You are on page 1of 14

OUTPUT INVARIANCE ACROSS PRODUCTS

Richard Carson Economics Dept. Carleton University 1125 Colonel By Drive Ottawa, Ontario, Canada, K1S 5B6 rcarson@ccs.carleton.ca Tel.: (613) 520-2600, x4360 Fax.: (613) 520-3906

OUTPUT INVARIANCE ACROSS PRODUCTS

ABSTRACT

This note gives a method that can sometimes be used to find the long-run equilibrium output of a firm operating under imperfect competition. It also differs from standard methods of output determination and sheds light on the nature of equilibrium. Let X be a product supplied under oligopoly or monopolistic competition and suppose that, by changing certain properties of X, we arrive at a new product, Y, which is sold under perfect competition. For example, Y might be a generic version of a good or service, while X is a differentiated version. Under a basic assumption given in the paper, this firms equilibrium output of X will then be where the average cost of Y reaches its minimum. At the level of the firm, the equilibrium outputs of the two goods will be the same, in other words, as long as each product is viable. While the implied economywide equilibrium is not a first-best welfare maximum, it does have a desirable second-best property.

JEL Classifications: D24, D43, D61. Key words: Output invariance, imperfect competition, second best.

OUTPUT INVARIANCE ACROSS PRODUCTS

Introduction

This note gives a method that can sometimes be used to find the long-run equilibrium output of a firm operating under imperfect competition. It also differs from standard approaches to output determination and sheds light on the nature of equilibrium.1 Let X be a product supplied under oligopoly or monopolistic competition and suppose that, by changing certain properties of X, we arrive at a new product, Y, which is sold under perfect competition at a different price. For example, Y might be a generic version of a good or service, while X is a differentiated version. Under a basic assumption given below, this firms equilibrium output of X will be where the average cost of Y reaches its minimum, or where the production function for Y exhibits constant returns to scale. This is true as long as X is viable and the basic assumption is met. Thus the equilibrium outputs of X and Y are the same at the firm level. If we also assume that production of Y uses no inputs that are indivisible over the long run, either output is entirely determined by technical properties of the production function for Y. As a result, shifts in demand or cost will manifest themselves in entry and exit of firms and in price changes over the long run. Aside from entry and exit, however, equilibrium firm-level outputs will remain constant, as long as the basic assumption holds. While the implied economy-wide equilibrium is not a first best welfare maximum, it does have a desirable second-best property.

Basic Results

We shall use capital lettersX, Y, and Zto designate specific products and small letters to denote output quantities of these products. We wish to find the long-run equilibrium output of a profit-maximizing firm supplying X under oligopoly or monopolistic competition. To this end, let Px and Py be the prices of X and Y and x be the output of X by a firm supplying this good. X and Y are assumed to be measured in the same well-defined output units. Therefore, with x plotted on the horizontal axis, ACy will be the long-run average cost of Y that a producer of Y would face. It is assumed to take the conventional U shape with a unique minimum at x = x*. Let ACRx be the above firms rent-inclusive long-run average cost of X. This is its long-run average cost inclusive of all rents earned by inputs in the supply of X that persist over the long run. These include the rent to the suppliers market position, when this is protected by entry barriers, as well as earnings of product-specialized inputs over and above the opportunity costs of these inputs in supplying other goods. When all inputs are mobile between firms including alternative potential suppliers of Xthe latter rents are part of the opportunity cost of any firm supplying X. Thus when inputs are mobile between firms and market positions are freely transferable from one firm to another, rent-inclusive and opportunity costs are the same for any supplier. This is the assumption we shall make here. A firm will supply X over the long run if and only if it expects to be able to produce where Px ACRx. In long-run equilibrium, Py = ACy for any firm supplying Y, and Px = ACRx for any firm supplying X. The outcomes, Px > ACRx or Py > ACy, are possible only in the short run when sellers have the prospect of earning disequilibrium or entrepreneurial profits (quasi-rents) that attract entry and are competed away over the long run via expansion of supply by competitors.

Finally, although I do not explicitly model gaming between firms, I shall make one basic assumption about price-setting. To wit, starting from any equilibrium in which Px = ACRx, a firm is assumed to be able to reduce its output without also having to cut its price. Thus long-run equilibrium never occurs where rent-inclusive average cost is upward-sloping, since this would imply a lower output at which price would exceed this average cost (eg., at which Px > ACRx). As a result, Px = ACRx MCx must hold in long-run equilibrium, where MCx is the marginal cost of X, and similarly for every other product. Consider first the case in which the differentiated product is of higher quality than the generic version, with the result that Px > Py. Then (Px Py) (ACRx ACy) must hold in longrun equilibrium for the firm supplying X. Since Px = ACRx in equilibrium, (Px Py) < (ACRx ACy) would imply that Py > ACy is possible. In general, either (Px Py) > (ACRx ACy) or (Px Py) = (ACRx ACy) can prevail, equality holding if and only if x* = xe, the long-run equilibrium value of x. Figures 1(a) and 1(b) below illustrate these possibilities. In Figure 1(a), the equilibrium price and output of X are assumed to be at E1, where xe < x*, and (Px Py) > (ACRx ACy) therefore holds. In Figure 1(b), equilibrium price and output are at E2, where xe = x*, and thus (Px Py) = (ACRx ACy). For example, under monopolistic competition [1], E1 and E2 are two possible points of tangency between demand and ACRx. (Of course, an output at which tangency occurs is also an output at which marginal revenue equals marginal cost.) Whether E1 or E2 best describes equilibrium will depend on the kinds of competitive pressures that a supplier of X faces, and our basic assumption deals with these. One could say that a supplier of X adds value to Y internallyor within its production processby adding quality. However, a supplier could also add value to Y externally by producing a third product, Z, that is used with Y. For example, let Y be a generic bread with a bland taste and X be a

FIGURE 1
6

differentiated bread with a unique taste. Suppose an entrepreneur markets a spread, Z, that improves the taste of the generic bread, but clashes with the taste of the differentiated bread. More generally, the basic assumption says that at least one product, Z, is available in long-run equilibrium that meets two conditionsa substitutability condition and a cost-effectiveness condition. When such a Z exists, we shall say that these conditions hold or are satisfied for X. The substitutability condition requires Z and X to be substitutes in demand, under the assumption that income effects on the demand for any individual product are small enough to ignore. In addition, products that compete in the same industry with Z (the Z industry) are assumed to be substitutes for X, and an expansion of supply of these substitutes that reduces Pz, the price of Z, at any given output, z, is also assumed to lower Px at any given x. The cost-effectiveness condition requires a supplier of Z to be as cost effective as a supplier of X in adding value to Y. Let Z have rent-inclusive average cost, ACRz, defined as for X above. At any given x, the quality increase from Y raises price by (Px Py) and rent-inclusive average cost by (ACRx ACy). Thus to meet the cost-effectiveness condition, a supplier of Z must be able to produce an output scaled to have a price equal to (Px Py) at an average cost no greater than (ACRx ACy). More precisely, suppose that the X industry is in an equilibrium which may be only a partial equilibriumin which Px = ACRx. Then for any supplier of Z, the cost-effectiveness condition requires an output to exist at which: ACRz (ACRx ACy), (1).

when z is scaled in such a way that Pz = (Px Py). Here Px, ACRx, and ACy are evaluated at the equilibrium output of X, and Pz and ACRz are evaluated at the indicated output of Z. Inequality (1) is assumed to hold, no matter where the equilibrium value of x lies, although by what was said above, ACRx can not be upward-sloping at this equilibrium. We note

in passing that a weaker condition can be substituted for (1). In addition, (1) reflects not only the success of a supplier of Z in controlling costs, but equally, limits placed by competitive pressures on this suppliers ability to translate the resulting savings into long-run rent.2 A specific example of a function, ACRz, that meets the cost-effectiveness condition is given below. It should be mentioned, however, that (1) requires competition in the Z industry to affect Px as well as Pz. If equilibrium in the X industry were completely independent of outputs and prices in the Z industry, (1) could not hold at every possible equilibrium value of x, as will be explained shortly. When the cost-effectiveness condition is met and the Y industry is in equilibrium, a supplier of X must produce on a scale that is efficient for Y in order to cost compete with the composite good whose units consist of a unit of Z plus a unit of Y. The average cost of a unit of this composite equals (ACRz + Py). Thus suppose that the X industry is in an equilibrium in which Px = ACRx holds and that x is set where (Px Py) > (ACRx ACy), or, equivalently, ACy > Py. But then long-run general equilibrium can not prevail, since (1) implies that Pz > ACRz is possible for any product, Z, meeting the cost-effectiveness and substitutability conditions. Weak competition from the Z industry is what allows Px to be high enough to cover ACRx when x is set where ACy > Py. When Pz > ACRz, the prospect of quasi-rents will attract entry of new competitors for Z, which will generate downward pressure on both Pz and Px, without affecting ACRx or ACRz (since this entry would be foreseen). By the time Pz = ACRz is reached, increased competitive pressures on Px generated by entry of new competition into the Z industry will have pushed Px below ACRx. The supplier of X must change its output or exit the market. In short, when equilibrium prevails in the Z industry, and x is set where ACy > Py, Px < ACRx must hold as well. If the supplier of X survives over the long run, therefore, we must have:

Py = MCy = ACy

(2).

in general equilibrium, where MCy is the marginal cost of Y. Thus xe = x*, the output at which the production function for Y exhibits constant returns to scale. The location of xe depends only on technical properties of this function, and the equilibrium outputs of X and Y are the same at the firm level.3 It follows that xe will be invariant to shifts in demand and cost, as long as the supplier of X can cover its costs and at least one product is available that meets the substitutability and cost-effectiveness conditions. These shifts will affect Px and the rents earned from producing X, but not xe. Let Fx be a fixed cost in the form of rent to the firms market position and to its productspecialized inputs. We think of any differentiated product as possessing unique attributes, features, or properties whose cost-effective supply requires inputs specialized to these unique elements and thus to the product. For example suppose that a restaurant supplies cuisine with a unique flavor and ambience. Its product-specialized inputs are then its recipes, together with the tacit knowledge and the talentembodied in a master chef and his or her teamrequired to create the cuisine from the recipes at lowest possible cost. These inputs could be purchased by another owner or restaurant which would then acquire the unique cuisine in question. Thus the rent earned by these product-specialized inputs is part of the opportunity cost of the restaurant supplying the cuisine that they produce. Moreover, these inputs are wholly or partly indivisible, and all or part of their remuneration is therefore a fixed cost which does not vary with output. With this in mind, let ACx = ACRx Fx/x be average cost exclusive of these rents, and Lex = (Px MCx)/Px be the Lerner index of market power for X. If we measure returns to scale using ACx rather than ACRx, we remove the effect of Fx on economies of scale. When the supplier of X faces fixed input prices and uses no indivisible inputs other than those which are specialized to

X, measuring returns to scale from ACx is the same as measuring them from the production function for X. Suppose, moreover, that differentiating the firms product (replacing Y with X) does not increase returns to scale at any output when these returns are measured using ACy and ACx. Then ACx can not be downward-sloping at x = x*, and Lex Fx/Pxx* must hold. When freedom of entry and exit leads to competitive pressures on X that are strongin the sense of keeping the share of product-specialized rent in product value lowthese same pressures will ensure approximate marginal-cost pricing. In this case the supplier of X may well produce where increasing returns to scale prevail, as measured by ACRx, but must also produce where constant or decreasing returns prevail, as measured by ACRx. If the substitutability condition did not hold for X, and the X and Z industries reached their respective equilibria independently, the cost-effectiveness condition could not hold either as a general rule. To see this, start again at a partial equilibrium for the X industry at which Px = ACRx and (Px Py) > (ACRx ACy) both hold. Inequality (1) again implies that Pz > ACRz is possible, and entry into the Z industry will occur which pushes down Pz. Without the substitutability condition, however, this will have no effect on Px or on equilibrium in the X industry. Once equilibrium in the Z industry is reached, ACRz > (ACRx ACy) must hold, after we have re-scaled output units of Z so that Pz again equals (Px Py) following the fall in Pz. Thus when no product meets both the substitutability and cost-effectiveness conditions, xe may be where (Px Py) > (ACRx ACy). But then there is a profit incentive to discover and market new substitutes for X that satisfy (1) or new technologies that lower costs and enable existing substitutes to satisfy (1). These products will be viable as long as X remains viable, and their appearance will give rise to positive quasi-rents in the short run, as long as (Px Py) >

10

(ACRx ACy) prevails, which implies Pz > ACRz as we have seen. Moreover, X may well remain viable if the rental component of ACRx is initially positive and flexible downward. In general, Z can be supplied under either perfect or imperfect competition, and in the latter case, a symmetrical relation potentially holds between X and Z. Suppose that Z is a differentiated product and that V is a generic version of Z. Let z* be the output of Z at which ACv reaches its minimum. In the bread example, V could be a generic spread, which complements the differentiated bread (X). By interchanging the roles of X and Z in the analysis above, we get ze = z*. Finally, if Px < Py in long-run equilibrium, the roles of X and Y are reversed. The combination of X and Z competes with Y, and the cost-effectiveness condition becomes ACRz (ACy ACRx) when z is scaled in such a way that Pz = (Py Px). Arguing as above, it is straightforward to show that the long-run equilibrium output, ye, of a supplier of Y is where Px = ACRx, which is to say that ye = xe. But since ye = x* always holds, we again have xe = x*.

Example

To give a simple example of how the entry of Z can change equilibrium, suppose that at first no such product as Z is present. Let ACy = .01x + 99 + 100/x and ACRx = ACy + 50 + Fx/x. This yields x* = 100 and Py = 101. Also MCy = 99 + .02x, and MCx = 50 + MCy. Suppose further that there are n firms in the X industrywith each firm supplying a differentiated productthat these firms reach a Nash equilibrium and share the market equally, and that total market demand in equilibrium is 1000 units. Finally, let Fx = 1000, and suppose that the demand for X is linear, with slope equal to 8.6/n. (A larger number of competitors will tend to make

11

this slope flatter.) With these data, we can solve to get n = 20, xe = 50 and Px = ACRx = 171.5. Here (Px Py) = 70.5 > (ACRx ACy) = 50 + 1000/50 = 70. Thus xe < x*. In addition, the Lerner index of market power equals .177, greater than Fx/Pxx* = .117 because ACx = .01x + 149 + 100/x is downward-sloping in equilibrium, and greater than MCx. When no product such as Z is present, this is quite possible. Even if Fx were zero at this value of xe, Lex would still be positive. Because (Px Py) > (ACRx ACy) in the equilibrium above, entry of products with the properties of Z can be profitable. Here their appearance is likely to cause some of the original 20 firms to exit, as those able to survive reduce price and expand output to stay competitive. Suppose the entry of Z and of products competing with Z halves Fx to 500, but does not affect ACy. Then ACRz could take the form, .005z + 49 + 550/z, which gives MCz = 49 + .01z. Moreover, suppose demand is such that Pz = Px Py = Px 101 whenever z = x. Then ze = xe and ACRz is nearly equal to (ACRx ACy) as long as ACRx is downward-sloping (or for all x 331).4 From earlier discussion, the new equilibrium following entry of Z will be where Pz = Px 101 = ACRz = 50 + Fx/xe and Px = ACRx = .01xe + 149 + 100/xe + Fx/xe. This gives 101 = .01xe + 99 + 100/xe. Thus xe = 100 = x*, and Px = 151 + Fx/100 = 156. When z = x, it is straightforward to check that ACRz + Py < ACRx at each x except x = 100, where the two are equal. When Fx > 0, Px > MCx at x = x*, and Lex = Fx/Pxx*. In this example, differentiating the productor replacing Y with Xdoes not affect returns to scale at any given output when the latter are measured by ACy and ACx. The entry of Z reduces Lex from .177 to .03 (with Lez = .09), while MRx (marginal revenue) = Px(1 Lex) rises from 171.5(1 .177) = 141.14 to 151. If V is a generic version of Z, and there is a symmetrical relation between X and Z, as indicated above, ACv could take the form, .005v + 25 + 50/v.

12

Conclusion

If X, Y, and Z are as described above and the substitutability and cost-effectiveness conditions hold for each supplier of X, the long-run equilibrium output of X at the firm level will be where the average cost of Y reaches its minimum and therefore the same as that of Y. As such, it depends only on technical properties of the production function for Y. Thus, if these conditions continue to hold, xe will remain constant and equal to x* in the face of shifts in Xs cost or demand, as long as X remains viable. From the standpoint of welfare, the implied general equilibrium is not a first best, but it does have a desirable welfare property. To wit, suppose that differentiating a firms product does not increase returns to scale, when these returns are measured by rent-exclusive average cost for the differentiated product. Then if freedom of entry and exit leads to competitive pressures on X that are strongin the sense of keeping the share of product-specialized rent in product value lowthese same pressures will ensure approximate marginal-cost pricing.

NOTES

1. For surveys of models of imperfect competition, see [2] and [3] and see [1] for the basic model of monopolistic competition. 2. We can replace (1) with the inequality: ACRz (ACRx ACy) + m(x)(ACy Py), (1n).

where m(x) is any function of x such that 0 < m(x) < 1 for every x. Inequality (1n) allows a supplier of Z to have a cost disadvantage equal to a share (which can be nearly all) of the

13

difference between the values of ACy at x and x*. When (1n) holds, Px = ACRx and (Px Py) > (ACRx ACy) again imply that Pz > ACRz. Equation (2) follows as in the text. 3. Equations (2) also follow from (1) when Y is supplied under imperfect competition. The output of any product, X, obtained by changing the properties of Y, will be where Py = ACRywith ACRy defined like ACRx aboveas long as there is a third product, Z, that meets the substitutability and cost-effectiveness conditions. However, this equilibrium output will now depend on prices and other economic variables, as well as on technical properties of the production function for Y. 4. In fact, ACRz is slightly greater than (ACRx ACy), but meets the weaker condition in note 2.

REFERENCES

1. Chamberlin, E. The Theory of Monopolistic Competition. Cambridge, MA: Harvard University Press, 1933. 2. Nicholson, W. Microeconomic Theory. Fort Worth, TX: The Dryden Press, 1998, 7th ed., Chs. 18-20, pp. 543-632. 3. Shapiro, C. Theories of Oligopoly Behavior. Ch. 6 of Schmalensee, R. and Willig, R.D. Handbook of Industrial Organization. Amsterdam: North-Holland, 1989, vol. I, pp. 329-414.

14

You might also like