COST THEORY. I What costs matter? A Opportunity Costs

COST THEORY Cost theory is related to production theory, they are often used together. However, the question is how much to produce, as opposed to which inputs to use. That is, assume that we use production theory to choose the optimal ratio of inputs (e.g. 2 fewer engineers than technicians), how much should we produce in order to minimize costs and/or maximize profits? We can also learn a lot about what kinds of costs matter for decisions made by managers, and what kinds of costs do not. I What costs matter? A Opportunity Costs Remember from the Introduction, Section (IV) of the Introduction, that in addition to accounting profit, managers must consider the cost of inputs supplied by the owners (owners capital and labor). Definition 16 Accounting Costs: Costs that would appear as costs in an accounting statement. Definition 17 Opportunity Costs: Accounting costs plus the value of using inputs in their next best alternative. Recall the example from Section (IV), we include opportunity costs such as the cost of time and capital that could be spent on another project. Throughout, costs are always opportunity costs. B Fixed costs, variable costs, and sunk costs First, costs may be fixed or variable: Definition 18 Total Variable Cost The total cost of all inputs that change with the amount produced (all variable inputs). Definition 19 Total fixed costs The total cost of all inputs that do not vary with the amount produced (all fixed inputs). 28

Some inputs vary with the amount produced and others do not. The firm s computer system and accountants may be able to handle a large volume of sales without increasing the number of computers or accountants, for example. Inputs that do not vary with the amount produced, like accountants and computers, are called fixed inputs. Most inputs are fixed only for a certain range of production. A medical office may be able to handle many additional patients without adding and office assistant or extra phones. The phones and office assistants are fixed inputs. But, if the number of extra patients is large enough, the firm needs extra office staff. Reasons for fixed costs: 1. Salaried workers. Salaried workers are a fixed input if the worker can work overtime without additional compensation (doctors paid a fee for service are variable inputs, salaried medical staff are fixed inputs). 2. Fixed hours at work. An hourly worker sometimes cannot be sent home early if not enough work is available. Therefore, workers may not be busy and be able to handle extra work without additional hours. 3. Time to adjust. Some inputs, like machines, take time to purchase and install. Conversely, unskilled labor may be adjusted more quickly through overtime, temps, etc. Therefore, by necessity the firm may only be able to vary production by increasing labor in the short run. Consider the Thompson machine company. The firm uses 5 machines to make machine parts. Because of the time to adjust, machines are a fixed cost, while the number of workers varies with the amount produced. Labor is a variable cost. Definition 20 Sunk costs Are costs that have been incurred and cannot be reversed. Anyirreversiblecostincurredinthepast, orindeedanyfixedcostforwhichpaymentmust be made regardless of any decision is irrelevant for any managerial decision. Suppose you hire an executive with a $100,000 signing bonus, plus $200,000 salary. After hiring, you may find the executive does not live up to expectations. However, if the executive s marginal revenue product is $200,001, the executive still generates $1 in profits relative to his salary and therefore should be retained. But if his MRP is $199,999, the firm loses an extra $1 each year they keep him, so he should be let go. The bonus is a sunk cost and does not affect the retention decision. 29

The principle of sunk costs is equivalent to the saying don t throw good money after bad. Sometimes a decision can be made to recover part of a fixed cost. Perhaps one could sell a factory and recover part of the fixed costs. Then only the difference is sunk. For example, if we can sell a building for which we paid $500,000 for $300,000, then only $200,000 is sunk. So fixed costs may or may not be sunk, but sunk costs are always fixed. Sunk costs are psychologically difficult things to ignore. Examples: 1. Finance. Studies show investors let sunk costs enter their decision making. What price the stock was purchased at is sunk and therefore irrelevant. What matters is only whether or not this stock offers the best return for the risk. Yet, investors are reluctant to sell stocks whose price has gone down. 2. Sunk Costs of Investment Consider fad toys. Most often such products have a short lifespan. The temptation will be try to and hang on, even after the business is no longer profitable, because of all the time and capital sunk into the business (sunk costs). Silly Bandz is a great example. 3. Pricing in high rent districts. Consider restaurants in a high rent district (say an airport). Should they take the rent into account when setting prices? No. In fact, prices are high not because of the rent, but typically because of the lack of competitors. II Short run costs We use short run costs primarily to compute how much to produce while maximizing profits. We use long run costs to answer questions like should the firm expand, contract, merge, etc. Definition 21 Average Costs: Costs divided by output. Definition 22 Marginal Costs: The cost of one additional unit of an input. Here is the notation: Type of Cost Total Cost equals Variable Costs Plus Fixed Costs Total TC = TVC +TFC Average ATC = TC = Q TVC AVC = Q Marginal MC = dtc dq Q Properties of cost functions in the short run: +AFC = TFC Q 30

1. Total costs of increase with Q, the quantity produced. 2. Average Total costs decline with Q, but eventually rise. The fixed costs are spread over many more units of production at high Q, reducing average costs. All of the extra workers required for producing additional units when the factory is near capacity starts to increase average costs eventually. 3. Marginal costs usually decline then increase, but must eventually increase. At first, producing one additional unit is may cheaper than the last unit, due to specialization. However, eventually diminishing returns sets in and the workers just get in each other s way. Then a very large number of additional workers might be needed to produce an additional unit. Here is a graph of the cost curves. 500 Total Cost Functions 450 Total Costs Total Fixed Costs Total Variable Costs 400 350 300 Cost ($) 250 200 150 100 50 0 0 1 2 3 4 5 6 7 8 9 10 Q Figure 5: Typical short run total cost curves. 31

120 Short Run Cost Functions 100 Average Total Costs Average Fixed Costs Average Variable Costs Marginal Costs 80 Costs per unit ($) 60 40 20 0 0 1 2 3 4 5 6 7 8 9 10 Q Figure 6: Typical short run average and marginal cost curves. III Examples using short run cost curves. A Profit maximization with perfect competition Let us suppose that you are a hypothetical manager of a group of cruise ships. Using data from previous years, you estimate the short run cost function is (we will see how to do this estimation below): TC = 60+ Q2 20 (73) Here Q is the number of cruises the ship takes (not the number of passengers). The cost of the ship is $60 million, which is sunk. Notice in equation (73) that the cost of the ship does not depend on the number of cruises the ship takes. Suppose further that each cruise generates revenue of $3 million. Maximize profits: maxπ = TR TC = 3Q 60 Q2 20 (74) 32

Take the derivative to get the slope and set the slope equal to zero: 3 Q 10 = 0 Q = 30. (75) Notice that the sunk costs have dropped out. The math agrees: sunk costs do not matter for our decision. In general, to maximize profits we set marginal revenue (here $3) equal to marginal costs (here Q/10). MR = MC. (76) In a competitive industry, firms have no ability to influence the price. Examples include market makers in stock markets, commodities, and large food markets. For example, an individual farmer may produce as much corn as desired and sell the corn on the commodities markets with no change in price. Regardless of the amount produced by an individual firm, the price is unchanged and MR = P. So in competitive industries, we set: P = MC. (77) The firm should produce one more unit if we can sell it for more than it costs to produce. The firm makes profits in this case. In the next set of notes, we will do the case where firms have pricing power. Using a table: Cruises (Q) Total variable costs (T V C) Marginal costs (MC) 0 0 20 = (20 2 )/20 = 20 = (20 0)/(20 0) = 1 25 = (25 2 )/20 = 31.25 = (31.25 20)/(25 20) = 2.25 30 45 2.75 35 61.25 3.25 40 80 3.75 Table 6: Variable costs on a cruise ship. The marginal revenue is the price of the cruise, equal to $3 million. We can see that marginal revenue equals marginal costs somewhere between 30 and 35 cruises. 1 Producing the 30th cruise gives us $3 of revenue, enough to cover the costs of producing the 30th cruise, which (using table 6) is approximately $2.75. However, the 35th cruise loses 1 The table is an approximation, however. In fact, using the true marginal cost of the 30th cruise is MC = dtc/dq = Q/10 = 3. 33

money. The table estimates that cruise would cost $3.25, and since revenues are $3, the cruise would lose an estimated $0.25 million. The sunk costs are irrelevant here. When considering the sunk costs, the firm has negative profits regardless of how many cruises the firm takes. The maximum profits occurs at 30 cruises: π = TR TC = 3 30 ) (60+ 302 = $15. (78) 20 We have already paid the sunk costs, so we might as well lose as little as possible. 2 B Break Even Analysis An important consideration when deciding whether to continue operations in a particular market, expand into a market, or start a new business is a break even analysis. We can do a break even analysis with a cost function. In a break even analysis, the question is: how much profit is required to exactly pay off all fixed costs? Alternatively, how much revenue is required to pay off the average variable costs and the fixed costs: π = 0 = TR TC (79) 0 = P Q TFC TVC (80) 0 = P Q TFC AVC Q (81) Q = TFC P AVC (82) Here I have assumed linear total costs, so that average variable cost is constant. One could assume (more realistically) that average variable costs depend on Q, and use a table to get the break even point. 2 Note that it is irrelevant how the ship is financed. Interest and principle payments are also sunk costs. 34

C Average Costs Often average cost data is easy to get. It is relatively easy to measure total costs and the quantity sold to get average costs. However, many managers then incorrectly base decisions on average costs. Consider the following data: Typical Data Calculate this! Gas Production Total costs Average Total Marginal Costs Profits (π) (Q) (TC) costs (ATC) (MC) 20 1271 = 1271/20 = 63.6 89 22 1359 61.8 = (1359 1271)/ 137 (22 20) = 44.0 24 1456 60.7 48.5 176 26 1562 60.1 53.0 206 28 1675 59.8 56.5 229 30 1797 59.9 61.0 243 32 1928 60.3 65.5 248 34 2067 60.8 69.5 245 36 2214 61.5 73.5 234 38 2370 62.4 78.0 214 40 2534 63.4 82.0 186 42 2707 64.5 86.5 149 44 2888 65.6 90.5 104 46 3077 66.9 94.5 51 48 3275 68.2 99.0-11 Table 7: Average and marginal costs in the gas industry. The first 3 columns of table 7 give a typical data set one might have in a real business situation. For example, a gas refinery might produce different quantities monthly depending on variations in the price of gas and demand. Suppose the price of gas is currently $68. How much gas should be produced? Setting price equal to average cost (48 units) actually produces a loss. We could try to minimize average costs, which occurs at 28 units. Still, this is not the maximum profits. The maximum profits occurs when price equals marginal costs, about 32-34 units. Instead, given the first 3 columns, one can easily calculate the marginal costs and calculate the profit maximizing quantity (between 32 and 34 units). 35

IV Long Run Costs We use long run costs to decide scale issues, for example mergers, but also the optimal size of our operations (e.g. optimal plant size). The long run is when all costs are variable. In the long run, we can build any size factory we wish, based on anticipated demand, profits, and other considerations. Once the plant is built, we move back to the short run as described above. Therefore, it is important to accurately forecast demand. Too small a factory and marginal costs will be high as the factory is stretched to over produce. Conversely too large a factory results in large fixed costs (eg. air conditioning, or taxes). A Definition Definition 23 Long Run Average Costs: The minimum cost per unit of producing a given output level when all costs are variable. Average cost ($/unit) Short Run Average Costs small plant medium size plant Long Run Average Costs Q d Cost of making Q d with a small plant Q Cost of making Q d with a medium plant Figure 7: Long run average cost curve. 36

Think of each short run average cost curve as a factory of different size. We build the factory, and then operate within the short run average cost curve for the factory that is built. For each size factory, average total costs are initially decreasing as the fixed costs are spread over more and more units. Eventually, however, average total costs rise as the factory becomes over crowded. Suppose we forecast demand at Q d on the graph. Then, from the graph, we create a small sized factory, as that is the cheapest way to build Q d units. Therefore, we must have that the long run average cost curve is tangent to the minimum of the short run average cost curves. B Deriving a long run average cost curve: Example Suppose we can build plants of three sizes: Cost Type Plant Size AV C Small Medium Large Labor $3.70 $2.50 $1.10 Materials $1.80 $1.40 $0.90 Other $2.00 $2.60 $3.00 T F C $25,000 $75,000 $300,000 Capacity 50,000 100,000 200,000 Table 8: Plant size and LRAC. What type of plant produces 100,000 units at the lowest long run average costs? We can build either two small plants, 1 medium sized plant, or one large plant. The average costs of two small plants are: ATC = AVC + TFC Q, (83) ATC = $3.70+$1.80+$2.00+ $25,000 2 100,000 = $8. (84) Here, the fixed costs of the small plant is multiplied by two since we need two plants. For a single medium sized plant: ATC = $2.50+$1.40+$2.60+ $75,000 = $7.25. (85) 100,000 The medium sized plant is cheaper. Although the factory is more expensive to build than 37

two small factories, it can produce units at lower average variable cost. A large plant has: ATC = $1.10+$0.90+$3.00+ $300,000 100,000 = $8.00. (86) The large plant can produce at a very low average variable cost of $5 per unit, but is just too expensive to build. So for Q = 100,000 units, we build a single, medium sized factory. The point on the long run average cost curve for Q = 100,000 is LRAC = $7.25. Suppose instead expected demand is Q = 200,000 units. We can build 1 large, two medium, or 4 small plants. The cost of these options (per unit) are: $6.50, $7.25, and $8, respectively. The large plant is the cheapest, and LRAC = $6.50 for Q = 200,000. We could also build 1 medium and two small factories, but then the costs would be an average of $7.25 and $8, which is still worse than a large plant. Finally long run average costs are minimized when we don t have to build any small or medium sized plants, at Q = 200,000, 400,000, and so on. C Factors that affect long run average costs Long run average costs may be decreasing and then increasing, but also may be strictly decreasing. Here are some LRAC curves for some industries. 1. Nursing Homes have decreasing LRAC. Nursing homes have many fixed management costs. Further, larger nursing homes are able to negotiate lower prices for many materials inputs. 2. Cruise Ships have decreasing LRAC. Huge cruise ships have lower average costs than small cruise ships, economizing on many services provided on the ships. 3. Hospitals have U-shaped LRAC. Cowing and Holtmann (1983) in fact found many hospitals in New York should in the long run reduce both capital and physicians to lower average costs. When the LRAC curve is decreasing, it is in the interest of the industry to consolidate. A merger with another firm can increase the customer base but reduce the cost per unit, thus increasing profits. Reasons for decreasing long run average costs: 38

1. Specialization of labor. Producing more units requires more workers. Workers can then specialize in different parts of the production process, and produce more units. 2. Indivisibilities. A firm may require one accountant, even if there is less than 40 hours of work to do. But then the firm can increase units produced without increasing the size of the accounting staff. 3. Pricing Power. Large firms buy in bulk and therefore negotiate lower prices for inputs. Notice the reasons are similar to the reasons for increasing returns to scale. Reasons for increasing costs: 1. Regulation may exempt smaller firms. 2. Coordination and information problems. In a large firm, many individuals do not meaningfully affect profits and thus have the wrong incentives. Smaller operations may know their customers and production processes better. Similar to decreasing returns, when long run costs are increasing, spin-offs and divestments may be optimal. A compromise is franchising. Nationalize just the parts for which increasing returns works. D Long run competitive equilibrium In the long run, firms can enter and exit the market, or grow or shrink in size. As long as economic profits are positive, new entrants arrive as the industry gives higher profits than alternatives. New entrants increases competition and pushes down the price until economic profits are zero. Therefore, in the long run: π = TR TC = 0, (87) P Q = TC, (88) P = TC Q = LRAC. (89) 39

Now at the firm level, we set P = LRMC. But since also P = LRAC, marginal and average costs are equal in the long run. We can therefore minimize long run average costs and in the long run, we will have P = LRMC = LRAC. V Application of long run average costs: Banking mergers Consider two banks. The first bank services Q = 15 customers and the second (smaller) bank services Q = 5 customers. The long run average cost function in the industry is: LRAC = 700 40Q+Q 2 (90) Price is constant at $300 dollars of revenue per customer. Should these two firms merge? The size of the customer base (Q) which minimizes long run average costs is: minlrac = 700 40Q+Q 2 (91) 40+2Q = 0 Q = 20 (92) Costs per unit fall until Q = 20. Thus these two firms can reduce costs by merging from two firms of size Q = 15 and Q = 5) into one firm with Q = 20. Profit per unit in each case are: π = 300 ( 700 40Q+Q 2) = 400+40Q Q 2 (93) π(q = 15) = 400+40 15 15 2 = 25 (94) π(q = 5) = 400+40 5+5 2 = 225 (95) π(q = 20) = 400+40 20 20 2 = 0 (96) Individually, the two banks lose money but together they have zero economic profits. At zero economic profits, no incentive exists for firms to enter or exit the market (remember, π here is economic profits so zero here means that investing in this industry gives the same profits as investing in the next best industry). The choice is essentially to merge and get 40

bigger or exit the market. VI Measuring Cost Functions We use the same procedure as with production functions: Obtain data on total costs and quantity produced, and use Excel to fit the data. Both total cost and total quantity produced may appear to be easier to obtain than input data. However, one must remember that costs represent opportunity costs, which are not always straightforward. Some additional issues: A Choice of Cost Function One choice is whether to use a linear, quadratic, or cubic function: TC = a+bq (97) TC = a+bq+cq 2 (98) TC = a+bq+cq 2 +dq 3 (99) Under most circumstances, the linear cost function does a reasonable job over a narrow range of Q (for example in the short run), but the quadratic and cubic terms must matter theoretically, especially for a wider range of Q. A good strategy might therefore be to estimate the cubic or quadratic. If the t-stats are low for the quadratic and cubic terms, then predictions are likely to be unreliable for Q that falls outside the data. This indicates using some caution before, for example, committing to large mergers. The following graph illustrates the problem. B Data issues Some problems with the data that often need correcting: 1. Definition of cost: as mentioned earlier, we use opportunity costs not accounting costs. 2. Price level changes: Historical data is likely to be inaccurate if the price of some inputs or outputs have changed dramatically. 41

64 Possible Problem Estimating Cost Functions: Data is to homogeneous 63.5 data True Cost Function Estimated Cost Function 63 Cost ($) 62.5 62 Estimate is inaccurate here 61.5 Estimate is accurate here 61 60.5 1 2 3 4 5 6 7 Q Figure 8: Cost estimation problems. 3. What costs vary with output: Some costs have a very limited relationship with output. For example, the number of professionals required may vary in some limited way with output. A firm with $1 million in sales may have two accountants. The firm can obviously increase output to some degree without needing more accountants (so the cost would be fixed). But for larger Q additional accountants are needed(like a variable cost). 4. The cost data needs to match the output data. Often the cost of producing some output may be accounted for in some other period. 5. The firm s technology may change over time. When estimating long run costs, it is usually preferable to use a cross section of firms across an industry. An individual firm is unlikely to have changed size significantly enough to generate data for a wide range of Q. 42