Introduction to Calculus for Business and Economics

by Stephen J. Silver Department of Business Administration The Citadel

Introduction to Calculus for Business and Economics

0 -1 0 0.5 -0.8 -0.5 -0.3 0.25 0.75 1

Figure 1. y = 2 - 3x II. Limits and rules

Figure 2. The unit circle

Look at y = f(x) in Figure 3 below for the function f(x) = -x3 +9x2 -15x + 1
30 20 y = f(x) 10 0 0 -10 x 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

Figure 3. f(x) = -x^3 + 9x^2 -15x +1

We see f(4) = 21. Thus, we see that as x approaches 4, f(x) approaches 21. In general, if f(x) has a limit at x = a, then the limit of f(x) as x approaches the value a equals the value of f(x) at x = a. We write this as lim f(x) f(a). xa Sometimes it is not clear what the actual value at a point is, but if the limit of f(x) as x a is well-defined regardless of how we get there, then we can say the function is continuous at all values in its domain. Take, for example, the following case f(x) = (x 1)/(x 1). What is the limit of f(x) at x = 1. We see that the numerator and denominator both are 1-1 = 0. Now 0/0 is not well defined so we might believe that the function has no value at x = 1. But clearly f(x) = 1 for all x, so we define lim f(x) at x = 1 to be 1. As we stated above if the limit of f(x) for all x = a = f(a) in the domain of f(x), we say that f(x) is continuous. Sometimes this does not occur. For example the relationship between the weight of a letter and the postage required is discontinuous at each ounce. Up to and including one ounce the postage is 42 cents. [See rate schedule below] Greater than one ounce up to and including two ounces the postage is 59 cents, etc. In this case the limits differ depending upon the direction you come from; f(1) = limit from below (42 cents), but not the limit from above (59 cents).
First Class Postal Rates for Letters (April 2009)
Weight Not Over 1 ounce 2 ounces 3 ounces 3.5 ounces Price $0.42 $0.59 $0.76 $0.93

If f(x) and g(x) are functions of x, then h(x) = f(x) g(x) are also functions of x, as is

h(x) = f(x) g(x). Also, f(x)/g(x) is a function of x for all g(x) 0, unless f(x) also = 0 at the point and the value of h(x) is well defined as above. In addition, if lim g(x)0 f(x)/g(x) exists, then f(x)/g(x) is also a function of x.

Furthermore, the fact that lim f(x) as x a = f(a) also means that the following hold: lim f(x) g(x) = lim f(x) lim g(x) = f(a) + g(a), xa xa xa lim f(x) g(x) = lim f(x) lim g(x) = f(a) g(a), and xa xa xa

lim f(x) / g(x) = lim f(x) / lim g(x) = f(a)/g(a), if where g(a) = 0, f(a) also =0 and xa xa xa the limit at x = a is well defined.. [LHpital showed that, in this case, lim f(x) / g(x) = lim f(x) /lim g(x) = f(a)/g(a)] xa xa xa III. The derivative; maxima, minima, and points of inflection One very important application of the quotient property above is the special limit known as the derivative function. In Figure 3 above, we saw that f(4) = 21. We also see that f(5) = 26. The slope of the line joining the two points (4,21) and (5,26) is found by taking y/ x = (26-21)/(5-4) = 5. Now what happens as we calculate the limit
y/ x = f(x + x ) - f(x) , as x 0?

The slope of the line joining the two points becomes the slope of the curve at (x, f(x)), that is, the slope of the line tangent to f(x) at x. This limit is called the derivative of f(x) at x, and is written as either f (x) or dy/dx. Thus, we can define this limit as follows: Let y = f(x) be continuous on the interval [a,b], x0, x0 + both [a,b]. Then dy/dx = lim f(xo + ) - f(xo) . 0 Now we notice that the denominator of this function equals 0 at the limit. This would tend to indicate that this function is undefined (ie. = ) at this point. But the numerator also is identically equal to zero at = 0, so the function is indeterminate rather than undefined. To evaluate the function at the limit point, we need to calculate the limit itself. Let us look at some examples. Example 1: Let y = f(x) = a + bx, which we learned is a straight line with intercept at a and constant slope b, which result we will now prove. Then dy/dx = lim f(x + ) - f(x) = lim (a+ b(x + )) - (a + bx) 0 0 = lim b = b. [For simplicity, we use x rather than the specific value xo] 0 Example 2: Let y = f(x) = ax2. lim f(x + ) - f(x) = lim a(x + )2 - ax2 = lim ax2 + 2ax + 2 - ax2 0 0 0 2 lim 2ax + = 2ax. [Note how the third term of f(x + ) went out!] 0

Example 3: Let y = f(x) = axn, where n is a positive integer. Now f(x + ) = a(x + )n = a(xn + nxn-1 + n(n-1)/2xn-22 + + n!/(k!(n-k)!)xn-kk + + n). Thus, (f(x + ) - f(x))/ = anxn-1 + other terms with multiples of powers of . These later terms all equal 0 at = 0. Thus, dy/dx = anxn-1. It is also the case for all n, whether n is a positive integer or not. Example 4. Now, from the additive property of limits, it follows that if y = f(x) = anxn + an-1xn-1 + + akxk + + ao, then dy/dx = nanxn-1 + (n-1)an-1xn-2 + + kakxk-1 + + a1. Thus, the derivative of a polynomial of degree n is a polynomial of degree n-1. Let us return to the function in Figure 3, f(x) = -x3 + 9x2 - 15x +1. The derivative is dy/dx = -3x2 + 18x -15. This equation provides us with the slope of f(x) at each x. Now, the slope at any maximum point or minimum point obviously is zero. So we can set dy/dx = 0, solve the resulting polynomial for x and find these points. In this case we get -3x2 + 18x -15 = 0 x2 - 6x - 5 = 0; x = 5, 1 are the solutions. Indeed, we see that f(x) reaches a maximum at x = 5 and is minimized at x = 1. Since a quadratic function has at most two real solutions, there are no other turning points. In fact, a quadratic equation may have no, one (double), or two solutions. (To review the quadratic formula, go to http://en.wikipedia.org/wiki/Quadratic_equation) The one we just solved had two, and therefore a maximum and a minimum. Figures 4 and 5 below show the cases of a double
20 10 f(x)
f(x) 60 40 20 0
-2

1. 5

0. 5

2. 5

-1 .5

-0 .5

3. 5

-10 -20 -30

x
0 1 2 -1 3

-20 -40 -60

Figure 4. f(x) = x^3-3x^2+3x-5

Figure 5. f(x) = 2x^3-3x^2+6x-5

solution and no solution, respectively. Solving the first, f(x) = x3 - 3x2 + 3x - 5, gives 3x2 - 6x + 3 = 0; or x = 1, 1. This is a point of inflection with slope = 0 at x = 1. An inflection point is one at which the slope stops increasing and begins to decrease, or vice versa. The second function, f(x) = 2x3 - 3x2 + 6x - 5, yields dy/dx = 6x2 - 6x + 6 = 0, with no real roots; we note, however, that there is a point of inflection at x = .5. This is where the change in the slope = 0 (the curve stops getting flatter and begins getting steeper). Points of inflection are points at which the slope achieves a maximum or a minimum; thus, the change in the slope is zero. This means the second derivative of the original function is zero. The second derivative is useful in another capacity. Suppose we find a turning point. We cannot be sure whether this point is a maximum or a minimum. But the value of the

second derivative at the x-value will tell us which it is. If the value of the second derivative is positive, this implies that the slope is increasing; this can only be the case if the slope is changing from negative to positive, which occurs at a minimum point. Similarly, if the second derivative is negative, the function must be at a maximum. Finally, if the first derivative is zero and the second derivative is zero, we must be at an inflection point with zero slope. For the function in Figure 4 above, the value of the second derivative at x = 1 is zero, so the point is an inflection. The function in Figure 3 was f(x) = -x3 + 9x2 -15x + 1. The first derivative is dy/dx = -3x2 + 18x 15. Thus, we solve x2 6x + 5 = 0, for roots, to get x = 1 and 5. The second derivative is d2y/dx2 = -6x + 18, which is negative at x = 5, positive at x = 1, and 0 at x = 3. Thus, we have a minimum at x = 1 and a maximum at x = 5. Also, the second derivative is zero at x = 3; so halfway between 1 and 5, the min and max of the functions, lies the point of inflection. IV. Functions of more than one variable, partial derivatives, and the Lagrangean In the real world we often find that the value of a variable depends on more than one other variable. For example, consumers derive satisfaction, or utility, from the consumption of many different goods. Level of production depends upon the levels of inputs of many different factors: labor, capital, materials, etc. To take into account that the level of output Y depends upon the levels of inputs of several factors, X1 , X2, , Xn, we will write this as Y = f(X1 , X2, , Xn). We are often interested in the effect of a one unit change in one of the inputs, say input i. Will write this partial derivative as Y/ Xi = fi(X1 , X2, , Xn). Let us show how this is done for a few cases. Example 1. Let f(X1 , X2) = 2X1 - 3X2. The f1 = 2 and f2 = -3, since we assume in each case that the level of the other good is unchanged, that is, it is constant. Example 2. Let f(X1 , X2) = 2X1X2. Then f1 = 2X2 and f2 = 2X1. Example 3. Let the output of widgets depend upon the inputs of labor L and capital K as follows: Q = 2L.4K.6. Then QL = Q/ L = 2(.4)K.6L.4-1 = .8K.6L-.6 = .8(K/L).6. This implies that the more labor currently in use the less productive each unit of labor and the more productive each unit of capital. Thus, if you give a worker more capital to use, the more productive that worker is. Q/ L is called the marginal product of labor. Similarly, the marginal product of capital is found to be QK = 2(.6)L.4K.6-1 = 1.2(L/K).4; that is, the marginal product od capital is greater the more labor used and smaller the more capital used. Example 4. Let l = length of a rectangle, w = the width. How much does the area of the rectangle change if we increase the width one unit; the length by one unit? For any given perimeter, what is the maximum area achievable?

Area A = LW, so A/ W = L, A/ L = W. Thus if W = 10, L = 20, A/ W = 20 sq. and A/ L = 10 sq.. [To those purists that would argue that the unit for L and W is inches, not square inches, lets just say that the area is changing at the rate of 20 square inches per one inch change in the width and at the rate of 10 square inches per one inch change in the length. This was pointed out to me by Erik Max Francis. Thanks.] The perimeter of a rectangle P = 2L + 2W, so W = P/2 - L. Thus, A = LW = L(P/2 - L) = PL/2 - L2. To maximize A we set dA/dL = 0. dA/dL = P/2 -2L = 0, so L = P/4. That is L = W and we have a square. Note, in this case we did not need to use the partial derivative as we substituted the value of W in terms of P and L. We could not have solved the problem differently by solving A/ L = 0, and A/ W = 0 simultaneously and then used the relationship between P, L and W, since we would have gotten the optimal solution by setting both L and W equal to zero. We did not impose on the solution that the perimeter not equal zero. There is a method that allows us to impose restrictions on variables while optimizing the values of a function of these variables. This method is known as the Lagrangean method and requires the use of a dummy function called the lagrangean. In the simple example above, we are trying to maximize the objective function A = LW, subject to the condition that L + W = P/2, where P is some number > 0. This constraint can be written as P/2 - L - W = 0. We let the lagrangean L = LW - (P/2 - L - W). is known as the lagrangean multiplier and becomes another variable in the equation. We now maximize L with respect to the three variables of the equation, L, W, and . (1) L/L = W - ( -1) = W + = 0 (2) L/W = L - ( -1) = L + = 0 (3) L/ = -(P/2 - L - W) = 0 Solving (1) and (2) we get that W = L = - . (3) tells us that L + W = P/2; 2L = P/2. So L = W = P/4. V. The total differential The last topic of differential calculus is the so-called total differential function. Let us once again look at the Area function, A = LW and its partials A/ W = L, A/ L = W. A/ W = L means that dA = L dW if L is held constant and A/ L = W tells us that dA = W dL when W is fixed. [Here, we can say dA is in square inches; W and dL are both in inches] If both L and W vary at the same time the new area is given by (L + dL)(W + dW) = LW + L dW + W dL + dL*dW. The total effect on A can be approximated, for only small changes in L and W for which dL*dW is really small, by dA = (A/W) dL + (A/L) dW = L dW + W dL.

L W dW LW L dW

dL W dL dL*dW

Figure 6. The area of a rectangle If L = 20, W = 10 and dL = dW = .1, then dA 20(.1) + 10(.1) = (2 + 1) sq. = 3 sq.. To check, we calculate the actual area before and after the changes. LW = 20(10) = 300 sq. before, and (20.1)(10.1) = 203.01 sq. after the changes. Thus, our estimated change of 3 was approximately correct. The reason we were off is that we omitted the dL dW = (.1)(.1) = .01. If the changes are small relative to the original size of the variables, the error will be relatively small. These additional effects are often referred to as second order effects and, if these are small, may be ignored. Another useful application of the total differential approach is to estimate the effect on total product of an increase in the amount of labor and capital in the production function we studied earlier, Q = 2L.4K.6, where L is in hours of labor and K is the number of machines. Then dQ is given by QL dL + QK dK = .8(K/L).6 dL + 1.2(L/K).4 dK. Suppose initially L = 100,000 hrs., K = 32 machines. Then Q = 2(100,000).4(32).6 = 2(100)(8) = 1600 widgets. Now let dL = 100 hrs., dK = .01 machines. dQ = .8(32/100,000).6(100) + 1.2(100,000/32).4(.01) = .8(8/1000)(100)+1.2(100/4)(.01) = 6.4 + 0.3 = 6.7 widgets. To check: Q2 = 2(100,100).4(32.01).6 = 1606.682 increase of 6.7 widgets from 1600. VI. Integral calculus Elementary calculus consists of two parts: differential calculus covered in sections I through V, and integral calculus. Integral calculus is closely related to differential calculus, as we shall see shortly, but deals with a different issue, namely, the determination of the area under a curve.

30 f(x+2dx) 25 f(x+dx) f(x) 20 f(x) 15 10 5 0 2.5 3

3.5

xo

4.5

5.5

Figure 7. y = f(x)

Suppose we want to find the area under the curve y = f(x) between x = a and x = b in Figure 6 above. We could approximate that area by finding the areas of a series of very narrow rectangles, say of width x, from a to b. We see three such rectangles starting with the one centered at point x = x0 on the figure. The area of this rectangle is approximately f(x0)x. The next rectangle has an area of approximately f(x0 + x)x and the third rectangles area is about f(x0 + 2x)x. We could thus calculate the area from a to b by taking the sum x=b f(x + x)x. x=a In order to improve on the accuracy of the estimate, all we need to do is let the rectangles be narrower and narrower, that is, have x 0. Thus, the exact area is given as x=b lim f(x + x)x. x0 x = a This limit is written as a f(x) dx, called the integral of f(x) from x = a to x = b. Now, let F(x) be the general integral of f(x). What is the functional form of dF(x)/dx? This is the change in the area function F(x). But the change of the area is the height of the function, which is just f(x) itself. This means that the integral of f(x) is merely the anti-derivative of f(x); that is, the function whose derivative is f(x). It should be noted that the exact function is unknown to within a constant of integration. That is because the derivative of a constant is zero. Example. To find the area under the function f(x) = 2x - 3 between x = 2 and x = 5, we first find the function F(x) that has the derivative 2x - 3. We know it is a polynomial of degree 2; in fact, F(x) = x2 - 3x + c, where c is any constant, has the desired derivative. Now since the lower limit of possible values of x is 2, F(2) must equal 0.
b

F(2) = 22 - 3(2) + c = 0 c = 6 - 4 = 2. Thus, F(x) = x2 - 3x + 2 and F(5) = 25 - 3(5) + 2 = 12. To check, we calculate the area of the triangle and the rectangle that lie under the curve between 2 and 5. Since f(2) = 1 and f(5) = 7, the area of the triangle is bh = (5-2)(7-1) = 9; the area of the rectangle is (5-2)(1-0) = 3; and 9 + 3 = 12.

Appendix: The derivatives and integrals of some common functions. Function ex erx ax ln(x) x sin x cos x Derivative ex rerx ln(a)ax 1/x /x cos x -sin x Integral ex ( /r)erx ax/ln(a) x(ln(x)-1) (2/3)x3/2 -cos x sin x
1