Mankiw, Romer, Weil, QJE 1992
Transcripción
Mankiw, Romer, Weil, QJE 1992
Does the Solow Model Explain the International Variation in the Standard of Living? (Mankiw, Romer, Weil, QJE 1992) Solow Model with Technological Progress • Treat productivity (A term in our model) as due to level of technology. • Assume that technology is labor-enhancing. • So, can view labor input in terms of “effective units of labor.” • Assume growth rate of technology is exogenous and equal to “g.” Solow Model with Technological Progress • Similar to way we modeled human capital, except now we assume this “quality” feature is due to technology. • Later we will enter human capital into the model as a separate input. • Specification allows production function to be expressed in terms of output per effective labor units and capital per effective labor units. data. growth, focus on the model model'stakes implications the ratesforofcross-country saving, population Solow's and technological progress as exogenous. There are two inputs, A. The Model capital and labor, which are paid their marginal products. We of saving, model takes the rates population growth, ECONOMICS OF so JOURNAL QUARTERLY 410Solow's a Cobb-Douglas production at time function, assume production and technological progress as exogenous. There are two inputs, t is given by We s, is andmodel which arethat products. paida their marginal labor, assumes capitalThe ECONOMICS of output, OF fraction QUARTERLY JOURNAL constant OF 0 < so a.production <ECONOMICS 1. 410 a Cobb-Douglas Y(t) = QUARTERLY K(t)a(A(t)L(t))l(1) at time function, assume productionJOURNAL invested. Defining k as the stock of capital per effective unit of t is given by is standard: Y is output, K capital, L labor, and A the The notation per effective unit of the level of as output, and yof k = KIAL, labor, is s,0tooutput fraction a(1) constant odel assumes that level fraction of a constant that assumes model The L A at output, s, is exogenously of and are assumed grow technology. = < a. < 1. Y(t) K(t)a(A(t)L(t))llabor, y = Y/AL, the evolution of k is governed by ECONOMICS OF JOURNAL QUARTERLY 410 of unit per effective of capital stock as the Defining k n and g: rates unit of perAeffective of capital the stock k Yas invested. K+ L labor, and the output, capital, The notation Defining is standard: issy(t) = + 8)k (t) g (n k(t) (4) per effective output level KIAL, and y as the (2) unit of per effective of output level the y are = KIAL, and =as kmodel L labor, ()entunit L A toof exogenously ofof and assumed grow (t) technology. level s, atis of output, fraction a Lconstant that assumes The = + + 8)k(t), g (n sk(t)0 rates k isn and governed Y/AL, the evolution of k is governed =g:Y/AL, the y Defining labor, per effectivebyunit of of of capital = A (r)ent. as the kby invested. Aevolution (3) (t)stock unit of that k per effective asLthe k =8 KIAL, labor, =level (4) implies Equation depreciation. ratey of is the and where L of output ()ent (2) number of effective units(t) A- (t)L at rate of labor, + grows (t) n + g. g + 8)k (n(t), k(t) = (4) +to 8)k (n y+ =gY/AL, k(t) = sy(t) -The by ofsy(t) k isk* governed evolution the(t) labor, sk *a = (n + g + 8)k *, by defined value a steady-state converges = A (t) A (r)ent. (3) or(n + g + 8)k(t), = sk(t)0 (4) 8)k+ (t)g + 8)k(t), (n + g -+ (n k(t) = sy(t)= - sk(t)0 -The number of effective units of labor, A (t)L (t), grows at rate n + g. = sk(t)0 - (n + g + 8)k(t), k* =of [s/(n + g + 5)]1I(1-a) (5) depreciation. 8 is the rate where that k Equation (4) implies that k (4) implies Equation the rate of depreciation. k +rate implies =to (n g + 8)k *, by sk *athat k* of depreciation. rate defined value 8 is the a steady-state toby converges the related is ratio The steady-state *, (4) positively +Equation skcapital-labor *a = (n + g 8)k k* defined o a steady-statevaluewhere + g + 8)k *, by sk *a = (n growth. k*rate defined a steady-state converges or of population to the negativelyvalue of savingtoand or The central predictions of the Solow model concern the impact k* = +[s/(n + 5)]1I(1-a) + greal (5) k* = [s/(n growth income. Substituting (5) on and population g + 5)]1I(1-a) ofg saving (5) k* = [s/(n + + 5)]1I(1-a) we find logs, takingispositively and function the production intosteady-state to the rate positively related capital-labor The steady-state ratesteadyto thethat related is ratio ratio capital-labor The is income capita per state rate to the positively related is state capital-labor ratio growth. of population ratethe to the to andand negatively ofofsaving growth. of population rate negatively saving OF ECONOMIC GROWTH THE EMPIRICS impact411the411 theconcern concern the Solow ofECONOMIC predictions central growth. of population nd negatively to the rateThe OF GROWTH THE EMPIRICS impact model Solow of themodel predictions (6)tl [Ot]ln()gt central The (5)8). income. Substituting on 1realIn(s) growth and population of(6) saving In + + 1 + + ln(n = A In (0) g gt impact the concern growth model Solow ntral predictions of the income. Substituting on real andendowments, population ofbut saving so andsteadyon; it may (5) institutions, that we find taking logs, function andclimate, production theresource into so it may and on; climate, institutions, endowments, but resource that steadytaking (5) arelogs, function production income. the into on real nd population growth that We assume across countries. therefore differ is Substituting income capita per state marginal theirfind paid we thatand factors assumes model the Because assume that across countries. therefore differ isWe income capita per state that we findonly taking and oduction function the steadysigns not predicts itlogs, products, = a +but (6)tl [Ot]ln()gt E, also the magnitudes of lnA(O) + g + 8).Specifically, 1 + 1 + ln(n = A In In In(s) (0) (6) gt = and growth. a + population E, saving the coefficients onlnA(O) e per capita is (6)tl [Ot]ln()gt E is a country-specific log shock. Thus, where acapital's and is a constant the oneln(n third, in +income is roughly share (a) In(s) because - 1 marginal + + + 1 8). = A In In (0) g (6) gt E log shock. Thus, is a country-specific where a constant and a is their paid are capita that factors per model theper Because incomeimplies at a given capita with respect to the income Ot]ln()gt elasticity of time-time an assumes model 0 for simplicity-is also the magnitudes of the notln(n only it predicts for -ofa1approximately + g signs + 8). 1 capita = In A (0) +income In(s) time-time per given gt + products, 0but to 0.5 and ansimplicity-is elasticity with respect marginal rate at saving paid their that factors assumes model the on Because Specifically, growth.are and population saving coefficients the -0.5. + g + In n(7) 8 of approximately a+ in income In(s)-(a) In(n+g+8)+E. third, roughly share because capital's magnitudes of the the butonealso theis signs not only predicts it products, In a+ In(s)(7) In(n+g+8)+E. marginal their paid are factors e model assumes that the to respect with income capita per elasticity of an model implies Specifically, and population ingrowth. onbasic saving coefficients the this section. empirical is our Equation to with respect 0.5 and an specification elasticity of(7) rate but approximately saving of magnitudes the also B.signs Specification t predicts not only thebecause third, inspecification roughly income share capital's inisthis of saving(a) and are the section.one the rates empirical population growth is our basicthat assume Equation -0.5. n + g(7)+We 8 of approximately Specifically, growth. population support to the ients on saving and dataproduction the the is whether consider natural The respect with capita pershifting elasticity of income implies factors ofanquestion country-specific independent of to and are Wemodel that the rates population growth assume saving of the determinants concerning model's Solow predictions the the one third, is roughly income (a) pital's share in to with 0.5 and an elasticity of approximately saving n are the of e.respect that s and This independent function. That is, we assumefactors production shifting ofrate country-specific independent Solow Model with Technological Progress • Assume g and δ (depreciation rate) are constant across countries. • A(0) term reflects not only technology but resource endowments, climate, institutions, etc., so it may differ across countries. • Identifying assumption is that the country- specific shock is independent of s and n, so can estimate using OLS. Solow Model with Technological Progress • Coefficients on ln(s) and ln(n + g + δ) should be equal in magnitude and opposite in sign. • Using factor share of 1/3 for capital and 2/3 for labor, the coefficients should equal 0.5 and -0.5. • Assume value for g + δ equal to 0.05 and use data on population growth rates to compute ln(n + g + δ). Estimation Results forTextbook Solow Model • Coefficients on investment rate and population growth have the predicted signs and are significant in two of the three samples. • Equality restriction can’t be rejected. • Regressions explain large fraction of the cross-country variation in income per capita. 414 QUARTERLY JOURNAL OF ECONOMICS TABLE I ESTIMATION OF THETEXTBOOK SOLOWMODEL Dependent variable: log GDP per working-age person in 1985 Sample: Observations: CONSTANT ln(I/GDP) ln(n + g + 8) H2 s.e.e. Restricted regression: CONSTANT ln(I/GDP) - ln(n + g + 8) 1?2 s.e.e. Test of restriction: p-value Implied a Non-oil 98 5.48 (1.59) 1.42 (0.14) -1.97 (0.56) 0.59 0.69 Intermediate 75 5.36 (1.55) 1.31 (0.17) -2.01 (0.53) 0.59 0.61 OECD 22 7.97 (2.48) 0.50 (0.43) -0.76 (0.84) 0.01 0.38 6.87 (0.12) 1.48 (0.12) 0.59 0.69 7.10 (0.15) 1.43 (0.14) 0.59 0.61 8.62 (0.53) 0.56 (0.36) 0.06 0.37 0.38 0.60 (0.02) 0.26 0.59 (0.02) 0.79 0.36 (0.15) Note. Standard errors are in parentheses. The investment and population growth rates are averages for the period 1960-1985. (g + 8) is assumed to be 0.05. Three aspects of the results support the Solow model. First, the coefficients on saving and population growth have the predicted Estimation Results forTextbook Solow Model • Impacts of saving and labor force growth are much larger than model predicts since value of α implied by model for intermediate sample is 0.59 (SE = 0.02), much larger than the capital’s income share of 1/3. • Growth (development?) Accounting shows adjusted R-squared of only 0.28 in intermediate sample. Solow Model with Human Capital Accumulation • Add human capital to production function. • Include additional transition equation for adjustment of the stock of human capital. • Solve for steady-state values of k and h. analogous to those in Table I to see whether proxies for human capital can resolve the anomalies found in the first section.7 A. The Model Let the production function be A. The Model (8) Y(t) = K(t)H(t)P(A(t)L(t))1-a-, Let the production function be where H is the stock of human capital, and all other variables are (8) Y(t) = defined as before. Let SkK(t)H(t)P(A(t)L(t))1-a-, be the fraction of income invested in physical and Sh of thehuman fraction invested in all human where Hcapital is theTHE stock and capital, othercapital. variables EMPIRICS OF ECONOMIC GROWTH 417Theare evolution of the economy is determined by GROWTH defined as before. Let Sk be the fraction of income invested THE EMPIRICS OF ECONOMIC 417 in there is nocapital steady and stateSh thisfraction model. We discussinthis possibility in The physical the invested human capital. sky(t) - (n + g + 8)k(t), (9a) k(t) =for Section Equations and (9b)We imply economy in evolution of the economy is determined by thatthisthepossibility there is noIII.) steady state for(9a) this model. discuss = Shy(t) - (n + g + 8)h(t), h(t) (9b) converges to a steady state(9a) defined by(9b) imply that the economy Section - (n + g + 8)k(t), (9a) III.) Equations k(t) = sky(t)and = K/AL, where y =to Y/AL, k state converges a steady andbyh = H/AL are quantities per defined = Shy(t) - (n + g + 8)h(t), h(t) We (9b) effective unit of labor. are assuming that the same production +g + \n function to human applies capital, physical capital, consump-per = = y where k h = H/AL Y/AL, K/AL, and areand quantities (10) + + tion. In other oneWe unit of consumption transformed g assuming \n are effective unit words, of labor. that can the be same production a 1/(1-a-) -S a k ( * (10) into costlessly either one unit of or one physical capital unit of 418 QUARTERLY JOURNAL OFphysical ECONOMICS function to human applies capital, capital, and consump, n a +g+ a /assuming human In addition, 1/(1-a-) we are that capital k-S of (* one unit tion. Incapital. other words, canhuman consumption be transformed at(10) depreciates the same as Ifphysical capital. Lucas [1988] = 1/3, value than into a =Pfunction the coefficient onrate ln(sk). for the example, Substituting into and taking logs , the nproduction costlessly either one unit or one capital unit of +g+ of /physical coefficient on + g income + ln(n models function production for 8) would behuman -2. similar Incapital as fundamenthis model high gives anthe equation per capita to equation (6) human Infor capital. addition, we are assuming that human capital Substituting into the (10) production function and taking logs population growth lowers income pergoods. capita because the that, amounts different from that for tally other We believe at least above: at depreciates the same rate as physical capital. Lucas [1988] of an both physical and must be spread more gives aninitial equation forhuman income pernatural capita similar to equation (6) for itcapital is to examination, assume thatthinly the two the models production function for human capital as fundamen+ over the population. above: types of functions similar. -are Inproduction [L(t)| In (I ln(n (1 1) There A(O) different from that+ for tally other We gt way goods. g + aof human that, at least is an alternative to the+ believe role Y (t)]_ We assume that a + ,B < 1, express which implies that there are + for an initial it is natural to assume(11)that examination, in determining capital in model. withthe two - this (If decreasing returns to income all a(I + Combining capital. there ,B= +1,g then are In [L(t)| In + + a (1 1) A(O) ln(n gt the equation for the functions steady-stateare level of human capital given in of production types similar. Y (t)]_ constant to scale in+ the reproducible returns case, +factors. In this In(Sh) (10)We yields an equation aIln(Sk) as which function of thethat rate of assume that afor+ income there are ,B < 1, implies investment in physical capital, the rate of population growth, and decreasing returns to all + ,B= 1, (If a depends then there are +capital. This equation shows how income per capita on of population + 7.level Previous authors have provided evidence ofIln(Sk) the importance human capital In(Sh) the of human capital: constant to in returns scale the In reproducible factors. this for growth in income. Azariadis and Drazen [1990] find that no country able case, was to growth and accumulation of physical and human capital. grow quickly during the postwar period without a highly literate labor force. They Like the Solow model, model augmented predictswith This shows how capita depends onassociated population interpret this astextbook evidence that there a the threshold externality [L(t) (12)equation In = InA(O) +income + 1is_per gt ln(sk) human 7. capital accumulation. Previous authors have Similarly, Rauch provided [1988] finds evidence of the that among importance of human countries capital in coefficients are equation (11)ofthat of the functions factor shares. and accumulation growth and human physical capital. had achieved that 95 percent adultand in 1960, for growth in income. literacy for to strong Azariadis Drazen tendency [1990]there findwas thata no country was able aaL As before, isWi capital's of share somodel welabor athatThey expect Like textbook Solow the model, augmented predicts income per the capita tophysical converge over the period 1950-1985. Romer grow [1989b] quickly finds during the postwar period without aincome, highly literate force. 1 Gauging -aare +a greasonable +and literacy 1960 helps +the -iffactor explain subsequent that, oneln(h*). 8) corrects interpret a in evidence that there value in of this ofas about isinvestment an(n threshold one(11) third. externality value of P, forwith associated coefficients equation that of functions shares. measurement error, literacy has no impact on[1988] growth beyond its effect on human capital accumulation. Similarly, Rauch finds that among countries human capital's iscapital's moreliteracy difficult. Inthe the United States the investment. There isshare, also older a is physical stressing As role of so of human share capital income, we ain for expect had achieved thatbefore, 95 adultwork in 1960, percent there was a strong tendency Equation is (12) almost identical to in In Section I. equation (6) development; for example, see Krueger [1968] and Easterlin [1981]. minimum the return to wage-roughly labor without human income per capita to converge over theGauging period 1950-1985. Romer [1989b] finds that value ofina oftheabout onehuman third. a reasonable value of P, for that model levelexplain of ispercent a component capital of the error literacy 1960averaged helps subsequent investment and that, if one corrects to capital-has about 30 50 of the average wage human capital's share, is more In the United States the on measurement error, literacy has population nodifficult. impact growth on growth beyond its effect Because the and saving rates influence interm. manufacturing. This fact suggests that 50 to 70 percent of total investment. There is also older work stressing the role of human capital minimum the capital returnto to wage-roughly labor without human in h *, one should expect see human be Easterlin correlated positively development; example, Krueger and [1981]. labor incomeforrepresents the return[1968] to human capital, or that ,Bis with the saving rate and capital-has about 30 to 50correlated the population averaged percent ofwith negatively average wage Solow Model with Human Capital Accumulation • Sum of coefficients on ln(sk ) and ln(sh ) in equation (11) should be equal in magnitude and opposite in sign to the coefficient on ln(n + g + δ). • Use average percentage of working-age population in secondary school to proxy for human capital investment rate. • Again, assume value for g + δ equal to 0.05. And again use population growth rate and investment rate for physical capital. Estimation Results for Augmented Solow Model • Coefficients on investment rates and population growth have the predicted signs and are significantly different from zero in two of the three samples. For OECD only school variable is significantly different from zero. • • Equality restriction can’t be rejected in all samples. Regressions explain large fraction (over 3/4) of the cross-country variation in income per capita for two of the three samples. the percentage of the population in secondary school. The humancapital measure enters significantly in all three samples. It also TABLE II OF THEAUGMENTED SOLOWMODEL ESTIMATION Dependent variable: log GDP per working-age person in 1985 Sample: Observations: CONSTANT ln(I/GDP) ln(n + g +5) ln(SCHOOL) R2 s.e.e. Restricted regression: CONSTANT ln(I/GDP) - ln(n + g + 5) ln(SCHOOL) - ln(n + g + 5) R2 s.e.e. Test of restriction: p-value Implied a Implied , Non-oil 98 6.89 (1.17) 0.69 (0.13) -1.73 (0.41) 0.66 (0.07) 0.78 0.51 Intermediate 75 7.81 (1.19) 0.70 (0.15) -1.50 (0.40) 0.73 (0.10) 0.77 0.45 OECD 22 8.63 (2.19) 0.28 (0.39) -1.07 (0.75) 0.76 (0.29) 0.24 0.33 7.86 (0.14) 0.73 (0.12) 0.67 (0.07) 0.78 0.51 7.97 (0.15) 0.71 (0.14) 0.74 (0.09) 0.77 0.45 8.71 (0.47) 0.29 (0.33) 0.76 (0.28) 0.28 0.32 0.41 0.31 (0.04) 0.28 (0.03) 0.89 0.29 (0.05) 0.30 (0.04) 0.97 0.14 (0.15) 0.37 (0.12) Note. Standard errors are in parentheses. The investment and population growth rates are averages for the period 1960-1985. (g + 8) is assumed to be 0.05. SCHOOL is the average percentage of the working-age population in secondary school for the period 1960-1985. Estimation Results for Augmented Solow Model • Implied values for α and ß are close to 1/3 for the first two samples. • Hence, model fits better with other evidence that physical capital’s share is about 1/3. • Poor performance of OECD sample may be due to violation of assumption that economies are in steady state by 1985, because it is more likely that rich countries were further from steady state due to WWII’s effects being concentrated in those countries. • So consider analysis out of steady state. Convergence to Steady State • Distance from steady state determines rate of growth in output per worker. • Countries that are further from steady state will grow faster than those that are closer. • Introducing human capital accumulation implies slower rate of convergence for a given gap between current output per worker and steady-state output per worker. about the speed of convergence to steady state. Let y be the steady-state level of income per effective worker given by equation (11), and let y(t) be the actual value at time t. Approximating around the steady state, the speed of convergence is given by (13) d ln(y(t)) dt = X[ln(y*) -ln(y(t))], where A = (n + g + a) (1-a - THE EMPIRICS OF ECONOMIC GROWTH 423 For example, if a = P = 1/3 and n + g + 8 = 0.06, then the convergence rate (A) would equal 0.02. This implies that the economy moves halfway to steady state in about 35 years. Notice that the textbook Solow model, which excludes human capital, implies much faster convergence. If IB= 0, then X becomes 0.04, and the economy moves halfway to steady state in about seventeen years. The model suggests a natural regression to study the rate of convergence. Equation (13) implies that (14) ln(y(t)) = (1 - e-At) ln(y*) + e-Atln(y(0)), where y(O) is income per effective worker at some initial date. Convergence to Steady State • Equation (13) can be used to derive an estimating equation to test for convergence. • This leads to an equation that expresses growth in output per worker as a function of the steady state and the initial level of output per worker: about the speed of convergence to steady state. Let y * be the of income per effective worker given by equation steady-state level THE EMPIRICS OF ECONOMIC GROWTH 423423 THE EMPIRICS OFECONOMIC ECONOMIC GROWTH OF 423 be the actual value atGROWTH time t. Approximating (11), and let THE y(t) EMPIRICS around the steady state,= the speed of convergence is given by == For example, if a 1/3 and n + g + 8 = 0.06, then the P = For example, ifif aa = PP = 1/3 1/3 and and nn ++ gg ++ 88 == 0.06, 0.06, then then the the For example, convergence rate (A) would equal 0.02. This implies that the ln(y(t)) convergence rate d (A) would= equal equal 0.02. This This implies that that the the convergence rate (A) would 0.02. (13) dt X[ln(y*) -ln(y(t))], in economy moves halfway to steady state about implies 35 years. Notice economy moves moves halfway to to steady steady state state inabout about 35years. years.Notice Notice that the textbookhalfway Solow model, which in excludes 35 human capital, that the textbook textbook Solow model, model, which excludes human capital, capital, where implies much fasterSolow convergence. which If IB= excludes 0, then Xhuman becomes 0.04, implies much much faster faster convergence. convergence. IfIf IB IB then XXbecomes becomes0.04, 0.04, ==state 0,0,then and the economy moves halfway to steady in about seventeen (1-a state A = halfway (n + g +to and the economy economy moves moves halfway toa) steady state aboutseventeen seventeen ininabout steady years. years. The model suggests a natural regression to study the rate of model suggests natural that regression totostudy The model studythe therate rateofof suggests aa natural regression convergence. Equation (13) implies Equation convergence. Equation (13) (13) implies implies that Solving the differential equation (13)that yields: (14) ln(y(t)) = (1 - e-At) ln(y*) + e-Atln(y(0)), ln(y(t)) (14) ln(y(t)) == (1 (1 -- e-At) e-At)ln(y*) ln(y*) ++e-At e-Atln(y(0)), ln(y(0)), where y(O) is income per effective worker at some initial date. where y(O) is per effective worker at some initial date. is income income perboth effective from sides, worker at some initial date. Subtracting In (y(O)) In both In rewritten from both sides, Subtracting (y(O)) sides, (y(O))from which can be as: (15) ln(y(t)) - ln(y(0)) = (1 - e-t) ln(y*) - (1 - e-At) ln(y(0)). (15) ln(y(t)) ln(y(0)). (15) ln(y(t)) -- ln(y(0)) ln(y(0)) == (1 (1 -- e-t) e-t) ln(y*) ln(y*) - - (1(1- - e-At) e-At) ln(y(0)). Finally, substituting for y*: Finally, substituting Finally, substituting for fory*: y*: (16) (16) (16) ln(y(t)) - ln(y(0)) = (1 - e-t) ln(y(t)) ln(y(t)) -- ln(y(0)) ln(y(0)) == (1 (1 -- e-t) e-t) + (1 ++ (1 e-t) (1 - - e-t) e-t) - ln(sk) − λt ln(sk) ln(sk)+ [1− e ]ln A(0) ln(sh) ln(sh) ln(sh) _ lIn(n + g + 8) - (1 - e-t) ln(y(0)). __ lIn(n + g + 8) - (1 - e-t) ln(y(0)). lIn(n + g + 8) - (1 - e-t) ln(y(0)). Thus, in the Solow model the growth of income is a function of the of income of the Thus, in the Solow the growth is a initial function of themodel and the determinants ultimate level in the Solow model of income ofofthe Thus, the steady is a function growthstate determinants of the ultimate steady state and the initial level of income. of the ultimate steady state and the initial level of determinants - (1 - eAxt) _ (1 (1 -- eAxt) eAxt) __ Estimation Results for Convergence Hypothesis • Consider first unconditional convergence, where the steady state y* is assumed to be the same across all countries. • Coefficient on initial output per worker in 1960 should be negative and significantly different from zero. • Results show evidence of unconditional convergence only for the OECD sample. 425 THE EMPIRICS OF ECONOMIC GROWTH TABLE III TESTS FOR UNCONDITIONALCONVERGENCE Dependent variable: log difference GDP per working-age person 1960-1985 Sample: Observations: CONSTANT ln(Y60) R2 s.e.e. Implied X Non-oil 98 -0.266 (0.380) 0.0943 (0.0496) 0.03 0.44 -0.00360 (0.00219) Intermediate 75 0.587 (0.433) -0.00423 (0.05484) -0.01 0.41 0.00017 (0.00218) OECD 22 3.69 (0.68) -0.341 (0.079) 0.46 0.18 0.0167 (0.0023) Note. Standard errors are in parentheses. Y60 is GDP per working-age person in 1960. essentially zero. There is no tendency for poor countries to grow faster on average than rich countries. Table III does show, however, that there is a significant tendency toward convergence in the OECD sample. The coefficient Estimation Results for Convergence Hypothesis • • Allow for different steady states across countries. • Consider first controlling only for different investment rates in physical capital and in population growth. • Next consider controlling as well for different investment rates in human capital (using school variable). Estimate growth regression controlling for these different steady states--i.e., testing for “conditional convergence.” 426 QUARTERLYJOURNAL OF ECONOMICS TABLE IV TESTS FOR CONDITIONALCONVERGENCE Dependent variable: log difference GDP per working-age person 1960-1985 Sample: Observations: CONSTANT ln(Y60) ln(I/GDP) ln(n + g + 8) R72 s.e.e. Implied X Non-oil 98 1.93 (0.83) -0.141 (0.052) 0.647 (0.087) -0.299 (0.304) 0.38 0.35 0.00606 (0.00182) Intermediate 75 2.23 (0.86) -0.228 (0.057) 0.644 (0.104) -0.464 (0.307) 0.35 0.33 0.0104 (0.0019) OECD 22 2.19 (1.17) -0.351 (0.066) 0.392 (0.176) -0.753 (0.341) 0.62 0.15 0.0173 (0.0019) Note. Standard errors are in parentheses. Y60 is GDP per working-age person in 1960. The investment and population growth rates are averages for the period 1960-1985. (g + 8) is assumed to be 0.05. TABLE V TESTS FOR CONDITIONAL CONVERGENCE Dependent variable: log difference GDP per working-age person 1960-1985 TABLE V TESTS FOR CONDITIONAL CONVERGENCE Dependent variable: log difference GDP per working-age person 1960-1985 Sample: Observations: CONSTANT ln(Y60) ln(I/GDP) ln(n + g + 8) ln(SCHOOL) R2 s.e.e. Implied X Non-oil 98 3.04 (0.83) -0.289 (0.062) 0.524 (0.087) -0.505 (0.288) 0.233 (0.060) 0.46 0.33 0.0137 (0.0019) Intermediate 75 3.69 (0.91) -0.366 (0.067) 0.538 (0.102) -0.551 (0.288) 0.271 (0.081) 0.43 0.30 0.0182 (0.0020) OECD 22 2.81 (1.19) -0.398 (0.070) 0.335 (0.174) -0.844 (0.334) 0.223 (0.144) 0.65 0.15 0.0203 (0.0020) Note. Standard errors are in parentheses. Y60 is GDP per working-age person in 1960. The investment and population growth rates are averages for the period 1960-1985. (g + 8) is assumed to be 0.05. SCHOOL is the average percentage of the working-age population in secondary school for the period 1960-1985. THE EMPIRICS OF ECONOMICGROWTH A. Unconditional 0 co 66 (0 Q) 4 -- 427 ~~00 2 2 00 Oo 0O 0 ~~~~~0 o 0 -~0 0 0 0 R ? 0 0 0 o 0 00 Oo0 ? -2 5,5 s 6,5 7,5 8,5 9.5 10.5 Log output per working age adult:1960 B. Conditionalon saving and populationgrowth ~~ ~~0 0 0 0 o0c~ O? ???0 08 8 6 060 W, 4 _CP = 2 o 020 0 o D 0 0 5s,5 6.5 7, 5 8.5 9.5 10,5 Log output per working age adult:1960 LO C.Conditionalon saving, populationgrowthand humancapital 6 0~~~~~~ 0 ? 2 0~~~~~~ ? w ? 0 O??b 9 lb8 -2 5.5 6.5 7.5 8.5 9.5 Log output per working age adult:1960 FIGUREI Unconditional versusConditional Convergence 10.5 Estimation Results for Convergence Hypothesis • Coefficient on initial output per worker is negative and significant in all samples when controlling for differences in steady states, both with and without human capital accumulation, indicating substantial convergence. • But speed of adjustment parameter is much smaller than we’d expect (assuming physical capital share of 1/3) when controlling only for differences in investment rates for physical capital (and population growth). • When also control for differences in human capital accumulation, the speed of adjustment is closer to the value of 0.02 (for the intermediate and OECD samples), which is what we’d expect if we assume physical capital share of 1/3 and human capital share of 1/3. 429 THE EMPIRICS OF ECONOMICGROWTH TABLEVI TESTS FOR CONDITIONAL CONVERGENCE, RESTRICTED REGRESSION Dependentvariable:log differenceGDP per working-ageperson 1960-1985 Sample: Observations: CONSTANT ln(Y60) ln(I/GDP) - ln(n + g + 5) ln(SCHOOL) - ln(n + g + 5) R2 s.e.e. Test of restriction: p-value ImpliedX Implied(x Implied 13 Non-oil 98 2.46 (0.48) -0.299 (0.061) Intermediate 75 3.09 (0.53) -0.372 (0.067) OECD 22 3.55 (0.63) -0.402 (0.069) 0.500 0.506 0.396 (0.082) (0.095) (0.152) 0.238 0.266 0.236 (0.060) 0.46 0.33 (0.080) 0.44 0.30 (0.141) 0.66 0.15 0.40 0.0142 (0.0019) 0.48 (0.07) 0.23 (0.05) 0.42 0.0186 (0.0019) 0.44 (0.07) 0.23 (0.06) 0.47 0.0206 (0.0020) 0.38 (0.13) 0.23 (0.11) Note. Standard errors are in parentheses. Y60 is GDP per working-age person in 1960. The investment and population growth rates are averages for the period 1960-1985. (g + 5) is assumed to be 0.05. SCHOOL is the average percentage of the working-age population in secondary school for the period 1960-1985. specifications that do not consider out-of-steady-state dynamics. Similarly, the greater importance of departures from steady state for the OECD would explain the finding of greater unconditional Estimation Results for Convergence Hypothesis • Sum of coefficients on ln(sk ) and ln(sh ) in equation (16) should be equal in magnitude and opposite in sign to the coefficient on ln(n + g + δ). • • Cannot reject this restriction. • These convergence regression estimates give larger weight to physical capital and a smaller weight to human capital compared to the estimates explaining variation in output per worker (Table II). Implied values for income share of physical capital of range from 0.38 to 0.48 and the implied values for income share of human capital is 0.23 in all three samples.