Where does pride lead? Corporate managerial hubris and strategic emphasis

Scholars have used the terms “hubris” and “overconfidence” interchangeably (Hayward et al. 2006; Malmendier and Tate 2008; Li and Tang 2010). Another related concept is narcissism. Extant research argues that narcissistic individuals differ from hubristic ones in their strong need for external applause and affirmation; and hubris is a mental state, whereas narcissism is a type of personality that remains persistent (Chatterjee and Hambrick 2007; Li and Tang 2010; Tang et al. 2015b). Hence, measures of narcissism are largely based on an executive’s prominence or relative compensation (e.g., Chatterjee and Hambrick 2007; Kashmiri et al. 2017). We focus only on hubris in this study to develop theories and measures. Future research can compare the roles of narcissism vs. hubris in this context.

For example, some prior studies examined the factors that predict successful introductions of radical new products, such as the willingness to cannibalize (e.g., Chandy and Tellis 1998) and CEO narcissism (e.g., Kashmiri et al. 2017).

Tang et al. (2015a) specified a model in which both managerial hubris and R&D spending influence patents/new product sales, i.e., R&D intensity is treated as an explanatory variable instead of the dependent variable. Hence, their model is completely different from ours, which examines the impact of hubris on R&D spending. Furthermore, Tang et al. (2015a) reported an insignificant model-free correlation between hubris and R&D intensity based on a sample of Chinese firms, whereas we estimated a panel data model and affirm a significant effect of hubris on R&D intensity based on publicly listed U.S. firms.

Simon and Houghton (2003) examined small computer companies only, whereas Li and Tang (2010) and Tang et al. (2015a) studied Chinese firms in the manufacturing industry.

A firm’s maturity is not solely determined by its age but reflects the extent to which a firm is growing or being stagnant (Anthony and Ramesh 1992). Therefore, firm maturity is measured as a composite score of various signals (including firm age) of firm vitality. A detailed scoring system is described in the “Data and measures” section.

Loughran and McDonald (2011) developed a dictionary of positive and strong modal words (available at www3.​nd.​edu/​~mcdonald/​Word_​Lists.​html) by combining the words they identified in all 10-K filings from 1994 to 2008 and those words in the Harvard Psychosociological Dictionary (available at www.​wjh.​harvard.​edu/​~inquirer/​). A complete list of the positive and strong modal words is available at www3.​nd.​edu/​~mcdonald/​Word_​Lists.​html.

We thank the review team for the suggestions on checking face validity. The sequence of the press releases is randomized for each respondent to minimize potential anchoring bias (i.e., a subject’s rating may be biased depending on his/her experience of rating press releases presented earlier in the survey). In addition, the respondents have the option to go back to review the press releases that they have already rated and revise their ratings.

Not all firms listed in ExecuComp are included in the final sample because not all firms are included in the Investor Responsibility Research Center (IRRC) database, which provides data for corporate governance (one of the proposed moderators). The IRRC firm list is based on S&P 500 plus the largest corporations included in the annual lists of Forbes, Fortune, and Businessweek. It covers a majority of the value-weighted market (IRRC tracked over 93% of the total capitalization of NYSE, AMEX, and NASDAQ combined in 1990 and was expanded by several hundred in 1998). We conducted t-tests for advertising, R&D, and strategic emphasis between the two samples before and after removing the observations with missing corporate governance data. The t-test results indicate that the means of the key variables are not significantly different across the two samples, and our results are unlikely to be subject to sample selection. We also used the original sample based on ExecuComp to test H1, H2, H4, and H5 without including corporate governance as a moderator. We found consistent results on these hypotheses. Therefore, our results are not biased because of the loss of observations.

Our corporate managerial hubris has a mean of 2%, a between standard deviation of 0.5%, and a within standard deviation of 0.3%.

Anthony and Ramesh (1992) also proposed capital expenditure as an alternative indicator of firm maturity but found that it has relatively lower explanatory power. Thus, we did not use it in the main analysis. However, as a robustness check, we used it as an alternative measure to rerun the models and found substantially consistent results.

Based on the sum, Anthony and Ramesh (1992) further grouped the firms into five different stages of corporate life cycle, and assigned each stage a 1 to 5 score to indicate firm maturity. We found consistent results when using the 5-stage measure to replace the original summated score. Therefore, our results are robust to the alternative operationalization of firm maturity.

We also tested the models with the market capitalization log (i.e., log [total number of shares outstanding × closing price]) as a proxy for firm size. The results on the hypothesized variables remained consistent.

Josephson et al. (2016) examined a firm’s strategic slack, which is measured as the total revenue divided by R&D spending. We did not include strategic slack as a control variable for two reasons. First, strategic slack is the reciprocal of R&D intensity and is not conceptually meaningful to predict R&D intensity with its reciprocal. Second, including strategic slack reduces sample size because strategic slack becomes a missing value when R&D spending (the denominator) is zero. That said, when including strategic slack as an additional control variable in our model predicting strategic emphasis, we found consistent results on the hypothesized effects.

Kleibergen-Paaprk LM test indicated that the model is identified.

We tested the qualification of the IV (i.e., peer firms’ hubris) following Germann et al. (2015) and found that the instrument is a significant predictor of a focal firm’s hubris in the first-stage regression based on the Cragg-Donal Wald F-statistic (F = 108.4, p < 0.001). Moreover, we tested our model using additional instruments to check the robustness of the results to the selection of IVs. We added the average tenure of the executives (data obtained from ExecuComp) and CEO’s education level (data obtained from S&P NetAdvantage) as instruments because they may be associated with the executives’ hubris (e.g., Finkelstein and Hambrick 1990; Hayward and Hambrick 1997) but do not directly influence a firm’s relative focus on R&D versus advertising. The inclusion of these instruments yielded consistent results.

We thank an anonymous reviewer and the associated editor for their suggestion on the new IV.

We estimate the random-effects instead of fixed-effects model because some of our focal moderators exhibit low variability across time within the firm, making the fixed-effects model inappropriate. Unlike the fixed-effects model, the random-effects model cannot independently account for time-invariant firm-specific unobservables. Therefore, when estimating the random-effects model, we corrected for the potential bias of unobservables using 2SLS with the instrument variable.

We acknowledge that R&D and product innovation are vital for firm survival and growth. However, excessive focus on R&D and value creation, particularly without complementary value appropriation efforts, can have an adverse effect on a firm’s performance.