A key economic issue is whether poor countries or regions tend to grow faster than rich ones: are there automatic forces that lead to convergence over time in the levels of per capita income and product? In this paper we will try to answer this question by referring to absolute and conditional convergence models. Both models will be introduced, discussed and compared empirically. Meanwhile, quantitative measures for the speed of convergence will be launched for both Solow’s growth model and the growth model of RamseyCass-Koopmans. Finally, empirical results on convergence will be reviewed.

2. Transitional Dynamics

We will start with finding out whether or not growth rate of capital per capita differs throughout its time path. Growth rate of k, γ k , is given by

γk ≡

k 1 = [ s. f (k ) − (n + δ )k ] k k


γ k = s.

f (k ) − (n + δ ) k

If k < k * .We can analyze equation (1) graphically as follows: y Growth rate > 0 n +δ Growth rate > 0 s. the average product of capital f (k ) / k is relatively high. faster is the convergence to the steady state level. If k > k * . then the growth rate of k is positive and k increases toward k*. According to Solow’s growth model we assume that households save and 2 . f (k ) / k reflects the growth rate of k. k*. To source of these results is the diminishing returns to capital: when k is relatively low. Greater the referred distance.Absolute Convergence In the figure the vertical distance between the line n + δ and the curve s. according to equation (1). f (k ) k k k (0) poor k (0) rich k * Figure 1 . then the growth rate of k is negative and k decreases toward k*.

s. k. Analogously. Y = AK α L1−α (3) thus. the growth rate of output per capita can be calculated as: γy = y = f ′( k ). effectively depreciates at the constant rate n + δ . f (k ) / k is relatively high. s.k / f ( k ) = [ k . Hence when k is relatively low. the growth rate. Capital per worker. of this product.invest a constant fraction.γ k (5) 3 . Consequently.γ k y (2) In the Cobb-Douglas case. k / k . y = Ak α (4) where α is known as the capital share. If we use Cobb-Douglas production function in equation (2) we obtain: γ y = α . f ′( k ) / f ( k )]. is also relatively high. the gross investment per unit of capital.

Absolute and Conditional Convergence Now. However. γ k . Figure 1 shows us also the growth rate of per capita output. In order to remedy this problem heterogeneity across economies must be allowed. because there is a slight tendency for the initially richer countries to grow faster in per capita terms. the growth rate of per capita output is the fraction α of the growth rate of per capita capital. the assumption that all economies have the same parameters and therefore the same steady-state positions must be ignored. This result implies a form of convergence: regions or countries with lower starting values of the capital/labor ratio have higher per capita growth rates. empirical studies on a broad cross section of countries over the period 1960 to 1985 do not support absolute convergence. So. we can give an answer to our main question: do poor countries or regions tend to grow faster than rich ones? According the analysis by the help of Figure 1. If the steady-states differ. then we have to modify the analysis to consider a 4 . The hypothesis that poor economies tend to grow faster per capita than rich ones – without conditioning on any other characteristics of economies – is referred to as absolute convergence. 3. we can say that other thing equal. smaller values of k are associated with larger values of γ k . This is to say the behavior of γ y mimics that of γ k . and tend thereby to catch up or converge to those with higher capital/labor ratios. In particular.Thus.

y n +δ srich .concept of conditional convergence. Figure 2 reflects that if a rich economy has a higher saving rate than a poor economy. The main idea is that an economy grows faster the further it is from its own steady-state value. that is absolute convergence would not hold. s poor . the rich economy would be predicted to grow faster per capita than the poor economy. then the rich economy may be proportionally further from its steady-state position. In this case. The cross-country data support the hypothesis of conditional convergence. f (k ) k f (k ) k k k (0) poor k * poor k (0) rich k *rich Figure 2 .Conditional Convergence 5 .

for Solow’s growth model can be written as follows1: β = (1 − α )( x + n + δ ) (6) where α is the capital share. A(t ) A(t ) is the technology term defined as A(t ) = e xt by normalizing the initial level of technology A(0) to 1. p. the growh rate of technology is equal to x. approaches 1 2 For the derivation of this measure see Barro and Sala-I Martin (1995). Similarly capital per unit where of effective labor is defined as k K ˆ = k= A(t ) L. x is the growth rate of technology. n is the growth rate of population and δ is the rate of depreciation.4. y . Speed of Convergence In this section a quantitative measure or the speed of convergence will be introduced for both Solow’s growth model and Ramsey-Cass-Koopmans model. Thus. A(t ) 6 . β .1 Speed of Convergence for Solow’s Growth Model A quantitative measure of the speed of convergence. The term β ˆ indicates how rapidly an economy’s output per effective worker2. 4.1) ≡ f (k ) = A(t ) L.53. Barro and Sala-i-Martin defines output per effective worker as y Y ˆ ˆ ˆ = y = F (k .

β is independent of the level of the technology. firstly we will setup the model constructed by Ramsey (1928) and developed by Cass (1965) and Koopmans (1965). One property is that the saving rate. as given by U = ∫ u[c(t )].ˆ its steady-state value y * . β . then 5 percent of ˆ ˆ the gap between y and y * vanishes in 1 year. Each household wishes to maximize overall utility. the observed convergence value β = 2 can be obtained only with a capital share α = 0. U. The measure for the speed of convergence implies some results. For instance.e − ρt dt 0 ∞ (7) 7 . 4. does not affect the speed of convergence.75 . For instance if β = 0.2 Speed of Convergence for Ramsey-Cass-Koopmans Growth Model In order to derive a quantitative measure for speed of convergence for growth models with consumer optimization. Solow’s growth model can generate the rates of convergence that have been observed empirically only with a broad concept of capital. Moreover. s. A.05 per year.e nt .

w denotes the wage income per adult person. represents the adding up of utils for all family members alive at time t. c. and fall because of expansion of the population in accordance with the term na . u ′′(c) < 0 . 8 . We assume that u (c ) is L (t ) increasing in c and concave – u ′(c) > 0. The equation says that assets per person rise with per capita income. fall with per capita consumption.where c (t ) = C (t ) is the consumption per capita. e − ρ t . and u (c) → 0 as c → ∞ . L = e nt . r denotes the real rate of return. ρ > 0 . We assume that the utility function is in the form: c (1−θ ) − 1 u (c ) = (1 − θ ) (9) 3 We omit time subscripts for simplicity. The other multiplier. c denotes consumption per capita and n the growth rate of population3. We also assume that u (c ) satisfies Inada conditions: u (c ) → ∞ as c → 0 . w + ra . The multiplication of u (c ) by the family size. The flow budget constraint for the household is: a = w + ra − c − na (8) where a = A / L denotes the level of assets per capita. involves the rate of time preference.

marginal products of the factors are given by ∂Y ˆ = f ′(k ) ∂K (12) ∂Y ⎡ ˆ ˆ ˆ = f ′(k ) − k . we can analyze the production part of the model.where θ > 0 . Now. L) (10) ˆ where L = L. We therefore assume that the production function is in the form ˆ Y = F ( K . A steady-state coexists with technological process at a constant rate only if this process takes the labor-augmenting form. A(t ) is the effective amount of labor input. As we mentioned in the case of Solow’s growth model. and A(t). the level of the technology grows at the constant rate x. the intensive form of the production function is: ˆ ˆ y = f (k ) (11) Thus. f ′(k ) ⎤ e xt ⎦ ∂L ⎣ The representative firm’s flow of net receipts or profit at any point in time is given by 9 .

f (k ) ⎦ e = w (16) The household’s flow budget constraint in (8) determines a . L) − (r + δ ) K − wL (13) ˆ We can rewrite equation (13) by using the definition of L as follows: ˆ ˆ ˆ π = L ⎡ f (k ) − (r + δ )k − we− xt ⎤ ⎣ ⎦ (14) ˆ A competitive firm which takes r and w as given. ˆ k = ke − xt . the wage rate has to equal the marginal product of labor corresponding to the value ˆ of k that satisfies equation (15): ⎡ ′ ˆ ˆ ′ ˆ ⎤ xt ⎣ f (k ) − k . and the conditions for r and w in equations (15) and (16) to get 10 . the firm chooses the ratio of capital to effective labor to equate the marginal product of capital to the rental price.ˆ π = F ( K . maximizes profit for given L by setting ˆ f ′(k ) = r + δ (15) That is. In order for profit to be 0. Use a = k .

However. together with equation (17) forms a system of two differential ˆ ˆ equations in c and k . which implied the linear consumption ˆ function c = (1 − s ) f (k ) . In the present setting. Equation (17) is the resource constraint for the overall economy. If we log-linearise this system around steady-state we obtain: ˆ ˆ ˆ log [ y (t ) ] = e − β t .ˆ ˆ ˆ ˆ k = f (k ) − c − ( x + n + δ )k (17) ˆ ˆ where c = C / L . The advantage of this method is that it provides a closed-form solution for the convergence coefficient. Thus. The 11 . by using the conditions r = f ′(k ) − δ and c = ce − xt then we get ˆ c c ⎛1⎞ ˆ = − x = ⎜ ⎟ ⎡ f ′(k ) − δ − ρ − θ x ⎤ ⎦ ˆ c c ⎝θ ⎠ ⎣ (18) This equation. log [ y (t )] is the weighted average of the initial and steadyˆ ˆ state values. with the weight on the initial value declining exponentially at the rate β . log [ y (0) ] and log ( y *) . In Solow’s growth model the missing relation was provided by the assumption of a constant saving rate.log [ y (0) ] + (1 − e − β t ) log ( y *) (19) ˆ where β > 0 . the behavior of the saving rate is ˆ not so simple.

The speed of convergence. We use the Cobb-Douglas production function ˆ ˆ ˆ y = f (k ) = Ak α (20) Thus. β . the size of the capital share coefficient α – has a strong effect on β . β tends to zero. Because the crucial element for convergence in the neoclassical model is diminishing returns to capital. As α approaches to unity. over an interval from an initial time o to any future time T ≥ 0 is given by 4 For the derivation β see Barro and Sala-I Martin (1995).x > 0 .87. 12 . the speed of convergence can be written as follows4: ⎧ ⎛ 1−α ⎞ ⎡ ρ + δ +θ x ⎤⎫ 2 β = ⎨ζ 2 + 4 ⎜ − (n + x + δ ) ⎥ ⎬ ⎟ ( ρ + δ + θ x) ⎢ α ⎝ θ ⎠ ⎣ ⎦⎭ ⎩ 1/ 2 −ζ (21) where ζ = ρ − n − (1 − θ ). p.disadvantage is that it applies only as an approximation in the neighborhood of the steady state. depends on the parameters of technology and preferences. the extent of these diminishing returns – that is. y. diminishing returns to capital disappear. Equation (21) implies that the average growth rate of per capita output.

Another result is that the parameter A in equation (20) does not affect β . We had mentioned above that as α approaches to unity. The theory conforms to the empirical findings if we assume parameter values as in the following table. Quantitatively.8 0. β .(1 − e− β t ) . β tends to zero. y (0) . The speeds of adjustment.log ⎡ y * ⎤ ˆ ⎡ y (T ) ⎤ 1 log ⎢ ⎥ = x+ ⎢ y (0) ⎥ T T ⎣ y (0) ⎦ ⎣ˆ ⎦ (22) This equation says that the average per capita growth rate of output depends ˆ ˆ negatively on the ratio of y (0) to y * .02 per year. Thus. ˆ y * .02 0. α 0. the Ramsey-Cass-Koopmans model also predicts conditional rather than absolute convergence. Thus the convergence coefficient β can be similar across economies that differ greatly in levels of per capita product because of differences in the available technique. the most important effect is that a lower θ (increased willingness to substitute intertemporally) raises β . In other words.05 0. the ˆ effect of the initial position. Other results can be classified as follows. diminishing returns to capital disappear. that is estimated by Barro and Sala-iMartin (1992) is in the neighborhood of 0. as in the Solow’s growth model.02 1 δ n x θ Table 1 13 . is conditioned on the steady-state position.

states. 14 . Empirical Evidence on Convergence In this section we will discuss on the empirical studies on absolute convergence and conditional convergence. Thus. 5. states. It has been found that growth rates of countries are correlated with the initial position. An even more homogenous group can consist of the continental U. thus diminishing returns set in slowly. each viewed as a separate economy. according to Table 1. In the study of Barro and Sala-iMartin (1992) there are 118 countries studied over the period 1960-1985.5.However. In another study of Barro and Sala-i-Martin (1992). we talked about absolute convergence and mentioned that it holds for the empirical studies on OECD countries and U. there is a slight tendency for the initially richer countries to grow faster in per capita terms.S. They have found that growth rates of states are correlated with the initial position. we must assume high values of θ (in excess of 10) and a value of δ close to zero. This is because of the fact that these countries are members of a more homogeneous group. α = 0. Barro and Sala-i-Martin (1992) have studied the period 1880–1990. It has been found that growth rates of countries are essentially uncorrelated with the initial position: in fact.8 . But to do this. However. there are 20 OECD (Organization for Economic Cooperation and Development) countries studied over the period 1960-1985.S. So far. we can reduce α to around 0. In order to remedy this. this sample rejects the hypothesis of absolute convergence.

They claim that convergent growth can be achieved by all or virtually all countries that follow a reasonable set of political and economic policies. According to their empirical results. Government policies include the decisons on public spending on education. the relation between the per capita growth rate and the log of initial real per capita GDP becomes significantly negative. Barro and Sala-i-Martin (1995) show that the inclusion of variables that proxy for differences in steady-state positions makes a major difference in the results across the broad cross section of countries. and (most importantly) an open economy. as predicted by the neoclassical model. basic adherence to political and civil rights. export monopolies. The findings on conditional convergence suggest that we should hold constant these determinants of k* in order to isolate the predicted inverse relationship between growth rates and initial positions. the rule of law and the quality of political institutions. s. all developing countries that followed such a pattern achieved per capita growth between 1970 and 1989 of two percent per year of greater. government consumption. We know that the steady-state value. through the absence of trade quotas. Thus. the model of conditional convergence must be studied empirically in order to find whether or nor it can explain the growth behavior of heterogeneous countries. k*. the cross-country data support the hypothesis of conditional convergence. 5 15 . war and defense expenditures. For a complete list of these varibles see Barro and Sala-i-Martion (1995). including civil peace. another supporting empirical study has been made by Sachs and Warner (1995). political instability. depends on the saving rate. the level of production function f(k) and on various government policies5 that effectively shift the position of the production function. or inconvertible currencies. tariff rate.empirical studies show that absolute convergence does not hold for countries that are not homogeneous. On the other side of the coin. That is to say. when these additional variables are held constant. Here. Chapter 12. school-enrollment ratios.

So. If poor countries can make governmental reforms and catch up rich countries in the name of substructure for production. 16 . states. Conclusion We tried to answer the question that whether or not poor countries grow faster than rich ones.6. This is to say absolute convergence cannot explain us the differences between the growth rates of heterogeneous countries. The former states that countries with similar properties (call them homogeneous countries) approach to the same steady-state. Here. it is not easy to figure out which country (poor or rich) will grow faster than the other one because of the differences between countries. there is no reason for them to grow slower than rich countries and converge to their own steady-states rapidly. we underlined some useful lessons for poor countries. and poor countries grow faster than rich ones. This difference is explained only by conditional convergence. We showed that empirical studies support absolute convergence only for OECD countries and U. by holding constant these differences it has been showed that conditional convergence holds empirically. Lastly.S. We used two theoretical models in order to achieve this: absolute convergence and conditional convergence. and the growth rate of each country depends on relative distance between its initial position and its own steady-state. The latter states that each country approaches to its own steady-state because of heterogeneity.

New York. N. X. pp. and Warner.2. • Sachs. 1992. X. 1995. D. Economic Growth. R. A. J. 17 . William Brainard and George Perry. R.” Journal of Political Economy.223-251.” NBER Working Paper No. • Barro. 1:1995. 1995. ed.100. “Convergence..2 and 11. V. M. and Sala-i Martin. J. 108-118).References: • Barro. J. “Economic Convergence and Economic Policies. McGraw-Hill. Chapters 1. W5039 (Published: Brookings Papers on Economic Activity. and Sala-i-Martin. 1-95.

Sign up to vote on this title
UsefulNot useful