices If the quotinvestment barriersquot hypothesis is true then the coe icient on yppp should be negative But in the data it is insigni cant and the R2 is very low In some case investment prices appear more expensive in rich countries 7The direct effect follows because we can express economyiwide TFP as L 7 L L TFPJ 13ch J Lj 17 prpAIj Investment Ratios Spring 2004 The high PPP rates of investment in rich countries is due not to low prices of investment but high relative prices of consumption This is consistent with B S nontradables are relatively cheap in poor countries But why are tradable consumer goods cheaper in poor countries Perhaps it is distribution costs But why is investment relatively ine icient in poor countries 0 nancial underdevelopment leads to bad project selection 0 public investment is relatively higher and this is ine icient relative to private Whatever the explanation and more later this implies that poor countries do not appear to lack investment e ort 7 they lack investment e icieney8 Of course this should imply that developing countries should import investment goods In the model there are no tradable consumer goods to export but surely in practice developing countries import a large share of machinery 4 Digression Government Production of Investment Schmitz JME 2001 considers the impact of government production of investment goods If the government is less e icient at producing investment goods this will have a productivity impact on the economy He uses a similar framework He notes that there is a direct and indirect effect 0 direct e eet is lower output and labor productivity in sector I o indirect e eet is that on the rest of the economy through lower capital per worker 8To recap poor countries do not exhibit particularly low investment rates at domestic prices Nor do they exhibit high investment goods prices Instead they exhibit low consumption prices When consumption is valued at common PPP prices in all countries the investment rates in poor countries are lower than in rich countries Poor countries do not appear to suffer from lowisavings traps brought on by high discount rates or subsistence consumption needs If they did we would expect to see much lower domesticiprice investment rates in poor countries Nor do they appear to heavily tax the returns to capital If they did we would expect to see low domesticiprice investment rates in poor countries Finally poor countries do not appear to impose high taxes and tariffs on producing and importing investment goods If they did we would expect to see high investment good prices in poor countries Poor countries do not appear to lack investment effort but rather investment ef ciency Investment Ratios Spring 2004 Suppose we are in a steady state then from earlier we know that R Lyn11kg 1 aPCACkf l 16 i A where 1 have set 739 1 we also know that relative prices satisfy Then the steady state capital labor ratio satis es P I k A1045 17 Now suppose that the government produces all investment goods and that TFP is now given by MA p lt1 Then 4 E MA104 H1 18 so that the effect of this ownership policy is ME 19 Just to have an idea suppose that 04 13 and that M 12 Then E 035 Steady state capital stocks are 35 of what they would be without this policy This is the direct effect Notice that this policy will also change the relative price of investment goods This should not surprise given Hsieh Klenow Before the policy we had now g 41 which implies that the relative price of investment goods doubles What happens to aggregate TFP Let yZPPP be detrended output under policy i 17 2 where the latter is the policy of government ownership Then it is straightforward to show that PPP a y i i k lt17 n m 20 341 1 where n is the share of the labor force that works in the investment sector in the steady state Using 19 we can write ygppwv 1 aww 1 i nyP quotI Fm M i quotI TMnI M Investment Ratios Spring 2004 Now suppose that 80 of the labor force is in consumption and 20 in the investment sector And suppose that we continue to let M 05 Then we obtain it 08 020lt05 050395 Dynamo5 636 which means that the policy of government production of investment would reduce aggregate productivity to 64 of its potential level Of course any of these calculations depend on the value of M Schmitz cites independent studies for Turkey and Egypt which show that stateowned capital goods rms are half as productive as privateowned ones These may seem large but it is important to note that we are not talking about natural monopoly type industries in this case These differences perhaps then re ect some of the costs of regulations that the stateowned rms operate under that the private ones do not Of course there may also be some selection issues that arise from the de nition of industries too broad categories for example More studies needed Notice that if you take into account the fact that government ownership is not 100 then you face some modelling questions How do private and state rms coexist if the latter are less e icient Either they produce different goods which is not too inviting since we would then need to know how substitutable they are Or you could assume that taxes and subsidies are used to allow coexistence It may also be the case that the government mandates the share of the sector that will be stateowned 402 General issues 1 Are government owned rms inherently less productive It would seem so We know why for planned economies Some of these factors however do not apply to developing economies relative prices are market determined there is no central plan no seller s market The causes in this case are primarily organizational 7 absence of free entry and selection and perhaps rent seeking that ensues Could also be extra corruption 0 Indeed one might think that the reason why governments want investment goods sectors 11 Investment Ratios Spring 2004 to be stateowned is that it makes corruption easier to facilitate It is harder to know the price of capital goods so kickbacks are easier to conceal The more frontier the technology the harder to judge the price so bureaucrats may want to encourage imports of such goods to get kickbacks 2 Why is the problem of government ownership more important in developing economies It could partly be a sectoral priorities issue Heavy industry is more important Command ing heights Although government shares of GDP are similar across countries government ownership of manufacturing differs greatly And in countries where there is data this seems concentrated in investment goods sectors Rent seeking opportunities must be a part of this Also they may need protection to survive competition from imports from more developed economies Legacies of import substitution 4 Speci c HiK issues Sum up HampK can explain 0 The cross sectional patterns of correlations particularly for later years 0 The increasing gap between the two correlations over time For this the model should be extended to feature higher growth of 141140 in richer countries 0 Still in need of explanation is the declining correlation of investment rates in domestic prices and income over time Conclusions from this exercise 0 Poor countries have low levels of physical capital The paper claims that this is NOT because they have sacri ced little consumption or invested little in domestic There is some truth to this Poor countries did not save necessarily less in the 90 s in domestic However current levels of capital are the result of decades 12 Investment Ratios Spring 2004 of accumulation and it seems that a few decades ago poor countries were indeed sacri cing less consumption 0 So it can still be true that the low levels of capital today are the result of little sacri ce in the past We need ner development accounting Some other implications The benchmark model has one tradable 1nvestment goods are tradable Consumption goods are not tradable Hence no good is traded in equilibrium There is no motive for trade because there is only one tradable o What happens when trade is allowed for 1e when the poor can produce tradable consumption goods with a comparative advantage 0 With trade the prediction that Carrippp Y gt 0 weakens ie this correlation is lower with than without trade This is because the share of tradables in consumption declines with income Since in the data Corr 1R in PPP prices Ygt0 is robust this suggests that in practice trade between developed and developing countries is negligible Why is trade so low Barriers to trade in DC Protection of primary sectors Productivity of tradables too low in LDC Suggests another test Corr 1R in PPP prices Y should be higher for countries that trade little with DC9 1s P1 equal across countries While measured P1 can be equal effective P1 might differ This could occur due to higher risk in Developing Countries eg risk of expropriation is higher 91f you split the sample into two groups Low and High Trade the CorriPPPY is higher for lowetrade countries opencltmedian 7 CorriPPPY 071 if OpennessltMedian 1996 CorriPPPY 016 if OpennessgtMedian 1996 Good for HampKl But do check other years 13 Investment Ratios Spring 2004 o This higher risk constraints the technology choice set leading countries to adopt less e icient technologies This could be behind the choice of less productive technologies Why should productivity in nontradables rise slower than productivity in tradables Two possible explanations are 0 tradables have greater inherent capacity for productivity improvement because more mechanized and capital intensive o tradable goods have larger scale due to larger markets so the return to innovations may be greater Why is the production of investment goods so ine icient in poorer countries One expla nation is government production Another may relate to scale and markets 5 Conclusion Poor countries have low levels of physical capital 1n the 90 s they did not invest less than the rich in domestic But they seemed to have invested less in the 60 s 80 s So to some extent current low levels of capital might still be the result of low sacri ce of consumption early on References De Long J Bradford and Lawrence H Summers quot How Strongly to Developing Economies Bene t from Equipment Investmentquot Journal of Monetary Economics 32 3 December 1993 395 416 Resource Curse Slides Econ 570 Barry W Ickes The Pennsylvania State University Fall 2008 Big Picture GDPgrowtht 196092 Resource exports as share of GDP Income and Resource Wealth F ure11 Median GDP per capitaconstant1995 U55 esource rich and resourcepoor developing countries 0 1960 1965 1970 1975 1980 1985 1990 1995 39 Resourcerich ResourcePoor Growth and Re50urces a All usuurce rich countries GDP gmwlh 1n Growth and Institutions D C With Dad institutions With good ins luliuns 5 5 4 4 3 3 9 2 2 M l 1 U m D l 1 1 05 2 2 3 3 5 39 E 0va Initial Oil Production and Income Figure Growth Institutions and Natural Resources a m m m m m mmm m m w 7 n l um WWW m m m M W m m w 7 M Iny M Figure n 5 r Oil Wealth and GDP Levels Luge mm Acmngmelal sum V 2 a O 5 5 7 s 9 Hydmcazhan mm mm unw cqzmayc39zwxn ms nu U13 n m Mnl mm naw H m W W cm my a m w l m mm wwnm m 3 1m quot33 mm mm m krgmc nu My 420 u 27 1m fatummzmm um 150 21 um um mm m quotn a Mm xznw mam rum n97 ass 1 PP um um um um human 9mm 1mm mam mm mm nzm 115 Um x H Um um um u w m A law mm mw 12w vuvw 22w ms m m m m m m m Kueufl nun 119 109 Am mm lmnmenhclz n 15 11 Inn may n b5 u 12 m 1 m aa 59 Um rm 2 14 all 972 99 W7 VHarstsLansm a a 260 2a 25gt 2m 2 A m a 30 72gt us 25 Hm mm mndad arm 2m farmlhzzexcnnra1hnnthvwn Mrsgnl mmatmlevdMrsm mmamlcva rnzm mmztlwb am 4mm sq m a nbsnlnls lamude szmzrx39raml w Figure The Effect of Oil Wealth on Per Capita GDP Institutions Oil and GDP Dmrxumlunau mzuuawmtmmym NW 2 1 mung 1 a HS39LmErhnr an n amms 17 l am y Am in um rumquot 1 Cune Slldx mun inzo rum m 433 oilcur a 15m 66m comm c mm 552w mm yum E lull in own My 465m 473 myth llei U mu Abmme mm mum mm mm hiva Milt mu Uln Wu uxw W 43quot um 95 1quot m2 121 m A Eu npm i 204 m lS 3w lanELEgc ME A 95 L145 1911 mm munimgmm m 74 m 251 25 274 ALanAiumx 2427 mm 1957 r241 2m 2451 Na abs m 1m 113 lit 11 Ad R5 and 17 7 on m m lutiwlnum 111ml ummttmuniulmumi m r s sn nanat1late rsign nandti39hlcve rsigm cimal ms Figure Effect of Oil Wealth on Rule of Law a a n q gt Oil Reserves and Per Capita Income Figure 4 Oil Reserves 1970 and Per Capita Incume Levels 1998 w m Maa u w 39 ME qu w P m m w 5 10 1 20 Oil Reserves 1970 v v m lugs n 3 7 RC n em ma sum Coppe mamaquot Lac we r Production and Reserves must mummummm mimismmdcnle am mus momma 15451975 5mm c I Schmilz 1979 World NanFarm Mm Pro111mm and mm 17001976 Loudm Frank Cass pp 53773 andl70 72 Notes on Health and Development Econ 570 Spring 2005 1 Introduction Relationship between health and gdp growth is important and works in both directions This is very important for development and transition because health outcomes are measured much better than national income More important however health can be an important factor in economic performance What makes analysis complicated 7 and interesting 7 is that economic performance also impacts on the health environment1 2 A Brief Survey Life expectancy has risen dramatically during the 20th century and the 19th too This must be important for assessments of GDP and for assessing the determinants of GDP 1 consider the former below with the concept of full income The latter point is worth noting however 2 Health and Growth The effect of economic growth on health is well known Because the demand for health care is income elastic rising per capita income leads to increasing expenditures on health care and improved health status But the reverse link 7 from improved health care to economic growth has recently drawn increasing attention Understanding this link is crucial to the project though it is dif cult due to the presence of causation in both directions The natural mechanism that relates health status to economic performance is human capital theory Health status is a component of an individual s human capital Poor health reduces both the quality and quantity of labor supply It also results in low levels of human capital accumulation Low growth results in poor health outcomes We will do a comparative 1This reverse causality makes estimating the causal effect of health on economic performance dif cult but work in the area usually identi es the effect through timing using childhood health and nutrition inputs as a determinant of adult wages or taking population health in say 1960 as a factor in uencing economic growth during 1960795 Notes on Health Spring 2005 cross country study for transition economies with comparisons to similar income level non transition economies Growth and health are correlated Health is a normal good so greater income leads to greater expenditure on health care But the reverse is also true health status has an effect on growth lmproved health increases the quantity and the e iciency of labor lmproved health not only reduces lost time due to illness It also increases the intensity of work from a given quantity of labor2 It is an investment in human capital and it increases the return to investing in human capital3 Improvements in health have both level and growth rate effects on per capita income Level effects from improved health result from increases in effective labor inputs4 lmproved health contributes to this in two ways First by increasing the supply of labor input due to less time missed due to disease Second by the increase in the e iciency of labor input due to improvements in the quality of labor when individuals are healthier5 Growth rate effects occur because a lower incidence of disease increases the private and social rates of return to human capital investment which in turn leads to higher rates of economic growth Modern growth theory Cf Aghion and Howitt 1998 distinguishes two channels through which human capital accumulation affects the growth rate 0 One approach 7 in the tradition of Lucas 7 argues that differences in growth rates of per capita income are driven by differences in rates of human capital accumulation It is straightforward to show how improvements in health increase the return to human capital accumulation and thus have a direct effect on the growth rate 0 In the second approach 7 Schumpeterian 7 differences in the level stock of human capital affect the capacity of the economy to innovate and catch up with more advanced 2Improved health also affects the choice of occupation Improved allocation of labor can thus result from improvements in health 3Improvements in health status enhance the return to investing in human capital the same way that business insurance increases physical investment 4Economic historians have estimated that up to 50 of the increase in British per7capita income 179071950 due to improved nutrition Fogel 1994 388 5Greater nutrition for example increases the proportion of energy ingested that can be metabolized 7 Atwater effects Fogel 1994 386 Notes on Health Spring 2005 countries Since the stock of human capital is correlated with the level of health achieve ment it is again straightforward to explain the growth impact of health performance Although for technological growth is it the extremes that matter or the average Prob ably the extremes This suggests that there may be increasing returns to health investment lmproved health leads to more investment in human capital and faster growth in per capita income But faster growth in income leads to greater expenditures on health and thus improved health status A virtuous circle results6 0 Azariadis Drazen model has an externality from human capital accumulation but this could come from health Some evidence on the role of health on economic growth comes from cross country re gressions Controlling for other determinants of economic growth Bloom and Sachs 1998 table 6 show that an increase in life expectancy in 1965 of one percent accounted for an acceleration in GDP per capita growth of 3 for the subsequent quarter century One must be careful in assessing this nding however as causation runs both ways and life expectancy is an endogenous variable Still the nding is suggestive and begs for analysis that can sort through the complications of simultaneity lf life expectancy in low and middle income countries increases from 44 years in 1960 to 64 years in 2001 then this must help explain increases in per capita income People can work harder and longer even before we consider any behavioral implications of this 0 this is equivalent to an increase in labor augmenting technical change at least during the transition to the new level of health It should raise the steady state capital labor ratio measured in ef ciency units and thus accelerate growth in transition 6A crucial question for transition economies is whether such a process works in reverse when there is a negative shock to the economy Transition has led to a decline in health performance and in pericapita income Will the feedback from health to growth lead in this case to a vicious circle The effects of health on growth take very long periods to manifest but the key question is whether the current deterioration in the health system will have longelasting effects on economic performance in transition economies Notes on Health Spring 2005 Ratio 1 1950 1960 1970 1930 i990 2000 2010 20m 2030 Emma SunLhenstAsm Scum Aai Figure 1 Ratio of WorkingeAge to NoneWorking Age Population So in crossecountry regressions differences in life expectancy may explain differences in performance We should use initial levels of life expectancy as a conditioning variable lnitial income is insuf cient Two countries with equal 3 but unequal life expectancies ought to produce different grth outcomes This is very important if these life expectancies also impact on the choices of subsequent policies Notice that if there is a lag between improvements in health and increases in income then conditioning becomes all the more important For then countries with good health but low incomes would be predicted to grow faster than otherwise similar economies This is clearly the case for East Asian countries Much recent evidence attributes this grth to rapid accumulation of factor inputs labor human capital and physical capital All of these are impacted by health issues Life expectancy for example rose from 39 years in 1960 to 67 in 1990 The ratio of workingeage people 1564 to dependents 0714 and over 65 rose from 13 to over 2 See for example gure 1 This meant more factor inputs and more savings that could be plowed into capital7 Greater life expectancy led to greater saving for retirement Greater life expectancy led to more human capital accumulation Asia s demographic transition followed the stylized model by starting with a decline in 70f course this is not a steadyrstate effect but it had a major growth impact precisely when growth accelerated Notes on Health Spring 2005 mortality rates 0 By the late 1940s the crude death rate had begun to decline very rapidly throughout much of Asia The decline proceeded most rapidly in East Asia and was accompanied by an increase in life expectancy from 612 to 746 years from 1960 to 1992 Similar declines occurred in Southeast and South Asia where life expectancy improved from 516 to 672 years and from 469 to 606 years respectively 0 1n the 1950s and 1960s most of the aggregate decline in mortality was driven by declines in mortality among the youngest cohorts Bloom and Williamson 1997 0 There are a number of possible explanations for the rapid decline in child mortality in Asia in the middle of this century One possibility is that in the 1940s Asia escaped from some four or ve decades of relative isolation ushering in an era of transfer and diffusion of new public health programs technologies and techniques For example the medical advances that were implemented in postwar Asia had been accumulating on the technological shelf for at least two decades penicillin sulfa drugs and bacitracin in 1943 streptomycin was isolated in 1943 etc With the advent of these and other drugs diseases that had once killed hundreds of thousands and even millions became treatable at low cost 1n addition the pesticide DDT became available in 19438 0 Another possibility is that increased agricultural productivity and trade in food both improved nutrition suf ciently to lower infant mortality dramatically in less than a decade and did so everywhere in Asia This may be true but it seems unlikely given that the magnitude and timing of the decline in mortality were so similar everywhere in Asia regardless of level of development and agricultural productivity 8To cite just one example DDT spraying in the late 1940s dramatically reduced the incidence of malaria in Sri Lanka the crude death rate declined from 215 to 126 between 1945 and 1950 with the most precipitous drops in the most malarial areas LivieBacci 1992 Notes on Hesnn Spnng 2mm Frgure 2 A sryusue Model ofEasL Asrsn Mrrsde In any case en1hlysmn ell ss hddsuwiox rsres ruse wnsr occurs men rs rne dynamcs of demyaphy ss m gure 1 In L112 early penod hen by fdlmg snsrss ofwokagrage sduns drrnrnrsn the grnmn ofpex espus rnsnrne As the Hammer proceeds falhng youth dependency burdens snd nsmg snsrss of workmgrage sduns prnrnsre rne Wan ofpex espus rnsnrne T ady burden ofhavmg ew workers snd savers becomes s pnrenusx ya s dupmpormonately hgh snsre of workmgrage sduns Lsrer rne emmmm gm drssrpsres ss rne snsre ofeldexlynses Thus grmn rs rsr slowed due us the burden snd men seedersres due us she dmdend The rdes rs given m gure 2 2 1 arm m mum rnru favorable dernegrspnre changes nus leads to s dmdend Notes on Health Spring 2005 De ne the growth rate of per worker output as y ln then the growth rate 1 over the interval 1515 k is 1 Yi yittk E111 1 Y from the neoclassical growth model we assume that growth is a function of the distance from steady state yittk A M i where A gt 0 and Y is the steady state level of output per worker We assume that Y 6X where X is a matrix with Z determinants of the steady state We are interested in per capita income growth but we have an expression for per worker growth We note that the former can be written as If we take logs and differentiate with respect to time then 3 y 9L 9N Suppose we estimate a Barro type cross country regression for pericapz39ta output Then we have yittk 7T1X 7T2Yz39t WagL 7T4gN 4 where clearly 7T1 and 7T2 depend on the speed of convergence A The important point is that in steady state gL gN so we should expect 7T3 771394 1 Hence normally we ignore this effect 7 net demographic effects wash out But if the economy is demographically unstable then there could be something from these last two terms Notes on Health Spring 2005 many years ago demographers worried that too high a dependency ratio would slow growth But it could also be the case that per capita output can accelerate if gL gt gN Notice that this could happen due to increased nutrition or due to changing demographics 0 both the sources of population growth and the stage of the demographic transition do matter both a decline in child mortality and a baby boom raise the share of young depen dents in the population a decline in mortality among the elderly increases the share of the retired dependent age cohort immigration raises the working age population because it self selects young adultsl and improved mortality among the population at large has no impact on age structure at all 0 Because an economy s productive capacity is linked directly to the size of its working age population relative to its total population distinguishing between the two components when exploring the impact of demographic change on economic performance seems nat ural and worthwhile Consider then Bloom and Williamson s table 3 which con rms that the growth of the working age population has a powerful positive impact on growth of GDP per capita while growth of the total population has a powerful negative impact after controlling for other expected in uences 0 Consider the results reported in column 1b of table 3 The coe icient on the growth rate of the working age population is positive statis tically signi cant and large in magnitude an increase of 1 percent in the growth Notes on Hesnh Spring 2mm e ll hemmemsle m muss w WWW m W m we 5 he memelsllnl mlw e Flgllle a lsle nuhe Workmgrage poplllsllen ls ssssslslea mh sn lnslesse of 1 4s peran ln lhe yoth me 0mm pel ssplls e The coe nem snlhe yomh lsle nuhe mm population ls negsllve slsllsllssuy slgnl ssnl and slnnsl ss large sn lnslesse of peran ln lhe glowlh lsle nuhe overs poplllsllen ellesllvely lhe dependenl popllmlnn smce lhe emplnssl spear l cauon hnlds xed lhe glowlh lsle snhe warhngrage population ls ssssslslea Wllh s decrease of us pelsenl lnlhe yowthnle 0mm pel ssplls One should wally shells levelse ssllsshly hele om economic gromhm demyaphxc e hlll when lhey llse lnslnlnenls lhell esllnlsles do nel change They also show lhsl lz lhe dzpendency gully ls hmhen down mm ynulh and elderly lhsl lhe 1mm sle slslgel alsg The seemmgly dependan elderly eonlllhllle more lo Notes on Health Spring 2005 growth than the dependent youth 7 good for Russia if true 0 How large is the contribution of demographic change to economic miracle They say large two channels increase in hours worked per person increase in savings but very hard to control for endogeneity here 2 2 Full Income These considerations also imply that GDP may not be the best measure of welfare Con sider two economies with similar GDP s but different health levels They clearly do not produce similar levels of welfare for their inhabitants nor will projections of the future be the same To adjust for this some researchers use estimates of the value of a statistical life VSL The latter is derived from observed willingness to pay to reduce risk9 Using this one can look at changes in quot full incomequot by adding the value of changes in annual mortality rates calculated using VSL s to changes in gdp per capita These studies tend to give a different view on inequality full income inequality fell sooner in the developing world than income inequality itself Some studies nd strong evidence for full income convergence even if there is evidence of divergence in gdp per capita 3 Implications for Other Work This is interesting because of the settler mortality instrument Recall the idea is that settler mortality is correlated with institutional choice but not with other factors that predict 9Suppose for example a worker requires and is paid 500 a year of additional pay to accept a more risky but otherwise similar job where the increase in the mortality rate is 1 in 10000 a year the value placed on reducing risk by this magnitude is simply 500 The value of a statistical life is de ned as the observed amount required to accept a risk divided by the level of the riskithat is in the example we have chosen the VSL would be 5 000 000 a number in the range of estimates for the United States today Willingness to pay to avoid risks rises not surprisingly with income A reasonable range of values for a country s VSL appears to be 1007200 times GDP per capita with values estimated in richer countries more likely to occur toward the high end of the range Notes on Health spring 2005 1750 1730 1710 1690 1670 1650 Millimeters 1710 1720 1730 1740 1750 1760 1770 1780 1790 Elirl deca es Sweden us Slaves Us Free England Bavaria A Northern Italy Figure 4 lnternational Comparison of Adult Heights growth in income in later centuries The results on health nnay however alter how we view this consider that low settler nnortality irnphes greater average health levels Less disease This leads to better anthroponnetric outoonnes Healthier people can work longer and harder This directly raises perecapita income 0 The average height of soldiers who fought in the American Revolutionary War was an astonishing 173 cm 68 inches well above European standards for a very long time to conne see Figure 4 This result made it clear that the abundant natural resources of parts of the New World Combined with the propitious disease environrnent low population density and productivity of the American population conferred considerable e until then unknown 7 biological advantages on its inhabitants Even Arnerican adult slaves were tall by modern standards lndeed they were about as tall as American urban workers More importantly those born on American soil were actually taller e had a higher nutritional statusrthan the African populations from which they originated o Anthroponnetric research on heights is very interesting because heights are very related to the disease environrnent Eradication of early childhood disease is an important Contributor to increased height Notes on Health Spring 2005 Moreover healthier environments will impact on agents behavior decisions Healthier agents accumulate more human capital They also save more for retirement important pre cisely for the echo when population starts to rise This creates a better economic environment for adoption of new technologies Greater savings plus more educated workforce o Glaeser et al suggest that colonialists brought better human capital with them rather than institutions In addition 1 am arguing that better health environments allowed for further human capital accumulation in such regions 0 The Glaeser et al effect is that the initial stock of human capital is negatively correlated with SM 1 am arguing in addition that the ability to accumulate ho is negatively correlated with SM Thus even if colonists imposed the same institutions in dense and sparsely populated areas we would expect on predicted health outcomes alone that the sparsely populated regions would do better because of the better health environments This further suggests that the appropriate test would require estimating the impact of institutions as the residual effect once the better health environment is accounted for 0 notice that that human capital effect on growth like that of institutions is going to occur with a lag because human capital becomes important for growth later in the process ie during industrialization just like institutions We might also expect that fdi will be related to health outcomes But this could clearly be related to settler mortality and ow for the same reason If regions had high settler mortality they may get less fdi Improvements in health would then lead to more fdi and to subsequently higher growth Lags could be long however Notes on Institutions Econ 570 Spring 2005 1 Introduction We want to distinguish the role of geography institutions and policy integration but this is hard due to endogeneity This is evident in the Rodrik diagram gure 1 Geography has effects via several channels 0 direct effect via agricultural productivity and morbidity o indirect via integration with world economy eg greater costs of trade 0 indirect effect via institutions via settlers and then institutions via factor endowments and then political systems via the resource curse Of course the existence of the resource curse is itself open to question The resource curse literature assumes that resource abundance bad institutions But it could be that the impact of resource abundance on performance depends on institutional type In that case the impact of resources themselves would not be independent of the institutions choice 2 Rodrik et a1 Basic Speci cation The basic idea is to test 10gy MaINS 51NT f YGEOiJFEi 1 where y is income per capita in country i INS INT and GEO are respectively measures for institutions integration and geography and E is the random error term Throughout the paper we will be interested in the size sign and signi cance of the three coef cients 04 6 and 39y Institutions Thoughts Spring 2005 inc cine level endogenous ge ogl aplly exogenous Figure 1 The Rodrik Diagram How to measure these things For now we note that institutions are measured by rule of law index Kaufman et al1 and geography is measured as distance from the equator2 Integration is openness measured as the sum of imports plus exports to GDPa The problem with estimating 1 is of course endogeneity So they use twopstage least squares with instruments from other successful papers the AJ R paper for institutions and the FR paper for trade Hence they estimate INS A 58M neomvsr wGEO em 2 INT 6 aOONST TSM wGEO em 3 1Am use protectlon against exproprlatlon but thls choice ylelols a larger sample 2There are problems Wlth thls choice AccordJng to Sachs thls Hls an exceedJngly poor choice for a serlous test of geographical yarlables 1 It ls at best a proxy and a poor one at that for cllmate or posslbly for dlstanee from major markets a 1 9 0 m 9 a 5 m 0 a 9n r 4 0 E r 9o 5 m n m g E 5 m 0 a H g 9 and oul when much better alternatlyes are ayallable In any event most geographical analyses stress several factors dlmate geographical lsolatlon sease enylronment so that testing these yarlables one at a tlme ls subject of leftroutrvarlable error 2000 4 Notlee that olstanee from the equator could also be a proxy for settlement aThls ls also problernatlc Flrst Japan ls not that open by thls measure nor ls the US Second there could be problems olue to use of nomlnal gdp rather than PPP Suppose that traole ralses prooluetlylty ln traoables Nonrtradables prlees rlse relatlye to traolables Thls blases down the ratlo of traole to gdp n o 3 5 E Institutions Thoughts Spring 2005 where SM refers to settler mortality and CONST to the FR instrument for tradeGDP4 The exclusion restrictions are that SM and CONST do not appear in equation They use quotAER approvedquot instruments Suppose you estimated 1 with OLS ignoring endogeneity Then in the 80 country sample they obtain 237 686 161 geography is statistically signi cant and important What about 2SLS Now the power of GEO and 1NT go away 1nstitutions have a larger coe icient three times 32 2007 355 and it is very signi cant The others are not signi cant and have the wrong sign 1n the rst stage SM DISTEQ and the FR measure CONST all explain institutions Only settler mortality seems to explain openness surprisingly Greater openness also improves institutions so though no direct impact on income there is an indirect effect But this is about a tenth of the direct effect of institutions on income Why so much bigger RST then estimate how these fundamental variables impact growth determinants y k h and productivity A They simply regress these variables on DISTEQ INST and LCOPENI They nd that institutions are signi cant in all the regressions the other variables are usually not and that institutions have a larger impact on physical than six times greater than hu man capital accumulation or on 32 times greater than A This could be because institutions play their biggest role in preventing expropriation 1f institutions are bad best to invest in human than physical capital This is not surprising but it is intuitive 4This is the constructed trade share from Frankel and Rose They suggest that we can instrument for actual tradeGDP ratios by using tradeGDP shares constructed on the basis of a gravity equation for bilateral trade ows The FR approach consists of rst regressing bilateral trade ows as a share of a country s GDP on measures of country mass distance between the trade partners and a few other geographical variables and then constructing a predicted aggregate trade share for each country on the basis of the coef cients estimated This constructed trade share is then used as an instrument for actual trade shares in estimating the impact of trade on levels of income Institutions Thoughts Spring 2005 2 Robustness Robustness checks are needed because the instruments may not be appropriate Alternative geographical variables can be used 3 Institutions and policies Should we think of these as alternatives as in EL This is problematic Policies and institutions are related Good policies such as Meiji Restoration or what happened in South Korea in the 60 s or China after 1978 have big effects Policy innovations become institutions Or bad institutional quality could be the result of bad policies It is useful to think of policy as a ow variable and institutions as a stock Institutional quality is the cumulative outcome of past policies Let p be policy on dimension i scal trade etc let I be institutional quality and let 6 be the rate at which it depreciates Then we can express the evolution of institutional quality as i where 04 is the impact of a policy on institutional quality The idea here is that good policies improve institutional quality and without this they can decay If this is correct then it is not proper to include both p and I in regressions Not just because of frequency 7 EL average policies over four decades The problem is that information about the former is already contained in the latter Of course we really need a theory to develop 5 and this should be one of our goals but the implications for testing seem evident The idea here is that levels of income depend on I The impact of p should then be on changes in the level of income This is very different from Sachs s argument quot Economic theory suggests that the determi nation of per capita income should be speci ed as a dynamic process in which the growth of income during a time interval 0 T is a function of the income level at the start of the period Institutions Thoughts Spring 2005 and some kind of average of the values of the forcing variables during the time interval 3le 1 Vi 114 50 511mm gznom 53111360 6 6 10 Here Z represent other factors Sachs s point is that the quality of institutions in a given time period will affect the growth rate of the economy during that period controlling for initial income as opposed to the contemporaneous level of national income Is this true however Growth rate differences may be explained by high frequency phenomena 7 policies 7 and the degree of backwardness Institutional quality may impact growth rates if the correct other factors are there I suppose But the question we are after is what explains the large differences in income levels 5 1 Transition Issues It is likely that causation goes in both directions but that these operate on different horizons Comparing countries with very different income levels 7 differences that took a very long time to occur 7 differences in institutions are likely to play a key role In the shorter run however improvements in income are likely to make it less costly to improve institutions and to make such improvements stick5 The problem for transition economics is to nd instruments to separate out these two different channels The reason for this confusion is partly due to the way most of these tests are speci ed They typically estimate a cross country regression of the form In 6061I Q52Zi5i 7 where Y is income I is a measure of institutional quality and Z is a vector of other variables that could affect income Ignoring the problems of endogeneity involved in estimating this regression the important point for our purposes is that static nature of the hypothesis being tested that the current level of institutional quality determines the current level of income6 5Crises are also periods when institutional change is rapid but what is clearly crucial for performance is the stability and durability of institutional change 6This point is emphasized by Sachs 7 Institutions Thoughts Spring 2005 But the mechanisms through which the quality of institutions are supposed to impact on performance are more likely to affect the growth rate of the economy than its level7 It is much more likely that institutional quality impacts on the growth rate of the economy given some initial level of income The process is a dynamic one The reason why regressions like expression 7 quotworkquot is that cross country differences in income are very large and so re ect years of different growth rates while the indicator of institutional quality is really the accumulation of increments in performance What transition economists would really want to explain is something like yi0T 50 f 51l om f 62Zi0T 53111330 8139 8 where yMO T is the growth rate of income over the period QT and we know measure the other forcing variables over the time period of question There are two serious problems for transition economists in this framework One relates to difference in this institution variable and what it means The second relates to the initial income variable In the context of a global cross country regression a variable like I 30m has a straightfor ward interpretation It is the institutional setup that characterizes an economy during the period that we are explaining Cross country we expect that differences in these institutional setups will be larger than any time series changes in them during the period of observation But we certainly cannot make this assumption for transition economies Time series variation in I Q is really the whole point The question is then how to measure institutional quality and how to specify the dynamic relationship between this and performance Now a serious problem for transition economists is how to measure the initial level of income when estimating an expression like 1n development economics this is not really an issue because there is no belief that initial income is measures signi cantly worse than subsequent levels unless the time period is very long But for transition economies this is precisely the problem The initial income level at the start of transition is an item of great uncertainty in transition economics because of continuing controversies about the size of the 7Notice that while this is clearly true for development it may not be so for transition Think of replacing planning institutions with the market this could improve the static allocation of resources independent of any impact on the growth rate Institutions Thoughts Spring 2005 output fall Transition is the process of prices gaining meaning So over time the dependent variable is measured with increased precision But the starting level is a big problem Of course if the size of the output fall is equally mismeasured in all transition economies then estimating 8 will not be seriously compromised But this is hypothesis is very dif cult to maintain 32 Hard Work RST argue that all of this is just a prelude to the hard work Knowing that institutions matter does not tell policy makers much What these models do is estimate 6 the impact of I on 31 but what they need to know about is 04 This is especially dif cult as the identi cation strategies for I rely on exogenous sources of information They are good instruments but they are not quotinstrumentalquot To see this sharply consider the fact that the measures of I are related to observer ratings of likelihood to invest And property rights is clearly an important factor in this Yet if we compare China and Russia we observe that in the latter there is clear formal legal protection of property rights while in the former there is a socialist legal system and very informal protection Yet investors are more secure in China than in Russia So you could not interpret the correct policy to be establish formal property rights But then it is not clear exactly how to improve I Presumably this is a problem of formal versus informal institutions We need more theory of that It is also related to the politics of rejecting institutions Institutional antibodies 7 ala Khrushchev and the Sovnakhozy 4 Unbundling Institutions AJ point out that we need to unbundle institutions We know that good institutions promote performance but why Is it support for private contracting or protection against ex propriation They use legal origins 7 who settled 7 to determine the degree of legal formalism and settler mortality for the protection type institutions Multiple IV s Institutions Thoughts Spring 2005 They nd that legal origins explains degree of legal formalism But once you control for property rights protection legal formalism has little impact on y or Suggests that legal institutions have more impact on form of nancial intermediation than on growth etc Their intuition is that private contracting and reputation mechanisms can alleviate inef ciencies that are the result of insuf cient legal formalism But when property rights protection is weak there is often no easy ways to circumvent the problem Better for capital just to ea 5 Imposition versus Adoption 0 Is mortality a good instrument or does it proxy for something else Note that it is settler mortality before there was growth in per capita consumption so it is not a proxy for poverty Indeed it is positively correlated with their measure of wealth because it is European settler mortality 0 The implicit assumption here is that institutions are more easily adopted via settlement transplantation Even though institutions are non rivalrous they are not adopted easily Imitation is dif cult Japans are very few and hard to explain 0 Two problems with imitation of institutions rather than transplantation Resistance of locals to foreign institutions this could be due to threats to political power see below Dif culty in getting new institutions to work in foreign cultures 0 Where Europeans could leave their stamp they solve both problems but where the societies were already developed it is harder Of course all this begs why these institutions developed in Europe If these insti tutions are superior why must they be imposed rather than imitated Institutions Thoughts Spring 2005 5 Invasion versus imitation o This seems very much related to notions of technology transfer Sometimes it works and other times it does not Why AJR seems to imply that invasion is necessary Relate to Mokyr s examples of technology eg Da Vinci s inventions Relate it to informal institutions versus formal institutions 0 Mokyr s thesis for macroinventions to survive three conditions are essential The new idea must be technically feasible That is within the ability of contemporaries to reproduce it8 The new idea must be economically feasible The new invention has to be borne into a socially sympathetic environment in biological evolution it is sometimes necessary for an environmental shock eg earthquake to cut off a region so that speciation can occur What about political shocks to allow institutional development cutoff of cheap Soviet energy might have forced institutional change in Central Europe 0 Mokyr discusses the complementary microinventions that are necessary for a macroin vention to succeed Is there anything similar for institutional development Is this what informal institutions is all about 0 Why do powerful interest groups block the introduction of new technologies Acemoglu Robinson If they have such power why not allow the innovation that raises output and tax the proceeds 8This is where many of Da Vinci s ideas failed the workmen of the time were not skilled enough to carry them out Institutions Thoughts Spring 2005 could be limits to scal instruments more likely it is the fact that the innovation could threaten political power 9 9 it is precisely when the innovation threatens the continuation of political power that innovations will be blocked blocking thus depends on the prob of remaining in power if they block 1 versus the prob of remaining in power if they do not block 8 Notice if s gt q they would never block so it is the opposite case that is interesting notice also that if the new technology is su iciently better then the monopolist would never block if s q l Taxing would always be better Blocking arises because 8 lt 1 they argue that in Britain and Germany the aristocracy was not threatened so they allowed innovation In Russia and Austria political power was more fragile So they were more concerned to block new technologies The latter had more closed political systems so they were more fragile q 7 8 may have been higher 0 The role of informal institutions is important North contrasts this with formal rules That the informal constraints are important in themselves and not simply as appendages to formal rules can be observed from the evidence that the same formal rules and or constitutions imposed on different societies produce different outcomes And discontinuous institutional change such as revolution or military conquest and subjugation certainly produce new outcomes But what is most striking although seldom observed particularly by advocates of revolution is the persistence of so many aspects of a society in spite of a total change in the rules p 36 the key point here is that these informal constraints preexist the formal constraints They are the iceberg below the sea 7 they impact how far the formal rules can sink roots into the system Institutions Thoughts Spring 2005 o What determines the power and persistence of informal rules When will they be able to bloc reform and when not Perhaps most important of all the formal rules change but the informal con straints do not In consequence there develops an ongoing tension between infor mal constraints and the new formal rules An immediate tendency is to have new formal rules supplant the persisting informal constraints Such change is sometimes possible in particular in a partial equilibrium context but it ignores the deep seated cultural inheritance that underlies many informal constraints Al though a wholesale change in the formal rules may take place at the same time there will be many informal constraints that have great survival tenacity because they still resolve basic exchange problems among the participants be they social political or economic The result over time tends to be a restructuring of the overall constraints 7 in both directions 7 to produce a new equilibrium that is far less revolutionary p 91 the point here is that the informal rules are solving problems that are embedded in the society If the underlying structure is unchanged they will still have great value institutions typically change incrementally rather than in discontinuous fash ion How and why they change incrementally and why even discontinuous changes such as revolution and conquest are never completely discontinuous are a result of the imbeddedness of informal constraints in societies Although formal rules may change overnight as the result of political or judicial decisions informal con straints embodied in customs traditions and codes of conduct are much more impervious to deliberate policies These cultural constraints not only connect the past with the present and future but provide us with a key to explaining the path of historical change p 6 Institutions Thoughts Spring 2005 0 we can think of informal institutions as a reason why it is hard to import institutions Almost like antibodies rejecting foreign organisms o O ring theory gives another way to think about why technologies may not work as well in different countries 0 what this all suggests is that adopting what appear to be ef cient institutions that are successful in one society may be problematic in different societies if institutions are organizations plus expectations it is critical that the expectations beliefs be appropriate in the adopter society This of course depends on history because of the importance of cultural beliefs Hence to answer this question it is critical that we understand the conditions that make the institutions successful in particular we need to understand what collateral institutions are necessary to support them 0 Adoption takes time in a gametheoretic setting a change in the extensive form should change the equilibrium of the game instantly but in practice these changes often take a long time 9 9 9 this may be because expectations are historically dependent this would be especially important in settings with multiple equilibria coordination games it is also due to bounded rationality which limits the ability to fully comprehend the strategic possibilities inherent in a speci c situation the work on the emergence of norms is important here it is important to recognize that a whole range of factors 7 social political cultural 7 may affect equilibrium selection not just economic ef ciency think about the cross country work on the ultimatum game Institutions Thoughts Spring 2005 this would not be important if equilibria are unique but with multiple equilibria this is important the literature on the emergence of conventions is relevant here 52 Privatization of relational capital 0 1n the STE supply relations involved Gossnab obkom of cials and enterprise relations 1n transition the privatization of relational capital means essentially that Gossnab and the obkom of cials are out But the enterprise relations remain You are get ting rid of the formal rules of the old system which had already eroded under Gorbachev But during the Gorbachev period the informal relations had our ished o The reformers get rid of the formal system 7 all their attention is focused on that But what really mattered by then was the informal relations In particular Communist party had mediated relational capital Now relational capital is more direct 7 it is not mediated by other institutions Getting rid of the mediation made it easier more economical to invest in relational capital The director has more power Less people to please Returns to investment in relational capital are now all appropriated by me 7 before the State took it You could say that the tax service still takes it but liberalization makes it easier for me to evade the state So 1 have more incentive to invest in 7 Had the relational capital not previously existed privatization etc would have been correct The only action would have been to invest in reducing d But because the relational capital was there you choose to invest in r Institutions Thoughts Spring 2005 55 Transition and Institutions 0 Note that institutions arise to meet the challenges posed to agents If you change the rules it had better be the case that new institutions will solve this better than the old There is competition between the structures 0 There are two critical factors in this story First there is the inherited structure of the games agents are involved with This structure determines the suitability of institutions to help agents solve the problems they face The second factor is that informal rules exist that developed when the old rules and the inherited structure were predominant When the formal rules are changed the informal rules provide an alternative lifeboat for those agents that cannot play by the new rules So in our game if the market system is suf ciently advantageous it can invade and take over the Soviet environment That means that the new formal rules are not too out of sync with the problems agents face If the legacy from the old regime is so severe that the new rules are not consistent with the problems agents face then a change in the formal rules may not be suf cient to make the new rules dominate In these circumstances the informal rules may end up dominating if the perturbation is not large enough then the virtual economy becomes the equilibrium Conquest recognizes the dif culty but the metaphor is incorrect It is not as if a country can as it were be put in dry dock and equipped with new institutions in a careful and considered way The whole venture is more like trying to reequip a ship at sea in stormy waters with a new engine This metaphor misses the point because it has the institutions replaced still in one fell swoop but it is appealing because it explains that time does not sit still for reformers and Institutions Thoughts Spring 2005 that the key problem they face is the di icult environment in Which the neW institutions must take root Note on Technology and Critical Mass Econ 570 Spring 2005 1 Introduction These notes are inspired by Mokyr s book The Gifts of Athena The key to Modern economic growth is the transformation of negative into positive feedback A phase transition occurs But why The Malthusian mechanism is one such negative feedback We have argued that what is key in modern economic growth is that a threshold is achieved which leads to positive growth 7 a cumulative process How Notice that before the modern period technological improvements oc curred but they did not amount to a key force for growth Smithian factors were more important before the last couple of centuries1 Pre industrial rev olution growth was Smithian But then technology took over before the demographic revolution Of course there was interaction 7 growth in trade organization property rights affected the returns to innovation Also com munication had a big impact Before 1750 knowledge was insuf cient to make technological growth cu mulative An innovation occurs but then the system stabilizes Trial and error without mechanics So innovations were not improved on Techno logical change in this era is serendipitous not causal A critical point is that prior to 1750 even informed mechanics chemists and farmers knew not enough about the elds of knowledge they sought to apply Not enough was known to generate sustained economic growth based on technological change A similar problem occurs with institutions For much of history and perhaps for much of the world institutions are a conservative force lncreases 1EG Imperial Rome Venice Genoa and Flemish cities fourished on the basis of commerical progress in wealth were dissipated through rent seeking and other predation that cooked the geese laying golden eggs Often it was tax collectors and pirates or brigands Or Mercantilism The most insidious were institutions that resisted technological innovations2 On this idea the key to modern economic growth is the reversal of these processes How does it happen One key element of this is what Mokyr calls the quotlndustrial Enlighten mentquot What is meant by this lt is a change in the culture and environment within which technology and science operate 2 Knowledge To understand the industrial enlightenment we need to distinguish two types of knowledge propositional and prescriptive The latter is invented the former is discovered only DNA existed before Crick Watson Propositional knowledge forms the epistemic base of society lt limits what can be done 0 Propositional knowledge need not be correct the humoral theory of disease led to prescriptions that worked such as the draining of swamps to cure malaria o The wider the epistemic base the lower the cost of research and devel opment and the greater the likelihood of success 0 The wider the base the easier it is to improve an invention primarily because more is known about why something works lt makes recom bination with other techniques easier 3 Industrial Revolution Some key points about the 1R First it is wider than Britain And it is ap plication as much as invention Britain had much greater success at applying inventions than in generating them ln France for example much talent was diverted to war and politics Lavoisier was executed as a tax farmer 2Not just the church but also medical establishments which resisted Jenner s discoveries with regard to smallpox Britain was also special in the willingness to adopt what was invented else where gas lighting chlorine bleaching Jacquard loom Leblanc soda making process for example Britain was also fortunate in that it had workers who could take advan tage and political institutions that were more adaptable to change Conti nental countries had more medieval debris to remove More rent seeking and regulation and more di iculty in removing it The key point about the 1R was not just the clusters of inventions but the persistence of technical change after the rst wave No petering out after 1800 It is really the second wave that starts the acceleration of per capita income 31 Why Did it Succeed Key was developments of useful knowledge before and around 1750 quotWhat mattered was not so much scienti c knowledge itself but the method and cul ture involving the generation and diffusion of propositional knowledge The lndustrial Revolution and its aftermath were based on a set of propositional knowledge that was not only increasing in size but which was becoming in creasingly accessible and in which segments that were more effective were becoming tighter Mokyr 2003 23quot 0 scienti c knowledge becomes public 0 exogenous decline in access costs to propositional knowledge including use of vernacular 4 Technology The limitations to the knowledge base cease to act as much of a constraint by the 18th century Modern science helps to understand why things work Blind alleys are forestalled It is important to recognize this interaction Often we think of technology as the result of RampD input If demand is great enough innovation will occur But some innovations are beyond our current knowledge 7 there is a demand for AIDS cures or cheap 8 hour laptop batteries but we don t know yet What does knowing mean But the important point is that technology feeds back into science just as science advances technology 1 focusing devices Technology poses well de ned problems for science to look at lt in uences the research agenda further enhancing the value of the science As an example the Newcomen steam engine induced research into the laws governing energy e iciency which led to the laws of thermodynamics 3 arti cial revelation Technical advances in instruments overcomes the limitations of the senses Think of microscopes and telescopes 03 rhetoric of knowledge techniques are not true or false 7 they work or not This is a much more open system An important way to con rm knowledge 7 see if it works Combats doctrinal authority and hierarchy This helps science advance 5 Industrial Enlightenment This is the historical episode where it takes place Like the general enlight enment but more concentrated on the application of ideas to industry and agriculture lt is a movement the quotBacon Programquot Bacon was the rst to regard knowledge as something growing Mokyr sees it as a movement Three purposes 1 reduce access costs to technical knowledge3 2 understand why techniques worked 3 facilitate interaction between those who understood and those who did We see this in scienti c societies and societies of arts Also catalogs and de nition of units and terms Lowered barriers to diffusion of knowledge 3The Industrial Enlightenment realized instinctively that one of the great sources of technological stagnation was a social divide between those who knew things savants and those who made things fabricants To construct pipelines through which those two groups could communicate was at the very heart of the movement 0 Key institutional change made it possible discoveries were no longer hidden but diffused to achieve priority 0 notion of intellectual property came into being 0 How to reward ingenuity rst mover advantage secrecy limited by reverse engineering possibilities subsidies and prizes patents Mokyr s point is that this is the way that quot intellectual changes in the eigh teenth century gradually transformed the way in which institutions affected technology quot Other factors contributed by the key is the quotpositive feedback mechanisms within the sphere of useful knowledge and those between useful knowledge and institutions that changed the course of history 28quot This episode led to the breakout from concavity lt was a peculiar historical episode in Europe that created this industrial enlightenment The key economic question is why has it spread to some regions and not others Why are some countries able to take advantage and in others not To summarize then the lndustrial Revolution had intellectual roots that needed to be met if sustained economic growth could take place just as it had to satisfy economic and social conditions The importance of property rights incentives factor markets natural resources law and order market integration and many other economic elements is not in question But we need to realize that without understanding the growth of useful knowledge the technological elements will remain inside a black box 6 Technological Modernity Why did the notion that knowledge is not stagnant arise Why was there a phase transition where positive feedback arises Evolutionary models predict explosions of change when conditions are ripe as in the Cambrian explosion or the post KT event Most growth models miss what happened lt is not number of scientists4 or changing rates of return to human capital5 Britain was not the most literate place How many inventors and truly technically able people were needed to generate sustained technological progress6 Feedback between prescriptive and propositional knowledge is important 0 steam engine and thermodynamics o optics and batteries for chemical piles7 0 Flight to aerodynamics In addition to the positive feedback within the two types of knowledge one might add the obvious observation that access costs were themselves a function of improving techniques through better communications storage and travel techniques Greater and greater combinations of known things are possible But obviously knowledge only creates opportunity It does not guarantee action Why the decline in negative institutional feedback 7 Institutional Feedback Why did the 18th century also see a slow but uneven decline in negative institutional feedback 4It seems that only a small tail is responsible for the innovations and some deiskilling actually took place initially 5Fertility rates came down much later than the Industrial Revolution 6Production techniques became more modular and standar ized meaning that labor might become more specialized and that each worker had to know less rather than more If much of the new technology introduced after 1825 was like the selfiactori simpler to use if more complex to buildi it may well be that the best model to explain technological progress in the sense of inventing new techniques rather than implementing existing ones is not the mean level of human capital or as modelibuilders have it the level of human capital of a representative agent but just the density in the upper tail of the distribution that is the level of education and sophistication of a small and pivotal elite of engineers mechanics and chemists dexterous motivated imaginative wellitrained technically with some understanding of some of the science involved 7Without microscopes how would Pasteur have refuted the spontaneous generation theory to establish the germ theory Indeed the widespread use of glass in lenses and instruments in the West was itself something coincidental a giant accident possibly a byproduct of demand for wine and different construction technology What happened to make the view that knowledge will be socially useful even though it can be socially disruptive and politically dangerous The Lud dites did rebel but they were not successful everywhere Why Resistance was also important in the medical industry Jenner s discovery was resisted He was told quot not to risk his reputation by presenting to this learned body anything which appeared so much at variance with established knowledge and withal so incrediblequot8 Key process hard to understand Clearly uneven across countries The key event is a change in cultural beliefs that made possible the in dustrial enlightenment Not just cultural beliefs though but also the meta physical beliefs that people held about their environment and nature and their attitudes toward the relationship between production and useful knowl edge lt should also include their cultural beliefs about the possibility and desirability of progress and their notions of economic freedom property and novelty It is interesting to note that this cultural change occurs after many other institutional innovations in Europe such as corporations formal law indi vidualism self governance property rights These elements did not trigger modern growth because they occurred prior to the industrial enlightenment Useful knowledge grows because in each society there are people who are creative and original and motivated by some combination of greed ambi tion curiosity and altruism Yet in order to be translated from personal predilections to facts on the ground and from there to economic growth an environment that produced the correct incentives and the proper access to knowledge had to be there The uniqueness of the European Enlightenment was that it created that kind of environment 8Jenner s famous discovery of the smallpox vaccine ran into the opposition of the inoce ulators concerned about losing their lucrative trade Hopkins 1983 p 83 The source of the vaccine infected animals was a novelty and led to resistance in and of itself Clergy objected to the technique because of the quotiniquity of transferring disease from the beasts of the eld to Manquot Cartwright 1977 p 86 Cartoonists depicted people acquiring bovine traits and one woman complained that after he daughter was vaccinated she coughed like a cow and grew hairy Hopkins 1983 p 84 Despite all this of course the smallpox vaccine was one of the most successful macroinventions of the period of the Industrial Revolution and its inventor became an international celebrity 8 Interaction between Institutions and Tech nology There are interactions and feedbacks between I and T 81 FromT gtI o T creates markets think of reduction in transportation costs improvements in communications military advances created the nation state which helped make rule of law feasible 0 T creates forms of business appropriate to its needs factory system business corporation with separation of ownership and control with ability to raise capital and large xed investments clearly related to the rise of the railroad Chandler 0 Rise to technical universities needed due to advances in technology 82 From I gtT o Institutions can frame the agenda With the wrong institutional setup the brains in society are de voted to economically unproductive ideas Think of Confucian and Talmudic scholars o Institutions affect how we choose between competing theories What is an acceptable proof 9 lf commercial success is rewarded this leads to the rhetoric of knowledge effect mentioned above 9Interesting to ponder the link between open societies and scienti c discovery7 ala Popper o Institutions affect how knowledge is shared and communicated if at all accessible or secret10 do the informed interact with those who can productively use knowledge from the mounts to the fabrictmts ln technologically advanced societies quot intellectuals get dirt under their ngernailsquot o lnstitutions affect whether knowledge will create quottechniquesquot sets up incentives and payoffs for discoveries and applications patents are an obvious notion here 0 Diffusion institutions can block diffusion 7 act as barriers 7 via vested in terests and governments worried about political instability and unemployment fear of technology 7 the EU and genetically modi ed crops hierarchy can impede diffusion ala my paper 7 it exacerbates in centive problems that arise from incomplete information Hierar chy implies greater costs to reveal hidden information hence it is costlier to induce adoption of innovations in a hierarchy o coordination for innovations to be successful coordination may be required A rm may allow for this allowing capital and labor to collaborate to make the innovation work Lack of such institutional arrange ments may defeat it remark lnnovations open doors and institutions invite or prohibit the econ omy to walk through 0 think of the holy inquisition versus current US rewards to innova tors 10numbered copies of economics texts in USSR 9 Institutions and Income Differences If technology or know how was the only thing that explained differences in income it would be easy to teach Zimbabwe to be rich Knowledge can diffuse but educated labor goes the wrong way The key problem is that technology diffuses easier than institutions that are amenable to economic growth 0 note that Iraq and Pakistan are better at importing western technology than e icient institutions This suggests that institutions are critical to crossecountry differ ences 0 but over time Germany is richer now than in 1815 due to differences in knowledge so that is the key for time series differences 0 Notice that because of success we now demand that technology solve problems from AIDS to pollution The industrial enlightenment caused optimism to replace fatalism References l Mokyr Joel quotThinking About Technology and Institutionsquot October 2002 10 Lecture Note on the Growth and Development Econ 570 Spring 2004 1 Some Empirics A key point about economic growth is how recent it is at least in the modern sense To see this note the vivid description of Jones Conservative estimates suggest that humans were already distinguish able from other primates 1 million years ago 1magine placing a time line corresponding to this million year period along the length of a football eld On this time line humans were hunters and gatherers until the agricultural revolution perhaps 10000 years ago that is for the rst 99 yards of the eld The height of the Roman empire occurs only 7 inches from the right most goal line and the 1ndustrial Revo lution begins less than one inch from the eld s end Large sustained increases in standards of living have occurred during a relatively short time equivalent to the width of a golf ball resting at the end of a football eld We thus need to distinguish modem economic growth Kuznets from the pe riod of Malthusian stagnation 1n the latter growth was slow as per capita income growth led to population increase That is why the growth rate of per capita in come is so low prior to 1800 in gure 13 1n the post Malthusian period we see rapid economic growth due to technological change and positive feedback Something clearly changed That is why we think of modern economic growth Should one theory explain both Now some would argue that it is all determined by nature e e g Deemnna whe uguzs thet gengupl nc aetemtnente e unetele the etnhtng d exemes hetween the lungrtexm hetnuee nt penples ntthe d exem canhnems hewe heen due tn d exemes m thee enyhnnnnent 405 39 He hnhe mndexn pmepeuty tn the ennnhtnne tn the emexgence nt gneultme m Nenhthte hmes e Seehe eheee the skung genguphlul pnettnn emphasnmg nheeeee envlmnr nente tunepm enete He empheenee the etmng nle nt eneethnee But theee thenuee dn net exphnn the dumahc changes m the tntunee n egnne neeent pexmds Cnmxdex the hunme tehle guu 1 1 mm Maddmnn whteh enmpeue twn gmups nt enuntuee and hnhe he I 1 the Us Canada Auethe and Japan Gmup B Amzuca Eeten Emnpe and 1751 he excluding Jepen and Afnu MM n the1yeemenweeheemeennewee Figure 11 Nettee thet peneeptt mcnmz 1eye1e ett tn dwexge enmewhee hetween 1mm and 1500 hut thet the hey pexmd e thet Lust Thet e when the d exences heenme veryslgm um Fxgme 1 2 e very netneettye m the xegaxd The hey pntnt hee e thet the nee nt the Weet e yeyeeent One 1weye wente tn hnew why Chum het te enventegev We went tn hnew why Japan h gxnwn en teet smce Meneetnetnn The e whee the Dumnndrtype axgumzms 11 Itcannntexplamhnw1mthm the tempetennee eennenne peunmeene e en yenea Tn etntme the 1eeenne n11 2 we een hnh t peneeptt nutput m Weeten Eumpe h the 1eet 2mm yeene gure 1 a The euggeete hnw eeent the tehenes In nBcu neeemeenmeeh e ennmd te the heenehehenet er wtem Eurap en 18m Th he pnmz er Llrnn n 1139 cm we m eheee the Met n whathx wtn Eump we nehe uthx ene when eexenuetten enee 1t n thet t we Figure 12 China and Western Europe 400 to 1998 really is The sharp impression from gure 13 is that of a sharp break There is a fundamental non convexity here that occurs somewhere in the 19th century It is the most dramatic qualitative change in human history since at least the start of the bronze age Another way to think about this is cross country at a point in time In 1960 for example for 113 countries for which gdp per capita was available at PPP the riches country Switzerland 14980 was 39 times higher than the poorest Tanzania 381 The mean was 3390 1996 dollars The richest coun try was 39 times the poorest By 2000 we have date for 150 countries and the dispersion has grown as has the mean 8490 The richest Luxembourg is 91 times that of Tanzania 482 The US is second 33350 and this is 69 times Tanzania So dispersion has grown std deviation of the log of gdp increased from 089 to 112 This about this If Tanzania were to grow at the long term US growth rate of 18 1870 to 2000 it would take 235 years to reach the 2000 level of US gdp At the Japanese growth rate of 275 it would take 154 years 2Actually the Democratic Republic of Congo is poorer but there is no data 18000 16000 14000 12000 10000 8000 6000 4000 2000 0 250 500 750 1000 1250 1500 1750 2000 F gmc 3 Output PCI391VU m New Europe m the year 0200 Data xmvrcc Mmllliwn 200 Figure 13 PereCapita Output in Western Europe Comparing 1960 with 2000 we can see that 16 countries had negative growth rates of real pericapita gdp DRC Central African Republic Niger Angola Nicaragua Mozambique Madagascar Nigeria Zambia Chad Comors Venezuela Senegal Rwanda Togo Burundi Mali All but Nicaragua and Venezuela were in subesaharan Africa Of course if not for missing data there would be more data are more likely to missing from those that do worst a one thing to note about this list is that none of these countries with negative growth rates is rich they are all relatively low income a the oil shock increased the variation in growth rates as it reduced average levels I another thing is that if we had split the sample 1960772 and 19732000 only one country had a large 73 negative growth rate in the rst period Burundi several in the second Nicaragua over 74 and four over 73 a total of 32 countries had negative growth rates for this subesample Several countries also had growth rates over 5 during this period and re ally moved up in the rankings Taiwan Singapore South Korea Hong Kong Botswana Cyprus Thailand China has done real well for half the period So the differences in performance over 40 years are huge The problem is to explain why Let us start with some of the new stylized facts of growth SF 1 Factor accumulation does not account for the bulk of crossicountry di eri ences in the leuel or growth rate of GDP per capita Rather it is TFP whateuer that means TFP growth is the great free lunch With the same inputs we get more output 1n the Solow model it is really manna from heaven Now we want to endogenize it But rst we need to document its importance 1 1 Lucas Paradox The key question of growth and development economics is how to combine the notion of increasing returns 7 which is critical to raising per capita incomes 7 with diminishing returns which is the key to explaining allocation The Solow model explains this with exogenous technical change But this is unsatisfying for economists precisely because it is exogenous The Lucas paradox is a good way to start thinking about this Consider the standard production function in intensive form y AM 11 TFP is represented by A Countries differ not only in their capital labor ratios but in their levels of productivity as well Suppose this were not the case 1ndia s per capita income is about that of the US 1fA1ndm AUS it follows that 15 55 kI ndia Now a good estimate of 6 capital s share of national income would be 04 a rough average of the two countries This would imply that the capital labor ratio in the US is 152395 m 871 This is obviously way too high It would imply that we save at a rate 800 times that of India Since our savings rate may be 17 we know this cannot be true Moreover if the capital labor ratio were really 5 this much higher in the US than in 1ndia the return to capital in 1ndia would be about 58 times higher Why To see this note that if we ignore A the marginal product of capital per worker is 7 k l From expression 11 it follows that k yl Now using this in the expression for 7 we obtain 7 6y g1 Since yUS 15 yjndm we have Tjndl39a 71515 Now 151395 is about 58 so the rate of return would have to be 58 times higher in 1ndia than the US3 But this should mean that capital should ow from the US to 1ndia Some does but not that much4 Why One reason could be TFP differences Ajndm lt AUS would alter the rate of return calculation5 Explaining these differences is one of the most important issues in development economics But we will ignore them here for the most part Notice that because it is measured as a residual it is really many things that include real technological advances real cost reduction Harberger improve ments in institutions and policies This is an exercise in growth accounting Some results across regions are given in table 14 Note the importance of TFP growth in the advanced countries and the lesser role in the Tigers The results of Alwyn Young called this into question for Asian Tigers His well known claim is that factor accumulation played a much larger role in these cases This immediately raises the question of why they did not succumb to the extensive growth trap that the Soviet Union could not escape from One explanation might be that capital owed primarily into exporting sectors of the economy Since these were small open sectors they did not encounter diminishing returns A second point could be that nancial institutions prevented really bad investments this was easier to say before the Asian crisis But the interpretation of his results may be wrong 0 First interest is in per capita growth not output growth 3To see this note that if we ignore A the marginal product of capital per worker is 7 k l From expression 11 it follows that k y1 S Now using this in the expression for 7 we obtain 7 y I Since yUS 15 96 dem we have 71mm TUS15 Now 151 5 is about 58 so the rate of return would have to be 58 times higher in India than the US 4This is sometimes referred to as the Lucas Paradox Robert Lucas rst pointed out that capital ows to developing countries were too small compared with predictions of standard economic models 5You can see this by taking the opposite assumption TUS 71mm and letting differences in A explain the higher US output vanc mm mm munnnn mm In mm newquot a WWW snmcunnmumm mm mm 1 nzcnmi Prince 0 SA m 55 1 my one ssm lt1 a 55 m me am 34 2 m umquot me 2 m m m unumwnvanm m am m m m mum um mm m m m m Prince 2 3m 55 m m Gummy n am m s m m 22 mm m s 45 mquot 2 m m u 294 Unwmxmvanm 3n 2m m 4M 52 Umxausme u mM m m m We mum mm as am 42 g m mu m 540 w m m cm n52 am 4 2 m Mencn can 5m m 23 m vennu Mg 520 m m w asusmvssr m Hana Knny n 7 1 2m 3m mm a 53 a m m 7w x m mm m m mm m m m ozmrwmsmm mamquot Mm nndJnWmvsnn mammahnwmvlyuawl mm mm was m ms mm mm ersvmm mm mm mm 1 4 Growth Accoummg 0 Second and much more important some capital accumulation is induced by productivity growth Suppose y Aka then if there is a productivity shock 7 ie a shock to A then the marginal product of capital increases and this raises capital accumulation But this should be attributed to produc tivity growth not factor accumulation But if growth in capital is taken as exogenous then this is ignored To see the problem more clearly suppose that output and capital are growing at rate so Using growth accounting we attribute act of the steady state output growth to capital and 1 7 158 to TFP Now in the standard Solow model with exogenous technical change in the absence of TFP output and capital do not grow at all So in a elem sense we are under valuing the role of TFP in generating this growth It only occurred due to TFP The problem is that the growth of capital is endogenous but growth accounting treats it as erogenous Now if TFP itself is truly exogenous then one could say that differences in TFP show up one for one in differences in output growth This could help explain the difference with the Soviet case In that case factor accumulation was high in the face of negative productivity shocks 7 the economy was becoming less ef cient In the Asian Tigers case there were positive pro ductivity shocks that allowed capital accumulation to proceed at very fast rates without reducing returns Growth accounting exercises thus suggest that TFP plays the major role in explaining per capita output growth This is evident in gure 15 from Easterly and Levine Notice that TFP is more important in the fastest growing countries How to interpret the group of countries with positive growth in capital per worker but negative TFP growth and hence negative output growth Clearly it is not technical regress More likely it is very bad policies which cause inputs to be used much less ef ciently than before Increases in transaction costs would be one example What happens when human capital accumulation is added to the investigation Very little 1 1 1 Variance Decomposition Easterly and Levine perform a variance decomposition to see how much of the variation in the growth of per capita output y is explained by factor accumu lation versus TFP growth Start with y AW and let 04 04 Taking logs s N nmum tcnu quotrt mumquot nmrh Run u Rm ulnnl u ax u at u m u 02 u w my ruu Slunr an m rum mum mum mo mum uhdk Flgure 1 5 Growth Accouhtrhg by Country Groups and deferentJatmg w1th respect to tlme we have the farmhar growth accountmg expresslon y A k 0 4 y A 10 so the Vanance of output growth can be wrltten as cor va r g 0 42 var 20 4 00v 3 12 One can a1so conslder hurhah capltal by lookmg at factor accumulatroh per worker 1 f de ned as quot2 where h 15 educatlonal attamrheht per worker Expresr son12lsnow cor cor g 0 72 var 20 7 my estlmatlons Notlce that TFP The results are glven 1n tah1e 1 o for both growth accounts for almost 60 of the Varlatlon 1n growth rates and that factor ccumulatlon a1ohe 15 hot that rrhportaht at 1east 1n the 198071992 perlod Yahle 2 Vanancu Ducnmpusman I mm human annual Cnmnhullnn n sn nnnrull cenmnes ampk gm cavl mvk glkll no 195m ssz u as n m b 195D4992 u 55 n 21 n 13 H mm human capnal Cumnbmlnn n qukm glknl mvlgmvkn qlxml a 19634592 my net 7045 n 52 b 19En4587 50 n 20 m2 Figule l 5 Vazlance Decomposluon 2 Development Accounting Development eeeounung ls lne analog f0 levels co gzoWLh eeeounung The latte shows how lnpuc glowln and My gzoWLh detezmlne dlffezences In gzoWLh laces explaln dlffezences 111 levels of lncome Suppose we zelace lncome co factoxs and emeleney X FL E If factozs explaln dlffexen 5 15 L ees then me pzoblem fox developmenc 15 co explaln low seenrnuleuon e lln n a 5 and how no cazefully measuze lnpuls T mk of Solow zesldual When Solow st esmnsceol lns glowth equellon mosc onne was due no Lechnologlcal change A w measuxe mp d npucs Slmlla e may be anneal fox developmenc 1 ap oach asploneeze by Hall nd Jones Hall and Jones 7 showthat clossrcountz plodu m dlffexencesmeas lnlevelselelslgesndesnnoc beexp n human Lec oucpuc be ploduced accoxdlng co Coberouglas plodueuon functlon x Kf39AH 1 2 l 10 where Hi is the amount of human capital augmented labor used in production and Ai is a labor augmenting measure of productivity The former is given by Hi e E Li where the function re ects the ef ciency of a unit of labor with E years of schooling relative to one with no schooling O6 The derivative gt E is the return to schooling estimated in a Mincerian wage regression an additional year of schooling raises a worker s ef ciency proportionally by Z We can re write 1 in terms of output per worker K i y lt gt H where h E f The nice thing about 22 is that we can use it to decompose differences in output per worker into differences in capital output ratios levels of human capital and levels of productivity7 Hall and Jones use data from 1988 to measure productivity differences using Penn World Tables They correct for natural resource endowments by subtracting value added from mining from GDP This prevents Saudi Arabia from being the world leader in productivity8 They use a value of 04 g and assume that Z is 6The rationale for this functional form is as follows Given our production function7 perfect competition in factor and good markets implies that the wage of a worker with s years of ede ucation is proportional to his human capital Since the wageeschooling relationship is widely thought to be logelinear7 this calls for a logelinear relation between h and s as well7 or some thing like h E with ab a constant However7 international data on educationewage pro les Psacharopulos7 1994 suggests that in Sub7Saharan Africa which has the lowest levels of ede ucation the return to one extra year of education is about 134 percent7 the World average is 101 percent7 and the OECD average is 68 percent Hall and Jones s measure tries to reconcile the logelinearity at the country level with the convexity across countries 7Notice the use of capital output ratio rather than capitalelabor ratio This follows the lead of David 19777 Mankiw et al 1992 and Klenow and RodriguezeClare 1997 in writing the decomposition in terms of the capitaleoutput ratio rather than the capitalelabor ratio7 for two reasons First7 along a balanced growth path7 the capitaleoutput ratio is proportional to the investment rate7 so that this form of the decomposition also has a natural interpretation Second7 consider a country that experiences an exogenous increase in productivity7 holding its investment rate constant Over time7 the country s capitalelabor ratio will rise as a result of the increase in productivity Therefore7 some of the increase in output that is fundamentally due to the increase in productivity would be attributed to capital accumulation in a framework based on the capitalelabor ratio 8Is this procedure really correct This rationale is inherently dubious then why not sube stracting the value added of agriculture and forestry7 that also use natural resources abune dantly piecewise linear9 This analysis shows that productivity levels and output per worker are highly correlated Figure 21 plots these in logs l39igmv l l39l H39thH39llthHHHHHllv 1llvl w u 64 WW W in quot5 is My M Wm com a mu SMFW unzx mun mm mm gran mum xznon emuxwwm mm m mm 3 Dullavsl Figure 21 Productivity and Output Per Worker From table 22 we can see the role of productivity differences in explaining output differences Notice that the values are ratios to US levels To see what goes on in this table look at the row for the Soviet Union In the Soviet Union investment was extremely high as was the capitalioutput ratio In addition human capital intensity was also high But it also had a rather low productivity level For the developing countries in the table differences in productivity are the most important factor in explaining differences in output per worker For example Chinese output per worker is about 6 percent of that in the United States and the bulk of this difference is due to lower productivity without the difference in gWith respect to human capital Psacharopoulos 1994 surveys evidence from many countries on returnitoischooling estima es Based on is summary of cerian wage regressions we assume that is piecewise linear Specifcally for the rst 4 years of education we a lime a rate of return of 134 percent corresponding to the average Psacharopoulos reports for sub Saharan Africa For the next years we assume a value of 101 percent the average for the world as a whole Finally for education beyond the 8th year we use the value Psacharopoulos reports for the OECD 68 percent quot11111111 1 1111111111i1 391111111139111111S39 R1111m111 US 711111 Conmb 11111111 11 mi I L 111111111 YL I139397 quotquot 1 111111211 51111 111m 1 U170 11102 1 U170 01111 11 u 911 1 1102 n 9113 1 1111 111117 mm 1 11621 mm 1 227 1 11111111111 11 3111 1 118 n 3112 11912 13111112 11 311 1 111 11 Gun 1 1115 111111511 1111g1111111 11727 131191 0 3113 1 D11 1 mg 1111 n 7111 1 1 1a Smgnpme 1 m1 1 D78 11111111 1 119 u 638 ME 1 mm D 925 1gs11111111 u 953 0518 s 12 1 231 11 468 111 u axe may 1 7151 u 251 1111111 1 11110 n 1191 111112 0 1111 1 0 1155 u 117 01151 u111lt Zmre 11 013 u 199 1111011 11 150 A1 Erags 127 1111111119 11295 11352 5111111111111 Devmtinn 2611 u 211 unelminu 111 17 110g1 1 11110 n 5211 C nnelatmn 111 1 11ugs1 a m n 2111 Note The 9191119111111111111 table are leE e111p111m1 1111113111111 111 HIE 1111111011111 11139 equalmn 111 111 111911111911 11 11111111 111 111 11s 11117 11 T11 1 1 HIE 11111 c1111111111 111 11111 is the 11mm 111112 nthel thee Figure 22 Productivity Calculations productivity Chinese output per worker would be more than 50 percent of US output per worker The bottom half of Table 1 reports the average and standard deviation of the contribution of inputs and productivity to differences in output per worker According to either statistic differences in productivity across countries are sub stantial A simple calculation emphasizes this point Output per worker in the very countries in 1988 with the highest levels of output per worker was 317 times higher than output per worker in the very lowest countries based on a geomet ric average Relatively little of this difference was due to physical and human capital differences in capital intensity and human capital per worker contributed factors of 18 and 22 respectively to the difference in output per worker Produc tivity however contributed a factor of 83 to this difference with no differences in productivity output per worker in the very richest countries would have been only about four times larger than in the very poorest countries In this sense differences in physical capital and educational attainment explain only a modest amount of the difference in output per worker across countries The reason for the lesser importance of capital accumulation is that most of the variation in capital output ratios arises from variation in investment rates Aver age investment rates in the very richest countries are only 29 times larger than average investment rates in the very poorest countries Moreover this difference gets raised to the power which for a neoclassical production function with 04 13 is only 12 so it is the square root of the difference in investment rates that matters for output per worker Similarly average educational attainment in the very richest countries is about 81 years greater than average educational attainment in the very poorest countries and this difference also gets reduced when converted into an effect on output each year of schooling contributes only something like 10 percent the Mincerian return to schooling to differences in output per worker Given the relatively small variation in inputs across countries and the small elasticities implied by neoclassical assumptions it is hard to es cape the conclusion that differences in productivity the residual play a key role in generating the wide variation in output per worker across countries 21 Caselli Decomposition Caselli alters the decomposition Rather than use 22 he uses 34139 Aikahlia E Aiykh 23 The reason is that the impact of differences in A are cleaner 7 the Hall and Jones measure is not invariant to differences in tfp since A affects 3410 This measure allows one to ask quite strictly what would the world income distribution look like if all countries had the same level of A Using 23 we can decompose the variance in log y as va log y var logykh varlogA 20011 logA logykh 24 If all countries have the same level of A then the varlog A 0 thus one measure of success would be var logykh successl va log y Caselli uses PWT60 to estimate this counterfactual and nds var l0gykh 05 and the observed va log y 125 so successl 04 The only problem with this measure is that it is sensitive to outliers so it may be useful to look at inter percentile differences He de nes 34229119 yQOy10 where ya is the cum percentile of the distribution of y In the data the value of the 31223115 7 and the value of 31903110 20 so the value of SUCCESSZ 035 Notice that with either measure the variance of log y is much greater than the variance of log ykh which is why there is so much interest in TFP differences But it is interesting that there is also some signi cant variation in gym as well We will return to this when we talk about the relative price of investment for example successZ 2 1 1 Subsamples It is interesting to look at subsamples Consider table 23 We observe that the variance in per capita income is lower in richer countries 7 this is obvious In Africa 7 the poorest regions 7 the variance is highest More important it is clear that the simple model explains richer countries better than poorer countries compare the above and below median success variable for example The one apparent puzzle is Europe 7 but that is entirely due to the inclusion of Romania which has high human capital but low per capita income It has the typical transition legacy If you drop Romania Europe looks like the Americas 10Obviously m h kahl aA o suwmample Obs ar ogy V39ar ogy1H success Above the median 47 o 174 112 o 643 Below the meampmn 46 n ass 0 257 o 438 OECD 24 0 077 n 047 a 6 06 NonrOECD 59 1 004 0 373 0375 Africa 26 o 843 0 27 o 322 mama 25 0 844 n 175 a 513 Am and Oceania 25 n 632 o 292 o 448 Europe 17 0 128 m 033 0255 All 93 1 246 0 5m 0 400 gure 2 a Success In Subsamples One conclusion from this exercise is that the factors only model works worst where we need it most 7 in the poorest regions 22 Chipping Away Can we chip away at the differences That is can we improve the t of the factor only model to reduce the impact of A Notice rst that one important variable is 04 Since countries differe more in terms of k than in terms of h if you have a higher capital share you can explain more of the cross country differences lf capital received 60 of income you could do pretty well But this is just too high 221 Human Capital Adjustments There are various issues here First what are we really measuringThere are two views of human capital 1 typical view resource cost incurred in learning the world s stock of knowl edge ie as a factor of production 2 alternative view human capital allows one to absorb new techniques at lower cost Nelson and Phelp 1968 On this view schooling is learning how to learn not that much of what is learnt is useful Hard to say which view is correct Econometric evidence indicates that human capital has negligible effects on productivity but is positively associated with improvements in productivity11 On the second view low human capital impedes the ability to follow But perhaps causation runs the opposite way when the conditions exist for entrepre neurship and innovations there is a large demand for human capital to implement these innovations Klenow conducted an interesting test of these two hypotheses using US data on manufacturing industries He noted that if the rival human capital story is correct then those industries where labor intensity is high should have higher productivity growth Higher labor intensity would signal more human capital accumulation If the Romer idea is correct on the other hand then industries with 11In Communist economies human capital was high and productivity was low but so was the growth rate of productivity low labor intensity and high capital intensity would have the higher productivity growth Klenow tested this using growth rates 1959 1991 for 449 4 digit US manu facturing industries He found that TFP growth is faster in the industries that are more intensive in capital and intermediate goods and less intensive in labor 7 favoring the idea models A second big issue is measurement First results are sensitive to how M R W used secondary school enrollment This varies more across countries than other measures so it explains more Klenow and Rodriguez Clare added primary school enrollment for example and this re duced the explanatory power of human capital They also constructed an index that used returns to schooling with years of schooling This also reduced the explanatory power of human capital Aside from this we have practical issues First not all human capital is iden tical Education Most adjustments to education do not account for the differences For example hours worked varies across countries is inversely related to per capita income So accounting for this would actually raise labor input in the poor countries and leave more to explain Unemployment rates could differ but there seems to be no pattern Of course underemployment may differ but again this is probably higher in the poorer countries so again it does not help 23 Quality The quality of human and physical capital may differ Perhaps this can explain some of the difference It could be that h is lower in poor countries but this is hard to measure and it does not seem to explain much The most important element could be health and nutrition This could indicate that labor input is less in poorer countries Weil 2001 uses as a proxy for health the Adult Mortality Rate AMR which measures the fraction of current 15 year old people who will die before age 60 under the assumption that agespeci c 12Klenow uses an interesting analogy This paper could have been typed on a 1970 typewriter Correcting spelling errors would have been tedious using the human capital accumulated to type Even using the human capital accumulated in the PhD program But the word processing program made it trivially easy No change in typing human capital but ideas embedded in the pc and in the software dramatically improved productivity 18 death rates in the future will stay constant at current levels In practice this is a measure of the probability of dying young and is therefore a plausible inverse proxy for overall health status Suppose then that h Ahe E as before but that A e amrAMR 25 where bum lt 0 since a higher adult mortality ratio means a less energetic work force Using Weil s preferred value for foamAcclOO is 168 This allows for a big improvement in explanatory power of the factors only model 7 by at least one third Weil s estimate puts a high value on health 7 equal to a year s human capital and thus a year s human capital worth of wages13 ls this too high Hard to tell 24 Social versus Private Returns Note that the estimates of Z that are used measure private returns to schooling but what is important for growth are social returns What if there are externalities from a more educated workforce The question is which way do they go 0 what if rent seeking is higher in the poorer countries Governments may employ graduates in poor countries to a greater extent than in rich countries Then the social return might be lower in poor countries This could mean that h is higher in rich countries and could explain income variation 3 Quality of Physical Capital It could be that physical capital varies in quality across countries Since most countries import capital goods from a small nulnber of countries it is possible to use imports of capital as a proxy for investment Furthermore the RampD content of investment goods differs as well so this may proxy for the quality of capital goods 13Weil uses published micrdlevel estimates from three developing countries to infer the elasi ticity of human capital to height He then uses time series data from Korea and Sweden to estimate a relationship between height and the AMR He then combines these two pieces of information to infer the elasticity of human capital to the AMR In essence he is using the AMR to predict height and then applies to the predicted height the microeconomic estimate of the effect of height on wages Suppose that 1 20cm l 7 lt1 31 101 YB where x are intermediate goods and B is a TFP term Suppose further that intermediate goods are produced by 17 A prpyi pr 0 lt 0 lt1 32 where hpr is human capital augmented labor in sector p and AP is TFP in that sector The key assumption is that capital is heterogeneous there are P distinct types of capital and each type is product speci c in the sense that intermediate p can only be produced with capital of type p The assumption that I lt 1 implies that 7 in producing aggregate output 7 all these activities are imperfect substitutes Notice also that AP is product speci c 7 this implies that the embodied technol ogy content of good p may be greater because the industry producing equipment of type p is more RampD intensive Within a country the law of one price should be enough to insure that we are measuring physical differences But in a cross country setting we would also have to worry about the relative price of investment which is differs across countries Now suppose that we can let hp h for all sectors 7 labor is mobile Then it is possible to write 31 as 17 17a P r Y Kawml aB Zwm c 33 171 where 1 is the share of the capital stock in sector p The important point about this equation is evident if we compare this with the expression we have been using to explain output differences something like Y AKO hL1quotquot It is evident that 33 provides an expression for A in terms of the composition of the capital stock This suggests perhaps that variation in equipment shares could imply variation in the quality of capital and this could over and above variation in the quantity of K explain income variations The only problem is that if you look at expression 33 you can see that it is extremely sensitive to variation in 7 And it is hard to know what the proper value is 7 recall this measures how di icult it is to substitute different types of capital 20 Hence it is hard to know if capital differences account for income variation It could be the case but with current data it is still too hard to tell But it is a promising idea 31 Public vs Private Capital Another important point is that capital may differ in its productivity if it is the result of private or public investment PWT does not distinguish Government investment may not be as productive 7 rent seeking Or it could be more produc tive if there are large externalities 7 ala Aschauer The important point however is that in poorer countries governments play a larger role in investment decisions It poorer governments have more corruption or less effective cadres then we might expect capital to be less e icient in poorer countries If the data existed one could recalculate the capital stock in the manner of Kt Iprivaiei VIpublici 7 6Kt71 but it is hard to nd such series Moreover you would have to de ate them appro priately for purchasing power We will discuss some evidence on the productivity of state investment a bit later 0 Notice that what this exercise would be doing is to peer into the institutional differences that account for differences in A since one reason we expect corruption to matter is the e iciency of investment 32 Sectoral Differences Caselli also looks at sectoral and industry differences For example TFP could differ across industries and countries differ in the mix Similarly TFP could differ across sectors and poor countries have larger agricultural sectors But still one would want to know why the sectors or industries have different TFP s It is worth noting the interesting nding of Gregory Clark with respect to cotton mills He examined the productivity of cotton mills around the world in the early years of the twentieth century He shows that assuming constant capital labor ratios the textile industries of Britain and New England would have had a huge cost disadvantage relative to India Japan and many other countries Yet British cotton textiles dominated export markets Clark shows that the various countries industries used identical equipment and that the expertise to organize and run the mills could not have differed too much Rather the source 21 of the productivity differences boils down to the fact that each English worker was willing to tend to a much larger number of machines In low productivity countries workers were idle most of the time Why this was so remains a bit of a mystery and one shouold be cautious in assuming that this nding would still hold up one century later Nevertheless Clark s ndings reinforce the case that labor practices may be an important source of observed differences in productivity This is clearly somehow an institutional difference but we need models to understand how exactly 33 I ntative Conclusion With the evidence to date development accounting still shows that TFP accounts for most of the differences in income variation across countries It could be that low substitution of capital types or of capital for human capital explains a lot but not at the current level of knowledge 4 Growth and Externalities here are several key stylized facts of cross country growth that indicate the im portance of spillovers The growth slowdown that began in the mid 1970s was a world wide phe nomenon It hit both rich countries and poor countries and economies on every continent Richer OECD countries grew much more slowly from 1950 to around 1980 despite the fact that richer OECD economies invested at higher rates in physical and human capital Differences in country investment rates are far more persistent than differ ences in country growth rates 0 Countries with high investment rates tend to have high levels of income more than they tend to have high growth rates