Abstract
We study a widely used ordering process (“Early Bird Discounts”) whereby a profitmaximizing manufacturer permits his dealers to place advance orders at a discount before they set retail prices. We show that such discounts may be used to shift just enough channel profits to dealers to enable them to cover their fixed costs and stay in business. If the manufacturer instead simply cut his wholesale price in order to generate gross margins for his dealers, these margins would soon dissipate as price competition among dealers selling the same product forced retail prices back down to the perunit cost. We show that when dealer fixed costs are low, the manufacturer offers an Early Bird Discount to his multiple dealers that induces all but two of them to exit; when fixed costs are high, the manufacturer offers no preorder discount (i.e. switches to linear pricing) and induces all but one dealer to exit. Although uniform slotting allowances could also be used to reward dealers, a salesbased alternative like an Early Bird Discount sometimes has a key advantage when the manufacturer has dealers in cities of different sizes. If the same Early Bird Discount is offered, dealers in markets with more consumers, who typically have larger fixed costs, will preorder larger amounts and will automatically receive higher gross margins. To duplicate such payments with slotting allowances, nonuniform allowances would have to be offered to firms in different markets, which is divisive and possibly illegal.
Similar content being viewed by others
Notes
Slotting allowances are lumpsum transfers or negative access fees in a twopart pricing scheme between a manufacturer and a retailer (e.g., Shaffer 1991; Kuksov and Pazgal 2007). When a manufacturer offers the same amount of slotting allowance to dealers, we call it a “uniform slotting allowance.” This is to distinguish it from the notion of “nonuniform slotting allowances” defined in the text.
If accelerating production would raise perunit costs, then it is less costly for the manufacturer to have at least some production occur prior to the realization of uncertain demand (e.g., Eichenbaum 1989). Stockouts are costly to dealers due to the loss of both immediate and longterm sales (Anderson et al. 2006). Preordering and inventory commitment may eliminate stockouts, avoiding situations where consumers would switch to competing products (McCardle et al. 2004). None of these selfinsurance behaviors (Ehrlich and Becker 1972) involves strategic considerations on the manufacturer’s part.
At the owner’s request, we have replaced the name of his firm with the pseudonym “Computec.” Computec’s executives said that they intentionally installed the threestage sales process to alter the strategic interaction downstream and to increase dealer margins. Further evidence that production smoothing is not the company’s motivation is that the company allows its dealers to take deliveries immediately after the last opportunity to preorder (see Appendix A).
As is standard in their businesses, Computec, Kaspersky, Toshiba, Hitachi, and Carrier—at least in some of their international markets—prohibit dealer returns of preorders unless the products are defective. For highereducation book distributors/publishers such as NACSCORP and Kendall Hunt, returns policies for bookstores vary but can be restrictive, and some merchandise is simply nonreturnable. For fruit farms, according to those we interviewed, it is logistically difficult to return fresh fruits such as melons, and thus returns are rarely accepted by growers.
Of course, when accelerating production raises perunit costs, both cost and strategic considerations may motivate the manufacturer to offer preorder discounts. Hence, in some of the examples we first cited, we cannot rule out the possibility that the manufacturers were motivated to some extent by the desire to avoid higher perunit costs or dealer stockouts.
We define gross margin as the difference between the sales revenue earned by a dealer and his total variable cost of acquiring the merchandise. Gross margin does not include the dealer’s fixed costs.
This may seem paradoxical since the profits of a firm who has more customers than he can serve will be the same no matter which customers he turns away. However, his rival’s strategy may no longer be profitmaximizing depending on which customers the first firm turns away. In that case, what was an equilibrium under one rationing rule ceases to be an equilibrium under the other rationing rule.
The same issue of observability arises in, for instance, Kreps and Scheinkman (1983), Davidson and Deneckere (1986), Maggi (1996), Padmanabhan and Png (1997, 2004), and Wang (2004). Although the role of observability of firm capacities or preorders is not always explicitly discussed in these models, some mechanism through which each firm or dealer learns the capacities or preorders must be implicitly assumed. This is because, as Tirole emphasizes on capacityconstrained games in his textbook (Tirole 1988, p. 217), to influence subsequent behavior, prior actions must be observable. Dixit (1982) and Shapiro (1989) make the same point.
We are assuming that distribution is not the manufacturer’s forte. That is, even if channel profits were maximized, the revenue net of production costs would be insufficient to cover the additional costs the manufacturer would incur if he tried to distribute the product himself. Thus, he is dependent on one or more dealers for distribution (Coughlan et al. 2006).
In other words, we assume that dealers preorder and distribute the manufacturer’s product if and only if they expect their profit (the difference between gross margins and fixed costs) to be weakly positive. Symmetrically, we also assume the manufacturer remains in business if and only if he earns a weakly positive profit.
If customers could instead not observe the quantity preordered by each dealer, demand would more plausibly be divided equally across the lowestpriced dealers, so that each would have demand, D(p)/R.
The other known alternative in capacityconstrained oligopoly games such as ours is to assume one of several ad hoc rationing rules as in Kreps and Scheinkman (1983) and the subsequent literature (see Tirole 1988, p. 212–3). Our results would then depend on which arbitrary rationing rule we adopted (Davidson and Deneckere 1986, p. 404). We instead follow Maggi (1996) in incorporating into the “rules of the game” the assumption that dealers must augment to satisfy unmet demand. As a shorthand, we refer to the KrepsSchenkman assumption as the “no rainchecks” assumption and to Maggi’s alternative as the “rainchecks” assumption. We make the latter assumption because (i) augmentation is a key feature in many of the realworld examples cited in the introduction, and (ii) requiring it also avoids equilibrium predictions that, as DavidsonDeneckere showed, depend on the arbitrary rationing rule that must be specified under the “no rainchecks” approach. The equilibrium may change because an equilibrium strategy profile under one rationing rule may not form an equilibrium under a new rationing rule since under the new rationing rule a rival firm inherits a different set of discarded customers and may have a profitable deviation where none existed under the old rule. The third option – allowing dealers to choose whether or not to issue rainchecks – requires a dealer to compare the payoff from each alternative and therefore again involves the adoption of an arbitrary rationing rule in the “no rainchecks” subgames.
Nevertheless, it is logically possible that some dealer would strictly prefer to turn away a customer unilaterally rather than offer him a raincheck if the rules of the game permitted such behavior. In our equilibrium, every dealer preorders enough to serve every customer (see Proposition 1). If a dealer nonetheless turned away a customer unilaterally under these circumstances, the dealer would be strictly worse off: he would forego the retail price the customer would pay but could not get any refund from the manufacturer on the preordered merchandise since the manufacturer has a “no returns” policy. We thank the associate editor and an anonymous referee for encouraging us to clarify these issues.
As shown in Proposition 1, when retail prices are below the level that maximizes channel profits, those prices are Cournot prices. Therefore, they decrease in the number of dealers.
Suppose the unique solution to the manufacturer’s profitmaximization problem formalized above occurs on the Rdealer CB boundary. There is then also a continuum of optima that can be achieved by taking the same c but raising θ, since changing θ has no effect when the dealers are behaving like “Cournot competitors.”
Dealer gross margins in the Cournot region equal the profits of Cournot competitors with marginal cost c. As discussed in Kotchen and Salant (2010), Cournot profits increase as c decreases given our assumption that inverse demand is weakly concave.
Early Bird Discounts degenerate into linear pricing when θ = c. If a single dealer can preorder at c per unit and augment at the perunit cost θ = c, we envision him as preordering what is profitmaximizing for a monopolist dealer to sell if product can be acquired at marginal cost c. The dealer would make a strict loss by preordering more. If he preordered less, he would earn an unchanged profit since it would then be profitmaximizing to acquire the remainder by augmenting at θ = c. If F = 0 the equilibrium payoff of the manufacturer and his single dealer is the same as in Spengler (1950).
The last equality follows since θ^{*} = (a + m) / 2 ⇒ b(a – θ^{*}) / b = (θ^{*} – m) ⇒ D(θ^{*}) / –D'(θ^{*}) = (θ^{*}– m) ⇒ (D(θ^{*}))^{2} / –D'(θ^{*}) = (θ^{*}– m)D(θ^{*}) = П_{max}.
In the smaller city, the Cournot price solves the equation \( \mathrm{p}+\frac{\widehat{\mathrm{D}}\left(\mathrm{p}\right)}{2\widehat{\mathrm{D}}^{\prime}\left(\mathrm{p}\right)}=\mathrm{c} \). Since \( \widehat{\mathrm{D}}\left(\mathrm{p}\right) \) = χD(p), this equation can be rewritten as \( \mathrm{p}+\frac{\upchi \mathrm{D}\left(\mathrm{p}\right)}{2\upchi \mathrm{D}^{\prime}\left(\mathrm{p}\right)}=\mathrm{c} \). This simplifies to precisely the equation determining the Cournot price in the larger city. Hence the two Cournot prices are the same.
Notice that if demand were the same in the two cities, contrary to our assumption, then χ = 1 and the derivation shows that Early Bird Discounts and slotting allowances are equally profitable. This confirms our conclusion in the onecity case.
See Coughlan et al. (2006, pp. 83–89) for a detailed discussion on the process and the high cost of handling and restocking returned merchandise in marketing channels.
To make this as clear as possible, we have adopted Maggi’s notation with two minor exceptions. First, we denote the preorder cost as c, which is equivalent to his original cost of building capability, denoted as c_{0} in his paper. Second, Maggi has another variable c as firms’ selling cost, which we set as zero. Of course, the proof of our proposition is entirely different, because we are dealing with an arbitrary number of players, a wider class of inverse demand curves, and perfect substitutes.
At Cournot equilibrium, no oligopolist would strictly benefit from increasing or decreasing his production. So p^{Cournot}(c;R) + q^{Cournot}(c;R) · P'(R · q^{Cournot}(c;R)) – c = 0. Since θ < p^{Cournot}(c;R), D(θ)/R > q^{Cournot}(c;R), and P'(·) < 0 is strictly decreasing, it follows that θ + (D(θ)/R) · P'(D(θ)) – c < 0. This last expression is the marginal change in dealer i’s gross margin if he unilaterally increased his preorder in the neighborhood of his proposed equilibrium strategy. Given strict concavity of his grossmargin function, nonlocal increases in his preorder are also unprofitable.
Suppose the contrary: c – m = 0 in the interior of the Bertrand region. (E1) yields (θ – m)D'(θ) + D(θ) = 0, which by definition implies θ = θ^{*}. However, it is well known that the price charged in a Cournot oligopoly (with two or more dealers) is strictly smaller than the price a monopolist would charge, p^{Cournot} (m;R) < θ^{*}, so (E4) would be violated. Therefore, we must have c > m in the interior of the Bertrand region.
References
Anderson, E., Fitzsimons, G. J., & Simester, D. (2006). Measuring and mitigating the costs of stockouts. Management Science, 52(11), 1751–1763.
Bertrand, J. (1883). Revue de la théorie de la recherche sociale and recherches sur les principes mathématiques de la théorie des richesses. Journal des Savants. 499–508.
Blattberg, R., & Neslin, S. (1990). Sales promotion: Concepts, methods, and strategies. Englewood Cliffs: Prentice Hall.
Carlton, D., & Perloff, J. (2005). Modern industrial organization. New York: AddisonWesley.
Coughlan, A. T., Anderson, E., Stern, L. W., & ElAnsary, A. I. (2006). Marketing channels. Upper Saddle River: Prentice Hall.
Davidson, C., & Deneckere, R. (1986). Longterm competition in capacity, shortrun competition in price, and the Cournot model. RAND Journal of Economics, 17, 404–415.
Desai, P. S. (2000). Multiple messages to retain retailers: signaling new product demand. Marketing Science, 19(4), 381–389.
Dixit, A. (1982). Imperfect competition and public policy: recent developments in oligopoly competition. American Economic Review, 72, 12–17.
Dutta, S., Bergen, M., & John, G. (1994). The governance of exclusive territories when dealers can bootleg. Marketing Science, 13, 83–89.
Ehrlich, I., & Becker, G. (1972). Market insurance, selfinsurance, and selfprotection. Journal of Political Economy, 80, 623–648.
Eichenbaum, M. (1989). Some empirical evidence on the production level and production cost smoothing models of inventory investments. American Economic Review, 79, 853–864.
FTC (Federal Trade Commission) (2003). Slotting allowances in the retail grocery industry: selected case studies in five product categories.
Ingene, C., & Parry, M. (1995). Channel coordination when dealers compete. Marketing Science, 14(4), 360–377.
Katz, M. (1989). Vertical contractual relations. In R. Schmalensee, R. Willig, eds. Handbook of industrial organization. Vol. 1, Chapter 11. Elsevier Science, New York.
Kotchen, M., & Salant, S. (2010). A free lunch in the commons. Journal of Environmental Economics and Management, 61, 245–253.
Kreps, D. M., & Scheinkman, J. A. (1983). Quantity precommitment and Bertrand competition yield Cournot outcomes. Bell Journal of Economics, 14(2), 326–337.
Kuksov, D., & Pazgal, A. (2007). The effect of costs and competition on slotting allowances. Marketing Science, 26(2), 259–267.
Lo, D., Salant, S. (2015). Early bird discounts for dealers. Santa Clara University discussion paper.
Maggi, G. (1996). Strategic trade policies with endogenous mode of competition. American Economic Review, 86, 237–258.
Mathewson, R., & Winter, F. (1984). An economic theory of vertical restraints. RAND Journal of Economics, 15, 27–38.
McCardle, K., Rajaram, K., & Tang, C. (2004). Advance booking discount programs under retail competition. Management Science, 50(5), 701–708.
Padmanabhan, V., & Png, I. P. L. (1997). Manufacturer’s returns policies and retail competition. Marketing Science, 16(1), 81–94.
Padmanabhan, V., & Png, I. P. L. (2004). Reply to “Do returns policies intensify retail competition?”. Marketing Science, 23(4), 614–618.
Shaffer, G. (1991). Slotting allowances and resale price maintenance: a comparison of facilitating practices. RAND Journal of Economics, 22, 120–135.
Shapiro, C. (1989). Theories of oligopoly behavior. In R. Schmalensee, R. Willig, eds. Handbook of industrial organization. Vol. 1, Chapter 6. Elsevier Science, New York.
Spengler, J. (1950). Vertical integration and antitrust policy. Journal of Political Economy, 58, 347–352.
Tirole, J. (1988). The theory of industrial organization. Cambridge: MIT Press.
Viscusi, W., Vernon, J., & Harrington, J. (2005). Economics and regulation of antitrust. Cambridge: MIT Press.
Vives, X. (1985). On the efficiency of Bertrand and Cournot equilibria with product differentiation. Journal of Economic Theory, 36, 166–175.
Wang, H. (2004). Do returns policies intensify retail competition? Marketing Science, 23(4), 611–613.
Acknowledgments
We are indebted to Enoch Hui, Bora Kara, and the owner and sales executives at Computec as well as to other company executives who preferred to remain anonymous for many helpful discussions. We also thank the associate editor, Dmitri Kuksov, and an anonymous referee for constructive comments. The usual disclaimer applies.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A: Colgatepalmolive’s promotion on Ajax laundry detergent
Source: Adapted from Blattberg and Neslin (1990), pp. 319–321.
Note: The terms in this promotion correspond to the notation in our model as follows: θ = 19.90; c = 14.30.
Appendix B: Kaspersky’s 2009 midyear promotion
Following is the 2009 midyear promotion offered by Kaspersky for its sales promotion within the Beijing area of an antivirus consumer software product.
2.1 Beijing region Kaspersky “2010 version” product preorder policy
Respected Dealer Friends,
Kaspersky’s 2010 new fullfunction security software is soon to be launched. Our preorder policies are as follows:
Preorder price:

Kaspersky FullFunction Security Software 2010 Version: 130 yuan/unit
Payment: between July 17–22, 2009^{a}
Promotional details for Kaspersky FullFunction Security Software 2010 Version

Each 100unit order gets 4 additional free units^{b}
Kaspersky products strictly follow our regional sales policy. Orders from this promotion are restricted to sales within the Beijing region. We will not honor our incentives if your products are found to have been sold in other regions. Please contact our Beijing branch for details of this promotion.
Note: The terms in this promotion correspond to the notation in our model as follows: θ = 130; c = (100/104) (4 free additional units per 100 units as preorder discounts; excluding other discounts from which we abstract).
^{a} Delivery of the packaged software CDs, based on dealers’ requests, can start as soon as the preordering period ends.
^{b} In addition, Kaspersky offers other incentives on preorders. These additional terms are described in Kaspersky’s “DealershipDevelopment” document and may include sales quotas, monetary rewards, prizes, and dealerspecific support in advertising, consumer promotions, and promotional materials.
Appendix C: Computec’s 2005 second quarterly promotions
C1. Computec’s second quarterly promotion in 2005 for its consumer product.
3.1 Computec “new version” product sales policy
Respected Computec Dealers and Distributors:
After several years of hard work and cooperation between Computec and our dealers, Computec’s product has shown growing market shares. Our sales trend is strong and every region is exhibiting hot sales. This makes our product the unassailable leading brand in the market.
To enhance consumer demand in the summer, Computec is launching our “New Version” product, together with millions of dollars of marketing expenses, to organize this promotion campaign. At the same time, to reward our core channel partners and retail outlets during summer holidays, we will start the “1000store Cool Gift” campaign. We hope our dealers will seize this opportunity to set another sales record.
In this summer promotion campaign, we will stabilize channel prices to protect the profitability of our dealers by penalizing those who viciously lower channel prices.
Our sales policies for the promotion are:

A.
Product and Price

Product name: “Computec Consumer Product 2005 Version”

List wholesale price: 95 yuan/unit

Discounted wholesale price: 90 yuan/unit


B.
Duration

July 7–8, 2005 (based on the time shown on your bank telex deposits)

From July 9 onward, Computec 2005 Version reverts to its regular list wholesale price of 95 yuan/unit


C.
Ordering Policy

Product delivery starts on July 10, 2005 in the order in which we have received payments during the promotion period.^{a}

Computec’s sales managers may assign sales quotas.^{b} We will offer additional marketing and advertising support for those whose orders exceed these quotas.

Computec Technology Co. Ltd.
Note: The terms in this dealer promotion correspond to the notation in our model as follows: θ = 95; c = 90 (excluding quantity discounts from which we abstracted).
^{a} Although a dealer has to take all deliveries before a specified date, the dealer can divide his total preorder into smaller portions and decide when to take delivery of each portion.
^{b} The sales quota includes a minimum order of 1,000 units in this sales cycle.
C2. Computec’s second quarterly promotion in 2005 for its small and medium enterprise (SME) product line
3.2 “Computec product small & medium enterprise (SME) version” product promotion policies
Respected Computec Dealers and Distributors:
Thank you for your support and cooperation for the past many years.
Computec is becoming an integrated manufacturer in the computer accessory market. Our market share and sales are both increasing rapidly for our hardware and software.
The peak selling season has arrived. Again, Computec is organizing our “SME version” promotion. In this promotion, Computec is offering the following incentives for our longterm dealers and distributors:

1.
Promotion Period
From 13:30 08/17/2005 to 17:00 08/19/2005 (based on the time shown on your bank telex deposits)

2.
Ordering price during the promotion period:
For our assigned dealers, your ordering price is 2% off the standard list wholesale prices.

3.
Special incentives
Orders of more than 80000 yuan get 1 unit of Product X^{a} free.

Orders of more than 200000 yuan get 3 units of Product X free.

Orders of more than 500000 yuan get 8 units of Product X free.

All orders during the promotion period are also eligible for routine annual discounts.^{b}


4
Other regulations

Those who ordered from us during this promotion period are eligible to sign an SME agreement with Computec. We will register those company names in “Computec SME dealership list.”^{c}

We may offer advertising and other marketing support for those who participate in this promotion.


6.
After the promotion period, the ordering prices of “SME version” will revert to the list wholesale prices terms specified in the dealership agreement.

7.
After we receive your payment, we will start product delivery. Dealers must take possession of all ordered units on or before 17:30 09/02/2005.^{d}

8.
Computec reserves all rights to interpret the policies and regulations of this promotion.
Computec works with you for prosperity.
Computec Technology Co. Ltd.
Note: The terms in this promotion correspond to the notation in our model as follows: θ varies depending on the product; c = 0.98θ (2% preorder discounts; excluding quantity discounts from which we abstracted).
^{a} Computec also makes and sells an enterprise computer product, referred to here by the pseudonym Product X, which is complementary to the SME product line.
^{b} Annual discounts are on average 10% and based on preorders taken during the four quarterly promotion periods in 1 year. Together with the 2% discount mentioned above, the total discounts for preorders amount to 12%. At the same time, there are minimum orders (sales quotas) for the dealers for each promotion period.
^{c} Computec distributes this list to existing and potential customers so that the latter will solicit business from those that are listed.
^{d} Although a dealer has to take all deliveries before a specified date, the dealer can divide his total preorder into smaller portions and decide when to take delivery of each portion.
Appendix D: Proof of proposition 1
Suppose R ≥ 2 dealers remain after the exit stage. Denote the preorder of dealer i as q_{i} and the retail price it subsequently chooses as p_{i}. Denote the aggregate preorder as \( \mathrm{Q}={\displaystyle \sum_{\mathrm{i}=1}^{\mathrm{R}}{\mathrm{q}}_{\mathrm{i}}} \). In this appendix, we prove Proposition 1: If p^{Cournot}(c;R) > θ, then p = θ and Q = D(θ), whereas if p^{Cournot}(c;R) ≤ θ, then p = p^{Cournot}(c;R) and Q = Rq^{Cournot}(c;R).
Recall that under the “rules of the game,” if the lowest retail price is p, then the D(p) customers arrange themselves so that a lowestprice firm that preordered q_{i} would get q_{i}D(p)/Z customers, where Z = ∑_{j∊{lowest price firms}}q_{j.}
It follows that if D(p)/Z ≤ 1, every dealer has preordered weakly more than its customers demand, and if D(p)/Z > 1, each dealer has preordered less than its customers demand. In the latter case, each dealer is required to satisfy the remaining demand ([D(p)/Z] – 1)q_{i} > 0) by augmenting the preorder at the undiscounted wholesale price (θ). We denote the inverse demand curve P(·) and assume it is strictly decreasing and weakly concave.
4.1 Pricing subgames
Denote the number of firms remaining after the exit stage as R. Assume R > 1 and that the manufacturer has chosen wholesale price θ and discounted price c. Then, in the symmetric Nash equilibrium of each subgame indexed by {q_{i}, Q_{i}, R} for i = 1,…,R, each dealer will set its retail price equal to p_{i} = max (0, min(θ, P(Q)). To prove that this profile of retail prices forms a Nash equilibrium in the subgame, we consider three cases:

Case 1:
When observed total preorders are small; that is, P(Q) > θ. In the conjectured equilibrium, p_{i} = p_{i} = θ, Q = D(θ), dealer i would get in revenue θq_{i} + (θ – θ){[D(θ)/Z] – 1}q_{i}. The first term is the revenue from selling the preorder. The second term reflects the net margin per unit (its first factor) multiplied by the amount of augmenting required under the rules of the game (the product of the second and third factors). Since consumers would pay the dealer what it cost him to augment (θ per unit), augmenting earns zero margins and the firm’s revenue would be θq_{i} > 0. Dealer i cannot strictly increase his revenue by unilaterally deviating from p_{i} = θ. If he strictly raised his retail price, he would lose all his customers and earn zero revenue. Suppose instead he unilaterally reduced the retail price to θ – ε > 0. Then he would earn less on his preorders but would have to augment more to cover all the remaining demand. As a result, he would earn the negative amount (θ – ε) – θ < 0 on every additional unit.

Case 2:
When observed total preorders are large; that is, θ ≥ P(Q) > 0. In the conjectured equilibrium, p_{i} = p_{i} = P(Q), Q = Z. Since D(P(Q))/Z = 1, each dealer’s preorder exactly satisfies the demand of his customers, and there is no need to augment. Dealer i would get in revenue P(Q)q_{i} ≥ 0. Unilaterally raising his price would result in no customers and no revenue. Unilaterally reducing his price by ε > 0 would result in smaller revenue from his preorders. Moreover, as the firm charging the lowest price, he would be required to augment and would lose money doing so, since he would earn the negative amount (P(Q) – ε) – θ < 0 on each additional unit. Hence, neither deviation would strictly improve dealer i’s revenue.

Case 3:
When observed total preorders are very large; that is, Q ≥ D(0) ⇔ P(Q) = 0. In the conjectured equilibrium, p_{i} = p_{i} = 0. In this case, dealer i would earn zero revenue, but he would also earn nothing if he unilaterally increased his price. Reducing his price unilaterally is infeasible, but if it were not, it would be unprofitable.
We conclude therefore that if R > 1, every firm will charge the retail price p_{i} = max(0, min(θ, P(Q)) for i = 1, …, R.
4.2 Preordering subgames
Assume that the number of firms that did not exit is R > 1 and that the manufacturer has chosen wholesale price θ and discounted price c.

Case 1:
Suppose p^{Cournot}(c;R) > θ. Then in the proposed subgameperfect equilibrium, q_{i} = q_{i} = D(θ)/R. If this strategy profile is played, every dealer will subsequently set the retail price P(R · D(θ)/R) = θ, since this is the equilibrium price in the pricing subgame that follows. Dealer i would earn q_{i} D(θ)/R > 0 in the proposed equilibrium. If dealer i unilaterally reduced his preorder, he would incur a strict loss, since the common retail price would remain θ. So he would lose the margin θ – c > 0 on each unit he failed to preorder and would break even on the amount he was required to augment. If dealer i unilaterally increased his preorder instead, his gross margin would also unilaterally decline. To see this, note that his gross margin would equal q_{i}[P(q_{i} + (R – 1)D(θ))/R) – c]. Since the first and second derivatives of the inverse demand function are negative by assumption, this function is strictly concave. Moreover, the first derivative of the grossmargin function is strictly negative for q_{i} > D(θ)/R > q^{Cournot}(c;R).^{Footnote 23} Finally, if he unilaterally preordered a very large amount, the common retail price in the pricing stage would fall to zero. He would scrap and get zero gross margin from his preorders. Hence, no dealer has a strict incentive to deviate unilaterally in either direction from his proposed equilibrium strategy.

Case 2:
Suppose dealers observe that p^{Cournot}(c;R) ≤ θ. Then in the proposed equilibrium, q_{i} = q_{i} = q^{Cournot}(c;R). Since Q = R · q^{Cournot}(c;R) < D(0) and p^{Cournot}(c;R) = P(R · q^{Cournot}(c;R)), the common retail price in the last stage would be 0 < p_{i} = p_{i} = p^{Cournot}(c;R) ≤ θ. If dealer i deviated locally by preordering a larger amount, the common retail price in the final stage P(Q) would fall to clear the market. Since dealer i would no longer be choosing the Cournot best reply to the unchanged preorders of the other (R – 1) dealers, his net profit would strictly decline. If dealer i unilaterally preordered a very large amount, the common retail price in the pricing stage would fall to zero. Those dealers who had placed a preorder, including him, would scrap; he would get zero gross margin. On the other hand, if he unilaterally preordered an amount smaller than q^{Cournot}(c;R), the common retail price in the final stage would rise. Suppose it remained below θ. Then dealer i would no longer be choosing the best reply to the preorders of the other (R – 1) dealers, and his net profit would decline. Finally, any unilateral reduction in i’s preorder after the common retail price reached θ would also be harmful. Dealer i would lose the margin θ – c > 0 on each unit reduction in his preorders beyond that point; he would break even on each unit that he was required to augment, since the price customers pay would just cover the undiscounted wholesale price the manufacturer would charge dealer i.
To summarize our findings, in a symmetric subgameperfect equilibrium, if p^{Cournot}(c;R) > θ, then p = θ and Q = D(θ), whereas if p^{Cournot}(c;R) ≤ θ, then p = p^{Cournot}(c;R) and Q = Rq^{Cournot}(c;R).
Appendix E: Proof of proposition 2
To analyze the constrained profitmaximization problem of the manufacturer for the R > 1 case, we append the multiplier λ to constraint (1), γ to constraint (2), and δ to constraint (3). The Lagrangean of the problem is
At an optimum, the following conditions must hold:
The notation “c.s.” is an abbreviation for “complementary slackness.” The condition a ≥ 0, b ≥ 0, c.s. means that both a and b are nonnegative and, in addition, that at least one of these variables is zero; thus, the condition eliminates the possibility that both variables are strictly positive.
We begin with three preliminary observations:

1.
(E2) requires λ > 0, and therefore (E3) requires D(θ)(θ – c) – RF = 0.

2.
(E3) requires that θ > c > 0. Since θ > 0, (E1) implies (c – m)D'(θ) + λ[D'(θ)(θ – c) + D(θ)] – γ = 0.

3.
(E5) requires c > 0. Since c > 0, (E2) requires D(θ) – λD(θ) + γp'^{Cournot}(c;R) + δ = 0.
5.1 Optimum in the interior of the Bertrand region
If the optimum occurs in the interior of the Bertrand region, (E4) requires γ = 0, and since λ > 0, (E3) requires that D(θ)(θ – c) – RF = 0. Moreover, c – m > 0, implying that δ = 0.^{Footnote 24} We deduce from (E2) that λ = 1 and from (E1) that θ = θ^{*}, the implicit solution to (θ – m)D'(θ) + D(θ).
5.2 Optimum on the Rdealer CB boundary of the Bertrand region
On the CB boundary, p^{Cournot}(c;R) – θ = 0, so (E4) implies that γ ≥ 0. Recall that c^{*} is the solution to p^{Cournot}(c;R) – θ^{*} = 0. That is, (c^{*}, θ^{*}) is on the CB boundary at the channelprofit maximizing wholesale price. If this is not a solution, it is because D(p^{Cournot}(c^{*};R))(p^{Cournot}(c^{*};R) – c^{*}) – RF < 0. The lefthand side is the sum of dealer profits in Cournot equilibrium where each firm has constant marginal cost c = c^{*}. It has been shown (Kotchen and Salant 2010, p. 247) that, given our assumption that the inverse demand curve is weakly concave, Cournot industry profits are strictly decreasing in the common constant marginal cost. It follows that if D(p^{Cournot}(c^{*};R))(p^{Cournot}(c^{*};R) – c^{*}) – RF < 0, the constrained optimum occurs at θ < θ^{*}.
5.3 Existence and uniqueness
If RF is sufficiently large, the constraint set will be empty and the manufacturer goes out of business, earning a payoff of zero. If the constraint set is nonempty, the solution to the constrained maximization problem must be unique. Recall that this solution occurs either in the interior of the Bertrand region with θ = θ^{*} or on the Rdealer CB boundary. We first prove that if there exists one solution in the interior of the Bertrand region (respectively, on the Rdealer CB boundary), there cannot be any other solution in the interior of the Bertrand region (respectively, on the Rdealer CB boundary). We conclude by proving that if a solution occurs in one of these two regions, there cannot be an additional solution in the other region.
There can be at most one solution in the interior of the Bertrand region. The assumption that the inverse demand curve is weakly concave ensures that channel profits are a strictly concave function of θ, and hence there is a unique channelprofit maximizing wholesale price. Given that price, if one discounted price (c) generates dealer gross margins exactly equal to their fixed costs, no other discounted price will achieve exactly those margins, since dealer gross margins monotonically fall as the preorder cost rises.
There can also be at most one solution on the Rdealer CB boundary, for along that boundary, dealer gross margins increase monotonically as c decreases.
The optimum can occur in only one of these two regions, because as one reduces c and moves horizontally across the Bertrand region at θ^{*} and then continues down the Rdealer CB boundary (θ < θ^{*}), dealer gross margins monotonically increase. Hence, if they exactly covered fixed costs in the interior of the Bertrand region, they must exceed the fixed costs anywhere on the Rdealer CB boundary. Similarly, if they exactly covered the fixed costs at θ < θ^{*} somewhere on the Rdealer CB boundary, they must be strictly smaller at any point in the interior of the Bertrand region with the wholesale price set at the channelprofit maximizing price (θ^{*}).
Appendix F: Mathematical proof of example
6.1 The case of R = 1
We begin by deriving the profit function of the manufacturer when demand is linear and the one remaining dealer has fixed cost F.
Given D(p) = (a – p) / b, in the Spengler’s (1950) case, the single dealer’s optimal retail price for a given manufacturer’s wholesale price θ is
And the single dealer’s optimal quantity is D(p(θ)) = \( \frac{\mathrm{a}  \uptheta}{2\mathrm{b}} \). The manufacturer maximizes his profit, \( \left(\uptheta  \mathrm{m}\right)\frac{\left(\mathrm{a}  \uptheta \right)}{2\mathrm{b}} \), and obtains his optimal wholesale price θ^{*} = (a + m) / 2 . Maximized manufacturer profit is therefore
The Spengler solution holds provided the gross margin of the dealer weakly exceeds his fixed cost. Substituting the optimal wholesale price θ^{*} = (a + m) / 2 into the exclusive dealer’s profit function, we get
\( {\uppi}_{\mathrm{d}}=\left(\mathrm{p}\left(\uptheta \right)\ \hbox{}\ \uptheta \right)\mathrm{D}\left(\mathrm{p}\left(\uptheta \right)\right)\ \hbox{}\ \mathrm{F}=\left(\frac{\mathrm{a} + \uptheta}{2}  \uptheta \right)\left(\frac{\mathrm{a}  \uptheta}{2\mathrm{b}}\right)  \mathrm{F}=\frac{{\left(\mathrm{a}  \mathrm{m}\right)}^2}{16\mathrm{b}}  \mathrm{F} \). Thus, Spengler’s solution holds for any \( \mathrm{F}\le \frac{{\left(\mathrm{a}  \mathrm{m}\right)}^2}{16\mathrm{b}}=\frac{1}{4}{\Pi}_{\max }. \)
If the fixed cost is larger, however, the manufacturer must reduce his wholesale price or his single dealer will exit. In this case the endogenous variables (p, θ, π’s) are determined by the two constraints (4) and (5) in the main text and by the following equation defining manufacturer profit:
Hence, \( \frac{\mathrm{d}{\uppi}_{\mathrm{m}}^1}{\mathrm{d}\uptheta}=\frac{\mathrm{a}+\mathrm{m}  20}{2\mathrm{b}} \). On the other hand, substituting \( \mathrm{p}\left(\uptheta \right)=\frac{\mathrm{a} + \uptheta}{2} \) and \( \mathrm{D}\left(\mathrm{p}\left(\uptheta \right)\right)=\frac{\mathrm{a}  \uptheta}{2\mathrm{b}} \) into (5) and differentiating, we obtain \( \frac{\mathrm{d}\mathrm{F}}{\mathrm{d}\uptheta}=\frac{\left(\mathrm{a}  \uptheta \right)}{2\mathrm{b}}<0 \). Therefore,
Equation (F2) implies that the second portion of π ^{1}_{m} starts to descend at \( \mathrm{F}=\frac{\prod_{\max }}{4} \) and since θ^{*} = (a + m) / 2, \( \frac{\mathrm{d}{\uppi}_{\mathrm{m}}^1}{\mathrm{d}\mathrm{F}}=0 \), implying that there is no kink in π ^{1}_{m} at the transition point. Moreover, \( \frac{\mathrm{d}{\uppi}_{\mathrm{m}}^1}{\mathrm{d}\mathrm{F}} \) is larger than −1 because the first term in (F2) is positive; but since \( \frac{\mathrm{d}\mathrm{F}}{\mathrm{d}\uptheta}<0 \), this term is decreasing in F. Hence, the second portion of π ^{1}_{m} starts out flat and is decreasing and strictly concave in F until it reaches the horizontal axis (π ^{1}_{m} = 0). Since at that point θ = m, Eq. (F3) implies \( \frac{\mathrm{d}{\uppi}_{\mathrm{m}}^1}{\mathrm{d}\mathrm{F}}=1 \).
6.2 The case of R = 2
In the text, we show that in the Bertrand region the manufacturer’s profit function π ^{2}_{m} = П_{max} − 2F declines linearly until the transition point derived in the text. Here, we describe the profit function, π ^{2}_{m} , if F is larger, and hence the manufacturer operates in the Cournot region.
In the Cournot region, the common retail price is p^{Cournot}(c; 2) = \( \frac{\mathrm{a}+2\mathrm{c}}{3} \), and each dealer preorders and sells D(p^{Cournot}(c; 2)) / 2 = \( \frac{\mathrm{a}  \mathrm{c}}{3\mathrm{b}} \). Since dealers’ total gross margins just their fixed costs, (p^{Cournot}(c; 2) – c)D(p^{Cournot}(c; 2)) = 2(a – c)^{2}/(9b) = 2 F. Solving for F, we conclude that F = (a – c)^{2}/(9b). Thus a one dollar increase in fixed costs requires the manufacturer to cut the preorder cost at a rate of \( \frac{\mathrm{dc}}{\mathrm{dF}}=\frac{9\mathrm{b}}{2\left(\mathrm{a}  \mathrm{c}\right)}<0 \). Since the manufacturer’s profit is
we have \( \frac{\mathrm{d}{\uppi}_{\mathrm{m}}^2}{\mathrm{d}\mathrm{c}}=\frac{2\left(\mathrm{a} + \mathrm{m}  2\mathrm{c}\right)}{3\mathrm{b}} \). This is negative because c < θ^{*} = (a + m) / 2. Using these results, we have
The last equality uses the fact that \( \mathrm{c}=\mathrm{a}  3{\left(\mathrm{b}\mathrm{F}\right)}^{\raisebox{1ex}{$1$}\!\left/ \!\raisebox{1ex}{$2$}\right.} \) since \( \mathrm{F}=\frac{{\left(\mathrm{a}  \mathrm{c}\right)}^2}{9\mathrm{b}} \). Since \( \frac{\mathrm{d}{\uppi}_{\mathrm{m}}^2}{\mathrm{d}\mathrm{F}} \) in (F3) is the product of a positive derivative and a negative derivative, it is negative. Since the first term on the righthand side of (F3) is positive and decreasing in F, the manufacturer’s profit function in this region is not only strictly decreasing but also strictly concave.
Rights and permissions
About this article
Cite this article
Lo, DF., Salant, S.W. The strategic use of early bird discounts for dealers. Quant Mark Econ 14, 97–127 (2016). https://doi.org/10.1007/s1112901691674
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s1112901691674
Keywords
 Distribution channels
 Sales discounts
 Advance purchase
 Slotting allowances
 Twopart pricing
 Constrainedcapacity oligopoly games