Dividends from unrealized earnings and default risk

These studies by and large use firms that adopted the International Financial Reporting Standards (IFRS). In comparison to US GAAP, which allows the measurement of financial instruments only at fair value, the international standards permit the measurement of different financial statement items at fair value. As a result, the impact of fair value accounting on financial reporting is far more substantial in IFRS-adopting countries than in the US. See De George et al. (2016) for a review of the literature on the effects of IFRS adoption on financial reporting quality, corporate decision making, etc.

To illustrate the increase in the riskiness of the firm following a payout based on unrealized earnings, assume that a firm owns a building with a historical cost of 100 and cash in the amount of 100. Assume also that the market value of the building is 200. The adoption of IFRS allows the firm to recognize an unrealized profit of 100 by switching from historical cost to fair value. Now, assume that the firm decides to distribute this profit and uses its cash to pay the dividend. Following the payment, the firm owns the same building valued at 200 and has no cash. As a result, the riskiness of the firm increases because it is now 100% invested in the risky asset while it was only 67% invested prior to the dividend payment. We thank an anonymous referee for offering this illustration.

FASB Statement No. 115 Accounting for Certain Investments in Debt and Equity Securities (1993), FASB Statement No. 133 Accounting for Derivative Instruments and Hedging Activities (1998), and FASB Statement No. 159 The Fair Value Option for Financial Assets and Financial Liabilities (2007).

IAS No. 39 Financial Instruments: Recognition and Measurement (as revised in 2005; later replaced by IFRS 9 Financial Instruments); IAS No. 40 Investment Property (as revised in 2005); IAS No. 27 Consolidated and Separate Financial Statements (as revised in 2005); IAS No. 28 Investment in Associates and Joint Ventures (as revised in 2005); IFRS 3 Business Combinations (as revised in 2008).

Before their adoption of IFRS, the firms reported their financial statements in accordance with the Israeli GAAP, which was mainly influenced by US GAAP. For a detailed description of the differences between Israeli GAAP and IFRS, see Markelevich et al. (2011).

See Sections 302 and 303 of the Israeli Corporate Law. Later in this section, we provide examples of countries in which the IFRS amounts do not have to be modified to determine distributable profits.

Corporate laws worldwide by and large require dividend payments to be conditional on the firm’s ability to repay all of its liabilities.

There is consistent evidence of the increased default risk being mispriced according to the bond ratings and expected default frequencies (EDF) of the firms. Bond yield spreads provide some evidence of investors in the bond market possibly reacting (at least to some extent) to the distribution of risky dividends originating from unrealized gains, but (unlike in the case of the bond ratings and EDF results) these findings do not hold when we limit the analysis to firms with an ex ante similar inclination to pay such risky dividends.

E.g., KPMG Feasibility Study of Capital Maintenance: Main Report. http://​ec.​europa.​eu/​internal_​market/​company/​docs/​capital/​feasbility/​study_​en.​pdf

From the debtholders’ perspective, the payment of dividends reduces the firm’s value, thereby increasing the value of the implicit put option and the probability of default.

While Handjinicolaou and Kalay (1984) find evidence consistent with the information content hypothesis, Dhillon and Johnson (1994) present evidence in support of the wealth redistribution hypothesis, which, as they note, “does not rule out the information content hypothesis” (p. 281).

Disbursing cash as dividends can also mitigate minority shareholder concerns about expropriation. Indeed, in countries with good legal protection, minority shareholders can use their legal power to force firms to dispense cash as dividends to reduce the risk of expropriation, particularly if the firms lack alternative value-maximizing uses for their cash reserves (e.g., La Porta et al. 2000; Shleifer and Wolfenzon 2002). On that note, we point out that Israel is a common law country (similar to the United States and the United Kingdom), with effective legal protection of minority shareholders (La Porta et al. 1998) and prevailing concentrated insider holdings of closely held firms (the average insider holdings in our sample are 61%).

According to the pecking-order hypothesis, information asymmetries lead managers to initially prefer internal financing, as it is a cheaper and less risky source of capital. Only if there are insufficient internal funds will the manager resort to the costlier option of external financing (Myers and Majluf 1984).

Though IFRS was formally adopted in 2008, almost all Israeli public firms had already voluntarily adopted IFRS in 2007.

Recall that the US GAAP rule that allows the measurement of financial instruments only at fair value affects mainly financial firms, particularly in terms of the ability to recognize unrealized revaluation earnings. Neither of our GAAP-reporting firms is a financial firm. Therefore neither one is significantly affected by fair value accounting rules. (US GAAP-reporting firms in Israel are by and large high-technology firms.)

Firms with no tradable debt are excluded from the sample, because there is no documentation about whether they defaulted on their (nonpublic) debt.

Appendix A outlines the identification and measurement of unrealized revaluation earnings (including quantitative information about the earnings in our sample) in accordance with the relevant international standards.

As of 2008, the Bank of Israel began recording all of the public debt restructurings in Israel. Note that the one-year gap between the initial adoption of IFRS in Israel (effective December 31, 2007) and the beginning of the recording of debt restructurings by the Bank of Israel (January 1, 2008) does not harm our analyses, because the consequences of distributing dividends, based on the new rules, would not have appeared prior to 2008.

A firm enters a debt restructuring when 1) it announces to its bondholders that it cannot pay its debt as outlined in the terms of the bond; and/or 2) the firm has not paid the debt, as per the terms of the bond; and/or 3) a court determines that the firm will not be able to repay its bondholders, as per the terms of the bond. The date of entering a debt restructuring is the date of whichever one of the above three events occurs first.

The methodology of calculation by Moody’s KMV is described by Crosbie and Bohn (2003).

Appendix B provides a detailed example of the identification of a specific firm in our sample as DFUR/non-DFUR throughout the sample period.

The results of our study are robust to excluding firms that never paid dividends throughout the sample period.

We point out that prior to the adoption of IFRS, real-estate firms, like other firms, did not revalue their holdings but reported them at cost.

On the face of it, DFUR firms could double or triple their dividend payout ratio using realized earnings alone. Still, this group of firms chose to distribute an amount that exceeds their total realized earnings based on unrealized gains recognized. While this study examines the repercussions of such behavior for the firm, an investigation of the behavioral aspects of DFUR is beyond its scope.

We use Z-scores based on Altman et al. (1998). Our results are robust to using either Z-scores based on Altman (1968) or Z-scores adjusted for Israeli companies (Ingbar 1994).

A survival analysis using hazard models obviates the shortcomings of static risk models and enables the estimation of the effect of several explanatory variables on a firm’s likelihood of defaulting on its debt during the estimation period (Shumway 2001; Campbell et al. 2008). Most studies examining the variables affecting financial distress have estimated single-period static models, although the information used is usually multiple-period data about financial distress (bankruptcy, filing for Chapter 11, etc.). As Shumway (2001) explains, by ignoring the fact that firms change over time, static models produce biased and inconsistent estimates. Survival analysis using hazard models solves the problems of static models by accounting explicitly for time.

Observations of firm-years for which a default has already occurred during the sample period are excluded from the analysis (in all, a redundancy of 174 post-default firm-years). In other words, a firm leaves the sample when it first enters a debt restructuring. If a firm enters a restructuring more than once during the sample period, the count of years is up until the first restructuring event.

We also examined specifications with capital expenditures as a proxy for the firm’s investment strategy. In principle, firms may invest in assets in the post-IFRS period merely for the sake of recognizing unrealized holding gains. Such improper investments can boost both unrealized earnings (and hence dividends) and default risk. In our sample, however, the levels of a firm’s (average and median) capital expenditures decreased in the post-IFRS period for both DFUR and non-DFUR firms (untabulated for parsimony). Moreover, we find that capital expenditures do not incrementally contribute to the explanation of default risk over and above the risk determinants included in model 1.

We repeated the analyses using cash dividend payments divided by realized earnings only. This ratio is interesting because it captures the excess dividend payments better, if any took place, given the firm’s level of realized earnings. Results are robust to using either the dividend payout ratio from total earnings or from realized earnings only. Nevertheless, we use cash dividend payments divided by total earnings as our primary measure to ensure that our results are not driven by changes in the dividend payout ratio. We thank an anonymous referee for suggesting this clarification.

Since none of the non-DFUR firms that paid dividends in the post-IFRS period required debt restructuring during the sample period, the inclusion of an interaction variable between DFUR and DivPayout is technically impossible.

We do not predict whether the association between earnings and the probability of a default differs for realized and unrealized earnings.

We also use alternative proxies for liquidity: the firm’s cash position and operating cash flows. Neither of these proxies performs better than the current ratio variable commonly used in financial-distress and cost-of-debt models. The main results remain unchanged when we replace the current ratio with either or both of these variables.

All of the variables in model 1 appear in Appendix C together with their definitions.

Interestingly, leverage does not contribute significantly to the explanation of the likelihood of a firm needing debt restructuring, beyond the impact of distributing dividends based on unrealized earnings, size, profitability, liquidity, earnings volatility, or bond maturity. The coefficient on Leverage in the default regressions remains insignificant, even if we exclude the other accounting items from the equation. Nevertheless, the coefficient on DFUR remains strongly significant and positive in all specifications. Note that in the cost-of-debt regressions presented later on, leverage is priced by rating agencies as well as by investors in the market, as reflected in significantly positive associations of leverage with bond ratings, yield spreads, and expected default frequencies.

Ownership concentration may have a significant effect on corporate governance by either mitigating agency problems (e.g., block holders may have greater incentive and power to monitor management; Shleifer and Vishny 1986) or exacerbating them (e.g., via expropriation of minority shareholders’ wealth; Shleifer and Vishny 1997). Thus the association between ownership concentration and DFUR is unpredictable.

Allowing for a nonlinear relationship is in accordance with Morck et al. (1988).

The discrepancy between ownership and control rights—a main feature of business groups—may create incentives for control holders to transfer resources from firms where they have fewer rights to firms where they have greater rights. (This transfer of resources is called “tunneling”; Johnson et al. 2000.) Control holders may take advantage of the new rules allowing recognition of unrealized revaluation earnings to increase the payment of dividends by companies situated lower down the pyramid within the business group.

As in the case of ownership concentration, the association between the adoption of a corporate social responsibility (CSR) policy and DFUR is also unpredictable. On the one hand, CSR may signal more management agency conflicts and less effective corporate governance, e.g., if managers choose to engage in CSR activities merely for the sake of building their personal reputations as good citizens (Barnea and Rubin 2010), to reduce the probability of their replacement (Cespa and Cestone 2007), or both. On the other hand, CSR adoption may signal strong, effective corporate governance, requiring managers to act in the best interests of shareholders, inter alia, by using CSR to reduce the conflict of interests between the former and the latter (e.g., Harjoto and Jo 2011). We obtained information about firms adopting a CSR policy from the annual “Maala Ranking of Corporate Social Responsibility” reports for the sample years. The Maala ranking includes categories about business ethics, corporate governance, management, and reporting (as well as community relations, work environment, and environmental protection).

http://​maala.​org.​il/​he/​company/​ranking/​faq/​Default.​aspx?​ContentID=​168

The estimation of the probit model for predicting DFUR is based on all of the Israeli nonfinancial and nondually listed public companies on the Tel Aviv Stock Exchange and is not restricted to firms with traded bonds. Nevertheless, the results are qualitatively the same when firms with traded bonds only are used. Furthermore, we use the average values (from 2004 to 2006) of the continuous variables in the probit analysis. We also run the probit model using the data for the most recent year prior to IFRS adoption only (2006). All inferences remain unchanged.

See footnote 5.

Critically, in the first year of IFRS adoption, firms reported the current as well as the previous year’s figures in accordance with IFRS. Thus, for the year preceding the adoption of IFRS, there are two values available: one according to the IFRS rules and one according to the former rules. For example, if a firm first adopted IFRS in 2007, the figures of 2006 appear in the financial statements of 2007 in accordance with IFRS rules (for comparability with the 2007 figures), whereas in the financial statements of 2006, they appear in accordance with the Israeli GAAP rules. The ratio of total assets for 2006 as per IFRS to total assets for 2006 as per the former rules captures the extent to which the firm potentially gains value as a result of the transition to IFRS, all else equal.

As an alternative indicator of a firm’s potential ability to exploit fair value accounting to increase dividend payments following the adoption of IFRS, we use an early IFRS adoption indicator (EarlyAdopt). In 2006, 45 firms in Israel voluntarily adopted IFRS before all the other public firms did. Hence early adaptors could recognize revaluation gains before other firms could and hence were potentially able to distribute dividends based on these gains before other firms. We find that EarlyAdopt is positively associated with DFUR (0.480, p value <5%). Importantly, the results of the default and cost-of-debt regressions based on the propensity score-matched sample of firms (see Tables 4 and 5, respectively) are robust to using either ExAnteValueGain or EarlyAdopt in the DFUR probit model 2. On that note, we point out that our main analyses for the post-IFRS period are robust to either including or excluding the 45 firms in Israel that early adopted the IFRS in 2006.

In an untabulated analysis, we examined whether intensified management agency conflicts, and/or poor corporate governance, are linked to a greater likelihood of the firm distributing dividends based on unrealized earnings. Univariate Pearson and Spearman correlations show that DFUR is insignificantly associated with our four controls: OwnerConc, OwnerConc_sq, B_Group, and CSR. Additionally, when included in the probit DFUR model, the coefficients on these controls are generally statistically insignificant.

According to Jensen and Meckling (1976), a high degree of leverage causes agency problems by creating incentives to shift risk and substitute assets.

When the cost of debt is proxied by bond ratings, the estimations of (3) exclude nonrated firm-years, resulting in the smaller number of observations in the Rating regressions. Nevertheless, our Rating variable has sufficient variation in the sample (see Table 2, Panel B) to allow a reliable statistical analysis. We repeated all of the analyses, including the debt restructuring analysis, using only firm-years for which we had both bond ratings and bond yield spreads. (In other words, we used the same number of observations throughout the study.) The results obtained from these analyses (untabulated for parsimony) are qualitatively similar to the results obtained from using all of the observations available for each regression separately (tabulated).

In an additional analysis, we ran the bond yield-spread regressions including the firm’s bond ratings as another control to explore the possibility that the adoption of fair value accounting affects a firm’s cost of debt through its impact on credit ratings (Anderson et al. 2003; Mansi et al. 2004; Magnan et al. 2016). Additionally, we repeated the cost-of-debt regressions, expanding the period prior to the adoption of IFRS (i.e., for 2004–2013). Untabulated results show that our main findings remain unchanged. Importantly, there is no empirical evidence that endogeneity affects our inferences.

In an additional analysis, we examined stock return volatility as an alternative risk indicator (e.g., Campbell and Taksler 2003), prior to the distribution of dividends originating from unrealized earnings and thereafter. We compared the stock volatility (our ReturnSD) for ex post DFUR and non-DFUR firms prior to the adoption of IFRS, when firms could not pay dividends based on unrealized earnings, and after the adoption, when such payments became possible. Untabulated results show that for both subperiods, ReturnSD is insignificantly different in the two groups of firms. The difference-in-difference is also insignificant. Thus there is no evidence that the increased default risk documented for DFUR firms in the post-IFRS period affects equity volatility.

See, for example, S&P Global Ratings, RatingsDirect, Methodology: Investment Holding Companies, http://​www.​maalot.​co.​il/​Publications/​MT20170124145506​.​pdf.

We thank an anonymous referee for proposing this additional analysis.

For consistency, tabulated results are for the models including industry and year fixed effects (see the explanation in the previous subsection). Untabulated results including firm, instead of industry, fixed effects and instead of the DFUR variable that is subsumed by the firm fixed effects, generally yield the same results. We also point out that in the difference-in-difference design, the variable EarnSD was excluded from the model, due to the small number of observations in the pre-IFRS period.

The inclusion of a dummy for the post-IFRS period is redundant, given that year fixed effects are controlled for in the equation.

We use daily bond return data from the Bank of Israel. Dividend announcement dates were obtained from the Bloomberg Professional database.

In the statistical tests for the cumulative two-day abnormal returns, we used the simplifying assumption of intertemporal independence of abnormal returns over the bonds’ holding periods.

In an additional analysis, we compared the abnormal returns on stocks around the dividend announcements made by DFUR versus non-DFUR firms. The abnormal stock returns are computed based on the market model (as explained in MacKinlay 1997), estimated using share prices over a period of 45 trading days ending 16 days before the dividend announcement date (as with the bonds). Untabulated results show no significant difference between the two groups of firms in abnormal stock returns around the announcement day. Thus, like the bond market, the stock market does not seem to respond differently to payouts made by DFUR versus non-DFUR firms.

Both measures were obtained from the Bank of Israel.

We thank an anonymous referee for suggesting this additional analysis.

Smaller (larger) firms are firms with total assets below (above) the sample median. Another control for information asymmetry used in previous studies is analyst coverage (generally proxied by the number of analysts covering the firm). In Israel, the extent of firm coverage by analysts is relatively limited. Moreover, none of the firms that needed debt restructuring during the sample period were covered by analysts. Hence this proxy cannot be used in the survival analysis.

Terza et al. (2008) show this method to be theoretically consistent and superior to the alternative (inconsistent) method of two-stage predictor substitution.

Inclusion of all of the control variables from the second-stage equation is consistent with the two-stage residual inclusion method (e.g., Chen et al. 2013). We point out that in an alternative analysis of the first-stage probit model, instead of using all of the second-stage controls, we included only the four controls used in our propensity-score matching procedure (Size, DivPayout, Leverage, and ExAnteValueGain; see Table 4). The results of the second-stage analysis remained unchanged.

The results from the first-stage analysis indicate that our ex ante valuation gain instrumental variable has a positive effect on the DFUR variable but that this effect is only marginally statistically significant (p value = 9.4%), suggesting a possible weak instrument problem.

We point out that the replacement of IAS 39 by IFRS 9 has no impact on those parts of the standard relevant to our research.

Revaluation earnings data are unavailable on financial databases such as Compustat or Bloomberg.

This type of financial asset/liability is referred to as “at fair value through profit or loss.”

Available-for-sale financial assets are those non-derivative financial assets not classified as (1) financial assets at fair value through profit or loss, (2) loans and receivables, or (3) held-to-maturity investments. Loans and receivables are non-derivative financial assets with fixed or determinable payments that are not quoted in an active market. Held-to-maturity investments are non-derivative financial assets with fixed or determinable payments and fixed maturity that an entity has the intention and ability to hold to maturity.

If a parent is required to measure its investment in a subsidiary at fair value through profit or loss, it shall also account for its investment in a subsidiary in the same way in its separate financial statements.

If, in accordance with IAS 28, an entity elects to measure its investments in associates or joint ventures at fair value through profit or loss, it shall also account for those investments in the same way in its separate financial statements. In compliance with IAS 28, many of the procedures appropriate for the application of the equity method are similar to the consolidation procedures described in IAS 27.

The investment holding firms in our sample are firms whose purpose is owning other firms’ stock. The holding firms themselves typically do not produce goods/services; rather, their purpose is the formation of corporate groups.

The amounts displayed in this appendix are in IS (Israeli Shekels) millions. During the sample period, the foreign exchange rate was in the range of 3.2–4.2 IS per 1 USD.