Rock your Thesis Logo
beta
search
menu

Academic Phrasebook

Our Academic Phrasebook provides you with sentence blocks to use in academic writing. It is organized according to the sections of your thesis structure, so that you can filter them by your current work status. Creativity and adjustments will be necessary to use the phrases in the context of your thesis.

Academic Phrasebook

Introducing the topic

This research illustrates how __________ can be leveraged to spur __________.
content_copy
We examine whether __________fall following __________.
content_copy
We study __________ to recognize __________ through __________.
content_copy
This research examines the effects of __________on __________.
content_copy
We study a __________ between __________, __________ in the presence of __________.
content_copy
In this paper, we study __________ and contrast these to __________.
content_copy
The success of __________ often depends on the ability to scale __________.
content_copy
To constrain the use of __________, many U.S. state governments adopted __________. __________reduce __________ from creating __________.
content_copy
In the presence of __________,arbitrageurs may engage in different strategies leading to __________.
content_copy

Stating the research gap

Specifically, this research introduces a previously overlooked yet practically relevant dimension on which __________ differ: __________.
content_copy
The lack of __________ has been at the heart of a number of on-going accounting debates.
content_copy
Existing studies have focused on a variety of approaches that __________ may use to scale but have not systematically considered that __________.
content_copy

Stating the methodology

Seven experiments, two of which were conducted in the field, support this hypothesis and illustrate conditions under which the effect __________.
content_copy
We depart from signaling theory and derive our hypotheses on the effects of __________. To test them, we implement a quasi experiment in the context of the __________.
content_copy
Across six studies, using both observational analyses of __________ and experimental manipulations of __________, the auhors show that __________increase __________, which improves __________.
content_copy
Using a sample of __________, we examine the effect of __________ on __________.
content_copy
Our hypothesis development departs from __________. One may argue that __________. We test the hypotheses in the context of __________.
content_copy
Using data from __________ we study __________, we compare __________.
content_copy
To do so, I construct __________. This framework models __________.
content_copy
Our empirical analyses employ a sample of __________ from __________. To measure __________. This output measure comprehensively captures both __________.
content_copy
For each __________, we run __________.
content_copy
We scrape __________ of __________ to measure __________. To measure __________. We focus on __________ because we expect that __________. Though __________. We implement __________ as our treatment to test our hypotheses.
content_copy

Reporting the findings

This research demonstrates that __________ can be more effective than __________.In particular, __________ are more likely than __________.
content_copy
Using __________, we find that __________negatively affects __________.
content_copy
Overall, our findings are consistend with __________negatively affecting __________.
content_copy
We report three findings First, __________.Second, __________.Finally, __________.
content_copy
We find that __________ when compared with __________, and they often __________.
content_copy
First, __________.Second, __________.Third, __________.
content_copy
We find that __________. We also find that __________.
content_copy
We find that: 1) __________, 2) __________, 3) __________. Our results indicate that __________.
content_copy
We provide evidence for __________. Results show that __________. We also identify theoretically and substantively relevant boundary conditions: __________.
content_copy
A key insight in this paper is that __________. This is because __________.
content_copy
These results can explain __________. They therefore shed light on a number of highly debated __________, such as (a) __________; (b) __________; and (c) __________. The analysis here suggests that __________.
content_copy
__________, we find that __________.
content_copy
Overall, our findings suggest that __________.
content_copy
Consistent with this expectation, we find that __________ has a significantly negative sentiment beta.
content_copy
In this paper, we document a positive __________ between __________. The relation is both statistically and economically significant based on a sample of __________ over the period __________.
content_copy
Our model demonstrates __________ and characterizes __________.
content_copy
Our main results are consistent with __________.
content_copy
In cross-sectional tests, we find results consistent with __________. We also predict and find that __________. Next, we predict and find that __________. Our results are consistent with __________.
content_copy
Additionally, we find evidence that __________. In an additional test, we find no evidence that __________. We also find evidence that __________.
content_copy

Stating the implications

On the whole, this work offers insights into __________, and it reveals __________.
content_copy
We contribute to __________ by __________.Additionally, our findings have theoretical implications for understanding __________.
content_copy
The results provide novel implications regarding __________. These include, among others, that __________.Additionally, __________.
content_copy
Overall, our findings suggest that __________.
content_copy
Our findings show that __________ by __________.
content_copy
Thus, this research lays the foundation for practitioners to understand __________.
content_copy
Thus, __________ - which managers can leverage to influence __________.
content_copy
This research provides important insights that are relevant to managers. Most importantly, the effects of __________ provides guidance to managers when constructing __________. __________. We find evidence that __________. However, we also show that __________.
content_copy
These results can add to our understanding of __________. The analysis also implies that in order to induce managers with __________, it is optimal to __________.
content_copy
Finally, we believe that our study informs policy makers who are interested in __________.
content_copy
Our results suggest that, beyond simply __________.
content_copy
Our result instead suggests that __________.
content_copy
Our results offer new insights and testable implications concerning how __________.
content_copy
The analysis gives rise to a number of novel insights and implications concerning the __________.
content_copy

Opening the thesis

A key challenge for firms in charge of __________ is __________. As noted by __________, __________ can __________.
content_copy
There is a growing prevalance of __________ that act as __________. The success of __________ is often shaped by __________.
content_copy
One of the most important and underestimated aspects of __________ is its __________. In most instances, __________.
content_copy
__________ is arguably the most influential means of persuasion and is one of the main predictiors of __________. Studies by __________find __________, and other reports suggest that __________. Indeed, __________consumer spending per year (Analytic Partners 2014), __________.
content_copy
Firms are not requried to disclose __________. In __________, the __________ determines the primary qualities of __________ and requires that the __________.
content_copy
The past three decades have witnessed significant increases in __________. As one of the most important __________,U.S. firms __________to shift __________.
content_copy
Private firms that wish to __________ often face uncertainty __________. This is particularly salient for firms in new industries for which __________.
content_copy

Outlining key terms, concepts and theories

__________ are __________.
content_copy

Explaining the practical relevance

Although the practice of __________ has become business as usual for __________, little systematic evidence exists regarding its potential effects. This raises the question of whether __________ may lead to undesirable outcomes for the __________.
content_copy
There are numerous examples of __________ making __________. In some cases, these are __________ to grow __________.For instance, __________. However, there are also examples of __________. While on the one hand, some of these examples are consistent with the ethos of __________, the remaining examples appear inconsistent with this logic.
content_copy
There are prominent examples of this phenomenon: When __________.Similarly, when __________.
content_copy
Given its persuasiveness, marketers invest in campaigns to stimulate __________. These campaigns typically take the form of __________ or __________ and involve giving consumers an incentive __________ to generate __________.
content_copy
For example, __________.
content_copy
Consequently, firms invest considerable resources in developing __________ to create __________.Moreover, as __________ are introduced daily, securing __________ is a growing challenge due to the __________. Perhaps for this reason, firms are increasingly turning to brand naming consulting firms to create __________.
content_copy
For example, __________.
content_copy
In the light of the rapid increase in __________ of firms in recent decades, there is an on-going debate about __________.
content_copy
Relatedly, there is an on-going discussion among __________ regarding __________.
content_copy
For example, __________. As a result, __________.
content_copy
__________, however, is costly as this may result in the loss of __________.
content_copy
A well-known example is __________. These examples indicate that __________.
content_copy
However, anecdotal and survey evidence implies that __________.
content_copy

Explaining the theoretical relevance

Work on __________ has investigated various mechanisms, including __________.
content_copy
Exisiting studies have identified a variety of approaches __________ may want to use __________.
content_copy
__________ are an important strategic activity of many companies and have drawn considerable scholarly interest in understanding __________.
content_copy

Demonstrating the research gap

Extant research is relatively silent on this issue. Work on __________ has investigated various mechanisms, including __________. However, these studies primarily assess __________. By contrast, __________ remain less understood.
content_copy
Existing studies have identified a variety of approaches that __________. One aspect that has not been systematically considered is __________. As a result, there is little understanding of __________.
content_copy
We focus on an overlooked but practically relevant dimension __________.
content_copy
To date, research on __________ has focused almost entirely on __________. However, __________. We use this expanded view to examine the conditions under which __________.
content_copy
A large theoretical and empirical literature examines the impact of __________. In this paper, we focus instead on __________.
content_copy
Unlike existing models of IPO timing __________ we consider __________.
content_copy
The extant literature on __________ does not generally focus on __________. However, anecdotal and survey evidence implies that __________. We extend prior literature by providing evidence that __________.
content_copy

Stating the research question or hypothesis

In this paper, we seek to understand __________. We investigate this question from the perspective of __________.
content_copy
In this paper, we study __________, and specifically focus on the question of __________.
content_copy
In this paper, I reflect on two broad questions related to __________: Q1 - __________? and Q2 - __________?
content_copy
We first hypothesize that __________. Second, we hypothesize that __________.
content_copy

Stating the purpose

In this work, we propose that __________. Specifically, we focus on the use of __________, a term we use to refer to __________.
content_copy
In the present research, we propose __________.
content_copy
The primary objective of this paper is to provide __________. Toward this end, this paper explores __________.
content_copy
In this study, we analyze __________. Specifically, we examine whether __________ and, if so, whether __________.
content_copy
Specifically, we investigate __________.
content_copy
We seek to study __________.
content_copy

Stating the contributions

Our study offers three theoretical contributions. The first and principal one is toward research on __________. The present work extends our understanding of one __________. Second, our study adds to the growing research on __________. Finally, this study adds to the literature on __________.
content_copy
This paper makes several contributions. First, existing studies have focused on a wide range of __________, but have not considered that __________. However, as our analysis shows, __________. This paper is among the first to consider __________, showing that __________.
content_copy
Our work contributes to both theory and practice by providing insights into __________. In doing so, this research contributes to the __________. Moreover, it contributes to the __________. Specifically, it suggests that __________.
content_copy
Our research makes important theoretical and substantive contributions. First, we contribute to the literature on __________. We introduce a more nuanced approach to studying how __________, distinct from prior research, which has focused primarily on __________. Second, this work adds to the __________ by demonstrating that __________.
content_copy
This study makes several significant contributions to the __________. First, our study contributes to the growing literature on __________. Prior studies focus on __________. Our findings suggest that __________. Thus, our findings help understand __________. Second, our study is related to the broad literature on __________, and more specifically the literature on __________.
content_copy
We contribute to the literature by providing evidence that __________ has an economically significant effect on __________.
content_copy
Our comprehensive analysis of __________ contributes to several strands of literature. First, we extend the seminal work of __________, who focus on __________, by providing evidence on __________. Consistent with __________, our evidence shows that __________. Importantly, we find that __________. Our results therefore suggest that __________. Our study also contributes to the literature by __________. Existing research documents __________. We provide strong evidence of __________. Our results suggest that, __________. Finally, our paper adds to the growing literature on __________. With few exceptions, most studies use __________. While our focus is on __________, we find that __________.
content_copy
Our study contributes to the literature on __________. Though we focus on __________. Prior studies in __________.
content_copy
We also contribute to the literature on __________. __________ finds evidence that __________. __________ examine __________. They find results consistent with __________. __________ find results consistent with __________. We contribute to this literature by providing evidence that, __________.
content_copy

Outlining the structure

This paper is structured as follows. In Section 2, __________. __________ in Section 3. In Section 4, we provide __________. In Section 5, we provide __________. In Section 6, we discuss __________.
content_copy
The paper proceeds as follows. The next section describes __________. Section 3 derives and analyzes __________. Section 4 provides __________. __________ in Section 5. __________ appear in the appendix.
content_copy
We organize the remainder of the paper as follows. Section 2 provides __________. Section 3 describes __________. __________ in Section 4 and __________. Section 6 concludes.
content_copy
The paper proceeds as follows. Section I describes __________. Section II presents __________. Section III investigates __________. Section IV concludes. __________ are provided in the Internet Appendix.
content_copy
The following section presents __________. Section 4 examines __________, while Section 5 presents __________. Section 6 discusses __________. __________ the Appendix, unless otherwise stated. __________ in the Online Appendix.
content_copy

Opening the theoretical background section

One strategy by which firms can profit from __________ is the __________.
content_copy
The term __________ is used to describe __________, as well as for __________.
content_copy
We propose that __________ differ in terms of __________. We use the term __________ to refer to __________, such as __________.
content_copy
The rapid increase in __________ by U.S. firms over the last three decades has attracted significant attention from researchers, politicians, and the public. For example, __________.
content_copy

Defining key terms, concepts and theories

The term __________ thereby refers to __________ that becomes more __________.
content_copy
The term __________ refers to __________.
content_copy
We use the term __________ to refer to __________.
content_copy
In this paper, we use the term __________ to refer to __________. These are companies which act as __________, rather than companies __________. This can include __________. We are only focusing on companies __________. Therefore, we use the term __________ throughout this paper.
content_copy
As a comparison group, we are referring to __________. Examples of __________ are __________.
content_copy
A key characteristic of __________ is that they are characterized by __________.
content_copy
__________ coined the term __________ to refer to this __________ and posited that it is __________.
content_copy
An omniscient observer econometrician is one who observes the true long-run mean θt of inflation.
content_copy
Similarly, in their application of the C-G framework, Bordalo et al. (2020) state that “Tests of individual beliefs are informative about departures from rationality.
content_copy

Comparing and contrasting existing definitions and sources

Our study relates to research on platform governance (Tiwana et al. 2010, Wareham et al. 2014, Parker and Van Alstyne 2018). This stream of research is based on the notion that platform owners’ activities go beyond development and marketing. Instead, the platform firm is in charge of a microeconomy, and therefore needs to design mechanisms that encourage desirable behaviors and outcomes.
content_copy
Multihoming refers to the decision to operate a complement on other platforms (Caillaud and Jullien 2003, Tanriverdi and Lee 2008, Landsman and Stremersch 2011). Multihoming contrasts with singlehoming, which is operating a complement only on the focal platform.
content_copy
The different strategies employed by complementors also have consequences for the platform firm (Bhargava et al. 2013, Foerderer et al. 2018). Arguably, new complement development is vital for platform firms because it can bring about innovation. Complement improvements are also important because they increase the quality of the complements and ensure that complementors address the needs of an existing user base. By contrast, multihoming is an undesirable outcome because it reduces platform firms’ differentiation from their rivals (Landsman and Stremersch 2011, Kang et al. 2019).
content_copy
Key to our research, prior literature suggests that sharing may also arise from an altruistic desire to help the company (Hennig-Thurau et al. 2004; Sundaram, Mitra, and Webster 1998).
content_copy
Consistent with Ijiri's (1975) view that softness of a measure captures how easily the measure can be “pushed in one direction or another”, we interpret lower levels of P as corresponding to ‘softer’ information.10 Similar to Glover et al. (2005), Roger (2013), and Bertomeu and Marinovic (2015), the information in this model is considered soft in that it cannot be verified with probability one.
content_copy
Prior literature on the effect of tax policies on innovation can be summarized into two streams. The first stream of studies examines the effect of R&D tax credits and allowances on R&D investments (e.g., Bloom et al., 2002; Hall, 1993; Hines, 1994; Hines and Jaffe, 2000; Rao, 2016). While most prior studies focus on country-level R&D tax credits, Wilson (2009) examines how state-level R&D tax credits affect firms' innovation behavior. He finds that the R&D tax credit in a state increases a firm's R&D investment in that state mainly by drawing the firm's R&D projects from other states. Thus, state R&D tax credits do not have a substantial effect on a firm's nationwide innovation. Another streamexamines the effect of statutory tax rates on innovation. Theoretical models (e.g., Hall and Jorgenson,1969; Jorgenson, 1963) predict that if R&D investment is tax deductible, statutory tax rates are not expected to have a major impact on R&D projectsdbecause the tax benefits from R&D investment deductions are canceled out by the taxes on taxable income. However, recent empirical studies (e.g., Atanassov and Liu, 2019; Mukherjee et al., 2017) find that higher state tax rates negatively affect patenting and R&D investment because of lower after-tax income available for future investment.
content_copy
Empirical studies of persistence in forecast errors are increasingly stepping back from these theoretical foundations and are framing their null hypotheses around the inference strategy proposed by Coibion and Gorodnichenko (2015) for their noisy-signal environment.
content_copy
In contrast, C-G directly posit that their variable of interest Pt is observable only up to a noisy signal St . This framing does not seem well matched to the learning challenges in bond markets, because yields are observed essentially without error.
content_copy
An undercurrent that runs throughout the literature is the connection between the survey respondents who form the basis of empirical work and the agents whose dispersion of beliefs in theory impact equilibrium bond prices. Within a noisy-signal framework, Patton and Timmerman (2010) argue that disagreement about macroeconomic fundamentals is better explained by heterogeneous priors versus heterogeneity in forecasters’ information signals, although the admissible priors did not encompass nonnested autoregressive dynamics. Bacchetta and Wincoop (2006) present evidence that disparate information sets across market participants explain several anomalies of exchange rate dynamics.
content_copy
Lamont (2002) argues that principal-agent problems incentivize older and more established forecasters to produce more radical forecasts. Similarly, Laster, Bennett, and Geoum (1999) argue that professional forecasters behave strategically, with those whose wages are tied to publicity-giving forecasts that are relatively far from the consensus (average).
content_copy
Prior studies examine whether tax avoidance affects firm and/or managerial reputations (see Wilde and Wilson 2018 for a review). Hanlon and Slemrod (2009) use an event study to document that shareholders react negatively to news about their firms’ involvement in tax shelters. They interpret their results as consistent with the costs (including reputation effects) of tax sheltering outweighing the shareholder wealth benefits. Moreover, they find more negative returns in retail firms compared to other firms, consistent with firms with marginally higher reputation effects facing stronger negative reactions to tax sheltering. On the other hand, Gallemore et al. (2014) find that the short-run effect documented in Hanlon and Slemrod (2009) reverses within 30 days. Similarly, Dhaliwal et al. (2019) provide evidence that firm value decreases in periods of negative market sentiment related to tax avoidance.
content_copy
Other studies examine whether other stakeholders react negatively to tax avoidance and news of tax avoidance. Gallemore et al. (2014) find little evidence that consumers respond negatively to firms’ tax shelter news using both sales and sales growth to measure consumer reactions. Similarly, Austin and Wilson (2017) find mixed evidence that firms with strong consumer brands engage in less tax avoidance. They argue that firms with strong consumer brands are expected to engage in little tax avoidance because they have the most to lose from negative publicity regarding their tax avoidance. Dyreng, Hoopes, and Wilde (2016) use publicity of firms’ international tax avoidance to document that firms reduce tax avoidance activities in response to public pressure.
content_copy
Chyz and Gaertner (2018) and Lanis et al. (2018) examine whether boards respond negatively to tax avoidance. Chyz and Gaertner (2018) hypothesize that “too much” or “too little” tax avoidance (relative to industry peers) contributes to boards’ decisions to fire CEOs. By contrast, Lanis et al. (2018) find that CEOs are rewarded for tax avoidance with increased outside board seats. They interpret their results as consistent with tax avoidance enhancing CEO reputations. Gallemore et al. (2014) find no evidence that news about firm participation in tax shelters increases CEO turnover.
content_copy

Explaining how your thesis fits to existing sources

Building on prior literature, we distinguish between three main strategies: new complement development, complement improvements (i.e., updates or enhancements), and multihoming (e.g., Bhargava et al. 2013, Cennamo et al. 2018). New complement development is the release of products that have not previously been released by the complementor to the market, in terms of being new to the complementor. Furthermore, new complements may be aimed at new markets or at customer groups not previously addressed by the complementor (Bhargava et al. 2013). New complement development is costly and risky because it requires market research, the generating of new ideas, prototyping, development, testing, and market launch (Brown and Eisenhardt 1995).
content_copy
The present study is also related to the extant theoretical literature on IPO timing, which largely consists of models that include either a single rm or multiple rms that move in an exogenously determined order.
content_copy

Describing findings of existing sources/results/definitions

There is growing research that takes the complementors’ perspective (Tanriverdi and Lee 2008, Bhargava et al. 2013, Cennamo et al. 2018). Some studies have investigated multihoming and updating (Cennamo et al. 2018, Kang et al. 2019). Cennamo et al. (2018), for example, found that the characteristics of the platform technology determine complementors’ tendency to multihome. Tiwana (2015) reports that increased rates of updating are related to platform control and the modularity of the complement. There is growing research that takes the complementors’ perspective (Tanriverdi and Lee 2008, Bhargava et al. 2013, Cennamo et al. 2018). Some studies have investigated multihoming and updating (Cennamo et al. 2018, Kang et al. 2019). Cennamo et al. (2018), for example, found that the characteristics of the platform technology determine complementors’ tendency to multihome. Tiwana (2015) reports that increased rates of updating are related to platform control and the modularity of the complement.
content_copy
There is some empirical evidence supporting this line of reasoning. Cennamo et al. (2018) find that complementors tend to release their complements on different platforms sequentially, potentially because of the costs linked to simultaneous launches.
content_copy
There is a vast literature that has studied why companies make acquisitions (Ahuja and Katila, 2001; Zollo and Singh, 2007; Capron and Shen, 2007; Wang and Zajac, 2007;). The explanations considered include: to realize cost synergies against increased competition, to increase pricing power, to address gaps in product mix and asset concentration, to create value more broadly (Bettinazzi and Zollo, 2017; Choi and McNamara, 2017; Maksimovic, Phillips and Prahala, 2008; Fan and Goyal, 2006). Other studies have looked at general antecedents to these acquisitions such as environmental factors (Berchicchi et al., 2017) and firm characteristics (Zhao, 2009).
content_copy
Studies have also identified behavioral factors driving acquisitions such as managerial self-interest (Halebian et al., 2009). An important factor identified by this literature is that platform companies may want to control key assets, in order to avoid being “held up” by critical suppliers (Silverman, 1999) that have bargaining power. This market power in one subsystem encourages companies to bundle with other subsystems to increase control and add more value (Pagani 2013) and this explain why companies may eventually want to control their suppliers1.
content_copy
A survey confirmed that consumers perceived these perks to vary in their contractuality (see Web Appendix A).
content_copy
Several streams of research suggest that experiencing high relational value induces people to engage in behaviors aimed to support and benefit others. According to sociometer theory (Leary and Baumeister 2000), individuals have a sociometer that monitors the social environment for cues about their relational value (i.e., signals that people appreciate a relationship with an individual).
content_copy
Prior research has identified a plethora of psychological motives that drive sharing (for a review, see Berger 2014), including the desire to selfenhance (De Angelis et al. 2012), restore control (Peluso et al. 2017), signal expertise (Packard and Wooten 2013), and foster social connections (Chen 2017; Dubois, Bonezzi, and De Angelis 2016).
content_copy
Research on the Stereotype Content Model, and gender research more generally, finds that women are associated with warmth to a greater degree than are men (Fiske 2010). People rate women as warmer than men (Eckes 2002), and on personality inventories, women rate themselves as warmer than men rate themselves (Costa, Terracciano, and McCrae 2001). Similarly, people evaluate brands in terms of warmth. Consumers often interact with brands as if they were people (Fournier 1998) and imbue brands with personality traits (Aaker 1997).
content_copy
Kervyn, Fiske, and Malone (2012) developed a framework that adapted the Stereotype Content Model to brands. They demonstrated that consumers assess brands in terms of warmth and are inclined to prefer brands that convey warm qualities. In particular, they found that “well-intentioned” (i.e., warmer) brands were rated higher on purchase and brand loyalty intentions.
content_copy
Gupta and Mills (2002) and Ma and Thomas (2020) both find that U.S. firms' state tax avoidance behavior reduces their state effective tax rates by approximately 3 percent (relative to the state statutory tax rate).
content_copy
Empirical studies find that firms use discretion when pricing intrafirm transactions to shift income to low-tax areas. Clausing (2003) finds that U.S. multinational firms use lower prices in intrafirm international transactions compared with their transactions with unrelated parties. Harris (1993) and Rego (2003) also find that U.S. multinational firms use foreign subsidiaries to avoid paying domestic taxes. Grubert (2003) estimates that the use of intangible assets accounts for half of the income shifted from high-tax to low-tax countries by U.S. multinational firms. Klassen and Laplante (2012) further show that such income shifting became even more aggressive in recent years.
content_copy
Le, Singleton, and Dai (2010) show that (3) is a good approximation in a representative agent model with habit-formation preferences, and Xiong and Yan (2009) find that the affine representation (2) is an accurate approximation to the actual pricing relation for their model with belief heterogeneity.
content_copy
Joslin, Singleton, and Zhu (2011) show within a canonical Gaussian model with fixed parameters 0 and no constraints on the MPR that the no-arbitrage structure of a dynamic term structure model has no impact on the estimation of the VAR parameters 0.
content_copy
There are numerous examples in the literature of situations in which transient effects can persist over long periods. Collin-Dufresne, Johannes, and Lochstoer (2016) find this in a model with rare disasters. Complementary research in progress by Farmer, Nakamura, and Steinsson (2021) documents persistent effects of learning on the biases and serial dependencies of consensus survey forecasts. Giacoletti, Laursen, and Singleton (2021) illustrate slow convergence within a constant-parameter version of the three-factor model adopted by BE (see their Figure 1).
content_copy
Recent studies suggest that employee perceptions of firms and managers affect firm value and success.5 For example, Guiso et al. (2015) find that firm performance is increasing in employee perceptions of managers as trustworthy and ethical. Edmans (2011) finds that stocks of firms with satisfied employees experience abnormal returns in future periods. Similarly, Gartenberg et al. (2018) find that firms in which employees believe in the purpose of the firm have stronger accounting and stock market performance. Thus, employee perceptions of managers and firms matter for organizational success.
content_copy
Hales, Moon, and Swenson (2018) and Huang, Li, and Markov (2019) also find results consistent with Glassdoor.com employee ratings on business outlook predicting earnings surprises and other income statement information. Similar to our study, Makridis and Zhou (2019) use changes in employee ratings of firms, managers and other variables as a measure of employee perceptions. Their findings are consistent with employees reacting negatively to accounting fraud and negatively perceiving their employing firms.
content_copy

Providing (practical) example

For example, Kov´acs and Sharkey (2014) observed that books attract more readers after winning an award. Hendricks and Singhal (1996) find that after receiving an award, firms experience an increase in market valuation and, among investors, a decreased perception of systematic risk faced by the firm.
content_copy
For example, Chen et al. (2020) find that firms whose competitors receive an award are more likely to show “catch-up” behaviors by expanding into their competitors’ product markets.
content_copy
Several extant empirical studies have documented such demand spillovers (e.g., Foerderer et al. 2018, Liang et al. 2019, Song et al. 2020). For example, Foerderer et al. (2018) observe that Google’s entry into the niche for photography apps on its Android platform attracted more new app releases to that market niche, because complementors expected to benefit from demand and attention spillovers. Liang et al. 2019 report that competitors of an endorsed complement benefitted from demand spillovers. The expectation of such spillover may attract entrants.
content_copy
To illustrate, consider these examples. Dunkin’ and Subway have required consumers to spend a set sum of money or make a certain number of purchases to earn free items. This framing is higher in contractuality, as the consumer is explicitly required to perform an action to receive the perk. In contrast, Chick-fil-A and Panera have awarded consumers free food or beverages after repeated purchases, but without any explicit contract as to when these perks would be awarded. This framing is lower in contractuality, as the consumer is not explicitly required to perform an action to earn the perk. Rather, the company observes the consumer and chooses when to provide the perk. As these examples illustrate, a perk can be identical in its cost to the company, provide the same benefit to the consumer (e.g., a free meal), and require the same amount of consumer effort (e.g., number of visits) but still vary in contractuality depending on how the company awards it.
content_copy
As another example, consider gifts such as birthday freebies. Some companies (e.g., Starbucks, Au Bon Pain) have required consumers to come in on the day of their birthday or spend a minimum amount of money to obtain the freebie, resulting in offerings that have a high degree of contractuality. In contrast, other companies (e.g., Pete’s Coffee, Sephora) have given consumers a free item for their birthday with limited or no contingencies— for example, it can be used even after one’s birthday—resulting in offerings that are lower in contractuality. Again, although such perks might have the same value (e.g., a free coffee) and require the same effort from the consumer, their perceived contractuality differs as a function of the way the company structures them.
content_copy
Prior literature suggests that people experience heightened relational value when others engage in intentional acts that signal benevolence toward them (Canevello and Crocker 2010; Lemay, Clark, and Feeney 2007) and are devoid of ulterior motives (Schopler and Thomposon 1968; Sirdeshmukh et al. 2002).
content_copy
For example, there is no obvious reason to consider the Nestl´e or Gap companies as feminine or masculine. Yet these names differ on three key properties: Nestl´e has more syllables than Gap, with stress divided between two syllables rather than one, and Nestl´e ends with a vowel sound, whereas Gap endswith a consonant. Thus,Nestl´e is linguistically feminine, whereas Gap is linguistically masculine. We propose that these automatically processed linguistic qualities will lead to implicit associations with feminine or masculine attributes, independent of explicit brand associations.
content_copy
The Institute on Taxation and Economic Policy (2017) shows that the average state effective tax rate was only 2.9 percent for 258 profitable Fortune 500 corporations in 2015, which is considerably lower than the average statutory state corporate tax rate of about 6.25 percent. Also, 92 out of the 258 profitable Fortune 500 corporations paid no state income tax in at least one year from 2008 to 2015.
content_copy
For example, if a subsidiary in Georgia pays royalty for using a patent created and held in another state that taxes intangible income, the firm would pay taxes for the intangible income in both states without this exception. Then, the firm would have incentives to assign the patent to either Georgia or another state without taxes on intangible income (e.g., Delaware) to avoid the double taxation. But such incentives to relocate the patent are mitigated by the subjectto- tax exception. With this exception, the firm pays the same amount of taxes no matter whether the patent is located in Georgia, a zero-tax state, or other states with the same tax rate as Georgia.13 Consequently, the subject-to-tax exception would further disincentivize patent relocations to zero-tax states.
content_copy

Describing existing limitations

content_copy

Describing existing research gaps

Nevertheless, the current studies center around a shared debate, namely, whether retrospective awards (hereafter, “awards”) have any effect on recipients’ behavior. This question is not trivial given that awards are symbolic, uncostly to arrange, and hold no monetary value for the recipient. The results mostly indicate positive effects, but some studies also indicate no effects or even negative effects (e.g., Gubler et al. 2016).
content_copy
There is growing research that takes the complementors’ perspective (Tanriverdi and Lee 2008, Bhargava et al. 2013, Cennamo et al. 2018). Some studies have investigated multihoming and updating (Cennamo et al. 2018, Kang et al. 2019). Cennamo et al. (2018), for example, found that the characteristics of the platform technology determine complementors’ tendency to multihome. Tiwana (2015) reports that increased rates of updating are related to platform control and the modularity of the complement. There is growing research that takes the complementors’ perspective (Tanriverdi and Lee 2008, Bhargava et al. 2013, Cennamo et al. 2018). Some studies have investigated multihoming and updating (Cennamo et al. 2018, Kang et al. 2019). Cennamo et al. (2018), for example, found that the characteristics of the platform technology determine complementors’ tendency to multihome. Tiwana (2015) reports that increased rates of updating are related to platform control and the modularity of the complement. These contributions notwithstanding, little work so far has considered the relationship between awards and complementors’ product strategies.
content_copy
Existing studies have considered a variety of approaches for scaling a platform, including how to set prices to attract users and suppliers (Parker and Van Alstyne, 2005), using promotion strategies (Rietveld et al., 2019) or establishing partnerships with key suppliers (Cennamo & Santalo, 2013). However, this literature has not considered that platforms may acquire companies as a part of these strategies. However, we might expect that platform companies may undertake acquisitions in a way that relates to the need to scale the platform, and exploit network effects. This is different from how non-platform companies acquire and highlights a need for a theory to understand the role of acquisitions in platform strategy.
content_copy
We know very little about professionals’ conditioning information sets or their underlying forecasting models, and (perhaps) only slightly more about the incentives that might guide construction of their forecasts.
content_copy
To our knowledge, prior studies do not examine whether employee perceptions of their firms or managers change following tax news.
content_copy

Developing hypotheses in quantitative research

To understand the consequences of awards, we draw on the theoretical concept of signaling (Spence 1973, Connelly et al. 2011). This departs from the notion that exchanges between market actors are characterized by information asymmetries (e.g., Jensen and Meckling 1976).
content_copy
Following Stiglitz (2000), we focus on asymmetries arising from the uncertainty regarding the qualities of an actor. Quality uncertainty refers to the “underlying, unobservable ability […] to fulfill the needs or demands” (Connelly et al. 2011, p. 43). A detailed discussion of signaling theory and an excellent overview of the literature appears in Connelly et al. (2011).
content_copy
As a result of this uncertainty, users’ decisions tend to rely on the judgment of others, as manifested in reviews, awards, or rankings (e.g., Dimoka et al. 2012, Li et al. 2020).
content_copy
There is some indirect evidence supporting these arguments. Various studies in product development literature document that firms focus their product development efforts on incremental improvements after initial product success (e.g., Brown and Eisenhardt 1997, Rothaermel and Deeds 2004). In addition, Voss et al. (2008) find that firms are more likely to focus on refining and improving their product portfolio when they have garnered trust and reputation in a market.
content_copy
Hence, we argue the following: Hypothesis 1. Awards increase recipients’ likelihood of releasing complement updates and reduce their likelihood of releasing new complements.
content_copy
In particular, we expect that for award winners, multihoming becomes feasible. An award may not only serve as a quality signal to users of the awarding platform but also to users of rival platforms.
content_copy
Given that an award is publicly bestowed, it is likely that rival platforms will take notice of it. Just as the award acts as a quality signal to platform users, rival platforms will likely interpret the award as a quality signal as well. We therefore expect that rival platforms are likely to poach award winners, providing a second strong incentive for award winners to multihome.
content_copy
There is some indirect evidence that supports this line of reasoning. Rietveld et al. (2019) find that, to differentiate their platforms, owners of video console platforms are more likely to endorse exclusive rather than multihomed games. Li and Zhu (2020) observe that after Groupon restricted public information about its online daily deals, its rival LivingSocial copied fewer Groupon deals and sought instead to source more new deals. We argue the following: Hypothesis 2. Awards increase recipients’ likelihood of multihoming
content_copy
In light of these arguments, we expect that complementors will release new complements in award winners’ market niches. Hence, we argue the following: Hypothesis 3. Awards encourage other complementors to release new complements in recipients’ market niches.
content_copy
We could expect that platforms might also acquire other platform companies, in order to build their base of users and suppliers. There are examples of this strategy behind the growth of many prominent platform companies. For instance, AirBnB acquired a number of similar “home sharing” platforms, in its early days and brought those suppliers (complementers) to their platform. Similarly, Match.com a prominent online dating platform has acquired more than twenty-four other dating platform applications (Gilbert, 2019). As a result, acquiring other platform companies in the same market niche can provide a strategy for platform companies to scale their base of suppliers and customers.
content_copy
At the same time, we may expect that as platform companies mature, they may have an incentive to acquire other types of companies. For instance, Microsoft and Sony both acquired a number of gaming studios that would develop content (video games) for their respective gaming consoles, often the canonical example of platform companies in scholarly studies (Corts & Lederman, 2009; Shankar & Bayus, 2003).
content_copy
In summary, we expect to see platform companies making acquisitions early in their life (shortly after founding) as a way of growing their platform.
content_copy
It is important to also highlight that these patterns are also likely not true for non-platform companies. As discussed in Section 2, there has not been a lot of research into when companies acquire, and very little work into the acquisition strategies of entrepreneurial companies. In fact, much of the existing literature on platforms has focused on large, established companies and has highlighted that acquisitions are a common growth strategy for such companies (Kim et al., 2011; Klepper, 1996). This suggests that non-platform companies often acquire when they are more mature and not as part of their early growth strategy. As a result, we expect the acquisition strategies of platform companies differ from those of non-platform companies that have typically been studied.
content_copy
Based on the arguments above, we summarize our empirical hypotheses below. H1: Digital platform companies make their first strategic acquisition earlier than nonplatform companies. H2. Digital platform companies are likely to first acquire companies from the same product market niches but then proceed to acquire companies from different product market niches as they mature. H3. Digital platform companies are likely to first acquire other platform companies but then proceed to also acquire digital non-platform companies as they mature.
content_copy
In summary, we propose that perks associated with lower contractuality heighten relational value, which increases consumers’ propensity to share WOM about the company in the absence of any explicit incentive to do so. In contrast, perks associated with higher contractuality are less likely to heighten relational value and thus are less likely to foster WOM. Stated formally, we propose a relational value account with the following hypotheses: H1: Perks lower (vs. higher) in contractuality increase WOM. H2: Perks lower (vs. higher) in contractuality signal higher relational value. Higher relational value, in turn, heightens consumers’ motivation to help the company, which explains the increase in WOM predicted in H1.
content_copy
Drawing on this research, we hypothesize that the linguistic femininity of a brand name will increase perceived brand warmth, which will, in turn, positively affect attitudes and choice. Further, the brand relationship literature finds that more favorable consumer attitudes can enhance brand performance (Puzakova, Kwak, and Rocereto 2013). Therefore, we also predict positive downstream associations between feminine brand name gender and brand performance. Formally: H1: Brand name femininity is associated with positive brand outcomes (attitudes, choice, and performance). H2: Perceived brand warmth mediates the feminine brand name advantage.
content_copy
Therefore, these arguments lead to the prediction that addback statutes impede corporate innovation. We state this prediction in the alternative form as follows: Hypothesis. The adoption of addback statutes negatively affects corporate innovation.
content_copy
We hypothesize that employee perceptions of senior managers and firms decrease following corporate tax avoidance news coverage for two reasons. First, employees do not clearly benefit from lower tax payments. Several studies on the incidence of the corporate income tax suggest that employees do not benefit from lower tax payments. In pioneering work, Harberger (1962) posits that shareholders bear the incidence of the corporate income tax, while labor bears little or no incidence. Recent work largely confirms this finding (e.g., Auerbach 2006, Gravelle 2010; Fuest, Peichl, and Siegloch 2018).6 Nallareddy, Rouen, and Serrato (2018) even find evidence that corporate income tax rate cuts harm rank-and-file employees by exacerbating income inequality. Moreover, employees are fixed claimants to the assets of the firm. As such, they generally do not prefer investments with risky cash flows, such as tax avoidance, which shareholders may prefer (Gleason et al. 2019). Thus, the weight of the evidence is consistent with lower tax payments providing little benefit to employees. Second, employees likely perceive corporate tax avoidance as unfair and not socially responsible. Anecdotal and survey evidence suggests that corporate tax avoidance is publicly perceived as inequitable and socially irresponsible (Rupp et al. 2006; Dowling 2014; Motel 2015; Pegg 2017; Elbra and Mikler 2017). Moreover, numerous studies in psychology provide evidence that tax avoidance and evasion are considered unfair or unjust, which leads “ordinary people” to desire fewer “loopholes.” (Kinsey 1984; Dornstein 1987; Spicer and Becker 1980; quoted language from Song and Yarborough 1978). We expect employees to share that perception because they are exposed to the same information and influences as the public and generally value fairness and social responsibility (Colquitt 2001; Rupp et al. 2006). Thus, our hypotheses are as follows: H1: Employee ratings of senior management are negatively related to news about firms’ tax avoidance activities. H2: Employee ratings of their employing firms are negatively related to news about these firms’ tax avoidance activities.
content_copy

Explaining how your thesis extends existing sources

Previous studies have researched various governance decisions, including pricing (Hagiu 2006), resourcing (Ghazawneh and Henfridsson 2013, Foerderer et al. 2019), restricting the number of complements (Boudreau 2012), matching (Bhargava et al. 2020), seeding (Huang et al. 2018), rule setting (Claussen et al. 2013), intellectual property rights (Ceccagnoli et al. 2012), performance investments (Anderson et al. 2014), interfirm exchange (Foerderer 2020), social norms (Burtch et al. 2018), and value capture (Foerderer et al. 2018). All these studies investigate mechanisms that require substantial financial commitment or the exertion of control in terms of a hard governance mechanism. By contrast, we study awards which represent a comparatively soft governance mechanism, in terms of being non-monetary and non-control. Research has started to investigate soft governance mechanisms only recently. Jullien and Pavan (2019) and Hukal et al. (2020) investigate the communication behavior of platform owners. For example, Hukal et al. (2020) observe that platform owners can direct complementor contributions to certain areas of the ecosystem by communicating their intention and interest in advancing these areas. Liang et al. (2019) investigate spillover effects from endorsements. Adding to these studies, we focus on understanding the consequences of awards on complementors’ product strategies.
content_copy
The work on retrospective awards is, as Frey and Gallus (2017) conclude in their survey of extant work, still in its early stages. The primary difference between our work and extant studies of awards is that we investigate awards that are conferred on firms, and not on individuals. Most previous studies have investigated awards for individuals and studied outcomes such as employee motivation (Kosfeld and Neckermann 2011), student attendance (Robinson et al. 2021), and community contributions (e.g., Gallus 2016, Burtch et al. 2021, Chen et al. 2010). Comparably few works have investigated awards for products or firms (e.g., Hendricks and Singhal 1996, Anand and Watson 2004, Kov´acs and Sharkey 2014).
content_copy
Of note, the goal in this research is not to argue that relational value is the only mechanism at play, but rather to evaluate whether relational value plays a unique role above and beyond other processes.
content_copy
The purpose of this research, therefore, is to understand if the linguistic characteristics of a brand name lead to the implicit transfer of gender-associated qualities to the brand and when this might be relevant to marketers. Warmth is one quality a name may convey that is of critical importance to brand perceptions. The present research investigates how qualities such as warmth, which are associated with gender, may be implicitly transferred to the brand, as a result of the linguistic properties of the name. We focus on warmth because it is of fundamental importance to human judgments, generally, and to brand attitudes, specifically.
content_copy
My framing of shifting endpoints differs in two important respects from most prior studies of bond yields. First, the typical assumption is that the risk factors share a single, common stochastic trend. This introduces the possibility that the level and slope of the yield curve are constrained to share a trend. In contrast, the long-run means of PC1 and PC2 are different linear combinations of the vector ˆκ0t . Second, prior studies typically presume that their common stochastic trend is observable to the econometrician and it is often proxied by a weighted average of past yield or macroeconomic information (e.g., a trend inflation series). Instead of linking κ0t to specific, observed macroeconomic time series, BE treats these shifting endpoints as latent, and constructs filtered ˆκ0t based on the past history of the yield PCs. Richer conditioning information could easily be accommodated as in GLS.
content_copy
Our study is related to the work of Persons and Warther (1997), who consider a model of technology adoption where multiple _rms must choose their adoption time. Each _rm observes the noisy cash ow returns of _rms who have already adopted and decides whether to adopt the innovation. They generate ooms in the adoption of the new technology, as each additional _rm that adopts the innovation may lead to another _rm's subsequent adoption. Our setting di_ers from Persons and Warther (1997) primarily in that, in our model, agents have private information over their types, the state variable follows a stochastic mean-reverting process, and agents are discrete.
content_copy
Our contribution relative to Persons and Warther (1997) and the extant literature is as follows. First, we examine how rms' learning incentives map into and a ect their IPO timing decisions when there is uncertainty about the types and IPO decisions of other rms. The presence of private information leads to a unique symmetric equilibrium which always exhibits strictly positive probability that no IPOs are observed in the rst period, whereas this kind of overall industry delay is absent in Persons and Warther (1997). This helps to explain, for example, delay in the emergence of pioneer IPO rms in new industries (and potential variation in this delay across new industries). Hence, the economic forces of our setting and the resulting equilibrium characterization are qualitatively distinct from Persons and Warther (1997). Second, our comparative statics analysis provides insights that are largely not captured in the framework of Persons and Warther (1997) or the other extant literature, particularly with regard to changes in the number of rms. These insights include predictions regarding IPO timing, IPO volume (and how volume uctuates over time), the expected incidence of sequential clustering or an IPO drought, and the variation of issue quality over time. In particular, we provide testable implications for variation in these outcomes with respect to industry concentration, the cost of delay, uncertainty, and persistence in sentiment. Third, we show that the introduction of a stochastic, mean-reverting state variable results in non- monotone e ects of the degree of mean-reversion on the incentive to delay. This leads to additional implications concerning the role of persistence in investor sentiment and helps to provide an explanation for delay between rms' IPO times. Moreover, our analysis more broadly contributes to the theoretical literature as few papers consider information spillovers with a stochastic state, and hence our study provides insights on how learning incentives are a ected when the underlying state evolves over time.
content_copy
Relatedly, Bustamante (2012) considers IPO timing as a real options problem with asymmetric information. She also nds that high-type rms go public earlier in some equilibria, and characterizes equilibria in which the presence of private information either speeds up or delays the IPO time relative to complete information. Our model varies from these studies in that we incorporate information spillovers between firms that a ects the timing of IPOs, whereas strategic interaction between rms is absent in Pastor and Veronesi (2005) and Bustamante (2012). This leads to additional insights and predictions that are not captured by these prior studies. Our model varies from the literature on dynamic voluntary disclosure literature (e.g., Dye and Sridhar (1995), Einhorn and Ziv (2008), Guttman et al. (2014), Aghamolla and An (2020)) in three ways. In our setting (i) the manager receives information with probability one and disclosure is costless, (ii) the entrepreneur is only concerned with the rm's value in the period of disclosure and IPO, and (iii) there are multiple rms whose decisions are interrelated.6
content_copy

Expressing formula and theorems

content_copy

Referring to tables and figures

Table 1 provides additional examples of how perks used by companies across a variety of industries and markets can differ in terms of contractuality depending on the way they are conferred, structured, or framed.
content_copy
Fig. 1 provides a timeline summarizing the main events in the model.
content_copy
Table 11 provides five robustness tests. In the first column, we re-estimate Model (1) after replacing the dependent variable with Ln_NPat5, which is the log of one plus the number of patents filed in year tþ5. Results remain similar to those in Table 2. In the second and third columns, we provide a falsification test by using the natural logarithm of one plus the number of patents filed in year t-1 (Ln_NPat_Past) and the natural logarithm of one plus the number of patents filed in year t-3 (Ln_NPat_Past3) as the dependent variables. The coefficient on Addback is insignificant in both columns. In the fourth column, we exclude firms that never file patents in our sample period. In the fifth column, we use the location of headquarters instead of subsidiaries for the definition of Addback.38 In both Column 4 and Column 5, the coefficients on Addback are negative and significant. Finally, in the last column, we exclude IT industries (Fama-French industry groups 34, 35, and 36) from our sample and find similar results. Thus, our findings are not due to the dot-com bubble.
content_copy
Table I reports descriptive statistics for the hedge fund sample based on fund-month observations. All variables are winsorized at the 1% and 99% levels.
content_copy
As shown in Panel B of Table II, the Baker-Wurgler sentiment changes index is positively correlated with changes in the Michigan consumer sentiment index and negatively correlated with the FEARS index, since the latter captures investors’ bearish attitude about economic conditions.
content_copy
Panel B of Table II shows that, overall, the Baker-Wurgler sentiment changes index is only modestly correlated with these risk factors. The three highest correlations are with the equity market, the size factor, and inflation (at 0.16, 0.22, and 0.22, respectively).
content_copy
Table III reports results for both hedge fund excess returns and alpha across the sentiment beta-sorted portfolios.
content_copy
Figure 1 plots the spread in one-month-ahead risk-adjusted returns (i.e., alpha) between the two extreme decile portfolios with high and low sentiment betas. The series start in January 1997, since we use a 36-month formation period.
content_copy
Table IV reports the results from the cross-sectional regressions, with the dependent variable being either the hedge fund excess return or alpha.
content_copy
In Table 3, columns (1) to (3) report the estimates for Equation (1) with UPDATE as the dependent variable. The coefficients are reported in log-odds notation. Column (1) shows that winning an award has a positive effect on the likelihood of an app update, with runners-up as a control benchmark. The odds of an app update are e0.472 1.60 times higher for app developerswhowon the award. Column (2) uses CEM matches as a control group, supporting the positive effects (e0.609 1.84). Column (3) uses PSM matches as a control group and documents odds comparable to the baseline (e0.641 1.90). Columns (4) to (6) estimate the effect of the award on NEWAPP. Column (4) indicates a significantly negative effect. For recipients, the odds of releasing a new app are e−2.837 0.06 times that of runners-up. Column (5) uses CEMmatches as the benchmark and suggests effects similar to the baseline (e−3.226 0.04). Column (6) uses PSM matches as the benchmark, and the coefficient differs only marginally fromthe baseline (e−3.461 0.03). Taken together, these data supportHypothesis 1. Taken together, these data support Hypothesis 1.
content_copy
Finally, we find support for Hypothesis 3. Column (4) of Table 4 is the baseline, predicting CAT_NUMAPPS. The DID estimate suggests that award recipients’ market categories experience an increase in app releases. Column (5) confirms the positive effect when using CEMmatches as the control group. Column (6) also confirms the positive effect when using the PSM control group. Online Appendix Table A5 reports that the estimate differs only marginally when restricting the control group to apps that are not in the same categories as award winners’ apps. Therefore, the composition of the categories within the groups seems to exert little influence on the results. Together, these analyses indicate support for Hypothesis 3.
content_copy
Table 6 reports the results.
content_copy
In Online Appendix Table A3, panel A reports tests for various further developer and app characteristics.
content_copy
To ease interpretation, Online Appendix Figure A4 plots the coefficients of the interaction term.
content_copy
Descriptive statistics for the variables used in the analysis are reported in Table 1 & 2. In Table 3, we report the average Firm Age (how many years after founding did an acquisition happen), as descriptive evidence for the various hypotheses.
content_copy
We report the results of this regression in Table 5.
content_copy
We report the results without any controls in Columns 1 and 2. In Column 3 we introduce the full set of control variables.
content_copy
Table 2 reports the regression results of Model (1) when Ln_NPat3 is the dependent variable. Standard errors are clustered by firm to mitigate serial correlation concerns. In the first column, we only control for firm, state, and year fixed effects.32 In the second column, we further include all the other variables in Model (1). The coefficient on Addback is 0.0501 in Column 1 and 0.0477 in Column 2, both significant at the 5 percent level, indicating that operating in a state that has adopted an addback statute is negatively associated with the number of patents filed three years later.
content_copy
The regression results are reported in Table 6. We use three measures of financial constraints: the KZ index of financial constraints (Farre-Mensa and Ljungqvist, 2016; Lamont et al., 2001), credit rating, and operating cash flows. Firms with lower values of the KZ index, credit ratings, and higher operating cash flows are less financially constrained. In Columns 1 to 3, we interact Addback with KZ index (KZ Index), an indicator for firms with credit ratings (Rated_Firm), and operating cash flows (CFO), respectively. The alternative explanation predicts a negative (positive) coefficient on the interaction term in Column 1 (Columns 2 and 3). In Columns 4 to 6, we re-estimate Model (1) using the subsample of observations with KZ index below the top decile, the subsample of observations with credit ratings, and the subsample of observations with positive operating cash flows, respectively. The reported regressions use patent count (Ln_NPat3) as the dependent variable. We find insignificant interaction effects in Columns 1 and 3. Further, the interaction effect is negative and significant in Column 2.35 Thus, inconsistent with the alternative explanation, the effect of addback statutes on innovation is not more pronounced for firms that are more financially constrained. Further, we find significantly negative coefficients on Addback in Columns 4 to 6. These findings suggest that the addback statutes have a negative effect on innovation among firms that are not financially constrained. Therefore, the effect on innovation cannot be simply attributed to crackdown on tax avoidance increasing financial constraints and thus reducing investment in innovation.
content_copy
Figure 1 plots the spread in one-month-ahead risk-adjusted returns (i.e., alpha) between the two extreme decile portfolios with high and low sentiment betas. The series start in January 1997, since we use a 36-month formation period.
content_copy
Table IV reports the results from the cross-sectional regressions, with the dependent variable being either the hedge fund excess return or alpha. From the univariate regression, the regression coefficient of fund excess return on sentiment beta is 0.17 (t-statistic = 3.03). Thus, an increase in sentiment beta from −1.07 (its average level in the bottom decile in Table III) to 0.96 (its average level in the top decile) leads to an increase in fund monthly excess returns of 0.35% (i.e., 0.17% × 2.03) on average.
content_copy
The results, reported in Table V, confirm a significantly positive relation between sentiment beta and hedge fund performance. For example, on average, the top sentiment beta hedge funds outperform the bottom sentiment beta hedge funds by 0.26% (t-statistic = 1.88) per month and by 0.58% (t-statistic = 2.17) per month on a risk-adjusted basis. The spreads in excess returns and alpha are very close in magnitude to those obtained from the Baker-Wurgler sentiment measure.
content_copy
Table VI shows a significantly positive relation between hedge funds’ gross returns and their sentiment betas. Based on gross fund returns, the top sentiment beta fund portfolio has an average excess return of 1.07% (t-statistic = 6.00) and an alpha of 1.11% (t-statistic = 5.95) per month, while the bottom sentiment beta fund portfolio has an average excess return of 0.72% (t-statistic = 4.57) and an alpha of 0.37% (t-statistic = 2.27) per month. Between the top and bottom portfolios, the spread in monthly fund returns is 0.35% (t-statistic = 3.04) while the spread in monthly alpha is 0.74% (t-statistic = 4.14). The analysis using gross fund returns therefore leads to the same inference about the relation between sentiment beta and subsequent hedge fund performance.
content_copy
Our results are reported in Figure 1. In Panel A, we use SeniorMgmt as our dependent variable. In the four quarters prior to treatment, the plotted coefficients do not appear to be increasing or decreasing prior to treatment. In Panel B, we use Firm as our dependent variable. Again, we discern no trend in the pre-treatment period, consistent with Glassdoor ratings of treated and control firms evolving in parallel in the pre-treatment period.
content_copy
In Panel A of Table 3, we report the results of estimating the effect of tax avoidance news on senior management and firm reputation using generalized difference-in-differences.
content_copy

Opening the methodology section

We conducted a quasi-experimental study in the context of the Google Android platform and the annual Google Play Awards. The Google Android platform is a mobile operating system. Complementor firms (“app developers”) provide ancillary mobile apps via the Google Play Store. The Google Play Award seeks to “recognize top app and game developers from around the world who lead the way in delighting users with incredible experiences on Android” (Google 2016). Its specific focus is to honor an outstanding app by an app developer firm. Online Appendix Table A1 provides further details on the Google Play Award.
content_copy
We conducted seven experiments (six main and one follow-up) to test our hypotheses. Experiment 1, a field study, tests whether awarding a perk in a less contractual way increases WOM. Experiments 2 and 3 test the proposed relational value account via mediation by using different operationalizations of contractuality and WOM. Experiments 4 and 5 provide evidence for the relational value account via moderation. Finally, Experiment 6, a second field study, tests the idea that lowcontractuality perks can be more effective at fostering WOM yet less effective at stimulating compliance with a request for a desired behavior.
content_copy
We test these hypotheses in six studies. In Study 1, we investigate whether brand name femininity is positively associated with brand performance by examining the linguistic characteristics of Interbrand Global Top Brand names (H1). Study 2 tests the hypothesis that warmth mediates the relation between linguistic name gender and brand attitude using a sample of real-world brands (H2). Studies 3a and 3b replicate the main effect and process account, providing additional support for H1 and H2 using actual product choices with consequences for time (Study 3a) and money (Study 3b). Studies 4 and 5 test the boundary conditions of typical user gender (H3) and product category (H4).
content_copy

Explaining the choice of research methods

This context is appealing for reasons of data availability. Compared with other contexts, the Google Play Award provides us with a setting in which we can observe complementor behavior over time. Other platforms nominate only a handful of firms per year and do not provide enough data for statistical analysis. Of particular value is the fact that Google publishes the entire short list of nominees for the award. This permits the construction of a quasi-experimental design, which we describe in the following section.
content_copy
The main empirical challenge is to isolate the effects of the award from pre-existing heterogeneity. Simple comparisons between award winners and other complementors would be misleading because they might capture not only changes due to winning the award but also pre-existing differences that correlate with the propensity to win an award in the first place. Figure 1 illustrates our identification strategy. The context of the Google Play Award offers the opportunity for a quasi-experimental research design that may mitigate the above concerns.
content_copy
We conducted interviews with three award winners. All interviewees were founders and executives of the firms. We conducted semistructured interviews to validate the assumptions of the research design. We also asked the award winners questions about the impact of the award on their businesses.
content_copy
One may argue that restricting the control group to apps that are not in the same categories as award winners’ apps would avoid comparing firms from the same niche that have received similar increases in new app releases. The robustness section confirms that the results hold for this alternative setup.
content_copy
Since our outcome variable is a time variable (i.e. Number of years after founding that a company makes their first acquisition), we use a survival model. This methodology takes into account the natural truncation that can occur with such a time-based variable.
content_copy

Comparing and contrasting with other research methods

In our quasi-experimental design, therefore, the treatment group consists of those app developers who ended up winning the award, whereas the control group consists of the runners-up. Econometrically, we infer the effects of award winning by estimating a DID design. By these means any bias caused by variables (observed or unobserved) common to award winners and runners-up is accounted for (Angrist and Pischke 2009).
content_copy
This design is consistent with other studies that implement staggered difference-in-differences (e.g. Bertrand and Mullainathan 2003; Giroud and Mueller 2010).
content_copy

Specifying quantitative data collection method(s)

Our dataset is a time-series panel on the app-month level. We conduct the analysis on the app-level because this allows accounting for app-level factors that influence product strategies, and it is in line with prior research on awards (Kov´acs and Sharkey 2014). We considered the Google Play Award in the years 2016, 2017, and 2018. We identified all apps nominated for the award in these years from the official Android Developers Blog run by Google. We removed apps from the sample that were nominated in consecutive years (see Online Appendix C). We obtained monthly data (e.g., on app ratings, updates, and new app releases) from AppBrain, App Annie, and the Internet Archive.Our dataset is a time-series panel on the app-month level. We conduct the analysis on the app-level because this allows accounting for app-level factors that influence product strategies, and it is in line with prior research on awards (Kov´acs and Sharkey 2014). We considered the Google Play Award in the years 2016, 2017, and 2018. We identified all apps nominated for the award in these years from the official Android Developers Blog run by Google. We removed apps from the sample that were nominated in consecutive years (see Online Appendix C). We obtained monthly data (e.g., on app ratings, updates, and new app releases) from AppBrain, App Annie, and the Internet Archive.
content_copy
We use data on acquisitions from CrunchBase. This data contains a profile for each company, which includes a text description of the business area of the company, information about the number of employees, date of founding, sources of venture financing and a list of acquisitions made by each company. This dataset does not contain SIC based industry classifications but does contain tags that describe the general business area of each company (e.g. Cloud Storage).
content_copy
In the field studies (Experiments 1 and 6), we targeted a number of approximately 50–100 responses per cell on the basis of funds and participants’ availability. In the MTurk studies (Experiments 2–5), the targeted number of responses per cell was 100–200.
content_copy
To test this hypothesis, we conducted a field experiment in a momand-pop bakery. We picked this bakery because it has a strong social media presence and is described by bloggers as being an “Insta-worthy” spot because of its aesthetically pleasing interiors and unique desserts.
content_copy
Participants (n ¼ 398, 170 female, mean age ¼ 37.19 years, SD ¼ 13.01) recruited via MTurk completed an online survey in exchange for monetary compensation.
content_copy
Five hundred twenty-four members of the MTurk online panel (51% men; Mage ¼ 36.79 years; 99% native English speakers) participated for a nominal fee.
content_copy
We obtain company financial data from Compustat and patent data from Kogan et al. (2017).18 Data on material subsidiaries disclosed in Exhibit 21 of Form 10-K are from Dyreng and Lindsey (2009) and Dyreng et al. (2013).19 Data on state statutory tax rates and state R&D tax credits are from the Federation of Tax Administrators and Wilson (2009).
content_copy
We begin with the sample of U.S. firms between 1997 and 2005 in Compustat. Our sample period starts from 1997, which is two years before the first state adopted an addback statute in 1999. Kogan et al. (2017) provide patent data matched with CRSP firms up to 2010. Patents filed before and in 2008 would have most likely been granted by 2010.20 Therefore, the sample for our primary tests ends in 2005, because we examine the number of patents filed three years ahead.
content_copy
I study the BCFF survey of yield and macroeconomic forecasts over the period from January 1985 through December 2018, with the start date determined by data availability. Of the 194 forecasters, 115 are categorized as financial institutions, 48 are consulting firms, and 31 represent other types of institutions. The forecasters submit forecasts of investment yields on U.S. Treasury bonds that have maturities of six months and one, two, three, five, seven, and 10 years. From these raw data, survey forecasts of zero-coupon bond yields of matching maturities are constructed as in Le and Singleton (2012). Survey data are released monthly at the beginning of the following month (usually the first business day), based on information collected over a two-day period (typically scheduled between the 20th and the 26th of the month). Disagreement is measured as the difference between the 90th and 10th percentiles of the cross-sectional distribution of BCFF zero-coupon yield forecasts.15
content_copy
Our hedge fund sample is from the Lipper TASS database. TASS classifies hedge funds into 11 strategy categories: convertible arbitrage, dedicated short bias, emerging markets, event driven, equity market neutral, fixed income arbitrage, funds of funds, global macro, long/short equity, managed futures, and multi strategy. Since our sentiment measure corresponds largely to U.S. stock markets, we focus on U.S. equity-oriented hedge funds and drop emerging markets, fixed income arbitrage, and managed futures. Dedicated short-bias funds are also excluded since only 42 such funds satisfy our data filters.8 The sample is free of survivorship bias, as TASS covers both live and defunct hedge funds since 1994 and we examine the period from 1994 onward.
content_copy
We incorporate data from three sources. First, we collect data on employee perceptions of their employing firms and senior management from Glassdoor.com. Glassdoor.com is a website that allows employees to anonymously provide their perceptions of the firm, senior management and various other aspects of working for a firm. […] To measure news coverage of tax avoidance, we hand-collect data on news coverage of S&P 500 firms’ tax avoidance activities. We searched for news about “tax evasion,” “tax avoidance,” “tax haven” and each company’s name in LexisNexis. We focus on corporate income tax avoidance (see Appendix C for a list of instructions), though we may unintentionally collect other forms of corporate tax avoidance due to human error in hand-collection (e.g., payroll tax avoidance). Our media sources include all worldwide news media sources (e.g., “newspapers,” “news,” “newsletters”) in LexisNexis. Finally, we rely on Compustat Quarterly to incorporate financial statement-based controls.
content_copy
We pretested to identify brands relevant to our participant population with a high degree of variability in terms of attitudes. One hundred members of the Amazon Mechanical Turk (MTurk) online panel (60% men; Mage ¼ 34.66 years) were paid a nominal fee to list ten brands they thought were “really cool” and ten brands they “would never use,” following the procedures outlined by Escalas and Bettman (2003).
content_copy
We conducted a pretest to identify hypothetical brand names that varied on name gender while being equally and generally devoid of meaning and unfamiliar to participants. We generated test name stimuli drawing on prior work (Klink 2009; Lowrey and Shrum 2007). In addition to 55 potential name stimuli, a set of 10 words selected on a priori grounds to carry some semantic associations were included to provide a benchmark for comparison and encourage participants to use the full range of response scales (Schmitt, Pan, and Tavassoli 1994; see Web Appendix E).
content_copy
We conducted a pretest to identify names that vary on gender score while being equal in length, equally and generally devoid of meaning, and unfamiliar to participants
content_copy
In the dataset, the pre-award period begins in January of each year and ends in March. Nominations are announced in April, which is why we exclude this month from the analysis. Across the years studied, the dates for nomination and award conferral fall into the same month of the year. The post-award period begins in May and ends in March of the following year so that it does not overlap with next year’s award period. The length of the post-award period involves trading off variance in the dependent variables— which gets larger as we extend the post-award period— with capturing the immediate effects. Because we removed app developers who were nominated in subsequent years, little bias is to be expected, but we additionally restrict the post-award period to end in March to avoid any potential effect (see Online Appendix Figure A2).
content_copy
We collect all daily ratings for the firms in the 2012 S&P 500. We are unable to identify firm information for five of the firms in the S&P 500 on Glassdoor and thus retain 495 of the 500 firms in the S&P500 in our sample. Our final sample spans all calendar-quarters from January of 2008 (Q1) to December of 2017 (Q4).7
content_copy
The final sample size differs depending on the control group. In the base setup (i.e., runners-up as the control group), the sample comprises 125 developers (30 winners and 95 runners-up) and their 793 apps, resulting in an unbalanced panel of 5,131 app-months. In the matched sample setup based on coarsened exact matching, given the employed procedures outlined below, the sample comprises 8,414 appmonths. In the matched sample setup based on propensity score matching, the sample comprises 8,126 app-months.
content_copy
The CrunchBase dataset contains information on a large number of companies across a variety of industries. We omitted companies from non-digital industries, as well as companies founded prior to the year 2000. We omitted companies that did not have text descriptions (or very short text descriptions) as we relied on these descriptions to calculate some of our variables. These were typically very small companies, which did not survive very long and typically did not make any acquisitions. Therefore, omitting these observations did not influence our conclusions.
content_copy
Our final sample consisted of 1,933 companies that made strategic acquisitions and were founded after the year 2000, out of a broader sample of 123,044 digital companies which were in our dataset but did not make any acquisitions. Of the acquiring companies, 278 were platform companies. Our final sample consisted of 3,062 acquisition events.
content_copy
One hundred fifty students (60% women; Mage ¼ 20.45 years; 77% native English speakers) from a public North American university participated for course credit. We collected the largest sample possible given subject pool constraints.
content_copy
Our final samples vary in size based on data availability of our dependent variables. We drop all missing observations with missing control variables. Our baseline regressions are performed on 14,840 firm-quarter observations when SeniorMgmt is the variable. When Firm is our dependent variable, we have 14,977 firm-quarter observations.
content_copy
We allow apps to drop out of the sample (e.g., due to being removed or due to missing data) and to enter the dataset (i.e., due to being released).
content_copy
Unless reported, no participant was excluded.
content_copy
Only participants who passed an initial attention check were eligible to participate.
content_copy
Forty-eight participants failed the attention check, which left us with 369 participants.
content_copy
We excluded six participants who did not fit the recruitment criteria, as they were visitors who were not affiliated with the university. In addition, three participants did not provide valid Instagram handles and were thus excluded. We were left with 193 participants, 140 of whom agreed to take the survey.
content_copy
Attention checks and/or IP address checking software were used to screen participants before they entered online experiments to ensure data quality (Dennis, Goodson, and Pearson 2020; Winter et al. 2019), except in Study 2, in which open-ended questions were embedded to discourage automated responses, and Study 3b, which was conducted in person in a lab. Data were analyzed only after collection was complete.
content_copy
We have several sample selection requirements. First, we remove firms that are not taxed as corporations as well as firms with missing Central Index Keys (CIKs).21 Second, we exclude single-state firms that have material subsidiaries in only one state, because addback statutes are supposed to affect the tax avoidance behavior of multistate firms. Third, we delete firmyear observations that have both negative state income tax and negative domestic pretax income, as these firms pay no state income taxes and thus are unlikely to be affected by state tax policies.22 Next, we exclude firms with missing industry code and firms in nonpatent industries.23 Further, to ensure enough within-firm variation for our analyses, we require each firm to have at least three observations in our sample period.24 Lastly, we restrict the sample to observations with nonmissing data to compute the variables used in the main tests. Our final sample includes 11,228 firm-year observations, which belong to 1946 unique firms.
content_copy
Following prior research, we apply several screens to the fund data. To address the concern that hedge funds may backfill returns when newly added to the database, we exclude the first 12 months of returns for each fund. We only include funds that report monthly net-of-fee returns in U.S. dollars and allow for redemption at a monthly or higher frequency.9 We also delete duplicate funds and funds with assets under management below $5 million.10 Finally, we require each fund to have at least 30 return observations. After these screens, our sample contains 4,073 hedge funds over the period 1994 to 2018.
content_copy
The four dependent variables are UPDATE, NEWAPP, MULTIHOMED, and CAT_NUMAPPS. The term UPDATE is an indicator that takes a value of one if app I was updated in month t. To identify updates, we obtained the version number of app i in month t. If the version number changed between two consecutive periods, we coded UPDATE as one (e.g., change from 1.1 to 1.2). We created a second variable, UPD_MINOR, coded as one if a so-called minor update was performed.
content_copy
The term NEWAPP is an indicator coded as one if developer j of app i released a new app in month t.
content_copy
The variable MULTIHOMED is an indicator coded as one if app i was available on Apple iOS in month t. For each app, the variable is coded as zero until an app is multihomed, and then is coded as one for the remaining periods. There is variation in the variable across apps, within developers, and over time.
content_copy
Regarding the independent variables, our empirical framework required two main indicators. AWARD, coded as one for apps of award winners, and AFTER, coded as one if month t is after the ceremony.
content_copy
Choosing control variables is not trivial because of the risk of including “bad controls.”2 We nevertheless deem it necessary to control for two variables. First, we control for pricing. Product strategies are likely to be influenced by app pricing. The term PRICE holds app i’s purchase price in USD in month t. We (log + 1)-transformed the variable to account for skewed distribution. We use alternative price variables in the robustness section to account for potential concerns over the empirical distribution of PRICE. Second, we control for app quality.
content_copy
We rely on two further variables in additional analyses. The termNUMRATINGS is the total number of ratings submitted for app i as of month t, and EMPLOYEES is a proxy of firm size in terms of the number of employees of the developer. We infer the number of employees from the number of LinkedIn members that state they worked for each firm. We obtain these data fromeach firm’s LinkedIn page. The termINAPP is coded as one if app i offers in-app purchases inmonth t. The termAPP AGE is the age of app i in t inmonths.
content_copy
In addition to using runners-up as a control group, we constructed two alternative control groups based on matching. Matching is based on the idea that units—in our case developers—are selected and placed into an artificial control group based on their observational similarity. We selected developers based on their similarity before the award.
content_copy
As a first step, we take the text descriptions of each company and convert the description into a TF-IDF vector. We remove keywords that occur frequently. These vectors reflect how much particular keywords occur in each text description. We then compare each pair of acquirer and target firms to identify to what extent they overlap in terms of these keywords using cosine similarity. We also test the robustness of this approach by using topic modeling to reduce the dimensionality of these vectors (following Shi et al., 2016), or comparing the overlap based on industry tags (e.g. Cloud Storage, etc.).
content_copy
We control for industries or business areas (Industry FE) using dummy variables to indicate whether a company is assigned a particular tag in its profile. These tags are used to describe the business of the company, as described above (e.g. Cloud Storage).
content_copy
We control for Firm Size based on the number of employees that the company employees. This is reported as range (e.g. 50 – 100 employees) in our dataset, and therefore we use dummy variables to indicate the different groups.
content_copy
As an additional control, we include Funding Controls, which contains dummy variables for the number of venture funing rounds that the firm has received. This provides a measure of the financial resources that a company has at its disposal to undertake acquisitions.
content_copy
The comparison data set consisted of an equal number of companies randomly drawn from the Thompson Reuters Eikon database.
content_copy
We calculated the linguistic name gender of the Interbrand and Thomson Reuters Eikon brands using a method developed by Barry and Harper (1995; Appendix). The name gender score quantifies the degree to which a name is masculine or feminine based on its length, sounds, and stress as discussed in the introduction, with scores ranging from 2 (very masculine) to þ2 (very feminine).
content_copy
We identify a firm as an affected firm if it has at least one subsidiary in a state during the year in which the state adopts the addback statutes.
content_copy
Further, to identify firm-year observations impacted by the adoption of addback statutes, we construct an indicator variable, Addback. Specifically, for an affected firm, we set Addback to 1 for the adoption year and all the subsequent years, unless the firm no longer has any subsidiary in states with the addback statutes. If a firm has no subsidiaries in states with the addback statutes in a given year, Addback equals 0.
content_copy
Following prior literature on innovation (e.g., Griliches et al., 1987), we use patent-based innovation measures for two reasons. First, patent is an output measure that captures both observable and unobservable inputs into innovation (He and Tian, 2013), whereas R&D expense only reflects observable inputs. Second, reported R&D expenditures contain significant measurement errors. Koh and Reeb (2015) show that almost one half of firms in Compustat report missing R&D expenditures, and about 10 percent of firms with missing R&D expenditures actually file patents.We also find that R&D expense is missing for 58.4 percent of the Compustat population during our sample period. Therefore, we use patent count and citation count to capture the amount and quality of innovation. Our innovation variables are constructed using patent data provided by Kogan et al. (2017).
content_copy
Following He and Tian (2013), our first innovation variable is Ln_NPat3, which is measured as the natural logarithm of one plus the number of patents filed three years after the year in which the key independent variable Addback is measured.
content_copy
Our second innovation variable is Ln_NCite3, which is measured as the natural logarithm of one plus the number of non-self-citations received on patents that are filed three years ahead.
content_copy
To measure risk-adjusted returns (i.e., alpha), we control for exposures to standard risk factors identified in the hedge fund literature. We start with Fung and Hsieh’s (2004) seven factors: an equity market factor, a smallminus-big size factor, the change in the constant-maturity yield of the 10-year Treasury, the change in the yield spread between Moody’s Baa bond and the 10-year Treasury bond, and three trend-following factors for bonds, currencies, and commodities.13 These factors are commonly used to evaluate hedge fund performance (e.g., Kosowski, Naik, and Teo (2007), Fung et al. (2008), Jagannathan, Malakhov, and Novikov (2010), Sadka (2010), and Cao et al. (2013)). We also control for the inflation rate and default spread, as Bali, Brown, and Caglayan (2011) find that exposures to these two factors are significantly related to hedge fund returns. We further include the momentum factor, as Griffin and Xu (2009) find that hedge funds engage in momentum strategies. Finally, we control for illiquidity risk using Pastor and Stambaugh’s (2003) liquidity factor.
content_copy
We perform a battery of sensitivity tests. First, instead of tracking returns from the month immediately following portfolio formation, we skip one month. Second, to address concerns about the precision of sentiment beta estimates, we use different combinations of risk factors as control variables in regression 1).
content_copy
To control for known determinants of hedge fund performance, we perform Fama-MacBeth (1973) cross-sectional regressions of fund excess returns or alpha on sentiment beta, along with various fund characteristics and style dummies. Specifically, we run the following cross-sectional regression of fund excess returns on sentiment beta: where ri,t+1 is the fund excess return in month t + 1, and βˆS i,t is fund i’s sentiment beta estimated from regression model (1) using fund returns in the 36-month rolling window from month t – 35 to month t. That is, the key independent variable—sentiment beta—is estimated from a backward-looking window prior to the return evaluation period for the dependent variable of the regression. The control variables x are predetermined fund characteristics including fund size, fund age, management fee, incentive fee, high-water mark dummy, lockup period, redemption notice period, and fund style dummies.
content_copy
We focus our collection efforts on employee ratings of their firm and of senior management. We then use the median rating across all employee ratings in a given fiscal quarter, resulting in two variables (SeniorMgmt and Firm). Both variables range from 1 (the lowest rating) to 5. Decreases in ratings imply reductions in perceptions.
content_copy
We control for return-on-assets because employee satisfaction may be increasing in firm profitability. Moreover, the media may be more likely to cover profitable firms (e.g. Google or Apple). Thus, controlling for profitability limits bias related to media coverage. We control for market-to-book because employees at non-growth firms may rate their firms lower relative to employees at growth firms. We control for leverage because employees at highly levered firms may rate their firms lower because they are concerned about bankruptcy risks. We control for size because large firms receive more media coverage than small firms. The media may “target” large firms for scrutiny more than other firms because larger firms are more well-known (e.g., Chen et al. 2019). Thus, we limit media coverage bias by controlling for size. Moreover, employees at large firms may rate their employers higher than employees at small firms because their salaries are high and/or their jobs are secure. We control for buy-and-hold returns to control for any public information or sentiment that may influence employee ratings.13 Return on assets, leverage and size are seasonally lagged to ensure that we do not control for our hypothesized effect.
content_copy
At this point, the research assistant pulled out a basket filled with sameflavored macaroons and randomly administered one of two treatments. In one condition, the research assistant invited customers to pick a macaroon, explaining that the bakery had committed to give a free macaroon as a gift in exchange for filling out the survey. Thus, in this condition, the contractual nature of the perk was salient (i.e., high contractuality). In the other condition, the research assistant invited customers to pick a macaroon, explaining that it was a gift from the bakery. Thus, in this condition, the perk was not portrayed as being given out of contractual obligation (i.e., low contractuality). Note that customers in both conditions received an identical gift that was equally unexpected, thus any effect observed is unlikely to stem from differences in the perceived value of the perk or in how surprising the perk was.
content_copy
Then, they were randomly assigned to two experimental conditions in which we manipulated the perceived contractuality of a perk.
content_copy
In the high contractuality condition, participants were told that their order was delivered and were given a note informing them that the restaurant included a $15 bonus gift card that they could redeem in the next three days between 10 A.M. and noon. In the low contractuality condition, participants received the same information, except the bonus gift card had no redemption limitations (see Web Appendix B for all stimuli).
content_copy
We tested the robustness of our analysis to various approaches for constructing this variable to ensure that this particular definition was not driving our results.
content_copy
For robustness, we consider two alternative measures of sentiment fluctuations: the monthly change in the University of Michigan consumer sentiment index, which is based on surveys of household confidence in the economy, and the FEARS index of Da, Engelberg, and Gao (2015), which captures sentiment changes based on Internet search volume for keywords that reveal investor concerns about the economy.
content_copy
Our results are robust to including lagged (by one quarter) and contemporaneous forms of these variables.
content_copy
We expected this manipulation to affect perceived contractuality because the redemption requirements for the high-contractuality gift card were more specific and restrictive than those for the low-contractuality gift card.
content_copy
Each participant was presented with a randomly selected subset of 10 of the 50 brands and, for each brand, answered warmth and brand attitude measures in random order
content_copy
The experiment was a one-factor (brand name gender: feminine, masculine) between-subjects design.
content_copy
Participants were told they would be evaluating a one-minute video from a channel of their choice. All participants were offered a choice between watching a channel with one of our pretested names or watching a video of equal length from a randomly selected YouTube channel. In the feminine name condition, participants chose between the “Nimilia YouTube Channel” and a randomly selected YouTube channel; in the masculine name condition, participants chose between the “Nimeld YouTube Channel” and a randomly selected YouTube channel. They next completed the same four-item warmth measure used in Study 2, along with a single-item measure of pleasantness (“To what extent does the Nimilia [Nimeld] Channel sound pleasant?”), in random order. The dependent variable was channel choice.
content_copy
The experiment employed a one-factor (participation incentive: cash, feminine-named product, masculine-named product) within-subject design
content_copy
Students were welcomed to the study and told that, as a thank-you for participating, they could choose either $.50 or one bottle of small-batch hand sanitizer. The hand sanitizers were commercially available customizable products with either a linguistically masculine (Nimeld) or feminine (Nimilia) name on the bottle and the label. (Note: this study was conducted before onset of the COVID-19 pandemic, and the small colorful bottles presented as both fun and functional).
content_copy
Finally, we asked whether they had discussed the lab session or incentive choice with anyone else before participating.
content_copy
Participants were randomly assigned to conditions in a 2 (brand name gender: feminine or masculine) 2 (typical user: male or female) between-subjects design.
content_copy
We manipulated typical user gender by telling participants that the sneakers were either for men or for women, which allowed us to hold the product constant.
content_copy
We included a suspicion check to rule out possible demand effects and manipulation checks asking who the most typical user of the product was (men: 1, women: þ1, both: 0) and how masculine or feminine the brand name seemed (1 ¼ “very masculine,” and 5 ¼ “very feminine”). The suspicion check indicated that less than 1.5% of participants suspected that the purpose of the study was to examine brand name and user gender.
content_copy
We conducted 2 (brand name gender: feminine or masculine) 2 (typical user: male or female) analyses of variance (ANOVAs) on the brand name gender and typical user gender manipulation check measures.
content_copy
Participants were randomly assigned to conditions in a 2 (brand name gender: feminine or masculine) 2 (product category: utilitarian or hedonic) between-subjects design.
content_copy
We measured warmth by asking participants to indicate the extent to which each brand sounded tolerant, warm, good-natured, and sincere (Fiske et al. 1999; a ¼ .96). Brand attitude was assessed with five items adapted from Chaudhuri and Holbrook (2001) and Brakus, Schmitt, and Zarantonello (2009), asking the extent to which participants were or would be loyal to, committed to, buy, choose, and recommend the brand (a ¼ .95). The brand attitude measure incorporated elements of loyalty because Kervyn, Fiske, and Malone (2012) find that warmth is positively related to brand loyalty. (We also measured competence but do not examine it here; for full details, see Web Appendix C). All items were measured on seven-point scales (1 ¼ “not at all,” and 7 ¼ “very much”).
content_copy

Specifying qualitative data collection method(s)

cc
content_copy

Specifying quantitative data analysis method(s)

We test the hypotheses in a fixed-effects DID framework (Angrist and Pischke 2009): Yi,j,t β0 + β1AFTERt × AWARDi,j + κi + φt + ψi,j,t + i,j,t: Yi,j,t is the dependent variable of interest in month t for complementor j’s app i, AFTERt equals one if month t is after the award, AWARDi,j is an indicator variable for whether complementor j’s app i received an award, κi are app fixed effects, and φt are month fixed effects. The coefficient of interest is β1. When comparing winners to runners-up, β1 is the relative change in Y due to award winning, whereas nomination effects are differenced out. When comparing winners to the matched developers, β1 captures both the effects of the nomination and of award winning. The vector ψi,j,t contains time-variant controls outlined above. We omit the regressor AFTER from the model because award time is homogenous and would be collinear to time fixed effects. The main term AWARDi,j is not included because it is collinear to the app fixed effects. To estimate the binary dependent variables, we use a conditional logit estimator. We cluster standard errors around app developers to adjust for the developer-app structure.
content_copy
First, we estimate the plain models without controls. The results are consistent (panel A). Second, we estimate the models using ordinary least squares (OLS), and the hypotheses remain confirmed (panel B).
content_copy
Next, we provide a similar comparison using a regression analysis. Our goal is to account for differences in terms of industry (market niche), Firm Size, Funding and Competition Controls which may in part explain the difference between platform and non-platform companies. Since our outcome variable is a time variable (i.e. Number of years after founding that a company makes their first acquisition), we use a survival model. This methodology takes into account the natural truncation that can occur with such a time-based variable. A linear representation of our estimating equation is as follows: Our analysis is at the level of each individual company (i) and using a Cox-Proportional Hazzard model, we estimate whether the Hazard Ratio, the likelihood of a company acquiring in any given year, differs for platform and non-platform companies. This difference is inferred from the coefficient ?. @ represents the vector of control variables described above.
content_copy
To control for potentially confounding factors, we move to a regression analysis. The unit of observation is at the level of each individual acquisition.
content_copy
We conducted a stepwise logistic regression, first regressing name gender score onto a binary outcome variable indicating whether the brand appears on the Interbrand Top Brands list or not, and then adding control variables for industry, number of employees, and years in business (Table 1, Panel B).
content_copy
A chi-square test showed that participants were more likely to choose the Nimilia hand sanitizer than any other option (49.3%), followed by the $.50 (36.0%), and Nimeld hand sanitizer (14.7%; w2(2, 150) ¼ 27.52, p < .001, d ¼ .94; Figure 3b), in support of H1.
content_copy
A 2 (brand name gender: feminine or masculine) 2 (typical user: male or female) ANOVA with attitude as the dependent variable revealed a nonsignificant main effect of typical user gender and a main effect of brand name gender (F(1,871) ¼ 41.97, p ¼ .001, Z2 ¼ .04), such that attitudes toward feminine brand names were more positive compared with those for masculine brand names (Mfeminine ¼ 3.95, SD ¼ 1.23 vs. Mmasculine ¼ 3.39, SD ¼ 1.32).
content_copy
A 2 (brand name gender: feminine or masculine) 2 (typical user: male or female) ANOVA with perceived warmth as the dependent variable yielded a nonsignificant main effect of typical user gender and a significant main effect of brand name gender (F(1, 871) ¼ 108.90, p < .001, Z2 ¼ .11), such that feminine brand names were generally perceived as warmer than masculine names (M ¼ 4.46, SD ¼ 1.38 vs. M ¼ 3.50, SD ¼ 1.33).
content_copy
Analysis with PROCESS Model 4 shows that warmth mediates the effect of name gender on brand attitude (B ¼ .55, 95% CI ¼ [.43, .67] with 5,000 bootstrapped resamples). To determine whether typical user moderates this indirect effect, we tested the moderating role of typical user gender on the link from name gender to warmth (path a;Hayes [2013]Model 7) and on the link from warmth to brand attitude (path b; Hayes [2013] Model 14).
content_copy
Mediation analyses (PROCESS Model 4) show that warmth mediates the effect of name gender on attitudes (B ¼ .37, 95% CI ¼ [.24, .49] with 5,000 bootstrapped samples). We tested moderated mediation on the a and b paths, separately and together (Hayes [2013] PROCESS Models 7, 14, and 58, each with 5,000 bootstrap resamples). Only Models 7 and 58 were significant, but Model 58 produced a larger index of moderated mediation (Figure 5).
content_copy
We conducted mixed within- and between-subjects analyses using the linear mixed model function in the nlme package (Bates et al. 2015) in the statistical software R to examine possible rater effects, following the recommendation of Bates et al. (2015). A linear mixed-effect model with brand name gender and warmth as fixed effects and separate intercepts for each participant as random effects was consistent with the results in the PROCESS model.
content_copy
We test our Hypothesis of the effect of addback statutes on future innovation using the following OLS models (1) and (2). All the subscripts are suppressed, as all independent variables are measured in the same time perioddyear t. The dependent variables in Models (1) and (2) are patent count (Ln_NPat3) and citation count (Ln_NCite3), respectively. Patent count is the natural logarithm of one plus the number of patents filed by a firm in year tþ3, while citation count is the natural logarithm of one plus the truncation-adjusted number of non-self-citations received by patents that are filed in year tþ3. We use the logged form of patent count and citation count as the dependent variables to address skewness. In both models, the key independent variable is Addback, which equals 1 if a firm is affected by addback statutes in year t. The coefficient on Addback indicates how the innovation of affected firms changes after being affected by the addback statutes compared with that of the other firms.
content_copy
We first use portfolio sorts to examine the relation between sentiment beta and hedge fund returns. Each month starting in December 1996, we form 10 equal-weighted portfolios of hedge funds based on the fund sentiment beta (i.e., the loading on the sentiment changes index) estimated from a rolling window of the most recent 36 months (including the current month), with the first rolling window spanning the period from January 1994 to December 1996.14 We then track the portfolio returns over the next month (starting in January 1997). These portfolios are rebalanced each month to generate a time series of returns from January 1997 to December 2018. Specifically, each fund’s sentiment beta is estimated by regressing fund excess returns on the sentiment changes index controlling for standard risk factors. In month t, for each fund with at least 30 return observations during the 36-month rolling window, we perform the following time series regression: where ri,t is the excess return (in excess of the one-month T-bill rate) on fund i in month t, sentiment is the sentiment changes index, βS is sentiment beta, and the vector f contains the Fung-Hsieh factors, momentum, liquidity, inflation, and the default spread. Thus, for month t, the rolling window covers month t – 35 to month t. The use of a rolling window allows for time variation in the beta estimates.
content_copy
We track the returns for the decile portfolios over the next month after portfolio formation. These portfolios are rebalanced each month. Finally, we estimate alpha (i.e., the risk-adjusted return) by regressing the time series of the excess returns of each decile portfolio on the Fung-Hsieh seven factors, the momentum factor, and the liquidity factors.15 Accordingly, the spread in alpha between the two extreme decile portfolios (i.e., portfolios 10 and 1) reveals performance dispersion attributed to sentiment beta. In the test, we calculate t-statistics using Newey-West (1987) standard errors with two lags, where the number of lags is based on autocorrelations in monthly hedge fund returns.16
content_copy
To test our hypotheses, we perform generalized difference-in-differences regressions with staggered treatment. The generalized difference-in-differences approach replaces the post and treatment main effect variables (as would be used in a traditional difference-in-differences regression) with unit and time fixed effects. We use generalized difference-in-differences because our treatment is staggered. In an additional test, we use a propensity-score matched sample.
content_copy
In our context, we use the following specification to test whether employee ratings decrease after tax avoidance news (H1 and H2, respectively): In this equation, Rating is (1) median employee ratings of senior management (SeniorMgmt) or (2) median employee ratings of the firm (Firm) in a given quarter.14 TaxNews takes a value of 1 in the first quarter we identify tax avoidance news for firm i and in all subsequent quarters. All other quarters are coded to 0.15 This variable is equivalent to a post variable interacted with a treatment variable in a traditional difference-in-differences strategy.
content_copy

Specifying qualitative data analysis method(s)

content_copy

Opening the results section

Tables 3 and 4 report the results of the hypothesis tests.
content_copy
Table III reports results for both hedge fund excess returns and alpha across the sentiment beta-sorted portfolios.
content_copy
In Table 1, we provide the number of companies receiving tax avoidance news coverage for the first time in our sample period.
content_copy

Reporting descriptive statistics in quantitative research

We begin with a descriptive analysis reporting the years after founding when companies make their acquisitions in Table 3: Part B. Platform companies acquire companies from the same market niche, on average, 4.2 years after founding, while they acquire companies from different market niches, on average, 5.1 years after founding. This difference is significant (Δ= - 0.97; S.E. = 0.37; t = 2.61; p<0.01), suggesting that platform companies acquire companies from the same market niche, on average, one year earlier than they acquire companies from different market niches. In contrast, nonplatform companies acquire companies from the same and from different market niches, on average, 6.1 years after founding. This suggests support for Hypothesis 2, and the lack of any difference in the case of non-platform companies, suggests that this is only the case for platform companies.
content_copy
We begin with a descriptive analysis comparing the years after founding when companies make their acquisitions. These results are reported in Table 3: Part C. Platform companies acquire other platform companies on average 4.1 years after founding, while they acquire non-platform companies, on average, 5.1 years after founding. This difference is significant (Δ=1.01; S.E. = 0.376; p<0.001), suggesting that platform companies do acquire other platform companies before they acquire nonplatform companies, on average. On the other hand, non-platform companies acquire non-platform companies (i.e. Same business model), on average, 6.2 years after founding, while they acquire platform companies (i.e. Different business model), on average 5.4 years after founding. This suggests that platform companies are likely to acquire companies with the same business model earlier than nonplatform companies.
content_copy
The descriptive statistics of the variables used in our regression analyses are presented in Table 1. All variables (except TobinQ) are winsorized at the top and bottom 1 percent. As we can see in Panel A, 26 percent of firm-year observations in the sample have Addback¼1. Consistent with prior studies on innovation (e.g., Atanassov and Liu, 2019; He and Tian, 2013), the distributions of patent count and citation count are highly skewed, as there are many firm-year observations with zero patents. The mean of NPat3 (the number of patents filed by a firm in year tþ3) is 13.40, and the mean of NCite3 (the truncation-adjusted number of citations received on patents filed by a firm in year tþ3) is 4.43. Fig. 1 further shows the percentage of firms affected by addback statutes by year. As expected, we find that more firms are affected by the addback statutes in later years. In the year of the first state's adoption of an addback statute (1999), 19 percent of the observations are affected. In the last year of our sample (2005), the percentage is almost tripled and reaches 57 percent. This sharp increase in the percentage provides a powerful setting to examine the consequences of the addback statutes. Also, these are consistent with our expectation because more states adopted the addback statutes in the later part of our sample period.
content_copy
The average fund return is 0.59% per month. The mean (median) assets under management is $181 million ($56 million), and the average fund age is 81 months. The mean (median) management fee is 1.35% (1.50%), while the mean (median) incentive fee is 15.05% (20%). About two-thirds of the funds use a high-water mark provision that requires the funds to recover previous losses before charging the incentive fee. In addition, 32% of the funds require a lockup period, and the redemption notice period is 1.44 months on average.
content_copy
In total, 143 of the 495 firms in our sample receive tax avoidance news coverage at some point in our sample. Overall, tax avoidance news coverage is fairly dispersed across time. Descriptive statistics and correlations are reported in Table 2. In Panel A, we report descriptive statistics for variables used in our analyses. SeniorMgmt has a mean value of 2.989 and a median value of 3. These values suggest employees are fairly neutral towards senior management on average. Firm has a mean (median) value of around 3.402 (3.5), which suggest employees hold slightly positive perceptions of their firms on average. TaxNews, our binary regressor of interest, has a mean value of .203. Thus, 20.3 percent of firm-quarters are treated observations.16 ln(AT) has a mean (median) value of 9.747 (9.620), which suggest that firms in our sample are generally large, as expected of S&P 500 firms. MTB has a mean (median) value of 3.806 (2.722), respectively. Leverage takes a mean (median) value of .234 (.212), which is roughly 23 percent of seasonally lagged total assets on average in our sample. We find that average ROA is around .041 and median ROA is around .037, suggesting that firms in our sample are generally profitable, as expected of S&P 500 firms. Average and median BHR are 13.2 percent and 12.4 percent, respectively.
content_copy
In Panel B, we report descriptive statistics on our Glassdoor.com ratings to describe the characteristics of employee raters in our sample. Num_Reviews is the number of reviews of a given firm in a given quarter. On average, firms receive 64 ratings per quarter. Median Num_Reviews is around 18 on average. We find that 82 percent of employee-raters are full-time employees (%FullTime) for an average firm-quarter. 59 percent of employee-raters are current employees of the firm they rate (%Current) in an average firm-quarter. We find that average tenure (Tenure) across our employee-raters at their current employer is around 2.5 years. We find that the average number of tax-related reviews in a given firm-quarter is .085. In Panel C, we report our correlation matrix. We find that TaxNews is positively and significantly (at the 10 percent level) related to both SeniorMgmt and Firm. However, the correlation coefficients are close to zero.
content_copy

Reporting inferential statistics in quantitative research

We find support for Hypothesis 1.
content_copy
In Table 3, columns (1) to (3) report the estimates for Equation (1) with UPDATE as the dependent variable. The coefficients are reported in log-odds notation. Column (1) shows that winning an award has a positive effect on the likelihood of an app update, with runners-up as a control benchmark. The odds of an app update are e0.472 1.60 times higher for app developerswhowon the award. Column (2) uses CEM matches as a control group, supporting the positive effects (e0.609 1.84). Column (3) uses PSM matches as a control group and documents odds comparable to the baseline (e0.641 1.90). Columns (4) to (6) estimate the effect of the award on NEWAPP. Column (4) indicates a significantly negative effect. For recipients, the odds of releasing a new app are e−2.837 0.06 times that of runners-up. Column (5) uses CEMmatches as the benchmark and suggests effects similar to the baseline (e−3.226 0.04). Column (6) uses PSM matches as the benchmark, and the coefficient differs only marginally fromthe baseline (e−3.461 0.03). Taken together, these data supportHypothesis 1. Taken together, these data support Hypothesis 1.
content_copy
We can also observe evidence in line with Hypothesis 2.
content_copy
Table 4 reports the estimates. Column (1) is the baseline, predicting MULTIHOMED and using runners-up as a benchmark. For award recipients, the odds of multihoming an app are e1.052 2.86 times that of runnersup. Column (2) uses CEM matches as the control benchmark, further corroborating the positive effects (e1.012 2.75). Column (3) uses PSMmatches as the control group, also corroborating the baseline in magnitude and significance (e1.055 2.87). Taken together, these data provide evidence in line with Hypothesis 2.
content_copy
Finally, we find support for Hypothesis 3. Column (4) of Table 4 is the baseline, predicting CAT_NUMAPPS. The DID estimate suggests that award recipients’ market categories experience an increase in app releases. Column (5) confirms the positive effect when using CEMmatches as the control group. Column (6) also confirms the positive effect when using the PSM control group. Online Appendix Table A5 reports that the estimate differs only marginally when restricting the control group to apps that are not in the same categories as award winners’ apps. Therefore, the composition of the categories within the groups seems to exert little influence on the results. Together, these analyses indicate support for Hypothesis 3.
content_copy
Hypothesis 1 predicts that platform companies are likely to make acquisitions earlier than nonplatform companies. Empirically this implies that Firm Age (years since founding) of when a platform firm makes its first acquisition, we would be lower on average, than for non-platform companies. In Table 3: Part A, we report these values. On average, platform companies that acquired made their first acquisition 3.6 years after founding, while non-platform companies that acquired made their first acquisition on average 5.5 years after founding. This difference is statistically significant (Δ=1.8; S.E. = 0.27; t = 6.52) based on a simple t-test providing evidence in support of Hypothesis 1.
content_copy
We report the results in Table 4. In Columns 1 through 4 we introduce the different control variables. The difference between platform and non-platform companies remains statistically significant (Table 3, Column 4: ? = 0. 256, S.E.= 0.083, p<0.01). The hazard ratio reported in Column 4 indicates that the likelihood of a platform company acquiring is 50% greater than that of a non-platform company in each period. Additional specifications are reported in Appendix A, Table A1. These results provide support for Hypothesis 1.
content_copy
The results provide further support for Hypothesis 2, indicating that platform companies are likely to acquire companies from the same market niches earlier, but then acquire companies from different market niches later.
content_copy
We report the results of this regression in Table 6. We report the results of a logistic regression without any controls in Columns 1 and 2. In Column 3 we introduce the full set of control variables. The coefficients (b=-1.48; S.E. = 0.053; p<0.001) indicate that if platform companies were to acquire in the first year after founding, there is a 75% likelihood that it would be another platform company. However, in the ten years after founding, this is likely to decline by 25%. In contrast, non-platform companies within our sample are approximately 80% likely to acquire a non-platform company and there is no negligible change over the first ten years. In Column 4 we report the results using an OLS regression which is more easily interpretable. This also provides a comparison to the Heckman results in the next column, since we are not able to perform this on a logistic regression. In Column 5, we report the Heckman correction with the excluded instrument based on the probability of an acquisition from Table 3. Comparing Column 5 and Column 6 can also allow us to interpret how much this selection approach is influencing the regression results. The results remain consistent across both specifications. The coefficients when accounting for selection suggest an even larger decline of 4.6% per year decrease in the likelihood of a platform company acquiring another platform company. These results provide support for Hypothesis 3.
content_copy
We report OLS results, but results are consistent with logistic regression. In Columns 1 and 2, we introduce the baseline results and in Column 3 we introduce the full set of controls. We omit industry controls as this sample was based on a narrow set of industries. The coefficient for the interaction term in Column 3, indicates a 3.5% decline (b=-0.035; S.E. = 0.013; p < 0.01) each year in the likelihood of a platform company acquiring a competitor from the same market niche, in comparison to a non-platform company. This is consistent with the results from Section 5.3 and supports Hypothesis 2.
content_copy
The results for platform companies indicate a decrease of approximately 4.5% (b=-0.045; S.E. = 0.013; p<0.001) per year, while the results for non-platform companies indicate an increase of approximately 5.2% per year (b=0.052; S.E. = 0.023; p<0.05). This provides further support for Hypothesis 2.
content_copy
WOM revealed a significant effect of contractuality (w2 (1, N ¼ 398) ¼ 17.17, p < .001, Cramer’s V ¼ .21): 71% of participants in the low contractuality condition were willing to share WOM about the restaurant, relative to 51% in the high contractuality condition (see Table 2).
content_copy
A one-way ANOVA on relational value revealed a significant effect of contractuality (F(1, 396) ¼ 11.85, p < .001, Zp 2 ¼ .029): Participants in the low contractuality condition reported greater perceptions of relational value (M ¼ 6.19, SD ¼ .88) than participants in the high contractuality condition (M ¼ 5.84, SD ¼ 1.17). Contractuality also had a significant effect on motivation to help (F(1, 396) ¼ 5.66, p ¼ .018, Zp 2 ¼ .014): Participants in the low contractuality condition reported greater motivation to help the restaurant (M ¼ 6.05, SD ¼ 1.04) than those in the high contractuality condition (M ¼ 5.79, SD ¼ 1.15).
content_copy
Consistent with our prediction, the analysis (based on 5,000 bootstrap samples) revealed that relational value and motivation to help serially mediated the effect of contractuality on WOM (b ¼ .14, 95% CI ¼ .037 to .294). The results remained unchanged after controlling for reciprocity, reactance, surprise, effort, and monetary value (b ¼ .05, 95% CI ¼ .002 to .135).
content_copy
In addition, we tested whether any of these processes mediated the effect of contractuality on WOM in parallel. We found that while relational value significantly mediated the effect of contractuality on WOM, reciprocity, effort, surprise, reactance, and monetary value did not.
content_copy
A one-way ANOVA on likelihood to generate WOM revealed a significant effect of contractuality (F(1, 301) ¼ 11.05, p ¼ .001, Zp 2 ¼ .035; Table 2). As predicted, participants in the low contractuality condition were more likely to engage in WOM (M ¼ 6.23, SD ¼ .81) than those in the high contractuality condition (M¼5.89, SD¼.94).
content_copy
Participants in the low contractuality condition reported greater perceptions of relational value (M ¼ 6.18, SD ¼ .85) than participants in the high contractuality condition (M¼ 5.80, SD¼ 1.02). Consistent with our theoretical account, contractuality also had a significant effect on motivation to help (F(1, 295) ¼ 8.96, p ¼ .003, Zp 2 ¼ .029, six responses missing): participants in the low contractuality condition reported greater motivation to help the restaurant (M ¼ 5.86, SD ¼ .99) than participants in the high contractuality condition (M ¼ 5.50, SD ¼ 1.11).
content_copy
As predicted, the analysis (based on 5,000 bootstrap samples) revealed that relational value and motivation to help serially mediated the effect of contractuality on WOM (b ¼ .09, 95% CI ¼ .037 to .158). This result remained significant after controlling for reciprocity, reactance, surprise, effort, and monetary value (b ¼ .05, 95% CI ¼ .011 to .096).
content_copy
As we hypothesized, linguistically feminine brand names predicted inclusion on the performance-based Interbrand Global Top Brands list (vs. the non-performance-based Thompson Reuters list; p ¼ .031). These findings offer preliminary, real-world evidence of a feminine brand name advantage (H1).
content_copy
Overall brand familiarity was above the scale midpoint, indicating that participants were generally familiar with the brands (M ¼ 5.35, SD ¼ 1.83; t(5,239) ¼ 53.15, p < .001). Familiarity was a significant covariate when included in the model (t(5,234) ¼ 26.39, p < .001) but did not alter the direction or significance of the results.
content_copy
Both models produced significant indexes of moderatedmediation (Model 7:B¼.29, 95%CI¼[.08, .50];Model 14: B¼.14, 95%CI¼[.04, .25]).
content_copy
The results of Study 4 support H3, demonstrating the moderating role of typical user gender: the feminine brand name advantage is neutralized when the typical user is male, and attitudes toward linguistically masculine and feminine brand names are equally positive.
content_copy
As we predicted, linguistically feminine names produced more favorable attitudes toward hedonic than utilitarian products, in support of H4.
content_copy
The coefficient on Addback is 0.0501 in Column 1 and 0.0477 in Column 2, both significant at the 5 percent level, indicating that operating in a state that has adopted an addback statute is negatively associated with the number of patents filed three years later. Therefore, after the adoption of an addback statute in a state, the number of patents filed by affected firms in year tþ3 decreases by 4.77 percentage points, which is equivalent to 0.639 patents (¼13.4 4.77 percent) and 0.536 percent of its standard deviation (¼0.639 ÷ 119.30).
content_copy
Overall, these findings are consistent with our Hypothesis that the adoption of addback statutes significantly reduces the level of corporate innovation of affected firms, reflected in the 4.77 percentage point decrease in patent count and the 5.12 percentage point decrease in citation count.
content_copy
The decile portfolio with the highest sentiment beta (i.e., portfolio 10) delivers an average excess return of 0.58% (tstatistic = 3.48) permonth and an alpha of 0.51% (t-statistic = 2.98) per month, indicating significantly positive abnormal performance, while the decile portfolio with the lowest sentiment beta (i.e., portfolio 1) shows an average excess return of 0.27% (t-statistic = 1.77) per month and an alpha of −0.08 (t-statistic = −0.51) per month. The return spread between the two extreme decile portfolios is 0.31% (t-statistic = 3.16) per month and both economically and statistically significant.
content_copy
While the portfolio return does not increase strictly monotonically with sentiment beta, the top three portfolios with the highest sentiment betas are also the top three portfolios with the highest average excess returns and alphas, while the portfolio with the lowest sentiment beta also has the lowest average excess return and alpha. Thus, the results from portfolio sorts indicate that sentiment beta is significantly positively related to both hedge fund excess returns next month and alpha after adjusting for standard risk factors.
content_copy
As shown in Panel A, the spread in alpha is positive for more than two-thirds of the months over our sample period. Panel B plots the cumulative spread in alpha between the two extreme decile portfolios sorted by sentiment beta. Again, the plot shows a persistent difference in risk-adjusted returns between hedge funds with high and low sentiment betas.
content_copy
Finally, we find that hedge fund sentiment beta displays a fair amount of persistence in the short run. For example, about 59% (45%) of the hedge funds placed in the top sentiment beta decile in a given month will continue to be in the top decile six months (one year) later. These additional findings are reported in the Internet Appendix.
content_copy
These results are consistent with those from portfolio sorts and indicate that sentiment beta can positively and significantly predict hedge fund performance in the cross-section.
content_copy
We obtain similar evidence from multivariate regressions. After including fund characteristics and style dummies, we continue to find that sentiment beta is significantly and positively associated with hedge fund performance next month. For example, in the regression with fund excess return as the dependent variable, the slope coefficient on sentiment beta is 0.16 (t-statistic = 2.84). In the regression with alpha as the dependent variable, adding fund characteristics as controls reduces the coefficient on sentiment beta slightly from 0.14 to 0.12 (t-statistic = 3.16).
content_copy
Our finding therefore suggests that these fund characteristics, traditionally used as proxies for hedge fund skill in the literature, account for only a small portion of the effect of sentiment beta on hedge fund performance.
content_copy
In sum, our results from both portfolio sorts and cross-sectional regressions show a significant and positive relation between sentiment beta and hedge fund performance, even after adjusting for common risk exposures and controlling for fund characteristics.
content_copy
We find that the coefficient on TaxNews is negative and significant at the 1 percent level across both Columns (1) and (2). In Columns (3) and (4), we find that the coefficient on TaxNews is negative and significant at the 5 percent level. Overall, these results are consistent with employee perceptions of their senior management and their firms decreasing following tax avoidance news, consistent with H1 and H2.18
content_copy
Across all columns, we find that the coefficient on Num_mention is negative and significant at the 1 percent level. These results are consistent with media coverage intensity increasing perception changes on senior managers and firms following tax avoidance news.
content_copy
Our results are reported in Panels A and B of Table 4. In Panel A, we use Consumer as our measure of consumer-facing firms. In Columns (1), (2) and (3), we find that the coefficient on 𝑇𝑎𝑥𝑁𝑒𝑤𝑠 × 𝐶𝑜𝑛𝑠𝑢𝑚𝑒𝑟 is negative and significant at the 1 percent level. In Column (4), we find that the same coefficient is negative and significant at the 5 percent level. In Panel B, we use Retail as our measure of consumer-facing industries. In Columns (1) and (2) ((3) and (4)), we find that the coefficient on 𝑇𝑎𝑥𝑁𝑒𝑤𝑠 × 𝑅𝑒𝑡𝑎𝑖𝑙 is negative and significant at the 10 percent (5 percent) level. Overall, these results are consistent with firms and their senior management in consumerfacing industries facing larger decreases in employee perceptions compared to firms in other industries.
content_copy
This test provides evidence that tax news is more likely to elicit a response from employees compared to tax information from financial statements.
content_copy
These results are consistent with employees not reacting to tax avoidance from financial statements.25
content_copy
These results are consistent with both accounting and non-accounting related employees perceiving senior managers negatively following tax avoidance news.
content_copy
These results provide evidence that non-accounting employees’ perceptions of firms fall following tax news. We also find that accounting employees perceive senior managers significantly more negatively than they perceive firms following tax news (F statistics of 5.52 and 5.11 with p-values of .02 in both specifications). F tests on non-accounting employees-sample reveal weak or no statistically significant difference in the negative ratings between senior managers and firms following tax news (F statistics of 2.89 and 2.56 with p-values of .09 and .11, respectively). Thus, we find evidence that accounting employees perceive senior managers significantly more negatively than firms following tax news. However, non-accounting employees do not seem to differentiate between their senior managers and firms in attributing blame for tax news.
content_copy
Next, we provide evidence that our results are not substantially affected by tax news that occurred prior to our sample period using three approaches. First, we limit our sample to the news events identified by Chen et al. (2019) in the S&P 250 to provide a benchmark and also to provide evidence that our results are not sensitive to our identified tax avoidance news. Second, we find that our results are robust after dropping firms that received tax news coverage in the three years prior to our sample period (2005-2007) (see Appendix A in Chen et al. 2019). Last, we find evidence that our results are robust to explicitly controlling for prior tax news based on the Chen et al. (2019) sample.
content_copy
The coefficient on NEWAPP drops below the significance level (p 0.056). Finally, we estimate the models using a different set of controls for app prices.
content_copy
A total of 19.6% of customers in the low contractuality condition started following the bakery after the intervention, whereas only 8% of customers in the high contractuality condition did so (exp(b) ¼ 3.37, Wald (1, N ¼ 101) ¼ 3.53, p ¼ .060). No other effects were significant.
content_copy
Study 2 supports the hypothesized process in H2: feminine brand names are positively associated with warmth, which is positively associated with brand attitudes. However, the direct effect of brand name gender on attitude was not significant.
content_copy
However, when the product category was utilitarian, there was no significant difference in warmth perceptions between the linguistically feminine and masculine brands (Mfeminine ¼ 4.16, SD ¼ 1.31 vs. Mmasculine ¼ 3.98, SD ¼ 1.25; F(1, 434) ¼ 2.14, p ¼ .14., d ¼ .14).
content_copy
We find insignificant interaction effects in Columns 1 and 3. Further, the interaction effect is negative and significant in Column 2.35 Thus, inconsistent with the alternative explanation, the effect of addback statutes on innovation is not more pronounced for firms that are more financially constrained. Further, we find significantly negative coefficients on Addback in Columns 4 to 6. These findings suggest that the addback statutes have a negative effect on innovation among firms that are not financially constrained. Therefore, the effect on innovation cannot be simply attributed to crackdown on tax avoidance increasing financial constraints and thus reducing investment in innovation.
content_copy
As shown in Panel B of Table 7, none of the economic conditions significantly explains the adoption decisions.
content_copy
As shown in Column 1, the coefficient on Addback is insignificant, which suggests no significant change in income shifting to foreign countries after the adoption of addback statutes.
content_copy

Reporting statistical power and robustness in quantitative research

In this section, we report several tests for the model assumptions, rival explanations, and sensitivity, as well as additional analyses. Table 5 summarizes them.
content_copy
Although our sample size is comparable to those of prior studies of awards (e.g., Kov´acs and Sharkey 2014), it still represents a small sample. To help interpret the data, we conduct an ex post power simulation for our own sample size and compare it with other sample sizes, as applied, for instance, in Gubler et al. (2018). More specifically, we randomly generate datasets, each with a structure identical to our panel and using the same ratio between treatment and control units. As minimum detectable effects, we take OLS estimates of Equation (1). Online Appendix Figure A3 plots the power for different sample sizes and a significance level of 5%. The statistical power associatedwith our sample is above the conventional threshold of 0.8 forHypothesis 2 (i.e.,MULTIHOMED) and Hypothesis 3 (i.e., CAT_NUMAPPS). For UPDATE, the power is 0.56, and a sample size of approximately 233 developers would be required to make a judgment of confidence above the conventional power threshold. For NEWAPP, the power is 0.67, and a sample of approximately 170 developers would be required for a judgment above the conventional power threshold. Nevertheless, it is important to consider the power for Hypothesis 1 in light of the fact that it was corroborated across different model specifications, an alternative variable, and in interviews.
content_copy
We also test the robustness of our results to alternative approaches, including the approach of Shi et al. (2016).
content_copy
As a robustness check, we manually validated a sample of the data using human coders. Because the process of using human coders is time consuming, we could only validate a smaller number of observations, and focused only on companies that made multiple acquisitions starting with platform companies. This manual validation yielded a sample of 764 acquisitions (target acquirer pairs). Of the acquisitions in this dataset, 311 were by platform companies and 262 were by nonplatform companies. Student coders were instructed to code the following for each acquisition in the sample: 1) Whether the acquiring company was in fact a platform or non-platform company, 2) Whether the acquired company was a platform or non-platform company, and 3) Whether the platform and nonplatform companies were competitors from the same market niche.
content_copy
We note that these tests may suffer from low power.
content_copy

Reporting observations and patterns in qualitative research

content_copy

Reporting individual responses of subjects in qualitative research

To further investigate the model assumptions and tests, we report anecdotal evidence from the interviews with award winners. The interviews confirm that there is some unexpectedness in award-winning. For example, one interviewee stated that winning “was a surprise. […] They certainly did not give us any indication that we would win anything.” Another interviewee said, “We had no idea we’d win it. […] When the winner was announced we could hardly believe it.” Interviewed winners also stated that even after receiving the award, they were not sure of the exact criteria for award winning or nomination. The interviews also provide support for Hypotheses 1 and 2.
content_copy
Asked about the business impact of the award, one interviewee said, “this major recognition helped us feel that we were on the right track, and it greatly boosted our confidence and productivity. […] It was essential to stay aligned with […] the overall trend.”
content_copy

Reporting agreement within subjects in qualitative research

content_copy

Reporting disagreement within subjects in qualitative research

content_copy

Reporting key findings

content_copy

Reporting positive findings

content_copy

Reporting negative findings

content_copy

Reporting neutral findings

To demonstrate discriminant validity, we calculated average variance extracted. The analysis showed that the average variance extracted between perceived warmth and brand attitude was greater than their squared correlation, indicating that the variables represent distinct constructs. These results were consistent for all variables in all subsequent studies.
content_copy
Model 58 mediation analysis with product category as a moderator showed that warmth is a significant mediator for hedonic products (B ¼ .65, 95% CI ¼ [.45, .86]), but not utilitarian ones (B ¼ .11, 95% CI ¼ [.04, .26]).
content_copy

Reporting unexpected findings

We observed these results even though customers received the exact same perk (i.e., a free macaroon) and the perk was equally unexpected in both conditions.
content_copy
While this finding is not ideal, it should not alter the validity of the findings, given that the masculine name was evaluated as significantly masculine and the feminine name was evaluated as significantly feminine regardless of condition (all ps < .001).
content_copy
Although not expected, the data also suggest that product category alters perceptions of brand name warmth, such that linguistically feminine names are not perceived as warmer when the product is utilitarian.
content_copy

Reporting similarities and contrasts to findings of previous studies

content_copy

Summarizing the key findings and closing the results section

content_copy

Opening the discussion section

This study investigates the decision of platform firms to honor complementary innovation ex post with awards. In an empirical study of Google’s Android platform and the Google Play Award, we report three findings. First, awards increase recipients’ likelihood of releasing complement improvements and decrease their likelihood of releasing new complements. Second, awards increase recipients’ likelihood of multihoming. Finally, awards encourage other complementors to release new complements in recipients’ market niches.
content_copy
In this paper, we study the acquisition of platform companies, and ask when and which type of companies do platform companies acquire. Existing studies have not considered that acquiring other platform companies may be a way for platforms to scale their base of customers and suppliers. This study is the first to our knowledge, that focuses on the role of acquisitions in this process.
content_copy
Companies often engage in efforts aimed to spur WOM. Typically, these efforts involve the use of incentives (e.g., paid WOM agents, “refer-a-friend” promotions). Although incentivized WOM can be beneficial, it also has risks and costs for companies (e.g., lower message persuasiveness). As such, marketers have begun to seek ways to promote WOM in the absence of direct incentives. In this spirit, this research examines how marketers can tailor common marketing perks that are already used to serve other objectives to fuel WOM.
content_copy
The results of six studies offer convergent evidence for the feminine brand name advantage. Specifically, feminine brand names increase perceived warmth, which is associated with more favorable attitudes and increased choice for both real and fictitious brands. These findings suggest that the prevalence of feminine names among Interbrand top brands may be due in part to the ability of brand name linguistics to convey warmth. The feminine brand name advantage is neutralized when the typical product user is male and is stronger for hedonic than for utilitarian products. We demonstrate these effects empirically, using both real and hypothetical brands, as well as observationally, by examining secondary data.
content_copy
In the last few decades, as the world has shifted from a production-based economy to a knowledge-based one, corporate value and growth have become increasingly driven by intangible assets (such as patents, human resources, and information technology) and less by tangible assets (such as plants, inventory, and machines). As a result, traditional accounting measures that are subject to the reliability criterion, reflecting mainly physical and financial assets that can be measured in a reliable way, have become less relevant for the users of financial statements. This naturally raises the question of whether it is beneficial to relax the reliability requirement and mandate that firms supplement their financial reports with forwardlooking information about the value of important intangible value drivers. While such information is soft in nature and unlikely to meet the reliability requirement, it is certainly very relevant for the users of financial statements. This paper studies how the softness of information affects the desirability of mandating its disclosure in a model where the existence and the content of information is private knowledge of the manager.
content_copy
Financial theory, along with the practicalities arising from informational frictions, implies that rational agents learning about the dynamics of the risk factors in bond markets will have serially correlated yield forecast errors. Accordingly, in assessing whether professional forecasters hold rational beliefs, it is useful to compare the properties of individual forecasts to those of a benchmark rational forecaster. For this purpose, I use the learning rule of GLS’s “outside observer-econometrician” BE who follows a (constrained) Bayesian rule for learning about the dynamics of the PCs of bond yields.
content_copy
In this study, we investigate a dynamic timing model where many firms must decide the time in which they disclose and sell shares of a firm or a project. Firms who delay their IPO times stand to gain from observing the IPOs of other rms. This informational rent is captured through a common component, such as investor sentiment, which is observed through the pricing of IPOs of other rms. The importance of investor sentiment in IPO pricing has been discussed in media outlets and is consistent with ndings in the empirical literature (e.g., Cornelli et al. (2006)).
content_copy

Stating the main conclusion and linking it to the research question(s)

In an empirical study of Google’s Android platform and the Google Play Award, we report three findings. First, awards increase recipients’ likelihood of releasing complement improvements and decrease their likelihood of releasing new complements. Second, awards increase recipients’ likelihood of multihoming. Finally, awards encourage other complementors to release new complements in recipients’ market niches.
content_copy
As our results show, platform companies often acquire other platform companies earlier, and then as they mature proceed to acquire other non-platform companies such as suppliers or complementors.
content_copy
Additionally, the results show that there is a substantial difference between the acquisition strategies of platform and non-platform companies. This suggests that unlike the literature on acquisitions by non-platform companies, which often centers on large and established companies (Kim et al., 2011; Ahuja & Katila, 2001), acquisition patterns by platform companies are distinctive and an integral part of the entrepreneurial strategy of platform companies.
content_copy
Six main experiments and one supplemental experiment show that perks lower in contractuality foster more WOM than perks higher in contractuality. This result occurs because lowcontractuality perks are more likely to convey a relational signal than perks higher in contractuality. This effect was robust across different operationalizations of contractuality, different perks, different measures of WOM, and different populations. Our studies also show that when ulterior motives for the perk were salient, the beneficial effect of lower (vs. higher) contractuality was attenuated or even reversed.
content_copy
It is shown that mandating disclosure of private information is desirable only when the information is verifiable with high probability.When the information is verifiable with lowprobability, mandating its disclosure leads to an overall decrease in the information content of the managerial disclosure. The intuition behind this result is that while mandating disclosure naturally increases the frequency of disclosure, it also reduces the quality of the disclosure. This is because when disclosure is voluntary, managers with relatively unfavorable information withhold disclosure instead of participating in the right tail disclosure pool. As a result, they are exposed to less risk and the information content of right tail reports is improved. These effects are stronger when the right tail disclosure pool is not likely to be otherwise separated by the auditing process. Consequently, if the information is verifiable with relatively low probability, voluntary disclosure yields higher welfare. If, instead, the information is verifiablewith relatively high probability, mandatory disclosure is weakly preferred. Strict dominance ofmandatory disclosure, though, can be achieved only if there is some friction that makes voluntary disclosure costly.
content_copy
Our study examines whether the adoption of the addback statutes has a negative effect on corporate innovation. We find that the adoption of addback statutes significantly reduces a firm's innovation. Specifically, the number of patents decreases by 4.77 percentage points and the number of patent citations decreases by 5.12 percentage points. Thus, the decline in innovation is economically significant. The magnitude is comparable to the effects of other state tax policies on innovation. For example, both Mukherjee et al. (2017) and Atanassov and Liu (2019) find that a 1-percentage-point increase in state income tax rate on average decreases patenting activities by approximately 5 percentage points. Moreover, the patents that disappear because of addback statutes have economic value and do not seem to be of lower quality than other patents. Furthermore, we find that after a state adopts the addback statutes, a firm with material subsidiaries in that state assigns fewer patents to subsidiaries in states that have zero statutory tax rates or that do not tax intangible income. We do not find a decline in the number of patents that the firm assigns to the other states.
content_copy
The main results of our model show that pioneer rms emerge endogenously, even in the face of information spillovers.
content_copy

Explaining the theoretical contribution(s)

This study offers three theoretical contributions. First, this study contributes to research on platform governance by advancing our understanding of the mechanisms that influence complementor behavior (Parker and VanAlstyne 2005, Parker et al. 2017). In particular, we add that a non-financialmechanismwhich is relatively simple to implement—awards—may influence complementors’ product development strategies. We observe three desirable and undesirable consequences for platform firms, further enriching our understanding of the mechanisms of platform governance and the complex decisions faced by platform firms. More generally, our findings help in understanding the effectiveness of soft governancemechanisms. Second, this study contributes to the growing literature on awards (Kov´acs and Sharkey 2014, Gallus and Frey 2016, Frey and Gallus 2017), particularly the literature that has studied awards in online communities (e.g., Gallus 2016, Chen et al. 2010, Burtch et al. 2021). We add to the understanding of the consequences of awards, especially how a firm’s receipt of an award is linked to its choice of different product strategies. More importantly, our findings provide evidence that awards honoring outstanding product developments can disincentivize new product development and incentivize the release of product improvements and multihoming. Finally, this study adds to the extant body of knowledge studying complementors’ product strategies (Tiwana 2015, Cennamo et al. 2018). We contribute by observing that complementors’ product strategies may be driven by the receipt of an award. More broadly, our findings suggest that complementors’ ability to signal product quality may be linked to a greater focus on updating. In addition, our study integrates three different software product strategies, namely, complement improvements, new complement releases, and multihoming.
content_copy
These results contribute to the literature on platform strategy that has attempted to understand the factors that determine the success of digital platform companies, and the nuanced strategies through which digital platform companies grow their base of users and complementers (Karhu et al. 2018; Cennamo et al. 2018; Boudreau, 2010; Cennamo and Santalo 2013; Parker and Van Alstyne 2005, Wareham et al. 2014; Huang et al. 2017).
content_copy
This work contributes to two streams of research. First, it contributes to the WOM literature (Berger 2011) by identifying both a new psychological process and a practical trigger of WOM. Whereas past research has mostly focused on examining the benefits (e.g., Godes and Mayzlin 2009; Ryu and Feick 2007) and costs (e.g., Jin and Huang 2014; Tuk et al. 2009) of incentivized WOM, the present work provides insight into how companies can use existing perks to promote WOM that is not directly incentivized. This research shows that perks perceived as lower in contractuality carry a relational signal, which increases consumers’ desire to support the brand via WOM. By identifying contractuality as a novel antecedent of WOM, this research provides a response to You, Vadakkepatt, and Joshi’s (2015) call for understanding levers companies can use to facilitate nonincentivized WOM. Second, our research contributes to work on marketing perks (Liu, Lamberton, and Haws 2015; Reczek, Haws, and Summers 2014). Whereas prior literature has mostly focused on how to structure perks to increase sales (e.g., Kivetz 2005) and customer satisfaction (e.g., Jiang, Hoegg, and Dahl 2013), the present research is the first to examine how perks can be effectively framed to promote WOM. Furthermore, this research draws attention to an unexplored psychological mechanism that perks can activate: relational value. Research in social psychology (Reis, Clark, and Holmes 2004) shows that when one appraises another to have high relational value, it triggers a cascade of processes, including compassionate goals toward the partner (Canevello and Crocker 2010) and feelings of gratitude (Algoe, Haidt, and Gable 2008), which work in concert to foster positive interpersonal relationships (Reis, Clark, and Holmes 2004). This work provides the first evidence that marketing perks can signal relational value.
content_copy
Prior research has produced a rich literature on brand name linguistics (e.g., Klink 2000; Lowrey and Shrum 2007; Yorkston and Menon 2004). We extend this literature by introducing a measure of femininity that combines both morphology (name length and stress) and phonology (sounds). Previous research has typically focused only on the effects of single phonemes. One potential benefit of multiple linguistic indicators of femininity is that the effects may be stronger than for single indicators.
content_copy
Prior work has examined brand names and gender, based on name sounds, in the context of matching hypotheses (e.g., feminine names are appealing to women or for feminine products; Grohmann 2009; Yorkston and DeMello 2005). We show that linguistically feminine names, as measured by length, sounds, and stress, are most often an advantage because they convey warmth, which is a generally desirable attribute. Feminine names are preferred by both men and women for products that are used by both genders (i.e., the vast majority of products) and are equally desirable as masculine names for products used by men.
content_copy
The present work also contributes to the research on hedonic and utilitarian product marketing. Previous work has established the affective nature of hedonic goods (Holbrook and Hirschman 1982; Yeung and Wyer 2004). We build on this research stream by showing the benefit of warm product descriptors (e.g., names) for hedonic products.
content_copy
Extending the existing evidence of Brunnermeier and Nagel (2004), our results show that the bubble-riding type of sentiment trading that generates a positive sentiment beta can enhance fund performance beyond the socially useful function of betting against mispricing.
content_copy
Moreover, we find evidence of sentiment timing skill for a subset of hedge funds, with sentiment timers exhibiting both high sentiment beta and large alpha. Thus, although sentiment fluctuations can deter arbitrage activity, some skilled arbitrageurs are able to profit from such fluctuations (e.g., by predicting changes in sentiment).
content_copy
We present a number of novel results concerning IPO timing and volume. We find that the IPO order is determined by rm type: higher-value rms go public earlier and emerge as IPO pioneers, whereas low-value rms tend to delay their IPO times in order to rst observe market conditions or benet from information spillover.
content_copy
We show that sequential IPO clustering emerges after the realization of favorable investor sentiment. However, the results also show how IPOs can exhibit dispersion, where there is substantial delay between pioneer IPOs and followers|a ubiquitous phenomenon which has largely not been captured by previous theoretical studies.
content_copy
The results also provide predictions regarding how IPO volume fluctuates over time and can vary by industry concentration, uncertainty over the state, delay costs, and persistence in sentiment. We find that greater industry concentration (i.e., number of rms) increases the likelihood that a given rm delays its IPO in the rst period, but this is met with a higher likelihood of sequential clustering and higher IPO volume in the second period.
content_copy
Our model can also shed new light on the e ect of accounting quality on the timing patterns of IPOs. In particular, higher accounting quality strengthens the incentive to delay the IPO in order to glean information from the IPOs of other rms, improves the information content of IPOs, and improves rms' welfare.
content_copy

Explaining the practical contribution(s)

content_copy

Interpreting and discussing findings

content_copy

Explaining causality of findings

content_copy

Comparing and contrasting findings with previous studies

content_copy

Generalizing findings

content_copy

Explaining theoretical implications

Our findings offer several theoretical implications. For research on platform governance, the puzzling question that emerges from our work is whether award recipients’ focus on updates provides value to the platform owner. Recipients’ shift toward releasing improvements might have direct positive impacts on the platform owner by promoting improvement without requiring coordination or resources (e.g., Wareham et al. 2014, Huber et al. 2017). The shift toward updates might have even more direct impacts on value if the platform owner can participate the recipient’s success (e.g., via transaction fees or in-app sales) and if this value exceeds that obtained from a more even demand distribution.
content_copy
The finding that awarded complementors are more likely to multihome is novel when considered in the light of recent empirical studies which observed that platform participants are more loyal the higher their past success and activity on a platform (Koh and Fichman 2014). These differences may be related to the type of platform exchange studied, for which our data do not permit further investigation.
content_copy
The observation that more new complements were released in the recipients’ market is novel in the light of the literature on entry deterrence by established firms (e.g., Goolsbee and Syverson 2008) and recent empirical studies on platforms. One may argue that award winners may obtain a competitive advantage that deters entry by rival complementors. One explanation that may reconcile these different arguments is that complementors still expect greater opportunities from entering (e.g., Liang et al. 2019). If so our findings indicate the strength of the signal emitted by an award.
content_copy
Another potential implication for these findings, is that they can inform our understanding of industry evolution. The double-helix model of industry evolution, first developed by Fine (1998) and expanded on by Pagani (2013), considers how industries evolve with respect to their need for companies to balance competing needs for intermediation versus control. These studies have documented how companies have evolved from emphasizing control over their suppliers towards more loosely coupled coalitions, and increasingly towards being platforms. They theorize how the shift from vertically integrated structures towards platforms was motivated by a need for flexibility and the need for product variety (Pagani, 2013). However, they also theorize that these platforms also eventually require greater control of their suppliers, particularly as those suppliers grow in influence. As a result, they theorize that platforms may eventually return to a phase where they require greater control of complementors or suppliers (This evolution is framed in Figure 1). As a result, they predict that while platform companies may benefit from being an intermediary between different actors, they eventually may want to control certain elements of the ecosystem.
content_copy
As the present results suggest, platform companies may want to acquire other platform companies, early on, often from the same market niche. However, as they mature, platform companies may want to control non-platform companies including those from other market niches. This is consistent with the theory from the double-helix model which predicts that platform companies may eventually act less like intermediaries and be more likely to exert control over other companies including potential suppliers. This raises questions for future research that may consider the evolution of platform companies in a future where they want to exert greater control over various partners. This also highlights the need for future work to consider the challenges that platforms face in balancing between “intermediation” and “control”, and rather than acting as pure intermediaries, they often have a need to control other companies in their ecosystem.
content_copy

Explaining practical implications / Managerial implication

Firms can directly benefit from the analyses presented in this study. First, this study has implications for platform firms that are interested in utilizing awards as a governance mechanism on their platforms. This study’s results suggest that awards do have significant effects on complementors’ behavior despite representing only symbolic and non-financial mechanisms. Managers should consider the flexibility of awards as a particular strength of the mechanism. The precise information conveyed through the award is at the discretion of the platform owner. The platform owner has full freedom in signaling ex post whatever outcome is most valuable. An award can, for instance, be given to “the most innovative complement” with no need to define the criteria exactly. Thus, awards reflect an instrument for platform owners to influence the behavior of the recipient and the ecosystem, especially in cases where owners are unable or unwilling to specify desired outcomes. This represents an important capability of digital business strategy (e.g., Park and Mithas 2020).
content_copy
Nevertheless, managers should be aware that awards are complex mechanisms and therefore require a careful evaluation of whether their effects are desirable. On the one hand, awards may be an instrument to incentivize improvement in a particular complement. Moreover, awards boost the development of new complements in the niche market targeted by the award. Therefore, platform firms may use awards as a strategic instrument to guide complementors’ attention toward a certain market niche. On the other hand, awards can encourage multihoming, potentially by providing additional resources through demand increases, but also by increasing competing platform firms’ awareness of promising potential complementors. After all, awards may increase the transparency of complementor performance beyond the boundaries of the platform. Which of these effects prevail is a question that our study cannot address, and one that likely depends on the particular platform. Thus, one recommendation for platform firms is to employ additional mechanisms to increase recipients’ loyalty to the platform, for example, through contractual exclusivity arrangements.
content_copy
Second, this study has managerial implications for complementors. These firms typically “swim with sharks,” facing high competitive pressure both from other complementors and from appropriation or “Sherlocking” by platform owners (e.g., The Economist 2012). The findings of our study suggest that awards may represent both an opportunity and a challenge for the complementors who receive them. On the one hand, awards provide these firms with elevated status, which helps them gain visibility among their consumers, reduce transaction costs, and diversify on competing platforms. This may help complementors recoup their investments into costly complement development and enable improvement strategies.
content_copy
On the other hand, recipients should be careful not to overstate the potential advantages of an award. As our study indicates, awards may attract the entry of other complementors. Eventually, recipients may face increased competition through having more “sharks” in the tank. In addition, other complementors, or the platform owner, may feel encouraged to challenge award winners and engage in competitive attacks. Therefore, multihoming could be caused by a desire to escape niche competition. It may be crucial for award recipients to leverage their signaling advantages on a rival platform to mitigate the increased competitive risks. Award recipients may demand exclusivity fees from the focal platform owner to not multihome, giving them even greater advantages over rivals.
content_copy
The results for this paper have considerable implications for managers and their understanding of how platform companies grow. Academic studies have typically highlighted that platforms grow by attracting users and suppliers. As the present study has documented, acquiring other companies may also play a very important role, especially early in the lifecycle of companies when they need to scale the platform. Additionally, once companies become prominent platforms, they often have a need to manage the relationship between users and suppliers, and therefore need to shift away from simply attempting to scale the platform by acquiring competitors.
content_copy
Our research suggests that marketers can encourage WOM with easily implementable pivots that lower the contractuality of their perks. For example, contractuality can be lowered by reducing the restrictiveness of a perk, reducing the salience of a perk’s contingencies, or framing a perk as a gift from the company rather than a prize earned through a customer’s effort. In each of these instances, companies do not have to change what they are offering—they only need to change the way consumers perceive the perk.
content_copy
Our research also cautions against the idea that low contractuality is always beneficial. Our results suggest that sometimes consumers may be suspicious of perks low in contractuality, with potentially negative consequences on WOM. This seems particularly likely to occur when low-contractuality perks come from disliked or distrusted brands, or when they are awarded before rather than after purchase. For example, many consumers do not like utility providers or financial institutions (e.g., Accenture 2013). To the extent that dislike prompts consumers to make hostile attributions toward benevolent gestures, such companies might be better off using perks that are higher in contractuality or using low-contractuality perks in conjunction with other interventions, such as positioning themselves as underdogs (e.g., Paharia et al. 2010). Moreover, to the extent that awarding a low-contractuality perk before purchase may make ulterior motives salient, companies may be better off timing them so that they are awarded after the purchase. Finally, our research suggests that although highcontractuality perks can be effective at motivating consumers to engage in a desired behavior, they may also carry an opportunity cost. Indeed, the field study conducted at the university bookstore showed that although shoppers were more likely to comply to a request (i.e., filling out a survey) when given a high- (vs. low-) contractuality perk, they were less likely to share WOM about the company. These results suggest that when designing perks, marketers need to be cognizant of the potential trade-off between incentivizing first-order responses and nudging second-order responses such as WOM.
content_copy
The present findings have important substantive implications and should inform marketing strategy. Feminine brand names appear to have a marketplace advantage, particularly for hedonic products. Moreover, such linguistic elements have an automatic influence (Pogacar, Shrum, and Lowrey 2018), providing a source of affinity that consumers respond to instinctively. Brand managers should take careful note of the linguistic characteristics of new brand names and brand extensions and leverage the feminine brand name advantage as appropriate.
content_copy
Managers should be aware of two factors that might influence the brand outcomes observed here: differences in naming conventions between languages and differences in gender stereotypes between cultures. If women’s names are equally as likely as men’s names to end in vowels in a given language, for instance, then the Barry and Harper (1995) name gender scale should not guide naming strategy for target markets speaking this language. Rather, marketers should determine the linguistic cues that signal gender in the target market language and test brand names with those characteristics.
content_copy
Managers should employ target market surveys of naming conventions and gender perceptions to address this. Moreover, in instances when brand names are not fixed across languages but are instead translated, brand managers should pay close attention to the effects of translation (Zhang and Schmitt 2001) and to local naming preferences (Wu et al. 2019).
content_copy
A final empirical implication of the analysis in the paper is that when measuring the value of mandating disclosure, it is important to take into account the information that managers disclose voluntarily. It may be problematic to make regulatory prescriptions based on the information content of the financial reports alone. For example, the deterioration in the informativeness of financial reports in the last few decades due to the absence of information about intangibles does not necessary imply that firms should be required to disclose such information in their financial reports. Instead, as this paper shows, it may be the case that the voluntary disclosure of such information is more effective. Therefore, to draw conclusions about the desirability of disclosure regulation, it is necessary to examine the effect of disclosure regulation on the information content of the total information that managers provide e both mandatorily and voluntarily.
content_copy

Recommending future research

Future research is needed in order to better understand the various ways that companies make acquisitions, and perhaps a greater understanding of the reasons behind platform acquisitions.
content_copy
In this paper, we provide evidence for when and who platform companies acquire. However, this raises new questions for subsequent research. First, questions remain regarding the reasons behind acquisitions made by platform companies, and in particular the specific types of assets that platform companies need to acquire. Second, questions remain regarding the extent to which different types of acquisitions by platform companies impact performance, and allow companies to attract users through network effects. Third, the present paper signals how platform companies have evolved from pure intermediaries, to exercise greater control over suppliers and other members of the ecosystem, as predicted by the double helix framework (Pagani 2013). Future research can build further on this to understand how digital platform companies manage the balance between intermediation and control.
content_copy
Nonetheless, future research could examine whether conditions exist in which contractuality might affect the nature of consumers’ motivation.
content_copy
Future research could examine other circumstances under which perks higher versus lower in contractuality increase WOM. For example, prior research has argued that perks might strengthen the relationship between a consumer and a company if the acquisition of the perk is effortful (Kivetz and Simonson 2003). A perk that has low value yet requires considerable effort to be earned could prompt consumers to conclude that they really care about the company. In this research, we intentionally held effort constant, but future research could allow effort to vary. High-contractuality perks might be more effective at increasing commitment than low-contractuality perks to the extent they require more effort on consumers’ part and/or when the effort-to-benefit ratio is extremely salient (Dodson, Tybout, and Sternthal 1978). Similarly, additional research could examine whether our conceptual framework applies to other marketing tactics. For example, do less-contractual return policies and products (e.g., devices that are more compatible) convey a relational signal that spurs WOM?
content_copy
Finally, future research could examine whether marketing offerings that tap into distinct WOM motives have a differential effect on WOM persuasiveness. In the present research, participants appeared motivated to generate WOM because they wanted to help the firm after the firm gave them a marketing offering that was lower in contractuality. However, marketing offerings could also prompt consumers to generate WOM by way of triggering the desire to self-enhance (De Angelis et al. 2012). For example, preferential treatment that signals high status could boost a customer’s desire to selfenhance in front of others. Future research could explore how different motivations for sharing WOM ultimately shape how persuasive the WOM is to others.
content_copy
Future research should also investigate how established brands with masculine names can incorporate these findings. Brands with equity in their existing names may be ill-advised to discard a well-established masculine name, even if the brand markets hedonic products and typical users include women. However, it may be possible to imbue masculine brand names with warmth via feminine subbrands, brand extensions, or logos. For instance, Fiesta is a linguistically feminine subbrand of Ford that could add warmth to the masculine corporate brand.
content_copy
Future work might also examine whether linguistically masculine names are advantageous when warmth is not a desirable product attribute. Although warmth is generally desirable, or at least not harmful, there may be rare exceptions. For instance, some products are designed to destroy pests, weeds, molds, warts, or even people. In such instances the feminine brand name advantage may be reversed. Indeed, our preliminary analyses suggest that people evaluate guns with masculine names more favorably than guns with feminine names (Web Appendix H) and that successful firearms have more masculine names than successful products from a more neutral category (i.e., backpacking packs; Web Appendix I).
content_copy
Finally, I have left two central issues virtually unaddressed. First, what explains the breadth of the reactions of professionals to economic “news” that is relevant for pricing bonds? Do plausible differences in priors or attitudes toward ambiguity/uncertainty induce the diverse response coefficients documented here? Or might incentives beyond the production of efficient forecasts in the classical econometric sense partially underlie this heterogeneity? Second, this analysis looks at survey data through the lens of an outside observer. What are the properties of equilibria in bond markets that arise when market participants follow diverse Bayesian learning rules or show differing degrees of ambiguity aversion in formulating their heterogeneous forecasts and investment strategies? My descriptive evidence challenges the plausibility of some recent framings of informational frictions underlying equilibrium dispersion of beliefs. Hopefully it also points to fruitful paths for enriching extant heterogeneous-agent bond pricing models.
content_copy
It would be fruitful to further advance our understanding of the interaction between sentiment traders and arbitrageurs (e.g., hedge funds’ trading strategies based on mispricing and how arbitrageurs respond to sentiment fluctuations) when finer information on hedge fund positions becomes available. The pricing of market-wide sentiment risk in financial markets also deserves closer scrutiny. We leave these topics for future research.
content_copy
Several extensions can be considered for future work. We have assumed that the rm's type (idiosyncratic component) is constant over time. A potentially interesting research question is to investigate a model where the firm's value is also evolving. Relatedly, our model can be extended to a continuous-time setting with a nite number of rms, where the market condition follows a Brownian diu-sion process. We conjecture that, in the continuous-time analog, there exists a symmetric equilibrium in which the IPO timing strategy is decreasing in rm-type and further delay is decreasing in realizations of the state. The continuous-time analog thus seems to share the main characteristics of our discrete-time model.
content_copy

Listing the limitations and closing the discussion section

We cannot fully rule out confounding effects or omitted variables. Although the analyses did not support pre-existing differences, winners and runners-up may differ in characteristics that we were unable to observe. Also, Google may have selected award winners because they expected exactly those firms to focus on complement improvements. Our research may not account for this possibility. We also acknowledge that our research design is constrained by its sample size. Our power analysis provides a starting point for the design of future studies of awards.
content_copy
One limitation of the present research is its focus on English-speaking participants. The Interbrand names we examined are global, and prior research suggests that many linguistic effects are robust across languages (Blasi et al. 2016; Shrum et al. 2012). However, differences between languages have also been observed (Coulter and Coulter 2010; Kuehnl and Mantau 2013), so it is an empirical question whether a given linguistic effect manifests in a given language. Indeed, the results of Study 3b, the only study with a sizable number of nonnative English speakers, suggest there may be language differences that merit further investigation (note B in Web Appendix D).
content_copy

Opening the conclusion section

We investigated the effects of platform firms’ choices when conferring retrospective awards on complementors.
content_copy
To crack down on corporate state tax avoidance via intangibles-based income shifting, many state governments adopted addback statutes. In this study, we examine whether the adoption of such addback statutes by U.S. state governments impedes corporate innovation. Specifically, the addback statutes require firms within the adopting state to add back to their state taxable income intangible-related expenses paid to related parties in other states. These provisions prevent firms from using intangible assets to avoid taxes and consequently reduce the benefits that firms and managers can gain from creating intangible assets such as patents. In other words, the projected net present value (NPV) of patents and innovation projects will decrease.
content_copy
In this paper, we explore how hedge fund exposure to sentiment fluctuations (i.e., sentiment beta) is related to fund performance.
content_copy
In this study, we examine whether employee perceptions of managers and firms decrease following media coverage of tax avoidance. We posit that employees perceive tax avoidance news negatively because they do not clearly benefit from decreased tax payments and they perceive it as unfair and/or socially irresponsible. We use employee ratings data from Glassdoor.com to measure changes in employee perceptions of managers and firms. We use news coverage of firms’ tax avoidance as our treatment events.
content_copy

Summarizing the main conclusion

In an empirical study of Google’s Android platform, we found both desirable and undesirable outcomes for platform firms. The award increased recipients’ release of complement improvements as well as the overall number of apps released in the recipients’ market niche. However, the award also increased recipients’ likelihood of multihoming.
content_copy
Overall, the results suggest that restricting mandatory disclosure requirements to information that is relatively verifiable while allowing managers to voluntarily disclose estimates, projections and other forward-looking information may maximize the total information content of managerial disclosures.
content_copy
Overall, we find that the adoption of addback statutes has a significant negative effect on corporate innovation.
content_copy
We show robust evidence that hedge funds with large positive exposures to changes in investor sentiment significantly outperform other funds. The return spread between the top and bottom deciles of hedge funds ranked by sentiment beta is as large as 0.59% (t-statistic = 3.55) per month on a risk-adjusted basis.
content_copy
We find that employee ratings of both S&P 500 senior managers and firms decrease following tax avoidance news using generalized difference-in-differences. In addition, we find results consistent with senior manager and firm ratings decreasing in media coverage intensity. We provide evidence that perceptions of managers and firms fall more when the firm is in a consumer-facing industry compared to when it is not. Next, we find evidence that employees at high-performing firms react less negatively to tax avoidance news compared to other firms. We perform numerous additional tests, placebo tests, and falsification tests. However, as with any quasi-experimental study, we cannot rule out all possible confounds.
content_copy

Summarizing the main contribution(s)

Our findings have important implications for policy makers who are interested in understanding the consequences of policies that constrain tax-motivated income shifting using intangibles and prevent income base erosion. For example, the recent Tax Cut and Jobs Act of 2017 also includes anti-base-erosion provisions similar to the addback statutes, which aim to crack down on tax-motivated income shifting by U.S. multinational firms to foreign countries with low taxes.
content_copy
Our study contributes to the literature on taxes and reputation. We extend prior studies by examining whether tax avoidance news decreases employee perceptions of their managers and firms, rather than focusing on shareholders or other stakeholders, as prior studies do. We also contribute to the literature identifying the effects of tax avoidance and tax incentives on labor by providing evidence that employees perceive tax planning negatively and rate firms and managers lower following tax news.
content_copy

Recommending future research and closing the thesis

content_copy