Financial Reporting Quality and Investment Efficiency of Private Firms in Emerging Markets

Feng Chen Rotman School of Management University of Toronto feng.chen@rotman.utoronto.ca Ole-Kristian Hope Rotman School of Management University of Toronto okhope@rotman.utoronto.ca Qingyuan Li School of Economics and Management Wuhan University qyli@whu.edu.cn Xin Wang School of Accountancy Chinese University of Hong Kong wangx@cuhk.edu.hk March 16, 2010

Financial Reporting Quality and Investment Efficiency of Private Firms in Emerging Markets

Abstract Prior research shows that financial reporting quality is positively related to investment efficiency for large, U.S. publicly traded companies. However, it is not obvious whether this finding extends to smaller, privately-held companies in emerging markets (i.e., a very different environment characterized by low investor protection and more concentrated ownership structures). Using firm-level survey data from the World Banks Enterprise Survey, this paper studies the relation between accounting quality and investment efficiency for a sample of relatively small private firms in emerging markets. We find strong evidence that accounting quality positively affects investment efficiency (i.e., is negatively related to both underinvestment and overinvestment) for our sample of firms. Second, we show that the relation between financial reporting quality and investment efficiency is stronger if a firms investment is funded relatively more through bank financing than other sources of financing. Third, we document that, for firms with the strongest incentives to manage earnings for tax purposes, the positive association between accounting quality and investment efficiency is reduced. Such a connection between tax-minimization incentives and the informational role of earnings has often been asserted in the literature, but to date there is limited empirical evidence on this issue.

Keywords: Investment Efficiency, Over- and Under-Investment, Accounting Quality, Private Firms, Emerging Markets, Financing Sources, Tax Incentives

1. Introduction One objective of financial reporting information is to facilitate the efficient allocation of capital in the economy. An important aspect of this role is to improve firms investment decisions. Specifically, theory suggests that improved financial transparency has the potential to alleviate both over- and underinvestment problems and recent studies support this prediction (e.g., Biddle and Hilary 2006; Hope and Thomas 2008; McNichols and Stubben 2008; Biddle, Hilary, and Verdi 2009). This evidence, however, has been mostly limited to large, publicly traded companies in the United States. It is not obvious that the findings documented in prior studies extend to different settings. In this study, we extend the literature by examining the relation between financial reporting quality and investment efficiency for a sample of private firms in emerging markets. It is important to study private firms from emerging markets for several reasons. First, private firms (i.e., firms that are not traded on public stock exchanges) are the predominant organization in most countries. More than 99% of limited liability companies are not listed on a stock exchange for most countries (e.g., Nagar, Petroni, and Wolfenzon 2005; Berzins, Bhren, and Rydland 2008). In the aggregate, non-listed firms have about four times more employees, three times higher revenues, and twice the amount of assets than listed firms (Berzins et al. 2008). In spite of their economic importance and likely differences from public companies, comparatively little is known about financial reporting by private firms. In particular, based on prior literature discussed further in Section 2, it is not clear whether accounting information plays the same role in mitigating information asymmetries for private firms as it does for public firms. Second, compared with a large literature on developed countries accounting systems and

managers reporting incentives, much less is known about the role of accounting in developing countries. The extent to which findings for public firms from developed countries apply to our setting is unclear. Some authors (e.g., Ball, Kothari, and Robin 2000) argue that the role of accounting information is more limited in environments that are characterized by low investor protection and more concentrated ownership structures. However, due to the lack of alternative sources of information available, accounting information may be especially useful for our sample of relatively smaller, privately-held firms.1 Ultimately it is an empirical question whether accounting information and especially the quality of reported earnings matters for investment decisions in private firms from emerging markets. Our research also addresses an area of direct interest for the World Bank. Guidance as to investment efficiency in emerging markets, and what factors affect this efficiency, is directly relevant to the World Banks mission of offering assistance to developing countries around the world.2 Furthermore, financing and tax factors influence the financial reporting incentives of our sample firms. For example, high quality accounting information is likely more desirable to mitigate information asymmetry for private firms when they are in need of external financing. Thus, we examine whether private firms investment efficiency is more sensitive to the quality of accounting numbers when these firms seek external bank financing. In addition, some authors argue that, compared with public firms, private firms financial reporting is more influenced by taxation (e.g., Ball and Shivakumar 2005). It is also often asserted that greater emphasis on taxes (and tax minimization in particular) would reduce the informational role of earnings. However, there is little empirical evidence on this important issue to date. We have data that allow us to explicitly test the role of tax incentives.
1 2

Section 2 provides a fuller discussion of these issues. See http://www.enterprisesurveys.org.

We obtain our data from the World Banks Enterprise Survey (WBES), a major crosssectional firm survey conducted around the world by the World Bank. This database has been used in a number of prior studies (e.g., Beck and Demirguc-Kunt 2006; Beck, Demirguc-Kunt, and Maksimovic 2005, 2006, 2008; Carlin, Shaffer, and Seabright 2007; Brown, Chavis, and Klapper 2008; Hope, Thomas, and Vyas 2010).3 In contrast to studies on the role of accounting information on investment decisions that use data on public firms only, our data source allows us to examine small and medium-size private enterprises. Further, the WBES database includes information on both the sources of financing (e.g., bank financing, internal financing, informal financing, etc.) used in making new investments and the degree to which the firm faces income tax pressures (higher tax rates and stronger tax enforcement), without having to rely on proxies based on accounting data. This unique information allows us to investigate whether the usefulness of accounting information varies both with the source of financing and with the firms tax incentives. We use WBES data from 2002 to 2005 and proxy for financial reporting quality using both earnings- and revenue-based accounting quality measures. Our main findings are as follows. First, consistent with economic theory, we show that our proxies for financial reporting quality are statistically and economically significantly related to investment efficiency. Specifically, we find that accounting quality is significantly negatively associated with both over- and underinvestment. Second, we show that the importance of accounting information is increasing in the degree of bank financing. This likely reflects the use of financial statements by banks in granting credit. Third, we document that for firms facing greater income tax pressures, the relation between accounting information and investment efficiency is reduced. It is often argued

For a more extensive list, please see http://www.enterprisesurveys.org/ResearchPapers/.

in the literature that a focus on minimizing taxes can negatively affect the usefulness of financial statement information; however to date the evidence on this issue has been limited. Our results are robust to the inclusion of firm-level control variables, industry and country fixed effects, and to the use of alternative measures of both investment efficiency and earnings quality. Our study advances the existing literature by providing empirical evidence that accounting quality relates to the economic fundamentals of firm-level investment efficiency in a sample of small, private firms across 21 emerging markets. We also examine how the importance of accounting varies with financing sources and tax incentives. These findings complement and extend current research on the relation between accounting quality and investment efficiency (e.g., Biddle and Hilary 2006; McNichols and Stubben 2008; Biddle et al. 2009). Existing empirical studies rely on databases of listed firms so that even the smallest firms in their samples are relatively large. In contrast, we focus on a sample of private firms, for which there is limited extant research but which are important drivers of economic growth.4 In addition, research shows that small- and medium-sized enterprises (SMEs) not only incur higher financing obstacles than large firms, but that the effect of these financing constraints is stronger for SMEs than for large firms (e.g., Beck and Demirguc-Kunt 2006). Policymakers in governmental and international aid organizations believe that in developing countries, small firms have inadequate access to external financing due to market imperfections (World Bank 2007). In other words, compared with public firms in developed countries, our sample firms are more likely to be underinvestment firms, and our empirical evidence supports this idea. Given the importance of financing for these firms, financial reporting quality could play a major role in their investment decisions, and our results are consistent with this notion. Further, compared with

Our evidence may thus also be of importance to accounting standard setters, for example to the IASB and its SME working group.

publicly traded companies in the U.S. and other developed countries, our sample firms face fewer mandatory accounting recognition and measurement (as well as disclosure) requirements. Therefore, our sample firms have greater flexibility to choose financial reporting quality and their choices have the potential to be especially important for improving investment efficiency. The next section reviews related research and develops our hypotheses. Section 3 describes our data and our measures of accounting quality and investment efficiency. Section 4 presents our empirical tests and results. Section 5 contains several robustness tests and Section 6 concludes.

2. Background and Hypotheses Development Our study relates to the growing literature on the role that accounting information plays in corporate investment decisions. While in theory, all positive NPV projects would be funded and implemented, in practice, firms often face financing constraints that limit their ability to fund potential projects, suggesting that underinvestment may take place. 5 Even under the assumption that financing is available, managers allocation of resources may not be efficient, resulting in a reduction in investor value. In other words, there is prior research on managers overinvestment or empire building (e.g., Jensen 1986; Stein 2003; Hope and Thomas 2008) as well as underinvestment (e.g., Bertrand and Mullainathan 2003). The literature suggests that financial reporting and disclosure has the potential to mitigate both under- and overinvestment problems, and hence to increase overall investment efficiency. There are several mechanisms through which financial reporting can play such a role. First, accounting information can aid investment efficiency by reducing adverse selection, liquidity

Hubbard (1998) provides an overview of the extensive literature in this area.

risk, and information risk (Diamond and Verrecchia 1991; Leuz and Verrecchia 2000; Easley and OHara 2004; Lambert, Leuz, and Verrecchia 2007).6 Second, disclosed financial information aids corporate control mechanisms in preventing managers from expropriating wealth from investors or creditors (e.g., Fama and Jensen 1983), hence providing external suppliers of capital with greater assurance about managers activities as well as aiding in internal stewardship functions (e.g., board supervision of management). Third, improved accounting information can enhance investment decision-making by managers. As McNichols and Stubben (2008, 1571) point out, investment decisions depend on expectations of investment benefits. These benefits in turn depend on expectations of future growth and product demand. In other words, high quality information can help managers form more accurate expectations and identify better investment opportunities, thereby improving investment efficiency even in a world without adverse selection and/or moral hazard (Bushman and Smith 2001; McNichols and Stubben 2008). Several recent studies provide empirical evidence supporting the role of financial disclosure for investment decisions. Using a sample of U.S. publicly traded companies, Biddle, Hilary, and Verdi (2009) predict and find that higher quality financial reporting increases investment efficiency. Two related studies by McNichols and Stubben (2008) and Kedia and Philippon (2009) document that U.S. firms that manage their earnings upward tend to over-invest (see also Bens and Monahan 2004; Beatty, Liao, and Yu 2009; and Kang, Liu, and Qi 2009).7

As a derivative to the adverse selection problem, misleading accounting information, produced either within or

beyond the confines of GAAP, may allow outside capital suppliers to form unrealistic expectations, which in turn pressures managers to deviate from optimal operation decisions (e.g., Lai, Debo, and Nan 2008).
7

Similar to McNichols and Stubben (2008), Kedia and Philippon (2009) predict and find that firms overstating their

financial results pool with better performing firms to avoid detection. These firms over-invest to mimic the better performing firms.

To our knowledge, few studies examine international settings. Biddle and Hilary (2006) use country-level measures of accounting quality. In their full sample of firms from 34 countries, they find that higher quality of financial reporting lowers investment-cash flow sensitivity (a proxy for investment inefficiency).8,9 Biddle et al. (2009) note that investment-cash flow sensitivities can reflect either financing constraints or an excess of cash (see also Kaplan and Zingales 2000; Fazzari, Hubbard, and Peterson 2000). Similar to Biddle et al. (2009), we directly test whether higher financial reporting quality enhances investment efficiency by reducing overand/or underinvestment. In this study, we examine the role that financial reporting quality plays in private firms from emerging markets. Although private firms make up the vast majority of economic activity around the world (e.g., Pacter 2004; Chaney, Jeter, and Shivakumar 2004), they have received limited attention in academic research. We are not aware of prior research on the relation between accounting information and investment efficiency in a private firm setting. Private firms are different from publicly traded firms in several respects. Almost by definition, they tend to have more concentrated ownership, although some large private firms can have relatively more dispersed ownership. With greater ownership concentration, the large shareholder can take advantage of her controlling position and direct private benefits for personal consumption, which is the typical problem of expropriation of minority shareholders and

Biddle and Hilary (2006) hypothesize that the effect should be stronger when capital is provided through arms-

length transactions. Although they observe a more negative coefficient on accounting quality in the subsample of countries with above-median stock market value to credit market value, their tests do not allow them to test whether this difference is statistically significant.
9

Focusing on conditional conservatism as their proxy for accounting quality, Bushman, Piotroski, and Smith (2007)

find that countries with more timely loss recognitions exhibit higher propensities to abandon negative NPV projects. In another country-level study, Francis, Huang, Khurana, and Pereira (2009) find that corporate transparency facilitates the allocation of resources across industry sectors.

creditors (e.g., Morck, Shleifer, and Vishny 1988). Prior studies using data from developed countries find evidence suggesting that private firms have lower earnings quality on average than public firms (e.g., Ball and Shivakumar 2005; Burgstahler et al. 2006). It is important to note that it does not follow from this observation that accounting is less important for private firms than for public firms. Private firms typically have a weaker information environment compared with the relatively stronger disclosure environment of public firms (e.g., Burgstahler et al. 2006), which suggests that accounting information could play an important role since there are fewer competing sources of information. For example, Indjejikian and Matejka (2009) highlight the importance of accounting information for private firms in compensation contracts. As noted above, McNichols and Stubben (2008) emphasize the role that accounting information plays in internal decision making. Small firms are unlikely to have management accounting systems that are separate from financial accounting (e.g., Drury and Tayles 1995), potentially enhancing the role of accounting in internal decision making. McNichols and Stubben (2008) emphasize how low-quality financial information implies distorted information to decision makers that in turn leads to inefficient investment decisions.10 We are also interested in the role of financial information in emerging markets. There is far less research on the usefulness of accounting in such environments than in developed countries. Although prior studies do not include our exact sample of countries (which are mostly low-income countries as described in Section 3), there is some evidence in the literature suggesting that the value relevance of accounting information (for publicly traded companies) is lower in less developed economies. Again, this does not necessarily mean that financial

10

For example, investment decision makers within the firm could believe the misstated growth trend from

overstated revenues, either because they are over-optimistic or because they are unaware of the misstatement, and invest accordingly (McNichols and Stubben 2008, 1572).

reporting information has less value in such countries, and in fact it is possible that the lack of analyst coverage, lower media coverage, and overall lower-quality institutions makes accounting information a relatively greater component of the overall information set used for decision making by insiders or outsiders.11,12 Based on the above discussion, theory supports the idea that high-quality information leads to higher investment efficiency in private firms. But as discussed, based on prior research this relation is not a foregone conclusion and it is ultimately an empirical question what the outcome will be. For our first hypothesis, we predict that higher quality accounting information will mitigate both over- and underinvestment.

H1: Accounting quality is negatively associated with both underinvestment and overinvestment.

In addition to examining the overall effect of earnings quality on investment efficiency, we explore two conditional effects: financing sources and tax incentives. We first investigate the effect of financing sources on the importance of accounting earnings. Using the WBES data, we have access to a detailed breakdown of financing sources as a percentage of new investment. Private firms in emerging markets fund investment from external sources such as bank lending, issuance of private equity, leasing, trade credit, financing from special development agencies or

11

Abu-Nassar and Rutherford (1996), Mirshekary and Saudagaran (2005), and Al-Razeen and Karbhari (2007)

conduct surveys and conclude that accounting information is the most important information source for users in emerging markets.
12

Accounting information may be useful for bank financing, in attracting new equity capital (for the private firms

with more dispersed ownership and/or firms that are selling shares to new investors), to suppliers in their decisions to grant trade credit, and to other providers of finance such as leasing companies.

governments, and informal financing (i.e., financing from moneylenders, family, and friends), or from internal sources (i.e., retained earnings and additional contributions by owners) (Beck et al. 2008). A large body of literature has documented how the availability of external or internal funds affects investment decisions (Myers and Majluf 1984; Fazzari, Hubbard, and Peterson 1988; Blanchard, Lopez-de-Silanes, and Shleifer 1994). Prior studies have used samples of public firms that rely mainly on equity and debt financing. In contrast, for private firms, the external financing sources are usually limited and mainly consist of bank loans and trade credit. We argue that, given differential demand for accounting information, the financing source will affect the relation between accounting quality and investment efficiency. Specifically, we are interested in whether firms that rely more on bank financing have a relatively stronger relation between accounting and investment efficiency than do other private firms in emerging economies. Bank lending is the most common source of external capital for private firms in developing countries (Brown et al. 2008; Beck et al. 2008).13 Banks may well have access to additional information beyond the financial statements, potentially reducing the importance of accounting information. However, besides the large body of research documenting the role of accounting information for lending decisions in the U.S. and other developed countries, there is also extensive evidence that banks rely on borrowers financial reports in credit decisions in emerging markets and for small firms (e.g., Danos, Holt, and Imhoff 1989; Berry, Faulkner, Hughes, and Jarvis 1993; Berry, Grant, and Jarvis 2004; Kitindi, Magembe, and Sethibe 2007). Compared with other external sources of informal financing that rely more on mutual trust and private communication, banks are likely to more carefully screen the financial statements of

13

In our sample firms, on average bank financing accounts for 42% of all external financing of new investments.

10

corporate clients. Most importantly, banks not only lend larger amounts compared with other sources of informal financing, but also extend loans with longer maturities, which make them more vulnerable to information and incentive problems. Failure rates are also higher among smaller firms, and this further encourages banks to carefully examine financial information in their lending decisions. Examining a clients financial statements helps banks to determine the firms assets that can serve as collateral, to evaluate its future cash-flow generating capability, to gauge the firms debt capacity, and to analyze the riskiness of the firm in determining a lending rate. Based on this discussion, our second hypothesis is:14

H2: The relation between financial reporting quality and investment efficiency is stronger if a firms investment is mainly funded through bank financing.

Finally, we explore the conditional effect of tax incentives. The alignment between financial and tax accounting is likely higher for our sample than for publicly traded firms in developed countries. High alignment means that financial statements serve as the basis for taxation or that tax laws explicitly require an equal treatment for certain items in both sets of accounts (e.g., Burgstahler et al. 2006). In such an environment, a focus on minimizing taxes can directly impact financial reporting quality. That is, for firms that face especially high income tax rates and strong enforcement by tax authorities (i.e., high tax pressure), the primary objective of financial reporting may be to minimize income taxes rather than to provide information to

14

Note that our hypothesis differs from Biddle and Hilary (2006) who compare bank financing with public equity

financing across countries. In contrast, as private firms do not rely on public equity financing, we compare the role of bank financing primarily with other non-public debt, private equity, and informal financing.

11

suppliers of capital or to management, which may reduce the role of reported accounting information in informing users. Our final hypothesis is thus:

H3: For firms with higher tax pressure, accounting quality is less negatively associated with both underinvestment and overinvestment.

3. Sample and Measurement of Main Variables 3.1 Data Source We obtain our data from World Banks Enterprise Survey (WBES) conducted during 2002-2005 by the World Bank in 79 countries, including many low-income countries. Prior studies that have used this database include Beck and Demirguc-Kunt (2006), Beck et al. (2005, 2006, 2008), Carlin et al. (2007), Brown et al. (2008), and Hope et al. (2010). The primary goal of WBES is to provide quantitative data that allows an assessment of a countrys investment climate in an internationally comparable manner.15 The surveys are administered in face-to-face interviews with managing directors, accountants, human resource managers, and other relevant company staff.16 Samples are stratified by size, industry, and location. Although the surveys are conducted with the knowledge and support of relevant government authorities, governments are not provided with the raw data or other information that

15 16

This discussion draws on Productivity and Investment Climate Survey: Implementation Manual (2003). WBES begins from a minimum core set of questions that are common across all countries. However, survey

managers at the country or region level are allowed to extend the survey. To maintain cross-country comparability, core questions cannot be reworded except in translating to well-understood phrases with the same meaning. In addition, all core questions are asked using standardized instructions provided by the surveys designers.

12

allows them to identify the responses of individual firms, and businesses are informed of this confidentiality prior to interviews to encourage truthful responses.17 The dataset includes a large sample of firms across multiple sectors (manufacturing, services, agriculture, and construction). Usable data include both quantitative and qualitative information on firm characteristics, including sources of finance, barriers to growth, access to infrastructure services, legal difficulties, and corruption. The dataset also includes measures of firm performance, such as multiple years of historical data on capital investment and operating performance. A limitation of using the survey data is that financial statement information is necessarily restricted. Thus we do not have access to some potential control variables used in prior studies. However, the survey does contain data on a number of important firm characteristics, and consistent with prior research that employs the WBES data, we include these firm characteristics as well as industry and country fixed effects in our empirical tests. We describe our sample selection procedure in Table 1. The main constraints on our sample size are the availability of data to compute the investment efficiency measure and the two accounting quality measures. Each sample firm appears in only one survey year and hence the number of observations equals the number of unique firms. However, conditional on data availability, the dataset provides three consecutive years of accounting information from a given survey year. We end up with 6,862 unique firms for the test of H1 involving discretionary revenues (described below) as a proxy for accounting quality, and our sample size is reduced to 6,697 firms when we use discretionary accruals as another proxy for accounting quality. The

17

Carlin et al. (2007), Beck et al. (2008), and Hope et al. (2010) contain several validity tests of the survey data.

13

sample size is reduced further for the tests of H2 and H3 for which we require additional data for the firms financing sources and tax burden. Table 2 provides a distribution of the sample by country. We have a total of 21 countries represented, with the greatest number of observations coming from Thailand, Brazil, Pakistan, Vietnam, and India. These five countries make up 65% of the total sample. Although all of our sample countries are considered developing or emerging markets as per United Nations classifications, there is considerable variation in income per capita across our sample countries. We examine the effects of financial reporting quality on investment efficiency across different income groups in Section 5.

3.2 Proxy for Investment Efficiency The two key constructs in the analysis are investment efficiency and financial reporting quality, and we investigate how accounting quality in the current year affects next years investment efficiency. Conceptually, investment efficiency refers to firms undertaking all and only projects with positive net present value. Consistent with prior research (e.g., Biddle et al. 2009), we measure investment efficiency as deviations from expected investment using a model that predicts investment as a function of growth opportunities. Thus, both underinvestment (negative deviations from expected investment) and overinvestment (positive deviations from expected investment) are considered inefficient investments. Specifically, we estimate a parsimonious model for expected investment as a function of revenue growth (see, e.g., Modigliani and Miller 1958; Hubbard 1998). As the relation between investment and revenue growth could differ between revenue decreases and revenue increases (e.g., Eberly 1997; McNichols and Stubben 2008), we allow for differential predictability for revenue increases and

14

revenue decreases by employing a piecewise linear regression model:18

Invest i ,t = 0 + 1 N EG i ,t 1 + 2 RevGrow thi ,t 1 + 3 N EG * RevG row thi ,t 1 + i , t

(1)

Following Biddle et al. (2009), we define Invest i ,t as the sum of new investment in machinery, equipment, vehicles, land, buildings, and research and development expenditures, less the sale of fixed assets, and scaled by lagged total assets for firm i in year t. RevG row thi , t 1 is the annual revenue growth rate for firm i in year t-1. The indicator variable NEGi,t-1 takes the value of one for negative revenue growth, and zero otherwise. We estimate the investment model cross-sectionally with at least ten observations in each WBES industry by country. The sample consists of 9,992 firm-year observations with available data to estimate Equation (1). To mitigate the influence of outliers, we winsorize all variables at the 1% and 99% levels. We then classify firms into two groups based on the residuals of Equation (1) (i.e., the deviations from the predicted investment levels). To ease exposition, we multiply the underinvestment variable by -1 so that a higher value suggests a more severe underinvestment.

3.3 Proxies for Financial Reporting Quality There is no one universally agreed upon measure of accounting quality (e.g., Dechow, Ge, and Schrand 2009). To ensure the robustness of our conclusions, we apply two measures that have been widely used in prior research. The first measure is performance-adjusted discretionary

18

Note that we describe results of using two alternative investment models in Section 5.2 (including a model that

contains seven additional explanatory variables beyond revenue growth). Furthermore, adding lagged investment to Equation (1) does not change our inferences but significantly reduces our sample size.

15

accruals as developed by Kothari, Leone, and Wasley (2005).19 Specifically, we estimate the following model by country and for each industry that has at least 16 observations:
Accri ,t = 0 + 1 (1 / Assets i ,t 1 ) + 2 Sales i , t + 3 PPE i ,t + 4 RO Ai ,t + i ,t

(2)

where Accri , t is measured as the change in non-cash current assets minus the change in current liabilities, minus depreciation and amortization expense for firm i at year t, scaled by lagged total assets ( Assetsi ,t 1 ); Salesi ,t is the annual change in sales revenues scaled by lagged total assets;
PP E i ,t is property, plant, and equipment for firm i at year t, scaled by lagged total assets; RO Ai , t

is return on assets for firm i at year t. The residuals from the regression model are discretionary accruals (DisAccr). In our main tests, we use the absolute values of discretionary accruals as a proxy for accounting quality. 20 We multiply the absolute values of discretionary accruals by -1. Thus, higher values of DisAccr represent higher earnings quality. To calculate the second proxy, we follow McNichols and Stubben (2008) and Stubben (2010) and estimate discretionary revenues. Specifically, we use the following regression:

ARit = + Salesit + it

(3)

where AR represents the annual change in accounts receivable and Sales is the annual change in revenues, each scaled by lagged total assets. Discretionary revenues (DisRev) are the residuals from Equation (3) which is estimated separately for each industry-country group that has at least eight observations. We multiply the absolute values of discretionary revenues by -1. Thus, higher values of DisRev represent higher accounting quality.

19 20

Data availability precludes the use of the Dechow and Dichev (2002) measure of accrual quality. We redo all tests using signed values of discretionary accruals. Results are presented in Section 5.1.

16

3.4 Bank Financing and Tax Burden Variables For the purpose of examining H2, we use an indicator variable (Bank) to capture bank financing as a major external financing source for firms. Specifically, this indicator takes a value of one if bank financing is the dominant source of external financing and constitutes more than 5% of all sources that fund new investments.21 Bank financing includes financing from both local and foreign banks. Our measure of income tax pressure refers to managers perceptions of whether tax rates and tax administration impose a major or severe obstacle in the operation and growth of their business. In the WBES database, higher ratings of tax rates and/or tax administration suggest a heavier tax burden borne by the firm, implying that firms have greater incentives to manage earnings for tax purposes. We use an indicator variable, Tax, to test H3. The indicator takes on the value of one when the average rating of managers responses is 3 or greater.22

3.5 Descriptive Statistics Table 3 provides descriptive statistics for our measures of investment efficiency, earnings quality, bank financing, tax burden, as well as for our main control variables. Panel A shows that only about a quarter of firms belong to the overinvestment group, while an overriding majority of sample firms belong to the underinvestment group. This intuitive result confirms that private firms in emerging markets, due to their difficulty in securing external financing, are more likely subject to the problem of underinvestment rather than overinvestment.

21

The 5% criterion ensures that bank financing is substantial compared with internal financing. As a sensitivity test Specifically, in the WBES database, a rating of zero denotes no obstacle; one denotes a minor obstacle; two, a

we change the threshold to 10% and no inferences are affected.

22

moderate obstacle; three, a major obstacle; and four, a very severe obstacle.

17

For the 4,657 firms that provide information about their financing sources, about 38% of these firms have bank financing as their primary source of external financing (and that constitutes more than 5% of all sources that fund new investments). For the 6,830 firms that provide information about the firm-level tax burden, about 17% of these firms report facing severe tax obstacles. The average firm has been in existence for 18 years. As expected, there is a wide variation in firm size measured by total assets. But most firms are quite small, with half the firms having total assets of $912,110 or less. In all tests we use log transformations of both firm age and firm size. Panel B reports Pearson correlations. As predicted, the two proxies of accounting quality are significantly negatively correlated with the proxy for investment inefficiency. In addition, the two accounting quality proxies are positively and significantly correlated. As correlation results do not control for differences in firm, industry, or country characteristics, we now turn to multivariate tests.

4. Research Design and Results 4.1 Basic Empirical Model Since we are interested in how accounting quality affects investment efficiency, we test Hypothesis 1 by regressing the measure of investment efficiency in year t+1 on the measures of accounting quality in year t. As well, similar to Biddle et al. (2009), we estimate equation (4) separately for overinvestment and underinvestment. This research design allows us to test whether higher financial reporting quality mitigates over- as well as underinvestment, which is important given that our sample firms presumably are more likely to experience underinvestment problems. The basic model is:

18

InvEffi ,t +1 = 0 + 1EQi ,t + nControl Variablesi ,t + Industries + Countries + i ,t

Where:

(4)

InvEff = Excess investment (overinvestment or underinvestment) is the residual of the investment model as described in the section 3.2. EQ = Accounting quality measured as either the absolute residuals of the Kothari et al. (2005) performance-adjusted discretionary accruals model, multiplied by -1; or the absolute residuals of the McNichols and Stubben (2008) discretionary revenue model, multiplied by -1.
Our main control variables are:

Size = Natural log of total assets. Age = Natural log of firm age. Tang = Asset tangibility, measured as property, plant and equipment scaled by lagged total assets. Slack = Financial slack, measured as cash divided by lagged total assets.
Motivated by prior research (e.g., Biddle et al. 2009), we include firm size, firm age, asset tangibility, and financial slack as control variables, as well as industry fixed effects. 23 We also include country fixed effects in all models, which is a common approach for controlling country-specific effects and addressing correlated omitted country-level variable problems (Doidge, Karolyi, and Stulz 2007). To the extent that lower earnings quality (EQ) (i.e., larger absolute residuals from the discretionary revenues or accruals models) leads to less efficient investment (H1) (i.e., larger magnitude of over- and underinvestment), 1 is expected to be negative.

23

Untabulated tests show that our results are almost identical using number of employees or revenues as alternative

size proxies. Note further that we in Section 5.3 report results of tests that include 11 additional control variables (which reduces the sample size by 54%).

19

4.2 Empirical Results Table 4 reports the regression results for the test of H1. The models have adjusted R2s of between 14% and 46%, with a higher explanatory power for underinvestment (the most prevalent scenario in our sample). Columns 1 and 3 (2 and 4) report the results using the DisAccr (DisRev) measure for overinvestment and underinvestment, respectively. 24 Across all four specifications, the conclusion is the same: accounting quality is negatively related to inefficient investments. Specifically, the estimated coefficients on DisAccr and DisRev are all negative and significant at the 1% level. These results are also economically significant. For example, in the overinvestment scenario, a one standard deviation decrease in the DisAccr measure implies a decrease of overinvestment by 1.07% of total assets. Firm size is negatively associated with both over- and underinvestment, consistent with expectations and prior research. Asset tangibility is positively correlated with deviations from expected investment, but is only significant for overinvestment cases. Financial slack is positively but insignificantly related to investment efficiency. Taken together, the results in Table 4 suggest that, although private firms in emerging markets could have different reporting incentives than the companies previously examined in the literature, we observe strong evidence that the quality of accounting numbers affects subsequent capital investment efficiency in accordance with theory. Specifically, our results support H1 and suggest that lower quality of financial reporting results in less efficient investment. In addition to complementing and extending prior academic research, our findings should be relevant to the World Bank, whose mission is to aid in improving living conditions in developing countries, as it is likely that more efficient investments should lead to higher social
24

Note that we report results of analyzing positive and negative residuals from the two accounting quality proxies

separately in Section 5.1

20

welfare. In fact, it is a reasonable conjecture that, compared with publicly traded companies in the U.S. and similar countries, our sample firms face fewer mandatory accounting recognition and measurement (as well as disclosure) requirements. Therefore, our sample firms choices regarding accounting quality have the potential to be especially important for improving investment efficiency. For the tests of the relative role of bank financing (H2) and the effect of tax incentives (H3), we add the main effects of these variables as well as interaction effects with the earnings quality measures. We predict a negative coefficient on BankEQ and a positive coefficient on

TaxEQ.
Results for the conditioning effect of bank financing are presented in Table 5. The additional data requirement for financing sources results in a reduced sample size, with 4,657 firms for the test employing discretionary revenues and 4,457 firms for the test using discretionary accruals. We first note that both EQ measures continue to load negatively and significantly after controlling for the effect of bank financing. The main effect of Bank is not significant. Our focus, however, is on BankEQ, and we observe negative estimated coefficients using both DisAccr and DisRev and for both overinvestment and underinvestment. Greater significance levels are achieved for the underinvestment case, possibly due to the larger sample size available for that test. The results suggest that the importance of accounting quality is increasing in the presence of bank financing, which is consistent with banks making use of financial statements in their lending decisions. In Table 6, we test H3, the effect of tax incentives. We have 6,830 firms for the discretionary revenues and 6,663 firms for the discretionary accruals tests. Recall that we argue that the informational role of financial information will be reduced when the overriding objective

21

is to minimize taxes. The main effect of EQ continues to be negative and significant in this specification. For the overinvestment sample, the coefficients on the interaction of EQ and Tax are positive and significant when using either the DisAccr or the DisRev accounting quality measures. In other words, consistent with H3, the effect of earnings quality in mitigating the overinvestment problem is reduced in the extent to which firms incentives are primarily tax oriented.25 We also observe positive but statistically weaker coefficients on EQTax in the underinvestment scenario. Thus, there is at least some support for the often-argued (but seldom tested) notion in the literature that tax motivations may distract from the usefulness of accounting information. Finally, to test whether bank financing and tax incentives have incremental effects in the presence of the other, in Table 7 we report tests in which we include both Bank and Tax jointly (and thus have a smaller sample). As before, the estimated coefficient on EQBank continues to be negative in all specifications and statistically significant in three out of four regressions. Moreover, the estimated coefficient on EQTax is significantly positive for the overinvestment case but not significant for the underinvestment case. Equally importantly, the main effect of EQ is yet again significantly negative across test specifications. In addition to being interesting results by themselves, our interaction tests also serve the function of providing additional credence to H1 (e.g., related to correlated omitted variables or causality). We observe that the estimated coefficients on the two interaction terms have the predicted signs and are statistically significant in several cases. In other words, the effect is more (less) pronounced in subsamples in which we predict the effect to be stronger (weaker). As Rajan and Zingales (1998) and Lang,

25

In fact, for both the DisAccr and DisRev measures of accounting quality, the sum of the coefficients on EQ and

EQTax (1+ 3) is not statistically different from zero in the overinvestment tests (the joint test results in Fstatistic=0.44 with p-value=0.51 for DisAccr; F-statistic=0.96 with p-value=0.33 for DisRev, respectively).

22

Lins, and Maffet (2009) point out, it would be difficult to envision a consistent theory in which causality is reversed yet the subsample results hold.

5. Additional Tests In this section, we report results of additional tests that both lend robustness and extend the reported results. For brevity, we only tabulate the results of the first sensitivity test.

5.1 Separate Regressions for Positive and Negative Values of DisAccr and DisRev To supplement our analyses based on the absolute value of the earnings quality measures, we use a signed earnings quality measure (e.g., Hribar and Nichols 2007). Again, we estimate the model for both over- and underinvestment. Unlike the EQ measures in the prior sections, we do not multiply the proxies of EQ by -1. Thus, for the main effect, we expect positive (negative) 1 for the sample with positive (negative) discretionary revenues or accruals. The results, which are reported in Table 8, are consistent with our expectations. Specifically, as reported in panel A, the positive discretionary accruals (DisAccr) are positively associated with both overinvestment and underinvestment (t=2.42 and 1.63, respectively) while negative discretionary accruals are negatively related to both overinvestment and underinvestment (t=-4.76 and -1.60, respectively). The results are similar for signed discretionary revenues (DisRev) as reported in panel B. Similarly, we obtain results consistent with H2 (the conditional effect of bank financing) and H3 (the conditional effect of tax burdens), although the coefficient estimates for some cases are not

23

as statistically significant as the results reported in Table 7. In general, the use of the signed earnings quality measure corroborates the conclusions above.26

5.2 Alternative (and Refined) Measures of Investment Efficiency We repeat all tests employing two alternative measures of investment efficiency. Our first alternative measure follows Biddle et al. (2009). Specifically, we sort firms, based on the positive and negative residuals, respectively, into deciles. Within the positive-residual group, we remove the bottom decile because these firms, whose unexpected investments are closest to zero among all overinvesting firms, are more likely to be affected by measurement error in the investment model (i.e., misclassified as overinvesting firms). Similarly, we remove the top decile from the negative-residual group. We repeat all the tests using the remaining observations. Results are similar to those tabulated in the paper, and no inferences are affected. The second alternative measure of investment efficiency comes from Richardson (2006). We first estimate optimal investment according to the following regression specification:

Investi ,t = 0 + 1NEGi ,t 1 + 2 RevGrowthi ,t 1 + 3 NEG * RevGrowthi ,t 1 + 4 Sizei ,t 1 + 5 Agei ,t 1 +6 Slacki ,t 1 + 7 Levi ,t 1 + 8 BConIndexi ,t + 9 LConIndexi ,t + 10 PEConIndexi ,t + i ,t In addition to the proxy for growth opportunity, we also include firm size, firm age, financial slack, and financial leverage as controls for financing constraints. Moreover, based on Brown et al. (2008), we augment the model with three additional firm-level variables related to constraints in the operating environment: business constraints, legal constraints, and

26

As an additional sensitivity test, we have used factor analysis to compute an aggregate measure of accounting

quality. All inferences are unchanged.

24

political/economic constraints.27 The unexplained portion (or residuals) is the estimate of inefficient investment. Our conclusions are unchanged when we employ this alternative measure of investment efficiency.

5.3 Additional Control Variables

Although we include several firm-level control variables as well as industry and country fixed effects in our main tests, it is always possible that there are some omitted (and correlated) variables. However, adding more control variables comes at the cost of reducing the sample size, and thus a trade-off exists between sample size (and hence generalizability) and model completeness. In untabulated tests, we add several additional controls to all regressions. Following Biddle et al (2009), we include (1) long-term liabilities divided by total assets (Lev); (2) the standard deviation of sales in the past three years (StdRev); (3) the standard deviation of investments in the past three years (StdInvest); (4) the operating cycle calculated by using both accounts receivables and sales (OperCycle); and (5) an indicator variable referring to whether a firm reports a loss (Loss). We also (6) control for profitability (ROA). Furthermore, we take advantage of the availability of certain unique data items from WBES and include: (7) an indicator variable referring to whether the firm exports (Export); (8) an indicator variable referring to whether any foreign company or individual has ownership interests in the firm (Foreign); (9) an indicator variable that takes on the value of one if financing obstacles (FinCon) are severe; (10) an indicator variable that takes on the value of one if the largest shareholder

27

The Appendix contains detailed definitions of these additional variables. Note that with the expanded set of

control variables, we require more observations to estimate the model than for the estimation of Equation (1). Therefore, we estimate this model by country and adjust both dependent and independent variables by the countryspecific industry median. The adjustment for industry median accounts for industry heterogeneity.

25

owns more than 50% of the firm (Control); and (11) an indicator variable referring to whether a firms financial statements are reviewed by independent auditors or not (Audit). Requiring data on these additional controls reduces the sample size by approximately 54%, with only 2,987 firm-years for the sample using the DisAccr measure and only 3,166 firmyears for the sample using the DisRev measure. Nevertheless, even with the reduced sample size our conclusions for H1 to H3 remain unaltered.

5.4 Possible Variations across Sample Countries

Although all our sample countries are emerging economies, they clearly vary in terms of income per capita and other factors, such as legal protection of investors. Recall that we control for such effects using country fixed effects in all reported tests (e.g., Doidge et al. 2007). In this section, we perform two additional tests. First, we replace country fixed effects with gross national income per capita (GNI), a country-level variable that is available for all our sample countries. No conclusions are altered with this specification (and in fact significance levels increase somewhat). Second, we separate the sample based on median GNI and estimate the regressions separately for the two groups. We find that our results hold in both sets of countries.

5.5 Possible Endogeneity Related to Bank Financing

Unlike our proxy for tax incentives, which we consider mostly outside of managers control, the importance of bank financing could be driven by other firm characteristics. For this reason we estimate a two-stage Heckman model in which the first stage (which is based on Beck et al. 2008) predicts Bank and the second stage controls for the Inverse Mills Ratio estimated from the first stage. As an instrument we include the approximate value of collateral as a

26

percentage of the loan value (Collateral). Banks may be less willing to lend to firms when the amount of assets that can be pledged as collateral is limited. In untabulated tests we find that Collateral is significantly correlated with the choice of bank financing but is not significantly associated with over- or underinvestment. After controlling for potential self-selection, our conclusions regarding H2 remain unchanged: the importance of earnings quality for investment efficiency is increasing in bank financing.

6. Conclusions

We study the role of accounting information for a set of firms for which there is very limited prior research evidence: private firms from emerging markets. On the one hand, prior research sometimes argues that the role of accounting is more limited for both private firms and for firms from emerging markets compared with publicly traded companies in developed countries. The argument put forth is usually that these firms do not face the same market demand for high-quality financial information and that tax incentives could be more important. On the other hand, it is also true that there are more limited alternative information sources other than accounting for private firms (and in countries with on-average lower disclosure levels). Also, the link between financial accounting and management accounting is likely to be strong in these settings, which suggests that the financial accounting information is largely the same set of information used in managers decisions. Thus, it is an empirical question whether financial statement quality is important for firms investment decisions for private firms in emerging markets. We find strong evidence that accounting quality is indeed positively associated with investment efficiency for our sample firms. This result complements and extends extant research.

27

In addition, we hypothesize that the source of financing as well as firms tax incentives will affect the role of accounting information. We find evidence supporting these hypotheses. Specifically, we find that greater use of bank financing increases the role that accounting information plays. In addition, for firms in which tax incentives are likely to dominate incentives to provide useful information for internal decision making as well as a source of information for outside providers of capital, the informational role of accounting is significantly diminished. Our results are robust to including a number of firm-level control variables as well as industry and country fixed effects, and to a variety of sensitivity analyses. We hope that our empirical findings will be of interest to researchers, standard setters and other regulators, government officials in emerging markets, managers, and importantly to the World Bank and others involved in improving economic conditions in developing countries.

28

References
Abu-Nassar, M., and B. A. Rutherford. External Users of Financial Reports in Less Developed Countries: the Case of Jordan. British Accounting Review 28 (1996): 73-87. Al-Razeen, A., and Y. Karbhari. An Empirical Investigation into the Importance, Use, and Technicality of Saudi Annual Corporate Information. Advances in International Accounting 30 (2007): 55-74. Ball, R., and Shivakumar, L. Earnings Quality in UK Private Firms: Comparative Loss Recognition Timeliness. Journal of Accounting and Economics 39 (2005): 83-128. Ball, R.; S. P. Kothari; and A. Robin. The Effect of International Institutional Factors on Properties of Accounting Earnings. Journal of Accounting and Economics 29 (2000): 1-51. Beatty, A.; S. Liao; and J. Yu. The Spillover Effect of Fraudulent Financial Reporting on Peer Firms Investment Efficiency. Working paper, Ohio State University, University of Toronto, and Southern Methodist University, 2009. Beck, T. and A. Demirguc-Kunt. Small and Medium-size Enterprises: Access to Finance as a Growth Constant. Journal of Banking and Finance 30 (2006): 2931-2943. Beck, T.; A. Demirguc-Kunt; and V. Maksimovic. Financial and Legal Constraints to Firm Growth: Does Size Matter? Journal of Finance 60 (2005): 137-177. Beck, T.; A. Demirguc-Kunt; and V. Maksimovic. The Influence of Financial and Legal Institutions on Firm Size. Journal of Banking and Finance 30 (2006): 2995-3015. Beck, T.; A. Demirguc-Kunt; and V. Maksimovic. Financing Patterns around the World: Are Small Firms Different? Journal of Financial Economics 89 (2008): 467-487. Bens, D. A., and S. J. Monahan. Disclosure Quality and the Excess Value of Diversification. Journal of Accounting Research 42 (2004): 691-730. Berry, A.; P. Grant; and R. Jarvis. European Bank Lending to the UK SME Sector. International Small Business Journal 22 (2004): 115-130. Berry, A.; S. Faulkner; M. Hughes; and R. Jarvis. Financial Information, the Banker and the Small Business. British Accounting Review 24 (1993): 131-150. Bertrand, M. and S. Mullainathan. Enjoying the Quiet Life? Corporate Governance and Managerial Preferences. Journal of Political Economy 111 (2003): 1043-1075. Berzins, J.; . Bhren; and P. Rydland. Corporate Finance and Governance in Firms with Limited Liability: Basic Characteristics. Working paper, Norwegian School of Management, 2008. Biddle, G., and G. Hilary. Accounting Quality and Firm-level Capital Investment. The Accounting Review 81 (2006): 963-982. Biddle, G.; G. Hilary; and R. S. Verdi. How Does Financial Reporting Quality Relate to Investments Efficiency? Journal of Accounting and Economics 48 (2009): 112-131. Blanchard, O.; F. Lopez-de-Silanez; and A. Shleifer. What do Firms do with Cash Windfalls? Journal of Financial Economics 36 (1994): 337-360. Brown, G. W.; L. W. Chavis; and L. F. Klapper. A New Lease on Life: Institutions, External Financing, and Business Growth. Working paper, University of North Carolina at Chapel Hill and World Bank, 2008.

29

Burgstahler, D.; L. Hail; and C. Leuz. The Importance of Reporting Incentives: Earnings Management in European Private and Public Firms. The Accounting Review 81 (2006): 983-1016. Bushman, R., and A. Smith. Financial Accounting Information and Corporate Governance. Journal of Accounting and Economics 31 (2001): 237-333. Bushman, R.; J. Piotroski; and A. Smith. Capital Allocation and Timely Accounting Recognition of Economic Losses: International Evidence. Working paper, University of North Carolina at Chapel Hill, University of Chicago and Stanford University, 2007. Carlin, W.; M. E. Schaffer; and P. Seabright. Where are the Real Bottlenecks? Evidence from 20,000 Firms in 60 Countries about the Shadow Costs of Constraints to Firm Performance. Working paper, CEPR, 2007. Chaney, P.; D. Jeter; and L. Shivakumar. Self Selection of Auditors and Audit Pricing in Private Firms. The Accounting Review 79 (2004): 51-72. Danos, P.; D. Holt; and E. A. Imhoff. The Use of Accounting Information in Bank Lending Decisions. Accounting Organizations and Society 14 (1989): 235-246. Dechow, P., and I. Dichev. The Quality of Accruals and Earnings: The Role of Accrual Estimation Errors. The Accounting Review 77 (2002): 35-59. Dechow, P.; W. Ge; and C. Schrand. Understanding Earnings Quality: A Review of the Proxies, Their Determinants and Their Consequences. Paper presented at the 2009 Journal of Accounting and Economics conference. Diamond, D. W., and R. E. Verrecchia. Disclosure, Liquidity, and the Cost of Capital. Journal of Finance 46 (1991): 1325-1359. Doidge, C.; A. Karolyi, and R. Stulz. Why Do Countries Matter So Much For Corporate Governance? Journal of Financial Economics 86 (2007): 1-39. Drury, C., and M. Tayles. Issues Arising from Surveys of Management Accounting Practice. Management Accounting Research 6 (1995): 267-280. Easley, D., and M. OHara. Information and the Cost of Capital. Journal of Finance 59 (2004): 15531583. Eberly, J C. International Evidence on Investment and Fundamentals. European Economic Review 41 (1997): 1055-1078. Fama, E. F. and M. C. Jensen. Separation of Ownership and Controls. Journal of Law and Economics 26 (1983): 301-325. Fazzari, S. M.; R. G. Hubbard; and B. C. Petersen. Investment and Finance Reconsidered. The Brookings Papers on Economic Activity (1988): 141-195. Fazzari, S. M.; R. G. Hubbard; and B. C. Petersen. Investment-Cash Flow Sensitivities are Useful: A Comment on Kaplan and Zingales. The Quarterly Journal of Economics 115 (2000): 695-705. Francis, J. R.; S. Huang; I. K. Khurana; and R. Pereira. Does Corporate Transparency Contribute to Efficient Resource Allocation? Journal of Accounting Research 47 (2009): 943-989. Hope, O.-K., and W. B. Thomas. Managerial Empire Building and Firm Disclosure Journal of Accounting Research 46 (2008): 591-626.

30

Hope, O.-K.; W. B. Thomas; and D. Vyas. Transparency, Ownership, and Financing Constraints in Private Firms. Working paper, University of Toronto and University of Oklahoma, 2010. Hribar, P., and D. C. Nichols. The Use of Unsigned Earnings Quality Measures in Tests of Earnings Management. Journal of Accounting Research 45 (2007): 1017-1053. Hubbard, R. G. Capital-market Imperfections and Investment. Journal of Economic Literature 36 (1998): 193-225. Indjejikian, R., and M. Matejka. CFO Fiduciary Responsibilities and Annual Bonus Incentives. Journal of Accounting Research 47 (2009): 1061-1093. Jensen, M. Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers. American Economic Review 76 (1986): 323-329 Kang, Q.; Q. Liu; and R. Qi. The Sarbanes-Oxley Act and Corporate Investment: a Structural Assessment. Forthcoming, Journal of Financial Economics, 2009. Kaplan, S., and L. Zingales. Investment-Cash Flow Sensitivities are not Valid Measures of Financing Constraints. The Quarterly Journal of Economics 112 (2000): 707-712. Kedia, S., and T. Philippon. The Economics of Fraudulent Accounting. Review of Financial Studies 22 (2009): 2169-2199. Kitindi, E. G.; B. S. Magembe; and A. Sethibe. Lending Decision Making and Financial Information: The Usefulness of Corporate Annual Reports to Lenders in Botswana. The International Journal of Applied Economics and Finance 1 (2007): 55-66. Kothari, S. P.; A. J. Leone; and C. E. Wasley. Performance Matched Discretionary Accrual Measures. Journal of Accounting and Economics 39 (2005): 163-197. Lai, G.; L. Debo; and L. Nan. Stock Market Pressure on Inventory Investment and Sales Reporting for Publicly Traded Firms. Working paper, Carnegie Mellon University, 2008. Lambert, R.; C. Leuz; and R. E. Verrecchia. Accounting Information, Disclosure, and the Cost of Capital. Journal of Accounting Research 45 (2007): 385-420. Lang, M.; K. V. Lins; and M. Maffett. Transparency, Liquidity, and Valuation: International Evidence. Working paper, University of North Carolina and University of Utah, 2009. Leuz, C., and R. E. Verrecchia. The Economic Consequences of Increased Disclosure. Journal of Accounting Research 38 (2000): 91-124. McNichols, M. F., and S. R. Stubben. Does Earnings Management Affect Firms Investment Decisions? The Accounting Review 83 (2008): 1571-1603. Mirshekary, S., and S. M. Saudagaran. Perceptions and Characteristics of Financial Statement Users in Developing Countries: Evidence from Iran. Journal of International Accounting, Auditing and Taxation 14 (2005): 33-54. Modigliani, F., and M. Miller. The Cost of Capital, Corporation Finance, and the Theory of Investment. American Economic Review 48 (1958): 261-297. Morck, R.; A. Shleifer; and R.W. Vishny. Management ownership and market valuation: an empirical analysis. Journal of Financial Economics 20 (1988): 293315.

31

Myers, S. C., and N. S. Majluf. Corporate Financing and Investment Decisions when Firms have Information that Investors do not have. Journal of Financial Economics 13 (1984): 187-221. Nagar, V.; K. Petroni; and D. Wolfenzon. Governance Problems in Close Corporations. Working paper, University of Michigan, Michigan State University, and New York University, 2005. Pacter, P. Financial Reporting: SMEs-Will the GAAP Widen for SMEs? Accountancy 133 (2004): 118119.

Rajan, R.G., and L. Zingales. Financial dependence and growth. American Economic Review 88 (3) (1998): 559-586.
Richardson, S. Over-Investment of Free Cash Flow. Review of Accounting Studies 11 (2006): 159-189. Stein, J. Agency, Information and Corporate Investment. in Handbook of the Economics of Finance, edited by G. Constantinides, M. Harris, and R. Stulz. Amsterdam: North-Holland, 2003. Stubben, S. 2010. Discretionary Revenues as a Measure of Earnings Management. Forthcoming. The Accounting Review. World Bank. Review of Small Business Activities. World Bank report, Washington, D.C., 2007.

32

Appendix: Definitions of Variables

Variables in the investment model Invest = The sum of new investment in machinery, equipment, vehicles, buildings, land, and R&D activities, less the sale of fixed assets in the current year, scaled by the lagged total assets. RevGrowth = The yearly growth rate of sales from year t-2 to year t-1. Major variables in the investment efficiency test InvEff = A measure of investment efficiency, which is measured as the absolute values of the residuals from the investment model:
Invest i ,t = 0 + 1 N E G i ,t 1 + 2 Re vG row thi ,t 1 + 3 N E G * R evG row thi ,t 1 + i , t .

Overinvestment Underinvestment EQ (measured by DisAccr) EQ (measured by DisRev) Bank

= = = = =

Tax

The indicator variable NEGi,t-1 takes the value of one for negative revenue growth, and zero otherwise. The investment model is estimated crosssectionally with at least eight observations in each industry by country. To mitigate the influence of outliers, we winsorize all variables at the 1% and 99% levels. Positive residuals from the investment model. Absolute value of the negative residuals from the investment model. The absolute residual of the Kothari et al. (2005) discretionary accrual model, multiplied by -1. The absolute value of the measure of discretionary revenues presented in McNichols and Stubben (2008), multiplied by -1. An indicator variable that takes on the value of one if bank financing accounts for the highest percentage of total financing among all external sources of financing, and its percentage is larger than 5% among all financing sources; and zero otherwise. An indicator variable that takes on the value of one if the average degree of obstacle is 3 or greater for both tax rates and tax administration; and zero otherwise.

Control variables in the investment efficiency test Size = The log of lagged total assets. Age = The log of the age of the firm in years. Tang = The ratio of PPE to total assets in year t-1. Slack = The ratio of cash to total assets in year t-1. Additional variables used in the additional tests BConIndex = The business constraint index, which is the average of indicator variables defined as equal to 1 if the firm perceives major or very severe obstacles in telecommunications, electricity, transportation, access to land, and skills and education of available workers, and zero otherwise. LConIndex = The legal constraint index, which is the average of indicator variables defined as equal to 1 if the firm perceives major or very severe obstacles in corruption, crime, and anticompetitive practices, and zero otherwise. PEConIndex = The political and economic index, which is the average of indicator 33

Lev StdRev StdInvest OperCycle

= = = = = = = = = = = =

Loss Export Foreign FinCon

Control Audit Collateral GNI

variables defined as equal to 1 if the firm perceives major or very severe obstacles in customs regulations, labor regulations, business licensing, regulatory policy uncertainty, and the macroeconomic situation, and zero otherwise. Financial leverage, measured as long-term liabilities divided by total assets. The standard deviation of sales in the past three years. The standard deviation of investments in the past three years. Operating cycle calculated by the following formula: (average accounts receivables/sales)365 + (average inventory/cost of goods sold)365. We use the log of operating cycle for our tests. An indicator variable referring to whether a firm reports a loss. An indicator variable referring to whether the firm exports. An indicator variable referring to whether any foreign company or individual has a financial stake in the ownership of the firm. An indicator variable that takes on the value of one if the degree of obstacle for both access to financing and cost of financing is larger than 3; and zero otherwise. An indicator variable that takes on the value of one if the largest shareholder owns more than 50% of the firm. An indicator variable that takes on the value of one if annual financial statement is reviewed by external auditor; and zero otherwise. The approximate value of collateral required as a percentage of the loan value. Gross national income per capita at a country level, which is provided by the World Bank using the Atlas method. The Atlas method smoothes exchange rate fluctuations by using a three-year moving average, priceadjusted conversion factor.

All data are from the World Banks Enterprise Survey.

34

Table 1: Sample Selection

All firm-years in the WBES database with listing status information Less firms residing in Other sectors Less publicly traded firms Private firms with all necessary industry information Less firms with missing data on revenue growth Less firms with missing data on investment Additional reduction due to the data requirement of the investment model (i.e., at least ten firms in each industry by country for Equation (1)) Firms with the investment efficiency variable (InvEff) 47,712 740 3,200 43,742 27,804 5,869 77 9,992

(1) Missing data on discretionary revenue (DisRev) and control variables (Table 4) Sample size for the main tests with DisRev as a proxy for accounting quality

3,130 6,862

(2) Missing data on discretionary accruals (DisAccr) and control variables (Table 4) Sample size for the main tests with DisAccr as a proxy for accounting quality

3,295 6,697

35

Table 2: Sample Distribution by Country Country

Bangladesh Brazil Ecuador El Salvador Eritrea Ethiopia Guatemala Honduras India Indonesia Nicaragua Oman Pakistan Philippines South Africa Sri Lanka Syria Tanzania Thailand Vietnam Zambia Total

N
489 1,035 56 88 24 39 67 43 528 337 39 37 855 373 391 297 44 77 1,342 591 110 6,862

Percentage
7.13% 15.08% 0.82% 1.28% 0.35% 0.57% 0.98% 0.63% 7.69% 4.91% 0.57% 0.54% 12.46% 5.44% 5.70% 4.33% 0.64% 1.12% 19.56% 8.61% 1.60% 100%

This table shows the sample distribution by country. There are a total number of 6,862 observations from 21 countries.

36

Table 3: Descriptive Statistics of the Main Sample Panel A: Descriptive Statistics

Variable InvEff Overinvestment Underinvestment DisRev DisAccr Bank Tax Age Size (in thousands) Tang Slack N 6,862 1,843 5,019 6,862 6,697 4,657 6,830 6,862 6,862 6,862 6,862 Mean 0.001 0.12 0.043 0 -0.001 0.384 0.171 17.734 60,900 0.364 0.069 25% -0.05 0.024 0.019 -0.029 -0.063 0 0 9 170 0.124 0.007 median -0.023 0.079 0.037 -0.005 -0.004 0 0 13 912 0.338 0.028 75% 0.004 0.215 0.06 0.018 0.048 1 0 22 6,073 0.563 0.09 STD 0.096 0.112 0.03 0.074 0.151 0.486 0.377 14.537 201,000 0.266 0.103

Panel B: Correlation Matrix

InvEff DisRev DisAccr Bank Tax Age Size Tang Slack -0.161*** -0.049*** -0.033** 0.014 -0.097*** -0.085** 0.038*** 0.040*** 0.426*** -0.038** 0.019** 0.073*** 0.076*** 0.181*** -0.018 -0.054*** 0.014* 0.068*** 0.137*** 0.201*** -0.039*** -0.085*** -0.042*** 0.085*** 0.095*** -0.102*** 0.017 0.077*** -0.008 -0.029*** 0.150*** -0.132*** -0.042*** 0.082*** -0.180*** -0.136*** DisRev DisAccr Bank Tax Age Size Tang

Panel A presents descriptive statistics for the variables used in the analyses. Panel B presents Pearson correlations for these variables. Variable computations are described in the Appendix. The exceptions are in Panel A, where InvEff is the signed residual from the investment model, and DisRev and DisAccr refer to the signed residuals from the discretionary revenue model and the discretionary accruals models, respectively. *, **, and *** represent significance at the 10%, 5%, and 1% level (two-tailed test).

37

Table 4: The Relation between Accounting Quality and Investment Efficiency

InvEffi ,t +1 = 0 + 1EQi ,t + nControl Variablesi ,t + Industries + Countries + i ,t

Dependent Variable = Overinvestment EQ measured by DisAccr Variables EQ Size Age Tang Slack Intercept Predictions - (H1) (1) -0.094*** (-4.04) -0.023*** (-7.01) -0.010 (-1.16) 0.0227* (1.73) 0.018 (0.70) 0.259*** (11.07) EQ measured by DisRev (2) -0.091*** (-2.49) -0.024*** (-7.19) -0.007 (-0.85) 0.023* (1.78) 0.024 (0.93) 0.265*** (11.08) Dependent Variable = Underinvestment EQ measured by DisAccr (3) -0.020*** (-6.09) -0.001** (-2.37) -0.001 (-0.37) 0.001 (0.60) 0.003 (1.05) 0.047*** (16.75)

(4)

EQ measured by DisRev (4) -0.060*** (-7.47) -0.001*** (-2.46) 0.001 (0.36) 0.001 (0.51) 0.001 (0.06) 0.047*** (16.12)

Country fixed effects Industry fixed effects Adjusted R2 N

Yes Yes

Yes Yes

Yes Yes

Yes Yes

0.141 1,878

0.136 1,843

0.428 4,819

0.458 5,019

The table reports OLS regression results. Variables are defined in the Appendix. *, **, and *** represent significance at the 10%, 5%, and 1% level, respectively, using one-tailed (two-tailed) tests for the test variables (control variables). Huber-White robust standard errors are used to control for heteroskedasticity.

38

Table 5: The Conditional Effect of Bank Financing on the Relation between Accounting Quality and Investment Efficiency
InvEff i ,t +1 = 0 + 1 EQi ,t + 2 Banki ,t + 3 Banki ,t EQi ,t + n Control Variablesi ,t + Industries + Countries + i ,t
Dependent Variable = Overinvestment EQ measured by DisAccr Variables EQ Bank EQBank Size Age Tang Slack Intercept Country fixed effects Industry fixed effects Adjusted R2 N - (H2) Predictions - (H1) (1) -0.112*** (-3.63) 0.004 (0.42) -0.053 (-1.01) -0.024*** (-6.41) -0.012 (-1.21) 0.032** (2.30) 0.028 (0.97) 0.261*** (10.13) Yes Yes EQ measured by DisRev (2) -0.139*** (-2.74) 0.008 (0.95) -0.141* (-1.60) -0.025*** (-6.60) -0.009 (-0.94) 0.032** (2.21) 0.032 (1.10) 0.268*** (10.12) Yes Yes Dependent Variable = Underinvestment EQ measured by DisAccr (3) -0.009** (-2.04) 0.001 (1.29) -0.015** (-2.24) -0.001 (-0.93) -0.001 (-0.51) 0.003 (1.43) 0.010*** (2.51) 0.043*** (12.07) Yes Yes EQ measured by DisRev (4) -0.034*** (-2.76) -0.002* (-1.49) -0.047*** (-2.46) -0.001 (-1.26) 0.001 (0.32) 0.002 (0.98) 0.004 (0.99) 0.045*** (12.28) Yes Yes

(5)

0.091 1,433

0.087 1,409

0.267 3,024

0.298 3,248

The table reports OLS regression results. Compared with Table 4, we further require availability of data for a firms financing sources. Variables are defined in the Appendix. *, **, and *** represent significance at the 10%, 5%, and 1% level, respectively, using one-tailed (two-tailed) tests for the test variables (control variables). Huber-White robust standard errors are used to control for heteroskedasticity.

39

Table 6: The Conditional Effect of Tax Burdens on the Relation between Accounting Quality and Investment Efficiency
InvEffi ,t +1 = 0 + 1 EQi ,t + 2Taxi ,t + 3Taxi ,t EQi ,t + n Control variablesi ,t + Industries + Countries + i ,t
Dependent Variable = Overinvestment EQ measured by DisAccr Variables EQ Tax EQTax Size Age Tang Slack Intercept Country fixed effects Industry fixed effects Adjusted R2 N + (H3) Predictions - (H1) (1) -0.121*** (-4.71) 0.018** (2.13) 0.150*** (3.06) -0.023*** (-6.93) -0.010 (-1.17) 0.020 (1.56) 0.021 (0.83) 0.255*** (10.87) Yes Yes EQ measured by DisRev (2) -0.122*** (-3.07) 0.012* (1.41) 0.194*** (2.41) -0.024*** (-7.20) -0.007 (-0.85) 0.022* (1.69) 0.026 (1.00) 0.263*** (10.97) Yes Yes Dependent Variable = Underinvestment EQ measured by DisAccr (3) -0.021*** (-6.13) -0.001 (-1.34) 0.004 (0.43) -0.001** (-2.20) -0.001 (-0.34) 0.001 (0.73) 0.003 (1.09) 0.047*** (16.58) Yes Yes EQ measured by DisRev (4) -0.067*** (-7.85) -0.001 (-0.28) 0.033* (1.53) -0.001** (-2.32) 0.001 (0.37) 0.001 (0.56) 0.001 (0.11) 0.046*** (15.95) Yes Yes

(6)

0.142 1,870

0.137 1,836

0.428 4,793

0.459 4,994

The table reports OLS regression results. Compared with Table 4, we further require availability of data for a firms tax burden. Variables are defined in the Appendix. *, **, and *** represent significance at the 10%, 5%, and 1% level, respectively, using one-tailed (two-tailed) tests for the test variables (control variables). Huber-White robust standard errors are used to control for heteroskedasticity.

40

Table 7: The Conditional Effect of Bank Financing and Tax Burdens on the Relation between Accounting Quality and Investment Efficiency
InvEff i ,t +1 = 0 + 1 EQi ,t + 2 Banki ,t + 3 Banki ,t EQi ,t + 4Taxi ,t + 5Taxi ,t EQi ,t + n Control variablesi ,t + Industries + Countries + i ,t
Dependent Variable = Overinvestment EQ measured by DisAccr Variables EQ Bank EQBank Tax EQTax Size Age Tang Slack Intercept Country fixed effects Industry fixed effects Adjusted R2 N + (H3) - (H2) Predictions - (H1) (1) -0.153*** (-4.48) 0.005 (0.54) -0.068 (-1.33) 0.017* (1.76) 0.182*** (3.44) -0.023*** (-6.28) -0.012 (-1.20) 0.030** (2.17) 0.031 (1.07) 0.255*** (9.87) Yes Yes EQ measured by DisRev (2) -0.187*** (-3.38) 0.010 (1.11) -0.167* (-1.63) 0.009 (0.99) 0.226*** (2.53) -0.025*** (-6.57) -0.009 (-0.94) 0.032** (2.21) 0.035 (1.17) 0.265*** (9.95) Yes Yes Dependent Variable = Underinvestment EQ measured by DisAccr (3) -0.010** (-1.96) 0.002* (1.44) -0.014** (-2.13) -0.002* (-1.40) 0.001 (0.04) -0.001 (-0.89) -0.001 (-0.42) 0.003 (1.56) 0.010*** (2.58) 0.042*** (11.99) Yes Yes EQ measured by DisRev (4) -0.039*** (-3.02) -0.002 (-1.30) -0.045** (-2.31) -0.001 (-0.66) 0.015 (0.54) -0.001 (-1.26) 0.001 (0.39) 0.002 (1.07) 0.005 (1.06) 0.045*** (12.20) Yes Yes

(7)

0.094 1,426

0.088 1,403

0.266 3,002

0.298 3,226

The table reports OLS regression results. Compared with Table 4, we further require availability of data for firms financing sources and tax burden. Variables are defined in the Appendix. *, **, and *** represent significance at the 10%, 5%, and 1% level, respectively, using one-tailed (two-tailed) tests for the test variables (control variables). Huber-White robust standard errors are used to control for heteroskedasticity.

41

Table 8: The Conditional Effect of Bank Financing and Tax Burdens on the Relation between Accounting Quality and Investment Efficiency
InvEff i ,t +1 = 0 + 1 EQi ,t + 2 Banki ,t + 3 Banki ,t EQi ,t + 4Taxi ,t + 5Taxi ,t EQi ,t + n Control variablesi ,t + Industries + Countries + i ,t

(7)

Panel A: Discretionary Accruals (DisAccr) as a proxy for EQ

Dependent Variable = Overinvestment DisAccr > 0 Variables DisAccr Predictions + if DisAccr>0; - if DisAccr<0 (H1) (1) 0.114*** (2.42) 0.006 (0.45) + if DisAccr>0; - if DisAccr<0 (H2) 0.050 (0.72) 0.018 (1.32) - if DisAccr>0; + if DisAccr<0 (H3) -0.171*** (-2.53) DisAccr < 0 (2) -0.233*** (-4.76) 0.003 (0.24) -0.113* (-1.46) 0.013 (0.90) 0.197*** (2.34) Dependent Variable = Underinvestment DisAccr > 0 (3) 0.012* (1.63) 0.002 (0.96) 0.012** (2.17) -0.001 (-0.36) -0.007 (-0.46) DisAccr < 0 (4) -0.009* (-1.60) 0.001 (0.84) -0.020** (-2.05) -0.003* (-1.44) 0.005 (0.30)

Bank

DisAccrBank Tax DisAccrTax

Control Variables Country fixed effects Industry fixed effects Adjusted R2 N

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

0.090 688

0.111 738

0.245 1,452

0.291 1,550

42

Panel B: Discretionary Revenue (DisRev) as a proxy for EQ

Dependent Variable = Overinvestment DisRev > 0 Variables DisRev Bank DisRevBank Tax DisRevTax - if DisRev>0; + if DisRev<0 (H3) + if DisRev >0; - if DisRev<0 (H2) Predictions + if DisRev>0; - if DisRev<0 (H1) (1) 0.073 (1.97) 0.003 (0.19) 0.047 (0.46) -0.001 (-0.01) -0.257** (-2.07)
**

Dependent Variable = Underinvestment DisRev > 0 (3) 0.042 (2.64)

***

DisRev < 0 (2) -0.317 (-3.65) 0.016* (1.39) -0.311* (-1.62) 0.018* (1.46) 0.270** (2.07)
***

DisRev < 0 (4) -0.039* (-1.58) -0.001 (-0.24) -0.031** (-1.93) -0.001 (-0.53) 0.012 (0.23)

-0.003** (-1.85) 0.062*** (2.54) -0.001 (-0.08) -0.031 (-0.98)

Control Variables Country fixed effects Industry fixed effects Adjusted R2 N

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

0.126 554

0.079 849

0.292 1,350

0.305 1,876

The table reports OLS regression results. Variables are defined in the Appendix. The variable EQ refers to either DisAccr (in Panel A) or DisRev (in Panel B). *, **, and *** represent significance at the 10%, 5%, and 1% level, respectively, using one-tailed tests for the test variables. Huber-White robust standard errors are used to control for heteroskedasticity.

43