The Effect of International Institutional Factors On Properties of Accounting Earnings

On Properties of Accounting Earnings

by

Ray Ball,* S. P. Kothari** and Ashok Robin***

*

William E. Simon Graduate School of Business Administration

University of Rochester Rochester NY 14627 ball@ssb.rochester.edu

**

Sloan School of Management, E52-325 Massachusetts Institute of Technology 50 Memorial Drive, Cambridge, MA 02142

kothari@MIT.edu

College of Business

Rochester Institute of Technology

Rochester NY 14623 ajrbbu@rit.edu

First Draft 19 September 1996 This Version: August 1999

We gratefully acknowledge the helpful comments of David Alexander, Eli Bartov, Sudipta Basu, Gary Biddle, Dan Collins, Peter Easton, Scott Keating, Christian Leuz, Gerhard Mueller, Christopher Nobes, Dieter Ordelheide, Peter Pope, Abbie Smith, Peter Taylor, Ross Watts, Greg Waymire, Steve Zeff, Jerry Zimmerman, the editor and referee, and seminar participants at: Carnegie-Mellon University, University of California at Berkeley, University of California at Los Angeles, CUNY Baruch College, EIASM Workshop in European Accounting in Krakow, 1999 Financial Accounting and Auditing Conference of the Institute of Chartered Accountants in England & Wales, Universität Frankfurt am Main, Hong Kong University of Science and Technology 1998 Summer Symposium, IAAER/CIERA 1998 Conference, Inquire Europe Autumn 1998 Seminar, University of Iowa, Harvard Business School 1999 Annual Workshop, KPMG-AAA International 1999 Accounting Conference, London Business School, London School of Economics, Massachusetts Institute of Technology, Melbourne Business School, Ohio State University, University of New South Wales, University of Technology in Sydney, and Washington University in St Louis. The paper has received a Vernon K. Zimmerman Award and an Inquire Europe Prize. Ball and Kothari received financial assistance from the John M. Olin Foundation and the Bradley Policy Research Center.

***

2

The Effect of International Institutional Factors

On Properties of Accounting Income

Abstract

International differences in the demand for accounting income predictably affect the way it incorporates economic income (dividend-adjusted change in market value) over time. We characterize the “shareholder” and “stakeholder” corporate governance models of common and code law countries respectively as resolving information asymmetry by public disclosure and private communication. Also, code law directly links accounting income to current payouts (to employees, managers, shareholders and governments). Consequently, code law accounting income is less timely, particularly in incorporating economic losses. Regulation, taxation and litigation cause variation among common law countries. The results have implications for security analysts, standard-setters, regulators, and corporate governance.

1. Introduction

We show that differences in the demand for accounting income in different institutional contexts

causes its properties to vary internationally. The properties of accounting income we study are timeliness and conservatism. Timeliness is defined as the extent to which current-period accounting income incorporates current-period economic income, our proxy for which is change in market value of stockholders’ equity. Conservatism is the extent to which current-period accounting income asymmetrically incorporates economic losses and gains.

The principal institutional variable we study is the extent of political influence on accounting. Our simplest proxy for political influence is a dichotomous classification of countries into code versus common law systems; that is, into whether accounting practices are determined primarily in the public sector or in the private sector, respectively. We argue that politicization of accounting standard setting and enforcement decreases the demand for timely and conservative accounting income, and increases the demand for an income variable with low volatility. In our sample, Australia, Canada, UK and USA are classified as common-law countries and France, Germany and Japan are classified as code-law. Seeking finer proxies, we then characterize the United Kingdom, among common-law countries, as exhibiting the least political influence on accounting, due to several institutional factors. Similarly, we study the effects of differential litigation rules, issuance of predominantly private debt, and laws (including tax laws) that cause accounting income to conform to current-period dividend policy.

In code-law countries, the comparatively strong political influence on accounting occurs at national and firm levels. Governments establish and enforce national accounting standards, typically with representation from major political groups such as labor unions, banks and business associations. At the firm level, politicization typically leads to a \"stakeholder\" governance model, involving agents for major groups contracting with the firm. Current-period accounting income then tends to be viewed as the pie to be divided among groups, as dividends to shareholders, taxes to governments, and bonuses to managers and perhaps also employees. Compared to common-law countries, the demand for accounting income under code law is influenced more by the payout preferences of agents for labor, capital and government, and less by the demand for public disclosure. Conversely, because agents for these groups are represented in corporate governance, insider communication solves the information asymmetry between managers and stakeholders. We hypothesize that their preferences penalize volatility in payouts and thus in

2

income. Managers respond by varying the application of accounting standards or by influencing operating, financing and investment decisions (for example, by deferring discretionary expenditures such as R&D in bad earnings years).

Under the “shareholder” governance model that is typical of common-law countries, shareholders alone elect members of the governing board, payouts are less closely linked to current-period accounting income, and public disclosure is a more likely solution for the information asymmetry problem. In comparison with the more political process in code law countries, the desirable properties of accounting income in common law countries are determined primarily in the disclosure market. We hypothesize those properties include timeliness in incorporating negative economic income (i.e., asymmetric conservatism).

We study over 40,000 annual firm’s incomes reported during 1985-95, under the accounting rules of seven countries. We show that code-law income is substantially less timely and less conservative than common-law income. Code-law income does not even exhibit more timeliness than dividends. Among common-law countries, there is less asymmetric conservatism in accounting income reported in the United Kingdom, a country we characterize in terms of lower political involvement in accounting, lower litigation costs and less issuance of public debt. In addition to a detailed analysis of seven countries, we also study properties of accounting income in an additional 18-country samples. The results are consistent with our general thesis, that properties of accounting income around the world are a function of the varying demands that reported income satisfy under different institutional arrangements.

Our research design addresses the incorporation of economic income into accounting income over

time, under different international institutions. This has several advantages over simply studying the international accounting standards. First, much accounting practice is not determined by rules, for reasons that include: practice is more detailed than rules; rules lag innovations in practice; and companies do not invariably follow the rules. The extent to which accounting practice is determined by formal standards varies internationally, and the incentive to follow accounting standards depends on penalties under different enforcement institutions, so studying accounting rules per se is incomplete and potentially misleading in an international context. Second, reported income is influenced by managers’ operating, financing and investment decisions, as well as by accounting rules and practices. For example, managers can reduce the volatility of accounting income by deferring discretionary expenditures (such as R&D) in bad years. Because the use of accounting income in corporate governance varies internationally, we expect

3

managers’ operating, financing and investment decisions to affect accounting income differentially across countries, and report evidence consistent with that expectation. For both these reasons, we study international variation in properties of the actually reported income numbers, as evidenced by the way they incorporate economic income over time.

The validity of this research design depends on two measures. First, the research design studies the flow of market-valued economic income into book-valued accounting income. It therefore uses the fiscal-year change in market value of equity (adjusted for dividends and capital transactions) as its proxy for economic income. The major concern is that the accuracy of this proxy is correlated with the institutional independent variables in the study, and in particular that code-law countries have endogenously lower market liquidity and public disclosure standards. Second, the research design requires us to infer independent variables, such as the degree of political versus market determination of reported income, from our characterization of salient institutional facts. While our characterization is based on surveying a wide range of sources, it undoubtedly is subject to error. In the concluding section, we argue that both types of measurement error create a bias against our hypotheses.

We contribute to a growing literature on the effects of international accounting differences, including Jacobson and Aaker (1993), Alford et al. (1993), Amir et al. (1993), Bandyopadhyay et al. (1994), Harris et al. (1994), Joos and Lang (1994), Barth and Clinch (1996), and Pope and Walker (1999). We also contribute to the literature on international corporate governance, including Baums et al. (1994) and La Porta et al. (1997).

The following section outlines the model used to test the timeliness and conservatism of accounting income. The third section describes the data. Section four surveys the salient institutional facts used to develop and then test hypotheses on properties of income internationally. Section five extends these tests to a comparison of income with dividends and cash flows. The sixth section reports specification tests and the concluding section discusses the research design and the implications of the results.

A Model of Incorporation of Economic Income in Accounting Income

4

The research design infers timeliness and conservatism from the way firms’ accounting incomes incorporate their economic incomes over time. Accounting income therefore is specified as the dependent variable. Firms’ economic incomes are measured as fiscal-year changes in market values of equity, adjusted for dividends and capital contributions

“Clean surplus” accounting (Ohlson 1988) implies two relevant identities for all firms: accounting income equals fiscal-year change in book value of equity, adjusted for dividends and capital contributions; and the sums of the firm’s lifetime accounting income and its lifetime economic income are identical. We investigate the temporal process by which economic income is incorporated in accounting income, i.e., the accounting model of income determination, and how it is affected by international institutional factors. Our research design allows for three fundamental features of the accounting model of income determination: accounting “recognition” principles that influence the timeliness of accounting income; the effectiveness of accounting accruals in ameliorating serial correlation in operating cash flows; and accounting conservatism. The most fundamental feature of accounting determining the incorporation of economic income in accounting income over time is the set of accounting “recognition” principles (FASB 1985, paras. 78-89), including the Revenue Realization and Expense Matching principles. Whereas economic income incorporates changes in expectations of the present values of future cash flows, to a first approximation (which we qualify below) the recognition principles incorporate such changes in accounting income only when the actual cash flow realizations occur. Hence, accounting income systematically lags economic income (Ball and Brown 1968). The lags extend over multiple periods (Beaver, Lambert and Morse 1980, Easton, Harris and Ohlson 1992, Kothari and Sloan 1992), because information that causes correlated changes in expectations of multiple future periods’ cash flows affects economic income immediately, but to a first approximation its effect is incorporated in accounting income in all periods in which the revised cash flows are realized. The recognition principles therefore cause economic income to be incorporated into accounting income in a lagged and “smoothed” fashion over time.

If timely information incorporation was the sole property demanded of accounting income then, for listed firms, economic income (change in market value of equity after dividend and capital contributions) would be the optimal accounting earnings variable. Because there is demand for accounting income with properties other than timeliness, accountants incorporate only a subset of the value-relevant information

5

available at the time. Notably, information asymmetry creates a demand for accounting income with the property of observability independently of managers. Accounting income thus incorporates only the subset of available value-relevant information that is independently observable, whereas economic income incorporates information that is not independent of managers, such as plans and forecasts.

The second fundamental feature of the accounting model of income determination is the effectiveness of accounting accruals in ameliorating serial correlation in operating cash flows. While the accounting model recognizes income to a first approximation at the time of cash flow realizations, it also provides for some anticipation of future cash flows via the rules of accrual accounting. For example, if managers pay an account for inventory early, then there is a decrease in current-period operating cash flow and, ceteris paribus, an offsetting increase in the subsequent-period operating cash flow. Accrual accounting rules attempt to insulate income from the effect of the prepayment, by expensing an amount in both periods that is based on inventory usage, not payments. In general, short term variation in firms’ operational financing and investment decisions (such as changes in inventories, accounts payable and accounts receivable) causes negative serial correlation in operating cash flows, which accrual accounting attempts to remove from accounting income (Dechow 1994). Hence, operating cash flow can be viewed as a noisy version of accounting income. Because accounting is costly, accruals are imperfect and accounting income does not remove all the noise in operating cash flows. Provided the residual cash flow effects that are not removed from income by accounting accruals contain no value-relevant information, this residual noise is independent of economic income. Noise also reduces the timeliness of accounting income, because accruals imperfectly incorporate the information contained in current and past cash flows about future cash flows.

The first two features of the accounting model of income determination together imply: Yit = fj(∆Vit, ∆Vit -1, ∆Vit-2, ∆Vit-3, … , νit )

(1)

where Y and ∆V respectively denote accounting and economic income (change in book and market values of equity, both adjusted for dividends and capital contributions), and ν denotes noise due to imperfect accounting accruals. We hypothesize that the accounting model is applied differently across countries, and assume the parameters of the model hold for all firms i that report under the accounting systems of country j. Assuming that ∆V is independent over time, this simplifies to:

Yit = gj(∆Vit, ηit)

(2)

6

The disturbance ηit incorporates lagged changes in market values (∆Vit -1, ∆Vit-2, ∆Vit-3, … ) as well as noise due to the residual serial correlation in cash flows not removed by accounting accruals (νit). Both components of the disturbance term affect the timeliness property of accounting income, which we measure as the R2 of regression (2), and its variants.

The independent variable (annual change in market value of equity, adjusted for dividends and capital contributions) is relatively free of short-term microstructure or mispricing effects. Scaled by opening market value Vit-1, the variables are annual rate of return (Rit ≡ ∆Vit/Vit-1) and earnings yield (NIit ≡ Yit/Vit-1

), and a linear specification gives:

NIit = α0j + α1jRit + ξit

(3)

The third fundamental feature of the accounting income model we study is conservatism. Income

conservatism is illustrated by new information about future cash flows from long-term assets. The recognition principles normally incorporate such changes in accounting income only when the actual cash flow realizations occur. However, a variety of accounting rules and practices cause immediate write-offs against income when expected future cash flows decrease, without waiting for the cash flow decreases to be realized. Longstanding write-off practices recently were formalized as asset impairment rules by SFAS 121 in the US. Upward revaluation is comparatively rare: it has not been practiced in the US since the Securities and Exchange Commission (SEC) was established in 1934, though it is practiced to a limited degree in some countries. Consequently, unrealized increases in asset values generally do not flow into income until approximately when the underlying cash flow increases occur, but unrealized decreases are more likely to be incorporated. Another example of this asymmetry is the “lower of cost or market” inventory rule, which incorporates inventory losses more quickly in income than gains.

We incorporate conservative asymmetry in accounting income timeliness by modifying (3) for

differential incorporation of negative and positive economic income (Basu, 1997):

NIit = β0j + β1j RDit + β2j Rit + β3j RitRDit + εit

(4)

The dummy RDt equals one if return Rt is negative and zero otherwise. β2j and (β2j+β3j) capture the incorporation of positive and negative economic income into current-year accounting income, under the accounting rules and practices of country j. The piecewise linear specification allows us to study international differences in the asymmetric timeliness, or conservatism, of accounting income. Initially, we estimate separate individual-country relations for each country j, pooling all firms i reporting under the

7

country’s accounting standards and all years t. International differences in income timeliness, for positive and negative economic income combined, are reflected in the R2s of individual-country regressions (4). 3. Data

Accounting income, cash flow and dividends over 1985-95 are from the Global Vantage Industrial/Commercial (IC) file. Accounting income NIt is net income before extraordinary items (IC data 32).1 Dividends (DIV) is dividends paid (IC data 36). Operating cashflow (OCF) is net income before extraordinary items (IC data 32) plus depreciation (IC data 11), minus the change in non-cash current assets (IC data 75 minus data 60), plus the change in current liabilities other than the current portion of long-term debt (IC data 104 minus data 94). These variables are scaled by market value of equity, calculated from the Global Vantage Issue file as price times number of outstanding shares, adjusted for stock splits and dividends using the Global Vantage adjustment factor. Change in accounting income ∆NIt is NIt - NIt-1. Stock return R is the holding-period return, including dividends, over the firm’s fiscal year (the same period as income). Each firm/year observation is assigned to a country based on Periodic Descriptor Array 13 on the IC file, indicating the accounting standards used in preparing the firm’s financial statements that year. This normally is the country of the firm’s home exchange.2

The two extreme percentiles of each variable (NI, ∆NI, DIV, OCF and R) are excluded. Next, each firm/year with a missing value for any variable is excluded, giving the same observation set for the various variables and models estimated. Finally, countries with less than 1000 firm/year observations over the thirteen years are excluded, leaving 40,359 firm/year observations in eleven years from seven countries: Australia, Canada, UK and USA (common-law countries) and France, Germany and Japan (code law countries). We also report abbreviated results for a secondary sample of 18 countries with at least 100 firm-year observations.

Table 1 contains sample descriptive statistics. Individual-country samples are pooled firm-years and range in size from 1054 (France) to 21225 (US). In Panel A, mean fiscal-year returns range from 1

German companies do not deduct minority interest from consolidated net income, but Global Vantage alters their numbers to comply with US practice. We therefore define German Y as IC data 32 + data 27, which reconciles to the numbers actually reported. The adjustment has only a trivial effect on the results. 2

In correspondence, Christian Leuz and Dieter Ordelheide have noted that some French German firms recently have issued consolidated financial statements – not individual-company statements – prepared under International Accounting Standards. This practice appears to largely post-date our sample. To the extent that Global Vantage classifies theses observations as “German,” the errors create a potential bias against our hypotheses.

8

3.7% (Japan) to 17.3% (Australia). Mean NI ranges from 1.7% (Japan) to 6.6% (UK). All countries’ mean and median OCF (from which capital investments are not deducted) exceed NI (from which depreciation, a weighted average of past capital investments, is deducted).

Consistent with asymmetric conservatism, accounting income is negatively skewed (all medians exceed means), in contrast with positively-skewed stock returns (our proxy for economic income). Conservative accounting tends to incorporate economic losses as transitory, capitalized amounts, but to incorporate economic gains as persistent flows over time. Also consistent with our model, in all seven countries the rank order of volatilities is (highest first): R, OCF, NI and DIV. Our interpretation is: (i) accounting income is a lagged function of present and past economic income, much like a moving (“smoothed”) average, and hence has lower volatility than returns; (ii) cash flow from operations is a noisy income, and hence is more volatile; and (iii) dividends is a lagged function of accounting income, and hence is the least volatile.

[Table 1 here]

Panel B reports the distribution of ending months for companies’ fiscal years, which varies internationally. December is the norm in Canada, France, Germany and US. March is the norm in Japan and June is the norm in Australia. Code law countries exhibit more conformity with their norm, presumably due to the greater influence of tax accounting in those countries, discussed below. UK, which we categorize below as the least regulated country in the sample, exhibits the greatest dispersion in fiscal year-ends. Panel C reports the sample distribution by calendar year.

4. Hypotheses and Tests: International Timeliness and Conservatism of Accounting Income

We develop hypotheses concerning the influence of institutional variables on the two properties of accounting income captured by our model: timeliness and conservatism (or asymmetric timeliness). There have been numerous attempts to classify nations’ accounting systems, based on a variety of institutional variables, but little empirical research has been directed at determining which variables explain differences in important properties of accounting income.3 3

The American Accounting Association classified accounting systems using eight variables: political system (traditional oligarchy, totalitarian oligarchy, modernizing oligarchy, tutelary democracy, political democracy), economic system, stage of political development, financial reporting objectives, source of accounting standards, education and licensing of accountants, mechanism to enforce standards, and accounting client (public or private). Nobes (1992) and Nobes and Parker (1995, ch.4) survey classification schemes. How these variables affect important properties of

9

4.1 Demand for Timely Accounting Income in Code versus Common Law Corporate Governance. Perhaps the most fundamental institutional variable causing accounting income to differ internationally is the extent of political influence on both standard setting and enforcement. An admittedly imperfect proxy for political influence – but a proxy our results legitimate – is whether standard setting and enforcement occur under codified law (a governmental process) or common law (a market process).4 We hypothesize that the demand for timely incorporation of economic income in accounting income is lower under the code-law “stakeholder” model of corporate governance than in the common-law “shareholder” model.

Common law arises from individual action in the private sector. It emphasizes following legal procedure over rules (David and Brierley 1985, p.24; Posner 1996). Common laws – including accounting standards – evolve by becoming commonly accepted in practice. While it might be efficient for private-sector bodies to codify generally-accepted accounting rules and make them binding on their members, such standards arise in an accounting market, not in government. Common law enforcement is a private matter, involving civil litigation. Common law originated in England and now is found in UK and many former British colonies. The common-law countries in our sample are Australia, Canada, UK and US.

Common law has evolved to meet the demands of contracting in markets. The “shareholder” model of corporate governance, in which shareholders alone elect the governing board, predominates in common-law countries. Alchian and Demsetz (1972) argue this is efficient due to the additional incentive of residual claimants to effectively monitor managers. Compared to code-law governance, board members are less likely to hold large blocks, there is more monitoring of managers by external debt and equity markets (including analysts), and lenders and employees seldom have board representation. We hypothesize that, because parties contracting with the firm operate at greater “arm’s length” from managers, information asymmetry in common-law countries is more likely to be resolved by timely public disclosure.

accounting information internationally is largely untested. International texts typically list variables to justify classifications, without correlating them with the national accounting standards listed in subsequent chapters, let alone with properties of the financial statements prepared under those standards. 4

David and Brierley (1985) provide a survey.

10

Code-law originates from collective planning in the public sector. Code-law accounting standards are established by governments or quasi-governmental bodies, such as France’s Conseil National de la Comptabilité or Japan’s Business Accounting Deliberation Council (which advises the Ministry of Finance). 5 The code prescribes regulations ranging from abstract principles (e.g., “prudence”) to detailed procedures (e.g., the format of financial statements). Code-law enforcement is a governmental function, involving administrative bodies undertaking criminal prosecution for code violation. Code-law countries in our sample are France, Germany and Japan.

A fundamental difference between common-law and code-law countries is the manner of resolving information asymmetry between managers and potential users of accounting income, including debt and equity investors, employees, suppliers and customers. Code-law corporate governance tends to be conducted by elected or appointed agents for these parties.6 These agents tend to be informed by private “inside” access to information. Thus, employees and stockholders each elect 50% of the supervisory board of German Aktiengesellschaft (stock corporations). Stock voting typically is dominated by banks, due to their large-block holdings and to the German practice of banks voting individuals’ stocks as agents (Köndgen 1994). The supervisory board appoints and monitors the managerial board, and approves the financial statements. For the system to be tractable, the number of contracting parties must be small, so managers have close relations with intermediaries: banks and other financial institutions, labor unions, governments, major customers and suppliers. There is no presumption that parties operate at a distance. Consequently, the demand for public disclosure in code-law countries is not as great as in common-law countries.7 We propose that this reduces the demand for accounting timeliness.

Conversely, we propose that code-law accounting income meets other demands. The stakeholder

model causes accounting income to be viewed as a common “pie” divided among stakeholders, as 5

Romans implanted code-law (also referred to as civil law, La Porta et al., 1997) in most continental European countries. The most detailed accounting code is the French Plan comptable général, adapted under German occupation from Goering’s 1937 standardization of German accounting for wartime planning (Standish 1997, p.60). Japan operates a code-law system, having adopted elements of the German legal system and the French accounting system during the Meiji Era (1868-1910). Scandinavian law is another code-law category with origins in the Roman law (David and Brierley, 1985, and La Porta et al. 1997). 6

See Nobes and Parker (1995, ch.12), Roe (1994) and Miwa (1996). 7

La Porta et al. (1997) predict that, in addition to differences in legal systems, differences in the extent of law enforcement affects the development of capital markets across countries. It is possible that variation in law enforcement has predictive power with respect to the demand and supply of accounting information that we ignore. This likely imparts a bias against our hypotheses.

11

dividends to shareholders, taxes to governments, and bonuses to managers and perhaps also to employees. The portfolio weights of managers and employees typically are skewed toward their employer firms, so their incentive is to reduce volatility in payouts. Regulation of bank leverage ratios penalizes volatility in bank income and thus in the accounting income and/or dividends on their equity investments. Code-law banking systems typically are hierarchical, so bank representatives are well aware of governments’ incentives to reduce volatility of tax receipts. Employee representatives typically are re-elected annually. While incentives to reduce volatility in accounting income exist in common-law countries (e.g., Healy 1985), we hypothesize that code-law governance features amplify them.

Volatility reduction is achieved, at the expense of timeliness, by accounting methods that “smooth” accounting income over time, incorporating economic income gradually over several periods. Code-law accounting gives managers considerably more latitude in timing income recognition. In good years, it is code-law practice to decrease income by asset write-downs (e.g., excessive allowances for bad debts), by provisions (e.g., excessive provisions for future losses or future expenses) and by transfers to reserves. In bad years, reversing these accounting adjustments increases income.8 We predict that code-law accounting income incorporates a lower proportion of current-period economic income, “smoothing” its incorporation to a greater extent over time.

H1: Code-law countries’ accounting incomes are more “smoothed” and less timely in incorporating current-period changes in market value than common-law countries’. Country regressions with observations pooled across time. Initially, we test hypothesis 1 by

2

comparing the adjusted Rs of the individual-country Basu regressions (4), each estimated from pooled

time series and cross-section data. Panel A of Table 2 reports the results and Figure 1 graphs them. With

2the possible exception of France, there is a clear difference between individual code-law country Rs

(ranging from 4.2% to 12.6%) and common-law country R2s (from 9.1% to 17.0%), consistent with the hypothesis. When countries are grouped, the R2 for the pooled code-law sample is 5.2%, approximately one-third the common-law equivalent of 14.4% (excluding UK). The international differences in the degree of income-return association we document are similar to those in Alford et al. (1993), despite slightly different time periods and models. 8

Daimler-Benz reported 1993 German-rule income of DM168 million, but subsequent US GAAP disclosure revealed that a loss of DM969 million had been hidden by various accounting adjustments (Ball 1998).

[Table 2 and figure 1 here]

12

2

To assess whether the difference in the Rs is statistically significant, we estimate the standard 2

deviation of estimated Rs, which Cramer (1987) shows is a function of sample size, the number of 2independent variables and the true R. For four independent variables including intercepts, a true R2 of

5%, and sample size 1,000 (5,000, 20,000), the standard deviation of the estimate is 1.3% (0.6%, 0.3%). For a true R2 of 15%, it is 2.0% (0.8%, 0.5%). Assuming independence across countries, differences in the order of 5% between our sample countries are significant, and the pooled common-law sample R2 of 14.4% significantly exceeds the code-law R2 of 5.2%. Since independence likely is violated, we report alternative tests below.

2Annual cross-sectional regressions. An alternative test estimates Rs from separate annual cross-

sectional regressions for each of three pooled country groups (code-law, common-law and UK), excluding the 6 of 77 country-years with fewer than 20 firm observations. For each group, we calculate from the annual regression R2s a sample mean and its t-statistic. The mean R2 for the common-law group is 16.7% (t-statistic = 12.85) compared to 7.0% (t-statistic = 6.07) for the code-law group and 16.8% (t-statistic = 8.81) for the UK. The common-law countries’ average R2 is significantly greater than that for the code-law countries at the 0.01 level. The t-statistics assume independence across time, and likely are conservative because the dependent variable, earnings yield, is mildly negatively serially correlated. The

2sampling error of the coefficient estimates and the Rs therefore could also be negatively correlated,

biasing the standard deviation of the sample R2 upward and the t-statistics downward.

Secondary sample of eighteen countries. A secondary sample of eighteen countries with between 250 and 1000 firm-year observations provides similar results. The Mueller et al. (1997, pp. 11-12, exhibits 1-4) classifications of countries as following a British-American or a Continental accounting model are used to proxy for our private-sector common-law and public-sector code law categories, respectively. We thereby classify eight additional countries as common-law (Hong Kong, India, Ireland, Malaysia, Netherlands, New Zealand, Singapore, and South Africa) and nine as code-law (Austria, Belgium, Denmark, Finland, Italy, Norway, Spain, Sweden, and Switzerland), adding Thailand as a code-law country based on our own assessment. While the Mueller et al. classifications agree with ours for each of the seven countries in our primary sample, we are less confident of the eighteen additional country classifications. Specifically, we suspect that many countries following British-American accounting rules

13

nevertheless lack common-law litigation enforcement and exhibit reduced demand for timely accounting income. Classification errors are expected to reduce the significance of the subsidiary sample results.

For each of the eight common-law and ten code-law countries, we estimate a pooled Basu regression model (4). Annual cross-sectional regressions for each country are not feasible because there

2

are too few observations in many years for most of the countries. The average Rfor the code-law

countries is 7.2%, the median is 7.0%, the minimum is 1%, and the maximum is 14.6%. In comparison, the average R2 for the common-law countries is 15.5%, the median is 12.4%, the minimum is 9.8%, and the maximum is 22.5%. The difference in mean R2s of the common and code-law countries is statistically significant. The results are consistent with those reported for the seven countries in table 2 and provide an independent confirmation of H1.

4.2 Universal Demand for Conservative Accounting Income. The second hypothesis is that conservatism, defined as asymmetric timeliness in incorporating economic gains and losses (Basu 1997), is a general property of accounting income.9 We propose that accountants contract to supply users with asymmetrically conservative income due to three properties of the market for accounting information. First, managers possess specific information (Alchian 1984) that is costly for external users to produce themselves (is not independently “verifiable”). Because managers have asymmetric incentives to disclose positive and negative specific information, uninformed accountants find information causing downward revisions of expected future cash flows (economic losses) more credible than upward revisions, and are more likely to incorporate it in accounting income. Second, lenders are important users of accounting information, including income and book value (a function of income), and they are asymmetrically affected by economic gains and losses (Watts and Zimmerman, 1986). Third, we propose that timely disclosure of economic losses is an important corporate governance mechanism. If reversing bad investment decisions and strategies is personally costly to managers, then informed monitoring by boards, analysts, investors, and lenders is a mechanism to force them to stem the losses. For these reasons, accountants supply income and book values that incorporate economic losses in a more timely fashion than economic gains. 9

We describe asymmetric timeliness in incorporating economic gains and losses as “income conservatism.” The concept is related to, but different from, balance sheet conservatism (reporting low book value of equity by under-stating assets and/or over-stating liabilities). Income conservatism implies balance sheet conservatism, but not vice versa: while code-law companies typically report conservative book values, they also boost income in bad years. This

14

Both of the above properties of the market for accounting information are universal (though we argue below that they vary in degree internationally). Consistent with this view, Basu (1997) reports that conservatism was practiced in medieval Europe and the lower-of-cost-or-market rule appeared by 1675, predating taxation and regulation accounting information. We therefore predict that accounting income is asymmetrically conservative in all countries.

H2: Accounting income is asymmetrically conservative in all countries, incorporating economic losses in a more timely fashion than economic gains.

This hypothesis predicts that accounting income exhibits higher R2s for negative fiscal-year return observations than for positive. Panel B of Table 2 reports that negative-return R2s exceed their positive-return counterparts in all common-law countries and in Germany.10 The exceptions are France and Japan, discussed below.

In all seven countries the Panel B coefficient on negative returns exceeds its counterpart on positive returns. Accounting income in the US on average is approximately ten times as sensitive to negative as to positive returns (estimated slopes of 0.32 and 0.03). The median country in terms of relative sensitivity to negative versus positive returns is the UK, with a ratio of approximately five (0.19: 0.04). A formal test of the asymmetry is provided in Panel A, in which the incremental slope β3 on negative fiscal-year returns (i.e., return interacted with the return dummy variable, R*RD) is significantly positive for all seven countries. As an alternative test, when the β3 coefficients are estimated from separate annual cross-sectional country regressions, the average incremental slope is significant for all seven countries except Germany.

[Figure 2 here]

Results for the secondary sample of 10 code-law and 8 common-law countries also are consistent

with asymmetric timeliness being a universal property of accounting income. The incremental slope on negative returns is statistically significant at the 10% level in 13 of the 18 countries and the point estimate is (insignificantly) negative in only one case.

reduces the asymmetric timeliness of accounting income and is difficult to describe as “conservative” in its effect on

income. Asymmetric timeliness is different from, but related to, Gray’s (1980) concept of conservatism. 10

For the separate positive and negative return regressions, we examined plots of the residuals against returns, and found no evidence of non-linearity (which was a concern for the negative-return sample in particular).

15

4.3 Demand for Greater Income Conservatism under Common-law Governance. We propose that common-law accounting income is more asymmetrically conservative than code-law, due to greater demand for timely disclosure of economic losses. Each of the three properties of the market for accounting information described above is qualitatively weaker in code-law governance. First, the closer code-law relation between managers and agents for contracting groups (banks, labor unions) reduces information asymmetry. Second, banks and other financial institutions tend to supply both debt and equity capital, making their loss function more symmetric. Third, there is less reliance on external monitoring of managers. Further, the demand for timely incorporation of economic losses is reduced by low expected litigation costs and the more direct relation between current-year accounting income and short-term payouts. For example, reluctance to cut dividends (Lintner, 1956) leads to a reluctance to incorporate economic losses in income as capitalized, transitory amounts, and a preference for gradual loss recognition as cash flow realizations occur.

H3: Common-law accounting income is more asymmetrically conservative than code-law.

The separate positive and negative return samples in Panel B of Table 2 reveal that the greater overall timeliness of common-law accounting income is due entirely to periods of economic losses. For

2

the common-law group, the 12.2% negative-return year R significantly exceeds the 0.4% in positive-

return years. In all common-law countries, R2s in negative-return years exceed those in positive-return years. Accounting income in Australian, Canadian and US over the sample period essentially ignores current-year economic gains. Common-law accounting income seems directed primarily toward incorporation of economic losses. With the exception of Germany, the same cannot be said of code-law accounting. For the code-law sample, the negative-return year R2 of 1.7% actually is lower than 3.3% for the positive-return years. These results are consistent with the hypothesis that, relative to code-law countries, the common-law demand for accounting income originates more in arm’s length corporate governance, debt contracts, and investor litigation, and less in determining short-term payouts.

A problem with cross-sectional data arises if the true regression slopes are heterogeneous. For

example, accounting income for start-up firms likely has a lower slope on economic income than for mature firms. The OLS cross-sectional R2 then is a downward biased estimator of the sample mean R2. We expect this constitutes a bias against H2 because the lower market liquidity and disclosure standards of code-law countries mitigates against listing of less mature firms and creates greater homogeneity.

16

However, the cross-sectional slope coefficient is an unbiased if inefficient estimator of the sample mean coefficient. Consequently, we also rely on the slopes in Panel A of Table 2 (graphed in Figure 2), which

2

reveal a similar pattern to the Rs. The β2j slopes show that accounting income in every common-law

country exhibits less sensitivity to positive returns than in France and Germany (accounting income in Japan exhibits the least timeliness for both return samples). In contrast, the incremental negative-return slopes β3j for common-law countries range from 0.15 to 0.40, considerably larger than the code-law range of 0.01-0.10. The pooled incremental slopes are 0.31 for the common-law group as a whole (excluding UK) and 0.01 for the code-law group. The difference is statistically significant at the 0.01 level.

Pooled Sample. We provide a formal test of international differences in the degree of conservatism by estimating (4) using data pooled across countries:

NI = β0 + Σj β0j CDj + β1 RD + Σj β1j RD.CDj + β2 R + Σj β2j R.CDj + β3 R.RD + Σj β3j R.RD.CDj (5)

Six dummy variables identify the country of the accounting standards used for each firm/year, with CDj = 1 for firm/years under country j and = 0 otherwise. US is the “base country,” with zero values for all country dummy variables. The coefficient β2 on return measures the incorporation of current economic income into US firms’ accounting incomes, and the coefficient β2j on the product of return and the country j dummy variable measures the country’s incremental incorporation, relative to the US. The coefficients β3 and β3j measure the asymmetric conservatism of accounting income under US standards and the incremental conservatism under other countries’ standards. Thus, β2+β2j+β3+β3j measures the incorporation of negative economic income in country j. Results are reported in column (1) of Table 3 and are consistent with those reported above. Accounting income in all sample code-law countries (France, Germany and Japan) exhibits significantly less incremental sensitivity to negative economic income than under US standards. Other columns of Table 3 report specification tests, discussed below.

[Table 3 here]

4.4 Regulation, Litigation and Debt Differences among Common-law Countries. The distinction between common-law and code-law countries provides useful insights, but the categories are by no means homogeneous. We consider two important institutional differences within the class of common-law countries: the method and extent of their regulation of accounting; and the extent to which their securities litigation rules favor plaintiffs.

17

Regulation. We propose that, among common-law countries, income conservatism increases with regulation of accounting standard-setting and enforcement. Building on Peltzman (1976), Watts and Zimmerman (1986, pp. 229-231) argue that the political process, and the SEC as agent, has an incentive to avoid voter perception of responsibility for investor losses. We argue that responsibility is attributed more to managers, and less to the political process, if losses are disclosed in the financial statements in a more timely fashion. Consequently, regulation adds criminal penalties (fines, incarceration) to the common-law civil remedy of damages for untimely disclosure of materially bad news. Accounting is regulated to some degree in all common-law countries, but the extent varies. For reasons summarized below, we characterize UK as having the least regulated accounting market among our sample common-law countries.

The Securities Exchange Act 1934 created the SEC as a US Government agency authorized to mandate and administer accounting standards. The SEC by and large has accepted the standards of the accounting profession, but nevertheless has retained a close supervisory role. It has intervened in standard-setting on several occasions, as has Congress on occasions. Rule 10b-5 of the 1934 Act states: “It shall be unlawful … to make any untrue statement of a material fact or to omit to state a material fact.” Section 32 imposes criminal penalties for willful violation of the Act, and Rule 10b-5 has been used as a basis for civil litigation (Coopers & Lybrand 1997, p.17). We conclude that the US operates a closely-regulated common-law system, with compounded criminal and civil penalties for non-disclosure of bad news.

Australia and Canada are widely viewed as having evolved from the loosely regulated UK model (described below) to the US regulatory model. Both commenced with a largely self-regulating profession, initially adopted provincial (rather than federal) regulation, and then moved to a system in which governmental or semi-governmental bodies set national accounting standards. Australia created a regulatory body (now constituted as the Australian Securities Commission) “obviously modeled on the SEC” (Nobes and Parker, 1992, p.106). It also removed standard-setting from the private sector, assigning it in 1984 to a government-appointed authority (now the Australian Accounting Standards Board), whose accounting standards can be overturned only by the Australian Parliament. These changes predate our sample period, so we characterize Australian accounting as highly regulated (see Choi and Mueller, 1992, p. 86).

18

Canada has evolved to a regulated federal model in a similar fashion. Accounting was increasingly regulated with the establishment of the provincial securities commissions, and the enactment of provincial and federal securities laws. Various revisions of the Ontario Companies Act have been particularly influential. In 1975, federal regulations required financial statements to comply with the accounting standards promulgated in the Canadian Institute of Chartered Accountants’ CICA Handbook, thereby giving them statutory status. The SEC has indirectly influenced Canadian accounting, due to large Canadian corporations listing in New York.

Among the common-law countries in our sample, we classify the UK as having the least regulated accounting market over our sample period. There is no UK regulatory body comparable to the SEC in the US, financial markets (the City of London) historically have been viewed as primarily “self-regulating,” and the UK parliament seldom has intervened in accounting matters (Choi and Mueller, 1992, ch. 3, Nobes and Parker, 1992, ch. 6, and Radebaugh and Gray, 1997, ch. 5). In 1990, UK established the Accounting Standards Board (ASB), modeled on the US FASB and accountable to a newly-created Financial Reporting Council. The ASB was authorized to issue Financial Reporting Standards (FRSs) that are backed by law (Radebaugh and Gray 1997, pp.91-92; Choi and Mueller 1992, pp.116-118). Nevertheless, the law allows a company to deviate from FRSs if it discloses the effect on its accounts. Even these changes did not take place until near the end of our sample period, 1985-95. We therefore treat UK accounting as less regulated than other sample common-law countries (Australia, Canada and US).

Litigation. Among common-law countries, we also propose that income conservatism increases in the expected costs to accounting firms and their clients from securities litigation. Expected litigation costs affect disclosure decisions by managers and accountants (Kothari, Lys, Smith and Watts 1988), and are a function of lawsuit probability, award size and legal fees. Securities lawsuits induce a demand for conservatism because the payoff function is asymmetric: they almost invariably allege investor losses arising from insufficiently conservative disclosures. We expect countries with higher expected litigation cost of non-disclosure are more likely to demand accounting income that incorporates economic losses in a timely fashion.

Our assumption is that expected litigation costs are lower in UK than in Australia, Canada and

US. Relevant UK institutional facts include: punitive damages are more difficult to obtain; juries are seldom

19

used in civil litigation; absence of class action suits; and the so-called “English rule,” under which losing plaintiffs pay part of defendants’ costs.11 In code-law countries, civil litigation is comparatively rare and the size of awards is comparatively small.

Private Debt. UK corporate debt is predominantly private. We conjecture there is less

information asymmetry between managers and private lenders than in the case of public debt, thus reducing the demand for timely incorporation of economic losses in UK accounting income.

Regulation, Litigation and Private Debt Considered Jointly. Overall, we classify UK as having lower regulatory and litigation costs and predominantly private debt, so we predict less conservatism in UK accounting income than in other common-law countries. Nevertheless, relative to code-law countries, UK is expected to be more income conservative.

H4: U.K. accounting income is less conservative than in other common-law countries and more conservative than in code-law countries.

The data weakly support this hypothesis. The UK negative-return R2 of 11.6% in panel B of table 2 falls between that of the common-law group (12.2%) and the code-law group (1.7%). The incremental negative-return slope β3j of 0.15 in panel A is significantly smaller than those for the other common-law countries (β3j from 0.29 to 0.40), and it is marginally significantly greater than those for the code-law countries (β3j from 0.01 to 0.10).12

4.6 Advantage of the Basu (1997) Piecewise Linear Model. The misspecification in linear models for

2

the common-law countries can be seen from their generally lower Rs in panel C of table 2, compared

with the Basu piecewise linear regressions in panel A. Code-law countries exhibit considerably less

2asymmetry, and hence their Rs differ little between the specifications. In contrast, the asymmetry is

11

The English rule is applied in UK, Australia, and most of Western Europe (Posner 1986, p.537), though not uniformly. It affects expected litigation costs via the frequency of litigation and the legal costs of the defendant, but not the size of awards, with unclear net effect. Katz (1987) argues: “While it is conceivable that the English rule would lower the total number of cases brought to trial, it would likely increase the average expenditure per case. … Even if the number of cases were to fall by as much as 30 percent, total expenditure could rise by more than 50 percent under the English rule.” Katz’ analysis does not take account of one institutional detail, however: English courts award only reasonable costs of successful defendants, thus reducing the incentive to spend. See also Shavell (1982), Posner (1986, pp. 534-540), Hughes and Snyder (1995), Posner (1996) and Miceli (1997). 12

Pope and Walker (1999) claim that the UK income before extraordinary items appears less conservative than US because U.K. firms enjoy greater discretion in reporting economic losses as extraordinary items. Pope and Walker (1999) use data from Datastream whereas we use the Compustat Gloabal Vantage database. Reconciliation of the differences between the two studies is beyond the scope of our research.

20

sufficiently strong in Australia, Canada and US to cause R2s from the Basu model to be approximately 1.6 to 3.2 times their linear model equivalents. These are depicted in Figure 3. Thus, linear models are potentially misleading in an international context, due to substantial international differences in the incorporation of economic losses.

[Figure 3 here]

5. Hypotheses and Tests: Accounting Income versus Dividends and Cash Flows

In this section, we compare accounting income with dividends and cash flows internationally. We argue that code-law institutional links between current-year income and dividends imply they incorporate similar information. Consistent with this hypothesis, we report that accounting income is more timely than dividends in common-law countries but not in code-law countries. Income is less timely than dividends in Germany and Japan. In contrast, accounting income is more timely than cash flows from operations in all countries.

5.1 Laws and Practices Linking Current-period Income and Dividends. Accounting income is influenced, in varying degrees across countries, by firms’ current dividend decisions. One influence results from laws on the taxation of corporate income and of dividend distributions.13 We address these linkages in code-law and common-law countries separately.

Code-law Links between Dividends and Accounting Income. In the code-law “stakeholder” corporate governance model, governments frequently are viewed as stakeholders, and tax payments are viewed as government’s share of the same income “pie” from which dividends and bonuses are paid. Politics requires reported and taxable incomes to converge, particularly since the government is responsible for both tax and accounting codes. Tax rules in the three sample code-law countries allow a deduction against taxable income only if also taken against reported accounting income (Choi and Mueller 1992, p.104). In code-law countries generally, it thus is considered imprudent to report income in excess of those required to justify dividends and bonuses, to minimize corporate tax. The consequence is that accounting income is influenced by short-term dividend policy.

The influence is particularly strong in Germany, where (Nobes and Parker 1995, pp.269-72) the Handelsgesetzbuch (\"commercial code\") includes the Massgeblichkeitsprinzip (\"authoritative principle\") 13

Except as indicated, taxation facts are from annual editions of Coopers & Lybrand International Tax Summaries.

21

that tax accounting be based on the firm's Handelsbilanz (\"commercial balance sheet\"). Choi and Mueller (1992, p.96) conclude of Germany: “The dominance of tax accounting rules means that there is literally no difference between financial statements prepared for tax purposes and financial statements published in financial reports. … Financial reports reflect tax laws – not primarily the information needs of investors and other financial market participants.” The relation between income and dividends is tightened in Germany by two additional institutional factors. First, Federal law forbids management from paying dividends less than 50% of income without stockholders voting approval. Second, undistributed profits are taxed at a higher rate than distributed profits (currently 45% versus 30%, excluding the “solidarity surcharge”).

These code-law institutional factors impose an additional role on the (already reduced) public-disclosure role of accounting income: corporate policy on current payouts to “stakeholders,” including governments (via taxation), shareholders (via dividends), and managers and employees (via bonuses). If these distributions per se are uninformative, then the testable implication is that code-law accounting income is noisier and less oriented to incorporating current economic income, relative to dividends, than common-law income.

H5: Code-law accounting income is less timely relative to dividends than common-law. Table 4 reports the contemporaneous association between returns and dividends, for the same sample of firm-years as for accounting income in table 2. Figure 4 depicts comparative R2s for income and dividends. Results generally support the hypothesis that common-law accounting income is more timely relative to dividends than code-law. In Panel A, the ratio of the returns/income R2 to the returns/dividends R2 exceeds unity for three of the four common-law countries. In contrast, dividends are more timely than income in two of the three code-law countries (Germany and Japan). In Germany, where there are particularly binding institutional links between them, accounting income captures only 56% as much current value relevant information as dividends. The equivalent ratio for Japan is 78%. For the pooled common-law countries, the ratio is 170%. In contrast, for the pooled code-law group, the ratio is 96%, meaning current-year income and dividends capture approximately the same amount of current-year economic income. The results are consistent with the timeliness of code-law accounting income being constrained by its direct role in determining current-year dividend payouts.

[Table 4 and Figure 4 here]

22

Panel B of table 4 reports separate dividend/return relations in positive and negative return years,

comparable to panel B of table 2 for accounting income. Since these are univariate regressions, we are able to report Vuong’s (1989) likelihood ratio statistic for selection among non-nested models, computed separately for positive and negative returns. The competing models have a common dependent variable (returns) but different independent variables (accounting income and dividends). The results are consistent with hypothesis 5. We also estimate, but do not report, a linear relation model with dividends levels and changes, comparable to the panel C of table 2. These results also support hypothesis 5.

Common-law Tax Imputation Link between Dividends and Accounting Income. Among common-law countries, there is considerable variation in the taxation of dividend distributions. We focus on “imputation”, which penalizes corporations for reporting taxable income in excess of distributed dividends. There is an incentive to structure transactions and use accounting standards that make taxable income conform to dividend policy, which affects accounting income because it is correlated with taxable income. Imputation systems are operated by Australia, Canada and UK (Coopers & Lybrand 1982-95), but not US.14 Provided dividends per se are uninformative, the testable implication is that US accounting income exhibits greater timeliness relative to dividends.15

H6: The differential timeliness of accounting income relative to dividends is greater in the US than in common-law countries with dividend imputation (Australia, Canada, UK). There are mixed results for this hypothesis. The results for Australia and UK reported in Panel A of Table 4 are consistent with the hypothesis, but those for Canada are not. One conjecture is that the marginal investor in Canada is a US resident and does not receive imputation credits, so that Canada effectively is in the same tax category as the US.

5.2 Accounting Income and Cash Flows from Operations. We propose that cash flows are noisier than accounting income in reflecting contemporaneous value-relevant information (Dechow, 1994) in all countries. Provided there is no new information in managers’ current financing and investment decisions, accruals are expected to make accounting income incorporate economic income in a more timely basis

14

The Australian imputation system is described in Hamson and Ziegler (1990). The Canadian system is dividend credits, not full imputation; and the marginal Canadian investor could pay US tax and thus not receive credits. 15

We expect this to be reinforced in US by stock repurchases to distribute cash to investors. Repurchases contain information about future income (Dann, Masulis and Mayers 1991), likely reducing information in dividends.

23

than cash flows (Dechow, 1994). This logic holds under the accrual-accounting systems of all countries, so we test the following hypothesis:

H7: For all countries, accounting income is more timely than cash flow from operations, incorporating more current-year economic income.

Table 5 reports results for the contemporaneous association between returns and cash flow from operations, for the same sample of firm/years as for income and dividends in tables 2 and 4. In all countries, the income R2s are higher than those of operating cash flow, the ratio ranging from 2.10 (Germany) to 4.78 (US). This result would be obtained in less than 1% of the cases by chance, assuming independence across the seven countries. The Vuong statistic and the ratio of explanatory powers of the income/return and cash flow/return relations reported in panel B indicate that the greater timeliness of accounting income than operating cash flows is due largely to the bad news years. Comparative R2s for accounting income and operating cash flows are summarized in figure 5. These findings generalize well-known US results on timeliness of accounting income compared to cash flows (Ball and Brown 1968, Dechow 1994, Basu 1997), and contrast sharply with results reported above for dividends.

[Table 5 and figure 5 here]

The incremental timeliness of accounting income (relative to cash flows) measures the extent to which accruals under a particular country’s accounting rules are oriented toward timely incorporation of economic income. We hypothesize that common-law accruals reflect a greater demand for timely accounting income, and hence predict that common-law accounting income has more incremental timeliness (relative to cash flows) than code-law accounting income.

H8: Common-law accounting income is more timely relative to operating cash flows. Results generally support hypothesis 8. In panel A of table 5, three of the top four ratios of the

2 2

returns/income Rto the returns/dividends R are from common-law countries. Accounting income in

Germany and Japan incorporates the least amount of current-period economic income, relative to cash flows. Thus, the greater timeliness of common-law income is not due entirely to properties of the underlying transactions.

6. Specification Tests: Income and Returns

The relation between accounting income and stock returns is robust with respect to a variety of specification tests, reported in columns (2) through (5) of table 3 and in table 6.

24

6.1 Variation in Expected Returns. When economic income is the independent variable, a possible misspecification arises from variation in expected returns (across time, countries and firms). Assume the true model for accounting income incorporates only the component of ∆Vt that is due to information about future cash flows a, not to variation in expected returns r. Assume these are independent, both of each other and over time. Our model (2) then can be simplified as:

NIt = h(at, ζt)

(6)

where the disturbance term ζt now incorporates expected return effects, in addition to lagged information about cash flows and accounting noise. Using ∆Vt as a proxy for at then would introduce measurement error in the independent variable, due to variation in expected returns.

Table 3 reports pooled regressions with two alternative controls for market-wide expected return effects. In column (2) the independent variable, annual return, is defined relative to the mean return for the firm’s country/year. In column (3) the dependent variable, accounting income, is scaled by the country/year long term interest rate. In both cases, results are similar to those in column (1). We conclude our results are not substantially influenced by market-wide effects.

6.2 Consolidated versus Parent Income. In countries where parent companies are not required to “equity account” their share of affiliated-company income, their accounting income omits a component that accrues to their stockholders. This could be viewed as introducing measurement error that likely is positively correlated with parent-company income, thereby biasing the regression slopes downward. A counter-argument is that we are interested in properties of accounting income as reported under different countries’ accounting rules, including any omission of affiliate income. Nevertheless, we replicate column (1) results for the 33,441 total firm/years for which Periodic Descriptor Array No. 13 on Global Vantage specifically labels income as “fully consolidated,” thus including parent companies’ proportional shares of affiliates’ incomes. The sample size reduction is due almost entirely to Japan, which loses most of its observations.16 Results (unreported) are almost identical except for Japan (due to the small sample size). 6.3 Extending the Lag in Accounting Income. Column (4) of Table 3 reports results with the dependent variable NI2yr = NIt+NIt+1, allowing an additional year to incorporate economic income in accounting 16

Excluding these observations is not clearly necessary because Japanese rules require equity accounting in the parent’s books (Nobes and Parker 1995, ch.13). Few observations are excluded for Germany where public companies have issued consolidated income since the Euopean Union’s Seventh Directive.

25

income. As expected, the coefficients generally increase: the median two-year coefficient is 1.6 times its one-year equivalent.17 Accounting income thus incorporates economic income over time (it is “smoothed,” and has “momentum” or “persistence”).

6.4 Control for Industry Composition. Another concern is correlated omitted variables. If the proportion of growth options relative to assets in place varies across countries, then common application of the Revenue Realization Rule will cause accounting income to vary internationally in timeliness, due to “real” rather than “accounting” effects. This concern is reduced by our result that international differences in timeliness are due in part to accounting accruals and not entirely to differences in timeliness of cash flows. A related concern is that our results hold only for listed corporations, which are less prevalent under code-law systems that do not facilitate public disclosure and public capital markets (see La Porta et al., 1997), to the same extent as common-law systems. For example, UK has approximately four times as many listed corporations as Germany. To alleviate concern that our results are due to different sample composition across countries, column (5) reports results after controlling for industry effects, in the form of interactive dummy variables for the ten most prevalent 2-digit SIC codes. Little change is apparent. The

2

regression adjusted R rises slightly and the F-statistic falls. The F-statistics for the countries’ dummy

slopes fall slightly. All remain significant at the 0.01 level. There is little change in the coefficients for both individual countries and country groups.

6.5 Subperiod Results. Splitting the sample into two subperiods, 1985-90 and 1991-95, reveals three interesting results, reported in table 6. First, the incremental coefficients on negative economic income increase in the second subperiod for six of the seven individual-country regressions (4), the exception being Japan. Several countries exhibit statistically significant increases, including Australia (0.48 vs. 0.26), UK (0.19 vs. 0.09), France (0.22 vs. 0.04) and Germany (0.19 vs. 0.05). The coefficient on negative economic income for US increases, but not significantly (0.33 vs. 0.28). One possibility is that timely incorporation of economic losses has become a more important corporate governance mechanism over time (though not in Japan). In any event, the results suggest Basu (1997) might have erred in attributing

17

The exception is the positive-return slopes for US and Australia. The sample size decreases, due mainly to losing one year. Results are for overlapping samples.

26

increased asymmetry over time in US income to changes in US litigation rules, since similar increases have occurred in France and Germany, where litigation is not an issue.

[Table 6 here]

Second, despite increased sensitivity to economic losses, the R2s in six of seven country regressions (4) decreased between subperiods, generalizing the US result [Ramesh and Thiagarajan (1995); Lev (1997)]. Third, the results for cash flow from operations, reported in panel C, contrast unexpectedly with those for accounting income. Between subperiods, the point estimates of the slopes on positive returns increased for six of seven countries, the incremental slopes on negative returns show no systematic change, and the R2s in all of the individual-country cash flow regressions increase. We have no explanation for this apparently systematic change. 7. Conclusions, implications, and limitations

Academic and professional interest in different national accounting models has been rekindled by

the worldwide trend toward “internationalization” of the markets, especially capital markets, which use accounting information. The properties of accounting information prepared under common-law accounting standards is of particular contemporary interest because the International Accounting Standards Committee (IASC) currently is completing a set of “international” accounting standards, which are widely viewed as reflecting a largely common-law view of “transparent,” timely disclosure.

We show that common-law accounting income does indeed exhibit significantly greater timeliness

than code-law accounting income, but that this is due entirely to greater sensitivity to economic losses (income conservatism). This result has important implications for corporate governance. We conjecture that early incorporation of economic losses, as distinct from gradual incorporation over time, increases managers’ incentives to attend to the sources of losses more quickly. It brings more and quicker pressure from security analysts, makes leverage and dividend restrictions binding more quickly, and affects managers’ and employees’ bonuses. In comparison with a system that allows economic losses to be reflected in accounting income gradually over time, it also makes managers’ optimistic statements less credible. Conservative common-law accounting thus facilitates monitoring of managers and is an important feature of code-law corporate governance. In contrast with Roe (1994), we conclude that enhanced

27

common-law disclosure standards reduce the agency costs of monitoring managers, thus countering the advantages of closer shareholder-manager contact in code-law countries.

Our results suggest an explanation for the emergence of a largely common-law model in international transacting, and in particular for the IASC adoption of a more “transparent” common-law approach to disclosure. The cost of cross-border transacting by parties who are geographically, culturally and linguistically separated from the firm’s management presumably is greater in a system that assumes they are informed insiders than in one with more timely public disclosure. Timely disclosure of economic losses is particularly important to cross-border lenders.

Greater common-law income conservatism should be no surprise, considering the use of accounting income in common-law arm’s-length debt and equity markets, and especially considering common-law litigation. Nevertheless, German accounting in particular is widely presumed to be more conservative, because German managers have unusual discretion to reduce reported income during good years. However, they also have unusual discretion to increase income in bad years, a practice common in the UK at the beginning of the twentieth century (Yamey 1962), but effectively extinguished in common-law countries after the Royal Mail case.18

Our research design is subject to several limitations. The most obvious concern is the validity of stock returns return as a proxy for economic income, particularly in code-law countries, which have endogenously lower liquidity and public disclosure standards. “Noise” in annual stock returns as a measure of market-valued income could be a correlated omitted variable. We counter with three observations. First, the flow of information into stock prices is not necessarily impeded in code-law countries by poor public disclosure, but occurs instead via the trading of informed insiders. In the absence of insider-trading laws, which are fundamentally incompatible with code-law governance, corporate insiders’ incentives are to trade on information and thereby incorporate it into prices. As Ramseyer (1993) observes, this explains the practice of Japanese banks trading in their clients’ stock. To some degree, equating poor public disclosure with uninformed stock prices involves projecting common-law precepts onto code-law institutions, which solve information asymmetry by private rather than public communication. Second, our 18

Rex v. Lord Kylsant 1932 1 KB 442. This case-law was codified in the 1948 UK Companies Act, which required companies to distinguish reserves from provisions, “making the creation of secret reserves more difficult.” (Nobes and Parker 1992, p.103). German practice is exemplified by the notorious Daimler-Benz case (Ball 1998).

28

results seem inconsistent with the hypothesis that stock returns reflect poor information flows in code law countries, which would imply less anticipation of income, a stronger income “surprise” effect, and a stronger return-income association. We observe the opposite. Third, we study stock returns over intervals of one and two years, not over short “event windows,” long enough for liquidity and disclosure timing effects to have minimal impact.

Correlated omitted variables are another obvious concern. For example, if the proportion of growth options relative to assets in place varies across countries, then common application of the Revenue Realization Rule will cause income to vary internationally in timeliness, due to “real” rather than “accounting” effects. Similarly, our results hold only for listed corporations. The number of listed code-law corporations is comparatively small, in part because code-law systems are not designed for the demands of public disclosure, and thus private corporations are comparatively more prevalent. Thus, our sample is less representative of code-law accounting in general. This concern is addressed in the table 3 control for SIC codes. Another concern is whether the sample period is representative. Japan, for example, went through boom and bust during 1985-95. This concern also is addressed by the specification tests in table 3, where the results change little when annual firm return is defined relative to the mean return for that country/year, and when income is scaled by the country/year interest rate.

A final concern is our characterization of the relevant institutional features of the seven countries in our primary sample. To some degree this is alleviated by the consistent evidence we report from our primary sample. In addition, we report consistent evidence from a secondary sample of eighteen countries, suggesting that our results are generalizable.

The research design also has its strengths. It studies the incorporation of economic income into accounting income over time, under different international institutions, using a model of accounting income determination that is based on fundamental properties of accounting. We believe the results to be of interest to accountants, analysts, standard-setters, regulators and students of corporate governance.

References

29

Alchian, A., 1984, Specificity, Specialization, and Coalitions, Journal of Institutional and Theoretical

Economics 140, 34-49.

Alchian, A. and H. Demsetz, 1972, Production, information costs and economic organization, American

Economic Review 62, 777-795.

Alford, A., J. Jones, R. Leftwich, and M. Zmijewski, 1993, The relative informativeness of accounting

disclosures in different countries, Journal of Accounting Research 31, Supplement, 183-223.

American Accounting Association, 1977, Report of the Committee on International Accounting,

Supplement to Accounting Review 52.

Amir, E., T. Harris and E. Venuti, 1993, A comparison of the value-relevance of U.S. versus non-U.S.

GAAP accounting measures using form 20-F reconciliations, Journal of Accounting Research 31, Supplement, 230-264.

Ball, R., 1995, Making accounting more international: Why, how, and how far will it go? Journal of

Applied Corporate Finance 8, Fall, 19-29.

Ball, R., 1998, Daimler-Benz AG: Evolution of corporate governance from a code-law “stakeholder” to

a common-law “shareholder value” system, unpublished manuscript, University of Rochester.

Ball, R. and P. Brown, 1968, An empirical evaluation of accounting income numbers, Journal of

Accounting Research 6, 159-178.

Ball, R. and R. Watts, 1972, Some time series properties of accounting income, Journal of Finance 27,

663-82.

Bandyopadhyay, S., D. Hanna and G. Richardson, 1994, Capital market effects of U.S.-Canada

GAAP differences, Journal of Accounting Research 32, 262-277.

Barth, M. and G. Clinch, 1996, International accounting differences and their relation to share prices:

Evidence from U.K., Australian, and Canadian firms, Contemporary Accounting Research 13, 135-170.

Basu, S., 1997, The conservatism principle and the asymmetric timeliness of earnings, Journal of

Accounting & Economics 24, 3-37.

Baums, T., R. Buxbaum, and K. Hopt, eds., 1994, Institutional Investors and Corporate Governance

(Berlin: Walter de Gruyter).

30

Beaver, W., R. Lambert, and D. Morse, 1980, The information content of security prices, Journal of

Accounting & Economics 2, 3-28.

Choi, F. and G. Mueller, 1992, International Accounting (Englewood Cliffs, N.J.: Prentice-Hall).

Coopers & Lybrand, 1996, Understanding IAS (London: Coopers & Lybrand).

Coopers & Lybrand, 1997, SEC Manual, 7th edn. (New York: John Wiley & Sons).

Coopers & Lybrand, 1982-95, International Tax Summaries, annual edns. (New York: John Wiley &

Sons).

Collins, D., E. Maydew and I. Weiss, 1997, Changes in the value-relevance of earnings and book

values over the past forty years, Journal of Accounting & Economics 24, 39-67.

2

Cramer, J., 1987, Mean and variance of R in small and moderate samples, Journal of Econometrics

35, 253-166.

Dann, L., R. Masulis and D. Mayers, 1991, Repurchase tender offers and earnings information, Journal

of Accounting & Economics 14, 217-251.

David, R. and J. Brierley, Major Legal Systems in the World Today (London: Stevens & Sons).

Dechow, P., 1994, Accounting earnings and cash flows as measures of firm performance: The role of

accounting accruals, Journal of Accounting & Economics 18, 3-42

Easton, P. and T. Harris, 1991, Earnings as an explanatory variable for returns, Journal of Accounting

Research 29, 19-36.

Easton, P., T. Harris and J. Ohlson, 1992, Aggregate accounting earnings can explain most of security

returns, Journal of Accounting & Economics 15, 119-142.

Financial Accounting Standards Board (FASB), 1985, Statement of financial accounting concepts no. 6

(FASB: Stamford, Connecticut).

Gray, S., 1980, International accounting differences from a security analysis perspective, Journal of

Accounting Research 18 , 64-76.

Hamson, D. and P. Ziegler, 1990, The impact of dividend imputation on firms’ financial decisions,

Accounting and Finance 30, 29-53.

Harris, T., M. Lang, and H. Möller, 1994, The value relevance of German accounting measures: An

empirical analysis, Journal of Accounting Research 32, 187-209.

31

Healy, P., 1985, The effect of bonus schemes on accounting decisions, Journal of Accounting and

Economics 7, 85-107.

Hughes, J. and E. Snyder, 1995, Litigation and settlement under the English and American rules: Theory

and evidence, Journal of Law and Economics 38, 225-250.

Jacobsen, R. and D. Aaker, Myopic management behavior with efficient, but imperfect, financial

markets, Journal of Accounting and Economics 16, 383-405.

Joos, P. and M. Lang, 1994, The effects of accounting diversity: Evidence from the European Union,

Journal of Accounting Research 32 (Supplement), 141-168.

Katz, A., 1987, Measuring the demand for litigation, Journal of Law, Economics, & Organization 3,

143-176.

Kothari, S. and R. Sloan, 1992, Information in earnings about future earnings: Implications for earnings

response coefficients, Journal of Accounting & Economics 15, 143-171.

Kothari, S., T. Lys, C. Smith and R. Watts, 1988, Auditor liability and information disclosure, Journal

of Accounting, Auditing and Finance 3, 307-339.

Köndgen, J., 1994, Duties of banks in voting their clients’ stock, Chapter 18 in Baums, Buxbaum and

Hopt (1994).

La Porta, L., F. Lopez-de-Silanes, A. Shleifer and R. Vishny, 1997, Legal determinants of external

finance, Journal of Finance 52, 1131-1150.

Lev, B., 1997, The boundaries of financial accounting and how to expand them, unpublished

manuscript, New York University.

Lintner, J., 1956, Distribution of incomes of corporations among dividends, retained earnings, and

taxes, American Economic Review 46, pp. 97-113.

Miceli, T., 1997, Economics of the law (Oxford: Oxford University Press).

Miwa, Y., 1996, Firms and industrial organization in Japan (NY: New York University Press).

Mueller, G., H. Gernon, and G. Meek, 1997, Accounting: An International Perspective (Chicago, IL.:

Richard D. Irwin).

Nobes, C., 1992, International Classification of Financial Reporting (Routledge, London).

32

Nobes, C. and R. Parker, 1995, Comparative International Accounting, 4th edn. (Englewood Cliffs, NJ:

Prentice-Hall).

Ohlson, J., 1988, Accounting earnings, book value and dividends: The theory of the clean surplus

equation, unpublished manuscript, Columbia University.

Peltzman, S., 1976, Toward a more general theory of regulation, Journal of Law and Economics 19,

211-240.

Pope, P. and M. Walker, 1999, International differences in the timeliness, conservatism, and

classification of earnings, working paper, Lancaster University, U.K.

Posner, R. A., 1986, Economic Analysis of Law, 3rd edn. (Boston, Mass.: Little, Brown).

Posner, R.A., 1996, Law and legal theory in the UK and USA (Oxford: Clarendon Press).

Radebaugh, L. and S. Gray, 1997, International Accounting and Multinational Enterprises, 4th edn.

(New York, NY: John Wiley & Sons).

Ramesh, K. and R. Thiagarajan, 1995, Inter-temporal decline in earnings response coefficients,

unpublished manuscript, Northwestern University.

Ramseyer, J., 1993, Columbian cartel launches bid for Japanese firms, Yale Law Journal 102, 2005-2020.

Roe, M., 1994, Some differences in corporate governance in Germany, Japan and America, chapter 2

in Baums, Buxbaum and Hopt (1994).

Shavell, S., 1982, Suit, settlement and trial: A theoretical analysis under alternative methods for the

allocation of legal costs, Journal of Legal Studies 11, 55-81.

Skinner, D., 1997, Earnings disclosures and stockholder lawsuits, Journal of Accounting & Economics

23, 249-282.

Standish, P., 1997, The French Plan Comptable (Paris: Expert Comptable Média).

Vuong, Q., 1989, Likelihood ratio tests for model selection and non-nested hypotheses, Econometrica

57, 307-333.

Watts, R. and J. Zimmerman, 1986, Positive Accounting Theory (Englewood Cliffs, NJ: Prentice-Hall).

Yamey, B., 1962, Some topics in the history of financial accounting in England, 1500-1900, in: W.T.

Baxter and S. Davidson, eds.: Studies in Accounting Theory (London: Sweet & Maxwell).

Table 1

Sample Characteristics

A. Descriptive Statistics

N R

µ Med σ Australia 1321 17.3 9.9 52.6 Canada 2901 12.1 6.0 46.8 US 21225 12.7 8.1 42.8 UK 5758 13.5 9.8 38.0 France 1054 14.9 6.6 43.0 Germany 1245 8.9 4.2 31.5 Japan 6855 3.7 -2.8 33.4 B. Observations by Fiscal Year End (Month) 1 2 3 4 Australia 5 0 32 4 Canada 113 26 131 51 US 964 405 872 401 UK 297 119 1222 333 France 0 2 23 0 Germany 0 19 4 0 Japan 94 363 5341 39 Total 1473 934 7625 828

µ 2.6 3.2 3.1 6.6 6.1 3.7 1.7 5 30 21 446 119 0 0 68 684 89 116 306 2159 621 98 90 841 4231

NI Med 6.7 5.3 6.2 7.2 6.6 4.2 1.8 6 964 105 2094 344 3 75 29 3614 90 165 325 2098 703 147 145 881 4464

σ 19.0 15.4 14.4 8.8 10.8 9.0 2.1 7 32 47 562 121 0 9 9 780 91 160 306 2006 722 155 180 850 4379

µ 3.5 2.0 1.9 3.4 2.3 2.0 0.7 8 2 144 506 143 12 0 14 821 92 164 311 2111 751 151 189 857 4534

DIV Med 3.6 1.4 1.0 3.3 2.1 2.0 0.7 9 45 155 1666 590 24 183 150 2813 93 181 317 2153 756 131 170 874 4582

σ 3.3 2.4 2.4 2.0 1.9 1.7 0.4 10 0 72 700 164 5 7 67 1015 94 178 325 2189 733 146 192 814 4577

µ 11.6 15.1 11.7 12.7 22.1 18.4 4.7 11 6 31 425 66 0 3 144 675 95 138 48 788 114 5 26 0 1119

OCF Med 10.9 12.2 10.3 10.4 15.7 14.3 4.2 12 201 2005 12184 2240 985 945 537 19097 Total 1321 2901 21225 5758 1054 1245 6855 40359

σ 33.3 28.9 21.4 18.8 36.2 26.8 7.8 Total 1321 2901 21225 5758 1054 1245 6855 40359

33

C. Observations by Year

Australia Canada US UK France Germany Japan Total

85 11 189 1586 98 9 20 3 1916

86 29 221 1661 314 66 74 328 2693

87 88 277 2234 438 74 83 668 3862

88 91 276 2240 508 72 76 739 4002

Sample consists of 40,359 firm-year observations selected from the Global Vantage industrial/Commercial and Issue files over 1985-95, using the following procedure. First, for each variable (see below) we eliminate the two extreme percentiles of firm-year observations. Second, we eliminate all firm-year observations with missing values for one or more variables, to facilitate comparability with results in previous tables. Third, we eliminate all firm-year observations from countries with less than 1,000 observations, leaving seven countries represented. Australia, Canada, United States, and the United Kingdom are the common-law countries, the rest are code-law countries.

R = security return over the fiscal year;

NI = annual earnings per share before extraordinary items deflated by beginning of period price; DIV = annual dividends per share deflated by beginning of period price;

OCF = annual operating cash flow per share deflated by beginning of period price; N = the number of firm/year observations.

34

Table 2

Contemporaneous Association between Earnings and Returns: Various Specifications

Statistics from pooled cross-section and time-series regressions using firm-year observations for each country. Intercepts are not reported

Panel A: NI = β0 + β1 RD + β2 R + β3 R*RD + ε

2

β2 t(β2)β3 t(β3)Adj. RN Australia -0.01 -0.530.37 8.639.1% 1321 Canada 0.00 0.120.40 17.2117.0% 2901 USA 0.03 8.570.29 34.0214.7% 21225 UK 0.04 10.140.15 13.3213.8% 5758 France 0.08 7.300.07 2.3012.6% 1054 Germany 0.05 4.280.10 3.275.4% 1245 Japan 0.01 5.950.01 2.584.2% 6855 Common 0.02 7.070.31 39.1014.4% 25447 UK 0.04 10.140.15 13.3213.8% 5758 Code 0.04 13.270.01 2.195.2% 9154 Panel B: NI = β0 + β1 R + ε (R ≥ 0) NI = β0 + β1 R + ε (R < 0) β1 Adj. R2N β1 Adj. R2N Australia -0.01 -0.1%813 0.36 10.1%508 Canada 0.00 -0.1%1688 0.40 17.9%1213 USA 0.03 0.8%12721 0.32 11.7%8504 UK 0.04 3.2%3612 0.19 11.6%2146 France 0.08 9.0%611 0.16 4.7%443 Germany 0.05 2.4%712 0.16 4.9%533 Japan 0.01 1.1%3141 0.02 0.8%3714 Common 0.02 0.4%15222 0.33 12.2%10225 UK 0.04 3.2%3612 0.19 11.6%2146 Code 0.04 3.3%4464 0.05 1.7%4690 Panel C: R = β0 + β1 NI + ε R = β0 + β1 NI + β2 ∆NI + ε β1 t(β1)Adj. R2 β1 t(β1) β2t(β2) Adj. R2N Australia 0.47 6.302.9% 0.44 5.75 0.122.84 3.4%1321 Canada 0.79 14.496.7% 0.72 13.06 0.266.44 8.0%2901 USA 0.90 46.089.1% 0.76 36.52 0.3718.58 10.5%21225 UK 1.44 26.6711.0% 1.21 20.40 0.469.07 12.2%5758 France 1.40 12.2212.3% 1.28 10.86 0.464.23 13.7%1054 Germany 0.76 7.844.6% 0.71 6.78 0.131.30 4.7%1245 Japan 3.16 16.824.0% 2.23 11.04 2.8511.83 5.9%6855 Common 0.85 47.528.2% 0.74 39.71 0.3018.92 9.4%25447 UK 1.44 26.6711.0% 1.21 20.40 0.469.07 12.2%5758 Code 1.44 22.445.2% 1.29 19.10 0.477.04 5.7%9154

35

Sample consists of 40,359 firm-year observations selected from the Global Vantage industrial/Commercial and Issue files over 1985-95, using the following procedure. First, for each variable (see below) we eliminate the two extreme percentiles of firm-year observations. Second, we eliminate all firm-year observations with missing values for one or more variables, to facilitate comparability with results in previous tables. Third, we eliminate all firm-year observations from countries with less than 1,000 observations, leaving seven countries represented. Australia, Canada, United States, and the United Kingdom are the common-law countries, the rest are code-law countries.

R = security return over the fiscal year;

NI = annual earnings per share before extraordinary items deflated by beginning of period price; DIV = annual dividends per share deflated by beginning of period price;

OCF = annual operating cash flow per share deflated by beginning of period price; N = the number of firm/year observations.

36

Table 3

Comparative Asymmetry in the Contemporaneous Returns-Earnings Relation:

Pooled Regressions with Individual-country Effects

Reported statistics are for the following model using the pooled cross-section and time-series of firm/year observations for all countries:

NI = β0 + Σj β0j CDj + β1 RD + Σj β1j RD.CDj + β2 R + Σj β2j R.CDj + β3 R.RD + Σj β3j R.RD.CDj

Results are not reported for the intercept, the negative return intercept, and their respective country dummies. The country category models use the common law countries of Australia, Canada and the US as the base category; dummies are used for (1) the UK and (2) the code law countries of France, Germany and Japan.

Model (1) (2) (3) (4) (5) Y: NI NI NI/RF NI2yr NI Control: Rjt SIC

Coeff t-stat Coeff Panel A: Country Dummies Model Earnings “Good News” sensitivity β2 (Return) 0.03 9.88 0.01 β2j (Return* Cntry Dums): Australia -0.03 -3.95 -0.04 Canada -0.03 -3.80 -0.01 UK 0.02 2.40 0.03 France 0.05 4.16 0.06 Germany 0.03 1.57 0.04 Japan -0.02 -2.64 0.00 F-stat for Country 11.35 Return Dummies p<.01

Incremental “Bad News” Sensitivity β3 (Rdum* 0.29 39.21 0.29 Return)

β3j (Rdum* Ret*Cntry Dums): Australia 0.07 2.64 0.01 Canada 0.11 5.28 0.05 UK -0.14 -8.26 -0.17 France -0.22 -5.81 -0.14 Germany -0.19 -4.41 -0.24 Japan -0.29 -15.74 -0.29 F-stat: Country 67.44 Negative Ret Dums p<.01 Regression: N 40,359 Adj. R2 15.4%

t-stat Coeff t-stat Coeff t-stat Coeff t-stat

3.96 -3.17 -1.63 3.74 3.48 1.81 -0.51 7.73 p<.01 47.25

0.36

-0.48 -0.32 0.14 0.65 0.55 -0.17

9.88 -0.05 -84.24

-4.24 -0.02 -3.02 -3.54 0.06 10.03 1.57 0.09 2.92 3.86 0.55 21.28 2.54 0.21 19.32 -1.69 0.07 6.82 10.28 164.13 p<.01 p<.01 3.66 38.11 0.43 26.13

0.12 1.850.12 2.43 -0.24 -2.47 -0.68 -7.40 -0.31 -7.48 -0.42 -10.70

37.81 p<.01 33,082 22.5%

0.04

-0.04 -0.02 0.01 0.05 0.02 -0.02

10.25

-4.15 -3.28 1.57 3.69 1.42 -3.12 9.75 p<.01 0.29 31.04

0.37 2.92 -11.11 -4.40 -5.74 -16.80 72.74 p<.01

40,359 16.8%

0.19 0.50 0.45 1.70 -2.05 -9.10 -2.87 -5.99 -2.49 -4.55 -3.43 -14.90

54.55 p<.01 39,240 13.8% 0.08 2.89 0.11 5.25 -0.14 -7.96 -0.21 -5.54 -0.20 -4.73 -0.28 -15.16

63.43 p<.01

40,359 16.2%

F

272.98

303.32

234.19

410.56

37

117.16

Panel B: Country Category Dummies Model Earnings “Good News” Sensitivity β2 (Return) 0.02 8.36 0.01 2.82 0.26 8.22 -0.05 -83.70 β2j (Return* Cntry Cat Dums): UK 0.02 3.52 0.04 4.35 0.23 2.75 0.20 19.12 Code 0.01 2.37 0.02 2.87 0.24 2.99 0.15 15.26 F-stat for Cntry Cat 7.94 12.24 7.20 297.97 Return Dums p<.01 p<.01 p<.01 p<.01 Incremental “Bad News” Sensitivity β3 (RDum* 0.31 46.23 0.30 53.64 3.76 43.28 0.46 30.27 Return)

β3j (RDum* Ret*Cntry Cat Dums): UK -0.16 -9.50 -0.17 -11.72 -2.14 -9.69 -0.34 -8.14 Code -0.30 -19.00 -0.28 -18.82 -3.47 -17.41 -0.48 -13.65

F-stat: Cntry Cat 200.92 218.66 174.82 111.56 Negative Ret Dums p<.01 p<.01 p<.01 p<.01 Regression: N 40,359 40,359 39,240 33,082 Adj. R2 14.9% 16.2% 13.5% 21.3% F 642.07 711.90 557.84 938.83

38

0.03

0.02 0.01

8.99

2.58 1.53 3.98 p<.01 0.31 36.68 -0.16 -9.25 -0.29 -18.31

187.17 p<.01

40,359 15.7% 148.07

Sample consists of 40,359 firm-year observations selected from the Global Vantage industrial/Commercial and Issue files over 1985-95, using the following procedure. First, for each variable (see below) we eliminate the two extreme percentiles of firm-year observations. Second, we eliminate all firm-year observations with missing values for one or more variables, to facilitate comparability with results in previous tables. Third, we eliminate all firm-year observations from countries with less than 1,000 observations, leaving seven countries represented. Australia, Canada, United States, and the United Kingdom are the common-law countries, the rest are code-law countries. The sample sizes for model (3) that uses NI/RF as the dependent variable and model (4) that uses only the two-year NI values are 39,240 and 33,082 respectively.

Variables and Controls:

R = holding-period security return over the fiscal year;

NI = annual earnings per share before extraordinary items deflated by beginning of period price; NI/RF = earnings deflated by the long-term interest rate; NI2yr = the sum of net income over the two years t and t+1; RD = the proxy for bad news = 1 if R < 0 and = 0 otherwise;

CDj = country identifier = 1 for firm/years in country j and = 0 otherwise. USA is the “base country” with CDj = 0 ∀ j.

In model (2) the regression is controlled for mean annual country return by scaling NI and by redefining RD; in model (5) the regression is controlled for the 10 most prevalent 2-digit SIC code by creating the appropriate dummy variables based on these SIC codes.

Interest rates: RF is measured as the longest term bond yield available in the IMF’s International Financial Statistics for the calendar year most coincident with the firm’s fiscal year [-11,0]. Rates for the individual countries are for: treasury bonds 2 years (Australia); govt. bond yield > 10 years (Canada); govt. bond yield (Germany); govt. bond yield, moyens (France); long term govt. bond yield (UK); govt. bond yield (Japan); govt. bond yield 10 year (USA).

Table 4

Contemporaneous Association between Dividends and Returns

39

Statistics are from regressions using the pooled cross-section and time-series of firm/year observations for each country. Intercepts are not reported.

Panel A: DIV = β0 + β1 RD + β2 R + β3 R*RD + ε β2 t(β2)β3 t(β3)Adj. R2 N Ratio Australia 0.00 -1.850.06 8.6711.1% 1321 0.82 Canada -0.01 -9.060.04 11.267.3% 2901 2.34 USA -0.02 -26.540.05 34.249.2% 21225 1.60 UK 0.01 8.690.02 9.259.7% 5758 1.42 France 0.01 4.220.02 3.908.8% 1054 1.43 Germany 0.02 6.660.01 0.999.6% 1245 0.56 Japan 0.00 2.280.01 9.725.5% 6855 0.78 Common -0.01 -25.530.05 35.678.4% 25447 1.70 UK 0.01 8.690.02 9.259.7% 5758 1.42 Code 0.01 10.080.01 3.785.5% 9154 0.96 Panel B: DIV = β0 + β1 R + ε (R≥0) DIV = β0 + β1 R + ε (R<0) β1 Adj. R2N Ratio Vuong β1 Adj. R2N Ratio Vuong Australia 0.00 0.2%813 -0.360.65 0.06 18.3%508 0.552.41 Canada -0.01 3.8%1688 -0.025.22 0.03 7.9%1213 2.26-3.69 USA -0.02 4.4%12721 0.189.67 0.04 10.4%8504 1.12-1.48 UK 0.01 1.9%3612 1.73-1.74 0.03 9.3%2146 1.24-1.48 France 0.01 2.2%611 4.12-2.46 0.03 9.5%443 0.491.88 Germany 0.02 4.8%712 0.501.43 0.02 3.8%533 1.31-0.6 Japan 0.00 0.1%3141 11.00-2.74 0.01 4.5%3714 0.196.06 Common -0.01 3.4%15222 0.138.71 0.04 10.3%10225 1.19-2.45 UK 0.01 1.9%3612 1.73-1.74 0.03 9.3%2146 1.24-1.48 Code 0.01 1.5%4464 2.17-2.87 0.01 2.7%4690 0.621.89

Sample consists of 40,359 firm-year observations selected from the Global Vantage industrial/Commercial and Issue files over 1985-95, using the following procedure. First, for each variable (see below) we eliminate the two extreme percentiles of firm-year observations. Second, we eliminate all firm-year observations with missing values for one or more variables, to facilitate comparability with results in previous tables. Third, we eliminate all firm-year observations from countries with less than 1,000 observations, leaving seven countries represented. Australia, Canada, United States, and the United Kingdom are the common-law countries, the rest are code-law countries. However, the common law country category in the third row from the bottom excludes the UK.

R = security return over the fiscal year; RD = 1 if R< = 0 and RD = 0 if R > 0;

DIV = annual dividends per share deflated by beginning of period price; N = the number of firm/year observations.

RATIO = the ratio of the R2s of the equivalent earnings (table 2) and dividends (this table) regressions.

Vuong statistic is a likelihood ratio statistic that compares the fit of the return/dividend model in this table in panel B against

the return/earnings model in panel B of table 2, both estimated separately using the good news and bad news return observations in panel B.

40

41

Table 5

Contemporaneous Association between Cashflows and Returns

Statistics are from regressions using the pooled cross-section and time-series of firm/year observations for each country. Intercepts are not reported.

Panel A: OCF = β0 + β1 RD + β2 R + β3 R*RD + ε

2

β2 t(β2)β3 t(β3)Adj. RN Ratio Australia 0.07 2.740.16 2.112.4% 1321 3.80 Canada 0.02 1.070.25 5.283.6% 2901 4.75 USA 0.04 7.810.10 7.393.1% 21225 4.78 UK 0.11 11.40-0.02 -0.873.7% 5758 3.69 France 0.19 4.78-0.16 -1.393.3% 1054 3.79 Germany 0.16 4.28-0.04 -0.372.6% 1245 2.10 Japan 0.01 1.170.08 7.181.7% 6855 2.49 Common 0.04 8.040.12 9.193.0% 25447 4.79 UK 0.11 11.40-0.02 -0.873.7% 5758 3.69 Code 0.08 8.920.00 -0.042.2% 9154 2.34 Panel B: OCF = β0 + β1 R + ε (R≥0) OCF = β0 + β1 R + ε (R<0) β1 Adj. R2N Ratio Vuong β1 Adj. R2N Ratio Vuong Australia 0.07 0.9%813 0.250.92 0.23 1.4%508 12.88-3.38 Canada 0.02 0.0%1688 384.000.42 0.27 2.8%1213 2.82-6.23 USA 0.04 0.5%12721 9.34-1.51 0.14 1.4%8504 7.31-12.88 UK 0.11 3.4%3612 0.550.18 0.09 0.7%2146 13.85-6.93 France 0.19 2.9%611 0.76-2.12 0.03 -0.2%443 -47.40-1.67 Germany 0.16 2.2%712 2.22-0.16 0.13 0.3%533 13.96-2.18 Japan 0.01 0.0%3141 10.00-2.79 0.09 2.2%3714 2.072.61 Common 0.04 0.4%15222 8.12-0.13 0.16 1.6%10225 6.45-14.62 UK 0.11 3.4%3612 0.550.18 0.09 0.7%2146 13.85-6.93 Code 0.08 1.3%4464 1.20-3.10 0.08 0.5%4690 5.49-2.18

Sample consists of 40,359 firm-year observations selected from the Global Vantage industrial/Commercial and Issue files over 1985-95, using the following procedure. First, for each variable (see below) we eliminate the two extreme percentiles of firm-year observations. Second, we eliminate all firm-year observations with missing values for one or more variables, to facilitate comparability with results in previous tables. Third, we eliminate all firm-year observations from countries with less than 1,000 observations, leaving seven countries represented. Australia, Canada, United States, and the United Kingdom are the common-law countries, the rest are code-law countries. However, the common law country category in the third row from the bottom excludes the UK.

R = security return over the fiscal year; RD = 1 if R< = 0 and RD = 0 if R > 0;

OCF = annual operating cash flow per share, where operating cashflow is defined as earnings plus decrease in non-cash current assets plus increase in non-debt current liabilities plus depreciation; N = the number of firm/year observations.

RATIO = the ratio of the R2s of the equivalent earnings (table 2) and dividends (this table) regressions.

Vuong statistic is a likelihood ratio statistic that compares the fit of the return/operating cash flow model in this table in panel B against the return/earnings model in panel B of table 2, both estimated separately using the good news and bad

news return observations in panel B.

42

Table 6

Contemporaneous Association between Earnings, Dividends, Cash flows and Returns:

Sub-period Analysis

Analysis: Statistics are from regressions using the pooled cross-section and time-series of firm/year observations for each country. Intercepts are not reported.

1985-90 1991-95

2 2

β2 t(β2) β3 t(β3) Adj. RN β2 t(β2) β3 t(β3) Adj. RN Panel A: NI = β0 + β1 RD + β2 R + β3 R*RD + ε

Australia 0.02 1.04 0.26 5.31 13.0% 500 -0.02 -0.99 0.48 7.55 9.4% 821 Canada 0.00 -0.23 0.38 12.94 19.8% 1594 0.01 1.30 0.44 11.63 15.8% 1307 USA 0.03 6.14 0.28 26.79 17.3% 11978 0.03 6.54 0.33 22.14 12.9% 9247 UK 0.04 7.49 0.09 6.92 13.7% 2682 0.06 9.02 0.19 11.64 17.4% 3076 France 0.09 5.96 0.04 1.09 19.1% 466 0.06 3.15 0.22 4.24 10.5% 588 Germany 0.05 3.86 0.04 0.97 6.1% 488 0.04 1.95 0.19 4.08 5.3% 757 Japan 0.01 5.21 0.02 4.96 7.0% 3460 0.00 0.16 0.01 1.56 1.0% 3395

Common 0.02 5.64 0.29 30.12 17.3% 14071 0.02 5.32 0.36 26.22 12.7% 11374 UK 0.04 7.49 0.09 6.92 13.7% 2682 0.06 9.02 0.19 11.64 17.4% 3076 Code 0.03 10.97 0.00 0.42 5.3% 4413 0.05 8.05 0.01 1.38 4.4% 4739 Panel B: DIV = β0 + β1 RD + β2 R + β3 R*RD + ε

Australia 0.00 -0.36 0.06 5.09 14.4% 500 -0.01 -1.89 0.07 7.30 10.3% 821 Canada -0.01 -5.37 0.04 8.12 7.4% 1594 -0.01 -6.60 0.05 7.69 7.0% 1307 USA -0.02 -18.02 0.05 25.51 9.8% 11978 -0.01 -19.45 0.05 23.09 8.9% 9247 UK 0.00 2.40 0.03 7.80 8.1% 2682 0.01 8.64 0.02 5.99 11.3% 3076 France 0.01 4.03 0.02 2.02 13.7% 466 0.01 1.42 0.04 4.24 6.7% 588 Germany 0.02 5.40 0.01 1.00 10.7% 488 0.01 3.69 0.01 1.34 8.2% 757 Japan 0.00 6.61 0.00 5.16 9.2% 3460 0.00 -0.61 0.01 7.81 7.2% 3395

Common -0.01 -17.25 0.05 26.29 9.2% 14071 -0.01 -18.40 0.05 24.33 8.0% 11374 UK 0.00 2.40 0.03 7.80 8.1% 2682 0.01 8.64 0.02 5.99 11.3% 3076 Code 0.01 9.31 0.00 -0.51 4.2% 4413 0.01 9.99 0.00 1.88 9.7% 4739 Panel C: OCF = β0 + β1 RD + β2 R + β3 R*RD + ε

Australia 0.01 0.21 0.14 1.31 1.0% 500 0.09 3.02 0.23 2.07 3.3% 821 Canada -0.02 -0.73 0.29 4.62 3.4% 1594 0.04 2.07 0.22 3.03 3.9% 1307 USA 0.03 3.82 0.09 5.13 2.5% 11978 0.05 7.15 0.13 5.99 3.8% 9247 UK 0.11 7.24 -0.02 -0.55 3.1% 2682 0.12 9.05 -0.03 -0.78 4.5% 3076 France 0.20 3.42 -0.15 -0.96 2.6% 466 0.23 3.95 -0.17 -1.00 4.3% 588 Germany 0.12 2.40 0.12 0.87 2.4% 488 0.25 4.08 -0.19 -1.40 3.0% 757 Japan 0.03 4.99 0.03 1.54 1.9% 3460 -0.01 -0.75 0.12 6.72 3.8% 3395

Common 0.02 2.87 0.12 6.88 2.5% 14071 0.05 8.13 0.15 6.98 3.7% 11374 UK 0.11 7.24 -0.02 -0.55 3.1% 2682 0.12 9.05 -0.03 -0.78 4.5% 3076 Code 0.08 7.30 -0.07 -2.34 1.6% 4413 0.21 11.49 -0.08 -2.39 6.3% 4739

43

44

Sample consists of 40,359 firm-year observations selected from the Global Vantage industrial/Commercial and Issue files over 1985-95, using the following procedure. First, for each variable (see below) we eliminate the two extreme percentiles of firm-year observations. Second, we eliminate all firm-year observations with missing values for one or more variables, to facilitate comparability with results in previous tables. Third, we eliminate all firm-year observations from countries with less than 1,000 observations, leaving seven countries represented. Australia, Canada, United States, and the United Kingdom are the common-law countries, the rest are code-law countries. However, the common law country category in the third row from the bottom excludes the UK.

R = security return over the fiscal year; RD = 1 if R< = 0 and RD = 0 if R > 0;

NI = annual earnings per share before extraordinary items deflated by beginning of period price; DIV = annual dividends per share deflated by beginning of period price;

OCF = annual operating cash flow per share deflated by beginning of period price, where operating cashflow is defined as earnings plus decrease in non-cash current assets plus increase in non-debt current liabilities plus depreciation; N = the number of firm/year observations.

RATIO = the ratio of the R2s of the equivalent earnings (table 2) and dividends (this table) regressions.

Vuong statistic is a likelihood ratio statistic that compares the fit of the return/operating cash flow model in this table in panel B against the return/earnings model in panel B of table 2, both estimated separately using the good news and bad news return observations in panel B.

Figure 1

International Differences in Earnings Timeliness

R-squares from Individual Country Regressions of Earnings on (a) Annual Return and (b) Annual Return times Negative Return Dummy

45

18%16%14%12%

R-squares10%8%6%4%2%0%

AUS

CAN

USA

UK

FRA

GER

JPN

Countries

Figure 2

International Differences in

Asymmetry in Earnings Response to Good and Bad News

Sensitivity to Positive and Negative Returns from Pooled Regression of Earnings on

(a) Annual Return and (b) Annual Return times Negative Return Dummy.

0.400.350.300.25

46

Sensitivity0.200.150.100.050.00

AUS-0.05

CANUSAUKFRAGERJPNCountries

Positive ReturnsNegative Returns

47

Figure 3

International Differences

in the Contemporaneous Relation Between Earnings and Returns:

R-squares from (1) the Pooled Reverse Regressions of Earnings Against

(a) Annual Return and (b) Annual Return times Negative Return Dummy and (2) Annual Return

against NI and ∆NI.

16%14%12%

R-squares10%8%6%4%2%0%

AUS

CAN

USA

UK

FRA

GER

JPN

Countries

NI=f(R,R*RD)R=f(NI,DNI)

48

Figure 4

Earnings Versus Dividends:

R-squares from the Individual Country Reverse Regressions of (1) NI and (2) DIV Against (a) Annual

Return and (b) Annual Return times Negative Return Dummy.

16%

14%

12%

10%

R-squares8%

6%

4%

2%

0%

AUS

CAN

USA

UK

FRA

GER

JPN

Countries

NI=f(R,R*RD)DIV=f(R,R*RD)

49

Figure 5

Earnings Versus Cashflows:

R-squares from the Individual Country Reverse Regressions of (1) NI and (2) OCF Against (a) Annual

Return and (b) Annual Return times Negative Return Dummy.

18%16%14%12%

R-squares10%8%6%4%2%0%

AUS

CAN

USA

UK

FRA

GER

JPN

Countries

NI=f(R,R*RD)OCF=f(R,R*RD)