Trademark Activity and the Market Performance of U.S. Commercial Banks

This empirical paper analyzes the effect of trademark activity on the market value and performance of U.S. commercial banks from two perspectives. First, a long-term perspective considers the effect of such activity on banks’ Tobin’s q. Second, with a short-term perspective, the authors analyze the effect of trademark activity on banks’ abnormal returns. An older portfolio of trademarks diminishes the ratio of market value to firm assets, but this ratio can be improved in the long term by abandoning old trademarks. Portfolios of trademarks with wide diversification do not help increase Tobin’s q. Furthermore, according to an event study, the creation of a trademark has a positive effect on cumulative abnormal returns compared with no event, whereas a cancellation event has a negative impact.


Introduction
Banks innovate and respond to customers' initiatives and needs by offering new products and services, yet few of these innovations are patentable. Rather they are quite easy for other banks to imitate or replicate. How then can innovative banks protect their innovations? According to Teece (1986), banks should control complementary assets, such as marketing or distribution channels. In reality, commercial banks are undertaking signifi cant investments in marketing and advertising; in 1999, they spent $5.1 billion on advertising and marketing, equal to an estimated 8.9% of their net income before taxes (Örs 2006). Furthermore, Örs (2006) reveals that advertising has a positive and economically signifi cant impact on bank profi tability and considers advertising an important element of bank competition that is linked to ownership of relevant portfolios of trademarks. A trademark-any sign that helps distinguish a product or service, whether a word, a fi gure, or an image-thus is closely related to advertising and the intangible value of the fi rm.
Marketing strategy is one more of the set of strategic tools that fi rms may use to generate cash fl ows. The value of the future cash fl ows will be refl ected in the bank market value. Poor trademarks strategies or weak marketing campaigns could lead to poor sales performance of the fi rm services. Given that the shareholder wealth can be viewed as the discounted value of expected cash fl ows, a revision of these expectations based on cancellation or introduction of trademarks could be manifested in the fi rm's stock price or even in a longer term bank performance.
In the retail banking sector, innovations also might relate to trademarks. In Rindova and Petkova's (2007) theoretical framework to explain the interaction of technological innovations with consumer perceptions, innovation enhances value creation but may not be perceived clearly by the demand side. Yet in retail banking, demand often changes before innovations appear. That is, customers' habits change fi rst (e.g., Internet access), then banks introduce innovations to cope with these changes (e.g., online banking), which they also might trademark.
The study here investigates the link between trademark activity and the performance and the stock prices of U.S. commercial banks. The focus centers on the value added by trade-or service marks owned by the main U.S. commercial banks between 1996 and 2006, because trademarks can alter the market value of the fi rm, which in turn infl uences how investors perceive fi rm value. The study concentrates on commercial banks because, whether out of necessity (e.g., new segments of customers) or merit (e.g., new forms of competition), the fi nancial sector has become an increasingly innovative industry.
Therefore, this paper estimates the impact of several explanatory variables related to banks' trademark activity on Tobin's q (Tobin 1978). Tobin's q, that is, the ratio of the market valuation of fi nancial claims relative to the cost of replacing the fi rm's assets has been used widely to assess profi table investment opportunities. The fi rms represent a set of assets and capabilities, such as knowledge, brand names, trade-and service marks, and other intangible factors, that get jointly valued on the market (see the seminal works of Grilliches (1981); Megna and Klock (1993); and the recent study of Krasnikov et al. (2009). This empirical study here focuses on the effect of specifi c characteristics of fi rm's trademark portfolio, such as age and scope, on fi rm's Tobin's q. Further implications about the fi rm's branding effort and readjustments of business strategies can be inferred at the light of the results.
Using quarterly data, the study here fi nds that an older portfolio of trademarks diminishes the ratio of market value to fi rm assets, though fi rms can improve this ratio by abandoning old trademarks. Also, trademarks with low ranges or application scopes add more value to the fi rm, in terms of a higher Tobin's q, than do trademarks with wide ranges. That is, specialization or specifi cation in the trademark portfolio is preferable to diversifi cation.
Finally, this paper also conducts a short-term analysis with a classical event study to investigate whether abnormal returns observed in banks depend on the banks' trade-mark activity. Using event study methodology (Campbell et al. 1997) we can analyze the impact of certain events related with trademark activity on the stock returns of a fi rm or set of fi rms. The event study developed in this study here relies on daily data and confi rms that the abnormal returns attributable to trademark creation differ from the abnormal returns attributable to a trademark cancellation. Specifi cally, an introduction exerts a positive effect compared with no event, whereas a cancellation event has a negative impact in terms of cumulative abnormal returns. As the sample represents a period with a strong activity of mergers and acquisitions in the banking sector, the event study is carried out in dates which are free of potential mergers and acquisitions effects to avoid any trends in the data that could be confused with abnormal returns. This robustness test confi rms the validity of the previous results.
By studying both the effect of trademark activity on fi rm value using quarterly data and the effect of introductions or cancellations of trademarks on abnormal stock returns, this work provides complementary analyses that offer results from both long-and short-term perspectives. The results therefore contribute to literature pertaining to trademark activity and commercial banking. The interesting implications regarding the effects of such activity on fi rm value should be of special interest for both investors and bank managers. For example, investors might infer bank value from its trademark policy, because certain events related to trademark activity have signifi cant effects on fi rms' value in the short and long run. Accordingly, investors could go beyond the effi cient market hypothesis, which asserts that they make trading decisions on the basis of the potential benefi ts of organizational initiatives announced by fi rms, and attend instead to the information revealed by the bank's trademarks policy. From a managerial perspective, these results suggest ways to manage trademark portfolios more effi ciently and with lower expenditures, while distinguishing short-and long-run effects. Specifi cally, managers should try to maintain a young, dynamic portfolio of trademarks and prefer the introduction of focused trademarks but the cancellation of old, diversifi ed ones.
To outline these fi ndings, this article is organized as follows: Section 2 presents the theoretical framework, then, Section 3 contains the description of the data, the methodology, and the variables for the empirical study. The empirical results are described in Section 4. Finally, Section 5 concludes and discusses the implications for both investors and bank managers.

Theoretical background
Advertising and branding has been analyzed in terms of market value and performance in some studies in the intersection between management and economic literature. For example, the importance of advertising has been documented by Mathur and Mathur (1995) who fi nd a signifi cant increase in the fi rm market value after the announcement of an advertising slogan change. Also, Reilly et al. (1997) show that strategic marketing planning (i.e. promotion management decisions) infl uence the sales, profi t, and stock market performance of fi rms. They fi nd a positive relation between the level of advertising expenditures and shareholder wealth over an eight-year period. Hozier and Schatzberg (2000) fi nd a signifi cant decline in stock market values before announce-ments of advertising agency terminations and accounts placed in review (potential terminations). However, the authors do not fi nd an improvement in fi rm performance following the agency termination of potential termination. Our study adds evidence to this interrelation and also examines the effect of marketing strategies focused on the banks' trademark activity on the performance and the stock prices.
Trademarks combine words, phrases, symbols, and designs to identify and distinguish the source of the goods or services. They also constitute important intangible assets of commercial banks. In their constant efforts to reach new segments of customers (e.g., immigrants, young people) and compete with electronic or savings banks, commercial banks have become very active and launched many trade-and service marks in recent years. Evidence of this importance can be found in Hasan et al. (2000) who investigate the impact of marketing on U.S. savings and loans industry and show a strong positive impact on the profi tability of their non-interest business activities over the period 1985 and 1989. They also indicate that promotional expenditures of U.S. savings and loans increased due to the increased competition in the local markets as a result of the introduction of new competitors during the period.
As we have already introduced, this study fi rst focuses on the value added by trade-or service marks by the main commercial banks of the United States between 1996 and 2006. To do so, this work estimates their impact on Tobin's q, a measure of performance traditionally employed in fi nancial and managerial literature to explain a wide variety of phenomena. For example, the measure has helped estimate the relative importance of industry or share effects for determining the market value of fi rms (for a review, see Hall et al. 2007; see also Lindenberg, Ross 1981). In similar approaches to ours, other researches examine the impact of R&D or patenting on fi rm performance or market value. Works such as those of Grilliches (1981), Megna and Klock (1993) and Hall (1993) also use Tobin's q to estimate intangible assets such as intellectual capital and technological assets. Wernerfelt and Montgomery (1988) use this ratio as a measure of returns from diversifi cation; fi nding that focused fi rms perform better than diversifi ed ones. Simon and Sullivan (1993) apply Tobin's q as a measure of brand equity in a similar way to our interpretation. Besides this latter study, in a related work, Krasnikov et al. (2009) analyze the fi nancial impact of branding using trademarks for different industries.
Although Tobin's q normally relates to non-fi nancial fi rms where physical capital is a reasonable indicator of their value and ability to produce, there are some papers in previous literature that use the Tobin's q to measure the performance of fi rms in the fi nancial sector. For instance, Bharadwaj et al. (1999) use Tobin's q to examine the association between Information Technology investments and fi rm value for a wide variety of industries including fi nancial services. Salas and Saurina (2003) evaluate the effects of deregulation in the market power of Spanish banks, measured in terms of their Tobin's q ratios. And recently, Greenhalgh and Rogers (2007) employ Tobin's q to analyze the effect of trade and service marks in the market value of UK manufacturing and service sector fi rms, including the fi nance sector and other similar sectors in terms of physical capital. In contrast with the latter works, this empirical study here focuses on the effect of specifi c characteristics of fi rm's trademark portfolio, such as age and scope, on fi rm's market value. Further implications about the fi rm's branding effort and readjustments of business strategies can be inferred at the light of the results. Rindova and Petkova (2007) note the importance of innovation for value creation, as well as the uncertainties associated with innovation on both supply and demand sides. Their focus is customers' and producers' adaptation to successful product innovation, such as through the creation of similarities between new products and existing ones to facilitate the use and spread of new technologies. However, there is not always a synchronized relation between technological innovation and product design; sometimes, an innovation emerges from external factors and changes the economy in general, as in the case of the widespread use of the Internet. As new technologies become more important and useful, new schemes for interpreting a new technology's value emerge and demand new product design. Existing products that are not appropriate for the new changes are abandoned. In the innovative bank sector for example, banks adapt their business lines to provide greater access to and promotion of banking services to more receptive customers who engage in a new kind of banking business.
Yet other managerial decisions also affect intangible assets, such as those related to product variety and diversifi cation, and thus affect market value. The market environment and its interaction with the fi rm determine the success of a fi rm's strategies. For example, product variety might be an important strategic variable, because according to Sorenson (2000), the likely success of a fi rm's product portfolio increases as the number of products increases, though product line culling increases organizational effi ciency when consumer preferences are stable. In the case of new technologies, consumer preferences should be clear, and the business strategy and trademark activity should refl ect these developments closely. Thus, the ideal path might be clear, in which case culling seems necessary. In a comparable argument, Gambardella and Torrisi (1998) point out that a narrower trademark portfolio is benefi cial, though other studies stress the importance of diversifi cation (Siggelkow 2003) and show that broad product offerings can be a valuable strategy for fi rm managers. Li and Jin (2006) study the effect of diversifi cation on fi rm returns in chemical and oil industries and fi nd that diversifi ed fi rms achieve higher returns than focused ones, though this fi nding may be specifi c to the industry context, in which the general products had low value prior to diversifi cation. Thus, the key to determining the role of diversifi cation is to test if such diversifi cation is well valued by businesses or not. This paper considers innovation and adaptation in terms of (1) the creation of trademarks, (2) the cancellation of trademarks, (3) the age of the stock of trademarks, and (4) the range of existing trademarks to study their effect on Tobin's q. That is, this paper analyzes the value that intangibles add to a fi rm, using quarterly data. In this sense, this approach undertakes a long-term analysis of the effect of trademark activity on banks' performance taking into account the theoretical framework detailed above.
In the second approach of this study, we turn our attention to banks' fi nancial returns. A growing stream of literature also explores the impact of product innovation and advertising on fi nancial returns and market value. R&D expenditures and advertising expenditures have been previously found to be associated with fi rm specifi c advantage, which may result in increased fi rm value (see Branch 1974;Grabowski, Mueller 1978). Chaney et al. (1991) study the effects of new product introductions on the market value of fi rms with successful results (see also Koh, Venkatraman 1991;Agrawal, Kamamura 1995). Lane and Jacobson (1995) cite stock market reactions to brand extension announcements and estimate a model that links excess returns to brand attitude and brand name familiarity. Subramani and Walden (2001) investigate the impact of electronic commerce announcements on the market value of fi rms. Recently, Bayus et al. (2003) and Srinivasan et al. (2009) show that a fi rm's product releases and fi led trademarks have a positive and signifi cant impact on stock returns. Other works consider the potential spillover effects from a rival's innovation, such as when Fosfuri and Giarratana (2009) investigate the impact of rivals' product innovation and new advertising on a fi rm's fi nancial market value in mature product markets.
In line with such research, this paper considers whether abnormal returns refl ect banks' trademark activity. Like in Chaney et al. (1991) and Lane and Jacobson (1995), a classical event study and daily data are employed to analyze the effect of trademark activity; therefore, in contrast with the fi rst empirical part, this second part adopts a short-term perspective in this case. The two analyses are complementary.
The effi cient market hypothesis claims that investors incorporate all possible benefi ts of the organizational initiatives that fi rms announce into their decision making. If certain events related to trademark activity seem likely to increase (diminish) a fi rm's future cash fl ows, the stock market should respond positively (negatively), and the market value of the fi rm changes accordingly. Trademark activity could be the exemplar of such events for this type of studies. Trademark activity may refl ect readjustment of business strategies and branding and could be considered as a proxy of investment and growth opportunities which should be derived in future cash fl ows. This information should be important in pricing equity securities and so, trademark activity news should affect to the fi rm value and the stock prices. This paper therefore defi nes two events: trademark introduction and trademark cancellation or extinction. An event study permits to determine if the creation and cancellation of trademarks leads to positive or negative abnormal returns in the short run.

Data
The The information about the banks' trade-and service marks comes from the U.S. Patent Offi ce (USPTO). Such trademarks protect intellectual property rights, and fi rms fi le them to secure legal protection of their brand names, recognizable designations, and symbols, as well as the fi rms' identities (e.g., Seethamraju 2003). The information extracted from the USPTO includes international and U.S. classes for each trademark, the mark draw, the fi ling date, and, if applicable, the cancellation date. Any single missing values in a series are fi lled by linear interpolation if information about previous and later observations is available or by the average growth rate if the gaps fall at one of extremes of the sample period. This interpolation procedure was used in very few cases. This procedure leads to a sample of 960 cancellations and 1560 introductions of trademarks by the 16 major U.S. commercial banks during the 11-year period from January 1996 to December 2006.

Methodology
This analysis uses two different approaches to study the effects of trademarks on stock market behavior for banks. First, this paper applies the classical Tobin's q linear model, as is commonly used in prior literature. Second, with a more dynamic scope, this paper studies the effect of banks' trademark policies on performance, according to abnormal daily returns, using a classical event study.

Linear model for Tobin's q Dependent variable: Tobin's q
The fi rst empirical analysis entails the effects of trademarks on the Tobin's q ratio, that is, the ratio of market value to tangible assets, which represents the added value of such assets for a given fi rm. The data source defi nes a bank's market value as the sum of market capitalization (close price multiplied by the quarterly common shares outstanding); tangible assets are the sum of gross fi xed assets plus inventory stocks. Therefore, the Tobin's q expression of fi rm i is: where V i , A i , and K i represent the market value, the value of physical assets, and the value of intangible (knowledge) assets, respectively. In equilibrium, this value should equal one, because there are incentives to invest when shares can be sold for more than the cost of the underlying assets and incentives to disinvest when shares can be purchased for less than the assets. However, Tobin's q defi ned in this way is diffi cult to observe, so instead, this study uses the replacement value of the fi rm's physical assets as the dependent variable; that is: In practice, this ratio excludes intangible assets from the denominator. The variable relates to fi rms' accounting features and their behavior with respect to trademark activity for the sample period.

Independent variables of theoretical interest
This study also uses the number of living trademarks in each period for each bank and per employee, denoted by Stock. This variable is a proxy measure of the innovation capacity of each fi rm. The stock of trademarks per employee provides a natural way to defi ne the size of the portfolio of trademarks in each quarter, as balanced by the number of employees, which provides a good proxy for the size of the fi rm. Additionally, we include the square of the number of living trademarks to employee to allow for a nonlinear relation between the innovation capacity and the fi rm performance.
The age of the stock of trademarks is measured according to the fi ling date and the cancellation date. If the trademark has not been cancelled, we compute the age according to the last date in the sample, which makes this variable right censored. In practice, this variable enters in the model as the logarithm of the average age of the stock of living trademarks, measured in days, for each quarter (log( )) live Age .
To modify stock age, bank's managers can discard trademarks of a certain age. In order to analyze this effect, the estimation procedure includes the logarithm of the average age of cancelled trademarks, measured in days, for each quarter (log( )) cancel Age .

Control variables
The use of control variables, mostly derived from prior literature, greatly reduces the likelihood of fi nding spurious effects. For example, the use of the one-period lagged Tobin's q, denoted by Tobin's q t-1 , helps to control for autoregressive effects. The previous value of the Tobin's q ratio clearly is relevant to the current value of Tobin's q.
The stock diversifi cation of living trademarks in each quarter (Diversifi cation) is the average number of different U.S. classes registered in each case. A diversifi ed portfolio means the stock of living trademarks has a wide range, which implies a high value for this variable.
Size should have effects on market value behavior. The size is measured by means of the logarithm of equity book value. Large fi rms might systematically have larger portfolios of trademarks; furthermore, this variable is a common market value factor for fi rms in classic corporate fi nance literature. For the same reason, this analysis also includes the return over income (ROI) lagged two periods, which helps us to control for the performance that the fi rm has realized two periods before.
The value of e-commerce in the United States measures the amount of internet transactions and is given in USD although it is incorporated in the model in log terms. It is obtained from the U.S. Census Bureau. This variable provides a proxy for the use of the Internet and new technologies by customers. This measure also refl ects changes in customers' habits. Therefore, the e-commerce variable helps us control for the interaction between trademark activity and changes in banks' activity due to increasing use of new technologies. That is, there is a large potential market for new online fi nancial services and products that could be responsible for a systematic increase in trademarks oriented towards this new market.
To control by the fact that bank's performance measure could be determined by its fi nancial health, we use the bank rating as an additional control variable. We defi ne the rating related variable as a discrete variable with values ranging between 1 and 9, such that 1 corresponds to rating BB+ (the lowest rating in the sample) and 9 to rating AA (the highest rating in the sample). The rest of the values for the rating variable cor-respond to the intermediate ratings. We expect a positive effect of this variable which would refl ect that fi rms with a better fi nancial health exhibit a better performance.
The number of mergers and acquisitions (M&A) during the sample period was remarkable. M&A may introduce some trends in the data which could be confused with unexpected and signifi cant changes in fi rm's performance as well as with abnormal returns. Besides the effects on fi rm's performance, M&A may create brand problems which could affect the trademark activity. For instance, after a given merger or acquisition the involved banks would need to decide what to do with overlapping and competing product brands. We try to isolate these effects from the ones shown by the trademarks related variables. For this reason, we control by the M&A activity of every bank in the sample introducing as an additional explanatory variable the number of M&A which were announced during a given quarter. The expected sign of this variable is not clear given that as Lubatkin (1983) state, ineffi ciencies associated to a merge may negate the possible benefi ts of merger and so, its overall effect is not signifi cant.
As a summary, Table 1 reports the main descriptive statistics for all variables employed in this analysis, including the mean, standard deviation (Std. dev.), minimum (Min), maximum (Max), and number of observations (N). The sample consists of quarterly data for sixteen banks over eleven year, that is, if there are no missing observations, N is equal to 704. The average Tobin's q is less than 1 for the sample and period considered. The average age of the stock of trademarks is approximately fi ve years. The average number of trademarks per bank and quarter is 136 and there is a signifi cant degree of variation: the maximum is 674 while the minimum is just 7 trademarks. This could be due to the banks' size and reinforces the necessity of standardize this variable by the bank size. The average rating category is A (6.27). M&A variable shows that on overage 0.43 M&A announcements took place by quarter and bank. Thus, on average every 2.33 quarters there was a M&A announcement. Next, we check the variables statistical properties analyzing the stationarity of the variables employed in our analysis to guess whether we must use the variable of interest (Tobin's q) in levels or fi rst differences. We perform a standard Augmented Dickey-Fuller unit-root test using drift and/or trend when their effects are signifi cantly different from zero at a 5% signifi cance level on the Tobin's q variable for each bank. We fi nd for all the banks in our sample that the dependent variable is stationary and thus, we use this variable in levels given that there should not be problems of cointegration with the explanatory variables. With the exception of the correlation between the Stock of trademarks and squared of the Stock of trademarks, which by defi nition is high (0.96), all the correlations among the remaining explanatory variables are always below 0.6. This level does not alert about potential problems of collinearity.

Linear model
The fi rst empirical study analyzes the effect of trademark activity on bank performance in a classical Tobin's q linear model. Thus, Tobin's q provides the measure of performance, and the analysis refers to the bank level. Using the defi nition of Tobin's q from Equation (2), this paper analyzes whether the added value in the market can be explained by the bank's particular management of its trademark activity. The linear regression model has the following form: The proxy for the value of innovation K i or intangible assets, is the number of each fi rm's living trademarks in quarter t, or Stock it . By including the other variables of theoretical interest and the control variables, the following econometric model for each fi rm i and each quarter t arises: where  0 is the intercept;  j , j = 1, 2, 3, 4, are the coeffi cients of the explanatory variables of theoretical interest (i.e. stock and age of the trademark portfolio); and X it is a vector whose components are the control variables (i.e. lagged Tobin's q, Size, Diversifi cation, e-commerce, lagged ROI, bank's rating, and M&A). To estimate the coeffi cients, this study employs a Prais-Winsten regression with correlated panels, corrected standard errors (PCSEs) and robust to heteroskedasticity and contemporaneous correlation across panels. Under this methodology, each element in the covariance matrix of the disturbances is computed with all available observations that are common to the two panels contributing to the covariance. Due to the nature of the sample and to the similarities in the banks behavior on a given date, it seems important to consider any cross-sectional correlation effect across the 16 banks. We have also checked for the existence of fi rst-order autocorrelation AR(1) in the residuals but we do not fi nd the residuals to be auto-correlated.

Trademark policy and abnormal returns: a classical event study
In a second approach, this study examines the effect of banks' trademark policy according to a more dynamic performance measure, namely, abnormal returns. This classical event study, instead of using quarterly data, employs daily data and thus emphasizes a short-term perspective.
Event study methodology (e.g., Campbell et al. 1997) can analyze the impact of certain events on the stock returns of a fi rm or set of fi rms. This paper employs this methodology to study the effects of trademark introductions or cancellations on the abnormal returns of commercial banks; thus, this study can establish a link between managerial actions and the market value of a fi rm. This paper defi nes two types of events: the introduction of a new trademark and the cancellation or extinction of an old one. To estimate their effects, we also need to measure the expected returns in the case of no event, which requires a fi nancial model for expected returns. In event studies of this type, the difference between the expected returns and the actual returns are the abnormal returns, denoted by AR. This work uses Fama and French's (1993) model to assess the fi rms' returns but also employs other specifi cations for the returns, such as an autoregressive model of order two, and obtains similar results.
The effi cient market hypothesis predicts that in their trading decisions, investors incorporate all potential benefi ts of organizational initiatives announced by fi rms. If certain events, related to trademark activity, are likely to increase (diminish) a fi rm's future cash fl ows, the stock market should respond positively (negatively), and the fi rm's market value should change. This estimation uses previous one-year data to estimate the abnormal returns for the data set of 16 banks. That is, abnormal returns for each day of the sample period are given by, where  is the number of days before and after the event, equal to 2+1, for a bank i. This aggregation occurs over time and across banks which leads to an average CAR for each event and for a certain event window.
According to the null hypothesis, the ARs are expected to be 0. To check this hypothesis this study conducts a test based on the J statistic. The J statistic follows a t-distribution with N-1 degrees of freedom and allows testing the impact of trademark activity on abnormal returns: where N is the length of the sample, and CAR is the average CAR across all fi rms.
In the next section, we proceed to present the results for both analysis, the linear model for Tobin's q and the event study with daily abnormal returns.

Linear model for Tobin's q
The fi rst empirical study defi ned in section 3.2.1 analyzes the effect of trademark activity on bank performance in a classical Tobin's q linear model. Thus, we present here the estimation of the linear regression model of equation (4) in section 3.2.1., considering any cross-sectional correlation effect across the 16 banks. Table 2 provides the estimation results of the model in Equation (4). In this table, each column shows the estimated coeffi cient, its degree of signifi cance, and the corresponding standard errors in parenthesis.  Model (1) in Table 2 represents the baseline model, with just the measure of trademark diversifi cation (range of the stock of living trademarks) in each quarter across the three independent variables. The coeffi cient for this variable is negative and signifi cant at a 1% level. Therefore, the ratio of market value to assets is lower when the diversifi cation of the trademark portfolio increases. The coeffi cients of Stock and Stock 2 show a convex form being the minimum at the value in which Stock is equal to 0.0085. This suggests the existence of an optimal level of trademark activity such that an increase in the stock of trademarks below this level (0.0085) affects negatively to the bank performance but increases above this level improve bank performance. The banks that are above this threshold in any period of the sample are: Capital One Financial Corp., Keycorp, State Street Corp., U.S. Bancorp.
Model (2) includes the age of living trademarks to investigate if an older stock implies a lower Tobin's q. The coeffi cient is negative and signifi cant at 5%, which supports the claim that an older portfolio of trademarks diminishes the ratio of market value to assets. A relatively younger stock of living trademarks thus is benefi cial. The coeffi cient and sign estimated for the diversifi cation variable match those obtained in Model (1), which reinforces the notion that specifi c trademarks add more value to the fi rm.
Model (3) includes the average age of cancelled trademarks, to test whether the best strategy for gaining market value through trademarks is to eliminate old trademarks from portfolios. Using the same baseline, namely, the controls from Models (1) and (2), the results indicate that the coeffi cient of log( ) cancel Age is positive and signifi cant at 5%. Thus, these results show support for the proposition that not only should the stock of living trademarks be young, but cancelled trademarks should be the oldest in the portfolio to improve Tobin's q.
Finally, Model (4) includes all the proposed variables at the same time. The results for the control variables do not vary signifi cantly, nor do the coeffi cients and signs for the variables of interest differ materially from those obtained by using the variables separately in Models (1), (2), and (3). Specifi cally, the estimation results of control variable coeffi cients are quite robust in all model specifi cations: the lagged Tobin's q coeffi cient is positive and very signifi cant, whereas the size coeffi cient is negative and signifi cant at the 1% level. As in Model (1), the stock per employee and its squared show a convex form. The e-commerce variable, which represents a common trend for all banks, displays a positive and signifi cant effect on Tobin's q at 1% signifi cance level. Finally, ROI t-2 , or fi rm performance in the two preceding periods, has positive coeffi cients in all specifi cations and a signifi cance of 10%. These results indicate a positive relationship between past performance and the ratio of current market value to assets. The rating and M&A are not signifi cant in any of the four specifi cations at any standard level of signifi cance.
Innovation and strategy management literature indicates an important role of product portfolio diversifi cation, so this study also considered several diversifi cation control variables, such as the Herfi ndahl index. In particular, the Herfi ndahl index of concentration for the type of trademark according to its international class is taken into account in this study. If this index equals one, all trademarks in that quarter belong to the same international class (the most common class in the sample is insurance and fi nancial ser-vices, IC = 36). However, none of the diversifi cation control variables was signifi cant. In an alternative specifi cation of the model, the estimation controls for the effect of systematic risk in Tobin's q using the quarterly fundamental beta, or the measure of the sensitivity of the bank's stock price to overall fl uctuations in Standard & Poor's 500. The beta equals the difference between the realized beta for a given bank and the beta associated with the market portfolio, equal to one, in absolute value which penalizes superior and inferior deviations in the same grade. However, this analysis does not include this measure of systematic risk in the fi nal results because of its collinearity with the other control variables.
According to the overall results, there is an optimal level of trademark activity such that an increase in the stock of trademarks below this level affects negatively to the bank performance but increases above this level improve bank performance. This optimal level of trademarks should be reached with new focused trademarks and should be accompanied by the cancellation of old and non-focused trademarks. It should also be noted that Tobin's q offers a long-term perspective, whereas fi rms act in response to short-term market reactions. For this reason, we next do a short-term analysis of the effect of trademark activity on banks returns.

Event study
Now, we present the results related with the effects of banks' trademark policy from a short-term perspective. For that purpose, we examine the CARs and their signifi cance defi ned in equations (6) and (7) Figures 1A, 1B, 1C, and 1D are, respectively, (-14, +14) days, (-7, +7) days, (-5, +5) days, and (-3, +3) days. The results show that an introduction has a positive effect compared with no event on the CAR. However, a cancellation event has a negative impact on the CAR. Table 3 presents these statistics for different event windows, ranging from 1-14 days before to 1-14 days after the event. In Table 3, the variable CAR intro denotes the cumulative abnormal returns for days with an introduction, while the variable CAR cancel denotes the cumulative abnormal returns corresponding to days with a cancellation. The variable CAR no event represents the cumulative abnormal returns for days with no events. In Table 3, the adjacent columns for each CAR indicate the corresponding J statistics.
Trademark activity has signifi cant impacts on the ARs in the sample. A trademark introduction exerts a positive impact compared with no event in the CARs, though CARs are signifi cant only at 10% for the (-14, +14) day window. For example, for this window, the mean CARs to trademarks introduction is approximately 7%, whereas the no event CAR is -5%. For a trademark cancellation, the J statistics are all negative and significant; in particular, the CARs for the (-3, +3), (-5, +5), (-7, +7), and (-14, +14) day windows, are approximately -5% (p < 0.10), -10% (p < 0.01), -15% (p < 0.01), and -14% (p < 0.01), respectively. As we have already mentioned, this sample represents a period with a strong activity of mergers and acquisitions in the banking sector. Therefore, as a robustness test and to avoid any trends in the data that could be confused with abnormal returns, we fi lter the fi rst 20 trading days from the announcement of a merger or acquisition, and then carry out the event study described above. Now, the CARs observed for the trademark activity taking into account mergers and acquisitions are presented in Table 4. The plots in Figure 2 represent the mean CARs to banks introducing and cancelling trademarks on each of the days in the 5-day and 7-day windows.  The results shown in Table 4 and Figure 2 indicate that the CARs to trademarks introduction are positive and signifi cant for event windows of (-5, +5), (-7, +7), and (-14, +14) days; having CARs equal to 4.5% (p < 0.10), 8.1% (p < 0.01) and 21.1% (p < 0.01), respectively. In contrast, the CARs for those banks that cancelled a trademark are negative and signifi cant at the end of the 5-day window around the cancellation (CARs = 7%, with p < 0.05), as well as at the end of the 7-day event window (CARs = 10.1%, with p < 0.01). Hence, this robustness test confi rms the validity of the previous results. Thus, the ARs attributable to the creation of a trademark are different from those attributable to a cancellation. The performance measures in Sections 4.1 and 4.2 differ in both conceptual and temporal features. The conceptual difference is obvious; regarding the temporal dimension, Section 4.1 is based on quarterly data, whereas Section 4.2 uses daily data. Therefore, some fi rm actions might not be perceived as advantageous for bank performance in the short run but would benefi t fi rm value in the long run, perhaps due to long-term intra-fi rm synergies.
The overall results suggest that a new trademark generates positive abnormal returns, which gives marketing managers incentives to fi le new trademarks. Cancelling a trademark generates negative abnormal returns; from a short-term perspective, a cancellation does not seem like a good strategy. Yet such a strategy would lead to excessive trademarks, based solely on short-term incentives. Even though the cancellation of trade- marks leads to negative and signifi cant abnormal returns in the short term, the results in Section 4.1 indicate that when controlling for age and other potential control variable, cancellations can lead to positive long-run effects on fi rm value, especially if they help focus the scope of the trademark portfolio.

Conclusions
Advertising plays an important role for commercial banks and their profi tability (Örs 2006). At the same time, advertising is closely linked to the possession of a relevant portfolio of trademarks. Because of the close relationship between trademarks and advertising, trademark activity should have a key infl uence on commercial banking. Following this line of reasoning, this empirical study tries to shed light on the relationship among trademark activity, market value, and fi rm performance.
This paper analyzes and measures empirically the effects of intangible assets, in the form of trademark activity, on the market value and the stock performance of commercial banks in the United States-one of the most innovative industries in modern economies. When Tobin's q is the dependent variable, results show that the ratio of market value to assets is lower when the diversifi cation in the trademark portfolio increases. In addition, the coeffi cient for the age of the trademarks is negative and signifi cant. Therefore, maintaining a relatively young stock of trademarks seems to be benefi cial.
To improve the Tobin's q, fi rms can also ensure that the trademarks they cancel are the oldest ones in their portfolio. Using abnormal returns on a daily basis, the results indicate a positive (negative) abnormal return around the date of the introduction (cancellation) of a trademark. Latter results are robust to including information of mergers and acquisitions dates.
These results thus have key implications for both investors and bank managers. An investor can infer, from a bank's trademark policy, the value of the fi rm. For example, if a bank cancels its divergent, old trademarks, investors should predict positive long-run effects on the bank's value. Thus, among many other economic indicators, investors should attend to the bank's trademark policy to determine their investment strategy, depending on their time horizon.
Bank managers also might use these results to guide decisions about managing the trademark portfolio more effi ciently and with lower expenditures. In the short run, the market perceives trademark introductions positively, leading to signifi cant positive abnormal returns. Effi cient portfolio management requires that the trademark activity is above the optimal level which guarantees that increases in the stock of trademarks above this level improve bank performance. This optimal level should be reached with new focused trademarks and should be accompanied by the cancellation of old and nonfocused trademarks. Such cancellations not only improve performance but also imply savings for the bank, because it is costly to maintain trademarks.
Among others, Lin et al. (2006) claim that the most challenging task of today's managers in the knowledge-based economy is to exploit the full value of corporate knowledge assets and to accumulate and commercialize knowledge assets effectively. Rutkauskas and Ginevicus (2011) also show the relevance for modern organizations of an integrated analysis of marketing effi ciency and risks. This paper here agrees with both statements. The current economic environment reinforces the importance and usefulness of new technologies, as well as the diminishing appropriateness of old products due to the emergence of new habits and behaviors. The bank sector is not an outside spectator of these changes; rather, this sector has become one of the most innovative industries. Consider an interesting question for further research. What is the optimal timing for cancelling trademarks? The results suggest that old, diversifi ed trademarks must be cancelled, but there could be a positive relation between a stable, nonvolatile economic scenario and trademark cancellation. This timing could refl ect a lesser impact in the short run but signifi cant effects on long-run bank performance.