Review of Accounting Studies

, Volume 19, Issue 4, pp 1309–1345

Product market competition and conditional conservatism

Authors

  • Dan Dhaliwal
    • Eller College of ManagementUniversity of Arizona
  • Shawn Huang
    • Arizona State University
    • Robert J. Trulaske, Sr. College of BusinessUniversity of Missouri-Columbia
  • Raynolde Pereira
    • Robert J. Trulaske, Sr. College of BusinessUniversity of Missouri-Columbia
Article

DOI: 10.1007/s11142-013-9267-2

Cite this article as:
Dhaliwal, D., Huang, S., Khurana, I.K. et al. Rev Account Stud (2014) 19: 1309. doi:10.1007/s11142-013-9267-2

Abstract

Based on pooled cross-sectional analysis, we find a robust positive relation between product market competition and conditional accounting conservatism. We also find evidence of an inter-temporal increase in conditional conservatism following industry deregulation and increased antitrust case filings. Distinguishing further between two dimensions of competition, we find conditional conservatism is greater when there is a higher threat of new entrants as well as stiff existing competition. Moreover, we find these results largely hold for industry followers as opposed to industry leaders, suggesting that strategic considerations shape the extent to which product market competition affects conditional conservatism.

Keywords

Market structureFirm strategyMonopoly rentsPolitical costs

JEL Classification

D4L1M41

1 Introduction

This paper examines the impact of product market competition on conditional accounting conservatism (i.e., asymmetric timely loss recognition or the more timely recognition of economic losses than gains in earnings). Several views motivate this inquiry. One view suggests that the relation between product market competition and conditional conservatism is shaped by strategic considerations. The argument here is that when choosing its reporting, a firm considers the impact on its competitive position. For example, firms in industries with low entry costs will find timely loss recognition beneficial, to the extent that this discourages new firms from entering the product market. In a similar vein, firms that face considerable competition from existing rival firms may recognize losses more quickly to discourage those firms’ over-production or encourage their under-production (Clinch and Verrecchia 1997; Li 2010). The implication is that a firm will likely adopt timely loss recognition to improve its competitive position vis-à-vis potential entrants and existing rivals.

An alternative view suggests that competition can attenuate manager-shareholder conflicts of interests. According to this view, managerial performance can be better evaluated in the presence of competing firms that serve as a benchmark. Intense competition also reduces managerial slack, forcing a manager to improve firm efficiency. The implication of this view is that product market competition reduces managerial slack and improves monitoring, and consequently strengthens the alignment of the interests of managers and shareholders. As a result, there will be less demand for conditional conservatism as a monitoring mechanism. A shortcoming of this view is that it narrowly emphasizes the benefit of competition in lowering agency conflicts while ignoring the potential costs involved, such as costly violation of contracts in place because of low firm slack. Also, this view ignores the potential benefits that shareholders can reap when a firm holds a monopolistic position.

There are additional arguments that speak to the relation between competition and conditional conservatism. For instance, monopolists face greater political costs than firms in competitive industries (Watts and Zimmerman 1986). Hence there is greater incentive for dominant firms in less competitive industries to undertake timelier loss recognition to avoid regulatory scrutiny. An empirical relation between competition and conditional conservatism can also arise from the fact that monopoly rents create an errors-in-variables problem in tests of conditional conservatism (Ball and Shivakumar 2006; Roychowdhury and Watts 2007). Specifically, monopoly rents, while embedded in stock returns, may not be reflected in a firm’s earnings, and this can affect the relation evaluated in this study. We consider these issues in our empirical analysis.

To analyze the relation between product market competition and conditional conservatism, we follow the literature in the construction of these measures. Our primary measure of conditional accounting conservatism is based on Basu’s (1997) model, which measures the asymmetric timeliness in the recognition (incorporation) of economic losses in accounting income (i.e., quicker recognition of economic losses than economic gains in current-period reported earnings). Following Hoberg and Phillips (2010), we use a fitted Herfindahl–Hirschman index (HHI) as a proxy for product market competition in three-digit Standard Industry Classification (SIC) code industries. We also use the HHI computed from the Compustat data as an alternative proxy for competition. Moreover, we follow Li (2010) and employ two common factors extracted from several industry-level competition variables to separately measure competition from potential entrants and existing rivals.

We find a positive association between product market competition and conditional accounting conservatism. This result holds even after controlling for several firm characteristics that are known to explain the variation in accounting conservatism across firms. To further evaluate alternative causal explanations, we classify firms as being either industry leaders or followers. Industry leaders have greater market share and hence greater market power (Nickell et al. 1992; Nickell 1996); therefore the reporting incentives induced through product market competition are likely weak or non-existent for them. Contrarily, these firms have higher incentives for reporting conservatively under the political cost argument. Moreover, the governance argument suggests that industry followers should be less conservative in more competitive industries because competition improves monitoring, thereby constraining managerial opportunism. Empirically, we find that the positive relation between product market competition and timely loss recognition holds for industry followers, but not for industry leaders, even after controlling for monopoly rents, suggesting that strategic considerations rather than political costs or improved monitoring of product market competition influence conditional conservatism.

We also conduct two time-series analyses to further evaluate the relation between product market competition and conditional accounting conservatism. First, we consider firms in the airline, telecommunications, and transportation industries, which experienced significant deregulation in the late 1970s and early 1980s. Using a difference-in-differences research design, we find that an increase in competition in these industries led to greater conditional accounting conservatism. Second, we exploit the variation in anti-trust enforcement across time and find conditional conservatism to be greater in periods of increased anti-trust enforcement actions.1 Collectively, our evidence is most consistent with the strategic consideration view.

Our paper is closely related to two recent papers that examine how the multi-dimensional nature of product market competition influences accounting and disclosure practices. Hui et al. (2012) study the role of a firm’s major customers and suppliers in influencing the demand for conditional conservatism. We differ in that we focus on the competitiveness of a firm’s own industry. Our approach in measuring product market competition is similar to that of Li (2010), who finds that firms manage their voluntary disclosure policy to discourage the entry of new firms and to encourage the under-production of rival firms. We find evidence of a similar behavior in the context of conditional conservatism. Taken together, the evidence suggests that disclosure and recognition are complements and that managers are prudent in the way that they provide information to outsiders.

The remainder of the paper is organized as follows. Section 2 describes the related literature and develops our testable hypothesis. Section 3 discusses the data and details the empirical methods. Section 4 presents the results, and Sect. 5 summarizes and concludes the paper.

2 Hypothesis development

Early research on product market competition emphasized its impact on prices and economic efficiency (Chamberlain 1933; Robinson 1933; Fellner 1949; Alchian 1950; Stigler 1958). More recent research has highlighted the influence of product market competition on corporate decisions, such as the choice of capital structure (Harris and Raviv 1991; Chevalier 1995; Xu 2012; Valta 2012). In this study, we examine the relation between product market competition and firm choice regarding accounting conservatism. We explore several channels through which product market competition can affect conditional conservatism.

One line of inquiry suggests that product market competition creates strategic incentives. There are two distinct schools of thought in this area. Porter (1980) outlines a market-centered approach under which firms formulate strategies by considering competitive forces that influence the industry in which they compete. Examples of these forces include the threat of new entrants and threat of substitutes. In contrast, the resource-based perspective notes that firms are heterogeneous in that they possess different resources/competencies (e.g., Wernerfelt 1984). The resource-based view starts with an inward look at a firm’s resources and then focuses on how these resources can be combined to obtain a competitive advantage. Accounting is relevant in both these perspectives in that it represents an intangible resource of the firm that has strategic value.

To understand the strategic value of conditional conservatism, consider its use as a barrier to entry. Given that the entry of new firms weakens its competitive position, a firm will recognize bad news quickly to discourage new entrants to an industry (Darrough and Stoughton 1990). Since the investments of rivals are sunk, one of the goals of firms is to induce reduction in its competitors’ production output (Clinch and Verrecchia 1997). To the extent that timely revelation of bad news induces uninformed rivals to under-produce, a firm will disclose and recognize bad news more quickly. In other words, bad news will be interpreted by uninformed competitors as an indicator of lower future demand. Under-production by rivals will benefit the firm if it can meet the production shortfall relative to the demand involved.2

Thus strategic considerations can shape the relation between product market competition and conditional conservatism. As product market competition increases, firms become asymmetrically timelier in recognizing losses into earnings to improve its competitive position vis-à-vis rivals and potential entrants. Moreover, the strategic consideration view predicts that the industry followers in more competitive industries are likely to exhibit more conditional conservatism because these firms are more susceptible to competitive pressures than industry leaders (Li 2010; Nickell 1996).

Another line of inquiry depicts product market competition as an informal governance mechanism that constrains managers from undertaking actions that are contrary to shareholder interests. This literature identifies two causal channels. First, intense competition can reduce managerial slack and force a manager to improve firm efficiency (Alchian 1950; Stigler 1958). Consistent with this argument, Nickell (1996) finds that competition, as measured by increased numbers of competitors or by lower levels of monopoly rents, is associated with a significantly higher rate of total factor productivity growth. Second, intense competition can facilitate better monitoring of managers (Holmstrom 1982; Nalebuff and Stigliz 1983; Hart 1983). In the presence of competition, the profits of other firms can be observed and used as a benchmark for evaluating managers. As such, managers are induced to increase firm profits (Fee and Hadlock 2004; Giroud and Mueller 2010). Consistent with this monitoring argument, Guadalupe and Perez-Gonzales (2006) find intense product market competition curtails the extent of private control benefits extracted by corporate insiders. Separately, Giroud and Mueller (2010) find that, while state-level adoption of anti-takeover provisions constrained the workings of the market for corporate control, operating performance declined only for firms in non-competitive industries and not for those in competitive industries.3,4 Taken together, these arguments suggest that intense product market competition will limit managers from taking actions that contradict shareholder interests (e.g., Hart 1983). An implication of this argument is that product market competition plays a governance role, and hence there is less need for conditional conservatism to constrain managers. This substitution argument predicts, on average, a negative association between competition and conditional conservatism. Moreover, this negative effect is likely to be concentrated among industry followers because they face more competitive pressure than industry leaders, and this will serve to constrain managerial opportunism.

However, this predicted relation may not empirically hold for three reasons. First, conditional conservatism and competition play distinct and complementary roles in constraining managers (Francis and Martin 2010; Hou and Robinson 2006). Second, Schmidt (1997) questions this governance view by noting that low profit margins that result from intense competition can discourage managerial effort. The implication is that intense competition can aggravate manager-shareholder conflicts of interest. Third, the governance view highlights the agency cost savings while ignoring benefits such as monopoly rents that can also accrue to shareholders. As such, it is not clear whether competition will have an empirically identifiable negative impact on conditional conservatism.

Besides these two views, there is also a political costs argument, which suggests monopolistic firms will undertake timelier recognition of losses in earnings to avoid regulatory scrutiny and sanctions. Garcia Lara et al. (2009) find an increase in conditional conservatism following regulatory changes. This political cost argument suggests a negative association between product market competition and conditional conservatism, i.e., the likelihood of conditional conservatism is expected to be higher in less competitive industries. Moreover, the political costs view predicts that the industry leaders in less competitive industries are likely to exhibit higher conditional conservatism because they are more likely to face regulatory scrutiny.

An empirical relation between competition and conditional conservatism can also arise from the fact that monopoly rents, while embedded in stock returns, may not be reflected in a firm’s earnings. Thus, monopoly rents create an errors-in-variables problem in tests of conditional conservatism (Ball and Shivakumar 2006; Roychowdhury and Watts 2007). Because product market competition reduces the error-in-variable problem arising from monopoly rents, the monopoly rent view predicts a positive relation between competition and conditional conservatism. While similar in prediction to that under the strategic consideration view, the monopoly rent view differs in that it predicts conditional conservatism for industry leaders to be low in less competitive industries. The argument here is that industry leaders are more able to reap greater monopoly rents when competition is lacking and that this creates greater errors-in-variable problem for the subset of firms in less competitive industries.

In summary, four views provide conflicting predictions on the direction of the relation between product market competition and conditional conservatism. Specifically, the strategic consideration and the monopoly rent-related views predict product market competition to have a positive impact, while the governance and the political cost views predict a negative relation. Further evaluation by distinguishing industry leaders from industry followers allows us to discriminate between these alternative causal explanations. Table 1 summarizes the testable predictions using a 2 × 2 matrix for industry competition (low vs. high) and firm position (industry leader vs. industry follower). Specifically, it depicts the direction and strength of the competition-conservatism relation across industry leaders and followers based on the four views noted above.
Table 1

Predictions for conditional conservatism (CC) under alternative explanations

 

Industry position of firm

Leader

Follower

Level of industry competition

Low

Political costs view: High CC

 

Monopoly rents view: Low CC

High

 

Strategic view: High CC

Governance view: Low CC

3 Sample selection and research design

3.1 Sample selection

Our sample includes firms with NYSE-, AMEX-, or NASDAQ-listed securities with share codes 10 or 11. Our sample starts with the intersection of CRSP monthly returns file and the Compustat industrial annual file. We merged this data set for each three-digit SIC code industry and year with concentration ratios obtained from the Hoberg–Phillips data library.5 We use the 1975–2005 period because the product market competition measures that are constructed using public and private firms in an industry are available only for this period. In addition, our sample is limited to firms with financial information data necessary for our analysis. We excluded firms in the financial (SIC 6000-6999) and utility (SIC 4900-4999) industries because these industries were regulated throughout our sample period. Our final sample comprises 71,627 firm-year observations covering 187 three-digit SIC industry groups over the 31 years from 1975 to 2005. Finally, we winsorized all continuous variables at 1 and 99 % of their empirical distribution.

Porter (1980) distinguishes between two dimensions of product market competition: competition from potential entrants and competition from existing rivals. To empirically evaluate the impact of these dimensions of product market competition on conditional conservatism, we follow Li (2010) and extract data from Compustat segments and annual databases. The period involved covers the years 1977 through 2005.6 We include only business segments with primary four-digit SIC codes and aggregate all financial items for segments with identical SIC codes belonging to the same firm. If a Compustat firm does not have segment data, we treat this firm as a single-segment firm. If a firm has multiple segments with different SIC codes, we identify the primary segment as the segment with the same SIC code as in the Compustat annual file. If none of the segments have the same SIC code as in the Compustat annual file, we define the primary segment as the segment with the largest sales.7 We also separately constructed seven industry-level competition variables as employed in Li (2010). The variables and their construction are described in Appendix 2. The sample involving these variables consists of 78,799 firm-year observations, covering 241 three-digit SIC industry groups over 29 years spanning 1977–2005. While this sample also excludes financial firms and utilities, it is still larger than the sample with available concentration data because the Hoberg–Phillips data library contains concentration data for fewer three-digit industry groups.

3.2 Measurement of conditional conservatism

To measure conditional conservatism, we begin with Basu’s (1997) model, which is:
$$ {\text{NI}}_{\text{t}} = \upbeta_{0} + \upbeta_{1} {\text{RET}}_{\text{t}} + \upbeta_{2} {\text{NEG}}_{\text{t}} + \upbeta_{3} {\text{RET}}_{\text{t}} * {\text{NEG}}_{\text{t}} + \varepsilon $$
(1)

In Model (1), the dependent variable NI is reported earnings (income after extraordinary items) deflated by the beginning of the fiscal-year market value of equity. Variable RET represents the buy-and-hold return over the fiscal year and is used to measure publicly available news about the firm’s current-year performance and includes information known to the market prior to the annual earnings announcement (Ball et al. 2000, Basu 1997). The explanatory variable NEG is an indicator variable equal to 1 if RET is negative and 0, otherwise. Negative returns (i.e., RET < 0) are used as a proxy for bad news, while positive returns (i.e., RET > = 0) represent good news. Asymmetric timeliness of bad news recognition imply greater sensitivity of earnings to bad news, which indicates that the slope coefficient on stock returns (variable RET) is expected to be higher for negative returns (“bad news”) than for positive returns (“good news”). The difference in the slope coefficients for variable RET, i.e., the difference in the sensitivity of earnings to negative and positive returns, is captured by the slope coefficient β3 of the interaction variable RET*NEG.

We extend Basu’s (1997) model to empirically examine the relation between product market competition and conditional conservatism. The empirical Model (2) incorporates our test variable along with other firm characteristics found in prior research to influence the extent of a firm’s conditional conservatism:
$$ \begin{aligned} {\text{NI}}_{\text{t}} &= \upbeta_{0} + \upbeta_{1} {\text{PMC}}_{\text{t - 1}} +\upbeta_{2} {\text{SIZE}}_{\text{t - 1}} + \upbeta_{3} {\text{BM}}_{\text{t - 1}} + \upbeta_{4} {\text{LEV}}_{\text{t - 1}} + \upbeta_{5} {\text{LIT}}_{\text{t - 1}} + \upbeta_{6} {\text{BIG}}\_{\text{N}}_{\text{t}} + \upbeta_{7} {\text{NEG}}_{\text{t}} \\ &\quad + \upbeta_{8} {\text{NEG}}_{\text{t}} * {\text{PMC}}_{\text{t - 1}} + \upbeta_{9} {\text{NEG}}_{\text{t}} * {\text{SIZE}}_{\text{t - 1}} + \upbeta_{10} {\text{NEG}}_{\text{t}} *{\text{BM}}_{\text{t - 1}} + \upbeta_{11} {\text{NEG}}_{\text{t}} * {\text{LEV}}_{\text{t - 1}} \\ &\quad + \upbeta_{12} {\text{NEG}}_{\text{t}} * {\text{LIT}}_{\text{t - 1}} + \upbeta_{13} {\text{NEG}}_{\text{t}} * {\text{BIG}}\_{\text{N}}_{\text{t}} + \upbeta_{14} {\text{RET}}_{\text{t}} + \upbeta_{15} {\text{RET}}_{\text{t}} *{\text{PMC}}_{\text{t - 1}} + \upbeta_{16} {\text{RET}}_{t} * {\text{SIZE}}_{\text{t - 1}} \\ &\quad + \upbeta_{17} {\text{RET}}_{t} * {\text{BM}}_{\text{t - 1}} + \upbeta_{18} {\text{RET}}_{t} * {\text{LEV}}_{\text{t - 1}} + \upbeta_{19} {\text{RET}}_{\text{t}} * {\text{LIT}}_{\text{t - 1}} + \upbeta_{20} {\text{RET}}_{\text{t}} * {\text{BIG}}\_{\text{N}}_{\text{t}} \\ &\quad + \upbeta_{21} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} + \upbeta_{22} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} * {\text{PMC}}_{\text{t - 1}} + \upbeta_{23} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} * {\text{SIZE}}_{\text{t - 1}} \\ &\quad + \upbeta_{24} {\text{NEG}}_{\text{t}} *{\text{RET}} * {\text{BM}}_{t - 1} + \upbeta_{25} {\text{NEG}}_{\text{t}} *{\text{RET}} * {\text{LEV}}_{\text{t - 1}} + \upbeta_{26} {\text{NEG}}_{\text{t}} *{\text{RET}}*{\text{LIT}}_{\text{t - 1}} \\ &\quad + \upbeta_{27} {\text{NEG}}_{\text{t}} *{\text{RET}} * {\text{BIG}}\_{\text{N}}_{\text{t}} + {\text{Firm}} - {\text{fixed effects}} + \upepsilon \\ \end{aligned} $$
(2)

The additional variables specified in Model (2) are proxies for conservatism demand (e.g., Khan and Watts 2009) and defined in Appendix 1. We follow Ball et al.’s (2012) suggestion and include firm-fixed effects in estimating Model (2).8 The significance tests for this estimation are based on robust t-statistics that are adjusted for residual correlation arising from pooling cross-sectional observations across time—i.e., the t-statistics are based on White’s (1980) heteroskedasticity-adjusted robust variance estimates that are further corrected for firm- and fiscal-year clustering (Petersen 2009; Thompson 2011).

In Model (2), the coefficient β14 reflects the timeliness of earnings with respect to good news, while the coefficient β21 measures the incremental timeliness of earnings in recognizing bad news. We do not predict a sign on the coefficient for PMC because of competing arguments. On the one hand, greater competition can reduce profit margin, which negatively affects earnings. On the other hand, if PMC improves governance, it should lead to higher profitability. For completeness, we pay attention to the coefficients β15 on RET*PMC and β22 on RET*NEG*PMC. The coefficient β15 reflects the impact of product market competition on how quickly earnings report good news. Similarly, the coefficient β22 measures the impact of product market competition on the incremental timeliness of earnings with respect to bad news. If product market competition induces firms to increase accounting conservatism, then we expect firms in competitive industries to impose stricter verification for gain recognition and induce asymmetrically timelier recognition of losses. Empirically, stricter verification of good news anticipates β15 to be negative, and asymmetrically timelier loss recognition predicts β22 to be positive. Alternatively, if product market competition reduces accounting conservatism, then β15 and β22 are expected to be positive and negative, respectively.

3.3 Test variable

The most commonly used measure of the degree of product market competition in the industrial organization literature is the Herfindahl–Hirschman index (HHI), which is defined as follows:
$$ {\text{H}}_{\text{j}} = \sum\limits_{i = 1}^{I} {s_{ij}^{2} } $$
where sij is the market share of firm i in industry j. For each firm, we compute its market share based on its net sales relative to the total net sales of the industry to which it belongs. The firm market shares are then squared and summed for each industry. The above calculations are carried out for each year and for each industry based on a three-digit SIC classification. Lower HHI values imply that the market is shared among many competing firms, while higher values imply that the market share is concentrated in the hands of a few large firms. For ease of exposition, we create a new variable PMC by multiplying Hj by negative one.
$$ {\text{PMC}}_{\text{j}} = \, \left( { - 1} \right)*{\text{H}}_{\text{j}} $$

Higher values of PMC reflect more intense product market competition, with each firm having a small product market share.

We compute our first measure of product market competition, denoted as PMC1(Sale), based on public firms covered in the Compustat annual file. A criticism of this measure is that it does not consider private firms in a given industry (Ali et al. 2009). In fact, industry concentration measures using U.S. Census Bureau data, which covers both private and public firms, are weakly correlated (13 %) to the industry concentration measures based on Compustat firms (Ali et al. 2009). In light of this concern, we follow Hoberg and Phillips (2010), who create a fitted Herfindahl–Hirschman Index (HHI) that considers both public and private firms. This variable is denoted as PMC2(Sale). The Hoberg and Phillips (2010) measure covers all three-digit SIC industries with the exception of firms from the financial and utilities industries. To construct their product market competition measure, Hoberg and Phillips (2010) combine the Compustat data with the Herfindahl–Hirschman Index data from the U.S. Commerce Department and the employee data from the U.S. Bureau of Labor Statistics to create the fitted HHI for all industries.9 They find this product market competition measure (i.e., fitted HHI) to be highly correlated with the actual HHI from the Commerce Department on manufacturing industries (correlation = 0.54). They also find it to be a vast improvement over the HHI measure based on the Compustat data.10

In addition to product market competition measures based on industry concentration, we follow Li (2010) to separately measure competition from potential entrants and competition from existing rivals. To this end, we conduct a principal component analysis on seven commonly employed proxies for competition listed in Appendix 2. Like Li (2010), we retain two components after using the orthogonal rotation method and requiring eigenvalues to be greater than 1.

Appendix 2 shows the rotated factor pattern and the standardized scoring coefficients of the variables categorized into two groups. Consistent with Li (2010), competition from existing rivals, denoted as EXIST-COMP, is the principal component extracted from the Herfindahl–Hirschman Index (IND-HHI), four-firm concentration ratio (IND-CON4), total number of firms operating in an industry (IND-NUM), and product market size (IND-MKTS). Moreover, competition from potential entrants, denoted as POTEN-COMP, is the principal component computed based on product market size (IND-MKTS), industry-average size of plant and equipment (IND-PPE), industry-average R&D expenditures (IND-R&D), and capital expenditures (IND-CPX). The signs of the standardized scoring coefficients for the seven variables are the same as those reported in Li (2010). We code the two dimensions of product market competition such that larger values of EXIST-COMP indicate greater competition from existing rivals, while larger values of POTEN-COMP indicate greater competition from potential entrants.

3.4 Control variables

In Model (2), we also control for several firm characteristics used in prior research to account for the demand for timely loss recognition. Each control variable is specified in Model (2) as a main effect, as two-way interactions with RET and NEG, and a three-way interaction with RET and NEG. The inclusion of control variables allows us to better assess the incremental effect of product market competition on conditional conservatism.

Prior studies show large firms tend to be less conditionally conservative than small firms (Basu 2001; Giner and Rees 2001; Givoly et al. 2007; LaFond and Watts 2008). We include SIZE to control for firm size. We also include BM to control for past conditional and unconditional conservatism because prior research (e.g., Basu 2001; Giner and Rees 2001; Pope and Walker 2003; Beaver and Ryan 2005; Pae et al. 2005) points out that Basu’s (1997) measure of conservatism is likely to be affected by past conditional and unconditional conservatism. To control for the impact of debt contracting on timely loss recognition, we include LEV in Model (2).11 Empirical evidence (Basu et al. 2001; Ball and Shivakumar 2005) documents that debtholders who face an asymmetric payoff structure create a demand for more conditional conservatism. We also include a proxy for litigation risk using a firm-specific measure as modeled by Rogers and Stocken (2005).12 Prior studies indicate that firms can use conditional conservatism to reduce expected litigation costs (Basu 1997; Holthausen and Watts 2001; Chung and Wynn 2008). We also include an indicator variable for Big N auditors because prior studies (e.g., Basu et al. 2001) argue and find that clients of Big N auditors report more conservatively. This is viewed as a rational response to the threat of investor class-action lawsuits against auditors.

4 Empirical results

4.1 Descriptive statistics

Table 2 reports the summary statistics for selected variables used in the empirical analysis. Panel A involves the sample for which we construct the two industry concentration variables, PMC1(Sale) and PMC2(Sale). Panel B involves the sample for which we construct two dimensions of product market competition, EXIST-COMP and POTENT-COMP, based on Li (2010).
Table 2

Descriptive statistics

Variable

Mean

Std Dev

Q1

Median

Q3

Panel A: Descriptive statistics for the sample with product market competition variables, PMC1(SALE) and PMC2(SALE) (N = 71,627)

PMC1 (Sale)

−0.204

0.158

−0.258

−0.153

−0.096

PMC2 (Sale)

−0.059

0.024

−0.067

−0.052

−0.044

NI

0.021

0.191

−0.007

0.053

0.100

NEG

0.412

0.492

0

0

1

RET

0.205

0.646

−0.185

0.092

0.424

SIZE

18.796

1.950

17.367

18.653

20.103

BM

0.701

0.602

0.304

0.542

0.913

LEV

0.215

0.186

0.045

0.191

0.330

LIT

−2.997

0.376

−3.250

−3.027

−2.782

BIG_N

0.889

0.314

1

1

1

Panel B: Descriptive statistics for the sample with two dimensions of product market competition, EXIST-COMP and POTENT-COMP, are available (N = 78,799)

EXIST_COMP

1.742

1.753

0.573

1.326

2.596

POTENT_COMP

−1.005

1.455

−1.869

−0.490

0.120

LEADERS

0.233

0.423

0

0

0

NI

0.017

0.187

−0.008

0.052

0.095

NEG

0.417

0.493

0

0

1

RET

0.194

0.636

−0.187

0.084

0.405

SIZE

18.872

1.942

17.449

18.738

20.178

BM

0.668

0.561

0.298

0.527

0.874

LEV

0.215

0.188

0.044

0.190

0.332

LIT

−2.989

0.373

−3.239

−3.017

−2.774

BIG_N

0.887

0.317

1

1

1

All variables are defined in Appendices 1 and 2

With respect to Panel A, the average firm in our sample belongs to an industry with mean PMC1(Sale) value of −0.204, which is a bit larger than that of PMC2(Sale). The PMC2(Sale) has a mean of -0.059 and a median of −0.052, similar to values reported by Hoberg and Phillips (2010). Consistent with prior research (Deakin 1976; Frecka and Hopwood 1983; Watson 1990; Basu 1995, and others), the dependent variable NI exhibits left-skewness, indicating the presence of a few firms that report large accounting losses. In contrast, the explanatory variable RET has a right-skewed distribution—the mean value of RET of 0.205 is greater than the median value of 0.092—which reflects the fact that shareholders have limited liability and that they cannot lose beyond their investment. The indicator variable NEG has a mean value of 0.412, indicating that 41.2 % of the annual stock returns of our sample firm-year observations are negative. The mean and median values of market capitalization are $1,023 and $126 million, respectively. The mean book-to-market ratio of 0.701 exceeds its median book-to-market ratio of 0.542. The mean and median value of leverage for our sample firms are 0.215 and 0.191, respectively. The variable LIT has a mean value of −2.997, which translates to a 0.001 probability of being sued. Finally, about 89 % of our sample firms are audited by Big N auditors. The descriptive statistics of firm characteristics reported in Panel A are similar to those reported in Panel B.

Table 3 reports the average of annual Spearman and Pearson correlations estimated in each year for the sample for which PMC1(Sale) and PMC2(Sale) can be computed.13 The two product market competition measures differ in that PMC2(Sale) accounts for private firm sales while PMC1(Sale) does not. As expected, these product market competition measures are negatively correlated with NI and positively correlated with NEG, suggesting that firms in highly competitive industries experience lower profit levels and more negative stock returns. The negative correlation between PMC and SIZE indicates that firms in competitive industries tend to be smaller than firms in concentrated industries. There is a negative correlation between PMC and BM, which could be due to monopoly rents not being reflected in the book value of equity. PMC and LEV also exhibit a negative correlation, which is not altogether surprising because intense product market competition, by constraining profit margins, can limit a firm’s debt-carrying capacity. BIG_N is not significantly correlated with PMC1(Sale). However, it is negatively correlated with PMC2(SALE), which considers private firms, which are less likely to appoint Big N auditors. Similar to Ball and Brown (1968), NI is significantly related to RET and NEG, which implies that earnings contain a certain amount of information conveyed by returns.
Table 3

Spearman and pearson correlations

Variables

PMC1 (Sale)

PMC2 (Sale)

NI

NEG

RET

SIZE

BM

LEV

LIT

BIG_N

PMC1 (Sale)

1

0.383

−0.101

0.037

−0.041

−0.012

−0.098

−0.044

0.073

0.001

(0.00)

(0.00)

(0.01)

(0.08)

(0.11)

(0.00)

(0.00)

(0.00)

(0.34)

PMC2 (Sale)

0.411

1

−0.118

0.050

−0.039

−0.164

−0.049

−0.051

0.106

−0.034

(0.00)

(0.00)

(0.00)

(0.06)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

NI

−0.048

−0.077

1

−0.386

0.468

0.053

−0.124

−0.058

−0.123

0.011

(0.00)

(0.00)

(0.00)

(0.00)

(0.32)

(0.00)

(0.00)

(0.00)

(0.05)

NEG

0.035

0.034

−0.277

1

−0.830

−0.047

−0.110

0.024

−0.026

−0.026

(0.01)

(0.01)

(0.00)

(0.00)

(0.06)

(0.00)

(0.69)

(0.82)

(0.00)

RET

−0.020

−0.005

0.296

−0.647

1

−0.021

0.148

0.010

−0.025

0.015

(0.34)

(0.94)

(0.00)

(0.00)

(0.09)

(0.00)

(0.68)

(0.04)

(0.39)

SIZE

−0.020

−0.167

0.118

−0.052

−0.102

1

−0.302

−0.022

0.620

0.233

(0.00)

(0.00)

(0.00)

(0.06)

(0.00)

(0.00)

(0.41)

(0.00)

(0.00)

BM

−0.056

−0.057

−0.092

−0.098

0.133

−0.333

1

0.100

−0.108

−0.031

(0.34)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

(0.00)

LEV

−0.037

−0.048

−0.080

0.034

0.016

−0.038

0.040

1

0.009

−0.016

(0.00)

(0.00)

(0.00)

(0.30)

(0.58)

(0.03)

(0.00)

(0.38)

(0.00)

LIT

0.065

0.091

−0.122

−0.025

−0.037

0.602

−0.076

0.008

1

0.137

(0.00)

(0.00)

(0.00)

(0.83)

(0.00)

(0.00)

(0.00)

(0.62)

(0.00)

BIG_N

0.009

−0.037

0.012

−0.026

1.500

0.231

−0.048

−0.009

0.125

1

(0.13)

(0.00)

(0.01)

(0.00)

(0.14)

(0.00)

(0.00)

(0.00)

(0.00)

Spearman (Pearson) correlations are above (below) the diagonal. All correlations are reported as the average of annual cross-sectional correlations estimated in each year 1975–2005 Two-tailed p-values are based on averages, and standard deviations of annual correlation coefficients and are in parentheses. All variables are defined in Appendix 1

4.2 Multiple regression results

4.2.1 Primary results using the Herfindahl–Hirschman index

Regression results for alternative specifications of Model (2) are presented in Table 4. Column (1) presents regression results for a model that includes all variables except our product market competition variables. The coefficient on negative returns, NEG*RET, is 0.531 and statistically significant at the 0.01 level, indicating that on average, sample firms report losses more quickly than gains. In terms of our control variables, the coefficient on NEG*RET*SIZE is negative and statistically significant at the 0.01 level, indicating large firms tend to be less conservative than small firms. Consistent with prior research, we also find the coefficient on NEG*RET*BM is positive and statistically significant, which is consistent with the view that conditional conservatism is affected by prior conditional and unconditional conservatism (Beaver and Ryan 2005; Giner and Rees 2001; Basu 2001). The significantly positive coefficient on NEG*RET*LEV suggests that firms with greater leverage are incrementally more timely in recognizing bad news. Moreover, the coefficients on the interaction NEG*RET*LIT and NEG*RET*BIG_N are positive and statistically significant, suggesting that conditional conservatism is higher for firms with high litigation risk and for firms audited by Big N auditors.
Table 4

Regression results—product market competition and conditional accounting conservatism

 

Pred.

(1) Controls only

(2) PMC1 (Sale) only

(3) PMC1 (Sale) with controls

(4) PMC2 (Sale) only

(5) PMC2 (Sale) with controls

Sign

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

PMC (Sale)

  

0.002

0.17

−0.030***

−2.99

−0.121

−1.56

−0.302***

−4.22

SIZE

?

−0.014***

−13.06

  

−0.014***

−12.96

  

−0.013***

−12.23

BM

?

−0.078***

−35.34

  

−0.078***

−35.16

  

−0.078***

−35.02

LEV

?

−0.097***

−13.56

  

−0.096***

−13.40

  

−0.095***

−13.30

LIT

?

−0.037***

−9.46

  

−0.038***

−9.76

  

−0.039***

−9.88

BIG_N

?

−0.006

−1.29

  

−0.005

−1.26

  

−0.005

−1.25

NEG

?

−0.023

−0.59

−0.016***

−4.94

−0.025

−0.63

−0.029***

−5.43

−0.022

−0.55

NEG * PMC (Sale)

?

  

−0.016

−1.29

0.005

0.43

−0.289***

−3.41

−0.120

−1.50

NEG * SIZE

?

−0.001

−0.11

  

−0.001

−0.02

  

−0.001

−0.26

NEG * BM

?

−0.036***

−9.98

  

−0.035***

−9.79

  

−0.036***

−9.91

NEG * LEV

?

0.050***

4.96

  

0.049***

4.92

  

0.048***

4.80

NEG * LIT

?

−0.007

−1.15

  

−0.008

−1.19

  

−0.006

−0.93

NEG * BIG_N

?

0.011*

1.85

  

0.011*

1.83

  

0.012*

1.89

RET

+

−0.180***

−8.08

0.009***

4.39

−0.179***

−8.02

−0.014***

−3.54

−0.178***

−8.00

RET * PMC (Sale)

  

−0.113***

−12.54

−0.064***

−7.51

−0.797***

−12.15

−0.463***

−7.47

RET * SIZE

?

0.002**

2.41

  

0.002**

2.55

  

0.001*

1.65

RET * BM

?

0.024***

13.63

  

0.023***

12.86

  

0.022***

12.35

RET * LEV

?

0.076***

12.47

  

0.072***

11.66

  

0.069***

11.24

RET* LIT

−0.047***

−13.81

  

−0.043***

−12.49

  

−0.043***

−12.54

RET* BIG_N

?

0.007*

1.76

  

0.006

1.56

  

0.006

1.62

NEG * RET

+

0.531***

5.89

0.081***

9.75

0.519***

5.76

0.043***

2.99

0.525***

5.82

NEG * RET * PMC (Sale)

+

  

0.100***

2.83

0.060**

2.19

1.028***

4.20

0.420**

2.32

NEG * RET * SIZE

−0.025***

−7.62

  

−0.025***

−7.53

  

−0.024***

−7.33

NEG * RET * BM

+

0.235***

25.08

  

0.236***

25.05

  

0.236***

24.86

NEG * RET * LEV

+

0.302***

13.00

  

0.305***

13.05

  

0.308***

13.13

NEG * RET * LIT

+

0.049***

3.70

  

0.044***

3.30

  

0.046***

3.39

NEG * RET *BIG_N

+

0.039***

2.65

  

0.041***

2.75

  

0.040***

2.70

Firm fixed effect

 

Yes

 

Yes

 

Yes

 

Yes

 

Yes

 

N

 

71,627

 

71,627

 

71,627

 

71,627

 

71,627

 

Adjusted R2

 

0.4163

 

0.3659

 

0.4713

 

0.3658

 

0.4716

 

For brevity, intercepts are not reported for the firm-fixed effect model. All variables are as defined in Appendix 1. *, **,*** are significant at the 0.10, 0.05, and 0.01 levels, respectively

In columns (2) and (4), we report regression results for Model (2) without the firm-level control variables but with the two proxies of product market competition—one based strictly on public companies (PMC1(Sale)) and the other based on both public and private companies (PMC2(Sale)). Moreover, columns (3) and (5) show results of estimating model (2) using PMC1(Sale) and PMC2(Sale) as proxies of product market competition, respectively. Consistent with the results reported in Column (1) in Table 3, the coefficient estimates on negative returns (NEG*RET) are positive and statistically significant. The adjusted R2 increases from 41.6 % in column (1) to about 47.1 % in columns (3) and (5), suggesting that our product market competition variables provide incremental explanatory power to the base model.

In columns (3) and (5), the coefficients on RET*PMC (β15) are negative and statistically significant at the 0.01 level, suggesting that product market competition results in delayed gain recognition. Moreover, the coefficients on NEG*RET*PMC (β22) are positive and statistically significant at the 0.05 level in both columns. Taken together, these results suggest that, as product market competition becomes more intense, earnings become less timely in recognizing gains and more asymmetrically timely in recognizing losses. These results are consistent with the notion that conditional conservatism is shaped by strategic considerations in the presence of intense product market competition; they are inconsistent with the governance view, which suggests that competition negatively impacts the extent of conditional conservatism,

4.2.2 Regression results using two dimensions of product market competition

In this section, we investigate how different levels of competition within the same industry can affect conditional conservatism by examining two dimensions of product market competition: competition from existing rivals (EXIST-COMP) and competition from potential entrants (POTENT-COMP). Given that these two factors constitute different dimensions of product market competition, we estimate Model (2) separately for each of the two dimensions. Moreover, we follow Li (2010) and divide firms within the same industry into subgroups by first sorting them into quartiles according to their market shares and then identifying those in the top quartile as industry leaders and those in the remaining three quartiles as industry followers. As noted previously, the industry leader/follower distinction is useful to discriminate between alternative causal explanations. For example, the strategic consideration view predicts industry followers to exhibit more conditional conservatism in more competitive industries because they face more competitive pressures than industry leaders. In contrast, the political cost view predicts industry leaders to exhibit more conditional conservatism in less competitive industries because they are more likely to face regulatory scrutiny.

To discriminate between alternative causal explanations, we report regression results for alternative specifications of Model (2) involving EXIST-COMP and POTENT-COMP in Table 5, Panels A and B, respectively. In Panel A of Table 5, we report a set of regressions with only control variables (columns 1 and 4), only test variables (column 2 and 5), and both sets of variables (column 3 and 6) for the subsamples of industry leaders and industry followers. Adjusted R2s are the lowest (largest) for estimations with only the test (both test and control) variables. The coefficients on the interaction of control variables with negative returns in Table 5 generally have the same sign and level of significance as that of the control variables in Table 4. The coefficients on RET*PMC are negative and statistically significant at the 0.01 level in columns (3) and (6), suggesting that product market competition results in delayed gain recognition for both industry leaders and industry followers. However, the coefficients on NEG*RET*PMC are positive and statistically significant at the 0.01 level only for industry followers, suggesting that competition from exiting rivals improves the incremental timeliness of loss recognition only for the industry followers.14
Table 5

Regression results by type of product market competition for full sample and subsamples classified by the level of firm-level competition

PMC proxied by EXIST_COMP

  

Leaders

Followers

Pred.

(1) Controls only

(2) PMC only

(3) PMC with controls

(4) Controls only

(5) PMC only

(6) PMC with controls

Sign

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

Panel A: Competition from existing rivals

PMC

  

−0.011***

−3.25

−0.001

−0.28

  

−0.009***

−4.84

−0.001

−0.57

SIZE

?

−0.012***

−6.73

  

−0.011***

−8.00

−0.012***

−8.79

  

−0.010***

−5.96

BM

?

−0.051***

−11.01

  

−0.093***

−36.17

−0.094***

−37.05

  

−0.048***

−10.33

LEV

?

−0.089***

−7.66

  

−0.086***

−10.56

−0.092***

−11.39

  

−0.084***

−7.20

LIT

?

−0.023***

−3.63

  

−0.044***

−9.79

−0.040***

−9.19

  

−0.027***

−4.17

BIG_N

?

−0.013

−1.41

  

0.007

1.50

0.007

1.51

  

−0.012

−1.32

NEG

?

0.083

1.11

−0.025***

−3.02

−0.012

−0.26

−0.021

−0.47

0.008*

1.91

0.106

1.39

NEG * PMC

?

  

−0.001

−0.18

−0.001

−0.74

  

0.002

0.90

−0.002

−1.25

NEG * SIZE

?

−0.003

−1.13

  

0.001

0.01

0.000

−0.04

  

−0.003

−1.25

NEG * BM

?

−0.031***

−4.48

  

−0.028***

−6.79

−0.028***

−6.80

  

−0.033***

−4.60

NEG * LEV

?

0.039**

2.33

  

0.035***

3.07

0.039***

3.56

  

0.036**

2.11

NEG * LIT

?

0.004

0.32

  

−0.004

−0.55

−0.006

−0.84

  

0.007

0.64

NEG * BIG_N

?

−0.010

−0.74

  

0.009

1.34

0.008

1.25

  

−0.011

−0.80

RET

+

0.149***

2.75

0.040***

7.93

−0.102***

−3.77

−0.157***

−6.04

0.039***

9.05

0.198***

3.56

RET * PMC

  

−0.011***

−6.67

−0.006***

−8.06

  

−0.005***

−3.61

−0.006***

−4.09

RET * SIZE

?

−0.007***

−3.65

  

0.002*

1.82

0.002**

2.23

  

−0.007***

−3.78

RET * BM

?

−0.012***

−2.71

  

0.016***

8.31

0.020***

10.34

  

−0.017***

−3.74

RET * LEV

?

0.004

0.32

  

0.055***

8.02

0.072***

10.78

  

−0.006

−0.46

RET* LIT

−0.013*

−1.71

  

−0.031***

−7.92

−0.040***

−10.64

  

−0.004

−0.56

RET* BIG_N

?

−0.011

−1.02

  

0.001

0.30

0.001

0.24

  

−0.012

−1.09

NEG * RET

+

0.822***

4.11

0.090***

4.75

0.687***

6.51

0.742***

7.15

0.202***

14.01

0.763***

3.77

NEG * RET * PMC

+

  

0.008

1.19

0.003

1.15

  

0.020***

4.37

0.009***

2.95

NEG * RET * SIZE

−0.024***

−3.57

  

−0.039***

−9.52

−0.040***

−9.86

  

−0.024***

−3.57

NEG * RET * BM

+

0.339***

16.86

  

0.216***

20.74

0.216***

21.27

  

0.349***

16.83

NEG * RET * LEV

+

0.302***

6.65

  

0.277***

10.64

0.274***

10.94

  

0.322***

6.90

NEG * RET * LIT

+

0.142***

5.31

  

0.021

1.40

0.028*

1.93

  

0.132***

4.83

NEG * RET *BIG_N

+

−0.007

−0.17

  

0.054***

3.54

0.053***

3.44

  

−0.009

−0.21

Firm Fixed Effect

 

Yes

 

Yes

 

Yes

 

Yes

 

Yes

 

Yes

 

N

 

18,348

 

18,348

 

18,348

 

60,451

 

60,451

 

60,451

 

Adjusted R2

 

0.4012

 

0.3625

 

0.4520

 

0.4035

 

0.3724

 

0.4845

 

PMC proxied by POTENT_COMP

  

Leaders

Followers

Pred.

(1) PMC only

(2) PMC with controls

(3) PMC only

(4) PMC with controls

Sign

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

Panel B: Competition from potential entrants

PMC

−0.014***

−7.38

0.005***

2.73

−0.006***

−4.18

0.002

1.11

SIZE

?

  

−0.010***

−5.46

  

−0.011***

−8.02

BM

?

  

−0.051***

−10.98

  

−0.094***

−36.87

LEV

?

  

−0.086***

−7.41

  

−0.090***

−11.21

LIT

?

  

−0.023***

−3.62

  

−0.042***

−9.43

BIG_N

?

  

−0.012

−1.33

  

0.007

1.58

NEG

?

0.001

0.36

0.085

1.13

−0.011***

−3.80

−0.016

−0.36

NEG * PMC

?

0.000

0.23

−0.001

−0.01

−0.002

−1.51

0.002

1.14

NEG * SIZE

?

  

−0.003

−1.11

  

0.001

0.07

NEG * BM

?

  

−0.031***

−4.48

  

−0.028***

−6.91

NEG * LEV

?

  

0.039**

2.33

  

0.037***

3.39

NEG * LIT

?

  

0.004

0.39

  

−0.005

−0.65

NEG * BIG_N

?

  

−0.011

−0.78

  

0.007

1.16

RET

+

0.026***

7.09

0.151***

2.78

0.037***

18.67

−0.140***

−5.37

RET * PMC

−0.001

−0.41

−0.001

−0.28

−0.009***

−8.69

0.005***

5.42

RET * SIZE

?

  

−0.007***

−3.72

  

0.003**

2.47

RET * BM

?

  

−0.012**

−2.58

  

0.019***

9.76

RET * LEV

?

  

0.004

0.29

  

0.068***

10.15

RET* LIT

  

−0.013*

−1.74

  

−0.036***

−9.41

RET* BIG_N

?

  

−0.011

−1.00

  

−0.001

−0.02

NEG * RET

+

0.165***

13.16

0.846***

4.21

0.134***

18.01

0.724***

6.98

NEG * RET * PMC

+

0.011*

1.70

−0.005

−0.81

0.029***

7.48

0.005**

1.98

NEG * RET * SIZE

  

−0.025***

−3.65

  

−0.039***

−9.44

NEG * RET * BM

+

  

0.339***

16.79

  

0.212***

20.77

NEG * RET * LEV

+

  

0.306***

6.73

  

0.269***

10.68

NEG * RET * LIT

+

  

0.143***

5.33

  

0.028*

1.94

NEG * RET *BIG_N

+

  

−0.010

−0.24

  

0.053***

3.46

Firm Fixed Effect

 

Yes

 

Yes

 

Yes

 

Yes

 

N

 

18,348

 

18,348

 

60,410

 

60,410

 

Adjusted R2

 

0.3606

 

0.4516

 

0.3781

 

0.4851

 

For brevity, intercepts are not reported for the firm-fixed effect model. All variables are as defined in appendices. Leaders (Followers) are firm-years classified as industry leader if its sales rank in the top quartile (bottom three quartiles) in an industry based on three-digit SIC code. *, **,*** are significant at the 0.10, 0.05, and 0.01 levels, respectively

Results in panel B of Table 5 mirror those reported in Panel A of Table 5. Overall, the results in Table 4 indicate that the association of asymmetric timeliness with the two dimensions of product market competition is muted for industry leaders, which is opposite of what one might expect under a political costs hypothesis.15 The absence of a statistically significant effect for industry leaders is also inconsistent with the monopoly rent argument, which predicts lower conditional conservatism for industry leaders in less competitive industries. Overall, the statistically significant effect for firms classified as industry followers supports the argument that strategic considerations shape firm accounting decisions.

4.2.3 Industry deregulation and conditional conservatism

To further identify the relation between product market competition and conditional conservatism, we examine firm reporting behavior surrounding industry deregulation, which contributes to an increase in the intensity of product market competition. This setting allows us to provide temporal, as opposed to the previous cross-sectional, evidence on the relation between accounting conservatism and product market competition. Similar to Winston (1998), we focus on three industries that experienced industry deregulation during our sample period: airlines (SIC 4512) in 1978, telecommunications (SIC 4813) in 1982, and transportation (SIC 4011 and 4013) in 1980.

In light of the evidence that conservatism increased over parts of the sample period (Basu 1997; Givoly and Hayn 2000; Holthausen and Watts 2001; Ryan and Zarowin 2003), we adopt a difference-in-differences approach. This approach allows us to isolate the effects of deregulation while at the same time accounting for time trends in reporting behavior. We consider not only firms in the deregulated industries, i.e., treatment firms, but also firms in industries that were not subject to deregulation, i.e., control firms. With respect to these control firms, we created a benchmark sample that consists of a matched sample of firm-years where the matched firm (1) is not in the airline, telecommunications, or transportation industry or regulated industries such as utilities and financial; (2) has the required Compustat and CRSP data; and (3) is in a 4-digit SIC industry that has an average Herfindahl index close to that of the comparison industry during the five-year period prior to deregulation.

Panel A of Table 6 compares the five-year mean values of product market competition and selected firm characteristics prior to and after deregulation for the treatment and benchmark samples.16,17 As expected, the mean values of product market competition after deregulation are significantly higher than that before deregulation for the treatment sample. These statistically significant differences in product market competition over time suggest that industry deregulation increases the level of product market competition that confronts our treatment firms. In contrast, there are no statistically significant differences between the mean values of product market competition over time for the benchmark sample. Moreover, firms in deregulated industries are more levered than other firms in the pre- and post-deregulation periods. We also see a dramatic increase in firm size and the use of Big N auditors over time. Further analysis reveals that the size differences over time disappear after we account for inflation. Moreover, the growth in the use of Big N auditor is also evident in the Compustat population of firms during the period covered in our analyses.18 Book-to-market ratios decline over time, which could be due to the increasing use of more conservative accounting policies over time or the effects of high inflation especially during 1970s where market values inflate but book values do not adjust under historical cost (Basu 1997, fn. 4). These changes in several firm characteristics over time highlight the importance of controlling for these variables in the multivariate tests.
Table 6

Descriptive statistics and regression results for firms in deregulated industries and a control sample

 

Deregulated industries (Nobs = 860)

Control industries (Nobs = 750)

Before

After

Diff

t-statistics

Before

After

Diff

t-statistics

Panel A: Descriptive statistics of product market competition and control variables pre- and post-deregulation

PMC1 (Sale)

−0.119

−0.102

0.017

(2.48)***

−0.132

−0.138

−0.005

(−0.47)

PMC2 (Sale)

−0.259

−0.176

0.083

(5.76)***

−0.203

−0.215

−0.012

(−0.70)

SIZE

18.825

19.191

0.367

(3.58)***

18.790

19.535

0.744

(4.71)***

BM

1.258

1.094

−0.164

(−5.26)***

1.093

0.856

−0.237

(−4.06)***

LEV

0.443

0.373

−0.070

(−10.14)***

0.329

0.313

−0.015

(−0.89)

LIT

−3.171

−3.118

0.053

(3.58)***

−3.148

−3.036

0.113

(4.80)***

BIG_N

0.366

0.550

0.183

(6.38)***

0.605

0.801

0.196

(4.63)***

 

Pred.

(1) Deregulation only

(2) Deregulation with controls

Sign

Coeff.

t-stat

Coeff.

t-stat

Panel B: Regression results

DeReg

  

0.021

1.16

Post

0.025

1.15

0.028

1.35

DeReg * Post

−0.017

−0.65

−0.021

−0.88

LEADERS

?

0.055**

2.41

0.042*

1.82

SIZE

?

  

−0.036***

−2.68

BM

?

  

0.003

0.17

LEV

?

  

−0.266***

−3.98

LIT

?

  

0.110**

2.51

BIG_N

?

  

−0.017

−1.06

NEG

?

0.049*

1.81

0.034

0.08

NEG * DeReg

?

−0.042

−1.33

−0.067**

−2.12

NEG * Post

?

0.009

0.21

−0.020

−0.49

NEG * DeReg * Post

?

0.028

0.58

0.049

1.06

NEG * LEADERS

?

−0.052**

−2.14

−0.026

−0.91

NEG * SIZE

?

  

−0.010

−0.82

NEG * BM

?

  

−0.020

−0.91

NEG * LEV

?

  

0.073

0.96

NEG * LIT

?

  

−0.061

−0.83

NEG * BIG_N

?

  

0.028

1.26

RET

+

0.214***

7.00

−1.315***

−3.46

RET * DeReg

−0.048

−1.27

−0.048

−1.28

RET * Post

−0.136***

−3.75

−0.109***

−3.09

RET * DeReg * Post

0.014

0.29

0.008

0.18

RET * LEADERS

−0.007

−1.41

0.004

0.11

RET * SIZE

?

  

0.017

1.52

RET * BM

?

  

−0.033*

−1.79

RET * LEV

?

  

0.099

1.38

RET* LIT

  

−0.382***

−6.20

RET* BIG_N

?

  

−0.037

−1.45

NEG * RET

+

0.002

0.02

5.975***

5.58

NEG * RET * DeReg

+

−0.005

−0.04

−0.182

−1.43

NEG * RET * Post

+

0.261*

1.86

0.041

0.29

NEG * RET * DeReg * Post

+

0.615***

3.67

0.639***

3.93

NEG * RET * LEADERS

+

−0.141

−1.45

−0.073

−0.66

NEG * RET * SIZE

  

−0.174***

−5.18

NEG * RET * BM

+

  

0.099

1.45

NEG * RET * LEV

+

  

0.502**

1.98

NEG * RET * LIT

+

  

0.946***

5.21

NEG * RET *BIG_N

+

  

0.310***

3.35

Firm Fixed Effect

 

Yes

 

Yes

 

N

 

1,610

 

1,610

 

Adjusted R2

 

0.5146

 

0.5695

 

For brevity, intercepts are not reported for the firm-fixed effect model. DeReg is 1 for the treatment firms and 0 for the benchmark sample; Post is 1 in the post-deregulation period and 0 otherwise. All other variables are as defined in Appendix 1. *, **,*** are significant at the 0.10, 0.05, and 0.01 levels, respectively

Figure 1 provides a graphical representation of the coefficient on NEG*RET based on Basu’s (1997) model during the pre-deregulation period (years t–5 through t–1), the deregulation year (year t), and the post-deregulation period (years t + 1 through t + 5) for the treatment and benchmark samples. We observe a steady increase in conditional conservatism after the deregulation year (Year t) for the treatment sample but not for the control sample. While the results in Panel A in Table 5 and Fig. 1 indicate that product market competition and conditional conservatism increased after deregulation, they should be considered as informal and complementary to the formal tests performed next because they do not discriminate between strategic considerations and alternative explanations. In other words, because firm financial reporting is likely to be shaped over time by other firm characteristics as well, it is important to design tests to rule out alternative explanations for the univariate results.
https://static-content.springer.com/image/art%3A10.1007%2Fs11142-013-9267-2/MediaObjects/11142_2013_9267_Fig1_HTML.gif
Fig. 1

Conditional conservatism around industry deregulation Y-axis shows the coefficients on NEG*RET based on Basu’s (1997) model, and X-axis depicts the pre-deregulation period (years t−5 through t−1), the deregulation year (year t), and the post-deregulation period (years t + 1 through t + 5). DeReg is 1 for the treatment firms and 0 for the benchmark sample

Specifically, we compare the difference in the conditional conservatism of firms in the deregulated industries over time to the difference in conditional conservatism of a benchmark sample over time by using the following regression:
$$ \begin{aligned} {\text{NI}}_{\text{t}} & = \upbeta_{0} + \upbeta_{1} {\text{De}}\text{Re} {\text{g}} + \upbeta_{2} {\text{Post}} + \upbeta_{3} {\text{DeReg}}*{\text{Post}} + \upbeta_{4} {\text{LEADERS}}_{\text{t - 1}} + \upbeta_{5} {\text{SIZE}}_{\text{t - 1}} + \upbeta_{6} {\text{BM}}_{\text{t - 1}} \\ &\quad + \upbeta_{7} {\text{LEV}}_{\text{t - 1}} + \upbeta_{8} {\text{LIT}}_{{{\text{t}} - 1}} + \upbeta_{9} {\text{BIG}}\_{\text{N}}_{\text{t}} + \upbeta_{10} {\text{NEG}}_{\text{t}} + \upbeta_{11} {\text{NEG}}_{\text{t}} * {\text{DeReg}} + \upbeta_{12} {\text{NEG}}_{\text{t}} * {\text{Post}} \\ &\quad + \upbeta_{13} {\text{NEG}}_{\text{t}} * {\text{DeReg}}*{\text{Post}} + \upbeta_{14} {\text{NEG}}_{\text{t}} * {\text{LEADERS}}_{\text{t - 1}} + \upbeta_{15} {\text{NEG}}_{\text{t}} * {\text{SIZE}}_{\text{t - 1}} + \upbeta_{16} {\text{NEG}}_{\text{t}} *{\text{BM}}_{\text{t - 1}} \\ & \quad + \upbeta_{17} {\text{NEG}}_{\text{t}} * {\text{LEV}}_{\text{t - 1}} + \upbeta_{18} {\text{NEG}}_{\text{t}} * {\text{LIT}}_{\text{t - 1}} + \upbeta_{19} {\text{NEG}}_{\text{t}} * {\text{BIG}}\_{\text{N}}_{\text{t}} + \upbeta_{20} {\text{RET}}_{t} + \upbeta_{21} {\text{RET}}_{\text{t}} *{\text{DeReg}} + \upbeta_{22} {\text{RET}}_{\text{t}} *{\text{Post}} \\ &\quad + \upbeta_{23} {\text{RET}}_{\text{t}} *{\text{DeReg}}*{\text{Post}} + \upbeta_{24} {\text{RET}}_{\text{t}} *{\text{LEADER}}_{\text{t - 1}} + \upbeta_{25} {\text{RET}}_{t} * {\text{SIZE}}_{\text{t - 1}} + \upbeta_{26} {\text{RET}}_{\text{t}} * {\text{BM}}_{\text{t - 1}} + \upbeta_{27} {\text{RET}}_{\text{t}} * {\text{LEV}}_{\text{t - 1}} \\ &\quad + \upbeta_{28} {\text{RET}}_{\text{t}} * {\text{LIT}}_{{{\text{t}} - 1}} + \upbeta_{29} {\text{RET}}_{\text{t}} * {\text{BIG}}\_{\text{N}}_{\text{t}} + \upbeta_{30} {\text{NEG}}_{\text{t}} *{\text{RET}}_{\text{t}} + \upbeta_{31} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} *{\text{DeReg}} + \upbeta_{32} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} *{\text{Post}} \\ &\quad + \upbeta_{33} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} *{\text{DeReg}}*{\text{Post}} + \upbeta_{34} {\text{NEG}}*{\text{RET}}_{\text{t}} *{\text{LEADERS}}_{\text{t - 1}} *{\text{Post}} + \upbeta_{35} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} * {\text{SIZE}}_{\text{t - 1}} \\ &\quad + \upbeta_{36} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} *{\text{BM}}_{\text{t - 1}} +\upbeta_{37} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} * {\text{LEV}}_{\text{t - 1}} + \upbeta_{38} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} * {\text{LIT}}_{\text{t - 1}} + \upbeta_{39} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} *{\text{BIG}}\_{\text{N}}_{\text{t}} + {\text{Firm}} - {\text{fixed effects}} + \upepsilon \\ \end{aligned} $$
(3)
where DeReg is 1 for the treatment firms and 0 for the benchmark sample; Post is 1 in the post-deregulation period and 0 otherwise; LEADERS is an indicator variable that equals 1 for a firm-year classified as industry leader if its sales rank in the top quartile in an industry based on three-digit SIC code and 0 otherwise; and all other variables are as defined before.19 Given the increased product market competition following deregulation, we expect greater accounting conservatism in the post-deregulation period for the treatment firms relative to the benchmark sample. Hence we expect the coefficient on RET*DeReg*Post to be negative and the coefficient on RET*NEG*DeReg*Post to be positive. These coefficients serve as our estimate of the effect of deregulation on conditional conservatism.

Panel B of Table 6 reports the regression results for alternative specifications of Model (3). Column (1) reports regressions with only test variables and column (2) reports regressions with test and control variables. In both columns, the coefficients on RET*DeReg*Post are not statistically significant at the 0.10 level, suggesting that deregulation in the airlines, telecommunications, and transportation industries did not result in less timely recognition of economic gains over time than our benchmark sample. However, the coefficients on NEG*RET*DeReg*POST are positive and statistically significant at the 0.01 level, suggesting that intense product market competition induces more timely accounting recognition of economic losses for the treatment firms than that for the benchmark sample in the post-deregulation era. Interestingly, the coefficients on NEG*RET*LEADERS are not statistically significant, which is inconsistent with the political cost argument. In summary, we find evidence of increases in the intensity of product market competition, inducing more timeliness in recognition of bad news but not affecting the recognition of good news.

4.2.4 Antitrust enforcement and conditional conservatism

In this section, we test for time-series variation in conditional conservatism for a broader sample than the above sample of three industries subject to deregulation. Specifically, we focus on variation in antitrust enforcement in the United States during the 1964–2006 period.20 In the United States, antitrust laws promote competition in the marketplace by prohibiting a variety of practices that restrain trade, such as price-fixing, horizontal mergers, and monopolization. Hence the extent of product market competition is affected by the enforcement of the antitrust laws in place. In other words, we expect the level of antitrust enforcement to have a bearing on conditional conservatism. To test for the time-series variation in conditional conservatism due to antitrust enforcement, we estimate the following model:
$$ \begin{aligned} {\text{NI}}_{\text{t}} &= \upbeta_{0} + \upbeta_{1} {\text{Antitrust}} + \upbeta_{2} {\text{HIGH}}\_{\text{LIT}} + \upbeta_{3} {\text{SIZE}}_{\text{t - 1}} + \upbeta_{4} {\text{BM}}_{{{\text{t}} - 1}} + \upbeta_{5} {\text{LEV}}_{\text{t - 1}} + \upbeta_{6} {\text{LIT}}_{{{\text{t}} - 1}} + \upbeta_{7} {\text{BIG}}\_{\text{N}}_{\text{t - 1}} + \upbeta_{8} {\text{NEG}}_{\text{t}} \\ &\quad + \upbeta_{9} {\text{NEG}}_{\text{t}} * {\text{Antitrust}} + \upbeta_{10} {\text{NEG}}_{\text{t}} * {\text{HIGH}}\_{\text{LIT}} + \upbeta_{11} {\text{NEG}}_{\text{t}} * {\text{SIZE}}_{\text{t - 1}} + \upbeta_{12} {\text{NEG}}_{\text{t}} *{\text{BM}}_{\text{t - 1}} + \upbeta_{13} {\text{NEG}}_{\text{t}} * {\text{LEV}}_{\text{t - 1}} \\ &\quad+ \upbeta_{14} {\text{NEG}}_{\text{t}} * {\text{LIT}}_{\text{t - 1}} + \upbeta_{15} {\text{NEG}}_{\text{t}} * {\text{BIG}}\_{\text{N}}_{\text{t - 1}} + \upbeta_{16} {\text{RET}}_{\text{t}} + \upbeta_{17} {\text{RET}}_{\text{t}} *{\text{Antitrust}} + \upbeta_{18} {\text{RET}}_{\text{t}} *{\text{HIGH}}\_{\text{LIT}} + \upbeta_{19} {\text{RET}}_{\text{t}} * {\text{SIZE}}_{\text{t - 1}} \\ &\quad + \upbeta_{20} {\text{RET}}_{\text{t}} * {\text{BM}}_{\text{t - 1}} + \upbeta_{21} {\text{RET}}_{\text{t}} * {\text{LEV}}_{\text{t - 1}} + \upbeta_{22} {\text{RET}}_{\text{t}} * {\text{LIT}}_{{{\text{t}} - 1}} + \upbeta_{23} {\text{RET}}_{\text{t}} * {\text{BIG}}\_{\text{N}}_{\text{t - 1}} + \upbeta_{24} {\text{NEG}}_{\text{t}} *{\text{RET}}_{\text{t}} \\ &\quad + \upbeta_{25} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} *{\text{Antitrust}} + \upbeta_{26} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} *{\text{HIGH}}\_{\text{LIT}} + \upbeta_{27} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} * {\text{SIZE}}_{\text{t - 1}} \\ &\quad+ \upbeta_{28} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} *{\text{BM}}_{\text{t - 1}} + \upbeta_{29} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} * {\text{LEV}}_{\text{t - 1}} + \upbeta_{30} {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} * {\text{LIT}}_{\text{t - 1}} + \upbeta_{31} * {\text{NEG}}_{\text{t}} * {\text{RET}}_{\text{t}} *{\text{BIG}}\_{\text{N}}_{\text{t - 1}} + {\text{Firm}} - {\text{fixed effects}} + \upepsilon \\ \end{aligned} $$
(4)

The variable Antitrust is a natural logarithm of the number of antitrust case filings in one year and hence represents the level of antitrust enforcement; HIGH_LIT is 1 if the sample year is within the period 1964–1975 or 1983–1995 or 2002–2006 and 0 otherwise; and all other variables are as defined earlier.21 If high levels of antitrust enforcement facilitate greater product market competition, accounting conservatism should increase under our argument. Furthermore, given that industry leaders face greater regulatory scrutiny, the effect of antitrust enforcement should be more pronounced for industry leaders than industry followers. Hence we expect the anti-trust enforcement effect on incremental timely loss recognition to be more pronounced for industry leaders than for industry followers.

Table 7 reports the regression results for Model (4) using the full sample and subsamples of industry leaders and industry followers. The coefficients on RET*Antitrust are positive but not statistically significant at the 0.10 level. The coefficients on NEG*RET*Antitrust are positive and statistically significant at the 0.05 level. Moreover, the untabulated F-test statistics of 3.79 indicates that anti-trust enforcement effect on incremental timely loss recognition is more pronounced for industry leaders than for followers. Taken together, these results are consistent with the argument that greater antitrust enforcement facilitates more intense product market competition and that this serves to induce firms to recognize bad news in a timely manner.22 These results hold after controlling for firm-specific litigation risk and the overall litigation environment.
Table 7

Regression results—changes in antitrust enforcement and conditional accounting conservatism

 

Pred.

(1) Full sample

(2) Leaders

(3) Followers

Sign

Coeff.

t-stat

Coeff.

t-stat

Coeff.

t-stat

Antitrust

?

0.059***

19.79

0.059***

12.50

0.055***

14.53

HIGH_LIT

?

−0.018***

−9.29

−0.010***

−3.40

−0.022***

−9.11

SIZE

?

−0.007***

−7.43

−0.005***

−2.78

−0.010***

−8.13

BM

?

−0.063***

−33.15

−0.027***

−7.30

−0.074***

−33.07

LEV

?

−0.084***

−13.60

−0.050***

−4.59

−0.096***

−12.75

LIT

?

−0.034***

−9.69

−0.026***

−4.18

−0.033***

−7.80

BIG_N

?

0.007***

2.76

0.007

1.61

0.009***

2.82

NEG

?

−0.125***

−2.65

−0.012

−0.14

−0.131**

−2.22

NEG * Antitrust

?

0.011**

2.17

0.014*

1.66

0.012*

1.78

NEG * HIGH_LIT

?

0.005

1.40

−0.001

−0.13

0.007

1.55

NEG * SIZE

?

0.001

1.12

−0.003

−1.16

0.001

0.92

NEG * BM

?

−0.020***

−6.50

−0.019***

−3.04

−0.021***

−5.65

NEG * LEV

?

0.043***

4.88

0.022

1.39

0.040***

3.83

NEG * LIT

?

−0.009

−1.62

0.006

0.56

−0.010

−1.38

NEG * BIG_N

?

−0.008**

−2.06

−0.005

−0.73

−0.007

−1.53

RET

+

−0.309***

−9.88

−0.143**

−1.96

−0.302***

−8.08

RET * Antitrust

0.029

1.29

0.033

1.19

0.029

1.23

RET * HIGH_LIT

−0.001

−0.54

−0.002

−0.50

0.001

0.12

RET * SIZE

?

−0.001

−0.09

−0.007***

−3.71

−0.001

−0.68

RET * BM

?

0.019***

11.73

−0.005

−1.39

0.022***

12.31

RET * LEV

?

0.066***

11.86

0.001

0.02

0.073***

11.46

RET* LIT

−0.044***

−13.86

−0.040***

−5.68

−0.043***

−11.66

RET* BIG_N

?

−0.002

−0.82

−0.006

−0.81

−0.001

−0.13

NEG * RET

+

−0.091

−0.73

−0.006

−0.02

0.117

0.78

NEG * RET * Antitrust

+

0.072***

4.78

0.097***

3.29

0.062***

3.51

NEG * RET * HIGH_LIT

+

0.020**

2.32

0.017*

1.68

0.012**

2.21

NEG * RET * SIZE

−0.018***

−6.18

−0.015**

−2.34

−0.028***

−7.05

NEG * RET * BM

+

0.193***

24.11

0.195***

11.47

0.186***

20.16

NEG * RET * LEV

+

0.282***

13.85

0.209***

4.79

0.268***

11.34

NEG * RET * LIT

+

0.060***

4.91

0.182***

7.08

0.044***

3.12

NEG * RET *BIG_N

+

0.100***

9.58

0.165***

7.06

0.091***

7.57

Firm Fixed Effect

 

Yes

 

Yes

 

Yes

 

N

 

92,703

 

21,802

 

70,901

 

Adjusted R2

 

0.4403

 

0.4099

 

0.4587

 

For brevity, intercepts are not reported for the firm-fixed effect model. Antitrust is the natural log of the number of antitrust case filings; HIGH_LIT is 1 if the sample year is within the period 1964–1975, 1983–1995, or 2002–2006 and 0 otherwise. All other variables are as defined in Appendix 1. Leaders (Followers) are firm-years classified as industry leader if its sales rank in the top quartile (bottom three quartiles) in an industry based on three-digit SIC code *, **,*** are significant at the 0.10, 0.05, and 0.01 levels, respectively

4.3 Robustness tests

  1. 1.

    We test our hypothesis with alternative conditional conservatism measures such as asymmetric persistence of earnings (Basu 1997) and the asymmetric accrual-cash flow relation (Ball and Shivakumar 2005). There is no qualitative change in our inferences.

     
  2. 2.

    In our empirical model (2), we use the book-to-market ratio to control for unconditional conservatism. Roychowdhury and Watts (2007) argue that the book-to-market ratio is likely to contain measurement errors. For example, stock returns reflect changes in both booked assets and (unbooked) monopoly rents, while earnings (asymmetrically) reflect changes only in booked assets (see also Ball and Shivakumar 2006). This measurement error is especially important in our setting, as monopoly rents vary with product market competition. To mitigate the effect of this measurement error in our analysis, we follow Sunder et al. (2009) and use the residual derived from the following model to isolate the component of the book-to-market ratio that is not related with monopoly rents, growth prospects, financial distress, and market sentiment.

    $$ \begin{aligned} {\text{BM}} & = \upbeta_{0} + \upbeta_{1} {\text{long}} - {\text{term growth forecasts}} + \upbeta_{2} {\text{sales growth}} + \upbeta_{3} {\text{industry concentration}} \\ &\quad + \upbeta_{4} {\text{industry concentration}}*{\text{indicator of top four companies}} + \upbeta_{5} {\text{S}}\& {\text{P index}} \\ &\quad + \upbeta_{6} {\text{consumer sentiment index}} + \upbeta_{7} {\text{ROA}} + \upbeta_{8} {\text{credit rating}} + \upbeta_{9} {\text{standard deviation of returns}} + \upepsilon \\ \end{aligned} $$
    (5)
    where Long-Term Growth Forecasts is the median of all long-term growth estimates made by analysts in the previous year; Sales Growth is defined as current sales scaled by sales in prior year; Industry Concentration is the Herfindahl Index defined as before; Indicator of Top Four Companies equals 1 if the firm is one of the top four companies based on sales in an industry and 0 otherwise; S&P Index is the S&P’s Composite Index from CRSP; Consumer Sentiment Index is the index of the consumer sentiment constructed by the University of Michigan; ROA is measured as earnings scaled by lagged total assets; Credit Rating is S&P long-term domestic issuer credit rating; Standard Deviation of Returns is the standard deviation of daily stock returns during the year. Because of additional data requirements, our sample reduces to 47,151 observations for this analysis. When we use the estimated residual of the book-to-market ratio to control for unconditional conservatism in Model (2), our inferences with respect to product market competition remain the same. We generally find firms in more competitive industries display more timely recognition of economic losses.
     
  1. 3.

    We also conducted additional analysis by explicitly controlling for accounting method choice, which serves to affect unconditional accounting conservatism.23 For example, firms can lower their earnings by adopting accelerated depreciation methods, the “last in, first out” (LIFO) inventory cost flow assumption, or both. To address the issue of accounting methods, we turn to footnote items on financial statements in Compustat. We create two dichotomous variables to capture the depreciation and inventory cost flow-related accounting method choices. The first variable is equal to 1 if a firm uses an accelerated depreciation method and 0 otherwise. The second variable takes a value of 1 if the firm adopts LIFO and 0 otherwise.24 We include these two variables, their two-way interactions with RET and with NEG, and the three-way interactions with RET and NEG into Model (2). We find our results with respect to conditional conservatism still continue to hold, i.e., firms in more competitive industries display more timely recognition of economic losses.

     
  2. 4.

    We also reestimate our models using auditor industry specialists based on prior research (e.g., Krishnan 2005; Reichelt and Wang 2010) as an alternative control variable for auditor quality. We find that doing so does not affect our inferences.

     
  3. 5.

    Hui et al. (2012) show that, when a firm’s contracting partners, primarily suppliers and customers, have greater bargaining power, the firm recognizes losses more quickly. Their argument is that a firm’s contracting partners demand conservative reporting to constrain post-contractual opportunism (Klein et al. 1978). Following Hui et al. (2012), we control for the influence of stakeholders such as consumers and suppliers. To this end, we use the average values of bargaining power in the form of relative size reported in Table 2 of Hui et al. (2012). The results for our product market competition measures were not affected by including supplier relative size or customer relative size in our empirical models. However, the results for the additional control variables (suppliers’ or customers’ relative size) were mixed and sometimes different from Hui et al.’s (2012) finding, likely due to the fact that we do not have a firm-level measure of the bargaining power of suppliers or customers. More importantly, our results suggest that, even though a firm’s contracting partners can influence the demand for conditional conservatism (Hui et al. 2012), firms with greater bargaining power, one proxy being less product market competition, can better resist such external demands for conservatism.

     

5 Conclusion

This paper examines the impact of product market competition on financial reporting decision in the form of conditional conservatism (i.e., asymmetric timely loss recognition). We find intense product market competition is positively associated with conditional conservatism. Generally, we find this relation to hold when we distinguish between competition from potential entrants and that from rivals. We also document time-series evidence on the product market competition-conditional conservatism relation. Specifically, we find an increase in conditional conservatism after deregulation and in periods of increased antitrust enforcement. We also examine whether the effects of product market competition differ between firms within the same industry with respect to their market share. We find the positive relation between product market competition and asymmetric timely loss recognition holds for industry followers but not for industry leaders.

The literature offers several alternative views on the competition-conservatism relation. Our findings rule out the political costs argument that firms undertake conservatism to avoid regulatory scrutiny when competition is lacking. They also contradict the governance argument that intense competition constrains managers such that it reduces the need for conservatism. Rather, our evidence suggests that firms undertake conservatism to improve their competitive position in the face of intense product market competition stemming from existing rivals as well as the threat of potential entrants. Support for this strategic consideration view is bolstered in that we find that the positive competition-conservatism relation largely holds only for industry followers, i.e., firms that are more susceptible to competitive pressures. The latter finding is inconsistent with the monopoly rent view, which predicts the relation to be stronger for industry leaders because they are more likely to reap monopoly rents when competition is less, thereby creating a greater error-in variable problem. Our evidence suggesting that strategic considerations shape the extent to which product market competition affects conditional conservatism is subject to the caveat that there may be other factors that are critical but missing in our analysis. While our results may not demonstrate direct causality, they shed light on how industry structure is related to financial reporting choices.

Footnotes
1

There are many reasons why anti-trust actions vary across time. One could be the philosophy of the administration in place. Alternatively, budget constraints and other regulatory issues may reduce the resources needed for anti-trust enforcement.

 
2

A competing view to this argument pertains to the presence of proprietary costs. This view also stresses competition among existing rivals (Verrecchia 1983). In the presence of proprietary costs, a firm may be less forthcoming with respect to firm-specific information out of concern that it will affect its competitive position. The upshot of this argument is that competitive pressures impose proprietary costs that may induce firms to recognize gains (losses) less (more) quickly. These effects may be more pronounced for industry followers because they are prone to greater competitive pressures than industry leaders.

 
3

Chhaochharia et al. (2009) provide additional evidence on the governance role of product market competition by examining its relation to formal governance mechanisms in place. They find firms in less competitive industries use more formal governance mechanisms such as having less anti-takeover provisions, greater pay-for-performance sensitivity, and greater managerial equity ownership. They find the converse to be true for firms in industries that face more intense product market competition.

 
4

Several studies have exploited this change in anti-takeover laws to examine its implication on financial reporting. Mehta (2010) finds financial reporting quality limits potential managerial excesses due to the passage of the anti-takeover laws. Other studies examine the impact of the adoption of the anti-takeover laws on financial reporting quality. For instance, Armstrong et al. (2012) find an improvement in financial statement informativeness following the passage of these laws. Other studies (e.g., Callen et al. 2010; Jayaraman and Shivakumar 2013) also find that conditional accounting conservatism increased significantly after the passage of state anti-takeover laws. Both studies argue that improvements in financial reporting serve to offset the potential weakening of corporate governance due to the stifling of the market for corporate control. In our context, the implication of this argument is that product market competition should reduce the demand for conservatism to the extent that it plays an effective governance role.

 
6

1977 is the first calendar year when segment data were available for all firms under SFAS No. 14, which became effective in 1976. We restrict our sample to the year 2005 because of the unavailability of the product market competition variable obtained from the Hoberg-Phillips data library for 2005 and thereafter.

 
7

Approximately 10 % of our firm-years’ segment SIC codes (three digit) did not match with the SIC codes in the Compustat annual file. To reduce the number of mismatches, we also matched the segment SIC codes with the SIC codes in the CRSP database and found even larger number of mismatches, which is consistent with Guenther and Rosman (1994), who find large differences between SIC codes assigned to companies by Compustat and CRSP and suggest using Compustat for SIC codes. As a result, we followed Li's (2010) approach of matching with the Compustat annual file instead of the CRSP database.

 
8

Ball et al. (2012) demonstrate that when firm-specific effects in earnings are taken into account, estimates of asymmetric timeliness do not exhibit the bias identified in Patatoukas and Thomas (2011) and are statistically and economically significant (though smaller in magnitude).

 
9

Specifically, they adopt a two-step method to create the fitted HHI. First, they regress the census HHI on three variables: the HHI calculated from Compustat data for those manufacturing companies, the average number of employees for each industry (including both public and private firms) using employee data from the Bureau of Labor Statistics, and the number of employees per firm for public companies in each industry using Compustat data. In addition, they also include interaction terms of each of the two employee size variables with the Compustat HHI. Then, they use the parameter estimates derived from the regression to estimate the fitted HHI for all industries.

 
10

While the Ali et al. (2009) and Hoberg and Phillips (2010) product market competition measure are improvements over the measure based on Compustat data, they are still incomplete in the sense that they do not include competition that can arise from not-for-profit and governmental entities or foreign firms. The omission of not-for-profit and governmental entities results in product market competition measures that may not completely reflect the extent of competition that exists in the marketplace. The measurement error is likely to be more severe for healthcare and education industries, in which non-profits dominate. To address this issue, we exclude healthcare and education industries from our sample and reestimate our regression models. Untabulated results indicate that the tenor of our results continue to hold after excluding these industries. To address the omission of foreign firms, we approximate competition from foreign firms by using import data by industry following Hui et al. (2011) and include it as an additional control variable in our empirical models; our inferences with respect to our test variable remain unchanged. However, we acknowledge that we still cannot completely rule out this measurement error and that it is important to recognize this limitation in interpreting our results.

 
11

Beaver and Ryan (2009) show that conservatism is overestimated in the presence of debt, suggesting an additional reason to control for leverage.

 
12

As a sensitivity test, we used the litigation risk metric developed by Kim and Skinner (2012), who supplement indicator variables for industry membership with measures of firm characteristics (such as size, growth, and stock volatility) to predict litigation risk. Untabulated results using Kim and Skinner’s (2012) measure yielded inferences with respect to our test variables that were similar to those reported in the paper.

 
13

Untabulated correlations for the sample based on Li’s (2010) methodology exhibit similar patterns and are not reported for brevity.

 
14

In terms of economic magnitude, when we multiply the standard deviation of 1.763 for PMC in the industry follower sample with the coefficient on NEG*RET*PMC (0.009), we get 0.016, which represents about a 2 % increase relative to the coefficient on NEG*RET of 0.763. Untabulated F-statistics to test for the difference in the incremental timely loss recognition coefficients for industry leaders and followers indicate a more pronounced effect for industry followers than leaders.

 
15

Under the political costs hypothesis, industry leaders might expect to be differentially scrutinized and therefore report more conservatively. Our results are consistent with the arguments in Basu (2005) that political costs/regulation are more likely to be circumvented by unconditional conservatism.

 
16

Our analysis focuses on the 10-year period surrounding the deregulation of the three industries covered in our study. We drop the year of deregulation from our sample and focus on the five years prior to and the five years after deregulation.

 
17

For the airline industry, the mean values of PMC2(Sale) are calculated based on three years before and after the deregulation year (1978) because the sample for our PMC2(Sale) measure is not available before 1975.

 
18

Analyses of the Compustat universe indicate that there is about a 10 % increase in firms with BIG N auditors during the same period.

 
19

Following Garcia Lara et al. (2009), we introduce the variable LEADER to capture political costs that create incentives for firms under regulators’ scrutiny to shift income to periods with a lower public visibility, inducing conditional conservatism.

 
20

We obtain the number of antitrust (government and private) case filings from Table 2 of Lin et al. (2000) with supplemental-years data provided by the Antitrust Division of the U.S. Department of Justice (Source: Maguire 2010; http://www.albany.edu/sourcebook/pdf/t5412010.pdf). We use the natural logarithm of the number of antitrust case filings each year to proxy for the level of antitrust enforcement. Because the antitrust data cases are not disaggregated into three-digit SIC codes, we use aggregated data only.

 
21

We include HIGH_LIT in the model to control for the effect of high litigation regimes because our antitrust variable may simply reflect different litigation regimes (Basu 1997; Ball et al. 2000; Holthausen and Watts 2001; Talley 2009).

 
22

We acknowledge that this finding could also be viewed as being consistent with political cost argument, in that more timely loss recognition during greater antitrust enforcement lowers political costs (Watts 2003).

 
23

One argument is that firms in concentrated industries may adopt conservative accounting methods to avoid regulatory scrutiny. This political cost argument implies that firms in concentrated industries, i.e., industries with limited product market competition, may adopt accounting methods that lead to lower reported earnings. We assess the validity of this claim by evaluating the relation between product market competition and unconditional conservatism in the form of accounting method choices. Principally, we examine whether firms in concentrated industries adopt conservative accounting methods in the form of accelerated depreciation and LIFO cost flow assumption. When we correlate these two accounting method choices with our product market competition variables, we find evidence supporting the argument that firms in concentrated industries tend to adopt accelerated depreciation methods (but not LIFO) that contribute to lower reported earnings.

 
24

One caveat for such categorical analysis is that the accelerated depreciation argument applies only when firms are growing and for LIFO when prices and inventory levels are not decreasing. Following Lev et al. (2005), we recode the two indicator variables to capture the depreciation and inventory cost flow-related accounting method choices under these specific scenarios. The first variable is set to be equal to 1 if a firm uses an accelerated depreciation method and when its depreciation expense growth exceeds its earnings growth and 0 otherwise. Similarly, the second variable takes a value of 1 if the firm adopts LIFO and when its growth in cost of goods sold is higher than its earnings growth and 0 otherwise. Untabulated results yield qualitatively similar inferences.

 

Acknowledgments

We gratefully acknowledge the thoughtful comments and suggestions of Lakshmanan Shivakumar (the editor), the referee, and John Wang. We also acknowledge the comments of workshop participants at the Florida International University, University of Technology-Sydney, University of Queensland, and Macquarie University.

Copyright information

© Springer Science+Business Media New York 2014