Economic Growth Second Edition
Convergence and the Dispersion of Per Capita Income
Download 0.79 Mb. Pdf ko'rish
|
BarroSalaIMartin2004Chap1-2
1.2.11
Convergence and the Dispersion of Per Capita Income The concept of convergence considered thus far is that economies with lower levels of per capita income (expressed relative to their steady-state levels of per capita income) tend to grow faster in per capita terms. This behavior is often confused with an alternative meaning of convergence, that the dispersion of real per capita income across a group of economies or individuals tends to fall over time. 22 We show now that, even if absolute convergence holds in our sense, the dispersion of per capita income need not decline over time. Suppose that absolute convergence holds for a group of economies i = 1, . . . , N, where N is a large number. In discrete time, corresponding for example to annual data, the real per capita income for economy i can then be approximated by the process log (y i t ) = a + (1 − b) · log(y i ,t−1 ) + u i t (1.31) where a and b are constants, with 0 < b < 1, and u i t is a disturbance term. The condition b > 0 implies absolute convergence because the annual growth rate, log (y i t /y i ,t−1 ), is inversely related to log (y i ,t−1 ). A higher coefficient b corresponds to a greater tendency toward convergence. 23 The disturbance term picks up temporary shocks to the production function, the saving rate, and so on. We assume that u i t has zero mean, the same variance σ 2 u for all economies, and is independent over time and across economies. One measure of the dispersion or inequality of per capita income is the sample variance of the log (y i t ): D t ≡ 1 N · N i =1 [log (y i t ) − µ t ] 2 22. See Sala-i-Martin (1990) and Barro and Sala-i-Martin (1992a) for further discussion of the two concepts of convergence. 23. The condition b < 1 rules out a leapfrogging or overshooting effect, whereby an economy that starts out behind another economy would be predicted systematically to get ahead of the other economy at some future date. This leapfrogging effect cannot occur in the neoclassical model but can arise in some models of technological adaptation that we discuss in chapter 8. Growth Models with Exogenous Saving Rates 51 where µ t is the sample mean of the log (y i t ). If there are a large number N of observations, the sample variance is close to the population variance, and we can use equation (1.31) to derive the evolution of D t over time: D t ≈ (1 − b) 2 · D t −1 + σ 2 u This first-order difference equation for dispersion has a steady state given by D ∗ = σ 2 u /[1 − (1 − b) 2 ] Hence, the steady-state dispersion falls with b (the strength of the convergence effect) but rises with the variance σ 2 u of the disturbance term. In particular, D ∗ > 0 even if b > 0, as long as σ 2 u > 0. The evolution of D t can be expressed as D t = D ∗ + (1 − b) 2 · (D t −1 − D ∗ ) = D ∗ + (1 − b) 2t · (D 0 − D ∗ ) (1.32) where D 0 is the dispersion at time 0. Since 0 < b < 1, D t monotonically approaches its steady-state value, D ∗ , over time. Equation (1.32) implies that D t rises or falls over time depending on whether D 0 begins below or above the steady-state value. 24 Note especially that a rising dispersion is consistent with absolute convergence (b > 0). These results about convergence and dispersion are analogous to Galton’s fallacy about the distribution of heights in a population (see Quah, 1993, and Hart, 1995, for discussions). The observation that heights in a family tend to regress toward the mean across generations (a property analogous to our convergence concept for per capita income) does not imply that the dispersion of heights across the full population (a measure that parallels the dispersion of per capita income across economies) tends to narrow over time. Download 0.79 Mb. Do'stlaringiz bilan baham: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2025
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling