Internal Control Weakness Disclosure and Firm Investment

Background and Literature Review

To enhance the accuracy and reliability of corporate financial reporting and to restore investor confidence in the financial markets, the U.S. Congress passed the SOX on July 30, 2002. Section 404 of SOX requires that firm management and the auditor assess and provide separate reports on the effectiveness of internal control over financial reporting ¹ as of fiscal year-end. The auditor issues an adverse opinion if the firm’s internal control over financial reporting has a material weakness, ² and a clean opinion otherwise. The primary objective of Section 404 is to provide useful disclosure to the public about the effectiveness of a company’s internal control over financial reporting and to protect the interests of shareholders and the public by preventing fraudulent practices and accounting inconsistencies. Although public companies in the United States are required to design and maintain adequate internal controls over accounting by the Foreign Corrupt Practices Act (U.S. Congress, 1977), auditor assessment and public disclosures of internal control weaknesses became mandatory only after the effective date of Section 404 of SOX. In other words, Section 404 internal control disclosures provide the public with information on a firm’s internal control quality that is not readily available otherwise.

With internal control disclosures, studies have examined the determinants of internal control weaknesses (e.g., Ashbaugh-Skaife, Collins, & Kinney, 2007; Doyle, Ge, & McVay, 2007b), the association between internal control effectiveness and the quality of accruals (e.g., Ashbaugh-Skaife et al., 2008; Chan et al., 2008; Doyle et al., 2007a), market reactions to the disclosures of internal control weaknesses (e.g., Hammersley, Myers, & Shakespeare, 2008), the association between internal control quality and cost of capital (e.g., Ashbaugh-Skaife, Collins, Kinney, & LaFond, 2009; Beneish, Billings, & Hodder, 2008; Costello & Wittenberg-Moerman, 2011; Gordon & Wilford, 2012; Ogneva, Subramanyam, & Raghunandan, 2007), the effect of internal control weaknesses on audit risk and audit fees (e.g., Ettredge, Li, & Sun, 2006; Raghunandan & Rama, 2006), the relation between internal control quality and the accuracy of management earnings guidance (e.g., Feng, Li, & McVay, 2009), and the effect of material weaknesses on shareholders’ decision to withhold votes on director re-election (Ye, Hermanson, & Krishnan, 2013). However, there is limited evidence on the consequences of internal control evaluation and disclosure as required by Section 404 of SOX for firms’ real decisions (e.g., firm investment). In this study, I use Section 404 auditor opinions on the effectiveness of internal control over financial reporting to examine the association between internal control weakness disclosures and the level of firm investment.

The Relation Between Internal Control Weakness Disclosure and Firm Investment

The role of financial reporting in capital allocation in the economy and within the firms is a fundamental question of accounting. High-quality accounting information is believed to improve capital allocation and enhance investment efficiency (Bushman & Smith, 2001; Healy & Palepu, 2001). Existence of weaknesses in internal control over financial reporting is expected to affect a firm’s information quality, investors’ assessment of future cash flows, investors’ demand for a risk premium, and thus firm investment.

In a capital market with no frictions, firm investment should be solely determined by investment opportunities (Modigliani & Miller, 1958). Firms invest if the expected benefits are greater than the costs. With a model of multi-security economy in which cash flows of different firms are correlated, Lambert et al. (2007) show that the quality of a firm’s accounting system influences its real decisions, such as investment, even when there are no agency problems. In their model, a firm’s accounting system provides noisy information about its future cash flows. The accounting system includes both the financial disclosures to external information users and the firm’s accounting internal control system. They show that the firm’s optimal investment level decreases with investors’ assessed variance of the firm’s cash flow and covariance of the firm’s cash flow with the cash flows of the other firms in the economy. Investors’ assessment is conditional on the information from the firm’s accounting system about its future cash flows. As the firm’s information quality deteriorates, investors’ assessed cash-flow variance and covariance increase, leading to a higher cost of capital and a lower equilibrium investment level by the firm.

When a firm’s internal control over financial reporting is revealed to have material weaknesses, investors may perceive the firm’s financial information to be noisy or unreliable as firms with internal control weaknesses have lower-quality accruals compared with firms with effective internal control (e.g., Ashbaugh-Skaife et al., 2008; Chan et al., 2008; Doyle et al., 2007a). As a result, outside capital providers will re-assess the firm’s information risk and future cash flows (Moody’s Investors Service, 2006). If outside capital providers respond to the disclosure of internal control weaknesses by demanding a greater risk premium and rationing capital, firm managers will have to decrease investment and may even choose to pass up good investment opportunities (Myers & Majluf, 1984; Stiglitz & Weiss, 1981), especially when the firm’s financial constraint becomes binding. As a result, ceteris paribus, the investment level of firms that receive adverse auditor internal control opinions is expected to be lower than that of firms that receive clean opinions.

Agency problems, such as adverse selection and moral hazard, arise when there is information asymmetry between managers and outside capital providers (Stein, 2003). The conflicts of interests and the information asymmetry between managers and investors may prevent firms from investing efficiently. Under adverse selection, managers’ better information about firm value and prospects may enable them to raise capital when the firm is overvalued, leading to more investments (e.g., Baker, Stein, & Wurgler, 2003). Under moral hazard, managers may make investment decisions that maximize their personal wealth but reduce their firm value (Jensen, 1986; Jensen & Meckling, 1976).

When there are agency problems, the disclosure of internal control weaknesses can have several effects. First, the likelihood for an ineffective-internal-control firm to sell overpriced shares is expected to be lower as investors question and re-assess the quality of the firm’s financial information and future prospects upon the revelation of internal control weaknesses. Second, the existence of internal control weaknesses may further increase the risk premium demanded by the capital providers because low-quality information exacerbates information asymmetry (Easley, Hvidkjaer, & O’Hara, 2002; Easley & O’Hara, 2004). And third, the disclosure of internal control weaknesses is expected to increase the scrutiny by capital providers on managers’ decisions to invest. These effects suggest a smaller amount of capital that a firm can raise, a higher cost of capital, and a lower likelihood for overinvestment, ³ leading to lower investment by firms that receive adverse internal control opinions compared with firms that receive clean opinions. Therefore, my first hypothesis (stated in the alternative form) is as follows:

There are economic reasons to believe that the revelation of internal control weaknesses has a different effect on different components of investment. The various components of investment are not equally risky. R&D expenditures are generally regarded more risky than capital expenditures because of greater uncertainty of future benefits for these investments (e.g., Kothari, Laguerre, & Leone, 2001). Acquisitions of companies are also more risky than regular capital expenditures on property, plant, and equipment because of the information asymmetry between the acquirer and the target company. More risky investments are more likely to lead to extreme poor performance and thus financial distress. If the knowledge of internal control weaknesses leads to greater monitoring on managers’ decisions by shareholders and creditors, firm managers are expected to be more cautious about their decisions for risky investments. Therefore, I expect that ineffective-internal-control firms are more likely to reduce investments that are more risky. My second hypothesis (stated in the alternative form) is as follows:

Association Between Internal Control Weakness Disclosure and Firm Investment

Firm investment decisions have been examined extensively in economics and finance literature. Following the corporate finance literature (Hubbard, 1998), I model firm investment as a linear function of growth opportunities, measured using lagged Tobin’s Q (Tobin, 1984) and control variables. I follow Richardson (2006) to use an accounting-based framework to measure investment as the sum of capital expenditures, R&D expenses, and acquisitions less sales of property, plant, and equipment, scaled by total assets at the beginning of the year. Q is measured as the ratio of market value of total assets to book value of total assets (Kaplan & Zingales, 1997).

To test H1 on the relation between internal control weakness disclosure and firm investment, I use Regression 1 to examine cross-sectionally whether firms that receive adverse auditor opinions on internal control have lower investment than firms that receive clean opinions.

INVES T i , t = α 0 + α 1 IC W i , t + β 1 Q i , t − 1 + β 2 LnT A i , t − 1 + β 3 AG E i , t − 1 + β 4 LEVERAG E i , t − 1 + β 5 CAS H i , t − 1 + β 6 CF O i , t + β 7 TANGIBL E i , t − 1 + β 8 Z i , t − 1 + β 9 DIVIDEN D i , t − 1 + β 10 StdINVES T i , t − 1 + β 11 A Q i , t − 1 + β 12 INSTITUT E i , t − 1 + β 13 ANALYS T i , t − 1 + Σ β m Yea r m + Σ β n Industr y n + ε i . t .

The variable of interest in Regression 1 is the disclosure of internal control weaknesses (ICW). It is defined as 1 if the firm receives an adverse auditor opinion on internal control in either the previous or the current year, and 0 otherwise. I define ICW using auditor opinions in both the current year and the previous year because the relation between weakness disclosure and firm investment may not be contemporaneous. A negative coefficient for ICW would suggest that firms that are revealed to have internal control weaknesses have lower investment, consistent with H1.

I include factors that have been shown to influence firm investment in prior studies as control variables. First, research shows that firm investment is lower if the firm is constrained in the ability to raise additional outside capital to fund new investment opportunities (e.g., Fazzari, Hubbard, & Petersen, 1988; Hubbard, 1998). I therefore include variables associated with financing constraints, such as firm size, firm age, and leverage. Firm size (LnTA) is measured as the natural logarithm of total assets. Firm age (AGE) is defined as the number of years that a firm has data on Center for Research in Security Prices (CRSP). On one hand, larger firms and older firms are more well-known and may have easier access to the capital markets to fund investment projects than smaller firms and younger firms. On the other hand, these firms are likely mature firms whose investment opportunity set shrinks. Thus, I do not have an expectation for the sign of the coefficient for firm size or age. Leverage (LEVERAGE) is the ratio of total debt to book value of equity. High-leverage firms may have difficulty raising additional capital because excessive leverage increases the probability of financial distress. The coefficient for leverage is thus expected to be negative. All three variables are measured at the beginning of the year.

Second, I include cash holding (CASH) at the beginning of the year and cash flows from operations (CFO) during the year, both scaled by beginning total assets, to control for differences in firms’ ability to finance projects using internal cash. Another reason to include CFO is that it may capture measurement errors in Tobin’s Q as a measure for investment opportunities because contemporaneous cash flows may have a better ability to capture a firm’s short-term profitability (Biddle & Hilary, 2006). Third, as a firm’s ability to obtain external capital increases with the amount of assets that can be pledged as collateral (Almeida & Campello, 2007), I control for the tangibility of a firm’s assets (TANGIBLE) at the beginning of the year, defined as the ratio of property, plant, and equipment to total assets.

Fourth, I control for bankruptcy risk and dividend payment. Bankruptcy risk is measured using Altman’s (1968) Z score (Z) computed at the beginning of the year. Firms with greater distress risk have a poorer ability to raise additional external capital, and thus may not be able to fund new investment. Dividend payment (DIVIDEND) is an indicator variable that takes the value of 1 if the firm pays dividends in the previous year and 0 otherwise. On one hand, firms with the ability to pay dividends likely have ample internal funds to meet debt obligations and to finance new projects. On the other hand, such firms are likely mature firms with fewer investment opportunities (Grullon, Michaely, & Swaminathan, 2002). Next, I include the standard deviation of firm investment in the previous 5 years (StdINVEST) to control for the variation of investment over time.

Prior studies (Biddle & Hilary, 2006; Biddle et al., 2009) find that firms with better accruals quality have lower investment-cash-flow sensitivity and are less likely to over- or underinvest. To examine whether the relation between internal control weakness disclosures and firm investment is incremental to the effect of accruals quality, I add accruals quality (AQ) as another control variable. ⁴ AQ is defined using a modified Dechow and Dichev (2002) measure proposed by Wysocki (2009). Following Biddle et al. (2009), I also control for other governance mechanisms that may affect firm investment, including the percentage of institutional ownership (INSTITUTE) and the number of analysts following the firm (ANALYST). ⁵ Both variables are measured at the beginning of the year. Finally, I include indicator variables for industries and years to capture cross-sectional and inter-temporal variation in investment. Industries are defined according to Fama and French (1997).

Association Between Internal Control Weakness Disclosure and Discretionary Investment

With Regression 1, I examine the association between internal control weakness disclosures and the actual amount of firm investment. To examine whether weakness disclosures also relate to a firm’s deviation from normal investment, I re-estimate regression 1 by replacing the dependent variable by discretionary investment. A negative coefficient for ICW will indicate that firms that are disclosed to have internal control weaknesses have a less positive or more negative deviation from the expected level of investment.

I use two approaches to measure discretionary investment. First, I estimate Regression 2 in which firm investment is modeled as a linear function of growth opportunities, measured using lagged Tobin’s Q (Tobin, 1984) that is widely used in the literature of corporate finance (Hubbard, 1998). INVEST and Q are defined in the same way as in Regression 1. Regression 2 is estimated using all COMPUSTAT firm-years with necessary data in each year and each Fama and French (1997) industry with at least 20 observations. The residual from Regression 2 is the discretionary investment of a firm-year observation, which measures the deviation from the expected investment level. ⁶

INVES T i , t = α 0 + β 1 Q i , t − 1 + ε i . t .

Second, I use average investment for each year and each Fama and French (1997) industry as another proxy for normal investment. The difference between a firm’s actual investment and the industry average is my second measure of discretionary investment.

Disclosure and Correction of Internal Control Weaknesses and Firm Investment

To assess the existence of a causal relation between control weakness disclosure and firm investment, I conduct both a time-series analysis and a change analysis.

First, I use Regression 3 to conduct a time-series analysis to examine whether firms decrease investment after the revelation of internal control weaknesses and whether these firms increase investment after the control problems are corrected. If lower investment of firms that receive adverse internal control opinions is indeed a result of the weakness disclosure, I expect these firms to increase their investment after internal control weaknesses are corrected as evidenced by the receipt of clean opinions.

INVES T i , t = α 0 + α 1 ICW 1 i , t + α 2 ICW 2 i , t + α 3 ICWCOR 1 i , t + α 4 ICWCOR 2 i , t + β 1 Q i , t − 1 + β 2 LnT A i , t − 1 + β 3 AG E i , t − 1 + β 4 LEVERAG E i , t − 1 + β 5 CAS H i , t − 1 + β 6 CF O i , t + β 7 TANGIBL E i , t − 1 + β 8 Z i , t − 1 + β 9 DIVIDEN D i , t − 1 + β 10 StdINVES T i , t − 1 + β 11 A Q i , t − 1 + β 12 INSTITUT E i , t − 1 + β 13 ANALYS T i , t − 1 + Σ β m Yea r m + Σ β n Industr y n + ε i . t .

ICW1 takes the value of 1 if year t is the first year that a firm receives an adverse auditor internal control opinion and 0 otherwise. ICW2 indicates the second or a later year that the firm receives an adverse opinion. ICWCOR1 takes the value of 1 if year t is the year during which the firm’s internal control weaknesses are corrected as evidenced by the receipt of a clean auditor opinion after the weakness revelation and 0 otherwise. ICWCOR2 indicates the second or a later year that an ineffective-internal-control firm receives a clean auditor opinion.

Section 404 of SOX requires the auditor to issue an opinion on the effectiveness of a firm’s internal control as of fiscal year-end. If, as I hypothesize, lower investment of firms that receive adverse opinions is indeed a result of the weakness disclosure, these firms are expected to reduce investment after, instead of during, the year that they receive the first adverse opinions. That is, the coefficient for ICW2 (i.e., the years after the receipt of the first adverse opinion) is expected to be significantly negative, whereas the coefficient for ICW1 (i.e., the year for the receipt of the first adverse opinion) is not. Similarly, if firm investment increases after control weaknesses are corrected, the increases are expected to take place after, instead of during, the year that these firms receive the first clean opinions. That is, the coefficient for ICWCOR1 (i.e., the year for weakness correction) is still expected to be negative. After the receipt of the first clean opinions, firm investment is expected to increase as investors re-assess the firms’ information risk. However, it is not clear how quickly these firms increase investment and whether firm investment will increase to the level before weaknesses revelation. Thus, an insignificant coefficient for ICWCOR2 or a coefficient that is less negative than the coefficients for the indicator variables for earlier years would both suggest that firm investment has increased after the receipt of clean internal control opinions.

Second, to better control for firm-level accounting and other factors that are stable over time, I conduct a change analysis that examines whether the change in firm-level investment is associated with the change in internal control quality using Regression 4. The change analysis also helps mitigate the concern of correlated omitted variables.

Δ INVES T i , t = α 0 + α 1 UNQ _ AD V i , t − 1 + α 2 ADV _ AD V i , t − 1 + α 3 ADV _ UN Q i , t − 1 + α 4 UNQ _ UN Q i , t − 1 + β 1 Δ Q i , t − 1 + β 2 Ln Δ T A i , t − 1 + β 3 Δ LEVERAG E i , t − 1 + β 4 Δ CAS H i , t − 1 + β 5 Δ CF O i , t + β 6 Δ TANGIBL E i , t − 1 + β 7 Δ Z i , t − 1 + β 8 Δ DIVIDEN D i , t − 1 + β 9 Δ StdINVES T i , t − 1 + β 10 Δ A Q i , t − 1 + β 11 Δ INSTITUT E i , t − 1 + β 12 Δ ANALYS T i , t − 1 + Σ β m Yea r m + Σ β n Industr y n + ε i . t .

I include in Regression 4 four indicator variables for different changes in auditor control opinions in the previous year (i.e., year t− 1). UNQ_ADV_t−1 takes the value of 1 if the firm receives a clean opinion in year t− 2, but an adverse opinion in year t− 1. ADV_ADV_t−1 equals 1 if the firm receives adverse opinions in both year t− 2 and year t− 1. ADV_UNQ_t−1 takes the value of 1 if the firm receives an adverse opinion in year t− 2, but a clean opinion in year t− 1. UNQ_UNQ_t−1 equals 1 if the firm receives clean opinions in both year t− 2 and year t− 1, but an adverse opinion before year t− 2. The dependent variable and the control variables are defined similarly to Regression 3, except that now they are the change from the previous year to the current year. Change in age (ΔAGE) is omitted from the regression because it is 1 for all observations.

Essentially, I use observations for firms that receive clean internal control opinions in all the years during the sample period and observations before the first adverse opinion for ineffective-internal-control firms as the control observations to examine the relation between the change in firm investment in the current year and the change in internal control quality in the previous year. If the investment change is indeed a response to internal control disclosure, I expect the coefficients for UNQ_ADV_t−1 and ADV_ADV_t−1 to be negative, suggesting a smaller investment increase or even an investment decrease after internal control weaknesses are revealed. I also expect the coefficients for ADV_UNQ_t−1 and UNQ_UNQ_t−1 to be less negative or even insignificantly different from 0, suggesting that the investment change is getting back to normal after internal control weaknesses are corrected.

Disclosure of Internal Control Weaknesses and Different Components of Firm Investment

I follow Richardson (2006) to use an accounting-based framework to measure a firm’s total investment that includes capital expenditures, R&D expenses, and acquisitions less sales of property, plant, and equipment. However, the various components of firm investment differ in terms of riskiness. Given the expected greater monitoring by existing and prospective capital providers on firm decisions after the weakness revelation, I expect that firm managers are more likely to reduce investments that are more risky because of the greater uncertainty in future benefits for these investments (H2).

To test H2, I replace the dependent variables for Regressions 1 and 3 by the three components of total investment: R&D expenses, acquisitions, and net capital expenditures (i.e., net of sale of property, plant, and equipment). I expect the investment decrease to be more significant for R&D expenses and acquisitions than for net capital expenditures that are necessary to support a company’s operations.

Sample Selection and Descriptive Statistics

I obtain financial data and auditors’ opinions on internal control from COMPUSTAT, analyst following from Institutional Brokers’ Estimate System (IBES), and institutional shareholding from Thomson Reuters. I start with 21,946 U.S. firm-year observations with non-missing data for the required variables for Regression 1, except for the internal control weakness variable, during 2004-2012. Then to define internal-control-disclosure variables, I delete 5,391 observations without auditors’ opinions on internal control. The final sample consists of 16,555 U.S. firm-year observations for 3,187 firms with necessary data. The sample period begins in 2004 because Section 404 of SOX became effective for fiscal years ending after November 14, 2004 for accelerated filers. ⁷ Six hundred eighty of the 3,187 firms receive at least one adverse auditor opinion on internal control during the sample period. Continuous variables are winsorized at the 1st and 99th percentiles of their respective distributions.

I report descriptive statistics of the sample in Table 1, Panel A. The mean (median) firm investment (INVEST) for the 16,555 sample firm-year observations is 15.28% (9.53%) of total assets at the beginning of the year. Among these 16,555 observations, 1,629 (9.84%) receive an adverse auditor internal control opinion in either the previous or the current year (ICW). In Panel B, I report the sample distribution by year and the number of ICW observations in each year. The number of sample observations is slightly higher between 2005 and 2009 probably because some non-accelerated filers chose to voluntarily have their internal control audited before the Securities and Exchange Commission (SEC) issued Rule 33-9142 to permanently exempt them from the Section 404 requirement for an internal control audit. ⁸ The ICW observations as a percentage of sample observations in each year decrease from 18.37% in 2005 to 5.55% in 2012, indicating that requirements of Section 404 have helped firms remediate their internal control weaknesses and that fewer firms get adverse internal control opinions over time. In Panel C, I report the distribution of the year of weakness correction for the 680 firms that receive adverse internal control opinions. Three hundred forty-four and 91 firms correct their control weaknesses in 1 and 2 years, respectively. The weaknesses for 193 firms are not corrected yet by 2012, the last year of the sample period. For the 487 firms that correct the weaknesses during the sample period, it takes a firm, on average, 1.47 years to remediate the problems, as evidenced by the receipt of a clean opinion.

OPEN IN VIEWER

Table 1.

Descriptive Statistics.

Panel A: Descriptive Statistics.
Variables	Mean	SD	1st quartile	Median	3rd quartile
INVESTt	15.2843	19.3023	4.7757	9.5323	18.6317
dINVESTt	−0.1846	15.3783	−7.6705	−3.0229	3.0178
INVEST_adjt	−3.8385	18.4908	−12.3079	−5.5175	0.5223
NetCAPEXt	5.2054	7.6507	1.5704	3.1474	6.1042
ACQRDt	10.1494	19.2928	0.0000	3.3829	12.4793
ICWt	0.0984	0.2979	0.0000	0.0000	0.0000
Qt −1	2.0880	1.6090	1.2171	1.6130	2.3817
LnTAt −1	6.6068	1.7490	5.3456	6.4711	7.7046
AGEt −1	19.1601	17.4875	7.0000	13.0000	25.0000
LEVERAGEt −1	0.7795	2.0156	0.0061	0.2779	0.7474
CASHt −1	0.2213	0.2294	0.0404	0.1366	0.3345
CFOt	0.0706	0.1725	0.0383	0.0926	0.1499
TANGIBLEt −1	0.2392	0.2221	0.0698	0.1615	0.3400
Zt −1	6.0283	11.3195	2.0841	3.5095	6.3448
DIVIDENDt −1	0.3877	0.4872	0.0000	0.0000	1.0000
StdINVESTt −1	13.2601	30.6675	2.6899	5.8839	13.0623
AQt− 1	1.1893	0.5240	0.9457	1.0910	1.2957
INSTITUTEt− 1	0.6812	0.2512	0.5188	0.7358	0.8869
ANALYSTt− 1	7.3934	6.5896	2.0000	6.0000	11.0000
NBSt	2.2834	1.6637	1.0000	1.0000	3.0000
FOREIGNSt	0.2606	0.4390	0.0000	0.0000	1.0000
MAt	0.2106	0.4077	0.0000	0.0000	0.0000
RESTRUCTt	0.3545	0.4784	0.0000	0.0000	1.0000
GROWTHt	0.5003	0.2887	0.2503	0.5003	0.7503
INVENTORYt	0.1041	0.1162	0.0068	0.0712	0.1598
LOSS% t	0.2715	0.3693	0.0000	0.0000	0.3333
AUDITORt	0.9321	0.2516	1.0000	1.0000	1.0000
LITIGATIONt	0.3678	0.4822	0.0000	0.0000	1.0000

Note. This table presents descriptive statistics for the 16,555 sample firm-year observations with necessary data during 2004-2012. Continuous variables are winsorized at the 1st and 99th percentiles of their respective distributions.

Definition of Variables: INVEST_t is the sum of capital expenditures, research and development expenses, and acquisitions less sales of property, plant, and equipment during year t, scaled by total assets at the beginning of the year; dINVEST_t is the firm’s discretionary investment during year t, defined as the residual from the regression of investment as a linear function of growth opportunities measured using lagged Tobin’s Q (Tobin, 1984); INVEST_adj_t is the firm’s industry-adjusted investment during year t, defined as the firm’s investment minus industry average in year t; NetCAPEX_t is the firm’s net capital expenditure during year t, defined as capital expenditures minus sales of property, plant, and equipment during year t, scaled by total assets at the beginning of the year; ACQRD_t is the firm’s acquisitions and research and development expenses during year t, scaled by total assets at the beginning of the year. If a firm has a missing value for research and development expenses but non-missing values for the other investment components, research and development expenses are set to 0; ICW _t is 1 if the firm receives an adverse auditor opinion on internal control in year t− 1 or year t and 0 otherwise; Q_t−1 is the ratio of market value of total assets to book value of total assets at the end of year t− 1; LnTA_t−1 is the natural logarithm transformation of the firm’s total assets at the end of year t− 1; AGE_t−1 is the number of years that the firm has data in Center for Research in Security Prices (CRSP) by the end of year t− 1; LEVERAGE_t−1 is the sum of long-term debt and short-term debt divided by total assets at the end of year t− 1; CASH_t−1 is the ratio of cash to total assets at the end of year t− 1; CFO _t is the cash flows from operation for year t divided by total assets at the beginning of the year; TANGIBLE_t−1 is the ratio of property, plant, and equipment to total assets at the end of year t− 1; Z_t−1 is the firm’s bankruptcy risk measured using Altman’s (1968) Z score at the end of year t− 1; DIVIDEND_t−1 is 1 if the firm pays dividends during year t− 1 and 0 otherwise; StdINVEST_t−1 is the standard deviation of the firm’s investment over year t− 1 and the previous 4 years; AQ_t−1 is the accruals quality for year t− 1 based on a modified Dechow and Dichev (2002) measure proposed by Wysocki (2009). It is measured as the ratio of the standard deviation of the residuals from a regression of working capital accruals on current cash flows to the standard deviation of the residuals from a regression of working capital accruals on past, current, and future cash flows from year t− 5 to year t− 1; INSTITUTE_t−1 is the percentage of the firm’s shares held by institutions at the end of year t− 1, computed as institutional shareholding at year-end as reported by Thomson Reuters divided by shares outstanding at year-end; ANALYST_t−1 is the number of analysts following the firm in year t− 1 as reported on Institutional Brokers’ Estimate System (IBES); NBS _t is the number of business segments for the firm in year t; FOREIGNS _t is 1 if the firm has foreign sales in year t and 0 otherwise; MA_t is 1 if the firm engages in a merger or acquisition in year t and 0 otherwise; RESTRUCT_t is 1 if the firm has a restructuring in year t and 0 otherwise. Restructuring is identified by a non-zero value for restructuring costs; GROWTH_t is the average sales growth rate over year t and the previous 2 years; INVENTORY_t is the average inventory to total assets over year t and the previous 2 years; LOSS% _t is the percentage of years with negative earnings over year t and the previous 2 years; AUDITORt is 1 if the firm engages one of the six largest auditors (PWC, Deloitte & Touche, Ernst and Young, KPMG, Grant Thornton, and BDO Seidman) in year t and 0 otherwise; LITIGATION _t is 1 if the firm belongs to a high litigation industry (Standard Industrial Classification [SIC]: 2833-2836, 3570-3577, 3600-3674, 5200-5961, and 7370) and 0 otherwise.

OPEN IN VIEWER

Panel B: Sample Distribution by Year.
Year	Sample observations	ICW observation	% of ICW observation	Year	Sample observations	ICW observation	% of ICW observations
2004	1,602	224	13.98	2009	1,820	85	4.67
2005	2,031	373	18.37	2010	1,740	52	2.99
2006	2,042	325	15.92	2011	1,712	58	3.39
2007	2,008	252	12.55	2012	1,677	93	5.55
2008	1,923	167	8.68	Total	16,555	1,629	9.84

Note.ICW = internal control weakness.

OPEN IN VIEWER

Panel C: Correction of Internal Control Weaknesses.
	Year of correction						Total
	t+ 1	t+ 2	t+ 3	t+ 4	t+ 5 ~ t+ 7	Other
n	344	91	31	15	6	193	680

Note. This table presents the distribution of the year of weakness correction for firms that receive an adverse internal control opinion in year t. “Other” refers to the number of firms that have not yet received a clean internal control opinion by 2012, the last year of the sample period.

OPEN IN VIEWER

Panel D: Comparison Between Firm-Years With Clean Auditor Internal Control Opinions (n = 14,926) and Firm-Years With Adverse Opinions (n = 1,629).
	Clean Mean	Adverse Mean	Difference in mean	Cleanmedian	Adverse median	Difference in median
INVESTt	15.3973	14.2485	1.1488**	9.5519	9.3339	0.2180*
dINVESTt	−0.0187	−1.7115	1.6929***	−2.8816	−4.4755	1.5940***
INVEST_adjt	−3.5669	−6.3265	2.7596***	−5.2527	−7.9313	2.6787***
NetCAPEXt	5.2241	5.0333	0.1909	3.1729	2.8817	0.2912***
ACQRDt	10.2502	9.2263	1.0239**	3.3766	3.4305	−0.0539
Qt− 1	2.0951	2.0237	0.0713	1.6189	1.5533	0.0656***
LnTAt− 1	6.6735	5.9953	0.6782***	6.5534	5.8369	0.7166***
AGEt− 1	19.5991	15.1381	4.4610***	14.0000	11.0000	3.0000***
LEVERAGEt− 1	0.7754	0.8167	−0.0413	0.2841	0.2336	0.0505
CASHt− 1	0.2208	0.2260	−0.0052	0.1347	0.1536	−0.0189***
CFOt	0.0736	0.0432	0.0304***	0.0956	0.0583	0.0373***
TANGIBLEt− 1	0.2422	0.2124	0.0298***	0.1632	0.1443	0.0190***
Zt− 1	6.0274	6.0372	−0.0099	3.5022	3.5618	−0.0596
DIVIDENDt− 1	0.4032	0.2455	0.1576***	0.0000	0.0000	0.0000***
StdINVESTt− 1	13.1145	14.5940	−1.4795*	5.7787	7.2449	−1.4662***
AQt− 1	1.1937	1.1494	0.0443***	1.0938	1.0706	0.0232***
INSTITUTEt− 1	0.6874	0.6242	0.0632***	0.7429	0.6613	0.0816***
ANALYSTt− 1	7.6381	5.1510	2.4871***	6.0000	4.0000	2.0000***

Note. This table compares key variables between firms that receive clean auditor internal control opinions (ICW = 0) and firms that receive adverse opinions (ICW = 1). Means and medians of these variables, differences in means, and differences in medians are reported. Tests for differences in means are based on t statistics. Tests for differences in medians are based on Z statistics from the Wilcoxon rank sum test. Variables are defined in Table 1, Panel A. ICW = internal control weakness.

***

, **, and * indicate statistical significance at the 1%, 5%, and 10% confidence levels, respectively.

In Panel D, I compare key variables between observations with clean opinions (i.e., ICW = 0) and those with adverse opinions (i.e., ICW = 1). I find that the average investment (INVEST) of firm-years with adverse opinions is lower by 1.1488% (15.3973% − 14.2485%) of beginning total assets than firm-years with clean opinions, consistent with H1. Furthermore, the average discretionary investment (dINVEST) of firm-years with weakness disclosures is lower by 1.6929% of beginning assets. In untabulated analysis, I find that 499 and 1,130 of the 1,629 firm-years with weakness disclosures (i.e., ICW = 1) have positive and negative dINVEST, respectively. For the components of total investment, firm-years with control weaknesses have lower net capital expenditures (NetCAPEX) and lower acquisitions and research and development expenses (ACQRD). Relative to firms with effective control, firms with control weaknesses are smaller (LnTA) and younger (AGE), consistent with the finding in prior research that large firms and mature firms are less likely to have internal control weaknesses (Ashbaugh-Skaife et al., 2007; Doyle et al., 2007b). The comparison in Panel D also indicates that firms with control weaknesses are less likely to pay dividends (DIVIDEND). They have lower operating cash flows (CFO), higher cash balance (CASH), fewer tangible assets (TANGIBLE), and a greater variation in investment over time (StdINVEST). Firms with effective internal control also have better accruals quality (AQ), greater institutional ownership (INSTITUTE), and more analysts (ANALYST), suggesting that these firms have better governance mechanisms that potentially help investors monitor and discipline firm investment.

Association Between Internal Control Weakness Disclosure and Firm Investment

To test H1 that firms that receive adverse auditor opinions on internal control have lower investment than firms that receive clean opinions, I estimate Regression 1 and report the results in Table 2, column 1. t statistics are based on standard errors clustered by firm and year to allow for both time-series and cross-sectional dependence (Gow, Ormazabal, & Taylor, 2012; Petersen, 2009). The coefficient for the internal control weakness indicator (ICW) is −2.1598, suggesting that investment of firms that are disclosed to have ineffective internal control is lower than that of firms with effective internal control by 2.1598% of assets at the beginning of the year, after controlling for other factors that influence firm investment. Given that average firm investment for observations with effective internal control is 15.3973% of beginning assets, 2.1598% represents a difference in investment by 14.03% (2.1598% / 15.3973%) of this average, which is not only statistically but also economically significant.

OPEN IN VIEWER

Table 2.

Internal Control Weakness Disclosure and Firm Investment.

Variables	Predicted sign	(1) INVEST		(2) INVEST
		Coefficient	t statistic	Coefficient	t statistic
Intercept		12.0041***	11.61	5.2678***	4.02
ICWt	−	−2.1598***	−4.96	−1.5645***	−3.58
Qt− 1	+	2.3427***	9.62	2.2269***	7.71
LnTAt− 1	?	−1.5394***	−11.74	−1.9890***	−10.76
AGEt− 1	?	−0.0250*	−1.85	0.0036	0.37
LEVERAGEt− 1	−	−0.2363***	−5.19	−0.1808***	−4.62
CASHt− 1	+	9.4888***	7.78	9.5544***	9.43
CFOt	?	−13.3489***	−5.35	−15.5051***	−4.71
TANGIBLEt− 1	+	12.4858***	8.67	15.5046***	8.46
Zt− 1	−	−0.0958**	−2.26	−0.0937**	−1.85
DIVIDENDt− 1	?	−1.3637***	−3.75	−0.8928*	−2.31
StdINVESTt− 1	?	0.0307***	4.30	0.0126*	1.67
AQt− 1	?	−0.0020	−0.01	−0.0978	−0.39
INSTITUTEt− 1	?	1.9971*	1.74	0.4486	0.45
ANALYSTt− 1	?	0.2302***	6.79	0.1693***	5.85
NBSt				0.0444	0.62
FOREIGNSt				0.1086	0.33
MAt				14.3824***	14.40
RESTRUCTt				1.0479***	4.42
GROWTHt				7.5250***	10.43
INVENTORYt				−7.6339***	−4.07
LOSS% t				−0.1153	−0.11
AUDITORt				3.7569***	4.88
LITIGATIONt				2.7555***	5.19
Adjusted R2		.244		.338
n		16,555		16,275

Note. This table presents the coefficient estimates from regressions of firm INVEST on ICW disclosure and control variables. The sample consists of 16,555 firm-year observations with necessary data during 2004-2012. Indicator variables for industries and years are included as control variables. t statistics are based on standard errors clustered by firm and year. Variables are defined in Table 1, Panel A. INVEST = investment; ICW = internal control weakness.

***

, **, and * indicate statistical significance at the 1%, 5%, and 10% confidence levels, respectively.

The results on the other factors that influence firm investment are consistent with the literature and my expectations. The coefficient for growth opportunities (Q) is significantly positive. The coefficients for firm size (LnTA), firm age (AGE), and dividend payment indicator (DIVIDEND) are significantly negative, indicating that mature firms have lower new investment because they face a shrinking investment opportunity set. LEVERAGE has a negative coefficient, suggesting that high-leverage firms are likely to have difficulty raising additional capital to fund investment because of risk of financial distress. Indeed, the coefficient for bankruptcy risk (Z) is negative. To the extent that property, plant, and equipment can be used as collateral and thus reduce the risk of lenders, asset tangibility (TANGIBLE) increases firm investment. The positive coefficient for CASH suggests that firms with a larger cash holding can fund investment internally, and thus rely on external capital to a smaller extent. The negative coefficient for cash flows from operations (CFO) is driven by observations with negative cash flows. After removing 2,682 observations with negative cash flows, the coefficient for CFO becomes significantly positive at 30.7522, suggesting that firms with greater operating cash flows are able to fund investment using internally generated cash. The positive coefficient for the number of analysts following the firm (ANALYST) suggests that firms with greater analyst coverage invest more possibly because it is relatively easy for such firms to raise capital for investment.

Ashbaugh-Skaife et al. (2007) and Doyle et al. (2007b) find that a firm’s probability to have internal control weaknesses is associated with the firm’s operation complexity as reflected by the number of business segments (NBS) and the existence of foreign sales (FOREIGNS), changes in business organization as a result of mergers or acquisitions (MA) or restructure (RESTRUCT), potential accounting measurement errors proxied by sales growth rate (GROWTH) and inventory level (INVENTORY), frequency of losses in previous years (LOSS%), auditor ability and scrutiny measured by the use of one of the largest six auditors (AUDITOR), monitoring by institutional investors (INSTITUTE), litigation risk (LITIGATION), firm size (LnTA), financial distress risk (Z), and firm age (AGE). To test whether the relation between internal control quality and firm investment is incremental to these risk factors, I re-estimate Regression 1 by adding those of the above variables that are not already included in the regression as additional control variables. Estimation results are reported in column 2 of Table 2. ⁹ The coefficient for the disclosure of internal control weaknesses (ICW) is still significantly negative, which is consistent with H1, though with a smaller magnitude at −1.5645. The coefficients for AGE and the standard deviation of investment over the previous 5 years (StdINVEST) become insignificant. The results for the other control variables are qualitatively similar to those in column 1.

Association Between Internal Control Weakness Disclosure and Discretionary Investment

The results in Table 2 indicate that the disclosure of internal control weakness is negatively associated with firm investment. I next examine whether weakness disclosure affects the deviation of firm investment from the expected level. First, I replace the dependent variable in Regression 1 by discretionary investment and report the estimation results in Table 3. Discretionary investment is measured by discretionary investment from Regression 2 (dINVEST) and industry-adjusted investment (INVEST_adj) in columns 1 and 2, respectively. The determinants for internal control quality are controlled for in both columns. The coefficient for the disclosure of internal control weaknesses (ICW) is significantly negative in both columns. For example, the coefficient for ICW is −1.4708 in column 1, suggesting that the discretionary investment of firms that receive adverse internal control opinions is lower than that of the other firms by 1.4708% of assets at the beginning of the year. Given that the first and third quartiles of discretionary investment are −7.6705% and 3.0178%, respectively, a decrease of 1.4708% represents 13.76% of the interquartile range. Similarly, the coefficient for ICW in column 2, −1.7177, represents a difference of 13.39% of the interquartile range of industry-adjusted investment. The results for control variables are similar to those reported in Table 2. The results in Table 3 provide further support to H1 that firms that receive adverse internal control opinions have lower investment than firms that receive clean opinions.

OPEN IN VIEWER

Table 3.

Internal Control Weakness Disclosure and Discretionary Investment.

		(1) dINVEST		(2) INVEST_adj
Variables	Predicted sign	Coefficient	t statistic	Coefficient	t statistic
Intercept		−0.7235	−0.51	−6.6748***	−4.29
ICWt	−	−1.4708***	−4.05	−1.7177***	−3.90
Qt− 1	+	0.7360***	3.49	2.2307***	7.41
LnTAt− 1	?	−2.0529***	−11.05	−1.9371***	−11.20
AGEt− 1	?	0.0059	0.70	0.0025	0.28
LEVERAGEt− 1	−	−0.1284***	−3.66	−0.1726***	−4.45
CASHt− 1	+	10.6716***	11.66	9.1399***	8.61
CFOt	?	−12.7191***	−6.16	−15.5970***	−4.68
TANGIBLEt− 1	+	14.9971***	9.87	15.6803***	8.36
Zt− 1	−	−0.0782	−1.50	−0.0934*	−1.70
DIVIDENDt− 1	?	−0.9787***	−3.21	−0.9422**	−2.37
StdINVESTt− 1	?	0.0137*	1.87	0.0133*	1.70
AQt− 1	?	−0.0835	−0.42	−0.0028	−0.01
INSTITUTEt− 1	?	0.4460	0.49	0.4038	0.43
ANALYSTt− 1	?	0.2163***	6.00	0.1615***	5.45
NBSt		0.0580	0.79	0.0465	0.66
FOREIGNSt		0.0313	0.11	0.1144	0.35
MAt		13.2753***	14.69	14.2757***	14.74
RESTRUCTt		1.0319***	4.08	1.0211***	4.08
GROWTHt		6.5753***	9.51	7.3602***	9.38
INVENTORYt		−7.0861***	−3.79	−8.0828***	−4.18
LOSS% t		0.2089	0.26	0.0086	0.01
AUDITORt		3.3775***	5.01	3.8862***	4.95
LITIGATIONt		2.3400***	3.82	2.8572***	5.31
Adjusted R2		.253		.282
n		16,180		16,243

Note. This table presents the coefficient estimates from regressions of discretionary investment on ICW disclosure and control variables. The dependent variables are dINVEST and INVEST_adj in columns 1 and 2, respectively. The sample consists of firm-year observations with necessary data during 2004-2012. Indicator variables for industries and years are included as control variables. t statistics are based on standard errors clustered by firm and year. Variables are defined in Table 1, Panel A. ICW = internal control weakness; dINVEST = discretionary investment; INVEST_adj = industry-adjusted investment.

***

, **, and * indicate statistical significance at the 1%, 5%, and 10% confidence levels, respectively.

Disclosure and Correction of Internal Control Weaknesses and Firm Investment

Time-series analysis

To assess the existence of a causal relation between internal control weakness disclosure and firm investment, I first conduct a time-series analysis. If the relatively low investment for firms that receive adverse control opinions is indeed a result of the weakness revelation, I expect these firms to decrease investment after, instead of during, the first year that they receive an adverse opinion. I further expect these firms to increase investment after the control weaknesses are remediated as reflected by the receipt of clean opinions.

Figure 1 presents the average firm investment (Panel A) and discretionary investment (Panel B) of ineffective-internal-control firms for the years around initial weakness revelation and subsequent weakness correction. The graphs show that firm investment (or discretionary investment) decreases slightly in the first year that the firms receive adverse internal control opinions (i.e., ICW1), but it decreases significantly in the second or a later year that these firms receive adverse opinions (i.e., ICW2). The investment increases but remains low during the year that these firms remediate the control weaknesses, as evidenced by the receipt of the first clean opinions (i.e., ICWCOR1). Discretionary investment increases further in the second or a later year that these firms get clean opinions (i.e., ICWCOR2). The pattern of investment changes shown in Figure 1 is consistent with my expectation.

OPEN IN VIEWER

Figure 1.

Firm investment around the initial weakness revelation and subsequent correction.

Estimation results for Regression 3, with the determinants for internal control quality controlled for, are reported in Table 4. This analysis is based on firm-year observations for ineffective-internal-control firms before the weakness revelation, during the weakness revelation or disclosure, and after the weakness remediation. ¹⁰ For firms that receive the first adverse auditor internal control opinions in 2004, I use 2003 as the year prior to weakness revelation. ¹¹ The dependent variable is actual investment (INVEST) and discretionary investment (dINVEST) in columns 1 and 2, respectively. Among the 3,825 observations in column 1, 628 have ICW1 = 1; 246 have ICW2 = 1; 449 have ICWCOR1 = 1; and 1,570 have ICWCOR2 = 1. Consistent with my expectation, the coefficient for the indicator variable for the first adverse internal control opinion (ICW1) is insignificant in both columns. After the control weaknesses are disclosed, firm investment is lower, no matter whether the control problems are corrected in the following year, as we can see from the negative coefficients for the second or later year that a firm receives adverse opinions (ICW2) and for the first year that an ineffective-internal-control firm receives a clean opinion (ICWCOR1).

OPEN IN VIEWER

Table 4.

Existence and Correction of Internal Control Weaknesses and Firm Investment.

Variables	Predicted sign	(1) INVEST		(2) dINVEST
		Coefficient	t statistic	Coefficient	t statistic
Intercept		9.5370***	3.88	4.6355**	2.15
ICW1t	?	−0.8336	−1.30	−0.6789	−1.13
ICW2t	−	−3.8586***	−3.29	−3.5635***	−3.35
ICWCOR1t	−	−2.7293***	−3.59	−2.2709***	−3.00
ICWCOR2t	?	−1.3601	−1.39	−1.4357	−1.34
Qt− 1	+	2.0870***	4.70	0.7724**	2.22
LnTAt− 1	?	−2.0597***	−5.68	−2.1910***	−6.23
AGEt− 1	?	0.0232	1.22	0.0221	1.15
LEVERAGEt− 1	−	−0.2772***	−2.69	−0.1955**	−2.10
CASHt− 1	+	11.7442***	5.53	12.1006***	6.84
CFOt	?	−8.6533	−1.14	−5.9591	−0.93
TANGIBLEt− 1	+	13.4382***	6.15	13.3013***	6.50
Zt− 1	−	−0.0848	−0.87	−0.1380*	−1.72
DIVIDENDt− 1	?	−0.8751	−1.34	−1.1773**	−2.26
StdINVESTt− 1	?	0.0234	1.35	0.0207	1.28
AQt− 1	?	−0.5322	−0.76	−0.3170	−0.51
INSTITUTEt− 1	?	0.5200	0.36	0.9065	0.67
ANALYSTt− 1	?	0.1438**	2.12	0.1950***	2.76
NBSt		−0.1490	−1.11	−0.1390	−1.00
FOREIGNSt		0.8341	1.00	0.9856	1.26
MAt		14.0675***	12.80	13.3943***	12.81
RESTRUCTt		1.2862***	2.61	1.4280***	3.57
GROWTHt		9.2424***	7.32	7.7706***	7.77
INVENTORYt		−0.7874	−0.27	−1.3503	−0.56
LOSS% t		−0.1989	−0.11	−0.2121	−0.15
AUDITORt		1.7209	1.54	1.4022	1.49
LITIGATIONt		2.0556*	1.93	1.7655*	1.84
Adjusted R2		.348		.286
n		3,825		3,793
Tests on coefficients		Difference	Wald statistic	Difference	Wald statistic
Test 1: ICW1−ICW2 > 0		3.0250***	6.660	2.8846***	8.160
Test 2: ICW2−ICWCOR1 < 0		−1.1293	0.850	−1.2926	1.480
Test 3: ICWCOR2−ICWCOR1 > 0		1.3692*	2.210	0.8352	1.030

Note. This table presents the coefficient estimates from regressions of firm INVEST (column 1) or dINVEST (column 2) on disclosure and correction of internal control weaknesses and control variables. The analysis is based on 3,825 firm-years for firms with internal control weaknesses during the sample period. Indicator variables for industries and years are included as control variables. t statistics are based on standard errors clustered by firm and year. ICW1 is equal to 1 if year t is the first year that a firm receives an adverse auditor opinion on internal control. ICW2 is equal to 1 if year t is the second or a later year that a firm receives an adverse auditor opinion on internal control. ICWCOR1 is equal to 1 if year t is the year during which a firm’s internal control weaknesses are corrected as evidenced by the receipt of a clean auditor opinion on internal control after the weakness disclosure. ICWCOR2 is equal to 1 if year t is the second or a later year that an ineffective-internal-control firm receives a clean auditor opinion on internal control. Other variables are defined in Table 1, Panel A. INVEST = investment; dINVEST = discretionary investment.

***

, **, and * indicate statistical significance at the 1%, 5%, and 10% confidence levels, respectively.

The coefficient for ICW2 is not only statistically lower than 0 but also significantly lower than the coefficient for ICW1. For example, the coefficient for ICW2 is lower than the coefficient for ICW1 by 3.0250 in column 1 (Test 1: ICW1−ICW2 > 0), indicating that investment of ineffective-internal-control firms is lower by 3.0250% of beginning assets during the second or a later year than the first year that they receive adverse opinions. ¹² This decrease is economically significant because it represents 19.65% (=3.0250% / 15.3973%) of average investment for firms that do not receive an adverse opinion in either the previous year or the current year. The coefficients for ICW1 and ICW2 and the difference between them suggest that ineffective-internal-control firms indeed decrease investment following, instead of during, the year with the first adverse control opinion. This result is consistent with the finding of Gordon and Wilford (2012) that the increase in the cost of equity is greater as the material weaknesses persist into multiple years.

During the year that ineffective-internal-control firms remediate the weaknesses as evidenced by the receipt of the first clean opinions, their investment is still significantly lower than that prior to the first weakness disclosure, as reflected by the negative coefficient, −2.7293, for ICWCOR1. Test 2 indicates that the coefficient for ICWCOR1 is not statistically greater than the coefficient for ICW2, suggesting that ineffective-internal-control firms do not increase investment during the year in which control weaknesses are corrected probably because outside investors do not know the weakness remediation until the auditor issues a clean opinion at or after the end of the year. Indeed, for the second or a later year that ineffective-internal-control firms receive clean opinions (i.e., ICWCOR2 = 1), firm investment increases and is not significantly different from the level prior to the first weakness disclosure, as we can see from the insignificant coefficient for ICWCOR2. Test 3 on the relation between the coefficients for ICWCOR2 and ICWCOR1 indicates that the coefficient for ICWCOR2 is greater than the coefficient for ICWCOR1 in column 1, suggesting that firm investment indeed increases after the receipt of clean internal control opinions. ¹³

Change analysis

I next conduct a change analysis to examine whether the change in firm investment in the current year is associated with the change in internal control quality in the previous year. Compared with the investment increase for firms that receive clean opinions throughout the sample period and observations for ineffective-internal-control firms before the first adverse opinions, I expect the investment increase to be smaller after a firm receives an adverse opinion for the first time (i.e., UNQ_ADV_t−1 = 1) or continues to receive adverse opinions (i.e., ADV_ADV_t−1 = 1). I also expect that, for ineffective-internal-control firms, firm investment change starts getting back to normal after they receive clean opinions (i.e., ADV_UNQ_t−1 = 1 or UNQ_UNQ_t−1 = 1).

The estimation results for Regression 4 are reported in Table 5. The sample period for the change analysis is 2006 to 2012 because auditor control opinions for the previous 2 years are used to define the opinion changes in year t− 1. The dependent variable is the change in total investment (ΔINVEST) in column 1 and the change in discretionary investment (ΔdINVEST) in column 2. Changes in the determinants for the existence of control weaknesses are omitted because they tend to be sticky from year to year. For example, most firms have the same NBS in both years. Consistent with my expectation, the coefficients for UNQ_ADV_t−1 and ADV_ADV_t−1 in column 1 are significantly negative at −2.6463 and −3.7027, respectively, suggesting that smaller investment changes after control weaknesses are revealed. The negative coefficient −2.6207 for ADV_UNQ_t−1 suggests that, after an ineffective-internal-control firm receives a clean opinion, the investment change is still lower than that for firms with effective internal control, but the difference gets smaller. After an ineffective-internal-control firm receives at least two clean opinions, the firm’s investment change is not significantly different from that for firms with effective internal control, as indicated by the insignificant coefficient for UNQ_UNQ_t−1. The results for the four internal-control-opinion change variables in column 2 are similar. For the control variables, I find that the change in investment is positively associated with the change in investment opportunity set (ΔQ), the change in cash holding (ΔCASH), the change in tangibility of assets (ΔTANGIBLE), and the changes in institutional investors’ shareholding (ΔINSTITUTE) and analyst coverage (ΔANALYST), while negatively associated with the change in leverage (ΔLEVERAGE).

OPEN IN VIEWER

Table 5.

Change in Internal Control Quality and Change in Firm Investment.

Variables	Predicted sign	(1) ΔINVEST		(2) ΔdINVEST
		Coefficient	t statistic	Coefficient	t statistic
Intercept		3.2216***	9.22	2.4186***	3.18
UNQ_ADVt− 1	−	−2.6463**	−2.08	−2.4541***	−2.62
ADV_ADVt− 1	−	−3.7027***	−4.69	−3.0968***	−7.66
ADV_UNQt− 1	?	−2.6207***	−3.14	−2.1694***	−2.84
UNQ_UNQt− 1	?	0.0086	0.01	0.0651	0.10
ΔQt−1	+	1.3795***	3.16	−0.0798	−0.23
ΔLnTAt−1	?	−34.3367***	−13.87	−29.6259***	−13.64
ΔLEVERAGEt−1	−	−0.8064***	−2.72	−0.5965***	−3.37
ΔCASHt−1	+	48.3268***	9.10	46.8704***	9.79
ΔCFOt	?	−0.6356	−0.16	0.7638	0.34
ΔTANGIBLEt−1	+	−7.3512	−0.82	−2.2553	−0.26
ΔZt−1	−	−0.0266	−0.43	−0.0236	−0.39
ΔDIVIDENDt−1	?	0.2320	0.38	0.4064	0.99
ΔStdINVESTt−1	?	−0.1577**	−2.25	−0.1051**	−2.17
ΔAQt−1	?	0.5496	1.09	0.4893*	1.78
ΔINSTITUTEt−1	?	3.8938***	1.89	4.4975**	2.25
ΔANALYSTt−1	?	0.2613***	2.86	0.2384***	3.95
Adjusted R2		.246		.225
n		11,720		11,612

Note. This table presents the coefficient estimates from regressions of change in firm ΔINVEST (column 1) or ΔdINVEST (column 2) on change in internal control opinions and control variables. The analysis is based on observations with necessary data during 2006-2012. Indicator variables for industries and years are included as control variables. t statistics are based on standard errors clustered by firm and year. UNQ_ADV_t−1 takes the value of 1 if the firm receives a clean internal control opinion in year t− 2 but an adverse opinion in year t− 1; ADV_ADV_t−1 takes the value of 1, if the firm receives adverse opinions in both t− 2 and t− 1; ADV_UNQ_t−1 takes the value of 1, if the firm receives an adverse opinion in year t− 2 but a clean opinion in year t− 1; UNQ_UNQ_t−1 takes the value of 1 if the firm gets an adverse opinion before year t− 2, but receives clean opinions in both year t− 2 and year t− 1. ΔX refers to the change in variable X from the previous year to the current year. Variables are defined in Table 1, Panel A. INVEST = investment; dINVEST = discretionary investment.

***

, **, and * indicate statistical significance at the 1%, 5%, and 10% confidence levels, respectively.

Taken together, the results in Tables 4 and 5 indicate that firm investment decreases after a firm’s internal control is revealed to be ineffective and increases after the control weaknesses are corrected. These results are consistent with the model by Lambert et al. (2007) that deterioration in the quality of accounting system has a negative impact on the firm’s optimal investment level. These results suggest that the disclosed quality of internal control over financial reporting required by Section 404 of SOX has an impact on firms’ real decisions.

Effect of Internal Control Disclosures on Different Components of Firm Investment

In H2, I expect that the disclosure of internal control weaknesses has a greater impact on firm investments that are more risky. To test H2, I divide total investment into two components: the less risky component (i.e., net capital expenditures) and the more risky component (i.e., acquisitions and R&D expenses). I re-estimate Regressions 1, 3, and 4 by using each component or the change in each component as the dependent variable and report the results in Table 6. In Panel A, I find that firms that are revealed to have internal control weaknesses (i.e., ICW = 1) do not have significantly different net capital expenditures, but they have significantly lower acquisitions and R&D expenses as we can see from the negative coefficient −1.6792 for ICW in column 2.

OPEN IN VIEWER

Table 6.

Internal Control Weakness Disclosure and Firm Investment Components.

Panel A: Existence of Internal Control Weaknesses.
Variables	Predicted sign	(1) NetCAPEX		(2) ACQRD
		Coefficient	t statistic	Coefficient	t statistic
Intercept		1.2556	1.11	3.6918***	2.81
ICWt	−	0.0035	0.01	−1.6792***	−4.65
Qt −1	+	0.7213*	1.86	1.7923***	7.24
LnTAt −1	?	−0.4327***	−3.16	−1.5217***	−9.69
AGEt −1	?	−0.0096*	−1.81	0.0116	1.41
LEVERAGEt −1	−	−0.1498***	−3.70	−0.0439	−1.44
CASHt −1	+	0.0479	0.07	8.0079***	6.36
CFOt	?	2.9751***	3.08	−21.8078***	−3.96
TANGIBLEt −1	+	15.0339***	12.30	0.0849	0.06
Zt −1	−	0.0015	0.07	−0.1000***	−2.62
DIVIDENDt− 1	?	−0.6589***	−2.69	−0.2158	−0.57
StdINVESTt− 1	?	0.0725***	3.38	0.0005***	4.12
AQt− 1	?	0.0442	0.49	−0.1112	−0.47
INSTITUTEt− 1	?	−0.0219	−0.07	0.5881	0.60
ANALYSTt− 1	?	0.0566**	2.22	0.1002***	3.72
NBSt		0.0267	0.76	0.0303***	0.47
FOREIGNSt		−0.0368	−0.34	0.1452	0.43
MAt		−0.0251	−0.23	14.5605***	14.52
RESTRUCTt		0.3147***	3.68	0.7494***	3.28
GROWTHt		2.5205***	7.39	4.8452***	8.31
INVENTORYt		−0.8489	−1.12	−7.5793***	−4.30
LOSS%t		−0.9822***	−4.01	0.3279	0.25
AUDITORt		−0.3499	−0.96	4.3500***	6.25
LITIGATIONt		−0.6524***	−2.98	3.6413***	6.02
Adjusted R2		.422		.342
n		16,275		16,275

Note. This table presents the coefficient estimates from regressions of NetCAPEX (column 1) or ACQRD expenses (column 2) on ICW disclosure and control variables. The analysis is based on 16,275 firm-year observations with necessary data during 2004-2012. Indicator variables for industries and years are included as control variables. t statistics are based on standard errors clustered by firm and year. StdINVEST is the standard deviation of NetCAPEX in column 1 and the standard deviation of ACQRD in column 2. Variables are defined in Table 1, Panel A. NetCAPEX = net capital expenditures; ACQRD = acquisitions and R&D; ICW = internal control weakness.

***

, **, and * indicate statistical significance at the 1%, 5%, and 10% confidence levels, respectively.

OPEN IN VIEWER

Panel B: Revelation and Correction of Control Weaknesses and Investment Components.
Variables	Predicted sign	(1) NetCAPEX		(2) ACQRD
		Coefficient	t statistic	Coefficient	t statistic
Intercept		1.1377	1.11	3.7836	1.32
ICW1t	?	−0.2561	−0.56	−0.6542	−1.43
ICW2t	−	−0.6422	−1.42	−3.2585***	−3.20
ICWCOR1t	−	−0.3386	−0.76	−2.3951***	−3.96
ICWCOR2t	?	−0.5479	−0.99	−0.8305	−0.87
Qt− 1	+	0.4243***	2.67	1.6974***	4.62
LnTAt− 1	?	−0.5490***	−3.42	−1.5015***	−4.36
AGEt− 1	?	0.0123	1.23	0.0089	0.49
LEVERAGEt− 1	−	−0.1375***	−2.78	−0.1505	−1.41
CASHt− 1	+	2.3815***	3.12	8.9860***	4.39
CFOt	?	5.2908***	4.77	−15.0319*	−1.83
TANGIBLEt− 1	+	13.1757***	7.22	0.1076	0.05
Zt− 1	−	−0.0262	−0.97	−0.0473	−0.53
DIVIDENDt− 1	?	−0.5855*	−1.80	0.0154	0.02
StdINVESTt− 1	?	0.0498	1.57	0.0063	0.67
AQt− 1	?	−0.0121	−0.04	−0.5733	−0.97
INSTITUTEt− 1	?	−0.3562	−0.51	0.8243	0.67
ANALYSTt− 1	?	0.0826**	2.01	0.0575	0.88
NBSt		0.0069	0.11	−0.1544	−1.17
FOREIGNSt		−0.0689	−0.36	0.8827	1.07
MAt		−0.1027	−0.51	14.4075***	12.86
RESTRUCTt		0.1778	1.21	0.9843*	1.92
GROWTHt		2.9779***	6.40	6.5611***	5.59
INVENTORYt		1.9541	1.41	−3.0985	−1.14
LOSS% t		−0.9378**	−2.18	0.5763	0.34
AUDITORt		0.0350	0.08	1.9339*	1.86
LITIGATIONt		−1.1541***	−3.47	3.4944***	2.82
Adjusted R2		.455		.340
n		3,825		3,825
Tests on coefficients		Difference	Wald statistic	Difference	Wald statistic
Test 1: ICW1−ICW2 > 0		0.3861	0.800	2.6043**	4.840
Test 2: ICW2−ICWCOR1 < 0		−0.3036	0.450	−0.8634	0.490
Test 3: ICWCOR2−ICWCOR1 > 0		−0.2093	0.450	1.5646**	2.900

Note. This table presents the coefficient estimates from regressions of NetCAPEX (column 1) or ACQRD expenses (column 2) on disclosure and correction of internal control weaknesses and control variables. The analysis is based on 3,825 firm-years for firms with internal control weaknesses during the sample period. Indicator variables for industries and years are included as control variables. t statistics are based on standard errors clustered by firm and year. StdINVEST is the standard deviation of NetCAPEX in column 1 and the standard deviation of ACQRD in column 2. ICW1 is equal to 1 if year t is the first year that a firm receives an adverse auditor opinion on internal control. ICW2 is equal to 1 if year t is the second or a later year that a firm receives an adverse auditor opinion on internal control. ICWCOR1 is equal to 1 if year t is the year during which a firm’s internal control weaknesses are corrected as evidenced by the receipt of a clean auditor opinion on internal control after the weakness disclosure. ICWCOR2 is equal to 1 if year t is the second or a later year that an ineffective-internal-control firm receives a clean auditor opinion on internal control. Other variables are defined in Table 1, Panel A. NetCAPEX = net capital expenditures; ACQRD = acquisitions and R&D; ICW = internal control weakness.

***

, **, and * indicate statistical significance at the 1%, 5%, and 10% confidence levels, respectively.

OPEN IN VIEWER

Panel C: Change in Internal Control Quality and Change in Firm Investment Components.
Variables	Predicted sign	(1) ΔNetCAPEX		(2) ΔACQRD
		Coefficient	t statistic	Coefficient	t statistic
Intercept		0.7534***	6.07	3.6778***	10.20
UNQ_ADVt− 1	−	−0.0931	−0.30	−2.5930**	−2.00
ADV_ADVt− 1	−	−0.5708***	−3.50	−3.4617***	−4.57
ADV_UNQt− 1	?	−0.4346	−1.55	−2.0189***	−3.07
UNQ_UNQt− 1	?	0.1862	1.40	−0.3869	−0.91
ΔQt−1	+	0.8970**	2.31	0.8624**	1.98
ΔLnTAt−1	?	−5.3208***	−5.29	−32.2640***	−13.16
ΔLEVERAGEt−1	−	−0.1684	−1.48	−0.8920***	−2.74
ΔCASHt−1	+	−2.1039	−1.60	50.5093***	7.42
ΔCFOt	?	0.2387	0.25	−3.8599	−0.50
ΔTANGIBLEt−1	+	−25.7841***	−4.46	19.4924**	1.96
ΔZt−1	−	0.0023	0.06	−0.0323	−0.91
ΔDIVIDENDt−1	?	−0.3087	−1.06	0.8968	1.43
ΔStdINVESTt−1	?	−0.1552	−1.34	−0.0133	−0.91
ΔAQt−1	?	0.1530	1.36	0.5217	1.11
ΔINSTITUTEt−1	?	0.4331	1.04	2.8287	1.06
ΔANALYSTt−1	?	0.0257	1.47	0.2245**	2.33
Adjusted R2		.149		.191
n		11,720		11,720

Note. This table presents the coefficient estimates from regressions of change in ΔNetCAPEX (column 1) or ΔACQRD expenses (column 2) on change in internal control opinions and control variables. The analysis is based on observations with necessary data during 2006-2012. Indicator variables for industries and years are included as control variables. t statistics are based on standard errors clustered by firm and year. UNQ_ADV_t−1 takes the value of 1 if the firm receives a clean internal control opinion in year t− 2 but an adverse opinion in year t− 1; ADV_ADV_t−1 takes the value of 1 if the firm receives adverse opinions in both t− 2 and t− 1; ADV_UNQ_t−1 takes the value of 1 if the firm receives an adverse opinion in year t− 2 but a clean opinion in year t− 1; UNQ_UNQ_t−1 takes the value of 1 if the firm gets an adverse opinion before year t− 2, but receives clean opinions in both year t− 2 and year t− 1. ΔX refers to the change in variable X from the previous year to the current year. StdINVEST is the standard deviation of NetCAPEX in column 1 and the standard deviation of ACQRD in column 2. Variables are defined in Table 1, Panel A. NetCAPEX = net capital expenditures; ACQRD = acquisitions and R&D.

***

, **, and * indicate statistical significance at the 1%, 5%, and 10% confidence levels, respectively.

For the time-series analysis in Panel B, I find that, in column 1, there is no significant change in net capital expenditures over time when there are changes in internal control quality. However, in column 2, acquisitions and R&D expenses decrease after a firm is disclosed to have internal control weaknesses (i.e., negative coefficients for ICW2 and ICWCOR1) and increase after the weaknesses are remediated (i.e., a greater coefficient for ICWCOR2 than for ICWCOR1). The insignificant coefficient for ICWCOR2 in column 2 indicates that acquisitions, on average, get back to the level before control weaknesses are disclosed after the firms have received at least one clean opinion, similar to the finding for total investment in Table 4.

For the change analysis, I examine whether the change in each of the two investment components in the current year is associated with the change in internal control quality in the previous year and report the results in Panel C. For net capital expenditures in column 1, I find an insignificant coefficient for UNQ_ADV_t−1 and a significantly negative coefficient −0.5708 for ADV_ADV_t−1, indicating that net capital expenditures decrease only after the firm repeatedly gets adverse control opinions. In contrast, for acquisitions and R&D expenses in column 2, the coefficients for UNQ_ADV_t−1 and ADV_ADV_t−1 are both significantly negative (−2.5930 and −3.4617, respectively), suggesting that firms’ risky investment components decrease right after the first weakness disclosure and become even lower if the firms get additional adverse control opinions. In addition, acquisitions and R&D expenses continue to be lower even after these firms receive the first clean opinions, as we can see from the negative coefficient −2.0189 for ADV_UNQ_t−1. The insignificant coefficient for UNQ_UNQ_t−1 suggests that acquisitions and R&D expenses get back to the normal level after ineffective-internal-control firms get at least two clean control opinions. These results suggest that the revelation of control weaknesses (UNQ_ADV_t−1) and the receipt of repeated adverse control opinions (ADV_ADV_t−1) have a greater effect on the more risky components of firm investment (i.e., acquisitions and R&D expenses). These results also indicate that the results in Table 5 for total investment are driven by acquisitions and R&D expenses.

Taken together, the results in Table 6 indicate that the negative impact of internal control weakness disclosures on firm investment is more significant for the investment components that are more risky, consistent with H2.

Additional Tests

My primary objective is to examine whether the disclosed information on internal control quality has any real effect on the level of firm investment. Two related questions are how this effect is related to investment efficiency and whether this effect varies with firm characteristics associated with investment efficiency. To address the first question, I divide the sample into overinvestment and underinvestment groups using discretionary investment in the year prior to the weakness revelation and find in untabulated analysis that the magnitude of discretionary investment decreases for the overinvestment group after the weakness revelation, while the magnitude does not change significantly for the underinvestment group. For the second question on whether the association between control weakness disclosures and firm investment varies across firms, I use variables suggested by prior studies that are associated with overinvestment to divide the sample into two groups to examine whether firm investment decrease is greater for firms that tend to overinvest. The partition variables that I use are firm size (i.e., total assets) and firm age that are used to measure financial constraints in the literature (e.g., Hadlock & Pierce, 2012; Kaplan & Zingales, 1997; Whited & Wu, 2006). ¹⁴ Firms with a size (or age) in the upper two quintiles are expected to be more likely to overinvest than the other firms. I re-estimate Regression 1 for firms that tend to overinvest and the other firms and report the results in Table 7. I find that the decrease in investment after the weakness revelation is greater for firms that are more likely to overinvest. These two additional tests suggest that internal control weakness disclosures help reduce overinvestment and thus improve firm investment efficiency.

OPEN IN VIEWER

Table 7.

Tendency to Overinvest and Firm Investment Response to Control Weakness Disclosures.

Panel A: Firm Size.
Variable	Firms likely to overinvest	Other firms	Difference
	Coefficient	Coefficient
ICWt	−2.9855	−1.8476	−1.1379*

Panel B: Firm Age.
Variable	Firms likely to overinvest	Other firms	Difference
	Coefficient	Coefficient
ICWt	−3.0894	−1.6844	−1.4050**

Note. This table presents and compares the coefficients for ICW from regressions of firm INVEST on ICW disclosure and control variables for firms that tend to overinvest and the other firms. Firms with a size (or age) in the upper two quintiles are defined as those that tend to overinvest in Panel A (or Panel B). Firm size is total assets. Firm age is defined as the current year minus the first year that the firm has non-missing stock price at CRSP. The sample consists of 16,555 firm-year observations with necessary data during 2004-2012. ICW = internal control weakness; INVEST = investment; CRSP = Center for Research in Security Prices.

***

, **, and * indicate statistical significance at the 1%, 5%, and 10% confidence levels, respectively.

Under Auditing Standard No. 2 (Public Company Accounting Oversight Board [PCAOB], 2004), besides assessing the internal control of the company, the auditor should also attest to and report on management assessment. This requirement is subsequently removed by Auditing Standard No. 5 (PCAOB, 2007). Studies have shown that the implementation of AS5 is associated with a reduction in audit fees, consistent with improvements in audit efficiency (Doogar, Sivadasan, & Solomon, 2010; Krishnan, Krishnan, & Song, 2011). To examine whether the move from AS2 to AS5 has any effect on the relation between internal control weakness disclosure (ICW) and firm investment, in untabulated analysis, I add an indicator variable AS5 that takes the value of 1 for sample observations with a fiscal year-end on or after November 15, 2007 (the AS5 implementation date) and an interaction term ICW×AS5 into the regressions. I find a positive coefficient on AS5, indicating a higher investment level in more recent years (i.e., post AS5), which is consistent with the sample distribution by the year reported in Table 1, Panel B (i.e., fewer firms receive adverse auditor internal control opinions in recent years) and the result for H1 (i.e., firms that receive adverse internal control opinions have lower investment). The coefficient on ICW×AS5 is insignificant. I continue to find that firms with internal control weaknesses have significantly lower investment or discretionary investment.

Prior studies (e.g., Krishnan, 2005; Zhang, Zhou, & Zhou, 2007) document that firms with weaker governance are more likely to have internal control weaknesses. To examine whether the relation between ICW and investment is affected by governance/monitoring mechanisms, I conduct two tests. First, I add interactions of ICW with analyst following (ANALYST) and institutional shareholding (INSTITUTE) to the regressions. Neither of these interaction terms has a significant coefficient. Second, Biddle et al. (2009) use the anti-takeover protection index (i.e., G-index) computed by Gompers, Ishii, and Metrick (2003) as another governance mechanism that potentially affects investment. I do not include G-index in the regressions reported in the tables because the latest G-index data are available only for years up to 2007, while my sample period is from 2004 to 2012. In untabulated tests, for the shorter period with G-index data, I add G-index and its interaction with ICW (ICW×G-index) to the regressions. The coefficient on ICW×G-index is insignificant. In both tests, ICW is still negatively associated with firm investment or discretionary investment.