Flexibility in Cash Flow Classification Under IFRS: Determinants and Consequences

International Financial Reporting Standards (IFRS) allow managers flexibility in classifying interest paid, interest received, and dividends received within operating, investing, or financing activities within the statement of cash flows. In contrast, U.S. Generally Accepted Accounting Principles (GAAP) requires these items to be classified as operating cash flows (OCF). Studying IFRS-reporting firms in 13 European countries, we document firms’ cash-flow classification choices vary, with about 76%, 60%, and 57% of our sample classifying interest paid, interest received, and dividends received, respectively, in OCF. Reported OCF under IFRS tends to exceed what would be reported under U.S. GAAP. We find the main determinants of OCF-enhancing classification choices are capital market incentives and other firm characteristics, including greater likelihood of financial distress, higher leverage, and accessing equity markets more frequently. In analyzing the consequences of reporting flexibility, we find some evidence that the market’s assessment of the persistence of operating cash flows and accruals varies with the firm’s classification choices, and the results of certain OCF prediction models are sensitive to classification choices.


Introduction
In this paper, we examine the determinants and consequences of comparative flexibility in classification choices within the statement of cash flows. International Financial Reporting Standards (IFRS) are perceived to allow managers more flexibility than generally accepted accounting principles in the United States (U.S. GAAP). This increased flexibility is apparent with regard to the classification of certain items within the statement of cash flows. U.S. GAAP requires that firms classify interest paid, interest received, and dividends received as operating cash flows. In contrast, IFRS allows firms the flexibility to report these items within operating cash flow (OCF) or, alternatively, to classify them as investing or financing. We describe variation in firms' cash flow classification choices under IFRS, identify capital market incentives and firm reporting environment characteristics associated with these choices, and document consequences of classification flexibility.
Cash flow, and particularly OCF, is well established as a basis for business valuation (e.g., Damodaran 2006, Imam et al. 2008), 1 contracting (e.g., Dichev and Skinner 2002;Mulford and Comiskey 2005), and financial analysis (e.g., Estridge and Lougee 2007). Although an extensive literature examines classification shifting within the income statement and within the balance sheet 1 Imam et al. (2008) present evidence that discounted cash flow models and price earnings multiples are the valuation models most preferred by analysts. Liu et al. (2007, 56), who present evidence that earnings multiples dominate cash flow multiples in predicting share price, nonetheless note that many practitioners prefer to use cash flows rather than earnings as a basis for valuation using multiples, "arguing that accruals involve discretion and are often used to manipulate earnings …. And expenses such as depreciation and amortization deviate substantially from actual declines in value because they are based on ad hoc estimates that are, in turn, derived from potentially meaningless historical costs." (Engel et al. 1999;Marquardt and Wiedman 2005;McVay 2006; Bartov and Mohanram 2014), less attention has been given to classification variations within the statement of cash flows (Lee 2012) and classification restatements (Hollie et al. 2011). IFRS reporting provides a setting where the accounting standards provide firms flexibility in classification choices within the statement of cash flows.
The effect of flexibility in cash flow classification and its consequences are important because both the International Accounting Standards Board (IASB) and Financial Accounting Standards Board (FASB) share the objective that financial information should enable financial statement users to better predict future cash flows. 2 Further, the Boards articulate the importance of both accrual accounting information and cash flow information in achieving this objective.
"Information about a reporting entity's cash flows during a period also helps users to assess the entity's ability to generate future net cash inflows. It indicates how the reporting entity obtains and spends cash, including information about its borrowing and repayment of debt, cash dividends or other cash distributions to investors, and other factors that may affect the entity's liquidity or solvency. Information about cash flows helps users understand a reporting entity's operations, evaluate its financing and investing activities, assess its liquidity or solvency and interpret other information about financial performance." 3 Despite identical objectives, standard setters have established different requirements for presentation of certain itemsinterest paid, interest received, and dividends receivedin the statement of cash flows. As a consequence, the amount of OCF reported by a given entity can differ under U.S. GAAP and IFRS. Theoretically, the appropriate classification of these items is open to 2 In IFRS, the Conceptual Framework, Chapter 1, The Objective Of General Purpose Financial Reporting ¶OB3 states: "Decisions by existing and potential investors about buying, selling or holding equity and debt instruments depend on the returns that they expect from an investment in those instruments, for example dividends, principal and interest payments or market price increases. Similarly, decisions by existing and potential lenders and other creditors about providing or settling loans and other forms of credit depend on the principal and interest payments or other returns that they expect. Investors', lenders' and other creditors' expectations about returns depend on their assessment of the amount, timing and uncertainty of (the prospects for) future net cash inflows to the entity. Consequently, existing and potential investors, lenders and other creditors need information to help them assess the prospects for future net cash inflows to an entity." In U.S. GAAP, Concepts Statement No. 8 ¶ OB3 is identical. 3 IFRS Conceptual Framework, Chapter 1, ¶OB20, which is identical to U.S. GAAP,Concepts Statement No. 8 ¶ OB20. debate. Even when deliberating the adoption of the statement of cash flows standard, Statement of Financial Accounting Standards No. 95 (SFAS 95), members of the FASB discussed the classifications of interest paid and interest received, ultimately opting to require these items be reported in the operating section. 4 We initially document variation in classification choices within a hand-collected sample of 798 non-financial IFRS firms in 13 European countries from 2005 to 2012. About 76%, 60%, and 57% of the sample classifies interest paid, interest received, and dividends received, respectively, in OCF. Only about 42% of the sample firms that report all three items opt to classify all three in OCF.
We document significant variation in classification across industries and most countries. The second set of analyses focuses on determinants of firms' cash flow classification choices from the perspective of OCF-increasing classifications. We examine incentives to inflate reported OCF, similar to Lee (2012), including capital market incentives, financial distress, the presence of analysts' cash flow forecasts, and profitability. Further, we explore characteristics associated with the reporting environment such as analyst following, classification choices of industry peers, cross-listing in the U.S., country and industry.
In our determinants analysis, we construct two dependent variables as proxies for OCFincreasing classification choices: (1) the amount of the difference in reported OCF under IFRS and a benchmark measure of what OCF would have been under U.S. GAAP, and (2) an indicator variable signifying a classification choice that would increase OCF under IFRS relative to U.S. GAAP. For the first of these variables, we create a hypothetical benchmark by adjusting each firm's OCF to include interest paid, interest received, and dividends received (i.e., consistent with U.S. GAAP requirements). That is, we consider a hypothetical U.S. GAAP benchmark assuming that managers' real operating activities would have remained the same even if cash flow classification choices had been restricted. We do not assert these items are appropriately classified as OCF.
Rather, we use U.S. GAAP classification as a benchmark because our main focus is on the differences between U.S. GAAP and IFRS. For the second of these dependent variables, we focus on the classification choice for one item, interest paid, which IFRS permits to be classified either in the operating or the financing section of the statement of cash flows. We focus on interest paid because it usually constitutes a relatively large amount relative to interest received and dividends received, is commonly reported, is typically reported separately and is thus easier to identify. It is also possible that a firm has more control over the amount and timing of cash outflows (i.e. payments) as opposed to cash inflows (i.e. receipts) thus making interest paid more susceptible to use as an OCF-increasing item. 5 When a firm classifies interest paid as financing, it follows that ceteris paribus reported OCF will be higher than if interest paid had been classified as operating.
Thus classification of interest paid as financing is an OCF-increasing classification choice.
A cross-sectional determinants analysis of all firms with consistent classification during the study period indicates that actually-reported OCF exceeds benchmark-OCF by a greater amount for firms with weaker financial positions (i.e., greater likelihood of financial distress, higher leverage, and lower profitability). Firms with higher amounts of equity-raising activity also make greater OCF-increasing classification choices. For the determinants analysis using an indicator variable signifying classification choice, we find that firms with higher leverage are more likely to make an OCF-enhancing choice and firms cross-listed in the United States are more likely to make a classification choice that is consistent with U.S. GAAP. We find no effects related to homogeneity of industry practice or to the presence of analysts' cash flow forecasts.
An examination of 99 firms that change classifications during our sample period reveals that 58% make OCF-increasing classification choices. The most common change is a reclassification of interest paid out of operating, an OCF-increasing choice. Analysis indicates that an OCF-increasing reclassification is more likely for firms with greater equity issuance and less likely for firms with more analyst coverage, homogeneity of industry practice, and a cross-listing in the U.S.
Variation in classification of cash flow items also introduces non-comparability into measurement of widely-used metrics such as accruals and free cash flow. 6 Therefore, the final set of analyses focuses on consequences of flexibility in classification choice. The first consequence we examine pertains to the market pricing of persistence of cash flows and accruals. We examine whether the persistence of cash flows and accruals differs for firms that report consistently with U.S. GAAP compared to those firms making classification choices permitted under IFRS. We find some evidence of differences in accrual pricing between the group of firms reporting consistently with U.S. GAAP and those using the classification flexibility allowed under IFRS, but results are sensitive to model specification. 6 Accruals are sometimes measured as the difference between earnings and cash flows from operating activities, and free cash flow is often measured as operating cash flow minus capital expenditures. The alternative Hribar and Collins (2002) measure of accruals based on the balance sheet, even if superior, is not always feasible in an international setting.
A second consequence we examine pertains to models for predicting cash flows that have been used in prior accounting research. We find that differences in cash classification choices affect results when the cash flow prediction model is based on prior sales (Dechow et al. 1998, Roychowdhury 2006, but not when the cash flow prediction model is based on prior cash flows (Barth et al. 2001, Givoly et al. 2009). One implication is that the latter type of model may be more useful in the international context in which flexibility in cash flow classification exists.
Our study contributes to literature on managerial discretion in non-earnings measures, especially in an international context. Although managerial discretion in cash flow classification could be potentially helpful to financial statement users, our evidence suggests that classification choices are associated with incentives to report higher OCF. We also find that the likelihood of making an OCF-increasing change in classification is positively associated with equity issuance but negatively associated with analysts' coverage, consistent with analysts serving some deterrent role.
Similarly, being cross-listed in the U.S. decreases the likelihood of making a cash-flow classification change.
Our study also contributes to the debate over costs and benefits of comparability and uniformity (De Franco et al. 2011). Flexibility in cash flow reporting may result in lower comparability and uniformity and thus theoretically creates costs for users which are potentially significant because of the use of cash flows in valuation and contracting. 7  arguably lead to more informative OCF, our findings indicate that such flexibility impacts the comparability of reported OCF. 8 Our study should be of interest to various audiences. Cash flow classification choices available under IFRS, but not under U.S. GAAP, potentially limit the generalizability of U.S. evidence relying on reported OCF. 9 Researchers comparing OCF and other performance measures should be interested in the effects of classification on their estimates (e.g., Bernard and Stober 1989;Sloan 1996;Ashbaugh and Olsson 2002;Orpurt and Zang 2009;Barton et al. 2010).
With an increasing number of countries permitting or requiring IFRS (De George et al. 2016), our findings should inform regulators, including the U.S. SEC which has expressed plans regarding convergence and potential adoption of IFRS (SEC 2011). As IFRS allows more flexibility than U.S. GAAP, U.S. regulators should be particularly interested in the variation in firms' classification choices and the factors associated with those choices. 10 Standard setters can potentially utilize an understanding of the factors associated with a firm's reporting choices when crafting standards that permit alternatives. In addition, financial statement users may benefit from understanding whether and how management's cash flow classification choices relate to reporting incentives and firm characteristics (Carslaw and Mills 1991).
The paper is organized as follows. Section 2 discusses the motivation and research design.
Section 3 describes our sample selection and presents a comprehensive description of cash flow classification of interest paid, interest received, and dividends received. Section 4 reports results of the determinants of firms' cash flow classification choices while Section 5 includes the analysis of specific consequences of flexibility in classification choice. Section 6 concludes.

Determinants of OCF classification choices
We follow Lee (2012) and explore incentives and reporting environment factors related to reporting higher OCF. 11 Incentives for reporting higher OCF relate broadly to capital access and contracting. Additionally, reporting environment factors affecting classification choice include industry and market aspects (analysts' forecasts and cross-listing).
Because OCF is an important measure in assessing credit and default risk (Beaver 1966;Ohlson 1980;DeFond and Hung 2003), we expect that firms closer to financial distress are motivated to report higher OCF. We create a proxy for financial distress based on Altman's Z-score (Altman and Hotchkiss 2006). 12 Arguably, firms accessing equity markets more frequently have stronger incentive to inflate OCF to increase their valuation and thus the amount of capital they can raise. Therefore, we expect these firms are more likely to make classifications that enhance their reported OCF. Our proxy for capital market incentives is equity issuances. We expect that the more firms opt to access the equity markets, the stronger incentives they have to report higher OCF. Thus, we expect a positive relation between equity issues and OCF-increasing classification choices.
We predict that firms with contracting concerns and costs involved in renegotiating debt covenants will also seek to report higher OCF. Our proxy for contracting concerns is leverage, computed as total liabilities divided by total assets. We predict a positive relation.
We expect that profitability will be associated with OCF-increasing classification choices.
On the one hand, less profitable firms could be more likely to make OCF-increasing classification choices, managing OCF upward to compensate for weakness in reported profits. On the other hand, more profitable firms could be likely to make OCF-enhancing classification choices to reflect better cash flow performance consistent with income performance. Therefore, we do not predict the sign of the association between profitability and OCF-increasing choices.
We examine three explanatory variables related to the firm's information environment: (1) availability of analysts' cash flow forecast, (2) industry practice, and (3)  Our second information-environment variable, industry practice, is relevant to classification choice because firms could be motivated to increase cross-sectional comparability by making classification choices consistent with those of their peer industry group. 13 For example, when considering the choice of where to report interest paid, a firm could be disadvantaged by classifying interest paid as operating and thus reporting comparatively lower OCF when, for example, the majority of its industry peers classify interest paid as financing. Alternatively, a firm could make a different choice to distinguish itself from its industry peers and, possibly, report higher OCF. In this case, OCF-increasing choices would not be expected to be associated with industry practice.
Therefore, we have no prediction on the sign of the homogeneity of firms' classification choices within an industry.
Our third information environment variable pertains to cross-listing. Bradshaw et al. (2004) argue that firms with cross-listings in the United States have stronger incentives to adopt similar reporting choices as U.S. companies. Empirically, their data show a positive correlation between U.S. GAAP conformity and cross-listing. Therefore we expect that cross-listed firms are less likely to classify items such as interest paid in financing, which is not allowed under U.S. GAAP.
We include firm size to capture financial reporting incentives, financial reporting expertise, and the financial reporting environment of large versus small firms. We do not have a prediction for its sign. Finally, we include indicator variables for country and industry. The regression model is as follows: OCF_Classification i = a 0 + a 1 Distress_Hi i + a 2 Equity Issues i + a 3 Leverage_Hi i + a 4 Profitability i + a 5 Analysts Cash Flow Forecast i (1) + a 6 Industry Homogeneity i + a 7 Cross-listed in US i + a 8 Size i + e i Where:

OCF_Classification i is either OCF_Reported t less OCF_ Pro forma_USGAAP t or Interest Paid in
Financing. OCF_Reported t less OCF_ Pro forma_USGAAP t = operating cash flows as reported by the firm for year t less operating cash flows for year t adjusted to include interest paid, interest received, and dividends received in operating cash flows if these items are not already reported in the operating section, averaged over the sample period. Interest Paid in Financing = 1 if the firm classifies interest paid in financing cash flows as of the last year reported, and zero otherwise. Distress_Hi = 1 if the firm's financial distress computed using Altman's Z-score is less than 1.81, indicative of high distress, and zero otherwise. Equity Issues = percent change in the firm's contributed capital over the sample period. Leverage_Hi = 1 if the firm's ratio of total liabilities over total assets at the beginning of the fiscal year, averaged over the sample period, is greater than the median, and zero otherwise. Profitability = the firm's net income divided by beginning total assets, averaged over the sample period. Analysts Cash Flow Forecast = 1 if at least one analyst's cash flow forecast is available on IBES, and zero otherwise, averaged over the sample period. Industry Homogeneity = the percent of firms within an industry that report interest paid in financing cash flows, with industry classifications based on Barth et al. (1998). Cross-listed in US = 1 if the firm is cross-listed in the United States, and zero otherwise. Size = the natural logarithm of the firm's beginning of year market capitalization in U.S. dollars, averaged over the sample period. Regressions include country, industry, and year controls.
We create one observation per firm summarizing data available during the sample period to compute the variables in the model. Firms with consistent classification over time are analyzed separately from firms that changed classification. 14 To examine the relation between the variables described above and the magnitude of the effect of IFRS-permitted classification choices, we estimate an OLS regression model using the dependent variable, OCF_Reported t less OCF_ Pro forma_USGAAP t . To examine the relation between the variables described above and the likelihood of an OCF-enhancing classification choice, we estimate a logistic regression in which the dependent variable is Interest Paid in Financing.

Determinants of OCF-increasing reclassifications
Because cross-sectional variations in the classification within the statement of cash flows might result from historical legacy for each firm, the subsample of firms that change classification 14 We examine these classification changers separately in Section 2.2. offers a potentially cleaner setting to examine the determinants of classification choice. The Appendix presents an illustrative example of one company that changed its classifications of interest paid and interest received. In 2007, Norse Energy Corp. ASA, a Norwegian gas explorer and producer, changed its classification of interest paid to financing from operating. Norse Energy also changed its classification of interest received from operating to investing. The net effect of these changes was to report positive, rather than negative operating cash flows, in both 2007 and

15
The various classification changes impact reported operating cash flow differently. To examine determinants of classification choice, we therefore focus on firms that increased OCF by making the classification change. We compare the OCF-increasing changers to a control group of firms that did not make a classification change, and specifically non-changing firms with existing Prior research shows that the cash flow component of earnings is more persistent than the accrual component, yet market pricing does not always correctly reflect the relatively greater persistence (Sloan 1996;Dechow et al. 2008;Pincus et al. 2007.) In the context of cash flow classification, the question remains whether investors anticipate the persistence of reported operating cash flows and accruals similarly, regardless of where cash flow items are classified. Our analysis focuses on a comparison of the persistence parameters for accruals and cash flow components of earnings with the parameters that are implied by stock returnssimilar to the approach in Sloan (1996) and Dechow et al. (2008).
where EARNINGS t+1 = the amount of net income for year t divided by average of total assets for year t. ACCR_Reported t = the amount of accruals, calculated as net income less reported operating cash flows for year t divided by average of total assets for year t. OCF_Reported t = the reported amount of operating cash flow for year t divided by average of total assets for year t. Controls = Size t , BM t and EP t .
Size t = the natural logarithm of the firm's market capitalization in U.S. dollars at the beginning of year t. BM t = the firm's book to market ratio, calculated as the ratio of the firm's shareholders' equity divided by its market capitalization at the beginning of year t. EP t = the firm's net income divided by its market capitalization at the beginning of year t.
We undertake this analysis separately for the subsample of firms with classification choices that reflect the flexibility under IFRS (FLEX = 1) and the subsample with classification choices similar to those under U.S. GAAP. The coefficients  1 and  2 from the forecasting equation (2) indicate the persistence of the two components of earnings: accruals and cash flow. Prior research has shown that the cash flow component of earnings is more persistent than the accruals component.
We examine whether the relationship, α 1 > α 2 , holds for both subsamples. An impact of differences in classification choice would be indicated by differences in comparative persistence parameters for accruals and OCF.
A comparison is also made between the coefficients from the market pricing equation 1 # and from the forecasting equation α i . Presence of the accrual anomaly, for example, is indicated by market underweighting cash flow ( 2 # < α 2 ) and overweighting accruals ( 1 # > α 2 ). In the international context, Pincus et al. (2007) provide evidence of the accrual anomaly only in certain countries; therefore, our focus is not on whether we find evidence of the accrual anomaly. Rather, we examine whether the comparative relationships between market pricing of the cash flow and accrual components differs for the two subsamples.

Consequences: Models of OCF prediction
Next, we examine models of future operating cash flow prediction. Cash flow prediction models are used both to develop expected cash flows (Dechow et al. 1998, Roychowdhury 2006, Kim and Park 2014 and to determine whether accounting measures are predictive of future cash flows (Barth et al. 2001, Givoly et al. 2009, Badertscher et al. 2012. We investigate whether the cash flow classification choices have different implications for the prediction of future cash flows. The first model we examine uses past sales and changes in sales to predict OCF based on Dechow et al. (1998): Where: 1/TA t = 1 divided by the average of total assets for year t. S t /TA t = sales revenue for year t divided by the average of total assets for year t. FLEX * S t /TA t is the interaction between the indicator variable FLEX and sales revenue for year t.
ΔS t /TA t is change in sales revenue from year t-1 to year t divided by the average of total assets for year t. FLEX * ΔS t /TA t is the interaction between the indicator variable FLEX and change in annual sales divided by the average of total assets for year t. Regressions include country, industry, and year controls.
In this model, the variables of interest are the FLEX interactions with sales and changes in sales,  3 and  5 . The coefficients on these variables will be significant if the firm's IFRS classification choices result in different predicted future OCF than would U.S. GAAP classification choices. Because OCF using FLEX classification choices is higher on average than OCF using U.S.
GAAP classification choices, we expect the FLEX interaction coefficients to be positive. If the classification does not relate to the future OCF, the FLEX interaction coefficients will not be significant. We expect the coefficients on sales and changes in sales to be positive and significant, consistent with prior research. The second prediction model uses past OCF and accruals to predict future OCF similar to Barth et al. (2001).
where all variables are as previously defined. These regressions include country, industry, and year controls.
In this model, the coefficients on the FLEX interactions with accruals and past OCF,  2 and  4 , will be significant if the predicted future OCF differs for firms using classification choices allowable under IFRS but not under U.S. GAAP. On one hand, we would expect the FLEX interaction coefficients to be positive because OCF using IFRS classification choices is higher than OCF using U.S. GAAP classifications. On the other hand, unlike the sales model, the independent variables are past cash flows and past accruals. Because past cash flows and past accruals capture Electronic copy available at: https://ssrn.com/abstract=2815123 the firm's classification choices in the prediction of future cash flows (using those same classification choices), these variables serve as controls for the classification choice also. In this case, the FLEX interaction coefficients will not be significant.

Sample selection and classification choices
3.1 Sample selection Because databases do not accurately report cash flow classification, we hand collect the detailed cash flow items from the financial statements. For those countries with 100 available firms or less, we select 100% of the firms. For those countries with over 100 available firms, we select the greater of 100 firms or 30% of the firms with available data. Because of the large number of firms 16 We select our sample based on data availability in 2008 to maximize coverage of firms with at least three years of data following the widespread mandatory adoption of IFRS in Europe starting in 2005. Our focus is on the post-2005 period because that is the time frame in which firms in our sample largely faced similar classification alternatives. Prior to 2005, some firms had already adopted IFRS or were using a home-country GAAP that permitted IFRS-allowable classifications. Cash flow reporting varied by country in the period before IFRS adoption. According to the Nobes (2001) report, the following countries' local GAAP had no specific rules requiring a cash flow statement: Austria, Belgium, Finland, Italy, and Spain. For Portugal, Nobes (2001) indicates there were no specific rules except for listed companies; our review of listed companies' pre-IFRS annual reports in Portugal indicates that the classifications for interest paid, interest received, and dividends received were financing, investing, and investing, respectively. The classification requirements were similar to IFRS in the UK (Davies et al. 1997) and in Germany (Leuz 2000). We were unable to document any local GAAP requirements for Denmark, the Netherlands, Norway, and Sweden, so we reviewed actual annual reports in the pre-IFRS period. In the annual reports reviewed, the classification used for all three items in those countries was operating. Nobes (2011) summarizes classification practices related to interest paid in five countries pre-IFRS as follows: Austria and Franceoperating; United Kingdomfinancing; and Germany and Spainoperating or financing.
in the United Kingdom, we selected 15% (or 146) of total potential firms to collect the cash flow data. When sampling from the available population of firms within a country, we utilize stratified sampling, first ranking within country by industry and size (total assets) and then selecting firms.
This  The choice of where to classify interest paid in the statement of cash flows varies by country ( Finally, for dividends received, 81% of firms in the extractive industries report dividends received 22 We follow the industry definitions in Barth et al. (1998). in operating, followed by durable manufacturers with 70% classifying dividends received in operating. Table 4 presents information on common classification-choice combinations for the 1,925 firm-year observations that clearly disclose classification choices for all three items. The most common classification-choice combination, selected by 42%, is classifying all items in OCF. The second most common combination is classifying interest paid in financing and both dividends received and interest received in investing.   In the left side of Table 8, panel A, we compare the 57 OCF-increasing changer sample to itself over time-before and after the reclassification for variables similar to those in the crosssectional regression. For variables created as averages over the sample period, averages are based on the periods before and after the reclassification. The significantly positive differences in the means and medians of the difference in OCF (reported minus pro forma) and interest paid reported in financing are a function of the criteria for inclusion as an OCF-increasing changer. In addition, we find that equity issues, and analysts' forecast coverage are higher in the period after the change than before. The mean and median profitability of changers is significantly lower after the change.
In the right side of Table 8, panel A, we compare the 57 OCF-increasing changer sample to the control sample. We find significant differences in the means and/or medians of the difference in OCF (reported minus pro forma), interest paid reported in financing, equity issues, analysts' forecast coverage, cross-listed in the US, and industry. We also examine the effect of including the following other variables but none are significant: credit risk, average market-to-book ratio, average returns, an indicator variable for earnings that are just positive, variability of OCF (computed as the standard deviation of the firm's OCF over the sample period), and capital intensity which captures structure of operations and potential financing needs.
When we include only observations with interest paid located on the face of or in the footnotes to the financial statements (about 70% of the sample), regression results are similar to the overall reported results.
We also reviewed the classification choices of a larger set of cross-listed firms to determine whether the results on the cross-listing variable are generalizable to a broader set of cross-listing firms. We collected data on 83 European Union cross-listed firms in 2006 (including some of the 40 cross-listed firms in our sample), and we find the classification choice for interest paid is similar to our overall sample: 78% reporting in operating and 22% in financing.

Market pricing of the persistence of cash flows
Results of the analysis comparing the persistence parameters for accruals and cash flow components of earnings are presented in Table 9. For both groups, accruals are significantly less persistent than operating cash flows (similar to findings in prior research (Sloan 1996;Dechow, Richardson, and Sloan 2008 The implications of the market-implied coefficients, however, differ for the two groups. The In contrast, the non-FLEX subsample's market-implied persistence of accruals (0.4020) is roughly equivalent (p = 0.6644) to the persistence parameter of accruals (0.4339) in the forecasting equation, while the market-implied persistence of cash flow (0.4039) is lower than the persistence parameter (0.6851), indicating underpricing only of the cash flow component. (Pincus et al. (2007) similarly find underweighting of OCF but not accruals in eight of the countries they study, five of which are European.) Further, the market-implied coefficient of accruals is also roughly equivalent to the market-implied coefficient of cash flow. In other words, unlike the FLEX sub-sample, the evidence does not reveal higher pricing for accruals relative to cash flow. Overall, these results could be interpreted to suggest that investors value accruals more highly than cash flowbut only for the FLEX subsample.  In the past cash flows and accruals models in Table 10, panel B, the FLEX interaction with OCF and accruals is not significant indicating that the classification choices do not contribute to the prediction of future OCF in this model. This finding is consistent with past OCF and accruals also controlling for the firm's classification choices. Further, this finding suggests that this type of model may be more useful in the international context in which flexibility in cash flow classification exists.

Additional analyses
Our market tests do not provide evidence consistent with the accruals anomaly. Pincus et al. (2007) provide evidence that the accrual anomaly occurs in common law countries rather than code law countries. Given that code law countries comprise 12 of the 13 countries in our sample, this finding is consistent. In the United Kingdom, the only common law country in our sample, we also find no evidence of the accruals anomaly. We explore whether the results of our market tests are sensitive to model specification. We find that results of our market pricing analysis in Table 9 are sensitive to model specification. 31 In particular, when the forecasting and valuation models exclude firm-specific control variables (Size t , BM t , and EP t ), the overall conclusions are similar for both subsamples. These conclusions, based on untabulated results excluding the control variables, are: accruals are significantly less persistent than operating cash flows as in the base analysis, the market-implied coefficients reflect underpricing of both accruals and operating cash flow as in the base analysis, but the comparative magnitude of the market-implied coefficients shows no indication of the accrual anomaly (i.e., the coefficient on accruals does not exceed the coefficient on operating cash flow) regardless of the firm's cash-flow classification choice.

Conclusion
Cash flow, and particularly OCF, is used in business valuation and contracting. However, OCF can be measured differently under IFRS and U.S. GAAP because of classification alternatives available only under IFRS. While previous international accounting research focuses on IFRS versus U.S. GAAP differences in earnings and shareholders' equity, little attention has been given to potential differences in OCF under the two sets of standards.
Using an international setting, we build on and extend certain findings from the U.S.-only setting in Lee (2012). We find that firms with a higher likelihood of financial distress, that issue more equity, with higher leverage, and that are less profitable are more likely to make OCFincreasing classification choices. Our findings further suggest that cross-listed firms tend to make classification choices consistent with U.S. GAAP. Firms are more likely to make OCF-increasing classification changes when they have issued equity and less likely to change when they have analysts following, more peers making similar choices, and are cross-listed in the U.S. Overall, OCF-enhancing classification choices are associated with both financial and informational factors.
The flexibility under IFRS also has consequences. We provide some evidence that the market's assessment of the persistence of OCF and accruals differs for groups of firms making different classification choices. However, results are sensitive to model specification. We also show that results of certain OCF prediction models differ for firms making different classification choices. When OCF prediction is based on past sales, results differ for firms making alternative classification choices. However, when OCF prediction is based on past OCF and accruals, results do not differ significantly for firms making alternative classification choices, likely because past OCF and accruals also control for firms' classification choices. Overall, the consequences of classification choices such as market reaction to OCF surprise/earnings surprise around earnings announcements offer an avenue for potentially fruitful future research.
Our paper contributes to the international accounting literature exploring the consequences of IFRS adoption and reporting. Given the recent adoption of IFRS in more than 120 countries and the consideration by U.S. regulators to adopt IFRS, our evidence on the classification of cash flows as operating, investing, and financing activities is important. Our results show that cash flow classification flexibility within IFRS creates a non-comparability that is absent under the more rigid classification requirements of U.S. GAAP. Flexibility in classification of cash flow items introduces potential non-comparability into measurement of widely-used metrics such as accruals and free cash flow. Understanding the impact of non-comparability under IFRS on such metrics will facilitate appropriate inferences from research incorporating these metrics.

APPENDIX EXAMPLE OF EFFECTS OF RECLASSIFICATION ON OPERATING CASH FLOWS
Norse Energy Corp. ASA, a Norwegian gas explorer and producer, changed its classifications of interest paid and interest received in 2007. It changed its classification of interest paid to financing from operating. It changed its classification of interest received to investing from operating. The net effect of these changes was to report positive, rather than negative operating cash flows, in both 2007 and 2008. The example below illustrates the computation of the net effect of the reclassifications.   *, **, *** denote statistical significance of difference between sample firms and non-selected firms. * p < 0.10, ** p < 0.05, *** p < 0.01.    OCF_Pro forma_USGAAP t = operating cash flows for year t adjusted to include interest paid, interest received, and dividends received in operating cash flows if these items are not already reported in the operating section. INV_Reported t = investing cash flows as reported by the firm for year t. INV_Pro forma_USGAAP t = investing cash flows for year t adjusted to exclude interest paid, interest received, and dividends received. FIN_Reported t = financing cash flows as reported by the firm for year t. FIN_Pro forma_USGAAP t = financing cash flows for year t adjusted to exclude interest paid, interest received, and dividends received. All firm subscripts are omitted. All variables are scaled by the firm's total assets.

Table 6
Descriptive statistics and regressions of the difference in operating cash flows and interest paid in financing on incentives and reporting environment

Table 6
Descriptive statistics and regressions of the difference in operating cash flows and interest paid in financing on incentives and reporting environment (continued) Variable Definitions: OCF_Reported t less OCF_ Pro forma_USGAAP t = the average by firm of operating cash flows as reported by the firm for year t less operating cash flows for year t adjusted to include interest paid, interest received, and dividends received in operating cash flows if these items are not already reported in the operating section. Interest Paid in Financing = 1 if the firm classifies interest paid in financing cash flows as of the last year reported, and zero otherwise. Distress_Hi = 1 if the firm's financial distress computed using Altman's Z-score is less than 1.81, indicative of high distress, and zero otherwise. Equity Issues = percent change in the firm's contributed capital over the sample period. Leverage_Hi = 1 if the firm's ratio of total liabilities over total assets at the beginning of the fiscal year, averaged over the sample period, is greater than the median, and zero otherwise. Profitability = the firm's net income divided by beginning total assets, averaged over the sample period. Analysts Cash Flow Forecast = 1 if at least one analyst's cash flow forecast for the period is available on IBES, and zero otherwise, averaged over the sample period. Industry Homogeneity = the percent of firms within an industry that report interest paid in financing cash flows, with industry classifications based on Barth et al. (1998). Cross-listed in US = 1 if the firm is cross-listed in the United States, and zero otherwise. Size = the natural logarithm of the firm's beginning of year market capitalization in U.S. dollars, averaged over the sample period. *, **, *** denote statistical significance of difference between firms that do not change classification and firms that change classification ("changer".) * p < 0.10, ** p < 0.05, *** p < 0.01. a Compares statistical significance of means and medians of pre-change and post-change variables. b Compares statistical significance of means and medians of pre-change and control samples. c Consists of 57 firms that made an OCF-increasing change and 109 firms that are not currently maximizing reported OCF but did not make a classification change. * p < 0.10, ** p < 0.05, *** p < 0.01. p-values are one-tailed for variables with directional hypotheses, and twotailed for all others. Standard errors are clustered by firm. Country controls and industry controls are included.
Variable Definitions: OCF-increasing classification change = 1 if a firm made an OCF-increasing classification firm, and 0 otherwise. See Table 6 for the remaining variable definitions.

Table 9
Simultaneous estimation of persistence parameters for accruals and operating cash flow and the parameters implied by stock returns, for subsamples making alternative classification choices EARNINGS t+1 = α 0 + α 1 ACC_Reported t + α 2 OCF_Reported t + Controls t + υ t Returns t+1 = β (EARNINGS t+1 -0 # -1 # ACC_Reported t + 2 # OCF_Reported t + Controls t ) + ε t ACCR_Reported t = the amount of accruals, calculated as net income less reported operating cash flows for year t divided by average total assets for year t. OCF_Reported t = the reported amount of operating cash flow for year t divided by average total assets for year t. Size t = the natural logarithm of the firm's market capitalization in U.S. dollars at the beginning of year t. BM t = the firm's book to market ratio, calculated as the ratio of the firm's shareholders' equity divided by its market capitalization at the beginning of year t. EP t = the firm's net income divided by its market capitalization at the beginning of year t. 0.5132 a The number of observations is based on the availability of accounting and market data to compute variables, including lagged variables, in the model. * p < 0.10, ** p < 0.05, *** p < 0.01. p-values are two-tailed. Errors are clustered by firm. Regressions include country, industry, and year controls.
Variable definitions: 1/TA t = 1 divided by average total assets for year t. S t /TA t = sales revenue for year t divided by average total assets. FLEX x S t /TA t is the interaction between the indicator variable FLEX and sales revenue for year t. ΔS t /TA t is change in annual sales revenue from year t-1 to year t divided by average total assets. FLEX x ΔS t /TA t is the interaction between the indicator variable FLEX and change in annual sales revenue for year t divided by average total assets. See Table 9 for remaining variable definitions.