Some Useful Mathematics for ECON10004 Many economic models start from the assumption that the economic decision maker is actively pursuing a goal. For example, firms in the economy may be interested in maximizing profit, governments may be interested in maximizing welfare, and individuals might be interested in maximizing their own well‐being. Starting with the assumption that decision makers are optimizing is useful for generating precise, solvable models, primarily because we can use a variety of mathematical techniques to understand them. Models based on optimization also turn out to match empirical behaviour well, and so they are viewed as a good foundation for models that seek to describe real behaviour. While we have tried to keep the mathematics simple throughout the course, our view is that you will develop a better understanding of the material if you are exposed to the mathematics that underlies the graphs in the lectures and book. The following guide will review some of the most common ideas with a few worked examples. Optimization Suppose that you are a manager of a charity that resells tickets to the local sports events and uses the proceeds to feed the homeless. The tickets are given to you free and the charity will receive any revenue that you earn. Suppose also that the revenue that you receive depends only on the price that you set. In most circumstances, the quantity of tickets that you sell will be related to the price that you set. If you set the price very low, you will sell all the tickets. If you raise the price, you will sell less tickets, but each ticket that you sell will have a higher revenue per ticket. Mathematically, we could represent this problem as R(P) = Q(P)P, where R(P) is the revenue that the charity receives if you choose a price P. This is equal to the quantity of tickets that you sell, Q(P), multiplied by the price you sell each ticket, P. Note that Q(P) is also a function that returns the number of tickets sold for a chosen price P. Figure 1 shows a graphical representation of your revenue for different chosen prices. As you might expect, if you set the price to zero, you would make no profit since you are giving the tickets away for free. Likewise, if you set a price that is too high, you won’t be able to sell many tickets and your profit will be low. Profits are thus hump shaped and the revenue maximizing price will be in the middle at a price denoted as P*. R(P) R ( P* ) P* P Figure 1: Revenue as a Function of Price As a manager of the charity, it is unlikely that you would know precisely where P* is without some experimentation. For instance, suppose that in week 1 you choose a price P1 = $40 and observe that the revenue is R(P1) = 100 and in week 2, you choose a price P2 = $41 and observe that revenue increased to R(P2) = 120. Based on the experiment, you could conclude that there was a change in revenue of $20 for a change in price of $1. Thus, revenue increased as price increased. Mathematically, we could represent this by R( P2 ) R( P1 ) R( P) 0 or 0, P2 P1 P where the delta is used to mean “the change in”. Having run the experiment, we know that the price of $40 was probably too low and we are likely to improve welfare if we continue to raise our price in small increments. R(P) Line Slope: R ( P2 ) R ( P1 ) R( P2 ) R( P1 ) 0 P2 P1 P P1 P2 Suppose instead that you started at a price P3 =$80 in week 3 and observe that the revenue is R(P3) = 120. You then increase the price in week 4 to P4 = $81 and observe that revenue decreases to R(P4) = 100. Based on the experiment, you could conclude that there was a change in revenue of ‐$20 for a change in price of $1. This experiment would suggest that your prices are too high and that you should lower them. R(P) Line Slope: R( P4 ) R( P3 ) 0 P4 P3 R ( P3 ) R ( P4 ) P3 P4 P As can be seen by these examples, the slopes of the policy experiment can inform you about what to do next. If the slope is positive, you want to increase prices. If the slope is negative, you want to decrease prices. Maximization essentially boils down to finding a policy experiment where the slope is zero for very small changes in P. Derivatives As you hopefully know, the limit of R( P ) for very small changes in P is called the derivative of the P dR ( P ) function, and is denoted by or R '( P ) . More formally, the derivative of R(P) is given by dP dR ( P ) R( P ) R( P ) lim 0 , dP (P ) P where lim 0 just says we are evaluating the function as we take the difference between the two prices, , to zero. Note that this is equivalent to our slope equation above if we were to set P2 P and P1 P. An equivalent expression that is more commonly used is: dR ( P ) R( P ) R( P ) lim 0 . dP Just like our example above, the derivative is simply the slope of the function at a given point. Thus, the derivative of dR( P1 ) is very different from the derivative at dR( P3 ) dP dP First‐order condition for a maximum As we saw in our example, if the derivative of revenue was positive, we would want to increase price. Vice versa, if the derivative of revenue was negative, we would want to decrease price. This intuition leads to a result that is quite general: for a function to attain its maximum value at a point, it must be that the derivative of the function has a slope of zero. Hence, in cases where we know the functional form, an easy way to find an optimum is to take the derivative and set this value equal to zero. Formally, we are looking for the optimal P* such that dR ( P * ) 0 dP One danger you will have to look out for later in your studies is that the slope of a function at a minimum is also zero. Thus, blindly following this rule can get you the lowest possible value. There are formal ways to test for this, but I find the easiest is to recognize that at a maximum, the slope just to the left of P* should be positive and the slope just to the right of P* should be negative. You can check for these by just plugging in slightly smaller and larger numbers and making sure the slopes are correct. Some Useful Derivatives The formula for the derivative can seem rather daunting. Luckily, if you know a few simple patterns, you will be able to derive most things without using it. Here, I will use f(x), g(x), and h(x) since these are the functions used most often in textbooks. I will also use the shorthand f’(x) instead of the more formal df(x)/dx, when expressing the derivative of a function f(x). Finally, I will use the large brackets [] to denote parenthesis. 1. If f(x) and g(x) are functions: d [ f ( x ) g x ) ] f '( x ) g '( x ) dx 2. If c is a constant, then dc 0 dx 3. If c is a constant, then d [cf ( x )] d [ f ( x )] c cf '( x ) dx dx 4. If n is a constant, d [ xn ] nx n 1 dx 5. Note that if n=1, rule (3) and (4) implies: d [cx] c dx 6. Product Rule: If f(x) and g(x) are functions: d [ f ( x ) g x ) ] f '( x ) g ( x ) f ( x ) g '( x ) dx 7. Chain Rule: If z=f(y) and y=g(x) and if both f’(y) and g’(x) exist, then: d [ f ( g ( x ))] [ f '( g ( x ))]g '( x ) dx Discussion: The first of these rules show that if we have a “+” or “‐“ sign anywhere in a function, we can treat each part separately when taking the derivative. This is extremely useful in economics because we are often interested in the relationship between profits, revenue, and costs: ( q) R( q) C ( q) If we are interested in maximizing profits, we are interested in finding the q* where '( q*) 0. Rule one says we can just treat R(q) and C(q) separately and thus we are looking for the place where R '(q ) C '(q) 0, Or, equivalently R '(q ) C '(q ). This is the marginal revenue equals marginal cost rule that you will see many times this year. Rules number 2 and 3 help us to deal with constants. If, for instance, C(q) = 10q+5, C '( q ) 10 d [ q] d [5] 10 0 10 dq dq Finally, the product rule helps us take derivatives of more complex functions and ones where we do not have explicit formulas. Recall, for instance, that in our example at the beginning we had R(P) = Q(P)P If we wanted to take the derivative of this, we could use the product rule to find that: R '( P ) Q '( P ) P Q ( P ). We will see in class that this equation has a nice economic interpretation when discussing monopolies. Partial Derivatives In the later part of the class, we will start to deal with problems where an outcome depends on more than one variable. For example, we may be interested in understanding what happens to the profit of a firm in an environment where the market price depends both on their own quantity q1 and the quantity of a second firm q2: ( q1 , q2 ) 100q1 q12 q1q2 cq1 50q2 In these cases, we would want to consider how profit changes with q1 holding q2 fixed. The partial derivative is exactly this operation. Formally, the partial derivative of profit is ( q1 , q2 ) ( q1 , q2 ) ( q1 , q2 ) lim 0 q1 where q2 means that we are holding q2 fixed at a certain amount. Notice that this is exactly the same as a single‐variable derivative where we treat all the other variables as constants. For the profit function above we would have ( q1 , q2 ) 100 2 q1 q2 c q1 To find the maximum q1 we would again find the place where this slope is equal to zero. Here it will be an implicit function that depends on the quantity chosen by the other firm: q1 100 q2 c 2