1 Introduction

Innovation helps a firm realize its valuable profits or growth in the long-run, but requires long-term efforts. It is difficult for a CEO (chief executive officer) to predict future outcomes since the process of innovation is uncertain and has the high probability of failure (Finkelstein & Boyd, 1998). CEOs commonly face a dilemma concerning whether to invest in innovation or not. For instance, CEOs would establish optimal decision-making by considering a trade-off between focusing on short-term performances to seek current gains and participating in innovation at the risk of uncertainty. Therefore, the CEOs’ willingness to take part in innovation may be determine successful innovation. This paper investigates the association between CEO age and firm innovation. In recent years, financial economists have been interested in psychological biases based on individual characteristics of CEOs and have reported that these characteristics affect overall managerial outcomes (Anderou et al., 2017; Galasso & Simcoe, 2011; Hirshleifer et al., 2012; Li et al., 2017; Yim, 2013). A firm’s strategic decision-making such as innovation activities may be determined by CEOs, since, as is well-known, investment in innovation is one of the most fundamental decisions made by them (Barker & Mueller, 2002).

Moreover, we note that there are two conflicting views on the economic effects of young CEO’s career concern. The first view predicts that young CEOs would be more sensitive to poor firm performances than older CEOs since achievements below par in early careers lead to a drop in their pay and early dismissals (Anderou et al., 2017; Holmstrom, 1999). A current good firm performance further increases the younger CEO’s pecuniary advantages in the future. Therefore, younger CEOs may have greater incentives to focus on short-term-oriented decision makings in order to increase their current firm performances by avoiding uncertain and/or risky projects, and by doing so, they are less likely to make innovative investments. The second view says that young CEOs tend to concern with their reputation in a labor market and thereby have incentives to modify the reputation the labor market has previously had (Prendergast & Stole, 1996), by participating in risky projects and then signaling to the market that they are excellent in innovation activities. Following Li et al. (2017), we name the first view market learning hypothesis, and the second view managerial signaling hypothesis, respectively. We test that which of these two hypotheses is supported for young CEOs.

Our dataset comprises 8589 firm-year observations in 199 unique IT firms and 625 unique non-IT firms listed on two Korean stock markets, the KSE (Korean Stock Exchange) and KOSDAQ (Korean Securities Dealer Quotation), between 2002 and 2016. In the sampling process, we control for the differences that young CEOs are likely to run different styles of firms with old CEOs by a propensity score matching. In this paper, moreover, we concentrate on comparing the effects of CEO age for IT (information and technology) firms and non-IT firms in Korea. From the existing studies, we find it hard to identify empirical evidence that describes industry-specific features. Exploring a mechanism for innovation by industry and comparing CEO age-innovation relations between IT and non-IT firms may provide some practical implications for in IT industry. Furthermore, Korea is known to a pioneer in the IT sector. According to the Information Society Report 2017 from the International Telecommunication Union, the Korea ranked second in the world and first in Asia based on Information and Communication Technology Development Index. In addition, with the advent of a new economy represented by a digital revolution in Korea and other economies around the world, the IT industry leads to the development of technology that is emerging as the most interesting issue for recent technology-related policies. Our paper may provide important implications for innovation with researchers or practitioners in IT firms in emerging economies.

We capture firm innovation with patent applications. An innovative project, however, is complicated than what is reflected in simple patents, which makes it difficult to capture the incentives for the scope of innovation. For instance, it is likely that some CEOs that file many patents apply for patents exploiting well-known technologies in order to maximize their wealth by increasing technological performances (Balsmeier et al., 2017). Meanwhile, for other CEOs, a small number of patents may be due to filing that are unfamiliar with them. Accordingly, we not only capture innovation productivity by the number of patent applications, but also characterize the technological scope of innovation via a distinction between exploitation of familiar technologies and exploration of unfamiliar ones (March, 1991).

Our primary results are as follows. We document a positive effect of CEO age on innovation productivity measured as the number of patent applications. Generating three categories by CEO age: under 50, between 50 and 59, and 60 and above, we also find that CEOs aged under 50 are more likely to apply for patents than CEOs aged 60 and above. In terms of the technological scope of innovation, CEO age significantly raises the proximity between current and past patent portfolios. In addition, CEOs aged under 50 tend to hold patent portfolios that have wider scope than past ones. These CEO age-innovation associations are more pronounced in IT firms than in non-IT firms. These findings imply that the positive effects from the managerial signaling hypothesis dominate the negative effects from market learning hypothesis, which may be interpreted that younger CEOs in IT firms have more incentives to signal their innovativeness to be regarded as ‘outstanding innovators’ or ‘adventurous innovators.’ We show in detail the process of deriving these results.

In this paper, we shed light on the role of CEO age playing in shaping firm innovation. Recent work identifies crucial effects of CEO age on R&D (research and development) investments. Barker and Mueller (2002) and Li et al. (2017) propose that a young CEO tends to take greater risk by increasing investment in R&D as her concern takes a long time. This paper differs from this contribution in focusing on the number of patent applications and the technological proximity. This paper may also be taken in as a study on implicit incentives for young CEOs in fulfilling firm innovation. The implicit incentive is from a CEO’s desire to acquire a reputation in the labor market (Tirole, 2006). Our findings would be empirical evidence that the incentives needed to manage the reputation for innovation are practical for young CEOs in IT firms.

This paper proceeds as follows. In Sect. 2, we review the related literature and propose two testable hypotheses. We describe our data, variables, and descriptive statistics in Sect. 3. Section 4 presents our empirical models and investigates the relationship between CEO age and firm innovation. Section 5 concludes.

2 Related literature and hypothesis development

2.1 Related literature

This paper relates to two strands of literature. First, this paper is associated with the literature that explores the effects of CEO age. Existing studies argue that a CEO is disciplined in a CEO labor market in which a poor firm performance leads to the reduction of future wage or early dismissal since the labor market builds a belief in the CEO’s ability with her current performance and then assesses her future pecuniary reward based on this belief (Tirole, 2006). A poor-performed CEO fears that her wealth will be damaged in the future since a firm performance affects wealth through her compensation package. The argument related to career concerns is based on the difference in the sensitivity between firm performance and compensation depending on CEO age. For instance, younger CEOs may more benefit from increased compensation for the remaining horizon (Bliss & Rosen, 2001), by increasing the amount of compensation at their early careers, and therefore be willing to enjoy a ‘busy life’ (Li et al., 2017). On the other hand, the increase in compensation is unlikely to benefit from older CEOs since they retire soon (Cheng, 2004). The older CEOs with career concerns, therefore, tend to enjoy a ‘quiet life’ (Bertrand & Mullainathan, 2003).

Second, this paper also relates to the determinants of firm innovation. Innovation frequently fails, while it is known to be main engine for firms to realize profits and to promote growth and survival (Aghion et al., 2013). In terms of corporate governance or incentive schemes, existing studies have found that how to provide managers with incentives for firm innovation that are uncertain or highly likely to fail. The studies suggest that shareholder monitoring induces managers to participate in firm innovation. For example, large shareholders can monitor managers from making myopic investment decisions (Stein, 1988). In addition, institutional investors may actively monitor managers since they are independent of conflicts of interest between shareholders and corporate management. Aghion et al. (2013) state that increased monitoring of institutional investors eases managers’ career concerns since it protects the managers from the failure in innovation processes. In regard to incentive schemes, Manso (2011) documents the incentives schemes tolerating for early failure and rewarding for long-term effort, such as stock options with longer vesting period, managerial entrenchment, or golden parachutes. Chang et al. (2015) report that employee stock options characterized by fewer downside risk or longer vesting period buffer the failure of innovation and compensate for long-term efforts of employees.

There are also studies that explore whether CEOs’ willingness to actively participate in firm innovation stems from their individual characteristics. According to upper echelons perspectives presented by Hambrick and Mason (1984), managers being close to retirement are likely to make conservative decisions because they enjoy less benefits from long-term investment. Supporting the perspectives, Barker and Mueller (2002) note that incentives for younger CEOs to invest in R&D are greater than for older CEOs. In terms of managerial overconfidence, i.e., a tendency to underestimate the probability of failure in an investment project, Galasso and Simcoe (2011) and Hirshleifer et al. (2012) report that an overconfidence CEO aggressively engage in innovation. They also state that successful innovation reveals information on a CEO’s competence for innovating whereas failure to innovate signals a lack of CEO skills.

2.2 Hypothesis development

Theoretical literature on CEO age shows two conflicting views on young CEOs’ career concerns. The first view predicts that younger CEOs put more efforts into improving their current firm performances rather than engaging in risky projects (Holmstrom et al., 1986; Holmstrom, 1999). With a situation in which internal or external labor markets assess a CEO’s ability based on her accomplishment, the CEO believes that a good firm performance leads to a good market assessment for her ability. According to the study of Holmstrom (1999), if a CEO makes significant results in her early career and thereby increases her wage, she benefits greatly from increased wage during the rest of her career horizon. Nonetheless, for young CEOs, their poor firm performances may be more distressing ones. Because the CEOs’ poor performances lead to their wage falling or early dismissals, the younger CEOs with longer career horizons would set their sights on boosting current firm performances. In other words, uncertain and risky projects may be not attractive for them. We name this prediction as market learning hypothesis, following Li et al. (2017).

We can find empirical vindications on the above-mentioned argument from the behaviors of fund managers or security analysts in environments in which their abilities are assessed directly from the capital market. For example, younger fund managers tend to hold portfolios with lower idiosyncratic risk and have more conventional portfolios because they are more sensitive on being early terminated of their careers (Chevalier & Ellison, 1999). Hong et al. (2000) show that young security analysts are likely to yield forecasts that do much not differ from consensus by worrying about being terminated for inaccurate earnings forecasts. Anderou et al. (2017) also report that younger CEOs are more likely to face stock price crash risk in that they tend intentionally not to disclose bad news to investors.

Overall, young CEOs who worry about possible damages in their careers may focus on setting policies that are able to improve current firm performances, and in doing so, they may have great incentives to benefit from increased pecuniary rewards over their career horizons. Therefore, under the market learning hypothesis, we expect that young CEOs will not actively participate in uncertain and/or risky firm innovation involving long-term efforts.

Hypothesis 1

Under the market learning hypothesis, younger CEOs are less likely to engage in firm innovation.

The second view is based on the prediction that a young CEO aggressively signals her ability to the labor market. Within the remaining career horizon, young CEOs who are not being close to retirement consider non-monetary incentives such as a reputation from a labor market as well as monetary incentives (Gibbons & Murphy, 1992). The study of Prendergast and Stole (1996) posits that a CEO’s decision-making reflects her competence, and this competence is signaled to the labor market in an implicit form. According to the study, CEOs may take a couple of actions on a receipt of new information; some are to exaggerate their own opinions on net present values of investment projects and others are to keep their existing behaviors conservatively unchanged. For young CEOs, the exaggeration is a good means of signaling their abilities to the market. On the other hand, old CEOs are reluctant to respond to new information indicating that previous experiences were wrong. We call this prediction as the managerial signaling hypothesis.

Empirical studies that support the managerial signaling hypothesis report that young CEOs are willing to perform acquisitions (Yim, 2013), risky corporate policies (Serfling, 2014), or restructuring (Li et al., 2017), even though these actions do not immediately lead to increase in compensation. Yim (2013) notes that young CEOs have incentives to gain careers in acquisitions and thereby want to enjoy the benefits from them. Serfling (2014) reports that young CEOs tend to actively make corporate policies with highly-risky natures such as investment in R&D, increasing operating leverage, and diversifications in order to signal their abilities to the market. Based on the second view, we hypothesize that younger CEOs have greater incentives to signal their abilities and make larger non-monetary reputations in their remaining career horizons if they take more active stance on firm innovation that requires long-term efforts.

Hypothesis 2

Under the managerial signaling hypothesis, younger CEOs are more likely to engage in firm innovation.

3 Variables, sample, and descriptive statistics

3.1 Variables

3.1.1 Firm innovation

We consider innovation variables in two dimensions: productivity and scope. First, innovation productivity captures how effectively innovative inputs (e.g., R&D expenditures) are converted into innovative outputs (e.g., patent applications). Following the innovation literature, we measure the innovation productivity (denoted as INNO_PROit) by adding one to the number of patents filed by a firm i and year t, and taking natural logarithm. In addition, a value of zero is assigned to the firm-year that has not applied for patents in order to prevent firms that do not have patent applications from being excluded in our sample. We get information on firms’ patent applications from the two databases available in Korea: KIPRIS (Korean Intellectual Property Rights Information Services) of the Korean Intellectual Property Office and WIPSON (World Intellectual Property Service Online). We can accurately identify an inventor firm, as these databases provide detailed administrative information (e.g., classification codes of patents, legal status, original assignee, etc.) for each patent.

Second, our innovation scope captures the degree to which a CEO exploits familiar technologies or explores unfamiliar technologies, for applying for patents. In order to measure our innovation scope, following Jaffe (1989) and Balsmeier et al. (2017), we focus on the technological proximity among patents filed by a firm. This variable indicates how similar a patent portfolio in at a given year is compared to the previous year, and then may capture whether a firm innovates in a familiar technological field. Our innovation scope variable (denoted as INNO_SCit+k) is calculated as follow:

$$INNO\_SC_{it + k} = 1 - \frac{{\sum\nolimits_{k = 1}^{n} {p_{ijt} p_{ijt + k} } }}{{\left( {\sum\nolimits_{j = 1}^{n} {p_{ijt}^{2} } \times \sum\nolimits_{j = 1}^{n} {p_{ijt + k}^{2} } } \right)^{1/2} }},$$
(1)

where pijt is the fraction of firm i’s patents belonging to jth technological class based on four-digit IPC (International Patent Classification) codes at year t. The second term of the numerator of Eq. (1) is large when patents filed at t and t + k years are in almost the same technological field expressed in the concordance of IPC codes. For instance, pijt will be one with full overlap in the classification of patents for a firm. This variable, therefore, captures how technologically similar a firm’s patent portfolio at year t + k is, compared to that in year t. The growing value of our innovation scope variable indicates that firms tend to apply for patents similar to those of last year, which also means an application for familiar technological field.

3.1.2 CEO age

Our explanatory variable is CEO age. Following extant literature, we calculate CEO age by taking the natural logarithm (denoted as CEO Age). Moreover, in order to explore whether younger CEOs are more likely to engage in firm innovation than older CEOs, we generate three categories of CEOs by ages: under 50, between 50 and 59, and 60 and above (Li et al., 2017), and then define Age under 50 and Age 50–59 as variables that indicate the CEOs are under 50 and between 50 and 59, respectively.Footnote 1

3.1.3 Control variables

Following Hall and Ziedonis (2001), we control for R&D expenditures, firm size, and capital intensity. R&D expenditure is the sum of R&D expenses (cost-counted R&D) and capitalized R&D expenditures (asset-counted R&D) normalized by total assets. Firm size is measured as the natural logarithms of total assets. Capital intensity is the ratio of net property, plant, and equipment to total asset. Following the innovation literature, we also control for leverage, defined as the ratio of total debt to total assets; market-to-book ratio, proxied by Tobin’s Q; and product market competition, calculated as one minus Herfindahl–Hirschman index (HHI) based on sales of firms. In order to address a nonlinear effect of competition in the product market, we also include the squared term.

3.2 Sample

We focus on firms listed on two Korean stock markets, the KSE and KOSDAQ, between 2002 and 2016. Financial and accounting data and stock data are acquired from the TS (Total Solution) 2000 of the Korean Listed Companies Association. We only consider the calendar year-end firms for improving comparability between variables. We also exclude financial institutions, such as bank, insurance, or securities, in which capital structures or public regulations are quite different from firms in non-financial industries.

In this paper, we define IT firms are firms belonging to IT-related manufacturing industry and IT service industries. Here, based on the KSIC (Korean Standard Industry Classification), the IT-related manufacturing industry includes the Manufacture of electronic components, computer; visual, sounding, and communication equipment (KSIC two-digit codes 26). The IT service industries are defined as industries categorized in the Information and communications (KSIC one-digit code J) including Publishing activities (KSIC two-digit codes 58), Motion pictures, video, and television program production (59), Broadcasting (60), Telecommunications (61), Computer programming, consultancy, and related activities (62), and Information service activities (63). Lastly, our sample consists of 3325 firm-years in IT industry and 7869 ones in non-IT industry, respectively.

3.3 Descriptive statistics

We control for the differences that young CEOs are likely to run different styles of firms with old CEOs. As shown in Panel A of Table 1, we find statistical significances between our control variables of firms that have young CEOs versus old CEOs. Probably, these differences are not adequately considered by our empirical model. Therefore, we employ a propensity score matching, in order to alleviate the differences in observable characteristics of firms with young and old CEOs (Li et al., 2017). We categorize CEOs aged under 60, i.e., CEOs aged under 50 and between 50 and 59, as young CEO group and CEOs aged 60 and above as old CEO group, respectively, and then match these two groups by one to one (Anderou et al. 2017).Footnote 2 After the matching, in Panel A, we find that significant differences between control variables in new group disappear. Therefore, we use this matched-sample, which eliminates differences in firm characteristics according to CEO ages.

Table 1 Descriptive statistics
Full size table

Panel B of Table 1 provides summary statistics for IT and non-IT firms. In order to minimize the effects of outliers, we winsorize our variables at the top and bottom 1% of each variable’s distribution. A sample IT firm has a higher mean number of patent applications and shows technologies with low proximity than non-IT firms. CEOs of IT firms are significantly younger than those of non-IT firms. On average, a sample IT (non-IT) firm has R&D expenditures normalized by its total assets of 3.1% (2.3%), total assets of 105 (140) billion KRW, tangible asset-to-total assets of 28.3% (31.5%), leverage of 37.5% (39.8%), and Tobin’s Q of 1.104 (0.937), and is 26.299 (30.045) years old.

Table 2 presents correlation coefficient matrix for our variables. The innovation variables, INNO_PRO and INNO_SC, have negative correlation coefficients with CEO Age at the 1% levels, respectively. These coefficients indicate that sample firms with older CEOs are associated with a lower number of patent applications. For the control variables, R&D, Firm Size, Firm Age, Tangible Asset, Tobin’s Q, and Competition have significantly positive correlations with two proxies for firm innovation. All of the absolute values of correlation coefficients do not exceed 0.5. Although not described in Table 2, the variance inflation factors are also quite lower than 10 which is widely used as the criterion for multi-collinearity.

Table 2 Correlation coefficient matrix
Full size table

Table 3 describes the yearly distribution of our matched-sample firms. To gain more insights, we classify our sample firms into some patent classes based on the numbers of patent applications and tabulate the numbers of firm-years for each patent class. As Table 3 shows, firm-years with zero patents consist of 55.71% of our sample. Of firm-years, 11.98% apply for one patent. The proportion of firms with 3–10 and 11–100 patent applications are 13.05% and 10.28%, respectively. The remaining 2.40% comprises firm-years with more than 100 patent applications.

Table 3 The yearly distributions of sample firms
Full size table

4 Empirical results

4.1 CEO age and innovation productivity

In order to examine the relationship between CEO age and innovation productivity, we estimate the following models as:

$$INNO_{ - } PRO_{it + k} = \alpha + \beta CEO{\text{ Age}}_{it} + \gamma Controls_{it} + ID + YD + \varepsilon_{it + k,}$$
(2)
$$INNO_{ - } PRO_{it + k} = \alpha + \beta_{1} Age{\text{ under 50}}_{it} + \beta_{2} {\text{Age50 - 59}}_{it} + \gamma Controls_{it} + ID + YD + \varepsilon_{it + k,}$$
(3)

where the subscripts i and t indicate firm and year, respectively. Controls is a vector including control variables. We present k-year-ahead (k = 1 and 2) dependent variables in order to address that CEO age has a long-term effect on firm innovation and also to account for an endogenous problem due to a potential reverse causality between CEO age and innovation. We include industry-fixed effects (denoted as ID) at the three-digit KSIC codes to account for omitted industry-specific characteristics that are constant over year, and year-fixed effects (denoted as YD) to control for inter-temporal variation that may make an effect on the relationship between CEO age and firm innovation. For addressing autocorrelation of innovation over year, we also report clustered standard errors at the firm level (Petersen, 2009).

Panel A of Table 4 examines the effect of CEO age on innovation productivity in one and two years. In column (1), the coefficient on CEO Age is negatively and statistically significant. For example, column (1) suggests that a one-standard-deviation (6.889) increasing in CEO age is associated with a 3.93% decrease in patent applications, relative to the sample average.Footnote 3 The negative coefficient estimate suggests that Korean listed firms employing older CEOs apply for fewer patents compared to the previous year. In other words, Korean listed firms with younger CEOs yield higher innovation productivity. This result support the empirical studies of Barker and Mueller (2002) and Li et al. (2017) that report younger CEOs invest more in R&D.

Table 4 CEO age and innovation productivity
Full size table

In columns (2) and (3), we specify our baseline models for IT and non-IT firm subsamples. Columns (2) and (3) say that the coefficients on CEO Ages continue to be negative and significant for IT and non-IT firms, respectively. Economically, one-standard-deviation aging of CEOs in IT firms (non-IT firms) leads to a 4.40% (2.59%) decrease in patent filed at their mean values. The negative coefficient on CEO Age implies that the older the CEOs, the lower the innovation productivity, regardless of the IT industry in which the firm belongs. Furthermore, the estimated coefficient on CEO age for IT firms has more significant value than that for non-IT firms at the 5% level. This difference implies that the negative age-innovation relation is amplified for IT firms, which also indicates that CEOs in IT firms more actively apply for patents.

Panel B reports the results using three age categories (Age under 50 and Age 50–59), in which the category indicating CEOs in 60 and above is used as the benchmark. Column (1) of Panel B says that firms with younger CEOs more actively yield innovation productivity, compared to those with CEOs aged 60 and above, for the full sample. We also present the p-values calculated by a test for the equality of the two coefficients on the two categories. Of the two categories representing younger CEOs, firms with CEOs aged under 50 tend to apply for more patents than those with CEOs aged between 50 and 59. That is, in a series of results from Panel A, the lower the CEO age, the more positively impact on innovation productivity. Furthermore, the coefficient on Age under 50 in IT firms is more significant than that of non-IT firms at the 1% level, which suggests that CEOs aged under 50 in IT firms more actively apply for patents than those in non-IT firms.Footnote 4

A possible interpretation on the positive effect of young CEOs on innovation productivity is that the managerial signaling hypothesis is more compatible with IT firms. The young CEOs who have longer career horizon may be encouraged to take part in firm innovation that requires long-term efforts and has highly probability of failure, rather than some current performance-improving strategies. That is, younger CEOs in IT firms have more incentives to gain a reputation as outstanding innovators in the future, by revealing their abilities to the market through successful innovation such as patent applications. Evidence on this finding can also be seen in studies that young CEOs willingly participate in risky investment (Serfling, 2014), acquisitions (Yim, 2013), or restructuring (Li et al., 2017), even though these activities do not lead immediate rises in monetary rewards.

4.2 CEO age and innovation scope

The empirical models for exploring the relationship between CEO age and innovation scope are as follows:

$$INNO_{ - } SC_{it + k} = \alpha + \beta CEO{\text{ Age}}_{it} + \gamma Controls_{it} + ID + YD + \varepsilon_{it + k,}$$
(4)
$$INNO_{ - } SC_{it + k} = \alpha + \beta_{1} Age{\text{ under 50}}_{it} + \beta_{2} {\text{Age50 - 59}}_{it} + \gamma Controls_{it} + ID + YD + \varepsilon_{it + k} .$$
(5)

We also present k-year-ahead dependent variables in order to address that CEO age has a long-term effect on innovation scope measured as one minus the technological proximity between patent portfolios at t and t + k years. The estimation methods of coefficients and standard errors in Eqs. (4) and (5) are the same as those given in Sect. 4.1.

Table 5 reports the effect of CEO age on the innovation scope. In column (1) of Panel A, the coefficient on CEO age has negative and significant value at the 10% level. Controlling for other factors, a one-standard-deviation (6.889) aging is associated with 0.93% decrease in one minus technological proximity measure at its mean value. This negative coefficient indicates that older CEOs tend to hold more analogous patent portfolios with respect to those of the previous year. In columns (2) and (3), the coefficient estimates on CEO Ages have negative and statistically significant values for IT and non-IT firms, respectively. For example, the coefficients indicate that one-standard-deviation aging of CEOs in IT firms (non-IT firms) decreases technological scopes of their patent portfolios by 0.54% (1.49%). In other words, CEOs tend to apply for patents exploiting familiar technologies with previous experiences as the CEOs grow older.

Table 5 CEO age and innovation scope
Full size table

Panel B reports the results using the age categories. In column (1) of Panel B, the positive coefficient on Age under 50 indicates that firms with CEOs aged under 50 hold patent portfolios that are technologically more distant to the past, compared to firms with CEOs aged 60 and above. For the two categories indicating younger CEOs, CEOs aged under 50 are significantly likely to apply for patents that are technologically unfamiliar for them to the previous year. Furthermore, the coefficient on Age under 50 for IT firms is more significant than for non-IT firms at the 5% level, which indicates that CEOs aged under 50 in IT firms more actively apply for patents exploring unfamiliar technological fields than those in non-IT firms. From the findings for the innovation scope, it is possible to interpret that young CEOs take an active stance in exploration into unfamiliar fields. That is, through active exploration, younger CEOs in IT industry may have greater incentives to acquire reputations as adventurous innovators who seek novel things.Footnote 5

4.3 Alternative specifications

Through the OLS specifications, we drive the main results that younger CEOs are more motivated to participate in applying for patents or to explore unfamiliar technological fields than older CEOs. However, there is a possibility that the coefficient estimates derived may be biased due to some econometric concerns. We address these concerns in two ways in order to confirm whether our main results remain robust.

4.3.1 Tobit, Poisson, and negative binomial regressions

As shown in Table 3, more than half of sample firms (55.71%) do not apply for patents. That is, our two dependent variables using patents have left-censored distributions, so THAT it is highly possible that our OLS specifications do not derive unbiased or inconsistent estimates due to a corner solution (Baltagi, 2013). To address this distribution concern for our dependent variables, we estimate the Tobit model. In addition, since the number of patent applications has the nature of non-negative and count variable, the residuals derived from OLS may be heterogeneous or non-normal (Blundell et al., 1995). To consider this nature, we estimate count variable models, such as Poisson and negative binomial models.

Panels A, B, and C of Table 6 show the results for estimating the Tobit, Poisson, and negative binomial models, respectively. For convenience in interpreting the estimated results, we present marginal effects of CEO age and two age categories (Age under 50 and Age 50–59) at their mean values. The results of Panel A say that our main results continue in spite of estimating the Tobit model. For instance, in column (1) of Panel A, the marginal effect on CEO Age (− 0.135) indicates that a one-standard-deviation aging for the full sample leads to a 13.5% decrease in patent applications at its mean value. In column (2), the positive marginal effect implies that CEO aged below 50 more apply for 29.9% patents than CEOs aged 60 and above in next year. For IT-firm subsample, results that CEOs under 50 are likely to file more patents remain unchanged. Panel B and Panel C also provide results that do not differ qualitatively from our main results. The marginal effects on CEO Ages in Poisson and negative binomial regressions are significantly negative. All dummy variables indicating CEOs are under 50 have statistically significant marginal effects in all regressions. Our main results supporting the managerial signaling hypothesis are robust even considering the left-censored distribution of our dependent variables and the nature of non-negative and count variable.

Table 6 Alternative specifications
Full size table

4.3.2 CEO-fixed effects

Next, we examine the possibility that our CEO age results are due to omitted-variable biases that may arise from unobservable CEOs’ individual characteristics not being included in our empirical models. For instance, Bertrand and Schoar (2003) document that younger CEOs are more likely to have MBA degrees related to aggressive firm policies. To consider unobservable characteristics, following Li et al. (2017), we include CEO-fixed effects in our empirical models and present the results in Panel D of Table 5. Our results remain robust in the inclusion of CEO-fixed effects.

4.4 Additional tests

In this section, we examine detailed mechanisms that support our managerial signaling hypothesis for the young CEOs. To do this, we conduct further analyses considering overconfidence and tenures of CEOs and firm age.Footnote 6

4.4.1 Impact of CEO tenure

From our main results, younger CEOs seem to be more motivated to successfully innovate, and in doing so, to reveal their innovation capabilities in the labor market and then to gain a reputation in the future. Nevertheless, the CEO’s tenure may affect CEO age-innovation relationships. It is interesting to investigate the impact of CEO tenure since there are extensive results on it. For example, externally hired CEOs have capabilities that are aligned with environmental conditions, and they have clear ideas about how to manage their role. Then, they may become overly committed to the earlier experiences, and find it difficult to execute novel projects. This may cause their innovation productivity and innovation scope to decrease with tenure (Hambrick & Fukutomi, 1991). Meanwhile, long-tenured CEOs become less open mind, firms led by long-tenured CEOs may continue to exploit the existing technologies (Miller & Shamsie, 2001). Our intuition for exploring the impact of CEO tenure is that, if the differences in the remaining career horizons among young CEOs generate a different impact on firm innovation, that is, if the young CEOs with longer career horizons have greater incentives to innovate (Yim, 2019), then this result may be understood as a result of the managerial signaling hypothesis being more consistent with the young CEOs.

To reflect this in our model, we add a CEO tenure variable (CEO Tenure) and interaction terms between CEO age variables and tenure of CEOs (i.e., Age under 50 × CEO Tenure). In Panel A of Table 7, we find that the interaction term makes significant and negative impacts on innovation productivity and innovation scope. The negative coefficients indicate that the positive impacts of Age under 50 on the innovation productivity and innovation scope are attenuated for longer-tenured young CEOs in IT firms than for shorter-tenured young CEOs. This result implies that the incentives for young CEOs in IT firms to participate in firm innovation are greater as they have shorter tenure—that is, the longer the remaining career horizons. This can be interpreted as evidence that shorter-tenured younger CEOs, i.e., younger CEOs with longer career horizons, in IT firms are more motivated to signal their innovativeness to the market to be regarded as outstanding or adventurous innovators.

Table 7 Additional tests
Full size table

4.4.2 Impact of firm age

There is also extensive evidence that innovation depends on firm age (e.g., Coad, 2018). In order to isolate the inherent CEO age-innovation relationships from the correlation between CEO age and firm age, we further control firm age and an interaction term between CEO age and firm age. Panel B of Table 7 shows the estimated results of them. We find that firm age has a positive effect on firm innovation. Nevertheless, the estimated coefficients of the interaction term (CEO Age × Firm Age) are statistically insignificant. For IT firms, moreover, the impact of the interaction term is amplified. These results indicate that the firm age does not affect our age-innovation relationships.

4.4.3 Impact of CEO overconfidence

Overconfidence of a CEO may make a positive impact on firm innovation since it reduces fear for failure in innovation (Galasso & Simcoe, 2011; Hirshleifer et al., 2012). We examine whether our findings supporting the dominance of the managerial signaling hypothesis are driven by younger CEOs’ overconfidence. To this end, we include interaction terms between CEO age variables and a dummy variable indicating that the CEO is overconfident.Footnote 7 Panel C of Table 7 shows that CEO overconfidence (Overconfidence) positively affects firm innovation, in line with the results of Galasso and Simcoe (2011) and Hirshleifer et al. (2012). We find that, however, the interaction terms between age variables (CEO Age and Age under 50) and CEO overconfidence are not statistically different from zero in all columns of Panel A. These results indicate that younger CEOs in IT firms have incentives to actively participate in firm innovation, regardless of overconfidence. Therefore, our results supporting the managerial signaling hypothesis are not driven by the CEO’s overconfidence.

5 Conclusions

This paper investigates the relationship between CEO age and firm innovation, proxied with the number of patent applications and the technological proximity, for firms listed on Korean stock markets between 2002 and 2016. We document that a firm’s innovation productivity or innovation scope decreases in CEO age. Furthermore, we also report that younger CEOs in IT firms are more likely to take part in firm innovation than those in non-IT firms. Our results remain robust in Tobit, Poisson, and negative binomial regressions. Omitted CEO characteristics and differences in characteristics of firms with young versus old CEOs cannot explain our CEO age-innovation relations. We present further evidence on the managerial signaling hypothesis, by showing that incentives for younger CEOs to innovate are not driven by their overconfidence and that shorter-tenured younger CEOs are more likely to engage in firm innovation than longer-tenured younger CEOs. In investigation for Chaebol firms, accounting for the special nature of Korea, we also get the result that young CEOs in IT firms actively participate in firm innovation. Taken together, younger CEOs in IT firms may be more motivated to signal their innovativeness to the market to be regarded as outstanding innovators or adventurous innovators who seek novel things, which supports managerial signaling hypothesis for young CEOs.