I am constantly brain storming challenge questions for my pre calculus students. Today I came up with this question, but upon finding the answer I don't think this one will be feasible for my students.

Suppose during month zero your credit card balance is $b_0$. Sometime during month zero you make the minimum payment, say $p$. The next month the bank calculates your new balance by $b_0 - p + I(b_0 - p)$ where $I$ represents your interest. Assuming you always pay the minimum payment every month, $p$, and your interest $I$ stays fixed, a recursive function that gives your balance in month $n$ could be given by $f(0) = b_0$ and $f(n) = f(n-1)-p + I(f(n-1) - p)$ for $n \geq 1$. Find a closed form for the function $n$.

As always, I set out to make sure I could even solve the problem before giving it to my students as a challenge problem. My solution took me quite a while. Here is the answer I came up with, that could possibly be further simplified:

$$f(n) = \frac{p(1+I)^{n+1} - Ib_0(1+I)^n - P(1+I)}{-I}$$

My method was to start by writing out the first few terms using the recursive definition, then start rearranging and looking for patterns. After a while I was able to spot and prove a binomial expansion, and a geometric sum of binomials was occurring which yields my solution. I will currently not write my full derivation unless requested because it is long and mostly just symbol crunching.

My question to the MSE community is, does anyone have a slick easy derivation of a solution to this problem? I did not thing it should be so tedious when I came up with it? Is there a better method?

I don't even know what to tag this, if anyone has any good suggestion for tags let me know.

  • $\begingroup$ As a practical matter, even if there is a slick derivation of the solution, the fact that you couldn't find it in a reasonable amount of time means it's unlikely that your students would either. $\endgroup$ – David Z Jan 25 '18 at 5:37
  • $\begingroup$ I’m aware, I mention this in post but I don’t think I made it quite precise that I’m not giving it to them just asking for my own curiosity if there’s a way to use more advanced techniques that make this problem easier. I suspect discrete differential equations could do it, maybe $\endgroup$ – Prince M Jan 25 '18 at 6:12

Rewrite the equation as $$f_n = f_{n-1}-p + I(f_{n-1} - p)=(1+I)f_{n-1}-(1+I)p$$ $$f_n = \alpha f_{n-1}+\beta\qquad \text{where}\qquad \alpha=1+I\qquad \text{and}\qquad \beta=-(1+I)p $$ $$f_n = \alpha f_{n-1}+\beta\implies f_n =\beta\,\frac{ \alpha ^n-1}{\alpha -1}+c_1\, \alpha ^{n-1}$$ leading to $$f_0=\frac {c_1} \alpha=b_0 \implies c_1=\alpha\,b_0 $$ making $$f_n=\beta\,\frac{ \alpha ^n-1}{\alpha -1}+ \alpha ^{n}b_0$$ Now, replace $\alpha $ and $\beta$ by their definitions.

  • $\begingroup$ Can you explain the first implication? $\endgroup$ – Prince M Jan 25 '18 at 6:31
  • $\begingroup$ @PrinceM. It is just a linear recurrence equation. $\endgroup$ – Claude Leibovici Jan 25 '18 at 6:50
  • 1
    $\begingroup$ I have no familiarity with such a thing, and I don’t understand yet. I will look into it today and see if I can fill in the blanks. I want to fully understand your answer before I accept it, but I will say thanks in advance for even responding! I am a first year grad student with an interest in Algebraic Geometry (still very novice), methods of this sort are definitely not familiar territory to me. $\endgroup$ – Prince M Jan 25 '18 at 19:47
  • $\begingroup$ Ok, I spent the day learning about recurrence relations. I still don't entirely understand your answer but I have a feeling if I went farther with learning the theory it would be obvious. Thanks for your answer and I must say, I had a lot of fun learning about this stuff today. The methods remind me so much of intro DE's, but cooler in my opinion. $\endgroup$ – Prince M Jan 26 '18 at 2:40

Ok! After finding out that this type of problem is well understood and falls under 'linear recurrence relations', I have spent the day learning about them and I am going to post a more detailed solution as an addendum to Claude's concise answer above.

First of all, this specific problem is an example of a non-homogeneous linear recurrence relation. In order to solve it, I will use methods that require knowing how to solve homogeneous linear recurrence relations which I will not elaborate on as much.

We seek a solution (closed form) to the relation $$x_n = (1+I)x_{n-1} - P(1+I), \text{ for } n\geq 1 (*) $$ with initial value $x_0 = b_0$. This initial value uniquely determines the sequence $x_n$. The corresponding homogeneous linear recurrence relation is $$x_n = (1+I)x_{n-1}. (**)$$

It is a theorem that if $\{b_n\}$ is a particular solution to (*) (not necessarily satisfying the initial condition) and $\{c_n\}$ is the generic solution to (**) then the generic solution to (*) is of the form $b_n + c_n$. So we seek to find such a $b_n$ and $c_n$.

For $c_n$, we use methods of solving homogeneous linear recurrence relations, in particular we find the characteristic polynomial of (**): \begin{align*} ar^n &= (1+I)ar^{n-1}\\ r &= (1+I)\\ r - (1+I) &= 0.\\ \end{align*}

The roots of this characteristic polynomial are apparent by inspection so we have the general solution to (**) $$c_n = k(1+I)^n, \text{ for some } k \in \mathbb{R}.$$

From my reading, finding a particular solution $b_n$ seems to be less systematic (in general), and it seems advised to guess something similar to the control term (which in our case is $-P(1+I)$. Since our control term is constant suppose we just guess $b_n = k_2$ ($k_2$ some constant) and use the definition of the relation (*) to solve for such a constant should one exists. We have:

$$k_2 = (1+I)k_2 - P(1+I) \implies k_2 = \frac{P(1+I)}{I}.$$

So set $b_n = k_2$, then we form the general solution $$x_n = c_n + b_n = k(1+I)^n + \frac{P(1+I)}{I}.$$ Recalling our initial condition $x_0 = b_0$ we evaluate and solve for $k$, $$x_0 = k + \frac{P(1+I)}{I} = b_0 \implies k = b_0 - \frac{P(1+I)}{I}$$.

Substituting back into our general solution yields our final solution:

$$\left(b_0 - \frac{P(1+I)}{I}\right)(1+I)^n + \frac{P(1+I)}{I} =$$

$$b_0(1+I)^n - \frac{P(1+I)^{n+1}}{I} + \frac{P(1+I)}{I} = x_n.$$

The really good news is that this is the original solution I found on my own with no theory at my disposal! I had a lot of fun learning about how to solve these recurrence relations today.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.