Entrepreneurial innovations and taxation

Journal of Public Economics 113 (2014) 13–31

Contents lists available at ScienceDirect

Journal of Public Economics journal homepage: www.elsevier.com/locate/jpube

Entrepreneurial innovations and taxation Andreas Haufler a,b,⁎, Pehr-Johan Norbäck c, Lars Persson b,c,d a

University of Munich, Germany CESifo, Germany c Research Institute of Industrial Economics, Stockholm, Sweden d CEPR, United Kingdom b

a r t i c l e

i n f o

Article history: Received 23 August 2012 Received in revised form 16 January 2014 Accepted 4 March 2014 Available online 12 March 2014 JEL classification: H25 L13 M13 O31

a b s t r a c t Stimulating entrepreneurship is high on the policy agenda of many countries. We study the effects of tax policies on entrepreneurs' choice of riskiness (or quality) of an innovation project, and on their mode of commercializing the innovation (market entry versus sale). Limited loss offset provisions in the tax system induce entrepreneurs innovating for entry to choose projects with inefficiently little risk. The same distortion does not arise when entrepreneurs sell their innovation in a competitive bidding process to an incumbent before the uncertainty is revealed. Tax systems which systematically favor market entry of entrepreneurs can thus lead to welfare losses due to inefficient quality choices, despite leading to more competition in the product market. © 2014 Elsevier B.V. All rights reserved.

Keywords: Business taxation Innovation Loss offset rules Market entry

1. Introduction How to stimulate entrepreneurship is one of the pressing policy challenges facing many countries in the world today. In Europe, the European Commission (2008) launched the “Small Business Act for Europe” in response to this challenge. This document sets out a comprehensive policy framework for the EU and its member states to create an environment that rewards entrepreneurship, specifically mentioning taxation in this context. At the same time, several EU countries are reforming their tax systems to make them more compatible with entrepreneurship. In the Netherlands, for example, the 2011 Tax Plan comprises a whole set of measures explicitly aimed at encouraging entrepreneurship. Another example is Sweden, which is currently making the largest investigation of its tax system in 20 years, with the main objective of making it more conducive to entrepreneurship.1

⁎ Corresponding author at: Seminar for Economic Policy, University of Munich, Akademiestr.1, D-80799 Munich, Germany. Tel.: +49 89 2180 3858, fax: +49 89 2180 6296. E-mail address: Andreas.Haufl[email protected] (A. Haufler). 1 As one of the measures already signed into law, half of the costs of buying shares in young and small firms have been made tax-deductible as of 1 December 2013. The taxdeductibility is granted up to thresholds of 1,300,000 Swedish kronors per individual and up to 20,000,000 kronors per firm.

http://dx.doi.org/10.1016/j.jpubeco.2014.03.002 0047-2727/© 2014 Elsevier B.V. All rights reserved.

One of the main reasons for the support of entrepreneurs is the important role they play in fostering ‘qualitative’ innovations. Cohen (2010, p. 137), who reviews the empirical literature on firms' size and R&D, concludes that small firms tend to account for a disproportionately large share of innovations, and that – where data exist – they also tend to pursue more significant innovations than incumbents.2 A further qualitative decision taken by entrepreneurs is that they commercialize their inventions or business ideas not only by entering the product market, but also by selling them to incumbent firms. Indeed, a substantial share of inventions made by independent innovators is commercialized through the sale or the licensing of a patent.3 Against this background, the present paper aims to increase our knowledge about how the tax system affects these qualitative dimensions of entrepreneurship. It is well known that taxation will reduce the amount of entrepreneurial activities; see Henrekson and Sanandaji (2011) for a recent survey of this literature. However, little is known about the effects of taxes 2 This conclusion is confirmed by Henkel et al. (2010) in a recent empirical study of the electronic design automation industry. The importance of small entrepreneurial firms in the innovation process in the United States is documented in Baumol (2004). 3 Serrano (2010, Table 1) reports for the United States that entrepreneurs sold 17.5% of their patents during the period 1983–2001, and this share increases to 24% if the patents are quality-weighted with the number of citations received. Furthermore, a large-scale survey carried out in six EU countries suggests that roughly 10% of all patents are licensed (Giuri et al., 2007).

14

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

on either the risk-taking behavior of entrepreneurs, or on their decision of how to commercialize their innovation. Our analysis incorporates both of these decision margins and we show that they are interrelated. In studying the effects of tax policies towards entrepreneurship, we focus on two fundamental features of existing tax systems. First, most countries provide various support schemes for start-ups and small businesses that cover all stages of the firms' development and range from initial research grants to the provision of subsidized loans and state guarantees. Financial support and cheap credit in the marketing stage are thus an important part of this policy package, in order to overcome asymmetric information in credit markets. However, entrepreneurs can typically only take advantage of subsidized loans when they enter the product market.4 These provisions imply that many countries in Europe and elsewhere grant fiscal advantages to entrepreneurs who market their inventions themselves, rather than selling out to an incumbent firm. Depending on country-specific tax rules, further tax advantages may arise when the entrepreneurial start-up firm is incorporated and the corporate tax rate on its profits in the product market is below the personal tax rate applicable on the firm's sale.5 The second important feature for our analysis is that tax systems in all OECD countries incorporate a fundamental asymmetry as positive earnings are taxed immediately, whereas losses can only be offset against positive incomes.6 The rationale for this asymmetry, which we take as given in our analysis, is to prevent tax revenue shortfalls and fraudulent claims for tax refunds. In an empirical study based on U.S. corporations for the period 1993–2003, Cooper and Knittel (2006) estimate that only about 50–60% of tax losses are used over a ten-year window, whereas the remaining 40–50% of tax losses are either never used, or are still unused after a period of 3–10 years. Incorporated start-up firms are particularly affected by loss offset restrictions, because they cannot carry back current losses and, in many countries, are also not allowed to offset corporate losses against positive incomes taxed under the personal income tax. Small, incorporated entrepreneurs, who are developing a single business idea, will therefore often not be able claim a tax refund when they invest in a project that turns out to be unsuccessful. In contrast, restrictions on loss offset opportunities are less important for mature and diversified firms, which are more likely to be able to offset the losses against positive profits in other business lines (see Cooper and Knittel, 2006; Mirrlees et al., 2011, p. 454). Therefore, selling a start-up firm to a mature incumbent increases the likelihood that tax losses can indeed be used. The tax and business conditions for entrepreneurs in Germany provide a good example of these effects. On the one hand, Germany subsidizes entrepreneurial market entry by business start-up grants and, in particular, by subsidized loans.7 At the same time, the possibility (since 2008) to incorporate as an ‘entrepreneurial firm’ (Unternehmergesellschaft, similar to the Limited Company in the United Kingdom), has proved to be the most attractive organizational form for entrepreneurial start-ups in Germany. A central issue for entrepreneurs in Germany arises, however, from limited loss offset opportunities. In addition to the general restrictions mentioned above, a specific

4 In the United States, for example, one of the main programs to promote small businesses is the Technological Innovation Programme (TIP), which subsidizes the commercialization of successful prototypes with up to USD 3 million. This support scheme is available only if the small or medium-sized enterprise (SME) markets the product itself, or is the leading company in a joint venture (OECD, 2010, p. 106). See OECD (2010) for a listing of similar support schemes for SMEs in all OECD states. 5 This is discussed in more detail in Section 2. 6 A business tax system that is neutral in the presence of uncertainty would require the government to pay a tax refund equal to the tax rate times the full value of the initial investment in case of default (Bond and Devereux, 2003). This full refundability is not granted by any OECD country, however. An early empirical analysis of the resulting tax asymmetries is Altshuler and Auerbach (1990). 7 These loans are made by the government-owned Kreditanstalt für Wiederaufbau (KfW), a national development bank, as well as by similar banks owned by the German states. In 2012, the KfW alone provided a total volume of Euro 24 billion in subsidized loans to SMEs (Kreditanstalt für Wiederaufbau, 2013).

feature of the German tax law is that tax losses cannot be transferred from one firm to another. This restriction also applies to venture capital firms that finance different stages of a start-up's development and it is particularly relevant for industries that incur high initial losses, such as the biotech industry. The German Expert Commission for Research and Innovation (EFI) therefore regard limitations on the tax deductibility of losses as one of the most important obstacles for the development and growth of small, innovative businesses (see EFI, 2011, p. 19). To capture these effects, we develop a model where a European-style tax system influences the decisions of the entrepreneur to select an R&D project, and to choose the mode of commercializing the innovation. In the initial research stage, the entrepreneur chooses the quality, or riskiness, of an innovation, trading off higher returns against an increased probability of failure. The entrepreneur can then sell the project at an early stage, or keep the innovation until its success is known. The trade-off faced by the entrepreneur is that selling before the uncertainty is lifted will ensure a positive sales price, thus allowing her to deduct the initial research investment with certainty. On the other hand, if the entrepreneur keeps the project and it turns out to be successful, she can add further value to the project in the ensuing development stage. At the end of the development stage, the entrepreneur can then once again sell her project to one of the incumbent firms, or enter the product market herself. A core element of our model is that the equilibrium sales price of the firm or patent is explicitly derived from the valuation of the risky innovation by both the incumbents and the entrepreneur. These valuations are affected by the features of the tax system described above. In particular, the entrepreneur can benefit from lower effective taxes when entering the product market, but only if the project has proven successful. We first establish that whenever the entrepreneur enters the market in equilibrium, she will choose an innovation that involves too little risk, in comparison to the socially desirable project. This distortion arises because there is an extra cost of failing for the entrepreneur in terms of forgone tax deductions, and hence she finds it optimal to reduce the risk of being left with non-deductible investment outlays.8 In contrast, when the entrepreneur sells her invention in a competitive bidding process before the uncertainty is revealed, she will be guaranteed a full deduction of her investment costs and hence chooses a discretely more risky project in the initial research stage. Based on these results, we then evaluate the welfare effects of tax policies towards entrepreneurship. We find that promoting entrepreneurial market entry by means of low effective profit tax rates is indeed the optimal policy when consumers value the additional variety produced by the entrepreneur, and when the value she can add in the development stage is potentially large. In contrast, if the entrepreneur has only a small comparative advantage in developing her own innovation, then the welfaremaximizing equilibrium is one where the entrepreneur sells her project at an early stage and chooses the more risky research project. Reaching this equilibrium requires, however, that there be no substantial tax disadvantage from selling the project to an incumbent firm, as compared to product market entry. Hence, we find that discriminatory tax policies aimed at promoting entrepreneurial market entry can be counterproductive in an environment with imperfect tax loss offset, if the quality dimension of entrepreneurial innovations is critical. Our findings are in line with recent empirical results showing that lower corporate taxes induce entry by new firms (Djankov et al., 2010; Da Rin et al., 2011), but also reduce the size and the capital intensity of the new entrants and thus lead to ‘weaker’ firms, on average (Da Rin et al., 2010). Moreover, our conclusions correspond to a recent stream in the entrepreneurship literature which questions policies 8 Econometric studies support the result that the asymmetric tax treatment of profits and losses has significant effects on entrepreneurial risk-taking (Cullen and Gordon, 2007), as well as on firms' investment incentives (Edgerton, 2010). In a recent study, Langenmayr and Lester (2013) show that firms which are unable to use loss carryback provisions (either because they have no previous profits, or because their home country does not allow loss carrybacks) take systematically fewer risks than firms which can use loss carrybacks.

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

aimed at increasing market entry and resulting in new firms that do not create stable employment and wealth (see e.g. Shane, 2008). Our model brings together two different strands in the literature. On the one hand is the public finance literature that analyzes the effects of taxes on various decision margins of entrepreneurs. Several contributions focus on the progressiveness of the personal income tax schedule as an obstacle to entrepreneurial activity (e.g. Gentry and Hubbard, 2000). In contrast, Gordon (1998) stresses that start-up enterprises have the option of incorporating, thus benefiting from the widespread fall in corporate tax rates over the last few decades. Fuest et al. (2002) show that the presence of a substantial positive tax gap between the personal income tax and the corporation tax is a second-best solution to the problem of asymmetric information faced by new firms. Keuschnigg and Nielsen (2002, 2004) focus on the effects of various tax policies when entrepreneurs face financial constraints and set up a contract with a venture capitalist under conditions of one-sided or two-sided moral hazard. None of these contributions, however, incorporates a choice between different R&D projects, nor the option for the entrepreneur to sell her invention to an incumbent firm.9 Secondly, this paper is also related to the literature on R&D and market structure, which mainly focuses on the level of R&D efforts.10 Several papers study the type of R&D project undertaken by firms and entrepreneurs (e.g. Bhattacharya and Mookherjee, 1986). There is also a literature on entrepreneurship and innovations, which is summarized in Acs and Audretsch (2005). A recent literature focuses on the quality of entrepreneurs, and on their allocation between large and small firms (Hvide, 2009; Friebel and Giannetti, 2009). Asoni and Sanandaji (forthcoming) examine how progressive taxes affect an entrepreneur's dynamic search for higher-quality projects, but do not consider the option to sell the invention. To our knowledge, the only analysis considering how the entry mode affects the type of R&D is Färnstrand Damsgaard et al. (2010). However, this paper focuses on the interaction between entrepreneurial and incumbent innovations and abstracts from tax policies, which are central to the present study. The remainder of the paper is organized as follows. Section 2 gives an overview of our model and its assumptions. In Section 3, we solve the different stages of the model and determine the equilibrium allocation in different tax regimes. Section 4 analyzes the effects of tax policy on equilibrium project choices and commercialization modes and studies the welfare consequences of the different equilibria. Section 5 discusses several extensions of our benchmark model and Section 6 concludes. 2. The framework We consider an imperfectly competitive market with n identical incumbent firms. Entry costs deter further firms from entering the market, unless they have a superior technology. The focus of our analysis lies on the decisions of an independent innovator, or entrepreneur, who can invest in an innovative project. In line with the evidence quoted in the Introduction, we focus on entrepreneurs as providers of major inventions and assume that incumbent firms do not innovate.11 Incumbents are nevertheless important in our framework as they may act as aquirers of the entrepreneur's innovation. Hence, they offer their pre9 Among the empirical contributions, Cullen and Gordon (2007) estimate the effects of imperfect loss offset provisions on entrepreneurial risk taking. Egger et al. (2009) provide empirical evidence that a positive tax gap between personal and corporate tax rates favors incorporation. Bloom et al. (2002) and Ernst and Spengel (2011) show that R&D tax credits have positive and significant effects on the volume of R&D, but do not address the choice of quality (or riskiness) of an innovation project. 10 See Reinganum (1989) and Gilbert (2006) for overviews, and Rosen (1991) and Cabral (2003) for specific models. 11 Similar results are obtained in a setting where incumbents are simultaneously allowed to innovate, and the entrepreneur can only make a profit when her innovation is successful, while the incumbents' innovation fails. In this setting – which is likely to hold in situations where the incumbents' initial advantage is sufficiently strong – the incumbents' ability to innovate does not affect the entrepreneur's choices, on which the focus of the present analysis lies (see Färnstrand Damsgaard et al., 2010).

15

existing channels to commercialize the innovation as an alternative to entrepreneurial market entry. Moreover, since incumbents have preexisting profits, selling to an incumbent firm will allow the entrepreneur to overcome the legal constraints on tax loss offsets that have been stressed in the Introduction. The sequence of events in our benchmark model is shown in Fig. 1. In Stage 1 the success of the project is still uncertain and the entrepreneur chooses the riskiness of the innovative project. In Stage 2, after the uncertainty has been revealed, the entrepreneur can undertake an additional investment to further develop the project. After each stage, the project can be sold to one of the incumbents. Alternatively, the entrepreneur can enter the market at the end of Stage 2. In Stage 3 the incumbents, and potentially also the entrepreneur, interact in the output market. In Period I of Stage 1, the entrepreneur makes a fixed, monetary research investment IR in order to develop a risky innovation. We suppose that there is an infinite number of independent research projects that the entrepreneur may undertake, all requiring the same investment costs IR. Hence, investment projects do not vary by the size of the investment, but by the riskiness of the chosen project. Each project k is characterized by a certain success probability pk. Along the technological frontier, entrepreneurs face a choice between projects that have a high success probability pk but deliver a small payoff – a cost reduction in our model – in case of success, and projects that are more risky but also have a larger payoff, if successful. Importantly, we assume that the entrepreneur is risk-neutral and thus chooses the project which maximizes the expected net payoff from the investment.12 Our benchmark model assumes that a successful invention reduces only the fixed costs of production. This assumption greatly simplifies the exposition, as it implies that product market competition between all firms remains symmetric and the product market price does not depend on the chosen project.13 To give an example, the fixed cost of producing a prototype part for a new airplane or a racing car can be reduced by small, low-risk improvements in existing technologies. A high-risk, high-return alternative is instead to develop a 3D printer which ‘prints’ the prototype part using titanium powder, causing virtually no waste of this precious material in the process.14 With projects differing by their degree of innovation, fixed production costs are F ðpk Þ ¼ F−Γðpk Þ;

ð1Þ

where Γk′(pk) b 0, pk ∈ (0, 1). We assume that the expected fixed cost reduction pΓ(p) is strictly concave in p. In Period II of Stage 1, the entrepreneur can decide on whether to keep the uncertain innovation project until its success is revealed (and then either sell in Stage 2, or enter the market), or to sell it to an incumbent at this early stage. The benefit of an early sale is that there will be a certain, positive income from the sale, irrespective of the eventual success of the project. Hence the entrepreneur will be able to tax-deduct her initial research outlays IR with certainty. At the beginning of Stage 2, the uncertainty is revealed. If the project fails, the entrepreneur's initial investment costs IR are not tax deductible. To protect the income tax base and prevent fraud, existing tax codes allow the deductibility of expenses only in combination with positive income, but do not pay out negative taxes to the taxable entity in case of a loss. Moreover, it is also not possible for the entrepreneur to sell her unused tax credit to an incumbent, as the tax authorities will 12 Hence, we eliminate the well-known effect that taxes may stimulate entrepreneurial risk-taking by making the government a silent partner in the risky activity (Domar and Musgrave, 1944). However, this effect is fully effective only when losses are taxdeductible. Since our analysis explicitly focuses on the limitations of loss offset provisions, the Domar–Musgrave effect is of lesser importance. 13 In Section 5 we consider the more general case where the invention reduces variable production costs. 14 See e.g. the article “The printed world” in The Economist, 10 February 2011.

16

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

Fig. 1. Overview of the game.

not accept a link between an incumbent's positive income from existing assets and the losses incurred by the R&D project. Finally, due to entry costs and fixed costs of production, entering the market will also be unprofitable when the innovation fails and fixed production costs thus remain high. Hence, if the project fails, the entrepreneur will lose all her research investment IR.15 In case of success the innovation can be further developed prior to commercialization. This might involve, for example, finding cheaper inputs, improving the reliability of the product, or improving its design. The entrepreneur is likely to know most of the features of her specific project and therefore has an advantage in this respect. To capture this in a simple way, we assume that after the uncertainty is revealed, only the entrepreneur can develop the project further by making an additional development investment ID.16 The benefits of this development are described by a further fixed cost reduction B(ID), which increases the value of the innovation. At the end of Stage 2, the entrepreneur

15 Our static model abstracts from the possibility that the entrepreneur can carry forward the loss for a certain number of years. Empirical evidence suggests that failed start-ups are rarely able to use loss carry-forward provisions in subsequent years (Cooper and Knittel, 2006). See also Auerbach (2007) for empirical evidence documenting the importance of unused tax credits among U.S. firms. 16 We thus exclude the possibility that an incumbent undertakes an early acquisition and then contracts the entrepreneur to finalize the innovation. The reason is that the contracts needed in such a setting are likely to suffer from strong moral hazard problems, as it is difficult in large organizations to monitor such tasks efficiently and appropriately reward them (see, e.g. Norbäck and Persson, 2009). Another reason is that the entrepreneur may have difficulties adjusting to the job culture in the large organization used by the incumbent. Indeed the business press frequently reports that entrepreneurs who become employees in the acquiring firm leave this firm after the acquisition (see e.g. http:// www.cnbc.com/100839046).

then makes a further commercialization decision of whether to sell the project at this late stage, or enter the market herself.17 In Stage 3, oligopolistic product market competition occurs between n or (n + 1) firms, depending on the commercialization decision of the entrepreneur and on the success of the project. If the entrepreneur decides to sell her project, the acquiring incumbent will replace his initial technology with the innovative one. In this case, there will thus still be n firms in the market, though one firm (the acquirer of the innovation) may have a superior technology. In the case where the entrepreneur decides to enter the market, there will be (n + 1) firms in the market, once more with one firm (the entrepreneur herself) having a superior technology. Equilibrium profits are paid out and taxes are collected on all income. Our framework thus allows the entrepreneur to sell her R&D project at two different points in time, either before or after the uncertainty about the success of the project is revealed.18 This choice will depend on the deductibility of the initial investment expenses, but also on the

17 An alternative set-up with very similar implications would be to introduce a transaction cost for an early sale in Stage 1, which captures the costs of communicating the qualities of a still uncertain project to an acquiring incumbent. In contrast, no transaction costs would arise under a sale of the certain project in Stage 2. In such a model the advantage of a late sale would thus lie in reduced transaction costs, rather than in an additional value created in the development stage. 18 Empirical evidence suggests that this conforms to actual practice when we regard the revelation of uncertainty as the legal granting of a patent. Thus, Svensson (2007a, Fig. 1) analyzes a sample of 530 patents granted to small and medium-sized Swedish firms in 1998 and shows that in a majority of cases the commercialization started before that date. Moreover, of the 61 patents in this sample that were sold directly, more than 50% were sold before the patent was granted (Svensson, 2007b).

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

17

Table 1 Tax treatment of entrepreneurial income (2011). European Tax Handbook (2011) and Global Individual Tax Handbook (2011). International Bureau of Fiscal Documentation, Amsterdam. OECD Tax Database: www.oecd.org/ctp/ taxdatabase, Table II.4 Country

(a)

(b) a

Austria Belgium Denmark Finland France Germany Ireland Italy Netherlands Norway Spain Sweden Un. Kingdom United States a b c d e f g h i k

(c) b

(d) c

(e) d

(f) e

Classification of business sale

Tax rate for business sale (%)

Corporate tax rate (%)

Total dividend tax rate (%)

Loss carryback

Loss carryforwarde

Capital gain Capital gain Ordinary income Ordinary income Capital gain Ordinary income Capital gain Ordinary income Ordinary income Ordinary income Ordinary income Ordinary income Capital gain Capital gain

25 16.5/33f 44.3 49 29 45 25 44.9 52 40 43 26.3/56.5g 10/28i 15k

25 34 25 26 34.4 30.2 12.5 27.5 25 28 30 26.3 28 39.2

43.8 43.9 56.5 40.5 57.6 48.6 48.4 36.6 43.8 48.2 43.3 48.4 52.7 52.1

0 0 0 0 3 1 1 0 1 0 0 0–6h 1 2

indef. indef. indef. 10 indef. indef. indef. 5 9 indef. 15 indef. indef. indef.

Including taxation of patent sales and royalties. For capital gains: capital gains tax rate; for ordinary income: top national and representative sub-central tax rate, not including social security contributions. Regular central and sub-central corporate tax rate. Corporate tax rate and personal income tax rate on dividends, net of imputation credits. Statutory periods under the corporate income tax, in years. Lower rate applicable to sale of business; higher rate applicable for sale of patents. Lower rate applicable to business income of individuals that is retained in a reserve fund. Loss carryback possible for up to 6 years through dissolution of reserve fund. Lower rate applicable to the first GBP 5 million from sale of a business. 20% as of 2012.

differential tax treatment of profits earned in the market vs. the profits made by selling the innovation to an incumbent firm. We denote by t the tax rate faced by the entrepreneur when she enters the market, whereas τ is the tax rate applicable for selling the project to an incumbent. A central assumption of our analysis is that the tax on the entrepreneur's own profits in the market, net of subsidies received, is lower than the tax that is due under project sale. This is stated in: Assumption 1. The effective tax rate (net of subsidies received) is lower for the entrepreneur when she enters the market herself than when she sells the innovation to an incumbent, t b τ. Assumption 1 is primarily motivated by the direct subsidies to entrepreneurial market entry that exist in most OECD countries (see OECD, 2010). As we discussed in the Introduction, an important part of these subsidies applies at the marketing stage. Typically, these packages are available only to small start-up firms, but not to (large) incumbents. Another source of a tax advantage of market entry is that the profit tax paid under entry is based only on current-period profits, whereas the income tax due under project sale incorporates the value of the entire firm or patent (see Landsman and Shackelford, 1995; Ayers et al., 2003). Hence market entry acts as a way to avoid this ‘lock-in effect’ of capital gains taxes and reduces the present value of aggregate tax payments. In addition, entrepreneurs producing for market entry have the option to incorporate their business and pay taxes under the corporate income tax. In contrast, entrepreneurs producing for sale will usually remain unincorporated and are thus subject to personal income taxation. Depending on the exact provisions in the host country's national tax system, this may provide a further advantage of market entry. Table 1 collects several indicators of national tax treatment of entrepreneurial incomes. With respect to the taxation of a company sale, a core distinction is between countries classifying the sale of a business as ordinary personal business income and those classifying it as a (taxprivileged) capital gain. As a result tax rates that apply to the profitable sale of a business diverge widely across countries (columns a and b). In

most countries, corporate tax rates are substantially below (top) personal income tax rates (column c). When dividend taxes accruing at the personal level are added to the corporate tax rate, however, a tax benefit of incorporation remains only in a few countries, and it is typically small (column d). Dividend taxes can be postponed, however, by keeping the profits in the incorporated business. Moreover, for profitable entrepreneurial firms, corporate taxes can be further reduced by registering the firm in a low-tax country, or by setting up a subsidiary in a low-tax country and shifting profits there.19 Therefore, incorporating the entrepreneurial firm may provide a tax advantage even in countries where the total dividend tax rate nominally exceeds the tax rate for a business sale. Finally, columns (e) and (f) show that business losses can be carried forward for extended periods, and in many countries even indefinitely. However, as we have discussed in the Introduction, the use of business losses is limited in practice by the need to generate enough positive profits in subsequent periods. Loss carrybacks, in contrast, are permitted only in a minority of countries, and they extend only to a few years. In sum, Table 1 shows that there may be a further tax advantage of market entry in those countries where business sales are taxed at high personal income tax rates. This is true for several countries in continental Europe, including Germany, Italy and the Nordic countries. In the following we will assume, for concreteness, that the entrepreneur incorporates when she enters the market herself.20 Importantly, however, entrepreneurial firms can always choose to remain non-corporate, if there are instead tax advantages to this organizational form.21 19 See e.g. Desai et al. (2004) for empirical evidence on profit-shifting in multinational firms. 20 There is substantial evidence that the larger and the more successful among the new firms are more likely to incorporate (see Da Rin et al., 2011). In Sweden, for example, only 25% of all firms which started up in 2005 and were still active in 2008 were incorporated (of a total of almost 30,000 start-ups). Among the incorporated start-ups, however, about 72% were high-growth firms, as compared to 34% high-growth firms in other groups (see Tillväxtanalys, 2010). 21 A special case arises in the United States, where various tax provisions give firms the opportunity to combine a legally corporate form with tax payments under the personal income tax or, conversely, choose a non-corporate form but pay taxes under corporate law. We thank an anonymous referee for pointing this out to us.

18

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

Therefore, our Assumption 1 is not tied to a specific organization of the entrepreneurial start-up, and it can be fulfilled for either corporate or non-corporate firms that receive subsidies for market entry from various government programs.

form profit expressions for each firm that are quadratic in equilibrium output and given by   2 πðmÞ ¼ q ðmÞ ¼



a−c 2 þ γ ½nðmÞ−1

2

;

ð4Þ

3. Equilibrium project choice and mode of commercialization thus satisfying π(i)Nπ(e). 3.1. Stage 3: product market interaction 3.2. Stage 2: development We solve the model by backward induction and start with the interaction of firms in the product market. The industry we consider is characterized by fixed costs of production F (in the absence of an innovation) and by entry costs G. Free entry to the industry ensures that product market profits (sales revenue minus variable costs) will, in equilibrium, not exceed the sum of fixed production costs and entry costs. We assume that, in the absence of an innovation that reduces fixed costs, the industry will support n symmetric, incumbent firms. In addition, there is a single entrepreneur, who may enter the market on account of an innovation that reduces fixed production costs. More specifically, let the set of firms in the industry be J ¼ fi1 ; i2 …in ; eg where i1 … in denotes the identical incumbent firms and e is the entrepreneur. The owner of the invention is denoted by m∈ J . In the product market interaction, firm j chooses an action xj, which represents the setting of either a quantity or a price, to maximize its product market profit πj(xj, x− j, m). This depends on its own and its rivals' market actions, xj and x− j, and on the identity of the owner of the invention, m. We assume that a unique Nash equilibrium {x∗j (m), x∗− j(m)} exists at this stage, which results in positive product market profits. We can then set up a reduced form for the product market profits of each firm j, taking as given ownership m: h i   π j ðmÞ ≡ π j x j ðmÞ; x− j ðmÞ; m

∀ j∈ J :

ð2Þ

Since incumbents are symmetric before the acquisition takes place, we need only distinguish between two types of ownership of the invention: entrepreneurial ownership (m = e) and incumbent ownership (m = i). Moreover, since the innovation affects only fixed production costs, the product market profit before deduction of fixed costs and taxes is always the same for all active firms. Hence, there are only two possible levels of such profits: π(i) is the profit of each incumbent when the entrepreneur does not enter the market, whereas π(e) is the product market profit of incumbents and the entrepreneur in case of entry. We assume that market entry by the entrepreneur will reduce the profit of each producer due to stronger competition, i.e. π(i) N π(e). This assumption is met in standard models of imperfect competition, such as the oligopoly model of quantity competition in a homogeneous good, or the model of price competition with differentiated products. As an example, which will be used in our welfare analysis in Section 4, consider quantity competition in differentiated products (Singh and Vives, 1984; Häckner, 2000). The utility of a consumer is then given by: U ðq; Z Þ ¼

n ðmÞ X j¼1

0 1 nðmÞ X 1 @X 2 aqi − q þ 2γ q j qk A þ Z; 2 j¼1 j k≠ j

ð3Þ

where q is the vector of outputs, qi denotes the output of a firm j, Z is a composite numeraire good and a is a constant. The parameter γ measures the substitutability between products, which is inversely related to the level of product differentiation. If γ = 0 each firm has monopolistic power, whereas if γ = 1 the products are perfect substitutes. Note finally that the number of firms is n for m = i and (n + 1) for m = e. Utility maximization yields inverse demand functions Pj = a − qj − γq− j, where Pj is the price of good j and q− j = ∑ k ≠ j qk. Quantity competition between firms with constant marginal costs c leads to reduced-

At the beginning of this stage, in Period I, it is revealed whether the innovation turns out to be successful or not, where ‘success’ can either be interpreted in a technological or in a commercial sense.22 For example, the uncertainty may be revealed by the granting of a patent, or by the results of mechanical or medical tests which determine whether a new, cost-saving technology is feasible. For other innovations, it may be revealed at this stage whether a small-scale market test shows a sufficient acceptance among prospective buyers to make the introduction of the new technology commercially viable. If the project proves to be unsuccessful, the failed innovation does not have any implications for the market equilibrium. If the entrepreneur has already sold her project to an incumbent in Stage 1, the incumbent's fixed cost remain at F and the Stage 3 profits of the acquiring incumbent are the same as for all other firms in the market. If the entrepreneur has not sold the project in Stage 1, market entry will be unprofitable in case of project failure owing to the entry costs G. To simplify the algebra in the following, it proves convenient to assume that the profits from market entry without an improved technology are negative by a very small amount ε, which we approximate by zero in the following.23 This is formalized in: Assumption 2. When the innovation fails, net profits from entry approach zero from below, π ðeÞ−F−G ¼ −εb0, and the entrepreneur does not enter the market. If the innovation project succeeds, we have to differentiate two cases, depending on whether the entrepreneur has sold her project in Stage 1 or not. If the project has been sold in Stage 1, the success of the innovation has no consequences other than affecting the profits of the acquiring incumbent in equilibrium. The superiority of the new technology over the existing ones is then reflected in reduced fixed costs, Γ(p), which follow from the Stage 1 investment of the entrepreneur. These increase the profits of the acquirer by the savings in fixed costs. The interesting case arises when the entrepreneur has not sold in Stage 1, and the project is successful. In this case the entrepreneur will further develop her project in a Period II. For this purpose she undertakes a development investment ID that delivers additional fixed cost savings B(ID). Hence, if the project is carried through to this stage, it confers total fixed cost savings of Γ(p) + B(ID) over the existing technology. In Period III, the entrepreneur finally decides on how to commercialize the innovation at this late stage. 3.2.1. Commercializing a developed innovation The commercialization process in Period III is depicted as an auction where n incumbents simultaneously post bids bi for the invention and b = (b1 , … b i …, bn ) is the vector of these bids. Following the announcement of b, the invention may be sold to one of the incumbents at the bid price, or remain in the ownership of the entrepreneur. If more than one bid is accepted, the bidder with the highest bid obtains the invention. If there is more than one incumbent with such a bid, each such 22 Formally, this is decided by nature through a lottery where the probability of success is the probability p associated with the entrepreneur's optimal project choice in Stage 1 (see also Fig. 1). 23 This condition implies that incumbents considered the probability of an additional entry as being low when they initially entered the product market. This could either be because incumbents regarded the probability of a future entrepreneur being successful with an innovation as low, or because they considered it unlikely that a successful entrepreneur would choose entry over project sale.

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

incumbent obtains the invention with equal probability. If the entrepreneur rejects all bids, she will enter the market herself, making positive net profits as given in Eq. (2). The acquisition game is solved for Nash equilibria in undominated pure strategies. 3.2.1.1. The entrepreneur's reservation price. As discussed in Section 2, the entrepreneur faces the (effective) tax rate t in case of market entry. We assume that this tax is levied at a proportional rate. If the entrepreneur instead innovates for sale, she will be taxed at the tax rate τ on her capital gains, which are defined as the excess of the sales price over the investment costs. In either case, all investment costs IR and ID are fully deductible from the entrepreneur's income at this stage.24 Using Assumption 2 and defining Δ2e (S2) as the entrepreneur's net gain from selling the invention in Stage 2 at price S2 over the alternative of market entry gives   2 2 2 Δe S ¼ ð1−τÞS þ τ ðIR þ ID Þ −fð1−t Þ½Γ ðpÞ þ BðID Þ þ t ðIR þ I D Þg: |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} after‐tax profit from sale

after‐tax profit from entry

Note that investment costs IR and ID are sunk when this commercialization decision is taken: they only impact the net gain of the entrepreneur by generating tax deductions. Let the reservation price of the entrepreneur be we = min S2, s.t. Δ2e (S2) ≥ 0. That is, we is the minimum price S2 at which the entrepreneur is willing to sell in this stage. Solving for Δ2e (S2) = 0, i.e., setting the net profit from sale equal to the net profit from entry, we get we ðpÞ ¼

ð1−t Þ ðτ−t Þ ½ΓðpÞ þ BðID Þ− ðI þ ID Þ: ð1−τÞ ð1−τ Þ R

ð5Þ

The reservation price we in Eq. (5) equals the entrepreneur's profits under market entry, adjusted for the differential tax treatment of earnings under the two commercialization modes. Moreover, recalling Assumption 1, the reservation price is reduced by the fact that the value of the tax deductions for IR and ID will be higher under project sale. Finally, note that the reservation price in Stage 2 is monotonously decreasing in the profit tax t: ∂we ðpÞ −1 ½ΓðpÞ þ BðID Þ−IR −I D b0: ¼ 1−τ ∂t

ð6Þ

3.2.1.2. Incumbents' valuations. We now turn to the incumbent firms' valuation of the invention. When an incumbent acquires the invention, it is certain that there will only be n firms in the market in Stage 3 and hence its product market profit is always given by π(i). When not acquiring the entrepreneurial firm, the invention can either remain in the hands of the entrepreneur (ownership index m = e), or it can be acquired by a rival incumbent (m = i). This difference will affect the profits of a non-acquiring incumbent. The profits of incumbent firms are taxed at the rate ti.25 Finally, as discussed above, the acquiring firm can always deduct the sales price S from its existing profits π(i).26 We denote by Δ2im (S 2) the net gain for an incumbent firm i of acquiring the entrepreneur's invention in Stage 2 when m = e,i is the alternative owner of the new technology. Given that total fixed cost savings from the innovation at this stage are Γ(p) + B(ID), this yields Δ2im(S2) = (1 − ti)[π(i) − π(m) + Γ(p) + B(ID) − S2]. Solving 24 In general, the investment ID will include both monetary investments and effort by the entrepreneur. We assume for simplicity that all effort-related costs are also tax-deductible. 25 The tax rate ti will typically exceed the tax rate t faced by the entrepreneur under market entry, because incumbents are not eligible for reduced tax rates or support schemes tied to small businesses. This difference is immaterial for our results, however, because competitive bidding ensures that the acquiring incumbent earns no extra profits from buying the invention. 26 Our static model assumes that the full sales price S is immediately tax-deductible for the acquiring firm. In actual tax law, the sales price is treated like an investment that can only be deducted pro rata over several years. Hence, in a dynamic model the present value of being able to deduct S would be somewhat lower for the acquiring firm, but this would not qualitatively alter our results.

19

for wim ≡ max S2, s.t. Δ2im(S2) ≥ 0, we obtain the incumbents' willingness to pay for the innovation as wii ðpÞ ¼ ΓðpÞ þ BðID Þ;

ð7Þ

wie ðpÞ ¼ ΓðpÞ þ BðI D Þ þ ½π ðiÞ−π ðeÞ:

ð8Þ

Here wii is the valuation of the certain innovation by an incumbent when a rival incumbent would otherwise obtain it. We label this the competitive valuation of a late acquisition. In contrast, wie is the incumbent's valuation when the entrepreneur would otherwise keep the innovation and enter the market. This is labeled a takeover valuation. Note that taxes have no direct effect on the incumbents' valuations wim(p) (which equals their maximum willingness to pay). The reason is that the sales price S2 is tax-deductible for the acquiring firm and, at its maximum willingness to pay, neither losses nor gains are created for the incumbent acquirer (cf. footnote 25). Note that wie(p) − wii(p) = π(i) − π(e) N 0 follows from the additional market entry in the product market stage. The incumbent's willingness to pay is always higher when the alternative is entrepreneurial market entry, as incumbents have an interest in reducing the number of competitors to keep profits high. Moreover, comparing Eqs. (5) and (8) gives we ðpÞ−wie ðpÞ ¼

ðτ−t Þ ½ΓðpÞ þ BðID Þ−ID −IR −½π ðiÞ−πðeÞ: ð1−τÞ

ð9Þ

The valuation differential we(p) − wie(p) in Eq. (9) can be positive or negative. Clearly, we(p) − wie(p) = − [π(i) − π(e)] b 0 holds when there is no tax disadvantage of a sale and t = τ. On the other hand, we assume that the sales tax rate τ is sufficiently high so that we(p) − wie(p) N 0 will hold when the profit tax rate t equals zero and the tax advantage of entrepreneurial market entry is therefore at a maximum. Fig. 2 illustrates this trade-off. For low profit tax rates t ∈ [0, tL) the tax advantage of market entry dominates and we(p) N wie(p). When the profit tax rises and reaches the critical value tL, the higher profits generated in the more concentrated market will overcompensate the tax disadvantage of a sale. Hence, for t ∈ [tL, τ) an incumbent's willingness to pay to deter market entry exceeds the entrepreneur's reservation price and we(p) ≤ wie(p). As shown in Fig. 2, the entrepreneur will thus enter the market when this confers a sufficiently large tax advantage, t ∈ [0, tL). When the tax advantage of entry is sufficiently small, t ∈ [tL, τ), a sale occurs where one incumbent will bid the entrepreneur's reservation price we(p) for the developed innovation, which is then accepted by the entrepreneur. Rival incumbents will not challenge such a bid since their competitive valuation is below the entrepreneur's reservation price. The latter follows from comparing we(p) in Eq. (5) with wii(p) in Eq. (7). This gives we ðpÞ−wii ðpÞ ¼

ðτ−t Þ ½ΓðpÞ þ BðI D Þ−ID −IR N0; ð1−τÞ

ð10Þ

which can be signed from Assumption 1.27 We can thus summarize: Lemma 1. Suppose that the tax rate τ when selling is sufficiently high so that we(p) N wie(p) holds at t = 0. Then there exists a profit tax tL such that the entrepreneur enters the market with a developed innovation for low profit taxes t ∈ [0, tL), whereas she sells the innovation at  the reservation price S2 ¼ we ðpÞ for high profit taxes, t ∈ [tL, τ). 27 Note that the gains from an entry deterring acquisition in this regime are unevenly distributed among incumbents, as the acquiring incumbent bears the cost of the entry deterrence while the other firms can free-ride on the acquisition. This raises the possibility of coordination failures among incumbents, if wie(p) N we(p) N wii(p). We assume, however, that incumbents are able to coordinate by randomly selecting one firm to buy the innovation. Since wie(p) N we(p), this selected firm will then find it profitable to purchase the innovation to prevent entrepreneurial market entry.

20

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

Fig. 2. Market entry vs. project sale in Stage 2.

3.2.2. Investing in developing the innovation If the entrepreneur's innovation project is successful in Period I of Stage 2, she can further develop the innovation in Period II. Specifically, she will invest ID to achieve an additional fixed cost reduction of B(ID). The investment costs ID are always tax-deductible from the proceeds of the developed innovation. If the entrepreneur chooses to enter in Period III, her objective function is (1 − t)[Γ(p) + B(ID) − ID] + tIR, where the last term captures the tax deduction that is available for the research investment in Period I. Moreover, we know from Eq. (5) that the net gain from developing the innovation is the same when the entrepreneur sells instead in Period III. For either commercialization mode, the optimal investment level in Period II is then implicitly defined from maxID : ð1−t Þ½ΓðpÞ þ BðID Þ−ID  þ tIR

⇒

′ 

B ID

where the n incumbents simultaneously post bids and the entrepreneur either accepts or rejects these bids. 3.3.1.1. The entrepreneur's reservation price. To determine the entrepreneur's reservation price in Stage 1, we need to keep in mind that there are two options for the entrepreneur, if the project turns out to be successful in Stage 2. Let us start with the case where the entrepreneur will enter the market, if successful. Using Assumption 2 and defining Δ1e (S1) as the entrepreneur's net gain from selling the invention at price S1 in Stage 1, as compared to keeping the project and entering the market in Stage 2, gives   1 1 1 Δe S ¼ ð1−τÞS þ τIR |fflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflffl}

Net profit from sale in Stage 1

¼ 1:

    −p ð1−t Þ Γ ðpÞ þ B ID −ID þ tIR |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}

:

Expected net profit from entry in Stage 2

ð11Þ

Hence the investment activity in this stage is carried to the point where the marginal fixed cost savings from developing the innovation equal the marginal investment cost. 3.3. Stage 1: research 3.3.1. Commercializing an uncertain research project In Stage 1, the success of the entrepreneur's research project is still uncertain. In Period II of this stage, the entrepreneur can decide on whether to keep the innovation until its success is revealed in Period I of Stage 2 (and then either sell, or enter the market), or whether to directly sell the uncertain innovation project at this early stage. We assume that the entrepreneur and the incumbents are symmetrically informed about the success probability of the project, and will both base their valuations on the same level of p.28 The commercialization process under uncertainty is again depicted as an auction 28 In the set-up of our model, information asymmetries could work in either direction. While the entrepreneur knows the technical properties of her invention better than incumbents do, incumbents may have more experience in evaluating the likelihood that the innovation succeeds in the market. Given this ambiguity, assuming that the information about p is symmetric represents an intermediate case that has the obvious benefit of analytical simplicity. For an analysis of a project sale where the entrepreneur has an informational advantage, see Norbäck et al. (2010).

Note that the research investment IR (taken in Period I of Stage 1) is sunk at this stage, whereas the investment in development ID (to be taken in Stage 2) is not. Moreover, and importantly, the outlays for the research investment IR are tax-deductible only with probability p when no sale occurs in Stage 1. This is because the entrepreneur will not achieve positive income when she keeps the project until Stage 2, and the project fails. The reservation price for a sale in Stage 1 is defined as v e = min S 1 , s.t. Δ 1e (S 1 ) ≥ 0. Solving for Δ 1e (S 1 ) = 0, and thus equating the net profits from the sale in Stage 1 with the expected net profits from market entry in Stage 2, gives the minimum price at which the entrepreneur is willing to sell: ve ¼

    ðτ−pt Þ pð1−t Þ  τ  ΓðpÞ þ B ID −ID − I ¼ pwe ðpÞ− pID þ ð1−pÞ I : ð1−τ Þ ð1−τ Þ R ð1−τ Þ R

ð12Þ The reservation price ve in Eq. (12) differs from the reservation price we in Stage 2 [Eq. (5)] in several respects. First, the success of the project is still uncertain at this early stage. Second, the entrepreneur's minimum sales price in Stage 1 is reduced, relative to the minimum sales price in Stage 2, because the (expected) costs of the development innovation ID do not have to be incurred under the early sale. Finally, and importantly, ve is reduced relative to we because the initial research investment IR is tax-deductible with certainty under the early sale, whereas it cannot

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31 Table 2 Equilibrium regimes of project commercialization. Regime

Definition

Commercialization

Acquisition price

R1

ve(p) N vie(p) N vii(p) and we N wie ve(p) N vie(p) N vii(p) and wie N we vie(p) N ve(p) N vii(p)

Market entry

–

R2.2 R2.1

Sale in Stage 2 (at reservation price) Sale in Stage 1 (at reservation price) Sale in Stage 1 (at competitive valuation)

vie(p) N vii(p) N ve(p)

R3

2

¼ we ðpÞ

1

¼ ve ðpÞ

1

¼ vii ðpÞ

S

S S

n h  i o 2 −p ð1−τÞ S −ID þ τIR |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}

Net profit from sale in Stage 1



;

Expected net profit from sale in Stage 2

where the tax-deductibility of the initial research investment IR in Stage 2 occurs again only with probability p. Solving for Δ1e (S1) = 0 and substituting the equilibrium sales price in Stage 2 [Eq. (5)] confirms that the entrepreneur's reservation price in Stage 1 is again given by Eq. (12). 3.3.1.2. Incumbents' valuations. The incumbents' valuation in Stage 1 will again depend on whether the alternative owner of the asset is another incumbent or the entrepreneur. Starting with the case where another incumbent would be the alternative acquirer of the research project in Stage 1 yields a net gain from acquiring the entrepreneur's invention at a price S1 equal to Δ1ii(S1) = (1 − ti)[pΓ(p) − S1]. Solving for vii ≡ max S1, s.t. Δ1im(S1) ≥ 0 gives an incumbent's competitive valuation in Stage 1 equal to    vii ¼ pΓðpÞ ¼ p wii ðpÞ−B ID ;

ð13Þ

where wii is given in Eq. (7). In comparison to the competitive valuation in Stage 2, the incumbent's willingness to pay for an uncertain project is reduced in proportion to the probability of failure. Moreover, when contemplating the acquisition of an innovation project in Stage 1, an incumbent takes into account that an early acquisition will not provide the additional cost savings B(ID) that would result from the entrepreneur's further development of a successful innovation in Stage 2. Furthermore, Δ1ie(S1) is the incumbent's net gain of buying in Stage 1 at price S1, if otherwise the entrepreneur will keep the innovation project in Stage 1, but then either enters the market or sells a successful developed innovation in Stage 2. Making use of Lemma 1 and again using the net gain function, we can express an incumbent's takeover valuation in Stage 1 as follows: ( vie ðpÞ ¼



pwie ðpÞ−pB ID 

pwe ðpÞ−pB ID

L

if t∈½0; t Þ and we ðpÞNwie ðpÞ L

if t∈½t ; τÞ and we ðpÞ≤wie ðpÞ

:

Then turn to the case where the entrepreneur would sell in Stage 2, which arises when we(p) − wie(p) b 0 holds in Eq. (9). The net gain from   o

n h  i acquiring now becomes Δ1ie S1 ¼ 1−t i p S2 −B ID −S1 . The first term in the curly brackets reflects the expected savings from not having to pay the acquisition price for a developed innovation in Stage 2,29 but net of the additional fixed cost savings that would arise from the further development of the innovation by the entrepreneur.

be deducted when the project fails in Stage 2 (which occurs with probability 1 − p). What is then the reservation price in Stage 1, if the entrepreneur could instead sell the successful, developed project in Stage 2? In this case the net gain Δ1e (S1) becomes   1 1 1 Δe S ¼ ð1−τ ÞS þ τI R |fflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflffl}

21

ð14Þ

To interpret Eq. (14), let us start with the case where the entrepreneur would enter in Stage 2. This case arises when we(p) − wie(p) N 0 holds in Eq. (9). In this case the Stage 1 valuation of the acquirer is given by the expected fixed cost savings from the innovation, plus the expected gain from avoiding increased competition through the entrepreneur's market entry. The net gain for an acquiring incumbent is therefore Δ1ie(S1) = (1 − t i){p[π(i) − π(e) + Γ(p)] − S1}. Solving for Δ1ie(S1) = 0 and using Eq. (8) gives the expression in the upper line of Eq. (14).

Solving for Δ1ie(S1) = 0 and using S2 ¼ we ðpÞ from Lemma 1 gives the expression in the lower line of Eq. (14). Comparing vie(p) in Eq. (14) with vii(p) in Eq. (13) shows that vie(p) N vii(p) since wii(p) b wie(p) from Eqs. (7) and (8) from the market power effect. Thus, just as in Stage 2, the takeover valuation always exceeds the competitive valuation. However, in contrast to Stage 2, the entrepreneur's reservation price ve(p) need not exceed the incumbents' competitive valuation, vii(p). Essentially, this arises due to the entrepreneur's limited loss offset opportunities if she keeps the project [see Eq. (12)]. Similar to Stage 2, the takeover valuation vie(p) may or may not exceed the reservation price ve(p). In our overall game, there are thus four different regimes that we need to consider: entrepreneurial market entry (Regime 1); a sale at the entrepreneur's reservation price, which may occur either in Stage 1 (Regime 2.1) or in Stage 2 (Regime 2.2); and a sale at the incumbents' competitive valuation, which always occurs in Stage 1 (Regime 3). These regimes and their properties are summarized in Table 2. In Regimes 1 and 2.2, the entrepreneur's reservation price associated with keeping the innovation project in Stage 1 exceeds the incumbents' valuations. If the project is successful, the entrepreneur will then either enter the market in Stage 2 (Regime 1), or sell the developed innovation at this later stage at her reservation price (Regime 2.2). In Regime 2.1, in contrast, the entrepreneur sells her innovation project directly in Stage 1, with the sales price being determined by her reservation price in this stage. One incumbent then bids ve(p) (which is below her takeover valuation vie), and the entrepreneur accepts (as she is indifferent between accepting and not accepting). Since incumbents' competitive valuations (their value of challenging a rival's bid) are below the reservation price (ve N vii), bidding competition does not arise in this regime. In Regime 3, the innovation project is also sold directly in Stage 1. However, since the competitive valuation exceeds the reservation price in this regime (vii N ve), bidding competition arises and the sales price will be driven up to the competitive valuation of incumbents. 3.3.2. Project choice In this section, we solve for the equilibrium project selected by the entrepreneur, given that she anticipates the mode of commercialization as summarized in Table 2. In each regime, the entrepreneur chooses the research project that maximizes her expected after-tax profit. Using the information from Table 2, the expected net rewards for the entrepreneur, denoted by Ω(p), are:    8  p ð1−t Þ Γ ðpÞ þ B ID −ID þ tIR > >  > > pfð1−τ Þ½we ðpÞ −ID  þ τIR g > > > |fflfflffl{zfflfflffl} > > < ¼S2 ð 1−τ Þve ðpÞ þ τIR ΩðpÞ ¼ |fflffl{zfflffl} > > > > ¼S1 > > > ð1−τÞvii ðpÞ þ τIR > > |fflffl{zfflffl} :

in R1 in R2:2 in R2:1 :

ð15Þ

in R3

¼S1

In Regime 1 the net reward is given by the entrepreneur's expected after-tax profits from entering with a developed innovation, which includes the expected tax deduction from the research investment IR (which is sunk when the project choice is taken). To obtain the expected net reward in Regime 2.2, we use the fact that the entrepreneur sells the 29 We assume, for simplicity, that the same incumbent is chosen in Stages 1 and 2 to buy the innovation. See footnote 27.

22

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31 

developed innovation in Stage 2 at her reservation price, S2 ¼ we ðpÞ. In Regime 2.1, the entrepreneur sells her project in Stage 1 at her res ervation price, S1 ¼ ve ðpÞ, whereas in Regime 3 the sales price equals  the competitive valuation of incumbents, S1 ¼ vii ðpÞ. In the sales Regimes 2 and 3, the net reward is the sales price net of the sales tax, plus the tax deduction on the initial investment outlay IR. Moreover, when the sale occurs in Stage 1 (Regimes 2.1 and 3), the entrepreneur obtains a certain tax reduction since she can write off the initial research investment from the certain income generated by the sales price.

3.3.2.1. Regime 3. To derive the equilibrium project choices we start with Regime 3. Substituting Eq. (13) into (15) implies that the entrepreneur will choose the project p∗S = argmaxp[(1 − τ)pΓ(p) + τIR], where the

project choice is denoted by the subscript S (for ‘sale’). The associated first-order condition is: 

 

Γ pS þ pS Γ′ pS ¼ 0:

ð16Þ

In this regime the entrepreneur will maximize the sales price by choosing the project that maximizes the Stage 1 value to incumbents. Since incumbents can fully deduct the investment costs from their taxable profits, the corporation tax does not distort the project choice in this regime. Hence the entrepreneur chooses the project that maximizes the expected fixed cost reduction from the research investment IR. This is shown by the point S in Fig. 3. 3.3.2.2. Regimes 1 and 2. Next, we consider the optimal project choice in Regime 1. As seen in Eq. (15), the net reward is based on the expected

Fig. 3. Optimal project choice.

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

after-tax profits that the entrepreneur can earn from market entry. Hence the net reward is maximized by incorporating the loss offset provisions that apply to the entrepreneur when she enters the market. The associated first-order condition for the optimal project choice is: 

 ′ 

Γ pE þ pE Γ pE ¼ −

   t I − B ID −ID ; ð1−t Þ R

ð17Þ

where the project choice is now denoted by the subscript E (for ‘entry’). The RHS of Eq. (17) now incorporates two negative terms at the entrepreneur's optimally chosen project p∗E, implying that the entrepreneur chooses a project with a lower level of risk than is true in Regime 3. The first term on the RHS arises because the entrepreneur cannot deduct her research investment costs IR from tax in case of project failure. As a result, she will trade off a lower probability of being left with nondeductible research expenditures in case of project failure against a reduced net profit when the project succeeds. Moreover, as the second term on the RHS of Eq. (17) shows, the entrepreneur can only develop the innovation further in Stage 2, and reap the additional profit B(I∗D) − I∗D, when the project succeeds. This gives a further incentive to the entrepreneur to choose a safer project than in Regime 3, where the net return from the additional development investment is not incorporated in the sales price. Substituting Eqs. (5) and (12) into Eq. (15) shows that Eq. (17) is also the first-order condition for the project choice in Regimes 2.1 and 2.2. Intuitively, in these regimes the sales price is determined by the entrepreneur's reservation price (either in Stage 1 or in Stage 2). Since the reservation price reflects the entrepreneur's opportunity cost of forgoing market entry, it incorporates both the limited loss offset opportunities in case of project failure and the extra benefit from the development innovation in Stage 2. Both of these effects lead to a safer project choice in the entrepreneur's optimum, just as in the market entry Regime 1. The expected fixed cost reduction that results from the project choices in Regimes 1 and 2 is thus unambiguously lower than in Regime 3, as shown by the point E in Fig. 3. Our analysis in this section is summarized in: Proposition 1. The entrepreneur chooses a more risky project p∗S b p∗E when the project is sold in Stage 1 at the competitive valuation of incumbents (Regime 3) than when the entrepreneur enters the market (Regime 1), or when the project is sold at the entrepreneur's reservation price (Regimes 2.1 and 2.2). 4. The effects of tax policy 4.1. Effects on project choice and commercialization mode In this section, we analyze how the system of taxing and subsidizing entrepreneurial incomes affects the mode of commercialization and the entrepreneur's project choice. We focus on exogenous variations in the effective rate of corporate profit taxation t that the entrepreneur faces in case of market entry. In this analysis, we hold constant the personal income tax τ, which is levied in case of project sale. In what follows, we will assume that all regimes in Table 2 emerge for the range of permissible profits taxes t ∈ [0,τ) (see Lemma 1). We thus examine which regimes and project choices are associated with different ranges of the profit tax t. Our results are summarized in the following proposition, which is proved below. Proposition 2. Suppose that all regimes in Table 2 are represented when t ∈ [0,τ). Then, in equilibrium, the following holds: (a) When the profit tax rate is low t ∈ [0, tC), the entrepreneur chooses the ‘safe’ innovation project p∗E. (i) For very low profit taxes t ∈ [0, tA), the entrepreneur keeps the project in Stage 1 and – if successful – enters the market with a developed innovation in Stage 2 (Regime 1).

23

(ii) For low profit taxes t ∈ [t A, tB), the entrepreneur sells  the project in Stage 1, receiving her reservation price S1 ¼ 

ve pE (Regime 2.1). (iii) For intermediate profit taxes t ∈ [tB, tC), the entrepreneur keeps the project in Stage 1 and – if successful – sells the developed innovation to an incumbent in Stage 2, at the reser

 vation price S2 ¼ we pE (Regime 2.2). (b) When the profit tax rate is high t ∈ [tC, τ], the entrepreneur chooses the ‘risky’ innovation project p∗S b p∗E. She then sells this project to an incumbent under bidding competition in Stage 1,

 receiving the sales price S1 ¼ vii pS (Regime 3). Proof. Part (b): To prove part (b), it suffices to compare the competitive valuation of incumbents vii(p) to the reservation price of the entrepreneur, ve(p). From Table 2 and Eqs. (16) and (17) it follows that the project p∗E must maximize the entrepreneur's reservation price ve(p), whereas the project p∗S must maximize the competitive valuation vii(p). Using Eq. (6) and applying the envelope theorem, we first note that the reservation price in Stage 1 must decrease in the profit tax rate:

   dve ðpE Þ t  

 ∂we ðpE Þ  ¼ pE Γ pE −IR þ B ID −I D b0: ¼ −pE dt 1−τ ∂t

ð18Þ

In contrast, the competitive valuation vii(p∗S) is unaffected by taxes since the project p∗S is invariant to the profit tax rate from Eq. (16). As shown in Fig. 4(i) there must then exist a unique cut-off tax level tC ∈ (0, τ) at which ve(p∗E) = vii(p∗S). Thus, Regime 3 with vii(p∗S) N ve(p∗E) must arise for high taxes, t ∈ [tC, τ). Part (a): To prove part (a) of Proposition 2, we first focus on the entrepreneur's optimal project choice, as a function of t. From our previous discussion we know that the first-order condition Eq. (17) for the optimal project choice p∗E applies for all t ≤ t C (i.e., in Regimes 1 and 2). Implicitly differentiating Eq. (17) shows that dp∗E/dt N 0. Intuitively, throughout Regimes 1 and 2 a safer project is chosen when t rises because a higher profit tax rate increases the potential tax losses from a failed investment project. The entrepreneur's optimal project choice as a function of t is shown in Fig. 4(ii). Next, we compare the takeover valuation of incumbents vie(p) to the reservation price of the entrepreneur, ve(p). Since the entrepreneur always chooses p∗E for t ∈ [0, tC), we can evaluate both valuations at p∗E. Using Eqs. (12) and (14) gives 8 τ  

 

>  

 < pE ½we −wie  þ pE B ID −ID − 1−pE ð1−τÞ IR ; ve pE −vie pE ¼ 



τ     > : I ; pE B I D −I D − 1−pE ð1−τÞ R

L

t∈½0; t Þ; L

C

t∈½t ; t Þ:

ð19Þ The first term in the upper line of Eq. (19) can be viewed as the expected net value of entry in Stage 2, which arises from the tax advantage of market entry [see Fig. 2, which is reproduced in Fig. 5(i)]. It is straightforward to derive from Eq. (9) that this term is decreasing in t; this is shown in Fig. 5(ii). The two following terms represent the net expected benefit of keeping the project p∗E in Stage 1: they reflect the trade-off between the opportunity to develop a successful innovation when succeeding and the cost of losing the tax benefit when failing. Since p∗E is rising in t [see Fig. 4(ii)], this term is increasing in the profit tax, as shown in Fig. 5(ii). Intuitively, when the project is more likely to succeed, this will not only increase the expected benefit from development, but also reduce the expected cost from lost tax deductions. From Lemma 1, we know that we(p∗E) = wie(p∗E) holds for t = t L. Hence, as shown by the lower line of Eq. (19), only the net expected benefit of keeping the innovation remains when taxes exceed t L.

24

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

Fig. 4. The effects of tax policy on project choice.

To determine the sequence of regimes as t is continuously raised, we differentiate ve(p∗E) − vie(p∗E) with respect to t. This gives:   8

> > B ID −ID þ t IR −½π ðiÞ−π ðeÞ dpE ;   d½ve ðpE Þ−vie ðpE Þ < 1−t dt   ¼

τ dpE > dt   > : B ID −ID þ I N0 ð1−τ Þ R dt

L

t∈½0; t Þ; L C t∈½t ; t Þ:

ð20Þ The expression in the lower line of Eq. (20) is always positive. Hence our assumption that all regimes must be represented translates into ve(p∗E) − vie(p∗E) being U-shaped, implying that the upper line in Eq. (20) must be negative. This in turn requires that the market power effect of an acquisition must be sufficiently strong, in relation to the remaining effects. Taking this as given, Fig. 5(iii) then illustrates the sequence of regimes as t is continuously increased. Our proof thus shows that Regime 1 appears for low tax rates t ∈ [0, tA), where the tax advantage of market entry is highest. As t increases, the net value of market entry is reduced. For t ∈ [tA, tB), the value of ve − vie turns negative in Fig. 5(iii) and Regime 2.1 emerges where the entrepreneurs sells her project in Stage 1. For this range of profit taxes, the lower line of Eq. (19) is negative, as the expected value of developing the innovation is less than the expected tax loss arising from the possibility of failure. As t continues to rise, and the equilibrium project has a higher success probability, the value of keeping the project unambiguously increases. For t ∈ [tB, tC) the value of ve − vie turns again positive in Fig. 5(iii) and Regime 2.2 emerges, with

the entrepreneur keeping her project and (if successful) selling the developed innovation in Stage 2. Finally, increasing the tax beyond t C reduces the tax advantage of market entry sufficiently to lead to a Regime 3 equilibrium with project sale in Stage 1 under bidding competition. We emphasize that, as t increases, a sequence of equilibria in both Regimes 2.1 and 2.2 will only emerge, if the exogenous parameters ensure that the net value of keeping the project in Stage 1 (to sell it in Stage 2) is negative for t ∈ [t A, t B), but positive for t ∈ [t B, t C). This requires that the counteracting terms in the lower line of Eq. (19) are roughly similar, so that the changing level of the profit tax, and hence of the research project's success probability in equilibrium, determines the sign of the net effect. If instead the benefit of further developing the project is much higher (much lower) than the tax loss in case of project failure, then only Regime 2.2 (only Regime 2.1) will occur. In other words, tax policy has only a limited potential to determine whether, within Regime 2, the project sale will occur in the early or in the late stage. Moreover, we can determine more precisely under which conditions an equilibrium with bidding competition in Regime 3 arises. For this purpose we evaluate ve(p∗E) − vii(p∗S) at the highest possible profit tax in our setting, t = τ (see Assumption 1). This gives:

 

   

    

   

ve pE −vii pS jt¼τ ¼ pE B I D −I D − pS Γ pS −pE Γ pE − 1−pE

τ I : ð1−τ Þ R

ð21Þ

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

25

Fig. 5. The effects of tax policy on the commercialization mode.

The entrepreneur's reservation value in Stage 1, ve(p∗E), incorporates the net value of the development innovation ID (the positive first term on the RHS). However, it also leads to lower expected cost savings from the research investment (see Fig. 3); this is the negative second term on the RHS. Finally, the negative third effect on the RHS is the expected (grossed-up) tax loss that the entrepreneur suffers due to the possibility of project failure. The sum of all effects on the RHS will thus be negative, and ve(p∗E)|t = τ b vii(p∗S)|t = τ holds, when the tax loss effect is large, relative to the difference in the combined cost savings arising in the research and development stages. Other things equal, this will be the case when the tax rate τ (which equals t in this case) is sufficiently high. In this case, the entrepreneur will thus prefer to sell at the competitive valuation of incumbents, leading to Regime 3.

To summarize, we have studied the effects of government policies that reduce the effective profit taxation of small businesses below the tax rate that applies for the sale of an innovative project. If the preferential tax and subsidy treatment of entrepreneurial profit income is very strong (i.e., t is very low), this will induce market entry by the entrepreneur (Regime 1). For a somewhat less discriminatory treatment, the project will be sold to an incumbent firm, but the sales price is determined by the entrepreneur's opportunity cost of forgone market entry (Regime 2). In all these regimes the entrepreneur will choose a comparably safe project, and thus accept a limited cost reduction in case of success, in order to maximize the sales price that she can obtain from the acquiring incumbent. Only when the effective tax rate for entrepreneurial profits approaches the tax rate for a project sale (Regime 3) will the

26

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

sales price be determined by the bidding competition of incumbents. In this case the entrepreneur will maximize her expected return by choosing a discretely more risky project. Which of these project choices and commercialization modes is preferred from the perspective of society's overall welfare is the subject of the following section. 4.2. Welfare effects of tax policy In this section we consider some of the welfare implications that tax policy has in our setting. Hence we ask which of the regimes, as characterized above, is preferred from a social welfare perspective under different economic conditions. We emphasize that this welfare analysis can only be a partial one. In particular, our framework does not allow us to endogenize the tax rates levied on either the sale of the entrepreneurial start-up firm (τ), or the effective tax rate on the entrepreneur's profits under market entry (t). We assume, however, that the flexibility of the tax and subsidy system is sufficiently large so that each of the different regimes can be induced by tax policy, if desired. In our analysis, we measure welfare as the sum of aggregate industry profits, consumer surplus and tax revenue. Note that the expected income loss for the entrepreneur that arises from the inability to deduct the research investment IR from tax in case the project fails is fully compensated from society's point of view by higher expected tax receipts. Hence the sum of tax revenue and the entrepreneur's net reward equals the gross-of-tax reward to the research and – in some regimes – development innovations, net of all investment costs. Moreover, the sales price S (in the regimes where the project is sold) represents a transfer from the incumbent to the entrepreneur, which does not affect aggregate welfare. This yields the following expressions for expected welfare in the different regimes30: 8  

 

  pE Γ pE −IR þ pE B I D −ID þ Ψ þ pE ξ > >  

 <  

pE Γ pE −IR þ pE B I D −ID þ Ψ E ½W  ¼   > pE Γ pE −IR þ Ψ > :  

pS Γ pS −IR þ Ψ

in in in in

R1 R2:2 R2:1 R3:

ð22Þ

Here, Ψ ≡ nπ(i) + CS(i) is the sum of incumbents' profits and consumer surplus in the case without market entry by the entrepreneur and ξ ≡ n½πðeÞ−πðiÞ þ CSðeÞ−CSðiÞ≶0 is the net externality arising from the entrepreneur's market entry in Regime 1, which increases the equilibrium number of firms from n to (n + 1). This causes a decline in the incumbents' aggregate profits given by n[π(e) − π(i)] b 0 and an increase in consumer surplus given by CS(e) – CS(i) N 0. Eq. (22) shows that there are three effects which cause expected welfare to differ in the various regimes: (i) the net externality from entrepreneurial market entry; (ii) the net benefit from further developing the innovation in Stage 2; and (iii) the efficiency of the choice of a research project in Stage 1. The net externality for society from the entrepreneur's market entry can be determined from the model of quantity competition in differentiated products, as introduced in Section 3.1 [see Eq. (3)]. The consumer surplus with or without entrepreneurial market entry (m = e,i) is given by: 

2 

2 i 1h  CSðmÞ ¼ aQ ðmÞ− nðmÞ q ðmÞ þ γnðmÞ½nðmÞ−1 q ðmÞ  2     −nðmÞ a−q ðmÞ−γðn−1Þq ðmÞ q ðmÞ ∀ m∈fi; eg;

30

ð23Þ

Note that the entrepreneur's entry costs G are not subtracted from the welfare expression for Regime 1. This is a result of Assumption 2, by which the market entry of the entrepreneur creates a rent that is (almost) equal to the entry cost.

where Q = ∑ j qj and n(m)[n(m) − 1]/2 is the number of unique crossproduct terms q∗j q∗k with j ≠ k in Eq. (3). From Eq. (23) and (4), the net externality ξ becomes h 1 2 ξ ¼ ða−cÞ 2

−n2 γ3 −n2 γ2 þ nγ 3 −3nγ2 þ γ 2 −4γ þ 4 2

2

ðnγ þ 2Þ ð−γ þ nγ þ 2Þ

i ≷0;

ð24Þ

where n is the number of incumbent firms. From (24) it is straightforward to infer that ξ N 0 for γ = 0, but ξ b 0 for γ = 1. Moreover, since ξ is continuous in γ, there must exist a critical level of product differentiation 0 b γc b 1 such that ξ = 0. Hence, if the goods produced by the incumbents and the entrepreneur are perfect substitutes (γ = 1), then entrepreneurial market entry has a negative net effect on surplus in the industry, because the reduction in producer surplus dominates the rise in consumer surplus. This is due to the additional fixed cost incurred by the new entrant, which leads to rising average costs in the industry (Mankiw and Whinston, 1986). If all firms are local monopolies, however (γ = 0), then the increase in consumer surplus will dominate, as consumers value the addition of a new product. The welfare trade-offs that exist between the different regimes are then directly inferred from Eq. (22). We start with comparing Regimes 1 and 2.2. Both regimes incorporate the value added from the entrepreneur's further development of the project, but both regimes also feature a project choice that is distorted by tax considerations [Eq. (17)].31 Hence the choice between these two regimes will primarily depend on the sign of the net externality from entrepreneurial market entry in Eq. (24). Turning to the comparison of Regime 2.2 and Regime 3, the trade-off is that Regime 2.2 incorporates the gains from developing the innovation, whereas Regime 3 maximizes the expected cost savings from the initial research project [see Eq. (16)]. Hence the choice between these two regimes depends on the relative importance of these two counteracting effects. Finally, it is obvious that Regime 3 dominates Regime 2.1, as these regimes differ only in the choice of the initial research project, and this is more efficient in Regime 3. We can then summarize that the socially optimal tax policy will induce a Regime 1 equilibrium, if the net externality from market entry and the net value of further developing the project are both large. An equilibrium in Regime 2.2 will be socially optimal when the net externality from market entry is negative, but the value of developing the innovation is large. A Regime 3 equilibrium will be socially preferred when the net externality from market entry and the value of development are both small, and are dominated by the welfare gain from choosing an efficient research project in Stage 1. Inducing an equilibrium in Regime 2.1 will never be socially optimal. We illustrate these results with the help of Fig. 6, which compares the expected welfare levels in the different regimes as a function of the research project p. The figure is drawn for the case where the net externality from market entry, given by ξ in Eq. (24), is positive. As shown in the upper panel of the figure, the expected welfare curve in Regime 1, E(W)R1, thus lies above the welfare curve in Regime 2.2. Moreover, since Regime 2.2 incorporates the positive net benefit from project development, this curve in turn lies above the (common) expected welfare curve in Regimes 3 and 2.1. The lower panel of the figure depicts the regime-specific project choices. Starting with Regime 3, we know from Eq. (16) that the choice of research project is not distorted in this regime, and the maximum of the expected welfare curve E(W)R3 in point S can therefore be reached. In Regime 2.2, the success probability of the socially optimal research project is higher than in Regime 3, because the additional benefit from development is only obtained when the project succeeds at the beginning of Stage 2.32 Hence the 31 Note that this distortion is quantitatively more important in Regime 2.2, because the profit tax rate is higher in this regime (see Proposition 2). 32 This is seen by differentiating the regime-specific welfare expression in Eq. (22) with respect to p.

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

27

Fig. 6. Comparing welfare across regimes.

maximum of E(W)R2.2 lies in point L, which is to the right of S. In Regime 1, the maximum of E(W)R1 lies still further to the right, in point W, because the socially optimal project in this regime also considers that the positive externality from market entry is generated only when the project is successful. In Regimes 1 and 2, however, the maxima of the expected welfare curves cannot be reached, because project choices are distorted by the entrepreneur's desire to minimize the expected tax loss in case of project failure. Moreover, we know from Eq. (17) that this distortion is the larger, the higher is the tax rate in the respective regime. In Regime 2.2, the minimal tax rate is tB [see Proposition 2(iii)], which leads to point B on the welfare curve E(W)R2.2 in the upper panel. Similarly, in Regime 2.1, the minimal tax rate is tA [see Proposition 2(ii)], resulting in point A on the welfare curve for this regime. The equilibrium in Regime 2.1 (point A) is dominated by Regime 3 (point S), however, whereas the outcome in Regime 2.2 is dominated by the equilibrium in

Regime 1. The latter must be true because Regime 1 incorporates the positive externality from market entry and it implies a lower tax rate from Proposition 2(i), and hence a smaller distortion of the efficient project choice. The tax rate in Regime 1 is denoted by tmin in the lower panel of the figure, resulting in point E on the welfare curve E(W)R1 in the upper panel.33 In the example shown in Fig. 6, the equilibrium in Regime 1 (point E) is dominated by the equilibrium in Regime 3 (point S), where the project choice is undistorted. This example therefore shows that an equilibrium where the project is sold in Stage 1 under bidding competition by

33 The figure thus assumes that a positive effective tax rate must be levied on the entrepreneurial profits in Regime 1, and hence the project choice in the research stage is indeed distorted. This condition will almost always be met in practice, due to the government's revenue needs and in order to avoid an overly discriminatory tax treatment of different firms.

28

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

incumbents can dominate all other regimes, even when there is a net value to entrepreneurial development in Stage 2 and when the net externality from entrepreneurial market entry is positive. Finally, we note that the argument for promoting entrepreneurial market entry is strengthened when there are market failures in labor markets and countries suffer from unemployment. Indeed, one of the policy objectives of a favorable tax treatment of new, small firms is the belief that they create additional jobs.34 It is straightforward to incorporate this argument into our analysis when we assume that aggregate employment in the industry is monotonously rising in aggregate output. The market entry of the entrepreneur will not displace incumbent firms, because the entry costs G are sunk for them. Hence, entrepreneurial market entry will increase aggregate output in the industry.35 Moreover, given macroeconomic unemployment, a higher employment level in the innovative industry is unlikely to be fully compensated by employment reductions in other parts of the economy. Summing up, a policy of promoting market entry of small, innovative firms is likely to lead to higher output and employment in the innovative sector, but the welfare effects of this policy are positive only when goods are sufficiently differentiated. If the externalities arising from the entrepreneur's market entry are positive and sufficiently large, then the optimal tax policy should indeed favor market entry by granting tax reductions and subsidized loans to entrepreneurs. The costs of an entry-promoting policy lie, however, in the distortion of the entrepreneur's project choice, in the direction of inefficiently low levels of risk-taking. When this distortion is sufficiently large, our model points to a basic policy trade-off between the declared goals of promoting small, innovative businesses and the fostering of major, risky innovations. In this case, welfare may indeed be maximized by a tax policy that does not tax-discriminate between entrepreneurs who enter the market themselves, and those who sell their product or firm in a competitive bidding process to one of the incumbents. 5. Extensions and discussion In this section, we briefly discuss several extensions and modifications of our benchmark model.36 1 Variable cost savings inventions A first extension of our benchmark model is to allow for variable cost reductions caused by the entrepreneurial innovations, where the fall in variable costs (in case of success) is the higher, the more risky a project is. This extension implies that the product market competition in Stage 3 becomes asymmetric, with the possessor of the invention having a larger market share than her competitors. Nonacquiring incumbents thus see their product market profits decreasing when the innovation succeeds, and this fall in their profits will be the stronger the more risky the innovation is. This will increase the incumbents' willingness to pay for the innovation in Regime 3, where the sales price is determined by the bidding competition of incumbents. As a result, there is an added incentive for the entrepreneur to choose a more risky project in Regime 3, implying that the difference in the riskiness of the projects chosen in Regimes 1 and 2 versus Regime 3 is even greater than in our benchmark model. Another core difference to our benchmark model is that the riskiness of the research project, by affecting the possessor's variable production costs, will now have direct implications for consumer surplus in the market. In particular, consumers may now prefer a Regime 3 34

A substantial part of the net employment growth in the SME sector comes from a small number of high-growth firms, so-called ‘Gazelles’, which are typically young and are represented in all industries (Henrekson and Johansson, 2010). 35 This is easily verified for the Cournot model with differentiated outputs by multiplying the per-firm outputs in Eq. (4) for m = i,e with n and (n + 1), respectively. 36 A detailed analysis of the first two extensions covered in this section is found in the online appendix.

equilibrium over a Regime 1 equilibrium with entrepreneurial market entry, even if products are differentiated and entrepreneurial market entry would be favored in our benchmark model [i.e., if ξ N 0 in Eq. (24)]. This is because the more risky project choice will increase expected variable cost savings, and hence expected output. Overall, therefore, introducing variable cost savings makes a Regime 3 equilibrium with a project sale under bidding competition of incumbents more attractive, relative to our benchmark model. Hence this extension strengthens the case against a policy that promotes entrepreneurial market entry by means of low effective profit taxes. 2 Effort by the entrepreneur An important aspect of entrepreneurial innovation is that a substantial share of the initial investment may consist of effort put in by the entrepreneur. It is straightforward to incorporate this into our model by introducing a zero stage of the game, where the entrepreneur chooses an endogenous level of effort in order to generate a basic innovative idea. In this setting it can be shown that increased levels of taxation reduce the entrepreneur's effort to create innovative ideas. While the effort-reducing effects of taxes have been stressed in previous literature (e.g. Keuschnigg and Nielsen, 2004), this result is specified in our setting because different taxes affect entrepreneurial effort under different equilibrium modes of commercialization. In particular, higher corporate tax rates are effort-reducing when the equilibrium is in Regimes 1 and 2, whereas higher taxes on the project sale reduce effort when the equilibrium is in Regime 3. All other results of our analysis remain unaffected by this model extension. 3 Product innovation Our benchmark model assumes that the project chosen by the entrepreneur is a cost-reducing process innovation. It is conceptually straightforward, however, to alternatively consider a product innovation where the entrepreneur develops a new product in the first stage. The degree of innovation, and the risk associated with it, would then both rise in the distance that the new product has to existing ones as, for instance, in Salop's (1979) circular-city model of product differentiation. While this model would be more complex than the one used here, a concave expected return curve similar to the one in Fig. 3(i) could also be obtained in this case. All other features of our model would also be preserved in such a setting. In particular, imperfect loss offset would again cause the entrepreneur to choose a project with too little risk, which would here imply that a project is “too similar” to existing ones. 4 Multi-firm licensing Our benchmark model has assumed that the innovation can be used only by a single firm. We relax this assumption now and examine the possibility of multi-firm licensing. We thus assume that the owner of a successful innovation can make a take-it-or-leave-it offer to potential licensees, where we restrict the potential licensees to incumbents. In this situation the possessor of the successful innovation, either the entrepreneur or an incumbent acquiring the innovation, could generate a revenue for each licensee equal to the fixed cost saving Γ(p). Adding these extra revenues to the entrepreneur's and the incumbents' valuations and assuming, for simplicity, that there is no value added in the development stage, B(I∗D) − I∗D = 0, (and late sales in Stage 2 do not arise) gives: pð1−t Þ ðτ−pt Þ ðn þ 1ÞΓðpÞ− I ; ð1−τÞ ð1−τ Þ R vie ¼ p½πðiÞ−πðeÞ þ pnΓðpÞ; vii ¼ pnΓðpÞ:

ve ¼

ð25Þ

Note that the valuations in Eq. (25) consist of the same terms as in our benchmark analysis, but the value of the invention is scaled up by either n + 1 or n. This implies that the qualitative effects identified in our main analysis are still valid, but the magnitudes may change. A first consequence of introducing multi-firm licensing is then that, in comparison to our benchmark model, market entry becomes more attractive for the entrepreneur, relative to project sale. This is

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

because, if the entrepreneur enters the market, a maximum of (n + 1) firms can use the innovation, whereas the maximum is n under project sale. Secondly, the difference between the optimal risk level in Regimes 1 and 2 (p∗E) and the optimal project risk in Regime 3 (p∗S) will decrease when multi-firm licensing is an option. The first-order condition for a sale at the competitive valuation vii is unchanged from our benchmark analysis [Eq. (16)]. In Regimes 1 and 2, however, it now becomes: 

 ′ 

Γ pE þ pE Γ pE ¼ −

t I : ð1−t Þðn þ 1Þ R

ð26Þ

For sufficiently high levels of n, the right-hand side in Eq. (26) approaches zero, and hence the optimal project under market entry approaches the efficient one. This is because limited loss offset possibilities for the initial research investment IR lose in importance from the perspective of the entrepreneur when the revenues from a successful project increase. However, one might argue that if the potential value of the innovation increases due to the option of multi-firm licensing, the competition to obtain an innovation might equally increase, and thereby also the initial investment costs IR. This isolated effect would then again strengthen the importance of loss offset restrictions and increase the wedge between the efficient project and the project chosen under market entry. Moreover, there are several reasons why multi-firm licensing might not be an optimal strategy, even if the invention reduces only fixed costs. A first reason is that if the sale occurs before the resolution of uncertainty, no ‘product’ can be licensed. Secondly, if the licensee and the licensor need to undertake post-licensing investments, free riding problems will likely be increased if there are many licensees. Third, if asymmetric information is present, then multi-firm licensing increases the risk that potential buyers can claim that an idea was already known, thus expropriating innovators once they have disclosed their invention to them (see Anton and Yao, 1994). These are only some of the reasons why multi-firm licensing rarely occurs in practice. 5 Venture capital Our benchmark analysis has assumed that the entrepreneur is not able to obtain tax rebates for the investment costs incurred when she attempts to enter the market herself, but the project fails. These loss offset restrictions imply that the entrepreneur could gain from the cooperation with a financial partner, who has other income sources. In this case the initial investment cost could be deducted as long as the partner has sufficient positive income, and the distortive effect on project choice could thus be avoided. It is well known, however, that such financial partnerships are associated with severe adverse selection and moral hazard problems. Therefore, where financial partners exist, they almost always involve a highly specialized venture capitalist. In most countries, however, only few projects are backed by venture capital. Even in the United States, where the venture capital market is by far the largest, only 14,000 portfolio companies worldwide received venture capital over a 30-year period from 1975 to 2005 (Lerner, 2010). Total venture capital investments in the United States amounted to 28.8 billion US-$ in 2008, corresponding to 0.20% of this country's GDP.37 In Europe, the share of venture capital investments in GDP is much lower, equaling 0.07% of national GDP in Sweden, 0.05% in the United Kingdom and 0.03% of GDP in Germany (EFI, 2011, p. 19). Moreover, even for start-up firms backed by venture capital (VC), the possibility to deduct losses for tax purposes is far from complete. On the one hand, entrepreneurs usually retain a substantial ownership share in the VC-backed company, in order to mitigate the moral 37

See “Investment funding rose 5 percent in 2008”, VentureBeat, 18 February 2009.

29

hazard problems inherent in the relationship with the venture capitalist (see Keuschnigg and Nielsen, 2002, 2004). This implies that loss offset restrictions will still apply for the ownership share of the entrepreneur. In addition, some countries' tax laws also restrict the tax deduction of losses for venture capitalists. In Germany, for example, unused tax losses cannot be transferred when the venture capitalist sells her shares in the company — for example to another VC firm, which specializes in financing a different development phase of the start-up. These restrictive tax provisions are widely believed to hinder VC financing, but also the establishment and growth of innovative companies in Germany (EFI, 2011, p. 19). In sum, bringing venture capital financing into the picture and accounting for its relative importance will introduce some qualifications to our assumption that an entrepreneur producing for market entry is not able to obtain any tax rebates for losses incurred in the research phase. However, this extension will clearly not overturn our basic argument that loss offset restrictions are more severe when the entrepreneur produces for market entry, as opposed to selling her innovation to an incumbent firm. 6. Conclusion In this paper we have focused on two important decision margins of entrepreneurs that have received little analysis so far in a tax policy context. These are the entrepreneur's choice between innovations with different risk and return characteristics, and her decision of how to commercialize the innovation. We have shown that policies which tax entrepreneurial profits at strongly reduced effective profit tax rates – typically brought about by subsidized loans and guarantees that are confined to small enterprises – face a fundamental trade-off. They promote market entry by entrepreneurs, which is welfare-enhancing if goods in the innovative sector are sufficiently differentiated. Moreover, such policies ensure that the entrepreneur will undertake development investments that add value to her innovation once the initial research project has proven successful. At the same time, however, entrepreneurial market entry leads to research projects that involve suboptimally low levels of risk-taking. This is due to the risk-reducing effect of limited loss offset provisions, which are particularly important for small start-ups. In that sense, the comparative advantage of new, entrepreneurial enterprises to create major innovations is thus compromised by such policies. Our results can be used to evaluate existing policies to support entrepreneurship in the European Union and elsewhere. The Small Business Act for Europe (European Commission, 2008), for example, simultaneously aims at fostering risky innovations and employment growth by promoting the market entry of small, entrepreneurial enterprises. Our analysis shows that these goals may be mutually incompatible when the interaction between the entrepreneur's project choice and her choice of commercialization mode is explicitly analyzed. An alternative policy would be to reduce or even eliminate the tax and subsidy preferences in favor of entrepreneurial market entry. This would imply to align the tax and subsidy rates that are applicable for the sale of a business or patent on the one hand, and for the profits earned by the entrepreneur under market entry on the other. This policy is recommended, for example, in the Mirrlees Review (2011), which stresses – in the context of UK tax policy – the case for more tax neutrality in business taxation. In addition to changed tax incentives, a further element of such a policy package would be an improved legal framework that reduces the obstacles for a patent sale to incumbent firms and ensures an effective bidding competition for target firms and their patents. Our analysis has shown that such a policy is particularly attractive when the net externalities from entrepreneurial market entry and the advantage of the entrepreneur in developing her own innovation are both small. In a model extension we have also shown that the case for a neutral tax policy with respect to the entrepreneur's commercialization mode is further strengthened when the innovation reduces variable production costs and thus affects aggregate industry output.

30

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31

In addition to its normative implications, our analysis also leads to several hypotheses that can be tested empirically. As subsidy programs have proliferated and effective tax rates for incorporated start-ups have fallen sharply in many countries, this has led to a rising tax-subsidy advantage of market entry by entrepreneurs over the alternative of project sale. According to our model, these policies should have led to a rising share of innovations that are commercialized by the market entry of entrepreneurs, and they should also have induced less risky innovation projects with a lower expected return. In concluding, we emphasize that our analysis is but a first step towards a more comprehensive study of the effects of public policies on the market for entrepreneurial innovations. A possible first extension of our framework is to endogenize optimal policies towards small businesses, including the extent of loss offset provisions that apply to new enterprises. In such a setting the disadvantages of restrictive loss offset opportunities that have been at the core of our study could be set against the possible tax revenue losses that arise from fraudulent claims for tax rebates. A second extension would be to incorporate a more complex interaction between entrepreneurial start-ups and incumbent firms, allowing the latter to also invest in research activities and giving them an independent advantage of developing the project further. Another relevant extension would be to focus on the financing of the entrepreneurial start-up, by incorporating banks or venture capitalists in a setting of asymmetric information. For instance, heavy reliance on debt financing can lead to excessive risk-taking by entrepreneurs, counteracting the distortion that arises from limited loss offset opportunities in the present paper. Finally, the analysis could be extended to a dynamic model, in order to capture the growth-promoting effect that is a major reason for the policy support to entrepreneurship. These extensions seem to be promising areas for future research. Acknowledgments We are grateful to two anonymous referees for their insightful and constructive comments. This paper was presented at conferences and seminars in Ann Arbor, Berlin, Frankfurt, Lund, Milan, Munich, Oxford and Stockholm. We have benefited from many helpful comments and suggestions by Massimo Bordignon, Michael Devereux, Judith Freeman, Clemens Fuest, Volker Grossmann, Dietmar Harhoff, Magnus Henrekson, Christian Keuschnigg, Andreas Oestreicher, Marco da Rin, Martin Ruf, Mikael Stenkula and Miranda Stewart. Thanks also to Christina Lönnblad for improving the language and to Nina Öhrn for the research assistance. Financial support from the Jan Wallander and Tom Hedelius' Research Foundations and Vinnova is gratefully acknowledged. This paper was started when the first author visited the Research Institute of Industrial Economics in Stockholm and continued when the third author visited CESifo in Munich. We wish to thank both institutions for their hospitality. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.jpubeco.2014.03.002. References Acs, Z.J., Audretsch, D.B., 2005. Entrepreneurship, innovation and technological change. Found. Trends Entrep. 1, 149–195. Altshuler, R., Auerbach, A.J., 1990. The significance of tax law asymmetries: an empirical investigation. Q. J. Econ. 105, 61–86. Anton, J.J., Yao, D.A., 1994. Expropriation and inventions: appropriable rents in the absence of property rights. Am. Econ. Rev. 84, 190–209. Asoni, A., Sanandaji, T., 2014. Taxation and the quality of entrepreneurship. J. Econ. http:// dx.doi.org/10.1007/s00712-013-0375-z (forthcoming). Auerbach, A.J., 2007. Why have corporate tax revenues declined? Another look. CESifo Econ. Stud. 53, 153–171. Ayers, B.C., Lefanowicz, C.E., Robinson, J.E., 2003. Shareholder taxes in acquisition premiums: the effect of capital gains taxation. J. Financ. 58, 2783–2801. Baumol, W.J., 2004. Entrepreneurial enterprises, large established firms and other components of the free-market growth machine. Small Bus. Econ. 23, 9–21.

Bhattacharya, S., Mookherjee, D., 1986. Portfolio choice in research and development. RAND J. Econ. 17, 594–605. Bloom, N., Griffith, R., Van Reenen, J., 2002. Do R&D tax credits work? Evidence from a panel of countries 1979–1997. J. Public Econ. 85, 1–31. Bond, S., Devereux, M., 2003. Generalised R-based and S-based taxes under uncertainty. J. Public Econ. 87, 1291–1311. Cabral, L., 2003. R&D competition when firms choose variance. J. Econ. Manag. Strategy 12, 139–150. Cohen, W., 2010. Fifty years of empirical studies of innovative activity and performance. In: Hall, B.H., Rosenberg, N. (Eds.), Handbook of the Economics of Innovation, vol. 1. North-Holland, Amsterdam, pp. 129–213. Cooper, M., Knittel, M., 2006. Partial loss refundability: how are corporate tax losses used? Natl. Tax J. 59, 651–663. Cullen, J.B., Gordon, R.H., 2007. Taxes and entrepreneurial risk-taking: theory and evidence for the U.S. J. Public Econ. 91, 1479–1505. Da Rin, M., Di Giacomo, M., Sembenelli, A., 2010. Corporate taxation and the size of new firms: evidence from Europe. J. Eur. Econ. Assoc. 8, 606–616. Da Rin, M., Di Giacomo, M., Sembenelli, A., 2011. Entrepreneurship, firm entry, and the taxation of corporate income: evidence from Europe. J. Public Econ. 95, 1048–1066. Desai, M., Foley, F., Hines, J., 2004. Foreign direct investment in a world of multiple taxes. J. Public Econ. 88, 2727–2744. Djankov, S., Ganser, T., McLiesh, C., Ramalho, R., Shleifer, A., 2010. The effect of corporate taxes on investment and entrepreneurship. Am. Econ. J. Macroecon. 2, 31–64. Domar, E.D., Musgrave, R.A., 1944. Proportional income taxation and risk-taking. Q. J. Econ. 58, 388–422. Edgerton, J., 2010. Investment incentives and corporate tax asymmetries. J. Public Econ. 94, 936–952. EFI (Expert Commission for Research and Innovation in Germany), 2011. Report 2011. http://www.e-fi.de/fileadmin/Gutachten/EFI_2011_en_final.pdf (English version). Egger, P., Keuschnigg, C., Winner, H., 2009. Incorporation and taxation: theory and firmlevel evidence. CESifo Working Paper No. 2685. CESifo, Munich. Ernst, C., Spengel, C., 2011. Taxation, R&D tax incentives and patent application in Europe. ZEW Discussion Paper No. 11-024. ZEW (Zentrum für Europäische Wirtschaftsforschung), Mannheim. European Commission, 2008. Think small first. A “Small Business Act” for Europe. Communication COM (2008) 394 (http://ec.europa.eu/enterprise/policies/sme/smallbusiness-act/index_en.htm). Färnstrand Damsgaard, E., Norbäck, P.-J., Persson, L., Vasconcelos, H. (2010). When do independent innovators make the break-through inventions? Mimeo, Research Institute of Industrial Economics, Stockholm. Friebel, G., Giannetti, M., 2009. Fighting for talent: risk-taking, corporate volatility and organisation change. Econ. J. 119, 1344–1373. Fuest, C., Huber, B., Nielsen, S.B., 2002. Why is the corporate tax rate lower than the personal tax rate? The role of new firms. J. Public Econ. 87, 157–174. Gentry, W.M., Hubbard, R.G., 2000. Tax policy and entrepreneurial entry. Am. Econ. Rev. 90, 283–287. Gilbert, R., 2006. Looking for Mr. Schumpeter: where are we in the competition–innovation debate? In: Jaffe, A.B., Lerner, J., Stern, S. (Eds.), Innovation Policy and the Economy, vol. 6. MIT Press, Cambridge and London, pp. 159–215. Giuri, P., Mariani, M., Brusoni, S., Crespi, G., Francoz, D., Gambardella, A., Barcia-Fontes, W., Geuna, A., Gonzales, R., Harhoff, D., Hoisl, K., Le Bas, Ch., Luzzi, A., Magazzini, L., Nesta, L., Nomaler, Ö., Palomeras, N., Patel, P., Romanelli, M., Verspagen, B., 2007. Inventors and invention processes in Europe: results from the PatVal-EU survey. Res. Policy 36, 1107–1127. Gordon, R.H., 1998. Can high personal income taxes encourage entrepreneurial activity? IMF Staff. Pap. 45, 49–80. Häckner, J., 2000. A note on price and quantity competition in differentiated oligopolies. J. Econ. Theory 93, 233–239. Henkel, J., Rønde, T., Wagner, M., 2010. And the winner is — acquired: entrepreneurship as a contest with acquisitions as the prize. CEPR Discussion Paper No. 8147. Henrekson, M., Johansson, D., 2010. Gazelles as job creators: a survey and interpretation of the evidence. Small Bus. Econ. 35, 227–244. Henrekson, M., Sanandaji, T., 2011. Entrepreneurship and the theory of taxation. Small Bus. Econ. 37, 167–185. Hvide, H.K., 2009. The quality of entrepreneurs. Econ. J. 119, 1010–1035. Keuschnigg, C., Nielsen, S.B., 2002. Tax policy, venture capital, and entrepreneurship. J. Public Econ. 87, 175–203. Keuschnigg, C., Nielsen, S.B., 2004. Start-ups, venture capitalists, and the capital gains tax. J. Public Econ. 88, 1011–1042. Kreditanstalt für Wiederaufbau, 2013. Geschäftsbericht 2012. Frankfurt/Main. Landsman, W., Shackelford, D., 1995. The lock-in effect of capital gains taxes: evidence from the RJR Nabisco leveraged buyout. Natl. Tax J. 48, 245–259. Langenmayr, D., Lester, R. 2013. Taxation and corporate risk taking. Working Paper 13/16. Oxford University Center of Business Taxation. Lerner, J., 2010. Geography, Venture Capital, and Public Policy. Rappaport Institute and Taubman Center Policy Briefs (March). Mankiw, N.G., Whinston, M.D., 1986. Free entry and social inefficiency. RAND J. Econ. 17, 48–58. Mirrlees, J., Adam, S., Besley, T., Blundell, R., Bond, S., Chote, R., Gammie, M., Johnson, P., Myles, G., Poterba, J., 2011. Small business taxation. Tax by Design: The Mirrlees Review. Oxford University Press, pp. 451–469. Norbäck, P.-J., Persson, L., 2009. The organization of the innovation industry: entrepreneurs, venture capitalists, and oligopolists. J. Eur. Econ. Assoc. 7, 1261–1290. Norbäck, P.-J., Persson, L., Svensson, R., 2010. Creative destruction and productive preemption. CEPR Discussion Paper No. 8281. CEPR (Centre for EconomicPolicy Research), London.

A. Haufler et al. / Journal of Public Economics 113 (2014) 13–31 OECD, 2010. SMEs, Entrepreneurship and Innovation (Paris). Reinganum, J., 1989. The timing of innovation: research, development, and diffusion. In: Schmalensee, R., Willig, R. (Eds.), Handbook of Industrial Organization. Elsevier, Amsterdam, pp. 849–908. Rosen, R., 1991. Research and development with asymmetric firm sizes. RAND J. Econ. 22, 411–429. Salop, S., 1979. Monopolistic competition with outside goods. Bell J. Econ. 10, 141–156. Serrano, C., 2010. The dynamics of the transfer and renewal of patents. RAND J. Econ. 41, 686–708. Shane, S., 2008. The Illusions of Entrepreneurship: The Costly Myths That Entrepreneurs, Investors, and Policymakers Live By. Yale University Press, New Haven.

31

Singh, N., Vives, X., 1984. Price and quantity competition in a differentiated duopoly. RAND J. Econ. 15, 546–554. Svensson, R., 2007a. Commercialization of patents and external financing during the R&D phase. Res. Policy 36, 1052–1069. Svensson, R. 2007b. Data sheet for ‘Commercialization of patents and external financing during the R&D phase’. Mimeo, Research Institute of Industrial Economics, Stockholm. Tillväxtanalys, 2010. Follow up of the newly established enterprises in 2005 — three years after the start. www.tillvaxtanalys.se/tua/export/sv/filer/statistik/nyforetagande/ Uppfoeljning_3_xr_nystartade_foretag_2005.pdf.