Discrete time: First order difference equations Introduction When dealing with discrete time, the value of variable y will change only when the variable changes from one integer to the next. The pattern of change is now represented by difference quotient counterpart of ππ¦ ππ‘ Δπ¦ Δπ‘ which is the discrete . t can only now take on integer values. Therefore, comparing the value of y in two consecutive periods, we must have βt = 1. This implies that the difference quotient Δπ¦ Δπ‘ can be simplified to βy. This is called the first difference of y. β means difference In our case, we define the first difference as follows. βyt = yt+ 1 - yt yt – means value of y in tth period yt +1 – means value y in t+1 period. With this then we can describe the pattern of change of y by an equation such as, βyt = 2 …………………. (i) OR βyt = -0.1yt ……………... (ii) Such equations are known as difference equations. We can rewrite equation (i) and (ii) as follows yt +1 - yt = 2 ………………(iii) OR yt+1 = yt + 2 yt+1 – yt = -0.1yt ---------------------(iv) OR yt+1 – 0.9yt = 0 1 yt+1 = 0.9yt Difference equations can be either linear or nonlinear, homogenous or nonhomogeneous and of the first or second (or higher) orders. Equations (i) for instance can be classified as linear for there is no y term raised to the second or higher power. Nonhomogeneous because the right-hand side is non zero. The first order because there exists only a first difference βyt. Equations (ii) can be characterized as having constant coefficients and constant term. We only consider difference equations with constant terms. Solving a first order difference equation. To solve a first order difference equation involves finding a time path yt. We can solve the difference equation using iterative method. This method can only be understood by working out examples. Example 1 Find the solution to difference equation yt+1 – yt = 2 (assuming yo initial value) = 15 It is more convenient to use the alternative π¦π‘+1 = π¦π‘ + 2 π¦1 = π¦0 + 2 π¦2 = π¦1 + 2 = (π¦0 + 2) + 2 = π¦0 + 2(2) π¦3 = π¦2 + 2 = (π¦0 + 2(2)) + 2 = π¦0 + 3(2) In general π¦π‘ = π¦0 + π‘(2) = 15 + 2π‘ This is the solution as it can be used to find the value of yt for any time period. Solve the difference equation βπ¦π‘ = −0.1π¦π‘ , with the initial value of y being yo 2 π¦π‘+1 − π¦π‘ = −0.1π¦π‘ π¦π‘+1 − 0.9π¦π‘ = 0 π¦π‘+1 = 0.9π¦π‘ π¦1 = 0.9π¦0 π¦2 = 0.9π¦1 = 0.9(0.9)π¦0 = (0.9)2 π¦0 π¦3 = 0.9π¦2 = 0.9(0.9)2 π¦0 = (0.9)3 π¦0 In general the solution can be written as π¦π‘ = (0.9)π‘ π¦0 Solve the homogeneous difference equation ππ¦π‘+1 − ππ¦π‘ = 0 ππ¦π‘+1 = ππ¦π‘ π π¦π‘+1 = π π¦π‘ The solution is π π¦π‘ = (π)π‘ π¦0 π It is through ( )π‘ that various values of t will lead to their corresponding values of y. It π therefore corresponds to π −ππ‘ in the solution of the differential equations. If we write it more generally as ππ‘ (b for base) and attach the more general multiplicative constant A (instead of yo) it gives the general solution for the homogenous difference equation. π¦π‘ = π΄ππ‘ This expression π΄ππ‘ plays the same important role in difference equations as the expression π΄π −ππ‘ in the differential equation. The general method Suppose that we are seeking the solution to the first difference equation π¦π‘+1 + ππ¦π‘ = π 3 Where a and c are constants. The general solution will consist of the sum of two components. 1. A particular integral yp – which is any solution of the complete non homogeneous equation. 2. A complementary function yc – which is the general solution of the reduced equation π¦π‘+1 + ππ¦π‘ = 0 yp represents intertemporal equilibrium level of y. yc represents deviations for the time path from that equilibrium. π¦π + π¦π constitute the general solution due to the presence of the arbitrary constants. First, find complementary function π¦π‘ = π΄ππ‘ π¦π‘+1 = π΄ππ‘+1 So that π¦π‘+1 + ππ¦π‘ = 0 π΄π π‘+1 + ππ΄ππ‘ = 0 So that π΄ππ‘ π1 + ππ΄π π‘ = 0 (π΄π1 + π΄π)ππ‘ = 0 Divide throughout by ππ‘ π΄π + π΄π = 0 π+π =0 hence π = −π Therefore set π = −π π¦π = π΄ππ‘ = π΄(−π)π‘ The particular integral Assume that π¦π‘ = π (constant) π¦π‘+1 = π 4 π + ππ = π π → π = 1+π Therefore, the particular integral can be written as π π¦π = π = 1+π (π ≠ −1) If it happens that π = −1, then π¦π is not defined and therefore another solution must be sought. In this case let, π¦π‘ = ππ‘ π¦π‘+1 = π(π‘ + 1) π (π‘ + 1) + πππ‘ = π π= π because a = -1 π‘+1+ππ‘ π¦π = ππ‘ = ππ‘ is a non-constant function of t. it represents the moving equilibrium. Adding π¦π πππ π¦π gives the general solution π π¦π‘ = π΄(−π)π‘ + 1+π (general solution a ≠ -1) π¦π‘ = π΄(−π)π‘ + ππ‘ = π΄ + ππ‘ (general solution a = -1) Corresponding definite solutions Let π‘ = 0 π π¦0 = π΄ + 1+π π¦π‘ = [π¦0 − π 1+π → ] (−π)π‘ + π¦π‘ = π΄ + ππ‘ π¦π‘ = π¦0 + ππ‘ → π π΄ = π¦0 − 1+π π 1+π (πππππππ‘π π πππ’π‘πππ π ≠ −1) π¦0 = π΄ (πππππππ‘π π πππ’π‘πππ π = −1) Example Solve π¦π‘+1 − 5π¦π‘ = 1 7 π¦0 = 4 5 π¦π = π΄(−π)π‘ = π΄(− − 5)π‘ = π΄(5)π‘ π 1 1 π¦π = 1+π = 1−5 = − 4 1 π¦π‘ = π¦π + π¦π = π΄(5)π‘ − 4 (πππππππ π πππ’π‘πππ) When t = 0 1 π¦0 = π΄ − 4 1 → π΄ = π¦0 + 4 1 1 4 4 π¦π‘ = (π¦0 + ) (5)π‘ − 7 1 1 = (4 + 4) (5)π‘ − 4 1 = 2(5)π‘ − 4 (πππππππ‘π π πππ’π‘πππ) The dynamic stability of equilibrium In the continuous time case, the dynamic stability of equilibrium depends on the π΄π −ππ‘ term in the complementary function. In period analysis, it depends on π΄ππ‘ the complementary function. However, its interpretation is more complicated than in the case of π΄π −ππ‘ . Significance of b Whether the equilibrium is dynamically stable or not is a question of whether the complementary function will tend to zero as t → ∞. So we must analyze the path of π΄ππ‘ as t increased indefinitely. The value of b is important in this regard. We first consider its significance alone disregarding the coefficient A (by assuming its equal to 1). We can divide the range of possible values of b (-∞, +∞) into seven distinct regions 1 2 3 4 5 6 7 b>1 b=1 0 < b< 1 b=0 -1< b< 0 b = -1 b < -1 |b|>1 |b|=1 |b|<1 |b|=0 |b|<1 |b| = 1 |b| > 1 π‘ (2) (1)π‘ (1/2)π‘ (−1/2)π‘ (−1)π‘ (−2)π‘ t=0 1 1 1 0 1 1 1 t=1 2 1 1/2 0 - 1/2 -1 -2 6 t=2 4 1 1/4 0 1/4 1 4 t=3 8 1 1/8 0 - 1/8 -1 -8 In each region, bt generates a different type of time path. i. ii. iii. iv. v. vi. vii. When b > 1, bt must increase, t increases to ∞ When b = 1, bt will remain at unity for all values of t. (0 < b< 1), bt represent a positive fraction raised to integer power. b t decreases as t increases but remains positive. When b = 0, bt remains at zero for all values of t (not of interest since b≠0) When (-1 < b< 0), bt alternates between positive and negative values from period to period. The alternating path tend closer and closer to the horizontal axis. When (b = -1), bt alternate between positive and negative one. When (b < -1), bt alternate between positive and negative values and deviates further and further from the horizontal axis In summary, the time path of bt (b ≠ 0) will be Non oscillatory if b>0 Oscillatory if b<0 Divergent if |b| > 1 Convergent if |b| < 1 The convergence of the expression bt depends on the absolute value b while that of the continuous case depends on the sign of a in π −ππ‘ The role of A First the magnitude of A can ‘blow up’ or ‘pare down’ the value of bt. It has a scale effect without changing the basic configuration of the time path. The sign of A does materially affect the change of the path because if multiplied by bt it will lead to change of bt if A is negative. Thus, a negative A produces a mirror effect as well as a scale effect. Convergence of equilibrium Based on the above discussion, we can interpret the Abt as the complementary function. Abt represents the deviations from some intertemporal equilibrium level. In summary, π¦π‘ = π΄ππ‘ + π¦π is convergent iff |b| < 1 7 Economic applications Cobweb model Illustrates first order difference equation application in economic analysis. In this model, Qs is a function of price of the preceding time period and not current price. The model considers a situation in which a producer’s output decision is made one period in advance of the actual sale – such as in agriculture production – where planting must precede by appropriate length of time to the harvesting and sale of the output. Assume the output decision in period t is based on the prevailing price Pt. Since the output is not available for sale until period (t+1). Pt determines not Qst but Qst+1. So, we have a lagged supply function. ππ ,π‘+1 = π(ππ‘ ) ππ π‘ = π(ππ‘−1 ) or When this supply function interacts with a demand function of the form πππ‘ = π·(ππ‘ ), an interesting dynamic price pattern will result. Let us assume a linear version of the (lagged) supply and (unlagged) demand functions and assume that in each time period, the market price is always set at a level which clears the market. The market model is given as πππ‘ = ππ π‘ πππ‘ = πΌ − π½ππ‘ ( πΌ, π½ > 0) ππ π‘ = −πΎ + πΏππ‘−1 (πΎ, πΏ > 0) πΌ − π½ππ‘ = −πΎ + πΏππ‘−1 π½ππ‘ + πΏππ‘−1 = πΌ + πΎ To solve the difference equation, normalize first and shift the time subscript ahead by one period. πΏ ππ‘ + π½ ππ‘−1 = πΌ+πΎ π½ πΏ πΌ+πΎ π½ π½ ππ‘+1 + ππ‘ = π¦π‘+1 + ππ¦π‘ = π which is similar to the equation πΏ π¦ = π, π = π½ , π = 8 πΌ+πΎ π½ Since πΏ πππ π½ are both positive → π ≠ −1, therefore we apply the following solution π π¦π‘ = π΄(−π)π‘ + 1+π πππππππ π πππ’π‘πππ πΌ+πΎ πΏ π‘ π½ π¦π‘ = π΄(− ) + π½ 1+π πΌ+πΎ πΏ 0 π½ π¦0 = π΄(− ) + π½ 1+π π¦0 = π΄ + πΌ+πΎ π½+πΏ πΏ πΌ+πΎ π¦π‘ = π΄(− π½)π‘ + π½+πΏ π the general solution π π¦π‘ = (π¦0 − 1+π) (−π)π‘ + 1+π π 1+π = πΌ+πΎ π½ πΏ 1+ π½ = πΌ+πΎ π½ × π½ π½+πΏ πΏ π‘ πΌ+πΎ πΌ+πΎ = π½+πΏ πΌ+πΎ ππ‘ = (π0 − π½+πΏ ) (− π½) + π½+πΏ πππππππ‘π π πππ’π‘πππ The cobwebs Three points regarding the time path πΌ+πΎ 1. The expression π½+πΏ which constitutes the particular integral of the difference equation can be taken as the intertemporal equilibrium price of the model. πΌ+πΎ πΜ = π½+πΏ it is a stationary equilibrium. If we substitute πΌ+πΎ π½+πΏ with πΜ in the equation, we have, π‘ πΏ ππ‘ = (ππ − πΜ ) (− π½) + πΜ 9 2. Significance of the expression (π0 − πΜ ) – corresponds to A in term π΄ππ‘ . It’s negative sign will have a mirror effect and its magnitude will have scale effects. 3. The expression −πΏ π½ which corresponds to b-component of π΄ππ‘ . π½, πΏ > 0. We can deduce an oscillatory time path. oscillatory time paths Explosive Uniform Damped It is this that gives rise to the cobweb phenomena. Three possible πΏ> π½ πΏ=π½ πΏ< π½ if if if An example: Find the time path of price given the following Qdt =10 – 2Pt Qst = -4 + 3Pt-1 Qdt= Qst 10 – 2Pt = -4 + 3Pt-1 10 + 4 = 2Pt + 3Pt-1 2Pt + 3Pt-1 = 14 Pt + (3/2)Pt-1 = 7 Pt+1 + (3/2)Pt = 7, the equation is equivalent to 3 π¦π‘+1 + ππ¦π‘ = π π¦ = π, π = 2 , π = 7 3 7 ππ‘ = π΄(− 2)π‘ + 1+3/2 3 ππ‘ = π΄(− 2)π‘ + 3 14 (π0 − 14 5 πππππππ π πππ’π‘πππ 5 π0 = π΄(− 2)0 + π0 = π΄ + πππππππ π πππ’π‘πππ 14 5 14 5 )=π΄ ππ‘ = (π0 − 14 5 3 )(− 2)π‘ + 14 5 πππππππ‘π π πππ’π‘πππ 10 11