Does investment efficiency improve after the disclosure of
material weaknesses in internal control over financial
reporting?$
Mei Cheng a,Dan Dhaliwal a,b,Yuan Zhang c,n
a The University of Arizona,Tucson,AZ85721,United States
b Korea University Business School,Seoul136-701,Republi
c of Korea
c University of Texas at Dallas,Richardson,TX75080,Unite
d States
a r t i c l e i n f o
Article history:
Received23September2009
Received in revised form
1March2013
Accepted5March2013
Available online16March2013
JEL classifications:
G31
G38
M41
M48
Keywords:
Effectiveness of internal control over
financial reporting
Investment efficiency
Disclosure
a b s t r a c t
We provide more direct evidence on the causal relation between the quality of financial
reporting and investment efficiency.We examine the investment behavior of a sample of firms
that disclosed internal control weaknesses under the Sarbanes-Oxley Act.We find that prior to
the disclosure,these firms under-invest(over-invest)when they are financially constrained
(unconstrained).More importantly,we find that after the disclosure,these firms’investment
efficiency improves significantly.
&2013Elsevier B.V.All rights reserved.
1.Introduction
Prior literature shows that firms with a lower quality of financial reporting under-invest(over-invest)when they are financially constrained(unconstrained)(Biddle et al.,2009).These results are important because they suggest that the quality of a firm's financial reporting has an association with real investment efficiency.However,the literature does not establish a causal relation for this association.In this study,we provide more direct evidence for this causal relation by taking advantage of a provision in the Sarbanes-Oxley(SOX)Act that requires a firm to disclose if it has a material internal control weakness(ICW)in its financial reporting(U.S.Congress,2002).
Contents lists available at SciVerse ScienceDirect
journal homepage:https://www.doczj.com/doc/1315148163.html,/locate/jae
Journal of Accounting and Economics
0165-4101/$-see front matter&2013Elsevier B.V.All rights reserved.
https://www.doczj.com/doc/1315148163.html,/10.1016/j.jacceco.2013.03.001
☆We thank the editor,John Core,and an anonymous referee for their invaluable suggestions.We also thank seminar participants at Columbia University, Nanyang Technological University,the University of Illinois at Chicago,and the University of Illinois at Urbana-Champaign for helpful comments.All errors are our
own.
n Corresponding author.
E-mail addresses:meicheng@https://www.doczj.com/doc/1315148163.html,(M.Cheng),dhaliwal@https://www.doczj.com/doc/1315148163.html,(D.Dhaliwal),yuan.zhang2@https://www.doczj.com/doc/1315148163.html,(Y.Zhang).
Journal of Accounting and Economics56(2013)1–18
An ICW suggests that there is an information problem in the firm 's financial reporting system.Given this information problem and the findings from Biddle et al.(2009),we predict that firms that disclose ICWs (ICW firms)exhibit inefficient investment behavior prior to the disclosure.More importantly,because an ICW provides an adverse public signal,these firms are expected to address their past financial reporting problems subsequent to the disclosure.Thus,firms should show an increase in the quality of their financial reporting from the pre-disclosure period to the post-disclosure period.If the improvement in the quality of financial reporting increases investment efficiency,we predict that the pre-disclosure inefficiency in investment by the ICW firms will be mitigated or eliminated in the post-disclosure period.
We test these predictions by examining the investment behavior of a sample of ICW firms surrounding their first disclosure of ICWs.Following Biddle et al.(2009),we focus on the relation between the effectiveness of the internal control and investment levels conditional on a given firm 's likelihood of over-investing or under-investing.We start our analyses with a pooled sample of ICW firms and control firms with effective internal control.Regression analyses show that in the year prior to the first disclosure of an ICW,relative to control firms with similar financial conditions,financially constrained ICW firms under-invest by about 1.79%(2.89%)of total assets,while financially unconstrained ICW firms over-invest by about 2.53%(2.76%)of total assets based on the pooled sample (pooled sample of survivors).These numbers represent about 14–23%of average investment levels of the sample (which is about 12.80%of total assets),suggesting that the magnitudes of the effects are economically significant.
Most importantly,we find that after the initial disclosure of material weaknesses,the investment inefficiency of ICW firms becomes small and insignificant relative to control firms.Regression analyses based on both the pooled sample and the pooled sample of survivors show that in the second year after the disclosure,the investment levels of ICW firms are no longer significantly different from those of the control firms with similar financial conditions.Further statistical tests also formally confirm significant reductions in both over-investment and under-investment.
Following Armstrong et al.(2010),we also employ a propensity-score matching procedure to generate a different control sample.This procedure provides a control sample that has similar characteristics to the ICW firms,but different levels of internal control effectiveness and hence financial reporting quality.When we examine all matched firms,regression analyses support both sets of our hypotheses:(1)in the year prior to disclosure,ICW firms significantly under-invest (over-invest)when firms are financially constrained (unconstrained);and (2)after the disclosure,there are significant reductions in both under-investment and over-investment.Two-sample statistical tests that compare ICW firms and control firms within groups of firms with high ex ante likelihood of over-and under-investment respectively are largely consistent with the regression analyses except that we find little evidence of decreases in under-investment.We also focus on a matched sample of surviving ICW and control firms.The survivorship requirement ensures that the ICW firms remain constant in our event period,which makes it more sensible for inferring over-time changes in investment efficiency.Both two-sample tests of the differences and regression tests based on this sample provide support for both sets of our hypotheses.Taken together,these results suggest that SOX disclosures of ICWs and the changes that follow reduce investment inefficiency.
Our study contributes to several streams of literature.First,we contribute to the emerging literature on the relation between the quality of financial reporting and investment efficiency (Bens and Monahan,2004;Biddle and Hilary,2006;McNichols and Stubben,2008;Biddle et al.,2009;Francis and Martin,2010;Bushman et al.,2011).By examining the changes around disclosures of ICWs,we are able to provide more direct evidence of the causal relation between financial reporting quality and investment efficiency than research based on cross-sectional analyses (e.g.,Biddle et al.,2009;Francis and Martin,2010;Bushman et al.,2011).1
Second,we shed light on the debate regarding the costs and benefits of SOX and,in particular,of the increased disclosure requirement under Section 404.The popular press and practitioners both argued that the requirement to disclose ICWs under Section 404is burdensome to corporate shareholders as well as to corporate managers and might lead to the misallocation of corporate resources (American Bankers ’Association (ABA),2005;Charles River Associates,2005).In line with these concerns,Berger et al.(2005),Zhang (2007),and Li et al.(2008)document the costs of implementing the SOX requirements with regard to auditing and reporting on internal controls.However,other studies document the benefits of these requirements such as providing information to the executive labor market (Li et al.,2010),improving the quality of financial reporting (Altamuro and Beatty,2010),and reducing the cost of capital for firms (Ashbaugh-Skaife et al.,2009;Dhaliwal et al.,2011).We add to this debate by showing that the changes following ICW disclosures increase real investment efficiency.
The remainder of the study proceeds as follows.Section 2provides background on Sections 302and 404of the SOX and develops our empirical predictions.Section 3introduces our research design and describes the samples.Section 4presents our empirical results.Section 5concludes.
1
Hope and Thomas (2008)also provide a time-series analysis of the effects of disclosure on investment efficiency.There are at least two important differences between our study and theirs.First,we focus on investment efficiency in terms of both over-investment and under-investment,while their focus is on over-investment (“empire building ”)only.Second,we examine the effects of disclosing internal control weaknesses while they examine the effects of geographic earnings disclosure.
M.Cheng et al./Journal of Accounting and Economics 56(2013)1–18
2
M.Cheng et al./Journal of Accounting and Economics56(2013)1–183 2.Background and hypotheses
2.1.Background on Sections302and404of the Sarbanes Oxley Act
The Sarbanes Oxley Act(SOX)became effective on July29,2002.Prior to the act,firms were only required to publicly disclose deficiencies in their internal control if they changed auditors(Doyle et al.,2007a).With the enactment of SOX, public firms were required to assess and disclose the effectiveness of their internal control systems.Specifically,Section302 mandates that a firm's CEO and CFO certify in periodic(interim and annual)SEC filings that they have“evaluated and presented in the report their conclusions about the effectiveness of their internal controls based on their evaluation.”Section404further requires that each annual report contains an internal control report that includes an assessment of the effectiveness of the issuer's internal control structure and procedures with respect to financial reporting.2
2.2.Hypothesis development
Whether suboptimal investments take the form of over-investment or under-investment depends not only on the managers’incentives and the monitoring environment,but also on the availability of capital.Firms with abundant financial resources are more likely to over-invest,while firms with constrained financial resources are more likely to under-invest (Jensen,1986;Myers,1997).Therefore,following Biddle et al.(2009),our prediction of the effects of ICWs on firms’investment is conditional on the availability of financial resources to the firm.
2.2.1.ICW and investment levels
We identify the year in which a firm first discloses an ICW as Year T.Our first set of hypotheses examines investment efficiency in the year immediately prior to the first disclosure(i.e.,Year T?1).3Researchers have documented that the reporting of internal control weakness/effectiveness is associated with accrual quality(e.g.,Doyle et al.,2007b),and they have used ICW reporting as a proxy for poor financial reporting quality(e.g.,Costello and Wittenberg-Moerman,2011).Thus, in the year prior to the first disclosure of an ICW,the firm is expected to have poor financial reporting quality.
In a world without frictions(as described in Modigliani and Miller,1958),funds flow such that the marginal product of capital is equal across all projects in the economy,leading to an optimal investment level.However,the literature recognizes that frictions do exist in the economy and documents,both theoretically and empirically,that these frictions can lead to investment inefficiency.Among these frictions,perhaps the most pervasive and important ones are those that arise from information asymmetry(Stein,2003).Information asymmetry between managers and outside suppliers of capital can result in adverse selection and moral hazard,both of which can affect investment efficiency.We argue that a weak internal control system can increase information asymmetry and exacerbate both of these problems,leading to inefficient investment.
Under adverse selection,managers are better informed than outside investors as to the true value of the firm's assets and growth opportunities.Managers are thus likely to issue capital when their firm is overvalued.Given that ICW firms have lower financial reporting quality than control firms(Doyle et al.,2007b),information asymmetry is higher and managers have greater incentives to time the issuance of capital.If these timing strategies are successful,managers can then over-invest the proceeds from these capital issuances(Biddle et al.,2009).On the other hand,rational investors anticipate this tendency and are likely to increase the firms’cost of capital.Ashbaugh-Skaife et al.(2009)and Dhaliwal et al.(2011) document that ICW firms have a higher cost of capital than control firms.In this case,we expect that the increased cost of capital associated with ICWs leads financially constrained firms to under-invest,compared with equally financially constrained control firms,because these firms have more difficulty raising the capital needed to fund their projects.
Moral hazard,on the other hand,suggests that when managers’and investors’interests are not well aligned,managers have incentives to over-invest so as to maximize their personal welfare(Williamson,1974;Jensen,1986).Without effective monitoring and control,managers of ICW firms have greater opportunities to provide upward-biased information to the board or to shareholders when seeking support for their investment plans.For example,managers might over-state revenue or under-state cost in order to depict a high-growth trend or strong competitive advantage in segments they hope to expand.Thus,under moral hazard,we expect ICW firms without financial constraints to over-invest.However,if outside suppliers of capital are able to anticipate this problem and can ration capital ex ante,moral hazard might lead ICW firms with significant financial constraints to under-invest ex post(Stiglitz and Weiss,1981).
To summarize,we expect that,in the year prior to the first disclosure of the ICWs(i.e.,Year T?1),the ICW firms have poor information quality.Given this low information quality and the findings of the literature(Biddle et al.,2009),we expect an ICW firm to invest inefficiently.Following Biddle et al.(2009),we formulate our first set of hypotheses regarding the effects of an ICW,conditional on a given firm's underlying financial condition,as follows:
2Section404came into effect for firms with market values greater than or equal to$75million(accelerated filers)for fiscal years ending on or after November15,2004.For smaller public companies(non-accelerated filers),the SEC extended the effective date to fiscal years ending on or after December 15,2007.On September15,2010,the SEC issued rule33–9142that permanently exempts registrants that are non-accelerated filers from the Section404(b) requirement for an internal control audit.
3Our empirical tests assume that the disclosed ICW also exists in the year immediately prior to that disclosure.This assumption is consistent with Doyle et al.(2007a),Ashbaugh-Skaife et al.(2008),and Dhaliwal et al.(2011).
H1a.Financially constrained ICW firms are more likely to under-invest in the year prior to the disclosures of their ICWs.H1b.Financially unconstrained ICW firms are more likely to over-invest in the year prior to the disclosures of their ICWs.
2.2.2.Disclosure of ICW and investment levels
Our second,and more innovative,set of hypotheses relates to the changes in a firm 's investment efficiency around their first disclosure of an ICW.We predict that disclosures of ICWs lead the disclosing firms to undergo important changes.These changes are expected to improve financial reporting quality and mitigate the agency problems of adverse selection and moral hazard,which,in turn,increases investment efficiency.
First,we expect a firm 's disclosure of an ICW to mitigate adverse selection.The disclosure of an ICW provides a signal to the board,shareholders,and other stakeholders that the firm has low information quality.The ICW firm is likely to improve its internal control systems over financial reporting subsequent to the disclosure,which should increase the quality of its financial information (Nicolaisen,2004).Consistent with this view,Altamuro and Beatty (2010)find that the requirements to report internal control effectiveness decrease earnings management and increase the validity of the loan-loss provision,persistence of earnings,and the predictability of cash flow.Furthermore,once investors and the board of directors recognize their internal control system is weak,they are likely to demand more and higher-quality disclosures from the managers (Feng et al.,2009).Thus,with both the enhanced disclosures and the improved quality of information,the extent of information asymmetry and hence adverse selection is expected to decrease.As a result,securities are less likely to be overpriced and managers are less likely to time the issuance of capital,which,in turn,decreases the amount of over-investment.On the other hand,in anticipation of these changes,investors are likely to reduce the cost of capital they demand from these firms (Lambert et al.,2007).This lower cost decreases the amount of under-investment because managers have better access to funding when there are good investment opportunities (Myers and Majluf,1984).
Second,the ICW disclosure should induce the board of directors,as well as market intermediaries such as credit rating agencies and financial analysts,to increase their level of monitoring,which,in turn,reduces the problem of moral hazard.Importantly,these monitoring entities are likely to increase their scrutiny not only of the financial information managers provide but also of the operating,investing,and financing decisions managers make.For example,the board of directors might seek more independent information,might crosscheck financial information,and might challenge managerial proposals more than they previously had.This enhanced scrutiny helps to reduce the amount of bias or the number of errors in financial reporting and to reduce possible empire-building investment activities.The possibility also exists that managers,now aware of the enhanced scrutiny,will make fewer investment decisions that are not aligned with investors ’interests.As investors anticipate this decrease in managers ’incentives to over-invest,their tendency to ration capital ex ante will also decrease,which will lead to less under-investment ex post in ICW firms with financial constraints.
To summarize,the identification and disclosure of an ICW is expected to mitigate the problems of adverse selection and moral hazard.Thus,we use an ICW disclosure as an instrument for an improvement in the quality of financial reporting and examine whether investment efficiency improves following the disclosure of ICWs.This setting helps us shed light on the causal relation between financial reporting quality and investment efficiency and leads to our second set of hypotheses:H2a.Financially constrained ICW firms under-invest less in the years following the disclosures of their ICWs.H2b.Financially unconstrained ICW firms over-invest less in the years following the disclosures of their ICWs.
It is important to note that while a firm 's disclosure of the material weaknesses might improve investment efficiency,the timing and extent of that improvement are not clear ex ante.For example,the board of directors and outside shareholders might not effectively detect and stop all over-investment or under-investment activities immediately,in which case ICW firms can still over-invest or under-invest,although to a lesser extent.3.Research design and sample selection 3.1.Research design
We use two different procedures to obtain control samples to compare with the ICW sample.The first procedure obtains control samples based on all firms that make no ICW disclosures around the ICW firms ’first disclosure of an ICW.We refer to this comparison as the pooled sample analysis.Our second method is based on a propensity-score matching procedure.We refer to this comparison as the matched sample analysis.
3.1.1.Pooled sample analysis
To test our two sets of hypotheses for the pooled sample,we examine investment efficiency for Years T ?1,T t1,and T t2separately.We cluster the standard errors at both the firm and year levels to obtain standard errors that are robust to heteroskedasticity,serial correlation,and cross-sectional correlation (Petersen,2009;Gow et al.,2010).Specifically,our
M.Cheng et al./Journal of Accounting and Economics 56(2013)1–18
4
M.Cheng et al./Journal of Accounting and Economics56(2013)1–185 regression model is as follows(firm subscripts are suppressed):
Investment t?a0ta1?Weakta2?Weak?OverFirm t?1ta3?OverFirm t?1t∑b i?WeakDetermin ant i,t?1
t∑c i?INVDetermin ant i,t?1t∑d1i GOV i,t?1t∑d2i GOV i,t?1?OverFirm t?1te t:e1TThe dependent variable Investment is the total investment measured as the sum of research and development,capital, and acquisition expenditures less the sale of property,plant,and equipment multiplied by100and scaled by the lagged total assets.Weak is an indicator variable,which is set to one for firms that disclosed an ICW and zero for the control firms.
Consistent with Biddle et al.(2009),we measure Investment in Year t,and our control variables at the end of Year t?1. As our hypotheses are conditional on the respective ex ante likelihoods of over-investment and under-investment(Opler et al.,1999;Biddle et al.,2009),we use a variable OverFirm to distinguish between situations in which a given firm is more likely to over-invest or under-invest.To construct OverFirm,we follow prior studies that suggest cash-rich and low-leverage firms are more likely to over-invest,and rank each of our sample firms’cash balances and negative leverage at the end of Year t?1into two decile ranks.We then average these two decile ranks and scale the average so that it ranges from zero to one.By focusing our analysis on firms that are prone to specific forms of suboptimal investment,we increase the power of our tests.
To test hypotheses H1a and H1b,we estimate Model(1)for Year T?1and focus on the indicator variable Weak and its interaction with OverFirm.Before the firms’initial disclosures of ICWs,if OverFirm equals zero,then firms are financially constrained and thus,ex ante,are more likely to under-invest.Under this scenario,if constrained ICW firms indeed under-invest more than control firms do as predicted in H1a,then the coefficient on Weak(i.e.,a1)is expected to be negative.On the other hand,if OverFirm equals one,then firms are financially unconstrained and thus,ex ante,are more likely to over-invest.Under this scenario,if unconstrained ICW firms over-invest more than control firms do as predicted in H1b,then the sum of the coefficients on Weak and Weak?OverFirm(i.e.,a1ta2)is expected to be positive.
In our second set of hypotheses,we focus on Years Tt1and Tt2instead of Year T itself.We choose this focus because how soon after the disclosure the board of directors and other stakeholders of the firm become aware of and respond to the ICWs is unclear.If a firm's disclosure of an ICW leads to elimination(reduction)of the under-investment and over-investment,then a1and a1ta2should be insignificant(of lower magnitude).
The control variables we include in Model(1)be categorized into one of three groups:(1)determinants of material ICW(WeakDeterminant);(2)determinants of investment level(INVDeterminant);and(3)other governance mechanisms (GOV).We base our first group of control variables on Doyle et al.(2007a),who find that smaller(LogAsset),4younger(Age), financially weaker(Losses,Z-score),and more complex(Nseg,Foreign)firms,as well as firms that are growing more rapidly (Extrgrow)or are undergoing restructuring(Rest),are more likely to have a material ICW.
We base our second and third groups of control variables on Biddle et al.(2009),who examine the cross-sectional relation between the quality of financial reporting and investment efficiency.Our determinants for investment levels comprise Tobin's Q(Q),cash flow(CFOsale),the standard deviation of cash flow(s(CFO)),the standard deviation of sales(s(Sales)),the standard deviation of investments(s(Investment)),the market-to-book ratio(Mkt-to-book), tangibility(Tangibility),market leverage(Leverage),industry market leverage(Ind Leverage),dividends(Dividend), operating cycle(OperatingCycle),firm age(Age),financial bankruptcy risk(Z-score),firm size(LogAsset),and a loss indicator(Loss).
Our third group of control variables captures other monitoring or governance mechanisms that could affect investment efficiency,and comprises the corporate governance index(G-score),5institutional holdings(Institutions),and analyst coverage(Analysts)(Jensen and Meckling,1976;Shleifer and Vishny,1997;Bhojraj and Sengupta,2003).More importantly, we also include the main proxy for the quality of financial reporting used in Biddle et al.(2009),namely,accrual quality (AQ).6,7The addition of these variables ensures that our findings are robust to their effects.To control for the differential effects of these variables on over-investment and under-investment,we include in the model these variables(i.e.,our monitoring/governance variables,particularly the accrual quality variable)as well as their interactions with OverFirm.All variables are defined in detail in the appendix.Further,consistent with Biddle et al.(2009),in addition to the above control variables,we add industry fixed effects using the Fama-French(1997)48-industry classification to control for industry-specific effects on investments.
4Doyle et al.(2007a)use Logmv to capture firm size.However,Logmv is highly correlated with LogAsset,which is a control variable in Biddle et al. (2009).We use LogAsset in our models.Replacing it with Logmv does not affect the inferences of this paper.
5To avoid a significant reduction in our sample size due to a missing G-score,we follow Biddle et al.(2009)and set the G-score to zero if missing and add an indicator variable for the availability of the corporate governance index(G-score dummy).
6Following Biddle et al.(2009),our AQ measure is measured as of Year t?2instead of Year t?1because of the requirement for one leading year of data in calculating AQ.
7We choose the accrual quality measure based on Dechow and Dichev(2002)as it is consistently the most significant measure in Biddle et al.(2009). Nonetheless,in additional robustness tests we use two alternative measures of earnings quality:a measure of discretionary revenue used in McNichols and Stubben(2008)and a dummy variable that indicates whether the firm-year's financial statements are subsequently restated according to AuditAnalytics data.Our results are generally robust to these alternative measures.
3.1.2.Propensity-score matched sample analysis
For our matched sample analysis,we follow Armstrong et al.(2010)and adopt the method of propensity-score matching to more effectively control for differences in relevant dimensions between the treatment and control samples.We attempt to match each ICW firm with a control firm that is similar across all observable relevant variables.Specifically,we include in the first stage of our regression all determinants of ICW and investment levels,and our governance variables.We estimate the following model for the first stage (firm subscripts are suppressed)by year:
Weak ?a 0ta 1?OverFirm t ?1t∑b i ?WeakDetermin ant i ,t ?1t∑c i ?INVDetermin ant i ,t ?1t∑d 1i GOV i ,t ?1te t :
e2T
The variables used in Model (2)are defined as in Model (1).We also add industry-fixed effects to Model (2)to account for industry-specific factors.We obtain the propensity score for each firm-year as the predicted value in Model (2).We then match each treatment firm (with no replacement)with the control firm that has the closest score in the same year within a distance of 0.01from the treatment firm 's propensity score.If the propensity-score matching is successful,then each ICW firm and its matching control firm are similar in all observable dimensions (including OverFirm ),with the exception of the effectiveness of internal control over financial reporting.Accordingly,in the second stage of our analyses,we compare the investment levels between the ICW firms and the control firms conditional on the ex ante likelihood of over-and under-investing separately to test H1a and H1b .We also examine the changes of the investment differences between ICW firms and control firms from Year T ?1to Year T t2to test H2a and H2b .In addition,we estimate the following model to test our hypotheses:
Investment t ?a 0ta 1?Weak ta 2?Weak ?OverFirm t ?1ta 3?OverFirm t ?1te t
e3T
We include OverFirm and its interaction with Weak because our identification of investment inefficiency is conditional on the ex ante likelihood that a given firm over-invests or under-invests.Similar to the design of the pooled sample,we expect a 1(a 1ta 2)to be significantly negative (positive)for the under-investment (over-investment)prediction in Year T ?1under H1a and H1b ,but insignificant in Years T t1and T t2under H2a and H2b .3.2.Sample selection
Following recent academic studies (Ogneva et al.,2007;Li et al.,2010;Costello and Wittenberg-Moerman,2011),we obtain information on firms ’disclosures of their internal control from the AuditAnalytics database.This database keeps track of SEC filings in interim and annual reports for both Sections 302and 404disclosures.While there are some differences between the types of disclosures Sections 302and 404require,many firms integrate the two procedures and,hence,reach similar assessments for both.Thus,following the literature (Doyle et al.,2007a,2007b;Ashbaugh-Skaife et al.,2008),we do not differentiate between the assessments of effectiveness under Section 302and those under Section 404.We restrict the disclosures to those occurring between 2004and 2007because disclosures occurring between 2002and 2003are scarce.8
Under SOX,internal control deficiencies can take the form of a significant deficiency or a material weakness,with the latter considered as the more severe one (Securities and Exchange Commission (SEC),2003).Consistent with Doyle et al.(2007a),we focus on firms ’disclosures of material ICWs in order to identify more severe deficiencies in internal control over financial reporting and provide greater power for our tests.Accordingly,we code each firm-year as containing material ICWs if (1)AuditAnalytics 302data identify the disclosure of material ICWs in any of its quarterly or annual reports;or (2)AuditAnalytics 404data identify the disclosure of ineffective internal control during that fiscal year.We focus on a firm 's first disclosure of an ICW only.We exclude disclosures made through Form 20-F or Form 10-K/QSB,because foreign or small firms have limited financial data available for our analyses.After merging this data with Compustat data using CIK identifiers,we obtain an initial sample of 1696firms that disclosed material ICWs for the first time between 2004and 2007(i.e.,Year T )via Form 10-K or Form 10-Q.
We further require the availability of the investment variable and various control variables used in Model (1).We obtain information on these variables from Compustat (for financial information and earnings quality variables),CRSP (for firm age),RiskMetrics (for corporate monitoring/governance variables),IBES (for analyst coverage),and Thomson Reuters (for institutional holding).Requiring these variables substantially reduces our sample of ICW firms to 545,439,and 388for Years T ?1,T t1,and T t2respectively.
Our main analyses also require a control sample of firms with effective internal control.We use two alternative sample selection procedures.The first procedure obtains all firms covered by AuditAnalytics that (1)disclosed no ICWs between 2004and 2007and (2)have information available for all variables used in our analyses.This leads to 4999,4050,and 3145control firms for the Years T ?1,T t1,and Year T t2respectively.This sample is larger than the matched sample described below and hence offers greater testing power.
8
The Securities and Exchange Commission (SEC)(2003)also requires auditors to provide an independent opinion on the effectiveness of the internal control system.Analyses of the AuditAnalytics data show that auditors ’assessments are predominantly consistent with managers ’assessments.For our sample,there are only four cases wherein these two opinions are inconsistent with each other.
M.Cheng et al./Journal of Accounting and Economics 56(2013)1–18
6
M.Cheng et al./Journal of Accounting and Economics56(2013)1–187
Table1
Descriptive statistics based on the pooled sample partitioned by Weak in Year T?1.
Variable ICW firms Control firms
N Mean Median N Mean Median
Investment54513.569.73499912.729.37 Weak54511499900 OverFirm5450.580.6149990.550.56 Institutions5450.480.4949990.520.55 Analysts545 1.09 1.14999 1.27 1.39
G-score545 3.5404999 4.60
G-score dummy5450.41049990.490
AQ545?0.06?0.054999?0.05?0.04 Mkt-to-book545 1.94 1.554999 2.04 1.66 LogAsset545 5.57 5.634999 6.11 6.13
s(CFO)5450.080.0649990.070.05
s(Sales)5450.20.1549990.160.12
s(Investment)5450.090.0649990.080.05
Z-score545 4.48 3.444999 5.09 3.87 Tangibility5450.230.1749990.260.19 Leverage5450.130.0549990.130.07
Ind Leverage5450.140.149990.160.12 CFOsale5450.050.0749990.060.09 Dividend5450.28049990.440 OperatingCycle545 4.72 4.814999 4.62 4.69 Losses5450.31049990.190
Nseg545 1.5814999 1.71
Q545 1.73 1.424999 1.95 1.58 Rest5450.01049990.010 Foreign5450.29049990.270 Extrgrow5450.17049990.140
Age545 2.6 2.564999 2.79 2.71
The table presents the descriptive statistics for the variables used in the pooled sample analyses separately for ICW and control firms during the year before the initial disclosures on the effectiveness of the internal control over financial reporting(i.e.,Year T?1).See the appendix for variable definitions.
Following the literature(e.g.,Armstrong et al.,2010;McInnis and Collins,2011),our second sample selection procedure is the propensity-score matching based on Model(2).This matched sample effectively controls for various differences between the ICW firms and the control firms,especially when the underlying specification is non-linear(Armstrong et al., 2010).We obtain matching control firms for every event year and get502,424,and374matched pairs in Years T?1,Tt1and Tt2respectively(“matched sample”).By matching in each event year,this method takes into account the changing nature of firm characteristics,especially the ex ante likelihood of over-or under-investment.
It is notable,however,that our sample sizes for the pooled and the matched samples decrease over the event years.We also examine a constant set of firms that survived through Year Tt2within each of these two samples and label them “pooled sample of survivors”and“matched sample of survivors”respectively.Including only surviving firms is likely to be subject to survivorship bias,but makes the analyses of over-time changes more meaningful because the sample stays constant.In the pooled sample of survivors,there are374ICW firms and3130control firms.To obtain the matched sample of survivors,we form propensity-score matched pairs from the pooled sample of survivors in Years T?1,Tt1and Tt2 respectively.If an ICW firm is successfully matched to a control firm in each of the three years,we retain the matched pairs. Thus,the ICW firms are constant throughout the event years.9These requirements yield282matched pairs in each event year for our matched sample of survivors.10
3.3.Descriptive statistics
In Table1,we present the descriptive statistics of our pooled sample separately for ICW and control firms.For the sake of brevity,we present the statistics for Year T?1only.As discussed above,for Year T?1,we have545ICW firms and4999
9Since we perform the match in Years T?1,Tt1,and Tt2separately,the matching control firms are not constant over the three years.Another possibility is to obtain the matched sample from the surviving firms,but only match once in Year T?1and then keep the matched pairs constant over time. When we use this approach,our results are generally similar to the findings based on the matched sample of survivors with the exception of insignificant difference-in-difference test results for under-investment.We do not use this approach in our main tests because the ICW firms and the control firms might differ in firm characteristics including OverFirm in Years Tt1and Tt2,given that the match is only made in Year T?1.
10The matched sample of survivors is smaller than the matched sample because the matched sample is obtained from the pooled sample,while the matched sample of survivors is obtained from the much smaller pooled sample of survivors.In addition,because we impose a maximum caliper of0.01for the distance between the propensity scores of an ICW firm and its matching firm in the same year,a surviving ICW firm may not have a matching surviving firm in all three years,which further restricts the sample size of the matched sample of survivors.
M.Cheng et al./Journal of Accounting and Economics56(2013)1–189 control firms,which respectively account for10%and90%of the pooled sample.The mean(median)of total investments is 13.56%(9.73%)of the lagged total assets for ICW firms and12.72%(9.37%)of the lagged total assets for control firms.These numbers are very similar to prior studies(e.g.,Biddle et al.,2009).The descriptive statistics of other control variables are also generally in line with these studies,which suggests that our sample is representative.
Panels A,B,and C in Table2present average Investment and OverFirm for ICW and control firms with different ex ante likelihoods of over-investing/under-investing in Years T?1,Tt1,and Tt2respectively.We partition the firms into ex ante over-investment and ex ante under-investment groups based on the medians of OverFirm in Year T?1and track these groups over the years.We report the statistics on both the pooled sample and the pooled sample of survivors,as we discuss in Section 3.2.In Panel A for Year T?1,for both samples,the average investment levels for ICW firms are higher in the over-investment group than in the under-investment group.A similar pattern also holds for control firms.Regarding the level of investment across ICW and control firms,in the over-investment group,for the pooled sample,ICW firms have significantly higher average levels of investment(based on t-tests)than control firms,which supports H1b that financially unconstrained ICW firms tend to over-invest.However,in the under-investment group,the investment levels of these two groups of firms are similar,which leaves H1a unsupported.For the pooled sample of survivors,the differences in investment levels between the ICW firms and control firms are small and insignificant in both the over-and under-investment groups.
In Panel B for Year Tt1,we continue to find that the investment levels are higher for the over-investment group than for the under-investment group for both samples.However,within the over-investment group,ICW firms no longer have higher levels of investment than control firms.Within the under-investment group,ICW firms have significantly lower levels of investment than control firms.When we move to Panel C for Year Tt2,we no longer observe clear systematic differences between the over-investment and under-investment groups.The investment level is no longer higher for the ICW firms than for the control firms in the over-investment group.There is no statistically significant difference between the two groups of firms in the under-investment group,either.Overall,these results are similar between the pooled sample and the pooled sample of survivors.They are consistent with the ICW disclosure mitigating over-investment,but inconsistent with this disclosure mitigating under-investment.We caution readers that these comparisons are based on simple univariate analyses.In Section4we employ regression analyses and further control for various other determinants of investment levels.
4.Empirical results
4.1.Testing conditional investment inefficiency in year T?1based on the pooled sample
Table3presents the results from estimating Model(1)based on the two pooled samples for Years T?1,Tt1,and Tt2 separately.11The dependent variable is Investment.Panel A,Column A is based on the pooled sample in Year T?1.The coefficient on Weak is?1.79and significantly negative.This suggests that when the ex ante likelihood of under-investment is high,ICW firms invest significantly less than control firms by1.79%of total assets,representing about14%of the average investment levels for the sample firms.These results lend strong support to H1a.
The coefficient on the interaction term between Weak and OverFirm is significantly positive.As discussed in the previous section,we are more interested in the sum of the coefficients on Weak and Weak*OverFirm that captures the extent to which ICW firms over-invest relative to control firms.Our estimate shows that the sum of these two terms is both statistically and economically significant in Year T?1.Specifically,ICW firms over-invest by2.53%of total assets in the pooled sample, representing about19%of the average investment level of the sample.The results lend support to H1b that when the ex ante likelihood of over-investment is high,ICW firms invest significantly more than control firms in the year prior to the disclosure.
Of the investment determinant variables and the ICW determinant variables,we find that total investment levels have a positive association with cash flow volatility,tangibility,Tobin's Q,extreme growth and have a negative association with firm size,sales volatility,industry leverage,cash flow,dividend payouts,and firm age.The effects of accrual quality and its interaction term with OverFirm are significantly positive and negative,respectively,which are consistent with the findings in Biddle et al.(2009).Of the other mechanisms that might affect investment activities,both institutional holding and its interaction term with OverFirm have significant coefficients in the pooled sample.These coefficients suggest that a higher percentage of institutional holding is an effective mechanism for reducing under-investment and over-investment.
In Column B of Table3,we present the results that are based on the pooled sample of survivors,as discussed in Section 3.2.This sample addresses the concern that the insignificant results we present for Year Tt2in Column A are due to its reduced sample size and/or sample attrition relative to Year T?1.The fact that the ICW and control firms remain constant through the event time makes the analyses of changes in investment inefficiency more sensible.For Year T?1in Panel A,
11As discussed in Section2,we focus on Year Tt1and Year Tt2because we expect that it will take some time after the disclosure of ICW for the investment inefficiency to be corrected.Further,it is also possible that some investment decisions have been made before the disclosure of the weaknesses in Year T.In untabulated analyses,we examine the investment efficiency for Year T(i.e.,the year of the disclosure).The results do not show any observable correction of investment inefficiency.
Table 3
Internal control weakness and total investment over the three years based on the pooled sample.
Variable
Investment inefficiency test
Predicted sign
Pooled sample (A)Pooled sample of survivors (B)Coefficient
t -Value
Coefficient
t -Value
Panel A:the year prior to the disclosure of internal control weakness Intercept 9.25nn 3.8910.38nnn 5.57Weak (1)T ?1:under-invest ??1.79nn ?2.61?2.89n
?2.12OverFirm
?1.09?0.51?0.290.39Weak ?OverFirm (2)t 4.32nn 2.85 5.65n 2.04(1)t(2)
T ?1:over ?invest
t 2.53nn 2.91 2.76nn
3.20Monitoring or governance variables Institutions
t9.41nnn 6.338.35nnn 5.91Institutions ?OverFirm ??10.50nn
?3.36?9.63nn ?2.93Analysts
t0.500.930.490.95Analysts ?OverFirm ??0.06?0.060.600.97G -score
??0.12?0.46?0.09?0.31G-score ?OverFirm t0.44 1.080.430.88G-score dummy
t 1.120.63 1.420.60G-score Dummy ?OverFirm ??3.94?1.14?5.63?1.18AQ
t28.11nn 3.4035.11nn 2.67AQ ?OverFirm
??32.88nn ?2.54?27.99n ?1.73Determinants of investments Mkt-to-book t0.020.09?0.23?0.72LogAsset ?
?0.52nn ?2.37?0.80nnn ?5.39s (CFO )12.18nn 3.978.91nn 2.29s (Sales )
?4.76nnn
?5.07?6.32nnn ?5.14s (Investment ) 3.64 2.288.78nn 4.34Z-score ??0.06?1.58?0.07n ?1.87Tangibility t
7.81nnn 8.779.04nnn 6.39Leverage 0.480.21 2.74 1.42Ind Leverage ?12.65nn ?4.49?13.32nnn
?5.40CFOsale ?3.66nn ?3.77?1.45?0.47Dividend
??1.17nn ?2.76?1.04nn ?2.87OperatingCycle ?0.25?0.64?0.09?0.26Losses ?0.030.030.630.49Q t 2.25nnn 10.76 2.86nnn 9.30Age
?
?0.94nnn ?4.81?0.96nn ?2.76Determinants of reporting ICW Nseg 0.24 1.270.23 1.51Rest ?23.86?2.19?45.83nnn
?5.47Foreign ?0.14?0.31?0.00?0.36Extrgrow 6.01nnn 21.31
4.57nnn
14.66
Industry FE
Yes Yes Firm/Year Cluster Yes Yes N
55443504R 2(%)
23.93
25.26
Panel B:the year after the disclosure of internal control weakness Intercept 12.87nnn 15.0111.24nnn 14.39Weak (1)T t1:under-invest ??3.37n ?2.460.270.15OverFirm
3.99nn 3.80?1.00?0.85Weak ?OverFirm (2)? 1.740.59?1.23?0.62(1)t(2)
T t1:over-invest
?
?1.63?0.99
?0.96n ?2.03
Controls Included Yes Yes Industry FE
Yes Yes Firm/Year Cluster Yes Yes N
44893504R 2(%)
22.18
20.93
Panel C:two years after the disclosure of internal control weakness Intercept 16.84nnn
9.6914.87nnn 9.17Weak (1)T t2:under-invest ?0.130.04?0.55?0.16OverFirm
?0.60?0.29?4.28?1.64Weak n OverFirm (2)??1.91?0.47?1.66?0.33(1)t(2)
T t2:over-invest
?
?1.78?1.59
?2.21?1.17
Controls Included
Yes
Yes
M.Cheng et al./Journal of Accounting and Economics 56(2013)1–18
10
M.Cheng et al./Journal of Accounting and Economics56(2013)1–1811 Table3(continued)
Variable Investment inefficiency test Predicted sign Pooled sample(A)Pooled sample of survivors(B)
Coefficient t-Value Coefficient t-Value
Industry FE Yes Yes
Firm/Year Cluster Yes Yes
N35333504
R2(%)20.2219.45
Test difference for under-investment between Year T?1and Year Tt2t 1.92n 1.58 2.34n 1.69
Test difference for over-investment between Year T?1and Year Tt2??4.31nn?2.07?4.97n?1.65
Panels A,B,and C of the table present the regression estimates for the model that examines the relation between internal control weakness(Weak) and total investment(Investment)during Year T?1,Year Tt1,and Year Tt2respectively.Column A is based on the sample with data available in the corresponding year.Column B is based on the sample with the same ICW and control firms in all three years.In both regressions,the model comprises industry fixed-effects based on the Fama-French(1997)48-industry classifications.t-Statistics are corrected for cross-sectional and time-series correlation using a two-way cluster at the firm and year level.The nnn,nn,and n denote statistical significance at1%,5%,and10%levels,respectively,based on one-tailed tests if the coefficient has a predicted sign and two-tailed tests otherwise.See the appendix for variable definitions.
similar to Column A,we find a significant,negative coefficient on Weak(?2.89)and a significant,positive sum of coefficients on Weak and OverFirm?Weak(2.76),again supporting H1a and H1b.
Overall,these results document that material ICWs adversely affect investment efficiency prior to their disclosure.These adverse effects take the form of either over-investment or under-investment,depending on the given firm's available financial resources,and are significant both economically and statistically.The effects of ICWs on investment inefficiency are also robust to the effects of accrual quality and various corporate governance mechanisms.
4.2.Testing changes in investment inefficiency after the ICW disclosure based on the pooled sample
We now turn to our main tests for the effects that the disclosure of ICW and the following changes have on investment efficiency.We re-estimate Model(1)for Years Tt1and Tt2and compare how they differ from the results for Year T?1.We report the results in Panels B and C of Table3,respectively,with the control variables omitted for brevity.In Column A of Panel B,the coefficient on Weak is significantly negative,suggesting that ICW firms continue to invest significantly less in Year Tt1when the ex ante likelihood of under-investment is high.The magnitude of the coefficient,at?3.37,remains economically significant.This finding suggests that these firms underinvest by22%on a relative basis given that the average investment is about15%of total assets during Year Tt1.The coefficient on Weak,however,is insignificant in Column B of Panel B.In both Columns A and B of Panel C for Year Tt2,the coefficient on Weak becomes insignificant and small in magnitude(0.13and?0.55respectively),which indicates that the under-investment has disappeared by Year Tt2.
As for over-investment,the sum of the coefficients on Weak and OverFirm?Weak in Year Tt1are negative in both columns in Panel B,inconsistent with over-investment.It is insignificant(marginally significant)based on the pooled sample(pooled sample of survivors).In both Columns of Panel C,the sum of the coefficients on Weak and OverFirm?Weak are also insignificant in Year Tt2.These results suggest that the over-investment among ICW firms is corrected by the second year after the initial disclosure.
To formally test whether the investment differences between ICW and control firms change from Year T?1to Year Tt2, we perform statistical tests on the change in over-investment(under-investment)for firms with high ex ante likelihoods of over-investing(under-investing)in Panel C.Specifically,we stack the observations of the three years(i.e.,Year T?1,Year Tt1,and Year Tt2)and compare the coefficient on Weak between Year T?1and Year Tt2to assess the changes in under-investment activities.12To assess the changes in over-investment activities,we compare the sum of the coefficients on Weak and OverFirm?Weak between Year T?1and Year Tt2.These tests confirm the statistical significance(at the0.10level or better)of these changes in cross-sectional differences,thereby providing additional support for our second set of hypotheses that the disclosure of ICW leads to decreases in both over-investment and under-investment among the ICW firms.
Overall,Table3provides robust evidence that the investment inefficiency in ICW firms during Year T?1disappears after their disclosure of this weakness,supporting both H1a,H1b,H2a,and H2b.It is also interesting to note that in our pooled sample analysis,over-investment completely disappears by the first year(Tt1)after the initial disclosure,while under-investment does not become statistically insignificant until the second year.Perhaps mitigating over-investment is easier than mitigating under-investment because the latter often involves raising additional capital.
12Our inferences for H1a,H1b,H2a,and H2b remain the same in the stacked model.Specifically,in Year T?1,there is significant evidence of over-investment and under-investment,while by Year Tt2both forms of investment inefficiency have disappeared.
4.3.The propensity-score matched sample results
Table 4presents information on the logit model we use for our propensity-score matching procedure.For brevity,we discuss and present the estimation for Year T ?1only.Because there are four years of data for Year T ?1(i.e.,2003–2006),we estimate the logit model by year.We report the average coefficients and aggregate z -statistics of the four by-year logit regression estimates of Model (2)based on the pooled sample (Panel A)and the pooled sample of survivors (Panel B)respectively.In the pooled sample,our results are generally consistent with prior literature and show that restructuring charges (Rest ),foreign income (Foreign ),and extreme growth (Extrgrow )have a positive association with the probability that a given firm reports internal control weaknesses.On the other hand,firm size (Logat )has a negative association with reporting ICW.The logit models have average pseudo-R -squared of 17%and 25%respectively for the two samples,suggesting good fit of the model.This procedure of estimating the propensity-score by year is repeated for Years T t1and T t2for both the pooled sample and the pooled sample of survivors.
As discussed in Section 3.1,we form our matched sample (matched sample of survivors)based on the propensity score generated from the first-stage estimate using the pooled sample (pooled sample of survivors).Within each event year (i.e.,Years T ?1,T t1,and T t2respectively),for each ICW firm we find one control firm that has the closest propensity score in the same year within a caliper of 0.01,with no replacements.This procedure generates 502,424and 374matched pairs in Years T ?1,T t1,and T t2respectively for the matched sample.For the matched sample of survivors,in addition to using the pooled sample of survivors to estimate the propensity score and form the matched pairs,we further require that a surviving ICW firm to have a matching surviving firm in all three years.Thus,the matched sample of survivors is smaller (282pairs in each of the three event years).
Table 4
Logit model stimates of the first-stage propensity-score matching procedure in Year T ?1.
Variable
Predicted sign
Dependent variable ?Weak
Pooled sample (A)
Pooled sample of survivors (B)
Average coefficient
Aggregate z -statistic
Average coefficient Aggregate z -statistic
Intercept
?2.9606 3.46?8.0743 2.44Monitoring or governance variables Institutions ?0.0322 1.78?1.2404 3.04Analysts 0.00010.480.059 1.16G-score
?0.0987 2.98?0.0709 1.78G-score dummy 0.5888 1.940.8004 1.49AQ
0.8952 2.317.2939 3.53Determinants of reporting ICW Logat ??0.1188 2.160.1869 1.82Nseg t?0.0791 1.15?0.4715 2.16Age ?0.0243 1.73?0.3479 1.66Losses t?0.3461 2.52?0.9623 2.4Rest t 4.4087 2.1123.1496 1.98Foreign t0.09930.950.5405 2.12Extrgrow t0.0375 2.50.4341 1.9Z-score t?0.0496 2.22?0.0167 1.38Determinants of investments OverFirm 0.8187 2.410.15170.7Mkt-to-book ?0.1227 2.61?0.03090.67s (CFO )?0.20530.799.903 3.21s (Sales )
1.2395 3.97
2.5997
3.47s (Investment )?0.2065 1.76?0.9633 1.51Tangibility ?0.11140.880.8006 1.88Leverage 1.6327 2.86 2.2467 1.28Ind Leverage ?2.7637 3.04?6.7467 2.39CFOsale 0.1184 2.51?2.7399 2.16Dividend
?0.2906 3.14?0.4771 2.8OperatingCycle 0.4343 3.80.8443 3.14Q
?0.0852 2.49
?0.1325 2.02
Industry FE
Yes Yes Average adj.pseudo-R -squared (%)16.6325.22N
5544
3504
The table presents the first-stage logit regression estimates of the determinants of the reporting of internal control weakness over financial reporting in Year T ?1.Column A is based on the pooled sample and column B is based on the pooled sample of survivors.Average coefficient reports the average coefficient estimates across year-specific estimations from 2003through 2006.Aggregate z -statistics reports the aggregate z -statistics,which are calculated as the sum of the annual z -statistic divided by the square root of the number of years over which the equation is estimated.These aggregated z -statistics assume that each annual estimation is independent of the other estimations.See the appendix for variable definitions.
M.Cheng et al./Journal of Accounting and Economics 56(2013)1–18
12
M.Cheng et al./Journal of Accounting and Economics56(2013)1–1813
Table5
Tests of internal control weakness and total investment over the three years based on the propensity-score matched sample.
Year T?1Year Tt2
Diff-in-Diff N ICW firms Control firms Difference N ICW firms Control firms Difference
Panel A:the means of Investment for the matched sample
Group110110.8414.58?3.74n7512.3715.18?2.81?0.93 Group210112.5814.03?1.457514.2214.62?0.04?1.41 Group310015.2113.70 1.517514.9012.64 2.26?0.75 Group410016.8315.73 1.107513.7117.26?3.55n 4.65nn Group510020.5915.44 5.15nn7417.0014.91 2.09 3.06
Panel B:the means of investment for the matched sample of survivors
Group15610.7817.27?6.49nn5613.7612.47 1.29?7.78nn Group25612.0215.54?3.525614.1514.52?0.37?3.15 Group35613.6811.60 2.085617.8314.43 3.40?1.32 Group45714.0013.820.185713.0316.00?2.97 3.15 Group55720.3211.708.62nnn5714.0915.50?1.4010.02nnn
The table presents the comparisons of the average Investment between propensity-score matched ICW firms and control firms for Year T?1and Year Tt2 separately.ICW firms and the matching control firms are sorted into five groups based on OverFirm in each event year.Panel A is based on the matched sample obtained from all firms with available data,while Panels B is based on the matched sample obtained from surviving firms only with the requirement that the ICW firms have matches in all three years.For both samples,the match is performed in Years T?1,Tt1,and Tt2separately.nnn,nn, and n denote statistical significance at1%,5%and10%based on t-tests for means.
Following Armstrong et al.(2010),we examine the covariate balance between the treatment and control samples to ensure that the observable dimensions of the matched pairs are similar with the exception of their internal control effectiveness.We compare the means of the independent variables that we use in the first stage of our estimation(i.e., Model(2))for the matched pairs.For both the matched sample and the matched sample of survivors in all three event years, we obtain desirable covariate balance.For example,for Year T?1based on the matched sample,untabulated results show that the difference in means(based on t-tests)is insignificant for all25variables.It is particularly noteworthy that the means of OverFirm for the treatment and control samples are very similar and that the differences are statistically insignificant.Further,the propensity scores for the two samples are also insignificantly different between the ICW firms and the control firms.Overall,as suggested by Armstrong et al.(2010),the absence of significant differences in the variables suggests that the covariates are balanced across the treatment and control samples and that differences in terms of these observed variables between the matched pairs are not likely to confound our estimates of the treatment effect.The differences between the matched pairs are theoretically due only to the difference in internal control effectiveness.
Table5presents the descriptive statistics as well as two-sample tests of Investment for our propensity-score matched samples.We partition the sample into five groups based on OverFirm.Panel A reports statistics for the matched sample.In Year T?1,within group1where the ex ante likelihood of under-investment is high,ICW firms have significantly lower average investment levels(10.84%)than control firms(14.58%).Within group5where the ex ante likelihood of over-investment is high,ICW firms have significantly higher average investment levels(20.59%)than control firms(15.44%). These statistics provide support for both H1a and H1b.In Year Tt2,the differences in investment levels between ICW firms and control firms within both group1and group5become statistically insignificant.We also compare the difference between ICW firms’and control firms’investment levels in Year T?1with that in Year Tt2,which effectively tests for the “difference-in-difference.”The difference-in-difference is significant for neither group1nor group5.We note,however,that the difference-in-difference is statistically positive for group4,providing partial support for reductions in over-investment.
Panel B reports descriptive statistics for Investment for the matched sample of survivors.In Year T?1,for group1where the ex-ante likelihood of under-investment is high,ICW firms significantly under-invest by?6.49%of total assets relative to effective firms.In contrast,in group5where the ex-ante likelihood of over-investment is high,ICW firms significantly over-invest by8.62%of total assets relative to effective firms.Both of these tests support H1a and H1b.In Year Tt2,we no longer observe any statistical differences in investment levels between ICW firms and effective firms in either group.13 In Panel B,we also test for the significance levels of the changes in the average differences between ICW firms and control firms from Year T?1to Year Tt2(i.e.,difference-in-difference)within the under-investment group and over-investment group,respectively.Relative to the matched sample in Panel A,this comparison based on the matched sample of survivors is more sensible for testing changes in investment efficiency because the ICW firms remain constant through the 13Note that the average investment levels in Panels A and B are different in both magnitudes and patterns across the five groups,especially for control
firms.Two factors contribute to these differences.First,Panel B is based on a much smaller sample than Panel A.Second and more importantly,while the ICW firms in Panel B are largely a subset of the ICW firms in Panel A,the control firms in Panels A and B are very different.This is because the two different matched samples are formed based on different propensity-scores estimated using different pooled samples.
event years.The fact that the matching is performed in each of the Years T ?1,T t1,and T t2also mitigates concerns over potential different changes in OverFirm over time between the ICW firms and the control firms.Based on two-sample t -tests,we find that the difference-in-difference is statistically significant in both extreme quintiles (i.e.,group1and group 5).These results provide direct support for our H2a and H2b .
Unlike simple univariate analysis based on the pooled sample,the analysis in Table 5based on the propensity-score matched samples controls for various firm characteristics and therefore is less subject to confounding effects.Relative to regression analysis,this design is also considered superior in that it does not impose a linearity assumption for the relations between these firm characteristics and investment levels.Nevertheless,in Table 6,we employ the traditional regression analysis which directly controls for time-specific OverFirm as in Biddle et al.(2009).Panels A,B,and C in Table 6present the results of our estimates of Model (3)in Year T ?1,Year T t1,and Year T t2respectively based on the two matched samples.14Column A is based on our initial matched sample wherein we obtain the matched pairs from all firms for Years T ?1,T t1,and T t2separately.Column B is based on the matched sample of survivors that requires the same set of ICW firms surviving through Year T t2but allows for different matching firms in each year.
In Panel A,we find that the coefficient on Weak is significantly negative and of an economically significant magnitude.For example,based on the matched sample in Column A,a difference in investment between ICW firms and control firms of
Table 6
Regression tests of internal control weakness and total investment over the three years based on the propensity-score matched sample.Variable
Investment inefficiency test
Predicted sign
Matched sample (A)Matched sample of survivors (B)Coefficient
t -Value
Coefficient
t –Value
Panel A:the year prior to the disclosure of internal control weakness Intercept 13.46nnn 7.5713.50nnn 6.85Weak (1)T ?1:under-invest ??4.09nn ?1.65?6.91nnn
?2.41OverFirm
2.050.760.840.27Weak ?OverFirm (2)t7.75nn 2.0412.39nnn 2.72(1)t(2)
T ?1:over-invest
t
3.66nn 2.01
5.48nnn
2.38
Firm/Year Cluster No No N
1004564R 2(%)
1.12
1.26
Panel B:the year after the disclosure of internal control weakness Intercept 13.47nnn 7.8419.04nnn 7.05Weak (1)T t1:under-invest ??3.93n ?1.64?7.92nn ?2.15OverFirm
2.770.99?5.18?1.24Weak ?OverFirm (2)?
3.180.8210.72n 1.85(1)t(2)
T t1:over-invest
?
?0.75?0.39
2.800.98
Firm/Year Cluster No No N
848564R 2(%)
0.89
0.31
Panel C:two years after the disclosure of internal control weakness Intercept 14.24nnn 7.7512.54nnn
5.72Weak (1)T t2:under-invest ??1.85?0.72 2.710.90OverFirm
1.150.40 3.49 1.02Weak n OverFirm (2)?
2.360.59?4.70?0.99(1)t(2)
T t2:over-invest
?
0.510.26
?1.99?0.85
Firm/Year Cluster No No N
748564R 2(%)
0.260.21Test difference for under-investment between Year T ?1and Year T t2t 2.24n 1.859.62nnn 5.01Test difference for over-investment between Year T ?1and Year T t2??3.15nn
?1.97?7.47nnn
?3.28
Panels A,B,and C of the table present the regression estimates of the model that examines the relation between internal control weakness (Weak )and total investment (Investment )during Years T ?1,T t1,and T t2respectively based on the propensity-score matched sample.Column A is based on the matched sample obtained from all firms with available data,while Column B is based on the matched sample obtained from surviving firms only with the requirement that the ICW firms have matches in all three years.For both samples,the match is performed in Years T ?1,T t1,and T t2separately.t -Statistics are corrected for cross-sectional and time-series correlation using a two-way cluster at the firm and year level.The nnn ,nn ,and n denote statistical significance at 1%,5%,and 10%levels,respectively,based on one-tailed tests if the coefficient has a predicted sign and two-tailed tests otherwise.See the appendix for variable definitions.
14
We note that with the propensity-score matched samples we are not computing standard errors correctly as we are ignoring estimation errors from
the first-stage logit model for the propensity score.Therefore,our t -statistics are likely to be overstated.
M.Cheng et al./Journal of Accounting and Economics 56(2013)1–18
14
M.Cheng et al./Journal of Accounting and Economics56(2013)1–1815?4.09%of total assets translates to?32%on a relative basis,given that the overall investment is around12.80%of total assets.Similarly,the sum of the coefficients on Weak and OverFirm?Weak is3.66%(29%on a relative basis)and significant. The corresponding numbers in Column B based on the matched sample of survivors are also in the predicted direction and statistically significant,with even higher magnitudes.These findings provide evidence that,in the year prior to their ICW disclosure and relative to control firms that are matched on various firm characteristics,firms with ICWs over-invest or under-invest,depending on the availability of financial resources.
Table6also presents the test results for our second set of hypotheses regarding the changes in investment efficiency over the three years.We compare the coefficients on Weak as well as the sum of Weak and OverFirm?Weak between Year T?1 and Year Tt2after stacking the observations of the three years(i.e.,Year T?1,Year Tt1,and Year Tt2).In Column A based on the matched sample,we find that although the coefficients on Weak are significantly negative in both Panels A and B for Years T?1and Tt1,the coefficient on Weak is insignificant and of a much smaller magnitude at?1.85in Year Tt2,roughly 45%of its counterpart in Year T?1.In Panel C,the test on the difference-in-difference between Year T?1and Year Tt2for H2a is statistically significant at the10%level.The inferences based on the matched sample of survivors in Column B are similar,with the difference-in-difference test in Panel C statistically significant at1%.
On the over-investment side,based on Column A with the matched sample,although the sum of the coefficients on Weak and OverFirm?Weak is positive and significant at3.66in Year T?1,the corresponding sums for Years Tt1and Tt2become insignificant and much smaller at?0.75and0.51,respectively.The inferences based on the matched sample of survivors in Column B are qualitatively similar.More importantly,the difference-in-difference tests suggest that the changes in the extent of over-investment from Year T?1to Year Tt2are significant at5%and1%respectively based on these two samples.
Thus,Table6presents results that are generally consistent with Table3,supporting both H1a,H1b,H2a,and H2b.15By performing a matched comparison between two groups of firms that theoretically differ only in terms of the effectiveness of their internal control over financial reporting,Table6provides further support for a significant improvement in investment efficiency following the disclosure of ICWs.The robustness of our results based on both the matched sample and a constant set of ICW firms over time suggests that our findings are neither due to decreased sample size(i.e.,reduced power)in Year Tt2nor due to sample attrition(i.e.,survivorship bias).
4.4.Additional analyses
4.4.1.Testing the effects of remediation versus disclosure firms
Our empirical analyses of the effects of the ICW disclosure as predicted in hypotheses H2a and H2b assume that this disclosure and the changes that follow cause the subsequent increase in investment efficiency.In this subsection,we specifically examine whether the increase in investment efficiency varies between firms that have remediated the ICWs in financial reporting and those that have not.
We begin by tracking the ICW disclosures in Year Tt2by the ICW firms.We find that about78%of the ICW firms have remedied these ICWs in Year Tt2while the remaining22%continue to report ICWs in Year Tt2.This finding is consistent with our earlier argument that most firms initiate various changes after their disclosures in order to remedy their ICWs.For each of the two groups of firms,we rerun the pooled analyses in Table3.
We find that in the remediated ICW firms,the improvement in investment efficiency documented in Table3remains virtually unchanged.Specifically,there is no evidence that these firms significantly under-invest in Year Tt2or significantly over-invest in Year Tt1or Year Tt2.In addition,statistical tests for the changes in investment inefficiency from Year T?1 to Year Tt2are also significant.For the ICW firms with continued ICW disclosures,we find no evidence of statistically significant under-investment in Year Tt2,but we do still find some evidence of over-investment in Year Tt2.Statistical tests for the changes in investment inefficiency from Year T?1to Year Tt2are insignificant.However,given the small number of these firms in this sample(i.e.,less than80in Years Tt1and Tt2),we are unable to conclusively determine whether our results derive from a lack of testing power or a lack of significant changes in investment efficiency.Overall, these findings are consistent with our prediction that an increase in the financial reporting quality leads to improved investment efficiency.
4.4.2.Testing Capex and non-Capex investment efficiency
Our main measure of total investment(Investment)follows Richardson(2006)and Biddle et al.(2009).In addition to this measure,we further examine the two components of total investment,namely,capital expenditure(Capex)and non-capital expenditure(Non-Capex),in order to test whether a specific type of investment is driving our main results.
15The test results in Table5are consistent with those in Table6for the matched sample of survivors and only partially consistent for the matched sample.Specifically,for the matched sample,while Table6shows significant reduction in both over-and under-investment,Table5shows partial evidence for reduction in over-investment but no evidence for reduction in under-investment.The differences can be attributable to two factors.The tests in Table5 do not require the assumptions about the functional form of the relation between Investment and its determinants,while the tests in Table6are regression analyses based on the linearity assumption.In addition,the regression analysis in Table6takes into account all observations in the sample,while the tests in Table5focus on specific groups individually.The different sample sizes used in calculating the test statistics may affect the power of the corresponding tests.
In untabulated tests based on the pooled sample that estimate Model (1)using Capex and Non-Capex as our dependent variables for Year T ?1,we find that the coefficients on Weak are significantly negative for both measures.Further,the sum of the coefficients on Weak and OverFirm ?Weak is significantly positive for both measures.These results once again confirm that ICW firms tend to over-invest (under-invest)when their financial resources are abundant (constrained),which supports H1a and H1b .
Furthermore,in estimating Model (1)for Capex and Non-Capex for Year T t2,we find that,consistent with the results presented in Table 3,the coefficients on Weak are insignificant in all specifications,which suggests that after the initial disclosure of material weaknesses,the associated under-investment has been eliminated by the second year.We also find that the over-investment has disappeared by this time,because the sums of the coefficients for Weak and OverFirm ?Weak are likewise statistically insignificant.Statistical tests for the changes in investment inefficiency from Year T ?1to Year T t2are generally significant,although in some specifications the results are slightly weaker than those reported in Table 3for total investments.Overall,these results are largely consistent with our second hypothesis;namely,that a firm 's disclosure of its ICW,as required by SOX,mitigates investment inefficiency in both capital and non-capital investments.Although,in this case,the extent of the mitigation is weaker than the mitigation reported for total investments.
4.4.3.Firm-level versus account-level ICW
Internal control weakness can be classified into two types:firm-level weakness and account-level weakness (Moody 's Investors Service,2004,2006,2007).Practitioners as well as researchers argue that account-level material weaknesses relate to controls over specific account balances or transaction-level processes and are relatively “auditable ”in comparison to firm-level material weaknesses.To the extent that account-level material weaknesses can be detected and corrected by auditors through increased substantive testing,we expect firm-level weaknesses to have greater adverse implications for corporate activities or the quality of financial reporting (Doyle et al.,2007b;Kim et al.,2009).Accordingly,we examine whether our main results,as presented in Table 3,vary for firm-level and account-level weaknesses.
In untabulated analyses,we find that only firm-level weaknesses drive investment inefficiency in the year prior to the disclosure.Firms with ICWs at the account level generally show no investment inefficiency relative to control firms.These results suggest that firm-level weaknesses are more likely to have adverse implications for corporate investments.In Year T t2,over-investment in firms with firm-level weaknesses completely disappears.However,there is still some evidence of under-investment in these firms,which is consistent with our finding (see Table 3)that over-investment is corrected sooner than under-investment.5.Conclusion
In this study we analyze the relation between weaknesses in internal control over financial reporting and investment efficiency both before and after a given firm 's initial disclosure of its ICW,as required by the Sarbanes Oxley Act of 2002.We find robust evidence that in the year prior to the disclosure,and relative to control firms,ICW firms over-invest (under-invest)when they operate in settings more prone to over-investment (under-investment).However,these investment inefficiencies generally disappear by the second year after the revelation of the material weaknesses,although our findings of the reduction in under-investment are weak in certain tests.
Our findings support our hypothesis that ineffective internal control over financial reporting has a significant adverse impact on investment efficiency.We predict that the ICW disclosures lead shareholders and other stakeholders in the firm to increase their monitoring and hence to improve firms ’financial reporting quality.These changes mitigate agency problems such as adverse selection and moral hazard and thereby increase the efficiency of investment.By examining the changes in investment efficiency surrounding ICW disclosures,our results provide more direct evidence for the causal relation between financial reporting quality and investment efficiency than prior research has provided.To our knowledge,our study is the first to link the SOX 's disclosure requirements for ICWs to real investment efficiency.Ours is also one of the few studies that present evidence on the potential benefits of regulations related to internal control over financial reporting under SOX.
Appendix.Variable de ?nitions Main variables
Investment The sum of research and development expenditure,capital expenditure,and acquisition expenditure less sale of
property,plant,and equipment (PPE)multiplied by 100and scaled by lagged total assets
Weak An indicator variable that takes the value of one if the firm belongs to the sample of firms with material weakness,
and zero if the firm belongs to the control sample
OverFirm The average rank of a ranked (deciles)measure of cash and leverage,as in Biddle et al.(2009).Leverage is
multiplied by minus one before ranking so that both variables are increasing in the likelihood of over-investment.This variable is scaled to range between zero and one
M.Cheng et al./Journal of Accounting and Economics 56(2013)1–18
16
M.Cheng et al./Journal of Accounting and Economics56(2013)1–1817 Monitoring or governance variables
Institutions percentage of institutional holding reported by the Thomson Reuters Ownership Database at the beginning of the current year
Analysts the number of analysts following the firm as provided by the IBES summary file
G-score corporate governance index based on Gompers et al.(2003)reported by the RiskMetrics database at the beginning of the current year;the score is coded zero if missing
G-score Dummy an indicator variable that takes the value of one if the G-score is missing and zero otherwise
AQ the standard deviation of the firm-level residuals from the augmented Dechow and Dichev(2002)model during the Years t?5to t?1that is multiplied by negative one.The model is a regression of working capital accruals on lagged,current,and future cash flows;change in revenue;and PPE.All variables are scaled by average total assets.
The model is estimated cross-sectionally for each industry with at least20observations in a given year based on the Fama and French(1997)48-industry classification
Determinants of investments
Mkt-to-book the ratio of the market value of total assets to the book value of total assets
LogAsset the log of total assets
s(CFO)standard deviation of cash flow from operations deflated by average total assets from years t?5to t?1
s(Sales)standard deviation of the sales deflated by average total assets from years t?5to t?1
s(Investment)standard deviation of investment from years t?5to t?1
Z-score the decile rank of the likelihood of bankruptcy from the Altman Z-score at the beginning of the current year Tangibility the ratio of PPE to total assets
Leverage the ratio of long-term debt to the sum of long-term debt to the market value of equity
Ind Leverage mean Leverage for firms in the same SIC three-digit industry
CFOsale net cash flow from operating activities scaled by total sales at the beginning of the year
Dividend an indicator variable that takes the value of one if the firm paid a dividend and zero otherwise OperatingCycle the log of receivables to sales plus inventory to COGS multiplied by360
Losses an indicator variable that takes the value of one if earnings before the extraordinary items in the current year and prior year sum to less than zero and zero otherwise
Q Tobin's Q at the beginning of the year measured as market value minus book value of shareholders’equity plus total assets divided by total assets(Moeller et al.,2004)
Age the log of the number of years the firm has been tracked by CRSP
Determinants of reporting ICW
Nseg the log of the number of operating and geographic segments reported by the Compustat Segments database in the current year
Rest the aggregate restructuring charges in the current year and prior year scaled by the firm's market capitalization Foreign an indicator variable that takes the value of one if the firm has a non-zero foreign currency translation in the current year and zero otherwise
Extrgrow an indicator variable that takes the value of one if year-over-year industry-adjusted sales growth falls into the top quintile and zero otherwise
References
Altamuro,J.,Beatty,A.,2010.How does internal control regulation affect financial reporting.Journal of Accounting and Economics49,58–74.
American Bankers’Association(ABA),https://www.doczj.com/doc/1315148163.html,ments Submitted to the SEC roundtable on Implementation of Internal Control Reporting Provisions.SEC, Washington,DC.
Armstrong,C.S.,Jagolinzer,A.D.,Larcker,D.F.,2010.Chief executive officer equity incentives and accounting irregularities.Journal of Accounting Research 48,225–271.
Ashbaugh-Skaife,H.,Collins,D.W.,Kinney,W.R.,Lafond,R.,2008.The effect of SOX internal control deficiencies and their remediation on accrual quality.
Accounting Review83,217–250.
Ashbaugh-Skaife,H.,Collins,D.W.,Kinney,W.R.,Lafond,R.,2009.The effect of SOX internal control deficiencies on firm risk and cost of equity.Journal of Accounting Research47,1–43.
Bens,D.,Monahan,S.,2004.Disclosure quality and the excess value of diversification.Journal of Accounting Research42,691–730.
Berger,P.,Li,F.,Wong,F.,2005.The Impact of Sarbanes-Oxley on Cross-listed Companies.Working Paper.University of Chicago.
Bhojraj,S.,Sengupta,P.,2003.The effect of corporate governance mechanisms on bond ratings and yields:the role of institutional investors and outside directors.Journal of Business76(3),455–475.
Biddle,G.C.,Hilary,G.,2006.Accounting quality and firm level capital investment.Accounting Review81,963–982.
Biddle,G.C.,Hilary,G.,Verdi,R.,2009.How does financial reporting quality relate to investment efficiency?Journal of Accounting and Economics48, 112–131.
Bushman,R.,Piotroski,J.,Smith,A.,2011.Capital allocation and timely accounting recognition of economic losses:international evidence.Journal of
Business,Finance and Accounting 38(1–2),1–33.
Charles River Associates,2005.Sarbanes Oxley Section 404Costs and Remediation of Deficiencies:Estimates From a Sample of Fortune 1000Companies.Costello,A.,Wittenberg-Moerman,R.,2011.The impact of financial reporting quality on debt contracting:evidence from internal control weakness reports.
Journal of Accounting Research 49(1),97–136.
Dhaliwal,D.,Hogan,C.,Trezevant,R.,Wilkins,M.,2011.Internal control disclosures,monitoring,and the cost of debt.Accounting Review 86,1131–1156.Dechow,P.,Dichev,I.,2002.The quality of accruals and earnings:the role of accrual estimation errors.Accounting Review 77,35–59.
Doyle,J.,Ge,W.,McVay,S.,2007a.Determinants of weaknesses in internal control over financial reporting.Journal of Accounting and Economics 44,
193–223.
Doyle,J.,Ge,W.,McVay,S.,2007b.Accruals quality and internal control over financial reporting.Accounting Review 82,1141–1170.Fama,E.,French,K.,1997.Industry costs of equity.Journal of Finance 43,153–193.
Feng,M.,Li,C.,McVay,S.,2009.Internal control and management guidance.Journal of Accounting and Economics 48,190–209.
Francis,J.R.,Martin,X.,2010.Acquisition profitability and timely loss recognition.Journal of Accounting and Economics 49,161–178.Gompers,P.,Ishii,J.,Metrick,A.,2003.Corporate governance and equity prices.Quarterly Journal of Economics 118,107–155.
Gow,I.,Ormazabal,G.,Taylor,D.,2010.Correcting for cross-sectional and time-series dependence in accounting research.Accounting Review 85(2),
483–513.
Hope,O.,Thomas,W.,2008.Managerial empire building and firm disclosure.Journal of Accounting Research 46,591–626.
Jensen,M.C.,1986.Agency costs of free cash flow,corporate finance,and takeovers.American Economic Review 76(May),323–329.
Jensen,M.,Meckling,W.,1976.Theory of the firm:managerial behavior,agency costs,and capital structure.Journal of Financial Economics 3,305–360.Kim,J.,Song,B.,Zhang,L.,2009.Internal Control Weakness and Bank Loan Contracting:Evidence From SOX Section 404Disclosures.Working Paper.
Concordia University.
Lambert,R.,Leuz,C.,Verrecchia,R.,2007.Accounting information,disclosure,and the cost of capital.Journal of Accounting Research 45,385–420.
Li,H.,Pincus,M.,Rego,S.,2008.Market reaction to events surrounding the Sarbanes-Oxley act of 2002and earnings management.Journal of Law and
Economics 51,111–134.
Li,S.,Sun,L.,Ettredge,M.,2010.Financial executive qualifications,financial executive turnover,and adverse SOX 404opinions.Journal of Accounting and
Economics 50,93–110.
McNichols,M.F.,Stubben,S.R.,2008.Does earnings management affect firms ’investment decisions?Accounting Review 83(6),1571–1603.
McInnis,J.,Collins,D.,2011.The effect of cash flow forecasts on accrual quality and benchmark beating.Journal of Accounting and Economics 51,219–239.Modigliani,F.,Miller,M.H.,1958.The cost of capital,corporation finance and the theory of investment.American Economic Review 48(3),261–297.Moeller,S.B.,Schlingemann,F.P.,Stulz,R.M.,2004.Firm size and the gains from acquisitions.Journal of Financial Economics 73(2),201–228.
Moody 's Investors Service,2004,2006,2007.First Year,Second Year,and Third Year Section 404Reports on Internal Control:Impact on Ratings Depend on
Nature of Material Weaknesses Reported.New York,NY.
Myers,S.,1997.Determinants of corporate borrowing.Journal of Financial Economics 5,147–175.
Myers,S.,Majluf,N.,1984.Corporate financing and investment decisions when firms have information that investors do not have.Journal of Financial
Economics 13,187–221.
Nicolaisen,D.T.,2004.Keynote Speech at 11th Annual Midwestern Financial Reporting Symposium.October 7,Chicago,IL.
Ogneva,M.,Subramanyam,K.,Raghunandan,K.,2007.Internal control weakness and cost of equity:evidence from SOX Section 404certifications.
Accounting Review 82,1255–1297.
Opler,T.,Pinkowitz,L.,Stulz,R.,Williamson,R.,1999.The determinants and implications of corporate cash holdings.Journal of Financial Economics 52,
3–46.
Petersen,M.,2009.Estimating standard errors in finance panel data sets:comparing approaches.Review of Financial Studies 22(1),435–480.Richardson,S.,2006.Over-investment of free cash flow.Review of Accounting Studies 11,159–189.
Securities and Exchange Commission (SEC),2003.Management 's Report on Internal Control Over Financial Reporting and Certification of Disclosure in
Exchange Act Periodic Reports.?https://www.doczj.com/doc/1315148163.html,/rules/final/33-8238.htm ?.Release nos.33-8238.Shleifer,A.,Vishny,R.W.,1997.A survey of corporate governance.Journal of Finance 52(2),737–783.
Stein,J.C.,2003.Agency,information and corporate investment.Handbook of the Economics of Finance 1(1),111–165.edition .Stiglitz,J.,Weiss,A.,1981.Credit rationing in markets with imperfect information.American Economic Review 71,393–410.U.S.Congress,2002.The Sarbanes-Oxley Act of 2002.Public Law https://www.doczj.com/doc/1315148163.html,ernment Printing Office,Washington,D.C.
Williamson,O.,1974.Economics of Discretionary Behavior:Managerial Objectives in a Theory of the Firm.Kershar Publishing,London,UK .Zhang,I.,2007.Economic consequences of the Sarbanes-Oxley Act of 2002.Journal of Accounting and Economics 44,74–115.
M.Cheng et al./Journal of Accounting and Economics 56(2013)1–18
18