Investment under Uncertainty and Regulation of New Access Networks

Transcription

1 Discssion Paper o. -00 Investment nder Uncertainty and Reglation of ew Access etworks Roman Inderst and Martin Peitz

2 Discssion Paper o. -00 Investment nder Uncertainty and Reglation of ew Access etworks Roman Inderst and Martin Peitz Download this ZEW Discssion Paper from or ftp server: Die Discssion Papers dienen einer möglichst schnellen Verbreitng von neeren Forschngsarbeiten des ZEW. Die Beiträge liegen in alleiniger Verantwortng der Atoren nd stellen nicht notwendigerweise die Meinng des ZEW dar. Discssion Papers are intended to make reslts of ZEW research promptly available to other economists in order to encorage discssion and sggestions for revisions. The athors are solely responsible for the contents which do not necessarily represent the opinion of the ZEW.

3 on-technical Smmary Contractal and reglatory provisions for access affect incentives to invest in an pgraded network and, in particlar, a next-generation access network. Investment decisions are made nder ncertainty and have to be made over time. This papers provides a framework for taking ncertainty, risk aversion, and the timing of investment explicitly into accont. First, it evalates varios access price policies in a framework in which the incremental vale over the legacy network is ncertain. Second, introdcing risk aversion, the access price strctre trns ot to be critical for the risk profile of the investing telecom operator and of the access-seeking alternative operator. Third, some implications of the time strctre of access payments are derived. Keywords: GA, investment nder ncertainty, access price rle, telecommnications

4 Das Wichtigste in Kürze Vertragliche nd reglatorische Zgangsbestimmngen beeinflssen die Anreize eines etzbetreibers, in ein leistngsfähigeres etz z investieren. Dies gilt insbesondere für `next-generation access networks. Investitionsentscheidngen müssen nter Unsicherheit über die Zeit hinweg getroffen werden. Diese Arbeit entwickelt einen Modellrahmen, innerhalb dessen Unsicherheit über die zkünftige Marktentwicklng, Risikoaversion der Marktteilnehmer sowie Investitionsentscheidngen im Zeitablaf explizit berücksichtigt. In einem ersten Schritt werden nterschiedliche Zgangsbestimmngen hinsichtlich der Investitionsanreize nter der Hypothese bewertet, dass der inkrementelle Wert der Investition gegenüber dem bestehenden etz nsicher ist. In einem zweiten Schritt wird Risikoaversion der etzbetreiber eingeführt. Hierbei wird bestimmt, wie die Strktr der Zgangspreise das Risikoprofil des investierenden nd des Zgang schenden etzbetreibers bestimmt. In einem dritten Schritt werden Investitionen im Zeitablaf betrachtet nd Implikationen der zeitlichen Strktr der Zgangspreise abgeleitet. Schlagwörter: GA, Investitionen nter Unsicherheit, Zgangspreise, Zgangsreglierng, Telekommnikation

5 Investment nder Uncertainty and Reglation of ew Access etworks Roman Inderst Goethe-University Frankfrt Martin Peitz University of Mannheim first version: Janary 0, this version March 0 Abstract Contractal and reglatory provisions for access affect incentives to invest in an pgraded network and, in particlar, a next-generation access network. Investment decisions are made nder ncertainty and have to be made over time. This papers provides a framework for taking ncertainty, risk aversion, and the timing of investment explicitly into accont. First, it evalates varios access price policies in a framework in which the incremental vale over the legacy network is ncertain. Second, introdcing risk aversion, the access price strctre trns ot to be critical for the risk profile of the investing telecom operator and of the access-seeking alternative operator. Third, some implications of the time strctre of access payments are derived. Keywords: GA, investment nder ncertainty, access price rle, telecommnications JEL-Classification: L, L5 This paper is part of a large project for the German Ministry of Economics. A German-langage and more-policy-oriented version of this research can be fond in Inderst et al. (0), which draws heavily on the present paper. Financial spport from the German Ministry of Economics is grateflly acknowledged. Johann Wolfgang Goethe University Frankfrt. inderst@finance.ni-frankfrt.de. Corresponding athor: Department of Economics, University of Mannheim, 68 Mannheim, Germany, martin.peitz@gmail.com, also affiliated with CEPR, CERRE, CESifo, MaCCI and ZEW.

6 Uncertainty, investments and reglation of GA. Introdction Uncertainty abot the sccess of an investment is argably an important obstacle to the roll-ot of fibre networks and other investments to pgrade an existing network. In particlar, investments in next-generation access networks are associated with highly npredictable ftre profits after making the investment. Policy makers have acknowledged this. In particlar, the Eropean Commission has observed that Erope lags behind the U.S. and Asia in sch developments. The Eropean Commission states: There are several reasons, the most evident being the ncertain commercial viability of sbstantial network investments, de to prevailing investment models and the EU market strctre. Bt also becase of dobts abot consmers' short-term willingness to pay more for higher speeds, as new high vale-added digital content and services are not necessarily available yet throghot the EU This qote also highlights the importance of the ncertain development of complementary services made feasible by the new technology, as well as consmers ncertain take-p of these services. An investment s profitability also depends, more generally, on the speed of market penetration since the investment, for the most part, constittes a snk cost, and an efficient roll-ot relies on qick take-p. 5 While there may be immediate revenes, profitability depends crcially on a continos revene stream over a long period. This long time horizon tends to frther increase the ncertainty that the investing party faces. Uncertainty over the key parameters of a firm s decision and its implications also plays an important role in the modern theory of reglation. A large part of the literatre considers private information held by the party that is sbject to reglation. 6 For example, a reglator may have to set access prices withot perfectly knowing the cost fnction of the reglated firm. While this is a relevant isse, this paper will abstract from private information of this type. We focs, instead, on ncertainty abot a new technology s potential. This ncertainty has direct implications for the way that particlar reglatory interventions affect market otcomes. For instance, the reglator may oblige the investing network operator to grant access at a given fixed fee. Whether other firms will make se of this possible access is ncertain, as it depends on the tility that consmers de- 4 Commnication from the Commission to the Eropean Parliament, the Concil, the Eropean Economic and Social Committee and the Committee of the Regions, The Digital Agenda for Erope - Driving Eropean growth digitally, December 8, 0, COM(0) 784 final, page 8. 5 A high degree of ncertainty may already have been relevant at an earlier point in the development of local telecommnications networks. Pindyck (007) emphasizes the relevance of ncertainty for sch investments. However, some economists have challenged the general presmption that the network operator sffers from ncertainty; see Economides (00). 6 For instance, Armstrong and Sappington (007) provide an overview.

7 rive from additional services associated with the investment. Ths, at the moment of making the investment, its overall vale depends both directly on its sccess with consmers and indirectly on the possibility of passing on costs to other firms. The extent to which this is possible depends, in trn, on the access reglation that is in place. This isse does not arise withot ncertainty. A key aspect in or analysis is the non-investing firm s decision of whether or not to se the new technology nder the prevailing access conditions. More generally, it has to decide the extent and the timing of offering the new technology to its consmers. Efficient reglation increases the probability that another firm will se the new technology early (here, we abstract from limited capacity). If firms se the new technology symmetrically, then competition in each downstream market is intense. This tends to lead to a high consmer srpls and only small allocative inefficiencies, measred by a small deadweight loss. This observation hints at an important disadvantage of fixed access fees, which are optional in the sense that access-seeking firms decide after the investment whether they are interested in accessing the new technology. The inefficiency may be redced if the payment for access depends on the qantity of the access prodct that the access-seeking firm demands. However, introdcing sch variable access fees may lead to higher retail prices, on average. Ths, we find a trade-off between fixed and variable optional access fees. This trade-off might be attenated by sing general, non-linear tariffs. Also, from a social perspective, access fees that condition on relevant changes in market characteristics may be preferable to those that do so only indirectly. As an extreme alternative, we also discss a fixed non-conditional pfront fee, which removes the optionality of the access payment. The marginal price of obtaining access can then be set eqal to the corresponding marginal costs. With this reglatory policy, however, one shold bear in mind several caveats. Most obviosly, if, at the moment of the investment, the access-seeking firm does not yet exist, the mechanism is simply not applicable. Frther, the fixed fee shold be based on the expected demand of the access-seeking firm, if this is feasible. A policy that applies a fee to all firms indiscriminately will not allow small firms to operate profitably in the market. Finally, an indced high degree of competition, while being desirable from an ex post perspective, cold lead to nderinvestment or inefficiently postponed investment from a social welfare perspective. Apart from the access price rles mentioned above, alternative soltions exist. For instance, an access-seeking firm may acqire the option to obtain access nder predefined conditions. This allows for a combination of optional and non-optional fixed payments. Also, the contract or the reglatory rle may specify certain qantities for which preferential access can be obtained by making a non-optional pfront payment. When a wait-and-see strategy becomes more attractive for the access-seeking firm, it may se the new technology rather late, to the detriment of social welfare. Then, the time strctre of the access tariff may be sed to improve dynamic efficiency by providing incentives for earlier and more-intensive se of the new technology. We discss when non-

8 4 Uncertainty, investments and reglation of GA linear tariffs based on access levels and front-loaded access tariffs can increase welfare. As a final contribtion, we discss the optimal allocation of risk between firms. The allocation of risk depends, for example, on the degree to which the access of the noninvesting firm remains optional and on the se of fixed fees verss incremental, sagebased payments. The allocation of risk may matter for efficiency, in particlar, when firms appear to be averse even with respect to idiosyncratic risk e.g., as they have limited access to capital markets. The rest of this paper is organized as follows. Section introdces the formal framework. Section considers firms optimal investment and contractal choice, as well as the impact of varios access policies. Section 4 introdces optimal risk-sharing, while Section 5 provides an extension to a dynamic investment path. Section 6 concldes. Gide to the literatre: This paper complements the analysis of Inderst and Peitz (0a, 0b), which investigates firms incentives to invest in a new technology when starting with an old technology (i.e., a legacy network). Others also have looked at this isse. For a recent srvey, see Cambini and Jiang (009). A recent addition to the literatre is itsche and Wiethas (0), a welfare assessment of the effect of access rles on the incmbent's investment incentives. Other recent work incldes Brito, Pereira and Vareda (0) and Borrea, Cambini and Dogan (0), which consider the migration from old to new technology and the role of access rles. Less closely related are stdies that investigate the effects of access prices on the otcome of dynamic investment races (e.g., Hori and Mizno, 006; Vareda and Hoernig, 00). The contribtion of this paper is to explore the role of ncertainty in investments and retail prices, distingishing among varios types of access contracts. We distingish, in particlar, among access contracts that are signed prior to investment and commit to payments (co-investments); those that are signed prior to investment bt in which payments by one firm are optional; and those that are signed after a firm has made the investment. Or paper also complements Borrea, Cambini, and Hoernig (0), which looks at co-investment nder ncertainty in a setting with a continm of regional markets. 7. The model This paper bilds on Inderst and Peitz (0a). In the main text, we adopt a redcedform presentation; in the Appendix, we develop a specific model that satisfies all of or 7 Since Borrea, Cambini, and Hoernig (0) allow for perfect price discrimination across regional markets, a model with a single market extends to a market with a continm of regional markets in a straightforward way, as long as costs are linear in the nmber of markets. However, they postlate that these costs are strictly convex, which makes the analysis of coverage meaningfl.

9 5 redced-form assmptions in the main text and, ths, provides a micro-fondation of or approach.. Model nder certainty We consider a dopoly model with firms i =,. Firm is the only firm that may ndertake an investment. In an otherwise symmetric environment, we consider markets in which network dplication is not economically feasible i.e., a sitation of natral monopoly with respect to the new technology. We largely abstract from other asymmetries that may contribte to an incmbency advantage (for a focs on asymmetries, albeit in an environment with certainty, we refer the reader to Inderst and Peitz, 0b). The ensing price competition between operators is explicitly modeled in the Appendix see, also, Inderst and Peitz (0a). Firm i s profits are denoted by π (, i ) where i denotes the gross tility delivered by firm i. Absent additional investments, we postlate that both firms offer the same service; they both have access to a legacy network. As a measre of network qality and, ths, consmer gross tility, we have. Sperscript O stands for the old technology. Withot O any investment, firms obtain profits π O, O ) and π O, O ), respectively. ( In Inderst and Peitz (0a), we consider an investment that can be made at cost I i O O and that increases the qality and associated tility from to > (where sperscript stands for the new technology). The investment can be made by one or both firms. To offer services with this qality for all consmers, it is not necessarily reqired that both operators invest. If the investing firm grants access to the other firm, both firms can offer the high qality available with the new technology. In the following, we are interested mainly in the case in which exactly one operator i.e., operator i = invests. If the other firm i = does not obtain access, profits (gross of the investment cost for firm ) are π (, O ) and π (, O ). ote that we postlate that firm contines to have access to the old technology, althogh this may not be tre if that access is throgh, for instance, an nbndling agreement, and the interconnection points are bypassed when investing in the new technology. We elaborate on this isse in Inderst and Peitz (0b) bt abstract from it in this paper. The difference from the stats qo is O that firm offers the better technology with its additional services sch that >. Ths, firm obtains a higher gross profit than in the stats qo, ( O O O O O O π, ) > π(, ), while firm obtains a lower profit, π (, ) < π (, ). O Sbtracting the investment cost, firm makes a net profit of π (, ). ( I If, however, firm obtains access at marginal costs to firm s high-qality network, profits gross of any investment costs and transfers are π, ) and π, ). ( (

10 6 Uncertainty, investments and reglation of GA Profits depend on the access agreement or access reglation, as we will elaborate on below.. Modeling ncertainty We introdce ncertainty abot the vale of the investment by making a random variable. This reflects the ncertainty abot the availability and qality of new services for which the new technology is reqired. Initially, firms have to form expectations abot ; the expected vale is denoted by E [ ]. We postlate that O can take any vale between and some pper bond. If O =, consmers do not derive any additional tility from sing the new technology. Frthermore, we postlate that variable costs are the same for the old and new technologies. Hence, for any strictly positive difference, the new technology will be O sed if available. Cost differences cold, however, be easily incorporated. If firm has invested in and ses the new technology while restricting access, its expected net profit is O E[ π (, )] I. () When firm i = does not obtain access and, ths, contines to se the old technology O O =, its expected profit is E[ π (, )]. If access is granted or if investment costs are shared, profits will be different. These profits will be analyzed in the following section when we consider varios alternative reglatory policies.. Risk-netral firms and investments In this section, we analyze a market in which firms are risk-netral. For simplicity, we also abstract from the timing of the investment and the dependence of profits on the timing decision. Ths, only the levels of investment costs and expected profits have to be considered. Looking at both access-seeking and access-granting firms, we investigate the role of ncertainty for the evalation of different reglatory policies and contracting environments. We distingish between optional and non-optional payments for access. With regard to the former, the access-seeking firm can decide whether to seek access after the investment (and after the ncertainty has been resolved). With regard to the latter, the access-seeking firm enters a binding agreement before ndertaking the investment and cannot simply walk away when the market environment trns ot to be nfavorable.

11 7 The second distinction concerns the access price strctre with linear tariff and fixed tariff as extreme cases and non-linear price strctres as intermediate cases.. on-optional fixed fees Or simplest benchmark case is a non-optional fixed access price F that the accessseeking firm has to pay to the access-granting firm (in addition to a marginal sage fee eqal to marginal cost). Ths, F implements a particlar cost sharing determined by reglation or throgh negotiations between the two firms. More precisely, the accessseeking firm contribtes F, and the access-granting firm contribtes I F. The fixed payment may depend on the expected sage of the new technology and on the bargaining power of the two firms. We will retrn to this isse below. If the investment is made and both firms se the new technology, profits are π (, ). The eqilibrim then implements the same allocation as in a sitation in i which both firms invest, bt with the difference that the dplication of fixed costs is avoided. Investment incentives We next address the qestion of whether the investment will be ndertaken. Here, we distingish between two scenarios. In the first scenario, the investment is not profitable when firm contines to se the old technology. In the second scenario, the contrary holds. In the first scenario, when the two parties reach no agreement, no investment takes place. In the second scenario, firm will invest regardless. Scenario prevails if O O O E π (, )] I < π (, ). () [ Both firms agree with the cost sharing if O O E[ π (, )] ( I F) π (, ) and () E[ π (, )] F π ( This can be satisfied only if indstry profits satisfy O O O O E π (, ) π (, )] I E[ π (, ) π (, )]. (4) O, [ Otherwise, one cannot find a contract with cost sharing ( I F, F) that is agreeable to both firms. Whether ineqality (4) holds depends on the reqired investment level I and the likelihood that high levels of are realized. It also depends on how the new technology affects total demand and competition. The basic model developed in the Appen- O ).

12 8 Uncertainty, investments and reglation of GA dix has the featre that total indstry profits are independent of the technology in se, which implies that ineqality (4) is not satisfied in that model. Ths, demand expansion or less competitive pressre as a reslt of the investment is reqired for ineqality (4) to hold. More-intense competition makes it more likely that we are in scenario, satisfying O O O E[ π (, )] I π (, ). (5) Here, the investment is profitable for firm if the new technology is not shared with firm. With more-intense competition, demand reacts more strongly to price changes (i.e., there is a strong bsiness-stealing effect), and firm can play ot its technological advantage. With intense competition and a low elasticity of total demand so that there are no benefits from market expansion that can be shared firm will neither obtain access nor dplicate the investment. These findings are in line with what is obtained nder certainty and, ths, do not merit the statement of a formal reslt. Cost sharing and sage We next address the qestion of how costs shold be shared, provided that there is an agreement, and how ncertainty impacts the expected profits nder sch cost-sharing rles. A reasonable rle for distribting the investment cost between the two firms is to base cost sharing on the expected sage of the technology i.e., the expected nmber of sbscribers for each firm. Given the realization, we denote the nmber of sbscribers of the two firms by q (, ) and q (, ). (For a particlar specification of these demand fnctions, we again refer the reader to the Appendix.) Denoting expected demand by E q ] and E q ], we obtain cost sharing according to flly distribted costs and [ [ E[ q] F = I, (6) E[ q ] E[ q ] I F = E[ q ] I E[ q ] E[ q ] for the two firms. We note that this cost-sharing rle does not necessarily satisfy the participation constraint () in scenario, even thogh a different distribtion might exist that does satisfy it. More precisely, we can always find sch a distribtion ( F', I F' ) if ineqality (4) is satisfied, even if () is not satisfied for the rle specified in (6).

13 9. Optional fixed fees We contine to consider access provision for a fixed fee. In contrast to the previos sbsection, we now postlate that firm can opt ot after ndertaking the investment and after was realized. This means that the non-investing firm makes its decision abot seeking access after the ncertainty has been resolved (while the optional contract is signed prior to the realization ). Inefficiency after an investment Making the demand for access conditional on the realization of introdces a possible inefficiency. We note that, given the investment in the new technology, it is always socially efficient that both firms se the new technology. However, it is not necessarily privately optimal for firm to obtain sch access becase the conditions of access are not favorable for it. Paying the fixed access fee F is optimal for firm only if π, O ) F π (, ). (7) ( We consider markets in which π, ) is weakly increasing in ( (see the Appendix for details): This means that profits do not decrease if the new technology and the associated services trn ot to be more attractive for consmers. By contrast, if firm contines to se the old technology, its profit π, O ) is postlated to be decreasing in the attractiveness of the new technology. 8 Since the left-hand side of (7) is weakly increasing in and the right-hand side is decreasing, there is, at most, one intersection point. With the appropriate bondary behavior, there is a niqe critical vale = * at which (7) is satisfied with eqality. For all realizations > *, firm will seek access at price F, while for all realizations < *, it prefers not to se the new technology. This leads to an inefficiency in the se of the new technology: Reslt : With an optional fixed access price greater than zero, a given investment is inefficiently sed becase the firm that may want to ask for access decides not to do so if the new technology trns ot to be not sfficiently sperior to the old technology. The optionality that firm enjoys has the conseqence that after a bad realization of, firm contines to se the old technology, and only firm, which has snk its investment cost, ses the new technology, althogh it wold be socially efficient if both firms did. For all realizations < *, firm is the only firm that offers the new technology. This tends to give rise to an additional inefficiency, as the reslting asymmetry of firms may redce competition. Ths, consmer srpls also may sffer. ( 8 We also note that this property is not always satisfied: This may happen if consmers are sfficiently heterogeneos with respect to their preference for the old verss the new technology. With sch vertical differentiation, the firm with the old technology may benefit if the new technology becomes more attractive, as this redces competition by increasing differentiation.

14 0 Uncertainty, investments and reglation of GA The investing firm shares part of the investment cost with positive probability namely, with the probability that is greater than *, Pr[ *]. Ths, the expected contribtion to costs is f = F Pr[ *]. To share a particlar fraction of the total investment cost, it becomes necessary to increase F beyond the level with a non-optional fee. However, the larger F is, the larger is the critical vale *( F), which is implicitly defined by ( O π *, *) F = π ( *, ). This again pshes p the level F that is necessary to realize a given f, and so on. Ultimately, it may, then, not be feasible at all to share costs in this way. Investment incentives Or preceding analysis has immediate implications for firm s investment incentives. In particlar, it may not be possible with optional fees to provide sfficient investment incentives for firm. If firm does not se the new technology for realizations *, this also redces the rent that the investing firm can potentially extract. (Also, the intensity of competition matters; we retrn to this isse below.) De to the optionality to access, firm will necessarily obtain a positive net srpls. This implies that it is impossible for firm to extract all the rents that are generated throgh the investment. More precisely, firm is, at = *, indifferent abot whether it shold pay F and obtain access to the new technology. For any realization > *, it does strictly better by obtaining access. Ths, the expected rent is strictly positive. There is, however, a contervailing force. So far, we have considered only rents at the wholesale level of firm and have not evalated the conseqences for downstream competition. In particlar, for all realizations < *, firm s market power increases relative to an environment in which firm always has access to the new technology; in a sense, a high fixed fee partially forecloses the market for firm. This leads to higher downstream profit. If symmetric competition is intense and the price elasticity of total demand low, firm might actally realize higher profits if it does not share the technology, even thogh it foregoes additional wholesale revenes.. Ex-post contracts with a fixed fee So far, we have restricted orselves to a conceptal discssion of simple ex-ante and ex-post (optional) fixed-fee contracts. In what follows, we discss how these parameters are determined with and withot reglation and make the respective contractal choices more flexible. Inderst and Peitz (0a) distingish between ex-ante and ex-post contracts. The latter are negotiated and signed only after the investment is made. In the

15 setting with certainty, the range of investment costs for which there is investment in the new technology is expanded by allowing for ex-ante contracting. Introdcing ncertainty in this paper, it is important to determine whether the contracting stage comes before or after has been realized. First, we abstract from reglation and also sppose that is already known at the contracting stage. With ex-ante ncertainty, ex-post contracts then allow for adjstments to the prevailing market conditions, which tend to redce or even remove inefficiencies in the sage of the new technology. Sbseqently, we introdce reglation. Flexible cost sharing, firm makes a take- Sppose that, after the investment and after the realization of it-or-leave-it offer to firm. Firm then sets the fixed fee F ( O ) = π (, ) π (, ). (8) Hence, the investing firm extracts all the vale added to firm as a reslt of obtaining access to the new technology. Access is provided whenever indstry profits increase de to access provision: O O π, ) π (, ) π (, ) π (, ). (9) ( By the preceding remarks, for this ineqality to hold, at least one of two properties has to be satisfied: a demand expansion effect de to the new technology being made available to firm or competition becoming less intense with a better technology. Otherwise, firm will not grant access nder conditions that firm will accept, as we will discss below. We note that it is possible, in general, that ineqality (8) holds only on a sbset of [ O, ]. Since the fixed fee has now throgh ex-post contracting indirectly become a fnction of, this more-flexible fee F ( ) can increase efficiency relative to an nconditional ex-ante fee. With a take-it-or-leave-it offer, firm can flly extract the increase in firm s profit when making the new technology available to its competitor. Its profit is O π (, ) [ π (, ) π (, )]. This implies that firm s expected profits from the investment are larger than in the setting in which access cannot be provided. Sppose that ineqality (9) holds for all feasible. Then, for the investment to be made, the expected profit from investing mst be larger than the eqilibrim profit when not investing, O O O Eπ(, ) [ Eπ (, ) Eπ (, )] I π(, ). (0) This defines a critical investment level above which firm does not invest. We can now compare the investment incentives with ex-post contracts to the ones with non-optional

16 Uncertainty, investments and reglation of GA ex-ante contracts. If firm makes a take-it-or-leave-it offer, and we are in scenario (cf. above), firm s expected profit when investing is O O Eπ, ) [ Eπ (, ) π (, )]. () ( I O O O Since π, ) > Eπ (, ), the expression in () is smaller than the left-hand ( side of (0). This implies that there is a larger range of investment costs sch that firm invests nder ex-post contracts. The logic is as follows: In scenario and with exante contracts, firm knows that if it rejects the proposed contract, there will be no investment and its profit will be π O, O ). In contrast, with ex-post contracts, firm ( has already invested. Therefore, rejecting the proposed ex-post contract implies that firm still ses the new technology, with the effect that firm will earn only π, O ). Ths, with ex-post contracts, firm is in a weaker position. 9 ( Reslt : Sppose that with ex-post contracts prescribing a fixed fee, the investing firm has all bargaining power. Then, this can provide stronger investment incentives than nder nconditional ex-ante contracts with a fixed fee. egotiations, hold-p and investment incentives The fixed fee F in (8) has been derived nder the assmption that firm has all the bargaining power and can, therefore, extract the maximal rent from firm. Yet another possibility is that there is a fixed sharing rle based on the two firms bargaining power: There is ash bargaining over the gain from access to the technology, so that the srpls gained is shared in eqal parts. The vales of the otside options of the two firms O are π (, ), i =,, respectively. The srpls is i S = O π (, ) π (, )] [ π (, ) π ( [ To implement eqal srpls sharing, the fee F ( ) mst satisfy O π (, ) F( ) = π (, ) S A hold-p problem may arise in this case: The investing firm has to bear all the investment costs, which, when negotiations take place, will be snk and, ths, will not be considered dring the negotiations. This redces firm s incentives to invest. Sch a hold-p-problem does not arise nder ex-ante contracting (i.e., prior to negotiations that take place prior to the investment) or a corresponding reglation of the fee F. Comparing the soltion nder ex-post bargaining with the (optional) ex-ante version, we see the following trade-off: On the one hand, throgh a flexible adjstment F ( ), the /., O )]. 9 Inderst and Peitz (0a) do not obtain this reslt, as they focs on sitations in which ineqality (9) is not satisfied. However, it can also be shown nder certainty.

17 ex-post contract permits a more efficient otcome with respect to sage. This leads to higher indstry profits and, ths, has a positive impact on investment incentives. On the other hand, efficiency may sffer de to the hold-p problem. What this discssion leaves nanswered, however, is the qestion of whether moreflexible ex-ante contracts (or corresponding reglation) cold sfficiently improve on the ex-ante contracts considered above. We trn to this isse in the following sbsection..4 Flexible ex-ante reglation and flexible ex ante contracts In Sbsection., we pointed ot that there are stronger investment incentives if the firm that decides whether to invest has all the bargaining power. This holds nder both certainty and ncertainty. If the firm does not have all the bargaining power, the pictre is more complicated becase of a hold-p problem for the investing firm. We also showed that ex-post contracts are a good instrment for extracting srpls for the investing firm in the presence of ncertainty, as they can be conditioned on the realization of the random variable. However, in principle, ex-ante contracts can be made more flexible too, allowing them to condition on the realization of the random variable. With expost contracts, it is reqired only that the realization is observable. For ex-ante con- tracts to directly condition on this realization, it is reqired that is not only observable, bt also verifiable. The simplest case is, ths, that where can be directly con- tracted pon, e.g., via a fee. To avoid allocative inefficiencies, given the considered optionality and a redction of investment incentives, this fee shold be increasing in. If the market performs better than expected (high ), the fraction of the investment cost I covered throgh F increases. This implies that the investing firm has to finance a larger part of the investment cost if the realization trns ot to be low. Alternatively, if one wants to condition the fee on the realized profits of the access-seeking firm, a possible rle cold take the following linear form: F ) = aπ (, ); () ( i.e., the access-seeking firm has to pay a fraction a of its gross profit π (, ) after has been realized. (This involves the ad hoc assmption that firms do not strategically react and, ths, distort competition in the prodct market to maniplate the fee payment. We trn to the isse of distorted competition in a related setting below.) We observe that sch a linear rle cannot ensre the efficient adoption of the new technology. That wold reqire that the fee is 0 when =. (However, a modified rle O O O F( ) = a[ π (, ) π (, )] wold satisfy this property.) A fee that does not directly condition on, bt on profits (as in eqation ()), is no longer a fixed payment. It depends (non-linearly) on the chosen qantities q and q (nmber of sbscribers). In the next sbsection, we consider, instead, a fee that depends linearly on these qantities.

18 4 Uncertainty, investments and reglation of GA.5 Usage-dependent srcharge: Cost sharing with a srcharge A special case of a contingent contract is an access fee that linearly depends on sage i.e., the nmber of sbscribers. The access price per sbscriber is denoted by w. Ths, firm pays for access wq. As in the previos sbsection, the access price per sbscriber can be determined ex-post i.e., after the investment has been made or exante. Absent any fixed payment, sch a payment is always optional, as firm can avoid any payment to firm by avoiding enrolling sbscribers to the new technology. 0 Distorted competition If both firms se the new technology i.e., = i and if the access price depends on the nmber of sers and is greater than the marginal cost, firm faces a higher marginal cost than firm. In the linear specification, this is the case if w is greater than the cost of providing the new technology to an additional consmer. Following the idea of cost sharing according to sage, in order to determine the level of the access price w, one has to make sre that, in expectation, the access-seeking firm pays the fraction of the investment cost for access, which corresponds to the fraction of access contracts it signs. Since there is a positive margin in the access market, competition is distorted in the following way: Firm faces a higher marginal cost, leading to higher prices and a weakly smaller qantity. Firm faces a higher opportnity cost: While a lower retail price increases its own sbscriber nmber and, ths, locally, its retail revene, it also redces the nmber of nits sold by the competitor and, ths, firm s revene in the access market. Therefore, both firms have little incentive to set low prices. In the special case of constant total demand, there is a one-to-one pass-throgh of access prices into retail prices (see de Bijl and Peitz, 006). Hence, a linear access rle wq leads to less competition and, ths, lower consmer welfare. For elastic total demand, there is also a deadweight loss (following from a lower eqilibrim qantity). However, as discssed in Inderst and Peitz (0a), comparing a fixed fee F with a price per ser can lead to a trade-off between allocative efficiency and investment incentives. The trade-off arises as follows: By setting w above marginal costs, the allocative efficiency is negatively affected if total demand is not constant and is not affect- 0 ote that we consider the choice of technology as a discrete choice i.e., we do not allow a firm to offer a men of contracts with different technologies to consmers. In the model developed in the Appendix, all sers have the same preferences over different technologies, which jstifies or discretechoice assmption. However, more generally, sers may have differences in taste with respect to the possible technologies, which wold make the discrete-choice assmption more restrictive. The calclation is more involved than the calclation of F in (6) since the expected qantities of q and q have to reflect the adoption probability.

19 5 ed if total demand is constant. In any case, indstry profits increase, which tends to increase the expected profitability of the investment since the investing firm might be able to appropriate a large fraction of the increase in indstry profits. By contrast, with a fixed fee, indstry profits are constant. In the special case of constant total demand, investment incentives are stronger with linear access tariffs. With price-dependent demand, the analysis is less straightforward; however, if demand is not very sensitive to price, the abovementioned trade-off prevails. Efficient adoption of the new technology Unlike the analysis with ncertainty in Inderst and Peitz (0a), there is an efficiency argment in favor of a linear access price w relative to a fixed fee F. Since the total access payment wq positively depends on qantity and, ths, on market conditions, the adoption of the new technology by firm becomes more likely. Reslt : Uncertainty over makes the adoption of the new technology by firm more likely nder a linear access fee than nder a fixed fee. Within or framework, this is seen as follows: Depending on the access price the expected qantity q (, ), at, firm asks for access if i and O π (, ; w) w q(, ) π (, ). () Here, we have introdced the notation π (, ; ) to captre profits gross of the w payment for access to the new technology. They still depend on w since its level affects competition in the retail market. Firm s profits when sing the old technology are on the right-hand side of (), which are decreasing in. Since firm chooses its demand for access by maximizing its profits, the right-hand side of () is typically strictly increasing in (for given w ). Ths, with a linear access tariff, the access seeker s profits are positively aligned with market conditions (higher ). Ths, from (), there is a ctoff * sch that access is reqested if *. A key observation is that with a linear access tariff, firm obtains larger access revenes for higher realizations of. Therefore, to cover a certain fraction of costs in expectation, the payment for a lower can be lower. This redces the ctoff * below which firm does not seek access (relative to an optional fixed fee). This has the welfareincreasing property that firm ses the new technology more freqently and works A caveat is in order: Absent the additional instrment of a fix fee (or some other non-linear access tariffs), it is possible that while a higher w redces the intensity of competition, incremental profits (inclding access revenes) of firm generates lower profits than a fix fee, the reason being that w may allow firm to appropriate only small part of the profit increase of firm. If an access price per sbscriber is reached throgh negotiations instead of throgh reglation, the danger of market foreclosre tends to be redced (in particlar, if the access tariff also incldes a fixed fee). Recall that investments sch that firm does not grant access happen only if indstry profits with joint se are less than indstry profits when firm does not obtain access.

20 6 Uncertainty, investments and reglation of GA against the negative effect of a higher access price w on competition between the two firms. on-linear sage prices By definition, a linear payment for access with the access price w has the property that the marginal cost to enroll an additional ser is constant for firm. An alternative is to allow for price discrimination between different demand profiles in the access market by, for example, allowing qantity disconts for firms that enroll a large nmber of sbscribers. This provides incentives to se the access prodct to a greater extent. In particlar, w may depend on q sch that the price per sbscriber is decreasing in q : Denoting this price by w ( q ), this fnction decreases in q. Given a realization (since qantities depend on total demand in the retail sector and, ths, on ), the deadweight loss wold become smaller. Both firms wold have an incentive to increase their qantities for firm, as the marginal price it obtains from granting access declines, which provides incentives to keep the competitor small; and for firm, as the marginal opportnity cost of increasing its qantity falls. 4 Sch a non-linear access price may rn conter to the goal of redcing the payment by the access-seeking firm in the case of bad realizations of. This makes it possible that the reslting ctoff * is lower and, hence, that both firms se the new technology. This, again, is de to the property that a low realization of reslts in low qantities q i(, ). 5 A higher realization of increases the average price paid for access, as long as w q ) is decreasing in qantity. ( To smmarize, an access price w( q ) that is decreasing in q exhibits two contervailing effects: On the one hand, if firm obtains access to the new technology, the nmber of sbscribers tends to go p relative to linear access prices. On the other hand, the likelihood that firm will se the new technology decreases. It is possible that this trade-off, introdced by a qantity discont, can be alleviated throgh more-complex access price tariffs that condition w not only on the qantity reqested by the accessseeking firm, bt also on other observables that depend positively on (sch as prof- its, in analogy to the access tariff for the fixed fee in eqation ()). 4 We abstract from the isse that small competitors may be at a disadvantage as a reslt of qantity disconts at the wholesale level. Ths, we do not inclde the important long-rn isse that the market may end p being more concentrated as a reslt of qantity disconts. 5 In the model in the Appendix, this reqires that demand in the hinterlands depends on prices.

21 7.6 Alternative access tariffs We next frther extend the set of considered (non-linear) access tariffs to discss access-price policies (as the otcome of negotiations or reglatory intervention) that also play a role in the policy debate. Access options The investing firm, prior to the investment, may want to sell an access option to firm. In or framework, this may take the following form: A fixed fee F to be paid initially is non-optional. After the realization of, firm can choose the qantity q ; the corresponding part of the access payment is optional. The advantages and disadvantages of this type of negotiated contract or reglated access tariff can be immediately inferred by or preceding analysis. The fixed fee F redces the inefficiency arising from firm s socially inefficient adoption of the new technology after has been realized. However, we recall that fixed ex-ante fees have the disadvantage that they exclde firms that enter the market at a later time. In addition, barriers of entry may arise for small firms. Frthermore, a higher w relaxes competition in the retail market, bt tends to make technology adoption by firm more likely. The lessening of competition may be welfare-increasing from a dynamic perspective, as it may make the investment profitable for certain levels of I. Possibly, offering a men of access contracts with one contract specifying the access option conditional on paying pfront F allows for cost sharing at an ex-ante stage among firms already present in the market, while still leaving entry possible at a later stage, albeit at less-favorable terms at the margin. Capacity limits and demand reqirements Another type of access contract specifies a capacity i.e., a certain maximal nmber of firm sbscribers that this firm can access for making a payment (in particlar, a fixed fee F ). If firm cold mix the new and old technologies (a fraction of sbscriptions to the old and another to the new), this can be seen as a commitment by firm to demand at least q. A higher qantity may be available for additional access payments. This means that marginal costs are higher for large qantities than for small ones; the opposite sitation prevails with qantity disconts. In general, there are no clear-ct lessons concerning the implications of non-linear access prices since reslts depend on the level q. If the capacity threshold is high relative to the realization of, then we expect competition to be more intense, while the opposite holds tre for low capacity relative to the realization of : Here, firm s high marginal costs for qantities q > q lead to high retail prices. Advantages and disadvantage of an additional fixed fee F were discssed above.

22 8 Uncertainty, investments and reglation of GA In practice, these kinds of contracts are also discssed with respect to (efficient) risk sharing between firms. The isses of risk sharing and, relatedly, the determination of the risk premim, depending on the types of contracting, and the chosen reglatory approach will be investigated in the following section. 4 Risk averse firms and the risk premim 4. Preliminary considerations Under risk aversion, a firm considers the expected cash flow to be less valable if it is more risky. 6 Pt differently, becase of ncertainty, the vale of expected profits has to be adjsted downward; here, we abstract from the time dimension, which is the topic of Section 5. Modern capital market theory sggests, however, that firms shold ask for a risk premim for systematic, bt not for idiosyncratic, risk since the former increases the costs of capital. If this were the only reason for a risk premim, we wold immediately obtain the following reslt: Risk sharing of the investment and the corresponding determination of, say, a fixed access charge F are irrelevant from an efficiency point of view. Pt differently, the premim reqired, in total, to afford the investment is the same no matter how the risk is shared between the two firms. This netrality no longer holds if there are firm-specific reasons for risk aversion. The finance literatre has identified a nmber of reasons why firms shold be averse to idiosyncratic risk. 7 In particlar, frictions in the capital market play a key role in explaining firms risk aversion. Broadly speaking, internal financing can be less costly than external financing. The efficient risk sharing between investing and access-seeking firms, then, depends on the firms abilities to bear the risk of the investment and, in particlar, on the firms external financing needs and their access to the capital market. This may depend on the particlar company; however, these parameters are endogenos. Reglation that conditions on these different abilities rns the risk of generating inefficiencies in the way that firms make their financing and other relevant firm decisions. In addition, financial intermediaries offer possible insrance mechanisms to firms. 8 6 In or simple framework, we do not need to distingish between cash flows and profits. 7 An overview and good access to this literatre is provided by Ross et al. (008). 8 In this section, we abstract from reglatory ncertainty. We cold introdce reglatory ncertainty in particlar, after presenting or dynamic extension in the following section. The term reglatory ncertainty refers to the ncertainty that prevails after existing reglation is revised. Given the long time horizon to recover investments in GA networks, this is an important isse. To the extent that reglatory ncertainty is part of the general ncertainty, one has to ask to what extent it contribtes to systematic risk and, ths, jstifies a higher risk premim. However, reglatory risk can also mean that ftre reglation is not a random event, bt depends on market conditions. Using that definition, reglation responds in a predictable way to market otcomes and redces expected profits of access-