«Tobias Schmidt (Deutsche Bundesbank) Wolfgang Sofka (Department of Organisation and Strategy, CentER, CIR, Tilburg University) Discussion Paper ...»
We construct a simple theoretical model to investigate the effect of the client portfolio of a firm’s main bank on the firm’s R&D investment which can be easily extended. We assume two identical firms is and js. Both operate in industry s. Bank A is the main bank of firm is, bank B is the main bank of firm js. Banks A and B are identical, except for their client portfolio. Each bank has only two firms in its portfolio. Bank A’s portfolio consists of firm is and identical firm ks which operates in the same industry; for short PA(is, ks). It is further assumed that industry s consists only of three firms is, js and ks. Bank B’s portfolio encompasses firm js and identical firm lt from a different industry t; in short PB(js, lt).
We will develop a purely comparative argument for the R&D investments of firms is and js.
We make the following assumptions: All firms have equal propensities to invest in R&D. The underlying uncertainties and adverse selection problems for the particular R&D investments are identical for bank A and B. They are fully exogenous to each bank. However, the portfolio composition can provide bank A with an information advantage over bank B. Firms can overcome endogenous uncertainties through various forms of R&D, market research, prototyping, simulations, etc. All firms reveal information to their main banks through their transactions and loan applications. Substantial parts of this information can be expected to be private and not available to the general public. This information may include competitive interactions, future product and market plans as well as revenue streams (Boot & Thakor, 2000). It is important to note that the private information is produced by the individual firm and only aggregated by its main bank which is uniquely positioned to do so. The bank does not necessarily learn in the narrow sense of exploring causal relationships but benefits from information externalities based on its client portfolio (Stiglitz, 2002). It is an externality because the bank does not directly and economically reimburse its other clients for the provision of this particular information. Hence, bank A and bank B draw from different pools of information in their lending decisions. The bank with the more relevant information can be expected to be in a position to assess risk premiums for individual firms more accurately than the general, high-risk premium. As a result, more funds will be available to its clients and, all other things being equal, the firm doing business with this bank should be able to make a comparatively greater investment in R&D.
The relevance of the information externality of the main bank for firms is and js is greatest if the information stems from a similar technological and market context, i.e. from competitors in the same industry (Dussauge, Garrette & Mitchell, 2000). Hence, PA(is, ks) can be expected to deliver more relevant information externalities than PB(js, lt) because bank A can obtain information from firm ks which operates in the same industry as is. Given that the pool of relevant information is finite, i.e. from all firms in a given sector, a bank that has a greater number of such firms as its clients is more likely to benefit from information externalities.
Hence, information asymmetries between bank A and bank B emerge from their market share with firms in sector s.
In a typical loan application process a bank will benchmark the information of a prospective borrower against key figures from its other clients in the same sector. This comparison is often times based on information stemming from other lending contracts which is not publicly available. The quality of such benchmarks is expected to be higher for banks that draw from a larger pool of industry information than banks with a comparatively narrower pool. In theory, firms could be expected to avoid certain banks in the first place because of the danger of unintentional knowledge spillovers to competitors. However, in reality, strong safeguards are in place to prevent banks from revealing information about one client to another client. The penalties would be high in terms of both legal liability and reputational losses (e.g. Degryse & Ongena, 2001).
However, the degree of specialization of its client portfolio is not an isolated information provision tool for the bank. A high degree of specialization in one industry would also imply that the risks involved from the technology or market side are highly correlated. This follows the basic rationale that banks manage the risks originating from their clients for the portfolio as a whole rather than individually (Markowitz, 1991). Banks can reduce the systemic risk of the overall portfolio by combining uncorrelated risks (Markowitz, 1952). Following this portfolio theory logic, PA(is, ks) contains more risk than PB(js, lt) because the risks originating from firms js and lt can be expected to be less correlated since they operate in different sectors, i.e. market and technology environments, respectively. Bank A can be expected to demand a higher risk premium from its client is than does bank B from js solely based soley on the risk exposure of its portfolio. As a result, available funds for is should be comparatively
lower, resulting in less R&D investment. We suggest:
Hypothesis 2: R&D investment of a firm decreases with the degree of specialization of its main bank’s corporate client portfolio in its industry.
Hypotheses 1 and 2 are not mutually exclusive. The former is denominated by the size of the industry the latter is dominated by the size of a bank’s client portfolio. Banks with a small portfolio can easily have a portfolio which is dominated by clients from a single industry (i.e.
high degree of specialization) while these clients represent only a small fraction of all firms in the industry (i.e. the bank has a small market share in the industry). The information externality logic suggests that banks benefit equally from every additional firm that they can take into their client portfolio. However, the aggregation of information from a bank’s portfolio entails costs for the bank. These costs are especially high if the information is dispersed and, hence, difficult to screen (Koput, 1997). This effect is most pronounced if a bank is highly diversified across a large number of industries, i.e. the degree of portfolio specialization is low. Information screening becomes increasingly effective and efficient if fewer industry information domains have to be covered. Information processing at a bank can be expected to be especially productive if it draws from a large pool of information based on a large market share in a given industry and a high degree of specialization in this industry making the screening of the information more efficient. A bank with these characteristics should possess superior information for setting adequate risk premiums for its clients in that particular industry. This in turn, should enable these client firms to invest more in R&D. We
Hypothesis 3: R&D investment of a firm increases if its main bank has both a large market share in its industry and a high degree of specialization of its main bank’s corporate client portfolio in its industry, i.e. there is a positive moderating effect.
Finally, the degree of uncertainty (both exogenous and endogenous) is not equally distributed across all industries. Especially at the “research” stage, which is not yet directed at a particular product, technological and market potentials are highly uncertain compared to the “development” stage, in which potential revenue streams are beginning to emerge (for a recent review, see Czarnitzki, Hottenrott & Thorwarth, 2011). It can take several years between the start of an R&D project and the generation of economic returns (e.g. in pharmaceuticals) or just several months (e.g. in service sectors where production and consumption are almost instantaneous) (Berry, Shankar, Parish, Cadwallader & Dotzel, 2006). Hence, the level of uncertainty of innovation activities in an industry is a function of the time it takes for an R&D project to be ready for application. This distance to application has often been linked to the importance of scientific knowledge from universities which is closer to academic research and further removed from industrial commercialization (e.g.
Siegel, Waldman, Atwater & Link, 2004; Agrawal, 2006). Cohen, Nelson & Walsh (2002) identify important differences among industries in the usage and importance of academic knowledge. We argue that the distance to application increases the uncertainty of the innovation activities in an industry. This, in turn, increases the potentials for benefitting from information asymmetries because the final resolution of fundamental uncertainties through observable market success is further removed in the future. At the same time, the risk of financing R&D increases if potential revenue streams are further delayed in the future (Czarnitzki et al., 2011).
Hypothesis 4: The positive effects of main bank industry market share and the negative effects of main bank industry specialization on firm R&D investment are greater in industries which rely heavily on knowledge from scientific sources.
3.2 Signaling through reputation and legitimacy So far, we have considered mechanisms only on the bank side and their ability to overcome information asymmetries through externalities. However, firms have additional opportunities to overcome the information asymmetries by signaling the value of their R&D activities. We follow Ndofor et al. (2004) and define a signal as “conduct and observable attributes that alter the beliefs of, or convey information to, other individuals in the market about unobservable attributes and intentions (p.688)”. This is a deviation from the theory outlined in the previous section as firms is and js in the model are no longer considered to be identical. They differentiate themselves through firm-specific signaling. A credible signal will allow a bank to provide a more accurate risk assessment on a firm’s R&D investment, resulting in more available funds and subsequently increased R&D investment. We will explore signals based on firms past actions (patenting) as well as legitimacy that can be transferred from ties to established actors and institutions (government R&D subsidies and venture capital investors).
The value of signaling through past actions is rooted in theory of firm reputation (Rindova, Williamson, Petkova & Sever, 2005). Levitas et al. (2009) investigates the value of patents as signals for attracting venture capital investors and corroborates it for a sample of firms from the pharmaceutical industry. Patents are a tangible representation of a successful innovation.
Moreover, the patent office requires a certain degree of novelty in order to grant a patent (Encaoua, Guellec & Martinez, 2006). The existence of a patent therefore also allows inferences to be drawn about the quality of the underlying R&D. Patents can be interpreted as signals of future revenue streams. These may come from possessing a temporary advantage on the product market or through generating license fees (Levitas & McFadyen, 2009).
Hence, a main bank’s risk concerns based on correlated risks in its client portfolio should be reduced.
Other potential signals are not rooted in a firm’s past actions but in being associated with authoritative actors (Rindova et al., 2005). This perspective is rooted in institutional theory.