A Tutorial on Simple First Order Linear Difference Equations (for Economics Part I Paper 3) Corrections to Dr Ian Rudy (http://people.pwf.cam.ac.uk/iar1/contact.
html) please. An example of a simple first order linear difference equation is: xt 2 xt 1 1800 [1]
The equation relates the value of x at time t to the value at time (t-1). Difference equations regard time as a discrete quantity, and are useful when data are supplied to us at discrete time intervals. Examples include unemployment or inflation data, which are published one a month or once a year. Difference equations are similar to differential equations, but the latter regard time as a continuous quantity. Equation [1] is known as a first order equation in that the maximum difference in time between the x terms ( xt and xt 1 ) is one unit. Second order equations involve xt , xt 1 and xt 2 . Equation [1] is known as linear, in that there are no powers of xt beyond the first power. There are various ways of solving difference equations. In lectures, you may simply be given a formula for the solution for a general difference equation. This is fine if you have a good memory, but is not terribly interesting. Another method begins from the assumption that we know x0 , and can then use [1] to find the value of x1 . Having done this, we can then use [1] again to find the value of x 2 , and so on. This method is very general in principle, but in practice its usefulness depends on whether we are able to sum the series that appear to get a general expression for xt . We will look at a third method of solving [1] in some detail. It is a two-stage process. We first of all look for any solution - no matter how simple it is, or whether it is the complete solution to the equation. When the right hand side of the equation is a constant, as it is in [1], this is quite simple: we just seek a solution: xt xt 1 x * [2]
This is often known as a steady-state or equilibrium solution. For equation [1], we get: x * 2 x * 1800 so x * 600 [3]
In stage two of the process, we look for a more sophisticated solution, such as: xt x * z t In our case, x * 600 , and by substituting [4] into [1], we get: [4]
600 z t 2600 z t 1 1800
so
z t 2 z t 1 0
[5]
It should be apparent that [5] will always be [1] with zero on the right hand side, and once you realise this, you can save time by jumping straight to [5] from [1]. Equation [5] can be solved in various ways. One way, which very usefully extends to second order equations, is to propose a trial solution of: z t A by substituting this into [5], one finds: A 2 A
t t 1 t
[6]
0 [7]
so, cancelling a factor A :
t 1
20 2
so Hence from [6], the solution is:
z t A 2
In the case of a second order equation, [7] is replaced by a quadratic in , from which you will get two values of (let's call them 1 , 2 ), and the solution for z t is: z t A1 B 2
t t
But returning to our first order equation [1], by putting together [4], [3] and [6], we find the solution is: xt 600 A 2
t
[8]
To find A , we need some information about xt at one value of t. Most commonly, we will know, or be given information about, x0 , known as an initial condition. For example, if x 0 601 , then from [8], A 1 , and so: xt 600 2
t
[9]
In summary, the solution to difference equations of the form of [1] is: xt x * z t
where x * is the steady state solution and z t is found by putting zero on the right hand side of the difference equation, replacing xt by z t and using a trial solution of z t A to find . Hence the general solution is:
t
xt x * A
[10]
The value of the constant A can be found from the initial condition(s). You will come across some other terminology in books: x * is also known as the particular solution or particular integral, and z t is known as the complementary solution or complementary function. We then have: general solution = particular solution + complementary solution. Having found the solution to [1], a question which often arises is how xt varies with t. By plotting [10] against time, you should be able to see that there are four situations we might encounter:
time
time
1 : unstable, oscillating
1 0 : stable, oscillating
time
time
0 1 : stable, not oscillating
1 : unstable, not oscillating
