The role of securities market governance on reporting conservatism for public

We use two traits of securities markets to examine relations between securities market governance mechanisms and financial reporting conservatism for public banks around the world. One is rules of securities laws and the other is securities market development.⁴⁵

Based on La Porta et al. (2006), there are two possibilities for the securities laws to influence bank financial reporting conservatism. First, private enforcement rules of securities laws that standardize the private contracting framework reduce the cost of private litigation for investors and improve their demand for conservative reporting on public banks. As Watts (2003a) and Bushman and Piotroski (2006) maintain, self-interested regulators may have incentives to apply regulation policies that encourage conservative financial reporting because

44 For further evidence on effects of government supervision policies, please refer to Barth et al. (2004) and Beck et al. (2005). For evidence on effects of bank supervisory actions, please refer to Berger and Davies (1998), DeYoung et al. (2001), Gunther and Moore (2003). For evidence on effects of private sector discipline, please refer to Martinez Peria and Schmukler (2001), Caprio and Honohan (2004) and Demirgüç-Kunt and Huizinga (2004).

45 For further discussions about securities market regulation, enforcement of laws, and securities market development, please refer to Shleifer (2005).

93 they are less likely to be blamed for understatements in financial reporting scandals. By empowering independent securities regulators to intervene in listed firms for violations to securities laws, public enforcement rules may increase bank litigation risks and force them to report more conservative financial information. Since more stringent rules of private or public enforcement increase public banks’ litigation risks, we expect they may improve securities holder’s reliance on contracting mechanism, lead to higher demand on verifiable information, and raise the reporting conservatism by public banks. Therefore, in countries with stringent private or public enforcement rules, public banks should be more likely to show a larger degree of reporting conservatism than their private counterparts.

It is also possible that, in countries with highly developed securities markets, market discipline and private litigation through contracting mechanism function well to ensure good conducts in the market without applying the rules in securities laws. Thus, we hypothesize that, in countries with more developed securities markets, public banks are more likely to show reporting conservatism than those in countries with less developed securities markets. We also expect that difference of conservatism across bank’s listing status should be more prevalent in countries with more developed securities markets than in countries with less developed securities markets. Furthermore, we compare the debt market demand (debt contracting) with the stock market demand (compensation contracting) for conservatism, using bond market development and stock market development measures as proxies. We expect that contracting mechanism works mainly through debt contracting rather than stock (compensation) contracting because other tools for compensation contracting, such as stock price information and stock options, may weaken its role on financial reporting conservatism.

94 3. Measures for bank conservatism in financial reporting

3.1 Conservative reporting on earnings changes

The work by Basu (1997) shows that conservative financial reporting exhibits a greater tendency for negative earnings changes to reverse in the next period than for positive earnings changes. By applying Basu’s work on conservative reporting for earnings changes as a measure of conservatism, Ball and Shivakumar (2005) and Nichols et al. (2005) examine difference in reporting conservatism across listing status for U.K. firms and U.S. banks, respectively. Both studies support that listing status is a crucial factor for financial reporting conservatism by showing that listed firms (banks) are more conservative in reporting earnings changes. We follow this line of literature to apply the serial dependence model in examining bank conservative reporting across listing status around the world. In this model, data for publicly traded and privately held banks are pooled together and a dummy variable, DPR, is used for private bank data to estimate difference in reporting conservatism across listing status.

Specifically, we estimate equation (4.1) for each country.

1 ,

where ΔNIit is change in net income from fiscal year t-1 to t, scaled by the total assets at the end of year t-1; D^ΔNIi,t-1<0 is a dummy variable that takes the value of one if the prior year’s earnings change (ΔNIi,t-1) is negative and it is zero otherwise. DPR is a dummy variable that takes the value of one for private banks and zero for public banks.

In equation (4.1), α3, named PUBCNSV_NI, is a measure of conservative reporting on negative earnings changes for public banks, and α7, named DIFCNSV_NI, is a measure of difference in conservative reporting across listing status on negative earnings changes.

According to the literature on financial reporting conservatism, we expect that public banks are

95 conservative in reporting negative earnings changes, i.e., a negative PUBCNSV_NI (α³). If private banks are less conservative than public banks in reporting negative earnings changes, we may observe a positive DIFCNSV_NI (α⁷).

3.2 Conservative reporting on loan losses

To measure bank conservative reporting for loan losses, we examine two behaviors of reporting on loan loss provisions. First, we analyze relations between loan loss provisions and changes in bank operating cash flows. Second, we examine relations between loan loss provisions and changes in problem loans and relations between loan loss provisions and net charge offs.

3.2.1 Conservative reporting on loan loss provisions to changes in bank operating cash flows

The relations between loan loss provisions and bank operating cash flows build on the literature for the relations between accruals and operating cash flows for firms. Accruals have two roles in financial reporting. First, they mitigate noise in operating cash flows, i.e. earnings stabilization (Dechow, 1994; Dechow et al., 1998). Second, they incorporate unrealized (expected) gains and losses into financial statements. The second role is stronger for loss recognition than for gain recognition, because firms with conservative reporting tend to recognize losses rather than gains as unrealized accrued charges (Ball and Shivakumar, 2005).

In the banking industry, loan loss provisions are bank accruals for expected changes of future loan loss realizations. The first stabilizing role indicates that banks increase their loan loss provisions when the performance goes up and decrease their loan loss provisions when the performance goes down. Since the role of loan loss provisions for unrealized loss recognition is stronger than that for unrealized gains recognition, when the performance goes down, the

96 second role indicates that banks with conservative reporting incorporate more expected loan losses by recognizing higher loan loss provisions. Based on this idea, we measure conservative reporting of loan loss provisions (LLP) relative to changes in bank operating cash flows (ΔCFt) through equation (4.2) for each country:

it proxied by income before loan loss provisions, scaled by lagged total assets; D^ΔCF^it<0 is a dummy variable that takes the value of one if earnings change in year t (ΔNIit) is negative and it is zero otherwise.

In equation (4.2), β³, named PUBCNSV_LLP_CF, is a measure of conservative reporting on loan loss provisions relative to negative changes in operating cash flows for public banks, and β⁷, named DIFCNSV_LLP_CF, is a measure of difference in conservative reporting across listing status on loan loss provisions relative to negative changes in operating cash flows. For the conservatism of public banks, if loan loss provisions play a stronger role in recognizing unrealized losses than unrealized gains, when operating cash flows decrease (the performance goes down) we may observe banks incorporate more loan loss provisions, i.e., a negative PUBCNSV_LLP_CF (β³). If private banks are less conservative in reporting loan losses than public banks, when operating cash flows decrease we may observe a smaller degree of increase in loan loss provisions for private banks, i.e., a positive DIFCNSV_LLP_CF (β⁷).

3.2.2 Conservative reporting on loan losses provisions to changes in problem loans and to net charge offs According to Hasan and Wall (2004), banks around the world follow analogous steps to determine their loan loss provisions. In many countries, banks apply systematic procedures to

97 classify loans as performing and non-performing (also called problem loans). The procedures are usually based on a combination of the following methods: the number of days that a loan is in arrears, or a forward looking estimate of default probability. Using the information on the problem loans in the existing loan portfolio and the loan loss reserves in the previous period, banks estimate expected changes in the value of loan losses and then determine their loan loss provisions at the end of each period. Loan loss reserves, the cumulative loan loss provisions, are a reduction for the outstanding loans on the balance sheet for a bank. Furthermore, banks charge off problem loans when they realize that the loans become uncollectible. The charge off reduces the level of loan loss reserves and the outstanding loans. When previously charged off loans are recovered, banks recognize gains in their income statements.

By examining U.S. banks, Nichols et al. (2005) show that banks with conservative reporting incorporate loan loss provisions for a larger amount prior to or at the same time when problem loans increase. We also expect banks with more conservative loan loss reporting charge off uncollectible loans at a greater speed than those with less conservative loan loss reporting. In equation (4.3), the measure of conservative loan loss reporting builds on Nichols et al. (2005), which mainly looks at the relationships across loan loss provisions, changes in problem loans, and net chare offs.

it and ΔPLi,t+1, are changes in problem loans scaled by lagged total assets at year t-1, t, and t+1;

NCOFit and NCOFi,t+1 are net charge offs scaled by lagged total assets at year t and t+1; LLRi,t-1 is loan loss reserves scaled by lagged total assets at year t-1 and it controls for the beginning level

98 of loan loss reserves in year t; TCLit is total customer loans scaled by lagged total assets at year t and it controls for different strategies for bank loan portfolios.

In equation (4.3), γ² andγ⁴ , named PUBCNSV_LLP_PL and PUBCNSV_LLP_COF, are measures of conservative reporting on loan loss provisions relative to changes in problem loans and to net charge offs for public banks, respectively. Furthermore,γ¹⁰ andγ¹² , named DIFCNSV_LLP_PL and DIFCNSV_LLP_COF, are measures of difference in conservative

reporting across listing status on loan loss provisions relative to changes in problem loans and to net charge offs, respectively. We expect banks that are conservatism in financial reporting recognize more loan loss provisions when their problem loans increase and charge off more problem loans when their loan loss provisions increase, i.e. positive PUBCNSV_LLP_PL (γ²) and PUBCNSV_LLP_COF (γ⁴). If private banks are less conservative than their public counterparts, we should observe a smaller degree of conservatism for private banks, i.e., negative DIFCNSV_LLP_PL (γ¹⁰) and PUBCNSV_LLP_COF (γ¹²).

4. Sample selection and summary descriptive statistics 4.1 Sample selection and data sources

Accounting variables are obtained from the September 2005 CD-ROM edition of BANKSCOPE database supplied by Bureau Van Dijk. Because accounting variables are only available for the past eight years in BANKSCOPE, we require sample countries with sufficient firm-level accounting data for banks over the period from 1997 to 2004 to estimate our models.

In addition, we exclude investment banks/securities houses, Islamic banks, specialized governmental credit institutions, central banks, and multi-lateral governmental banks from the sample, because their primary activities are very different from traditional banks.

We also require sample countries contain data for legal origin and rules of securities

99 laws. Data for legal origin and rules of securities laws are obtained from La Porta et al. (1998) and La Porta et al. (2006), respectively. New Zealand and Uruguay are excluded because no data on public banks are available. Ecuador is excluded due to its problematic extreme accounting figures.⁴⁶ Zimbabwe is excluded because, during the sample period, it experienced hyperinflation (an average inflation rate of 79.18% from 2000 to 2004) and involved in a war.⁴⁷ After these restrictions are applied, the final sample consists of 1,195 public traded banks and 6,404 private held banks around 45 countries.

4.2 Descriptive statistics and institution variables

Table 4.1 presents descriptive statistics of accounting variables for public and private banks by country. In many cases, public banks have larger size (TA) and better performance (ROA and NI) than their private counterparts. The data show that number of banks varies across listing status and across countries. Over 43% of the public banks comes from U.S.A.

(394 banks, or 33%) and Japan (122 banks, or 10%), and over 57% of the private banks comes from Germany (1,646 banks, or 26%), Italy (705 banks, or 11%), Japan (657 banks, or 10%) and U.S.A. (655 banks, or 10%). Among countries with large number of banks, German and Italy have very small number of public banks relative to private banks. In many other countries, both number of public banks and private banks are small. To make sure that our results are not dominated by some countries, we estimate country-by-country measures for conservatism and apply cross-country regressions to examine the effects of securities laws on bank reporting conservatism. The data also indicate that public banks may be more conservative in reporting

46 For example, its net income to lagged total assets (NI) and problem loans to lagged total assets (PL) are -105.8% and 378.75%, respectively. According to information from the CIA World Factbook, Ecuador suffers from economic crisis and political instability during this period. We conjecture that these problems lead to its inaccurate accounting information. However, the results for Ecuador are available upon request.

47 For more information about the economy and political situation in Zimbabwe during the sample years, please refer to the CIA World Factbook. Besides, the results for Ecuador are available upon request.

100 loan losses than their private counterparts. Public banks report a higher level of loan loss provisions (LLP) on average, although the average level of total customer loans (TCL) is similar between public and private banks. Furthermore, the large variation of total customer loans (TCL) indicates banks have very different strategy for their loan portfolios. Thus, we control for this factor when we examine the conservatism in reporting loan loss provisions to problem loans and to net charge offs.

101 Table 4.1 Summary statistics of accounting variables for public and private banks by country

Public banks (N = 1,195) Private banks (N = 6,404) Notes: Accounting variables are obtained from BANKSCOPE. The detailed definitions of variables are provided in Appendix C.

102 Table 4.2 presents law origins, economic conditions, and descriptive statistics for indexes that portray the stringencies for banking industry regulations and securities laws regulations in our sample countries. Sample countries are grouped into four commercial law origins which are English, French, German and Scandinavian origins. The averages of GDP per capita (GDP), GDP growth (GDPGR) and inflation rate (INFL), retrieved from the World Development Indicator (WDI) database, show economic conditions of sample countries for the period from 2000 to 2004.⁴⁸ It shows that our sample consists of countries with varieties of legal origins and economic conditions, and thus we control for them in the cross-country regressions. We use GDP per capita as a control for economic condition for a country and set it to 1 if the value of GDP per capita is higher than or equal to the median country value; otherwise, it is set to 0.

Columns 6-7 of Table 4.2 report indexes for banking industry regulations. Supervisory power index (SUPPWR), ranging from 0 to 14, measures supervisory powers to take actions to avid and to cure problem banks. It is a proxy for the stringency of direct government supervision on banks for a country. Private monitoring index (PRIIDX), ranging from 0 to 5, measures supervision policies for encouraging private sector monitoring, including monitors from depositors, certified auditors, and credit rating agencies. Higher values of SUPPWR and PRIIDX indicates more powerful supervisory actions and better environment for outside

monitoring and are expected to be positively associated with more conservative financial reporting. Data source for SUPPWR and PRIIDX is the 2003 edition of the World Bank’s bank regulation and supervision database. Further, in the cross-country regressions, we set these supervision variables to 1 if their values are higher then or equal to the median country value;

otherwise, they are set to 0.

48 The WDI database does not provide Taiwan’s data, so we download them from the website of the National Statistics (http://www.stat.gov.tw/mp.asp?mp=4) supported by the government of Taiwan.

103 Table 4.2 Law origins, economic conditions and descriptive statistics for bank supervision and securities

law stringencies by country

Notes: The detailed definitions of variables are provided in Appendix C.

104 Columns 8-11 of Table 4.2 present measures for securities market governance. They are Private enforcement rules (PRIVENF), public enforcement rules (PUBLENF), bond market capitalization to GDP (BONDCAP) and stock market capitalization to GDP (STKGDP).

Private enforcement rules (PRIVENF) assess the prospectus disclosure requirements for issuing firms and the liability standards for the issuer, its distributors and accountants when investors sue them for recovering losses due to misleading statements in the prospectus. A higher value for PRIVENF indicates more stringent rules that standardize the private contracting framework and lower costs of private litigation for investors. Public enforcement rules (PUBLENF) assess the independence of securities market regulators and their powers to make rules regarding securities offerings, to investigate violations of securities laws, and to issue non-criminal and criminal sanctions for violations of securities laws. A higher value for PUBLENF indicates that independent regulators with experience and expertise are better

empowered to protect investor’s rights. We expect that a county with better empowered regulators exhibits higher litigation risks for issuing firms and thus is associated with stronger conservative reporting by public banks. In the cross-country regressions, we set these variables for securities market governance to 1 if their values are higher then or equal to the median country value; otherwise, they are set to 0. Appendix C describes variable definitions and provides data sources in more detail.

5. Results of bank financial reporting conservatism across listing status around the globe 5.1 Financial reporting conservatism across listing status: Results from pooled regressions

Using the pooled sample banks from all 45 countries, we estimate the conservatism of bank financial reporting and examine whether public banks are more conservative than their private counterparts. The models are estimated by linear OLS regressions with the White and

105 the country clustered robust standard deviations. The results are presented in Table 4.3.

Panels A-C of Table 4.3 discuss the reporting conservatism across listing status on earnings changes, loan loss provisions relative to changes in operating cash flows, and loan loss provisions relative to changes in problem loans and to net charge offs, respectively. Number of observations differs for each model due to missing accounting variables. Our results for pooled sample generally support that banks are conservative in their financial reporting and that public banks are more conservative than their private counterparts.

Panel A of Table 4.3 estimates the conservative reporting across listing status on earnings changes. With a significantly negative α3 (-0.761 with t-stat = -4.18), public banks show a larger amount of reverse in earnings changes for negative earnings changes in the previous period than for positive earnings changes. This supports our argument that public banks are conservative in reporting earnings changes. Public banks also report a larger amount of reverse in earnings changes for negative earnings changes in the previous period than private banks do, with α7 (0.382 with t-stat =1.77) indicating the amount of difference in reverse. The parameters show that, for 1% of earnings decreases in a period, public banks report 0.765% of earnings increases in the next period, but private banks report only 0.333% of earnings increases.

Panel B of Table 4.3 estimates the conservative reporting across listing status on loan loss provisions relative to changes in operating cash flows. It shows that public banks report larger

Panel B of Table 4.3 estimates the conservative reporting across listing status on loan loss provisions relative to changes in operating cash flows. It shows that public banks report larger