Economics 685 Managerial Economics Notes

David L. Kelly Department of Economics University of Miami Box 248126 Coral Gables, FL 33134 dkelly@miami.edu Current Version: Fall 2013

INTRODUCTION
I A Managerial Economics what is Managerial Economics?

Denition 1 Managerial Economics is the application of economic theory to decisions made by managers and rms. Denition 2 Economics is the study of the allocation of scarce resources. Economics is the study of the allocation of scarce resources. Because all decisions are essentially about the allocation of scarce resources, economics is in fact the study of decision making and problem solving in general. So we will simply apply economic rules for decision making to problems faced by managers. B Allocating scarce business resources

Some typical business decisions viewed as an allocation of scarce resources. 1. Going Green: A trucking company can save fuel by reducing the speed at which the truckers drive. However, since truckers take longer to arrive at their destination, total labor costs rise (the trucking company must pay for more hours of labor by the truckers). So exactly what speed should the truckers drive? Or alternatively, how much labor and how much fuel produce truck miles traveled at the least possible cost? Both fuel and labor are scarce resources. Many companies have going green initiatives, which involve reducing fuel or electricity consumption, but often at the cost of using more of other inputs. 2. Budget Allocation: A marketing executive may use both traditional TV advertising and social media advertising. The advertising dollars in the executives budget are the scarce resource. Exactly how many dollars should be spent on each type? 3. Market Entry: In 2005, Walmart entered the DVD-by-mail market, challenging Netix. Wal-Mart is a larger operation and has lower costs. However, Netix was the rst mover and had already established a large customer base. Should Walmart enter or invest its scarce capital in another business?

II

Decisions Studied in Managerial Economics

Although managerial economics principles can be applied to any decision, managerial economics typically focuses on three types of decisions. 1. How much to produce? How many cruise passenger spots should Royal Caribbean oer per day? 2. What inputs to use? How many ships, sta, entertainment workers, etc. should Royal Caribbean use? 3. What price to charge? Should Royal Caribbean charge a xed markup over costs (cost-plus pricing) or some other method? Notice that these three decisions are very high-level decisions, typically made by CEOs and senior management, not newly minted MBAs (unless they are entrepreneurs or work in a small rm). This is one reason why the most common major among CEOs is economics. Nonetheless, the examples below illustrate that the economic decision making may be applied to virtually any decision. Further, some of you no doubt aspire to be CEOs or entrepreneurs, and so will eventually need this information. III Managerial Economics advice for decision making

Economic decision making gives many general principles for making business decisions. MBA students who forget the math may years later still nd the course helpful by remembering these principles. Beyond the principles, managerial economics provides a scientic method for making business decisions. A Steps 1. Formulate the problem: list the objective and the possible decisions. 2. Gather data on the objective and the decisions. 3. Using statistics, estimate the relationship between the objective and the decisions, known as the objective function. 4. Using calculus, spreadsheets, or game theory, choose the decision that maximizes the objective function. 2

B 1

Examples Example 1: Marketing Budget

Consider again the marketing executive deciding between TV and social media. It is reasonable to conjecture that diminishing returns exist: as more and more advertisements are made, each additional dollar gives less and less additional sales. The rst ad is fresh and reaches many previously unaware consumers. The 100th ad generates little additional sales, everyone is familiar with the product. We might conclude from the principle of diminishing returns that it is best to spread the advertising dollars around. But can we go further and say exactly how much to spend? Step 1. The objective is presumably to maximizes sales (if the advertising budget is xed and the goal is to get as much out of each marketing dollar as possible) or prots (if the executive has stock options or bonuses based on prots). Let us suppose the former. Here is an objective function that has diminishing returns:
2 S = a + b1 Atv b2 A2 tv + c1 Af b c2 Af b

(1)

Here A is advertising expenses (f b is Facebook), S is sales, and a, b, and c are unknown parameters (A parameter is a variable which the manager cannot change). Step 2. Our executive has data on past sales and advertising expenditures. We also of course need to know the cost of a TV ad and the cost of a Facebook ad. Step 3. We use statistics to nd the unknowns a, b and c. Suppose we get (sales are in millions of dollars):
2 S = 22 + 5Af b 1.5A2 f b + 24Atv 0.5Atv

(2)

We also have our budget numbers. We have $0.5 million to spend, a block of TV ads costs $5 million, while a set of Facebook ads costs $1 million. So TV ads are more expensive, but reach a wider audience. So the budget is: $0.5 = 5Atv + 1 Ag (3)

Step 4. Use calculus or a spreadsheet to nd the values of Atv and Ag that maximize S , given our scarce resources (budget dollars). We would nd that: Atv = 0.086 ad blocks or about $430,000 and Af b = 0.07 ad blocks or about $70, 000.1 Although TV ads are more
By the way, this executive should be arguing for a larger budget: spending $2 million on Facebook marketing and $128 million on TV ads (total budget of $130 million) would maximize prots.
1

expensive, the extra expense is more than oset by the larger audience. The example illustrates that we can use economics for some general helpful principles. In addition, we can use economics to make exact decisions which improve prots. Companies are increasingly making more precise decisions, especially with the advent of cheaper computers and large customer data sets. 2 Example 2: Airline Pricing

Airlines have customers which are both price sensitive (typically consumers) and customers which are not price sensitive (typically business travelers). How can the airline charge more to the business travelers? This is the concept of price discrimination, which we will look at in section III below. Recognizing that ways do exist to charge customers dierent prices is the economic principle. However, the more advanced question is exactly how much should the airline charge each type in order to get the highest possible sales revenue? Step 1. The objective is to maximize sales revenue, since the cost of a seat on the plane is the same regardless of who sits in it. Sales revenue is: S = pc qc (pc ) + pb qb (pb ) . (4)

Here q is the quantity of seats purchased at each price p, c are consumers and b are business travelers. Notice that the quantity purchased depends on the price. Steps 2-3. To compute the exact prices, an economist would use data mining of millions of ticket sales to compute statistically exactly how price sensitive consumer and business travelers are (the demand q (p)). Step 4. Then use calculus or a spreadsheet to compute the exact prices. Increasingly, this is what is being done. Not only for airlines, but seats at sports stadiums and other entertainment venues. Companies across the world are collecting vast customer databases. MBAs who know how to make use of such data will be in high demand, and their companies will be more protable. C Alternative Methods of Making Decisions

Alternative methods include relying on experience or expertise. One can actually do a pretty good job relying on experience, if you have it. Nonetheless, most MBA students would not be here if they had already 30 years of experience in their eld.

The movie Money Ball comes to mind here. In that movie a young economics major tells a baseball executive that his analysis revealed that the team should recruit players who walk rather than players who hit home runs. The team experts believed the opposite. The executive lled the team with players who walk and the strategy worked. Both decision methods require common sense to work. The data used above may have been collected in a record year, giving numbers that are unlikely to be repeated. The executive may have data on TV advertising in one location and falsely suppose all locations work the same. IV A Theory of the Firm Objective of the rm

What is the objective or goal that a manager has in mind when making decisions? At the most general level, managerial economists suppose that managers try to maximize the value of a rm. Denition 3 The value of a rm is the present value of the rms expected future prots. Prots are revenues less costs. Thus: 2 n 1 + = pv = 2 + ... 1 + i (1 + i) (1 + i)n t = T Rt T Ct
n

t=1

t (1 + i)t

(5)

(6)

Here pv is the present value of the rm, n is the planning horizon, i is the appropriate interest rate (or the rate of return that could be earned if the prots were invested elsewhere). Also, is prots and T Rt and T Ct are total revenues and total costs. B Economic Prots

Note that prots here are economic prots. Denition 4 Economic Prots are prots after taking into account capital and labor provided by owners. Prots as normally recorded are known as accounting prots. Economic prots are lower, they subtract from total revenues opportunity costs. 5

Denition 5 Opportunity Cost is the cost of a decision in terms of the next best alternative not chosen. Example. Suppose the market rate of return is 10%. A caterer is considering buying an additional kitchen for $240,000. The owner values the additional time required to supervise the new kitchen at $40,000 per year (or suppose she could hire someone at this price, but prefers to do it herself). She gures the accounting prot from the extra business is $60,000 per year. To get the economic prot: Economic Prot = accounting prot return on capital used owners labor (7) = $60, 000 $24, 000 $40, 000 = $4, 000. (8)

Here the next best alternative use for the caterers capital is to invest it in the market at 10%. Since the caterer values her time at $40,000, the next best alternative for her time must be worth $40,000. In this case, the economic prot is negative. Although the additional kitchen would make money, the owner would do better by investing the $240,000 in the market. Thus the optimal decision would be not to buy the kitchen. Clearly, accounting prots are good for maintaining records, but economic prots are needed to make decisions. In the long run, economic prots are generally driven to zero. If economic prots are negative, rms will drop out of the market and prices will rise, increasing prots. Conversely, if economic prots are positive rms will enter the market and competition will drive prices down. Thus unless the industry has barriers which prevent rms from entering and exiting (eg. monopolies or patents) or regulations on price setting exist (eg. electricity prices), economic prots tend to zero. C 1 Other objectives Other possible objectives of rms

We claim that rms maximize prots. Firms that do not tend to go out of business. CEOs who do not maximize prots tend to be replaced, since shareholders want CEOs to maximize prots. Many other objectives, such as maximizing market share, are just intermediate goals toward the nal goal of maximizing prots (remember, by setting a price of zero one could always maximize market share so it is doubtful if any rms, despite their claims, really do 6

this). Similarly, satiscing, or meeting prot goals, are approximations to maximizing prots done when it is not clear what maximum prots are. 2 Other possible objectives of managers

It is important to realize that individual managers or CEOs may have other objectives. The marketing executive trying to maximize sales generally has no incentive to return some of his budget if he gures that $1 of ads generates less than $1 of prots from extra sales. In general, managerial incentives should focus on maximizing prots. Prot incentives include stock options and stock incentives, and bonuses paid relative to protability. However, this may be dicult because: 1. Most managers make decisions that enter on the revenue or cost side, but not both. 2. In a large corporation, individual managers have little eect on prots. D Prot maximization, ethics, and welfare

Maximizing prots improves the welfare of society in two ways: 1. The rm provides products and services that consumers want (as evidenced by their willingness to pay for the products, and thus add to rm prots). 2. The rm stays in business, thus providing income to workers and stockholders. Famous theorem in economics: in general, maximizing prots maximizes the welfare of society. For example, consider price gouging of gasoline after a hurricane. By charging high prices when supply is low, gas retailers insure that the consumers who most need the gas (as evidenced by their willingness to pay high prices) get it. Further, retailers have a greater incentive to invest in costly inventories of gas if a higher price may be charged after the hurricane. In addition, the retailer generates income in the form of wages and prots for stockholders. Charging a low price means that whoever gets the small supply of gas is whoever gets to the station rst, for example. Another example is corporate giving. A corporation can give prots to charity or give prots to stockholders. It is not necessarily true that it is socially better for the corporation to give. It depends on what the stockholders would have done with the money. Perhaps stockholders really need the money, or are better at picking charities.

PRODUCTION THEORY
Production Theory helps managers decide what inputs to use. The book presents the material largely from a manufacturing prospective, for example how many machines and how many laborers to use. But the material could equally apply to services. For example, a nancial planning rm deciding how many assistants should be assigned to each nancial planner. I A The Production Function Denition

Denition 6 The Production Function is a graph, table, or equation showing the maximum output rate that can be achieved by any specied set of inputs. For example, consider the production function in the book, for Thompson Machine Co. Thompson has ve machines in a shop that produces machine parts. Lets suppose the input number of machines is xed in the short run. Let Q be hundreds of parts produced per year, and let L be the number of full time workers, again per year. Then: Q = 30L + 20L2 L3 Here is the information in table form. Full time laborers (L) 0 1 2 3 6 9 12 14 15 Parts Produced (Q) 0 49 132 243 684 1161 1512 1596 1575 (9)

Table 1: Production data for Thompson Machine Co. The production function tells us how much inputs to use.

1. Set inputs to produce a specied quantity of outputs. If the manager expects 700 orders this year, he knows to hire 7 workers (from the table, 6 produces only 684). 2. Hire enough workers to meet the objective of maximizing production. See section II. 3. Choose inputs to maximize prots. With information on wages, hire enough workers to meet the objective of maximizing prots. See section III Where does the production function come from? We can use statistics to estimate a production function. It is also possible to get data from engineering. For example, gas is an input to a trucking rm which produces output miles traveled. The production function is easily computed using the trucks MPG rating. For more examples of how to compute production functions see section VI. B Properties

Properties of the production function. 1. Zero inputs implies zero output. Production cannot take place without inputs. 2. Positive marginal product. Up to a point, adding additional workers increases output. 3. Denition 7 Law of diminishing marginal returns: if all other inputs are held constant, then the additional output from increasing one input eventually falls. Hold the number of machines constant at ve. Then going from four to ve workers is no problem: each can use one machine. Adding one more worker has a lower marginal return: he can only assist one of the machine operators. Adding (in this case) the 15th worker results in zero additional output: he can only stand and watch. Adding still more workers decreases output: extra workers now get in the way.

The graph below illustrates these concepts.

Output of Thompson Machine Co. 1600

1400

Slope or derivative of objective is zero at the maximum. Adding an 11th laborer increases production more than adding a 12th laborer (diminishing marginal product).

1200 Q=Parts Produced Per Year

1000

800 Adding Laborer increases production (Increasing Marginal Product).

600

400

200

8 10 12 L=Full Time Laborers Per Year

14

16

18

Figure 1: Typical production function using one input.

II

Application 1: Maximizing Production

Suppose you a manager assigned to a particular shop with ve machines. A shortage of machine parts exists. The orders are to crank out as many parts as possible. 1. Formulate the problem. The objective is to maximize production of parts. The decision is how many workers to hire. 2. Gather data on the objective and the decisions. 3. Estimate the objective function, which is the production function. Suppose your engineers have come up with the production function Q = 30L + 20L2 L3 . 4. Using calculus or a spreadsheet, determine the optimal decision.

10

So mathematically, the problem is: max Q = 30L + 20L2 L3

L

(10)

Solution 1: read the graph

By looking at gure 1, we can see that output is maximized when adding an additional worker adds zero output. This occurs at about 14 workers (1596 parts). B Solution 2: use the table

Denition 8 The Marginal Product is the additional output from an additional unit of an input. Therefore, the marginal product is the slope of the production function. For production data in table form, such as table 1, the marginal product is approximately: MP Q . L (11)

The marginal product is the slope of the production function. From gure 1, we want to nd where the slope or marginal product is zero. So we can compute the maximum production from Table 2: Full time laborers (L) 0 1 2 3 6 9 12 14 15 Parts Produced (Q) 0 49 132 243 684 1161 1512 1596 1575 Marginal Product (MP ) = (49-0)/(1-0) = 49 = (132-49)/(2-1) = 83 111 147 = (1161 - 684)/(9-6) = 159 117 42 -21

Table 2: Production function and marginal product. To maximize production, we set MP = 0. This occurs somewhere between 14 and 15 workers.

11

Solution 3: use calculus

Alternatively, if through overtime or part time, we can hire part of a worker, we could use calculus. Since the actual production function is available (Q = 30L + 20L2 L3 ) we can compute the actual marginal product using calculus: MP = dQ . dL (12)

From the graph, when output is maximized, the marginal product which is the slope or derivative is zero. Thus: marginal product of labor = The derivative is: Q = 30 + 40L 3L2 = 0 L The solution is given by the quadratic formula: 40 1600 4 (3) 30 b b2 4ac L= = = 14.05 2a 2 (3) (14) Q =0 L (13)

(15)

Notice I have discarded the negative solution. So in fact (if possible) the manager should pay some overtime so that the equivalent of 14.05 full time workers are employed to maximize production. It is clear that the easiest way to solve this problem is to look at the graph. This will depend on the exact problem we are solving and the data which is available, however. Data from a table is the most commonly found in business. III A Application 2: Maximizing Prots Example

Suppose now the manager is paid a bonus based on the prots of the company. To maximize prots we need to know how output aects total revenues and how inputs aect total costs. Suppose the parts can be sold for $500 each (per hundred parts) and workers earn a salary of $45,000 per year.

12

Total revenues are thus: T R = $500 Q = $500 30L + 20L2 L3 Total costs are: T C = $45, 000L Follow our steps again. 1. The objective is to maximize prots, the choice is the number of full time employees. 2. The only data we need are the cost of labor ($45,000) and the price of the parts ($500). 3. We again do not need to do any estimation beyond the production function. 4. Maximize prots, which are: = T R T C = $500 30L + 20L2 L3 $45, 000L We have: max $500 30L + 20L2 L3 $45, 000L
L

(16)

(17)

(18)

(19)

= max $15, 000L + $10, 000L2 $500L3 $45, 000L

L

(20)

Set the slope or derivative equal to zero: d = 15, 000 + 20, 000L 1, 500 L2 45, 000 = 0 dL Divide by 1000: 20L 1.5 L2 30 = 0 L= 20 202 4 (1.5) (30) = 11.6 2 (1.5) (22) (21)

(23)

Either answer works, I have chosen the larger. So to maximize prots, we should hire 11.6 workers. 13

Marginal Revenue Product

To help with the intuition, note that: = TR TC dT C dT R = dinput dinput We call the rst term the marginal revenue product (MRP). Denition 9 Marginal Revenue Product is the amount of additional revenue from an additional unit of an input. The second term is the marginal expenditure (ME). Denition 10 The Marginal Expenditure is the amount of additional costs from an additional unit of input. So we hire additional inputs until marginal revenue product equals marginal expenditure. Ignoring the fancy jargon, we see that we should hire a worker if the worker produces more revenue than the cost of hiring that person. Such a worker makes money for the rm. Finally, notice that: MRP = dT R dT R dQ = = MR MP dinput dQ dinput (26) dT R dT C d = =0 dinput dinput dinput (24)

(25)

Thus the additional revenue from an additional unit of input is equal to the marginal revenue (MR) times the marginal product. For prot maximization, therefore we have: MRP = MR MP = ME. C Table example using marginal revenue product (27)

We can compute the prot maximizing amount of labor from table 2 using the concept of marginal revenue product and marginal expenditure. First, compute the marginal revenue product:

14

Full time laborers (L) 0 1 2 3 6 9 12 14 15

Parts Produced (Q) 0 49 132 243 684 1161 1512 1596 1575

Marginal Product (MP ) 49 83 111 147 159 117 42 -21

Marginal revenue product (MRP = MR MP ) = $500 49 = $24, 500 = $500 83 = $41, 500 = $500 111 = $55, 500 $73,500 $79,500 $58,500 $21,000 $-10,500

Table 3: Production function and marginal revenue product. The marginal expenditure is the cost of one additional worker, ME = $45, 000. So prots are maximized where marginal revenue product is $45, 000, between 12 and 14. Now the true answer is 11.6, so the table gives an answer that is slightly o. This is because the formula for marginal product in the table is an approximation: MP = IV A Q dQ . dL L (28)

Multiple Inputs Isoquant

We now suppose there is more than one input to choose from. Everything will work the same, but there are more things to keep track of. Let L and K denote the two types of inputs, for example workers and machines (capital), or workers and managers. The price of capital is PK and the price of labor is PL . Denition 11 An Isoquant is a curve which shows all possible input levels capable of producing a given output level. Figure 2 gives a typical isoquant.

15

K K0

K1

Q = 30 Q = 20
K Slope = L or = dK dL

Q = 10 L

L0 L1

Figure 2: Isoquants for a typical production function corresponding to Q = 10, 20, 30. Looking at gure 2, we see that increasing either or both inputs results in a move to an isoquant with more production. More inputs means more production. Second, the isoquant is convex. Consider moving from the red point K0 , L0 to the blue point K1 , L1 . Because at K0 , L0 already a fair amount of capital is being used, we have diminishing returns. Extra units of capital above K0 add very little output. Therefore, replacing a little labor (L0 to L1 ) requires a lot of capital (K0 to K1 ) to keep the same production. B Marginal Rate of Technical Substitution and Price Ratios

The slope of the isoquant is critical: it represents the amount of capital required to replace one unit of labor. Denition 12 The Marginal Rate of Technical Substitution (MRTS) is the rate at which one input is substituted for another while keeping production constant. It is the slope of the isoquant.

16

It is best to think of the MRTS as the ratio of marginal products: MRT S = dK MPL dL = . = dQ dL MPK dK
dQ

(29)

That is, if we decrease L by 1 unit, MPL tells us how much production is lost. Then

1/MPK tells us how much K we need to get the same Q units of production back. Notice from the graph, if K is on the y-axis the MRTS has MPL /MPK , not the reverse. Now we know one critical piece of information: how many machines are needed to replace one worker. But to determine which inputs to use, we need to know also how much each input costs. It may be that we can replace many workers with a machine, but if the machine is expensive enough, it is not worth it. C Isocost

Suppose now the manager has a budget of M dollars. What are all possible ways to spend the budget? Denition 13 The Isocost curve shows all input combinations that have the same total budget. The manager could spend the entire budget on labor, in which case: pL L = M, L = M , pL (30)

where pL is the price of labor (the wage). Similarly, the manager could spend the whole M . Figure 3 gives the remaining combinations: budget on machines and buy K = p K

17

M PK

PL Slope = P K

M PL

Figure 3: Isocost for a budget of M . The slope of the isocost is the price ratio pL /pK , which gives the cost of replacing one unit of labor with capital. We now have enough information to get the optimal input choices. D Optimal input choice

I claim that optimal input choice requires setting the MRTS equal to the price ratio: MRT S = pL MPL = , MPK pK (31)

MPL pL = . MPK pK Suppose instead that MRT S <

pL . pK

(32) Then we can replace one unit of labor with pL /pK

units of capital. Costs are constant, but production increases. We needed only MRTS units of capital to keep production constant with the loss of one unit of labor but got more than that since pL /pK > MRT S .
pL . Reversing the argument, we can replace one unit of Conversely, suppose MRT S > p K capital with pK /pL units of labor. Costs are again unchanged. Production rises since we

need only 1/MRT S units of labor to replace one unit of capital and: pK 1 > . pL MRT S Graphically, 18 (33)

pL MRT S > p K

pL MRT S < p K

Q = 10 Q < 10 L
Figure 4: Ecient and inecient input combinations. Optimal input choice means the slopes of the isocost and isoquant equal, which means the curves are tangent. In gure 4, the curves are tangent at K and L . Any movement away from K and L results in less production (Q < 10). E Example: Multiple Inputs

Suppose an engineering analysis rm uses engineers and technicians to do their consulting. Engineers are paid $4,000 per month and technicians $2,000. The production function was found to be: Q = 20E E 2 + 12T 0.5 T 2 (34)

The rm charges $1,000 to do an engineering analysis. This is a very common situation, engineers are more productive, but also cost more than technicians. The same problem comes up with doctors and nurses, research faculty and lecturers, nancial planners and assistants, etc.

19

1. How many engineers and technicians should be hired if the manager is given a maximum wage bill of $28,000? We know that the ratio of marginal products equals the price ratio at the optimal input choice: MPE = MPT Or: $4000 20 2E = =2 12 T $2000 20 2E = 2 (12 T ) 2E = 4 2T E = T 2 The rm should always have two fewer engineers than technicians. Thus, since the wage bill is $28,000: $28, 000 = $4000E + $2000T (40) (36)
dQ(T,E ) dE dQ(T,E ) dT

PE PT

(35)

(37)

(38)

(39)

14 = 2E + T

(41)

14 = 2 (T 2) + T 18 = 3T T =6 E=4

(42)

(43)

2. How many engineers and technicians should be hired to maximize prots? We know already that we need two fewer engineers than technicians. Further, we know that for any input the marginal revenue product equals the marginal expenditure. For example, for engineers: T R = $1000Q, 20 (44)

MR =

dT R = $1000, dQ

(45)

MRPE = MR MPE = $1, 000 (20 2E ) . T C = $4000 E + $2000 T. MEE = Therefore: MRPE = $1, 000 (20 2E ) = MEE = $4000, 20 2E = 4, E = 8. Finally since we always have two fewer engineers than technicians, we have: T = E + 2 = 10. dT C = $4000. dE

(46)

(47)

(48)

(49)

(50)

(51)

3. How many engineers and technicians should be hired if the rm needs to perform an output of 166 engineering analysis? We need Q = 166 = 20E E 2 + 12T 0.5 T 2 Next we use that we always want two fewer engineers than technicians: 166 = 20 (T 2) (T 2)2 + 12T 0.5 T 2 166 = 20T 40 T 2 + 4T 4 + 12T 0.5T 2 166 = 36T 44 1.5T 2 0 = 1.5T 2 36T + 210 21 (53) (52)

(54)

(55)

(56)

T = V A

36 +

362 4 (210) 1.5 = 42/3 = 14 E = 12 2 1.5

(57)

Mergers and Spinos Returns to Scale

Here we think about the size of our production processes. Should the rm expand (say through a merger) or contract (say through a spin-o). Should the rm build a second factory or increase the size of the current factory? Should the rm close one plant and move operations to another? Should the rm outsource some production processes? The production function tells us the answer. Denition 14 The production function exhibits Increasing (decreasing, constant) returns to scale if increasing all inputs by a proportion increases output by more than (less than, exactly) the same proportion It is easiest to think of returns to scale by setting the proportional increase to 100%, that is, double all inputs and see if output more than, less than, or exactly doubles. Doubling all inputs can be thought of as building an identical factory or merging with an identical rm. If increasing returns to scale exists than the merger is benecial, since the total costs do not change (the costs are the same as the costs of the two rms acting separately), but output is greater. Thus the cost per unit falls. Similarly, with decreasing returns to scale, if half the rm is spun-o, then costs are split in half, but output falls by less than half. Thus cost per unit falls. Consider our previous production function where engineers and technicians produce engineering analyses, and let us double all inputs: Q = 20 (2E ) (2E )2 + 12 (2T ) 0.5 (2T )2 = 40E 4E 2 + 24T 2T 2 On the other hand, if we double output, we get: 2Q = 2 20E E 2 + 12T 0.5T 2 2Q = 40E 2E 2 + 24T T 2 22 (60) (58)

(59)

(61)

So Q < 2Q, doubling all inputs resulted in less than double the output. This rm has decreasing returns. In the long run, I would recommend splitting this plant into two smaller units. Reasons for decreasing returns to scale: 1. Extra layers of management are required in a large organization which do not directly contribute to overall production. 2. Information may not ow well to all employees in a large organization. 3. Individual workers have little impact on prots, and therefore have little incentive to engage in prot maximization. 4. Regulation costs may increase in larger rms. For example, many regulations exempt rms with less than 50 employees. Reasons for increasing returns to scale: 1. Indivisibilities. It may be dicult to hire part time accountants, one full time accountant must be used. But then doubling the size might not require any additional accountants. Note, however, that small rms may avoid indivisibility costs by outsourcing. 2. Engineering Reasons. It may be the case that doubling the size of the warehouse might not require double the steel, electricity, etc. 3. Specialization. Large rms can have employees become more ecient by specializing. A larger rm can hire an accountant, rather than have the head sales guy also do the accounting. Outsourcing can help small rms stay specialized as well. B A special production function and output elasticity

Denition 15 The output elasticity is the percentage increase in output from a one percent increase in inputs. For a special production function, the output elasticity is constant and very easy to calculate. This production function is the Cobb-Douglas production function and takes the form: Q = AK a Lb . 23 (62)

Here A, a, and b are parameters which vary across industries. Consider the telephone industry in Canada. The production function was found to be: Q = 0.70L0.70 K 0.41 A one percent increase in inputs gives: Q = 0.70 (1.01L)0.70 (1.01K )0.41 = 0.70 (1.01)0.70 L0.70 (1.01)0.41 K 0.41 = (1.01)0.70 (1.01)0.41 0.70L0.70 K 0.41 = (1.01)0.70+0.41 Q (64) (63)

(65)

(66)

(67)

= 1.0111Q

(68)

Thus if inputs increase by 1%, outputs increase by 1.11%, so the output elasticity is 1.11. The Cobb-Douglas production function has the special feature that the output elasticity is constant. Further, for the Cobb-Douglas production function the output elasticity is simply the sum of the exponents: a + b = 0.71 + 0.40 = 1.11. Does this function have increasing, decreasing, or constant returns to scale? How could a telephone monopoly increase or decrease its size? VI Appendix: How Do We Find the Production Function?

We can do many great thing with the production function. How do we obtain the production function? Several ways exist, one of which involves gathering company data (or even competitors data) and then using statistics (regressions). What we need is data on inputs used and how much output was obtained. A second method is to use engineering data. I will discuss here using company data. A Obtain input-output data

The rst step is to obtain some data. Here are several possibilities.

24

1. Time Series Data. Get historical data of the rms inputs and outputs. 2. Cross Section Data. Get data of all plants (or factories) owned by the rm in a single time frame. Or get data on all of the rms in the industry. 3. Use technical information supplied by engineers. 4. Conduct a randomized study. Select a random factory or factories and change the inputs. 5. Benchmarking. Observe rms outside the industry that specialize in this type of production.

25

Each has advantages and disadvantages. Time series data is often easy to come by. The manager does not need to request data from other managers or look up data on other rms. However, things change over time. Suppose the time series data is something like: Date April 3 April 4 Full time laborers (L) 6 7 Parts Produced (Q) 684 681

Table 4: Time series production data. One might look at this data and think that the MP becomes negative after 6 workers, that is, the rm should never hire more than six workers. But it is also possible that something odd happened on April 4 that did not happen on April 3. For example, it could have been someones birthday on April 4, and the rm wasted a few hours giving out cake. Cross sectional studies take more time to acquire the data. To a lesser degree, cross sectional data has the same problem. There might be something special about one factory that the manager does not know about. With this data, we would again conclude that hiring Plant West Palm Beach Miami Full time laborers (L) 6 7 Parts Produced (Q) 684 681

Table 5: Cross sectional production data. more than 6 workers is a mistake. But it could be that the Miami plant uses older equipment, and so the workers are not as productive. The randomized study, as is done in medicine, does not suer from these problems, but is the most expensive way to obtain data. Because the plant selected is random, dierences between plants are random and tend to average out. But you have to make the plant use odd combinations of inputs, possibly resulting in low production, just to gather data. B Choosing a production function

The second step is to choose a production function. Suppose we have two inputs, K and L again. We would like our function to have all of the properties listed above like diminishing marginal products. Here are two that have these properties. Q = aLK + bL2 K + cLK 2 dL3 K eLK 3 26 (69)

Q = aLb K c C Regression

(70)

We now use the data to nd values for a, b, c, d, and e. No production function is exactly like the above two. What the regression tells us is what values of a-e make the above production function closest to the data (ie closest to the real world). Excel has a simple function linest which will be sucient. On the website is an example Excel le that demonstrates how to do a linear regression. Statistics can tell us many things besides the values a - e. We will focus on two other things the Excel le tells us. 1. T-stat. The T-stat tells us if K , L, K 2 , etc. likely have any eect on Q. If the T-stat for d is small, then we may want to drop the term L3 K from the production function. 2. R2 . This tells us the percentage of the variation in output that is explained by variation in the inputs. If the number is close to one, the production function is doing a good job describing the real production process. If the number is close to zero, there are some aspects of production being missed. Perhaps another input.

27