Economic Growth Second Edition
The Degree of Commitment
Download 0.79 Mb. Pdf ko'rish
|
BarroSalaIMartin2004Chap1-2
2.7.5
The Degree of Commitment The analysis thus far considered a case of full commitment, as in equation (2.53), and ones of zero commitment, as in equation (2.76). Barro (1999) also considers intermediate cases in which commitment is possible over an interval of length T , where 0 ≤ T ≤ ∞. Increases in the extent of commitment—that is, higher T —lead in the long run to a lower effective rate of time preference and, hence, to lower interest rates and higher capital intensity. However, changes in T also imply transitional effects—initially an increase in T tends to make households less patient because they suddenly get the ability to constrain their “future selves” to save more. Thus the analysis implies that a rise in T initially lowers the saving rate but tends, in the longer run, to raise the willingness to save. If the parameter T can be identified with observable variables—such as the nature of legal and financial institutions or cultural characteristics that influence the extent of individual discipline—the new theoretical results might eventually have empirical application. In fact, from an empirical standpoint, the main new insights from the extended model concern the connection between the degree of commitment and variables such as interest rates and saving rates. For a given degree of commitment, the main result is that a nonconstant rate of time preference leaves intact the main implications of the neoclassical growth model. 2.8 Appendix 2A: Log-Linearization of the Ramsey Model The system of differential equations that characterizes the Ramsey model is given from equations (2.24) and (2.25) by ˙ˆk = f (ˆk) − ˆc − (x + n + δ) · ˆk ˙ˆc/ˆc = ˙c/c − x = (1/θ) · [ f (ˆk) − δ − ρ − θx] (2.78) We now log-linearize this system for the case in which the production function is Cobb– Douglas, f (ˆk) = A · ˆk α . Start by rewriting the system from equation (2.78) in terms of the logs of ˆc and ˆk: d[log (ˆk)]/dt = A · e −(1−α)·log(ˆk) − e log (ˆc/ˆk) − (x + n + δ) d[log (ˆc)]/dt = (1/θ) · [α A · e −(1−α)·log(ˆk) − (ρ + θx + δ)] (2.79) In the steady state, where d[log (ˆk)]/dt = d[log(ˆc)]/dt = 0, we have A · e −(1−α)·log(ˆk ∗ ) − e log (ˆc ∗ /ˆk ∗ ) = (x + n + δ) α A · e −(1−α)·log(ˆk ∗ ) = (ρ + θx + δ) (2.80) Growth Models with Consumer Optimization 133 We take a first-order Taylor expansion of equation (2.79) around the steady-state values determined by equation (2.80): d[log (ˆk)]/dt d[log (ˆc)]/dt = ζ x + n + δ − (ρ + θx + δ) α −(1 − α) · (ρ + θx + δ) θ 0 · log (ˆk/ˆk ∗ ) log (ˆc/ˆc ∗ ) (2.81) where ζ ≡ ρ − n − (1 − θ) · x. The determinant of the characteristic matrix equals −[(ρ + θx + δ)/α − (x + n + δ)] · (ρ + θx + δ) · (1 − α)/θ Since ρ + θx > x + n (from the transversality condition in equation [2.31]) and α < 1, the determinant is negative. This condition implies that the two eigenvalues of the system have opposite signs, a result that implies saddle-path stability. (See the discussion in the mathematics appendix at the end of the book.) To compute the eigenvalues, denoted by , we use the condition det ζ − x + n + δ − (ρ + θx + δ) α −(1 − α) · (ρ + θx + δ) θ − = 0 (2.82) This condition corresponds to a quadratic equation in : 2 − ζ · − [(ρ + θx + δ)/α − (x + n + δ)] · [(ρ + θx + δ) · (1 − α)/θ] = 0 (2.83) This equation has two solutions: 2 = ζ ± ζ 2 + 4 · 1 − α θ · (ρ + θx + δ) · [(ρ + θx + δ)/α − (x + n + δ)] 1 /2 (2.84) where 1 , the root with the positive sign, is positive, and 2 , the root with the negative sign, is negative. Note that 2 corresponds to −β in equation (2.41). The log-linearized solution for log (ˆk) takes the form log[ˆk (t)] = log(ˆk ∗ ) + ψ 1 · e 1 t + ψ 2 · e 2 t (2.85) where ψ 1 and ψ 2 are arbitrary constants of integration. Since 1 > 0, ψ 1 = 0 must hold for log[ˆk (t)] to tend asymptotically to log(ˆk ∗ ). (ψ 1 > 0 violates the transversality condition, 134 Chapter 2 and ψ 1 < 0 leads to ˆk → 0, which corresponds to cases in which the system hits the vertical axis in figure 2.1.) The other constant, ψ 2 , is determined from the initial condition: ψ 2 = log[ˆk(0)] − log(ˆk ∗ ) (2.86) If we substitute ψ 1 = 0, the value of ψ 2 from equation (2.86), and 2 = −β into equation (2.85), we get the time path for log[ˆk (t)]: log[ˆk (t)] = (1 − e −βt ) · log(ˆk ∗ ) + e −βt · log[ˆk(0)] (2.87) Since log[ ˆy (t)] = log(A) + α · log[ˆk(t)], the time path for log[ ˆy(t)] is given by log[ ˆy (t)] = (1 − e −βt ) · log( ˆy ∗ ) + e −βt · log[ ˆy(0)] (2.88) which corresponds to equation (2.40). 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